# ORIGINS OF HUMAN NEUROPATHOLOGY: THE SIGNIFICANCE OF TENEURIN-LATROPHILIN INTERACTION

EDITED BY : David A. Lovejoy, Antony A. Boucard and Richard P. Tucker PUBLISHED IN : Frontiers in Neuroscience and Frontiers in Endocrinology

#### Frontiers eBook Copyright Statement

The copyright in the text of individual articles in this eBook is the property of their respective authors or their respective institutions or funders. The copyright in graphics and images within each article may be subject to copyright of other parties. In both cases this is subject to a license granted to Frontiers. The compilation of articles constituting this eBook is the property of Frontiers.

Each article within this eBook, and the eBook itself, are published under the most recent version of the Creative Commons CC-BY licence. The version current at the date of publication of this eBook is CC-BY 4.0. If the CC-BY licence is updated, the licence granted by Frontiers is automatically updated to the new version.

When exercising any right under the CC-BY licence, Frontiers must be attributed as the original publisher of the article or eBook, as applicable.

Authors have the responsibility of ensuring that any graphics or other materials which are the property of others may be included in the CC-BY licence, but this should be checked before relying on the CC-BY licence to reproduce those materials. Any copyright notices relating to those materials must be complied with.

Copyright and source acknowledgement notices may not be removed and must be displayed in any copy, derivative work or partial copy which includes the elements in question.

All copyright, and all rights therein, are protected by national and international copyright laws. The above represents a summary only. For further information please read Frontiers' Conditions for Website Use and Copyright Statement, and the applicable CC-BY licence.

ISSN 1664-8714 ISBN 978-2-88963-858-1 DOI 10.3389/978-2-88963-858-1

#### About Frontiers

Frontiers is more than just an open-access publisher of scholarly articles: it is a pioneering approach to the world of academia, radically improving the way scholarly research is managed. The grand vision of Frontiers is a world where all people have an equal opportunity to seek, share and generate knowledge. Frontiers provides immediate and permanent online open access to all its publications, but this alone is not enough to realize our grand goals.

#### Frontiers Journal Series

The Frontiers Journal Series is a multi-tier and interdisciplinary set of open-access, online journals, promising a paradigm shift from the current review, selection and dissemination processes in academic publishing. All Frontiers journals are driven by researchers for researchers; therefore, they constitute a service to the scholarly community. At the same time, the Frontiers Journal Series operates on a revolutionary invention, the tiered publishing system, initially addressing specific communities of scholars, and gradually climbing up to broader public understanding, thus serving the interests of the lay society, too.

#### Dedication to Quality

Each Frontiers article is a landmark of the highest quality, thanks to genuinely collaborative interactions between authors and review editors, who include some of the world's best academicians. Research must be certified by peers before entering a stream of knowledge that may eventually reach the public - and shape society; therefore, Frontiers only applies the most rigorous and unbiased reviews.

Frontiers revolutionizes research publishing by freely delivering the most outstanding research, evaluated with no bias from both the academic and social point of view. By applying the most advanced information technologies, Frontiers is catapulting scholarly publishing into a new generation.

#### What are Frontiers Research Topics?

Frontiers Research Topics are very popular trademarks of the Frontiers Journals Series: they are collections of at least ten articles, all centered on a particular subject. With their unique mix of varied contributions from Original Research to Review Articles, Frontiers Research Topics unify the most influential researchers, the latest key findings and historical advances in a hot research area! Find out more on how to host your own Frontiers Research Topic or contribute to one as an author by contacting the Frontiers Editorial Office: researchtopics@frontiersin.org

# ORIGINS OF HUMAN NEUROPATHOLOGY: THE SIGNIFICANCE OF TENEURIN-LATROPHILIN INTERACTION

Topic Editors:

David A. Lovejoy, University of Toronto, Canada Antony A. Boucard, Centro de Investigación y Estudios Avanzados, Instituto Politécnico Nacional de México (CINVESTAV), Mexico Richard P. Tucker, University of California, Davis, United States

We are delighted to introduce this new special issue on "The Origins of Neuropathology: The Roles of Teneurins and Latrophilins". Although the title may seem particularly bold, and indeed, perhaps presumptuous, we the editors, think our title well warranted based on the findings and interpretation provided by a dedicated group of researchers who have developed this field over the last 25 years. In this publication, we introduce the readers to researchers whom have pioneered this field, and those whom have played an essential role in developing this research direction. Now, together, their combined work have elucidated a novel ligandreceptor network that evolved during the earliest period of animal evolution, and has fostered a new insight into the ancient evolutionary organization of the central nervous system (CNS). Specifically, this work offers a new understanding of several aspects of neuropathology including degenerative, psychiatric and mood disorders and, furthermore, illuminates a fundamental role that teneurins and latrophilins play in cell-to-cell metabolism that may be associated with various forms of cancer both within and outside of the brain.

In 1994, the laboratories of Professors Ron Wides in Israel and Ruth Chiquet-Ehrismann working in Switzerland, independently reported the existence of a novel transmembrane protein and its gene in Drosophila. A complex gene/protein, its closest homologue was that of the tenascins. The gene was named either odd oz (odz) or tenascin major (ten-m) by these researchers. Subsequent studies indicated that the gene was highly expressed in the brains of vertebrates and the term 'teneurin' was coined to reflect both its relationship with tenascins and with the CNS. Around the same time as these studies, a novel G protein-coupled receptor was identified by Yuri Ushkaryov and his team in the United Kingdom (in fact the latrophilins then named CIRL, calcium-independent receptor for a-latrotoxin, was first identified by the group of Petrenko at NYU Medical Center in New York, USA), which was subsequently established as a cognate receptor for the teneurins. This receptor was later termed as the latrophilins and more recently 'Adhesion receptor G-protein coupled receptor, family L or ADGRL.

In Part 1 of this publication, the early history on the origin and discovery of teneurins has been described by Stefan Baumgartner and Ron Wides; Ron Wides; and Richard Tucker. Recent structural studies by Verity Jackson and her colleagues, as well as Demet Arae¸ and Jingxian Li have provided molecular models to understand how teneurins are ensconced in the plasma membrane and play a role in synaptic interaction. In addition, their work integrates the molecular mechanisms with the early evolution of both teneurins and latrophilins.

In Part 2, four studies build upon the evolutionary development of teneurins by examining its role in nematodes by Ulrike Topf and Krzysztof Drabikowski, a model of teneurin action in the Drosophilia nervous system by Alison DePew and associates; and two studies on fish. Angela Cheung and her colleagues describe the neurological function and expression in zebrafish, whereas Ross Reid and his coworkers have described novel actions of the teneurins with respect to metabolism in fish.

Part 3 of this publication is focused on the latrophilins and is led off by Yuri Ushkaryov and his team describing the discovery, structure and function of the latrophilins. This work is followed by a review by Ana Moreno-Salinas and colleagues in Antony Boucard´s laboratory describing the structure of the latrophilins and its interaction with associated transmembrane proteins with respect to adhesion, neuronal function and pathology. The following paper, by Torsten Schönberg and Simone Prömel links the previous papers with a comparison of teneurin and latrophilin interactions in invertebrates and vertebrates. Finally, in this section, Peter Burbach and Dimphna Meijer provide an interesting overview of the relationship of teneurins and latrophilins with respect to other proteins described in these other papers. Together, these studies provide a novel understanding of how the teneurins and latrophilins interact in a complex set of associated proteins.

The next section (Part 4) of the publication focuses on the development and maintenance of the CNS in mammals. Here, Catherine Leamey and Atomu Sawatari lead off with a discussion of the role of teneurin-associated neuro-circuit formation using knockout studies in mice. A detailed review by Luciane Sita and her colleagues in the Bittencourt laboratory frames this and previous studies in a comparative neuroanatomical background, and in addition, provides a neuroanatomical rationale for new studies associated with other regions of the CNS. Building upon these studies, David Hogg and his coworkers include a review on the behavioral actions of the teneurin C-terminal associated peptide (TCAP) in mammals and its potential relationship to brain metabolism and forms of neuropathology. Finally, in this section, a study by Gesttner Tessarin in the Casatti laboratory shows for the first time, teneurins may be associated with astrocyte function, indicating a novel function for teneurins with respect to some glial-based disorders in the brain.

Finally in our last section, we have provided some studies on the potential roles of the teneurins and latrophilins with respect to carcinogenesis. Although these studies are somewhat removed from our treatise on the role of teneurins and latrophilins with respect to neuronal development, maintenance and pathology, they provide interesting observations that may be relevant to some types of CNS pathology. Thus, Boris Rebolledo-Jaramillo and Annemarie Ziegler include a review on the relationship of teneurins to several types of cancers. This is followed by a research report by Mia Husić and her colleagues suggesting that the TCAP region of the teneurins could play a role in modulating the adhesion of the cancer-like cell line, HEK293 and finally, Sussy Bastias-Candia and associates have provided novel data on the role of teneurin-3 with respect to Wnt signalling and have discussed its potential role in neural development and carcinogenesis.

Overall, we posit that the teneurins and latrophilins played a major role in the early evolution of the nervous system and may underlie the etiology of a number of neurological disorders that are thus-far misunderstood. Indeed, we hope that this publication will stimulate further research into the actions of teneurins and latrophilins and lead to novel approaches of understanding and ultimately treatment.

#### Obituary: Ruth Chiquet-Ehrismann (1954-2015): A Teneurin Pioneer

A major player in the discovery and characterization of teneurins was the Swiss scientist, Ruth Chiquet-Ehrismann. Dr. Chiquet-Ehrismann had a long-standing interest in cell-cell and cell-extracellular matrix interactions, particularly during development and tumorigenesis. She earned her Ph.D. at the ETH Zurich under the mentorship of David C. Turner, where she performed early work on the cell and heparin-binding sites of fibronectin. Shortly after joining the Friedrich Miescher Institute in Basel as a junior group leader in 1984, Ruth, in collaboration with Eleanor J. Mackie and Teruyo Sakakura, published a paper in Cell describing an extracellular matrix glycoprotein that she named "tenascin". A key observation made in this widely cited paper was the presence of tenascin in the extracellular matrix of embryonic tissues and the stroma of breast cancer, but its absence from most normal adult tissues. We now know that the original "tenascin" was the founding member of a diverse gene family, and that members of this family promote cell motility, proliferation and differentiation in a variety of tissue environments, both normal and pathological.

But in the early 1990s, it was unclear how tenascins functioned. Specifically, its receptors and binding partners were not understood. Subsequently, Ruth engaged in a multi-pronged approach to studying tenascin function in an attempt to identify its homologues in Drosophila. This work, led by her postdoctoral fellow Dr. Stefan Baumgartner, resulted in the discovery of a novel family of type-2 transmembrane proteins that they named ten-a and ten-m, for "tenascin-like proteins accessory and major". When the homologues of ten-a and ten-m were found in vertebrates and they were shown to be highly expressed in the nervous system, Ruth proposed the name "teneurins". This name combined the names of the original proteins from Drosophila with neurons, which appeared to be their most prominent site of expression.

From that point onward, Ruth's research group at the Friedrich Miescher Institute studied two topics: the roles of tenascins in cancer and the roles of teneurins in development. Using numerous model systems, her research included studies of teneurins in arthropods (Drosophila), nematodes (C. elegans) and chordates (birds and humans). Key firsts that came from Ruth's laboratory include the cloning and sequencing of human teneurins, experimental evidence of teneurin processing by furin and the potential nuclear localization of the intracellular domain, the ability of teneurins to promote growth cone spreading, patterning defects in teneurin knockout animals, a description of the ancient origins of teneurins via horizontal gene transfer, the complementary expression patterns of different teneurins during development, the cytotoxic properties of the teneurin C-terminal domain, and the presence of homotypic adhesion domains in teneurins.

Since 1994, Ruth's group published 24 papers on the cloning, expression, origins and functions of teneurins. Contributing to these papers were 15 graduate students and postdoctoral fellows, often with the expert technical guidance of Jacqueline Ferralli, Marianne Brown-Luedi and Doris Martin. This work has provided a foundation for a new generation of researchers in the field of teneurins.

Ruth Chiquet-Ehrismann passed away at her home near Basel on September 4, 2015. She is survived by her husband and collaborator Matthias Chiquet, three children, Daniel, Patrice and Fabian, and an expanding cohort of grandchildren.

Richard P. Tucker Davis, California

Citation: Lovejoy, D. A., Boucard, A. A., Tucker, R. P., eds. (2020). Origins of Human Neuropathology: The Significance of Teneurin-Latrophilin Interaction. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-858-1

# Table of Contents

*08 Editorial: Origins of Human Neuropathology: The Significance of Teneurin-Latrophilin Interaction*

David A. Lovejoy, Antony A. Boucard and Richard P. Tucker

#### PART 1

#### ORIGIN, DISCOVERY AND STRUCTURE OF TENEURINS


Richard P. Tucker


Demet Araç and Jingxian Li

#### PART 2

# TENEURIN EXPRESSION AND FUNCTION IN ESSENTIAL NON-MAMMALIAN MODELS


#### PART 3

#### DISCOVERY AND ORIGINS OF THE LATROPHILINS AND INTERACTION WITH TENEURINS

*100 Catching Latrophilin With Lasso: A Universal Mechanism for Axonal Attraction and Synapse Formation*

Yuri A. Ushkaryov, Vera Lelianova and Nickolai V. Vysokov


#### PART 4

# PHYSIOLOGICAL AND PATHOLOGICAL ROLES OF TENEURINS AND LATROPHILINS IN THE MAMMALIAN CENTRAL NERVOUS SYSTEM


Luciane V. Sita, Giovanne B. Diniz, José A. C. Horta-Junior, Claudio A. Casatti and Jackson C. Bittencourt


#### PART 5

#### TENEURINS AND LATROPHILINS: POTENTIAL ROLE IN CARCINOGENESIS IN BRAIN AND OTHER TISSUES

*208 Teneurins: An Integrative Molecular, Functional, and Biomedical Overview of Their Role in Cancer*

Boris Rebolledo-Jaramillo and Annemarie Ziegler

*228 Teneurin C-Terminal Associated Peptide (TCAP)-1 and Latrophilin Interaction in HEK293 Cells: Evidence for Modulation of Intercellular Adhesion*

Mia Husić, Dalia Barsyte-Lovejoy and David A. Lovejoy

*245 Wnt Signaling Upregulates Teneurin-3 Expression via Canonical and Non-canonical Wnt Pathway Crosstalk*

Sussy Bastías-Candia, Milka Martínez, Juan M. Zolezzi and Nibaldo C. Inestrosa

# Editorial: Origins of Human Neuropathology: The Significance of Teneurin-Latrophilin Interaction

David A. Lovejoy <sup>1</sup> \*, Antony A. Boucard<sup>2</sup> and Richard P. Tucker <sup>3</sup>

<sup>1</sup> Department of Cell and Systems Biology, University of Toronto, Toronto, ON, Canada, <sup>2</sup> Departamento de Biología Celular, Centro de Investigación y de Estudios Avanzados del Instituto Politécnico Nacional, Mexico, Mexico, <sup>3</sup> Department of Cell Biology and Human Anatomy, University of California, Davis, Davis, CA, United States

Keywords: G-protein couple receptor, adhesion, affective disorders, molecular evolution, transmembrane proteins, neural development, neuroplasticity, synapse formation

**Editorial on the Research Topic**

#### **Origins of Human Neuropathology: The Significance of Teneurin-Latrophilin Interaction**

We are delighted to introduce this new special issue on "The Origins of Neuropathology: The Significance of Teneurin-Latrophilin Interaction." Although the title may seem particularly bold, and indeed, perhaps presumptuous, we the editors, think our title well warranted based on the findings and interpretation provided by a dedicated group of researchers who have developed this field over the last 25 years. In this publication, we introduce the readers to researchers whom have pioneered this field, and those whom have played an essential role in developing this research direction. Now, together, their combined work have elucidated a novel ligand-receptor network that evolved during the earliest period of animal evolution, and has fostered a new insight into the ancient evolutionary organization of the central nervous system (CNS). Specifically, this work offers a new understanding of several aspects of neuropathology including degenerative, psychiatric and mood disorders and, furthermore, illuminates a fundamental role that teneurins and latrophilins play in cell-to-cell metabolism that may be associated with various forms of cancer both within and outside of the brain.

In 1994, the laboratories of Professors Ron Wides in Israel and Ruth Chiquet-Ehrismann working in Switzerland, independently reported the existence of a novel transmembrane protein and its gene in Drosophila. A complex gene/protein, its closest homolog was that of the tenascins. The gene was named either odd oz (odz) or tenascin major (ten-m) by these researchers. Subsequent studies indicated that the gene was highly expressed in the brains of vertebrates and the term "teneurin" was coined to reflect both its relationship with tenascins and with the CNS. Around the same time as these studies, a novel G protein-coupled receptor was identified by Ushkaryov et al. in the United Kingdom (in fact the latrophilins then named CIRL, calcium-independent receptor for α-latrotoxin, was first identified by the group of Petrenko at NYU Medical Center in New York, USA), which was subsequently established as a cognate receptor for the teneurins. This receptor was later termed as the latrophilins and more recently "Adhesion receptor G-protein coupled receptor, family L, or ADGRL".

In Part 1 of this publication, the early history on the origin and discovery of teneurins has been described by Baumgartner and Wides, Wides, and Tucker. Recent structural studies by Jackson et al., as well as Araç and Li have provided molecular models to understand how teneurins are ensconced in the plasma membrane and play a role in synaptic interaction. In addition, their work integrates the molecular mechanisms with the early evolution of both teneurins and latrophilins.

Edited and reviewed by:

Hubert Vaudry, Université de Rouen, France

\*Correspondence: David A. Lovejoy david.lovejoy@utoronto.ca

#### Specialty section:

This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Neuroscience

Received: 31 March 2020 Accepted: 21 April 2020 Published: 15 May 2020

#### Citation:

Lovejoy DA, Boucard AA and Tucker RP (2020) Editorial: Origins of Human Neuropathology: The Significance of Teneurin-Latrophilin Interaction. Front. Neurosci. 14:501. doi: 10.3389/fnins.2020.00501

**8**

In Part 2, four studies build upon the evolutionary development of teneurins by examining its role in nematodes by Topf and Drabikowski, a model of teneurin action in the Drosophilia nervous system by DePew et al.; and two studies on fish. Angela Cheung et al. describe the neurological function and expression in zebrafish, whereas Reid et al. have described novel actions of the teneurins with respect to metabolism in fish.

Part 3 of this publication is focused on the latrophilins and is led off by Ushkaryov et al. describing the discovery, structure, and function of the latrophilins. This work is followed by a review by Moreno-Salinas et al. in Antony Boucard's laboratory describing the structure of the latrophilins and its interaction with associated transmembrane proteins with respect to adhesion, neuronal function, and pathology. The following paper, by Schöneberg and Prömel links the previous papers with a comparison of teneurin and latrophilin interactions in invertebrates and vertebrates. Finally, in this section, Burbach and Meijer provide an interesting overview of the relationship of teneurins and latrophilins with respect to other proteins described in these other papers. Together, these studies provide a novel understanding of how the teneurins and latrophilins interact in a complex set of associated proteins.

The next section (Part 4) of the publication focuses on the development and maintenance of the CNS in mammals. Here, Leamey and Sawatari lead off with a discussion of the role of teneurin-associated neuro-circuit formation using knockout studies in mice. A detailed review by Sita et al. in the Bittencourt laboratory frames this and previous studies in a comparative neuroanatomical background, and in addition, provides a neuroanatomical rationale for new studies associated with other regions of the CNS. Building upon these studies, Hogg et al. include a review on the behavioral actions of the teneurin Cterminal associated peptide (TCAP) in mammals and its potential relationship to brain metabolism and forms of neuropathology. Finally, in this section, a study by Tessarin et al. in the Casatti laboratory shows for the first time, teneurins may be associated with astrocyte function, indicating a novel function for teneurins with respect to some glial-based disorders in the brain.

Finally in our last section, we have provided some studies on the potential roles of the teneurins and latrophilins with respect to carcinogenesis. Although these studies are somewhat removed from our treatise on the role of teneurins and latrophilins with respect to neuronal development, maintenance and pathology, they provide interesting observations that may be relevant to some types of CNS pathology. Thus, Rebolledo-Jaramillo and Zieglerinclude a review on the relationship of teneurins to several types of cancers. This is followed by a research report by Husic´ et al. suggesting that the TCAP region of the teneurins could play a role in modulating the adhesion of the cancer-like cell line, HEK293 and finally, Bastias-Candia et al. and associates have provided novel data on the role of teneurin-3 with respect to Wnt signaling and have discussed its potential role in neural development and carcinogenesis.

Overall, we posit that the teneurins and latrophilins played a major role in the early evolution of the nervous system and may underlie the etiology of a number of neurological disorders that are thus-far misunderstood. Indeed, we hope that this publication will stimulate further research into the actions of teneurins and latrophilins and lead to novel approaches of understanding and ultimately treatment.

# RUTH CHIQUET-EHRISMANN (1954-2015): A TENEURIN PIONEER

A major player in the discovery and characterization of teneurins was the Swiss scientist, Ruth Chiquet-Ehrismann. Dr. Chiquet-Ehrismann had a long-standing interest in cellcell and cell-extracellular matrix interactions, particularly during development and tumorigenesis. She earned her Ph.D. at the ETH Zurich under the mentorship of David C. Turner, where she performed early work on the cell and heparin-binding sites of fibronectin. Shortly after joining the Friedrich Miescher Institute in Basel as a junior group leader in 1984, Ruth, in collaboration with Eleanor J. Mackie and Teruyo Sakakura, published a paper in Cell describing an extracellular matrix glycoprotein that she named "tenascin." A key observation made in this widely cited paper was the presence of tenascin in the extracellular matrix of embryonic tissues and the stroma of breast cancer, but its absence from most normal adult tissues. We now know that the original "tenascin" was the founding member of a diverse gene family, and that members of this family promote cell motility, proliferation, and differentiation in a variety of tissue environments, both normal and pathological.

But in the early 1990s, it was unclear how tenascins functioned. Specifically, its receptors and binding partners were not understood. Subsequently, Ruth engaged in a multi-pronged approach to studying tenascin function in an attempt to identify its homologs in Drosophila. This work, led by her postdoctoral fellow Dr. Stefan Baumgartner, resulted in the discovery of a novel family of type-2 transmembrane proteins that they named ten-a and ten-m, for "tenascin-like proteins accessory and major." When the homologs of ten-a and ten-m were found in vertebrates and they were shown to be highly expressed in the nervous system, Ruth proposed the name "teneurins." This name combined the names of the original proteins from Drosophila with neurons, which appeared to be their most prominent site of expression.

From that point onward, Ruth's research group at the Friedrich Miescher Institute studied two topics: the roles of tenascins in cancer and the roles of teneurins in development. Using numerous model systems, her research included studies of teneurins in arthropods (Drosophila), nematodes (C. elegans) and chordates (birds and humans). Key firsts that came from Ruth's laboratory include the cloning and sequencing of human teneurins, experimental evidence of teneurin processing by furin and the potential nuclear localization of the intracellular domain, the ability of teneurins to promote growth cone spreading, patterning defects in teneurin knockout animals, a description of the ancient origins of teneurins via horizontal gene transfer, the complementary expression patterns of different teneurins during development, the cytotoxic properties of the teneurin C-terminal domain, and the presence of homotypic adhesion domains in teneurins.

Since 1994, Ruth's group published 24 papers on the cloning, expression, origins and functions of teneurins. Contributing to these papers were 15 graduate students and postdoctoral fellows, often with the expert technical guidance of Jacqueline Ferralli, Marianne Brown-Luedi, and Doris Martin. This work has provided a foundation for a new generation of researchers in the field of teneurins.

Ruth Chiquet-Ehrismann passed away at her home near Basel on September 4, 2015. She is survived by her husband and collaborator Matthias Chiquet, three children, Daniel, Patrice and Fabian, and an expanding cohort of grandchildren.

# AUTHOR CONTRIBUTIONS

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

**Conflict of Interest:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2020 Lovejoy, Boucard and Tucker. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Discovery of Teneurins

#### Stefan Baumgartner<sup>1</sup> and Ron Wides<sup>2</sup> \*

<sup>1</sup> Department of Experimental Medical Science, Faculty of Medicine, Lund University, Lund, Sweden, <sup>2</sup> The Mina and Everard Goodman Faculty of Life Sciences, Bar-Ilan University, Ramat Gan, Israel

Teneurins were first discovered and published in 1993 and 1994, in Drosophila melanogaster as Ten-a and Ten-m. They were initially described as cell surface proteins, and as pair-rule genes. Later, they proved to be type II transmembrane proteins, and not to be pair-rule genes. Ten-m might nonetheless have had an ancestral function in clockbased segmentation as a Ten-m oscillator. The turn of the millennium saw a watershed of vertebrate Teneurin discovery, which was soon complemented by Teneurin protein annotations from whole genome sequence publications. Teneurins encode proteins with essentially invariant domain order and size. The first years of Teneurin studies in many experimental systems led to key insights, and a unified picture, of Teneurin proteins.

Keywords: teneurin, ten-m, ten-a, TENM, ODZ, latrophilin, type II transmembrane protein, Drosophila

#### Edited by:

Richard P. Tucker, University of California, Davis, United States

#### Reviewed by:

Timothy Mosca, Thomas Jefferson University, United States Krzysztof Drabikowski, Institute of Biochemistry and Biophysics (PAN), Poland

> \*Correspondence: Ron Wides ronwides@yahoo.com

#### Specialty section:

This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Neuroscience

Received: 11 October 2018 Accepted: 27 February 2019 Published: 19 March 2019

#### Citation:

Baumgartner S and Wides R (2019) Discovery of Teneurins. Front. Neurosci. 13:230. doi: 10.3389/fnins.2019.00230

# FLY TENEURINS WERE FIRST DESCRIBED AS CELL SURFACE PROTEINS, AND AS PAIR-RULE GENES

#### Discovery of the Teneurins

The teneurins were discovered in the early 1990 when one of us (SB) tried to find the tenascin-C homologue in Drosophila. Tenascin-C is a six-armed extracellular matrix (ECM) molecule which displays many functions during development, morphogenesis and tissue homeostasis (Midwood et al., 2016). Since the Drosophila genome harbors a solid stock of basement membrane and other important ECM molecules (Broadie et al., 2011), it seemed conceivable to search for a Drosophila homologue of tenascin-C using PCR and degenerate primers. The tenascins are composed of several domains that appear in a repetitive manner such as the tenascin-type of EGF repeats or the fibronectin-type III (FN III) repeats. The carboxy terminus harbors a globular fibrinogen domain. Since all these above mentioned domains were found as parts of other Drosophila proteins, the question was which domain-specific primer pair would turn out to be fruitful. Of the many primers that were used in this approach, only the EGF-like domain proved successful, leading to the detection of the first Drosophila tenascin-type EGF-like repeats. These were then used to screen bacterial cDNA libraries that were optimized for long cDNAs (Brown and Kafatos, 1988; Brown et al., 1989) resulting in three overlapping cDNAs of 7.3 kb in length that altogether constituted a partial sequence of what had the potential to represent the Drosophila homologue of tenascin-C (Baumgartner and Chiquet-Ehrismann, 1993). The deduced amino acid (aa) sequence showed the presence of eight tenascin-type EGF repeats (**Figure 1**), as was the case in vertebrate tenascin-C (Midwood et al., 2016). At the amino terminus, a hydrophobic stretch of amino acids reminiscent of a signal or a transmembrane domain was found. C-terminally of the EGF-like repeat, an additional 100 aa were found that did not show any resemblance to FN III repeats, but soon the protein would run into a stop codon, leaving 4.3 kb of a putative 3<sup>0</sup> untranslated region (UTR). Based on the deduced sequence information, the isolated composite cDNA was

proposed to code for a 782 aa secreted protein and was subsequently called Ten-a (tenascin accessory) (Baumgartner and Chiquet-Ehrismann, 1993). In retrospect, the published Tena aa sequence from 1993 comprised only a partial sequence. This became also evident from comparing the transcript size on a Northern analysis which showed two large transcripts of 11 and 13 kb, respectively, which were developmentally regulated (Baumgartner and Chiquet-Ehrismann, 1993). The discrepancy of the length deduced from available cDNA and the actual transcript size by Northern analysis was attributed to an unusually long 5<sup>0</sup> untranslated region (UTR) which later turned out not to be true. Indeed, it would take years to realize that the protein was indeed much larger (Fascetti and Baumgartner, 2002), because its coding part extended considerably in the carboxy terminal direction. This carboxy extension was also confirmed by the advent of the fully sequenced Drosophila genome (Adams et al., 2000).

(Baumgartner and Chiquet-Ehrismann, 1993) also showed a zoo blot equipped with DNA from Drosophila, leech, zebrafish, chicken, mouse, and human origin, as probed with chicken tenascin-C EGF sequences under low stringency. The blot revealed that the majority of genomes analyzed showed crosshybridizing bands. These findings immediately opened the avenue for further quests/searches for tenascin-type EGF-like sequences not only in Drosophila, but later also in higher organisms (Mieda et al., 1999; Minet et al., 1999; Oohashi et al., 1999; Rubin et al., 1999; Feng et al., 2002).

The Drosophila lane in the zoo blot contained several crosshybridizing bands, two of which could readily be ascribed to Tena. The Drosophila lane, however, revealed further unidentified bands, hence the hunt for further tenascin-EGF-like sequences was continued. To this end, one of us (S. B.) used a Ten-a EGF-like repeat probe and screened Drosophila genomic libraries under low-stringency conditions (McGinnis et al., 1984). Several cross-hybridizing phages were isolated that all mapped to a new locus (Baumgartner et al., 1994). Subsequently, overlapping cDNAs were isolated from this locus and were assembled. These cDNA clones covered two slightly smaller transcripts compared to Ten-a, 10.5 and 11.5 kb in size, respectively. Due to the fact that the protein encoded by the transcript of this new locus was apparently larger than that of Ten-a, this gene was termed Ten-m (tenascin major) (Baumgartner et al., 1994). At the time, it was proposed that the gene encoded a large secreted proteoglycan ECM molecule. Ten-a and Ten-m proteins' structures and domains, as realized in 2018 terms (as described below), can be seen in **Figure 1**.

One of the two Teneurins was independently discovered in Drosophila melanogaster via an alternative approach: (Ten-m, as "odd Oz" by RW), in Levine et al. (1994). In 1990, a screen was carried out to uncover novel fly tyrosine kinase substrates of previously unknown classes. Drosophila proteins were highly immunopurified on an anti-phosphotyrosine antibody column, and the resulting phospho-protein collection was used to raise a bank of monoclonal antibodies. One of these specific monoclonals was directed against a greater than 300 kD protein that was later given the name Odd Oz (Odz, now Tenm) (see **Figure 1**). That monoclonal was used for expression cloning of the Odz/Ten-m's 11 kb transcript from an embryonic cDNA library (Zinn et al., 1988). Further mapping led to genomic cloning, chromosome mapping, mutant identification, and expression and phenotype characterizations (Levine et al., 1994). Two hydrophobic stretches in the predicted protein were interpreted as: (1), a signal peptide before a series of EGF-like repeats, followed by (2), a post-EGF transmembrane domain. The type-I transmembrane model was anchored by placement of the EGF-like repeats extracellularly. Yet this type-I model was also influenced by biases based on the phospho-tyrosine protein screen and consensus phosphorylation site motifs of the time. In fact, the second predicted "transmembrane" domain assignment was incorrect, and the assigned "signal peptide" sequence is the protein's true transmembrane stretch. Odz/Ten-m is instead a type-II transmembrane protein, as all further Teneurins proved to be (see below).

# Expression of the Founding Teneurins in Drosophila

Both Ten-a and Ten-m genes were extensively analyzed with respect to their expression patterns during early Drosophila embryogenesis (Baumgartner and Chiquet-Ehrismann, 1993; Baumgartner et al., 1994; Levine et al., 1994; Fascetti and Baumgartner, 2002; **Table 1**). Predominant expression of both genes was in the central nervous system (CNS). In general, the Ten-m gene showed far more progenitor tissue labeled. Apart from its prominent CNS expression, Ten-m was found in the future tracheal cells, heart cells, lymph gland and hemocytes. Hence, the expression profile demanded further claims to call it "major."

Other studies expanded the breadth of Ten-m expression profiles. Striking expression was documented in non-neuronal imaginal disk tissues, such as ring gland expression (Harvie et al., 1998), in the in sensory and motor neuron precursors in pupae, and in adult neuronal tissues (Levine et al., 1997). Expression in the eye, and influences of upsteam genes such as Glass on Ten-m expression, were observed (Treisman and Rubin, 1996). Hematopoietic cells showed Ten-m expression, such as plasmatocytes (Baumgartner et al., 1994; Braun et al., 1997).

### Drosophila Ten-m/odz and Segmentation

Phenotypically for odz/Ten-m, multiple alleles from independent screens proved allelic, and displayed different severities of a pairrule mutant phenotype (Baumgartner et al., 1994; Levine et al., 1994). This phenotype was very like that of odd paired (opa). Considerably later, the assigned pair-rule phenotype was instead attributed to mutant opa alleles in the genetic backgrounds of the odz/Ten-m strains (see below). On another note - in retrospect, previously created mutations in Ten-a existed that, appropriately, affect the fly brain and behavior. The gene central body defective (cbd), with several alleles known, had been isolated a decade before the gene was cloned and characterized (see **Table 1**; Heisenberg et al., 1985). The recognition that cbd mutations were Ten-a lesions occurred two decades later (Cheng et al., 2013).

In 2006, an indication that odz/Ten-m is not a pair-rule gene was published. Using new technologies that were developed, it

transthyretin; FN, fibronectin; NHL, NCL, HT2A Lin41; YD, YD-repeat motif; ABD, antibiotic binding domain; Tox GHH, Tox GHH fold (Zhang et al., 2012); TCAP, Teneurin C-terminal-associated Peptides; Ig, immunoglobulin.


TABLE 1 | Features of Drosophila Teneurins.

L3, third instar larvae; mat, maternal; cBL, cellular blastoderm; eGA, early gastrulation; TR tracheae, VM visceral mesoderm, MAS, muscle attachment sites; CNS, central nervous system; LG, lymph gland; HE, hemocytes; CB, cardioblast; GC gastric cecae; AMC antenno-maxillary complex; ALG, antennal lobe glomerulus, AORN, adult olfacory receptor neuron; AALPN, adult antennal lobe projection neuron; ED, eye disk; OS, optic stalk, WD, wing disk, CBD, central body defect (cbd) mutant phenotype; EL, embryonic lethality; SL semi-lethal; LL, larval lethality; V, viable; NOP, no obvious phenotype.

was found that an entire 133 kb genomic clone covering Tenm failed to rescue the attributed odz/Ten-m pair-rule phenotype (Venken et al., 2006). The concern arising from this finding led to a re-examination of all odz/Ten-m mutant lines displaying the pair-rule phenotype. Ultimately, the pair-rule phenotype proved to derive from odd paired (opa) mutations on the balancers in odz/Ten-m strains (Zheng et al., 2011).

The different odz/Ten-m mutations, and the balancers in their lines, came from separate mobilized-P-element screens (Cooley et al., 1988; Karpen and Spradling, 1992). The sources of the balancers for these screens were different. In addition, the many non-odz/Ten-m lines examined from these screens, with these balancers, displayed no pair-rule phenotype. The lines that were chosen to assess for odz/Ten-m lesions were based on genome position, and not pair-rule appearance, so phenotype was not a screening bias. Ten-m is deployed as seven stripes during late cellular blastoderm, but its mutants do not have pair-rule phenotypes. To this day, the reasons for the many co-incidences that led to the findings are still unclear. Unfortunately, a great deal of mis-directed work was subsequently carried out. A Ten-a maternal effect impact on segmentation was reported, then was later retracted (Rakovitsky et al., 2007; retracted 2012), despite the correct molecular data detailed there.

#### Ten-m Might Nonetheless Have a Segmentation Role: A Ten-m Oscillator?

One aspect of Ten-m expression was particularly interesting because it showed its transcripts relatively uniformly expressed during cellular blastoderm, while the Ten-m protein only minutes later was detected in seven stripes (Baumgartner et al., 1994; Levine et al., 1994; **Figure 2C**). This observation opened the avenue for proposing a function of Ten-m as an oscillator.

In the past and first documented in the chicken hairy gene, it could be shown that periodically waves can arise from the posterior end of the elongating embryo. These waves move toward the anterior end where they come to a halt and add a segment during each period, as depicted in **Figure 2A** (Palmeirim et al., 1997). Later, a model emerged involving oscillation of the zebrafish hairy/Enhancer of split-related genes, her1 and her7 (Lewis, 2003). This model proposed that the Her protein would bind to its own promoter and inhibit its own transcription. It

mathematical model. (C) Experimental evidence of emergence of Ten-m stripe formation during early Drosophila gastrulation, starting from ubiquitous Ten-m expression. Double-antibody staining reveals Ten-m in red and Fushi tarazu in green (for comparison). Top part shows the transition from ubiquitous Ten-m expression at early gastrulation to the formation of Ten-m stripes at somewhat later gastrulation (as exemplified of the boxed part comprising stripes 3 and 4 and indicated by an arrow). Bottom part shows enlargements of the formation of Ten-m stripes, again exemplified by stripe 3 and 4 formation and the boxed area. Note that Fushi tarazu (green) is already expressed in stripes from the very beginning, in contrast to Ten-m.

was then concluded that the delay would cause a biochemical oscillator because of the time difference between formation of the mRNA and the protein (Lewis, 2003). Posteriorly, cells are fed to the presomitic mesoderm (PSM) during this oscillation. When the embryo undergoes elongation, more oscillating cells are fed into the PSM that will reveal different phases. Subsequently, the cellular oscillation grows until the oscillation comes to a halt with the consequence that segmental borders will emerge. The mechanism is repeated when subsequent cells stop their oscillation. This mechanism enables that segments can be generated, starting from the anterior to the posterior.

The idea that Ten-m could be an oscillator originated from the observation that the Ten-m mRNA was uniformly expressed during nuclear cycle (nc) 14, but once the protein was synthesized, it started to emerge as seven stripes (Baumgartner et al., 1994; Levine et al., 1994; Hunding and Baumgartner, 2017). Since only the long nc 14 was long enough to synthesize the large primary transcript of Ten-m (115 kb, see **Table 2**) and to translate Ten-m (Prescott and Bender, 1962; Baumgartner et al., 1994), it appeared conceivable to assume that stripe formation was tightly linked to translation and to occur only during a limited time, i.e., during late nc 14. nc 14 is terminated once cellularization occurs, whereby the nuclei are wrapped by a membrane. Hence, once cellularization has taken place, signaling from the extracellular space is only possible with the help of a receptor residing at the surface of the cells.


#### TABLE 2 | Comparison of features of Teneurins across phyla.

Model systems used: roundworm, C. elegans; insect, Drosophila melanogaster; ascidian, Ciona intestinalis; chicken, Gallus gallus; mouse, Mus musculus; rat, Rattus norvegicus; human, Homo sapiens. aa, amino acids; kb, kilobase.

As stated above, the Drosophila Ten-m gene encodes a large type II transmembrane protein (**Figure 1**) hence, it is located at the cell surface. Ten-m becomes localized to the membrane which grows from the apical side to the basal side thereby ensheathing the syncytial nuclei (**Figure 2C**). The large extracellular domain of Ten-m may be involved in forming homodimers, as was shown for Ten-a (Fascetti and Baumgartner, 2002) and mouse Teneurins (Feng et al., 2002; Berns et al., 2018). The dynamics of this process has properties proposed to have the potential to create a biochemical oscillator (Hunding and Baumgartner, 2017). Ten-m interaction at the membrane could lead to intracellular cleavage of Ten-m. This cytoplasmic fragment then translocates to the nucleus. As alluded to above, Ten-m is not transcribed in seven stripes, but rather appears fairly homogeneous along the A-P axis. The mechanism to solve this apparent discrepancy is so far not clear. However, it was proposed that the intracellular mechanism of the interplay between the protein and the membrane may lead to a spontaneous pattern-forming mechanism, as was reported from other biochemical oscillators (Hunding and Baumgartner, 2017). In fact, Ten-m fulfills most criteria of stripe formation based on a model originally described for prokaryotic cell division (Hunding and Engelhardt, 1995) and further developed by (Meinhardt and de Boer, 2001). This model has recently been recapitulated using in vitro data and expanded models (Loose et al., 2008).

Thus, the Ten-m oscillator is not caused by delayed translation as in the case for the zebrafish her1/her7 genes (Lewis, 2003), but could arise from cooperative membrane binding (Hunding and Baumgartner, 2017). To enable Ten-m to function as a signaling molecule, a mechanism was proposed that would involve regulated intramembrane proteolysis (RIP) (Brown et al., 2000; McCarthy et al., 2017). Indeed, reports could show that an intracellular tail part of vertebrate Teneurin 2 protein is proteolytically cleaved off, possibly via RIP. This short peptide is then translocated to the nucleus where it represses zic-1, a vertebrate counterpart of odd-paired (opa), a pair-rule gene in Drosophila (Bagutti et al., 2003). On the other hand, ubiquitously expressed Zic-1 leads to fast degradation of the short Teneurin 2 signaling peptide. The model of Hunding and Baumgartner (2017) included thus cooperative interaction of Ten-m with the membrane, intracellular cleavage and degradation (**Figure 2B**).

In summary, what these data would like to suggest is that, despite the fact that Ten-m mutants do not show a segmental phenotype, there might be an ancestral function of Ten-m in clock-based segmentation. The one established by the Notch signaling system was probably when the insects evolved, due to the fact that Notch receptor does not show an involvement in Drosophila segmentation. This is where Ten-m might come in and the field is eagerly waiting for data that support this hypothesis.

### WATERSHED OF VERTEBRATE TENEURIN DISCOVERY, 1998–2000, IN THE PRE-VERTEBRATE-GENOME ERA

The first publications of vertebrate Teneurin genes emerged from screens searching for other phenomena: studies of cancer rearrangements and gene changes; olfaction-related genes; and ER stress-related CHOP genes. A gene rearrangement encoding a fusion protein containing Teneurin 4 and Neuregulin 1 domains was identified in human breast tumor tissue (Schaefer et al., 1997). The resulting fusion protein contained only the pre-EGF amino-terminal portion of TENM4, but beyond ESTs, was the first harbinger of vertebrate Teneurins, as was later recognized (Wang et al., 1999). Soon thereafter, human Teneurin 1 was sequenced and named TNM (**Figure 3**), when it was found adjacent to the X-linked lymphoproliferative syndrome causative gene SH2D1A (Coffey et al., 1998). Teneurin 4 in mouse was uncovered in a screen for CHOP - dependent stress-induced genes, and was named DOC4 (Wang et al., 1998). Tenm4 came up twice in that screen (as DOC4 and DOC5), and was the first non-fly Teneurin to receive non-cursory treatment, with a number of pivotal observations made for the protein family as a whole. Some time later, first phenotypes for mouse Teneurin 4 were documented, when it was established that existing l7Rn3 mice were mutants (Lossie et al., 2005). In a search based on homology to E2 cysteine rich loops of odorant receptors, rat Neurestin (Teneurin 2) was found as a novel, non-odorant receptor, protein (Otaki and Firestein, 1999a). The characterization of Teneurin 2 in Neurestin papers also contributed key observations made for the protein family as a whole (Otaki and Firestein, 1999a,b).

Meanwhile, efforts directed specifically at identifying and cloning vertebrate Teneurin by homology to the fly genes were underway in three species, and were reported in 1999 and 2000. The four paralog types in chicken were identified, Tenm1–Tenm4 (Minet et al., 1999; Rubin et al., 1999), and this work continued with many wide-ranging discoveries and publications. Four corresponding mouse paralogs were found and well characterized (Oohashi et al., 1999), and were also independently sequenced and mapped (Ben-Zur and Wides, 1999; Ben-Zur et al., 2000). Two of the four of these paralog types were also uncovered in zebrafish (Mieda et al., 1999). The rat and human Teneurin genes mentioned above were retrospectively assigned to their

paralog-type number. Thus, at the end of the "pre-vertebrate genome sequence" era, five vertebrate species had been proven to bear Teneurins, with a four-copy content apparent as the common, and likely conserved, paralog complement (**Figure 3**).

# ANALYSIS FROM THE FIRST COMPLETE GENOMES: TENEURINS FORM A DISTINCT, ANIMAL, FAMILY

With the completion of the Caenorhabditis elegans genome, a single full length Teneurin, Ten-1, was evident (**Figure 3**; C. elegans Sequencing Consortium, 1998). Its protein and function were characterized and described (Drabikowski et al., 2005). The Drosophila melanogaster genome encoded a Ten-a the length of the original Ten-m (Adams et al., 2000), as was later described (Fascetti and Baumgartner, 2002). A nematode singleton Teneurin, in contrast to a pair of paralogs, Ten-a and Ten-m, in insects, held true in the nematode C. briggsae (Stein et al., 2003), the mosquito Anopheles gambiae (Holt et al., 2002), and the silkworm Bombyx mori (Mita et al., 2004), genomes. Vertebrate genomes, including human (Lander et al., 2001; Venter et al., 2001), mouse (Mouse Genome Sequencing et al., 2002; Mural et al., 2002), rat (Gibbs et al., 2004), and chicken (International Chicken Genome Sequencing Consortium, 2004) validated that the four paralogs Tenm1, 2, 3 and 4 were a fixed vertebrate feature. Contemporaneously, non-vertebrate chordates, such as the ascidian Ciona intestinalis, proved to have a single Teneurin gene (Dehal et al., 2002). This indicated that Teneurins were quadruplicated sometime during chordate or early vertebrate evolution. A representative list of these Teneurin genes appears as **Table 2**. The proteins, which all maintain the same domain order, are of roughly the same size. Their protein lengths are reflected in their mature transcript sizes. Their nascent transcripts, however, are consistently of unusually large size, as is often seen for highly developmentally regulated genes. As a consequence, the Teneurin genes' lengths occupy an "over-sized" fraction of total genome sizes (**Table 2**).

In contrast, Teneurins were not found in the kingdoms of plants or fungi. The earliest sequenced genomes of: the yeasts Saccharomyces cerevisiae (Goffeau et al., 1996) and Saccharomyces pombe (Wood et al., 2002); plants Arabidopsis (The Arabidopsis Genome Initiative, 2000) and rice (Goff et al., 2002; Yu et al., 2002), and the first other protists and fungi revealed no Teneurins. No eukaryotic homologous sequences could be found at all, outside of those to the Teneurin's EGF-like domains. The only other Teneurin domains with homology to any proteins were rhs (recombination hot spot)-like elements otherwise found only in a small number of bacteria (Minet et al., 1999; Minet and Chiquet-Ehrismann, 2000).

Overall, in the first 10 years that Teneurins were studied, they were recognized as animal specific genes (**Figure 3**), with two paralogs in insects, and four paralogs in vertebrates. These were reviewed with an eye toward an evident ancient duplication, and an evident ancient quadruplication, in insects and vertebrates, respectively, (Tucker and Chiquet-Ehrismann, 2006; Tucker et al., 2007). These reviews also recognized that Teneurin proteins

are largely invariant throughout evolution, with no domain content or order variation. For a more recent evolutionary history of the family, see the Wides article in this volume. For more recent views of Teneurins and their structure, see the Tucker (2018), and DePew et al. (2019) article, in this volume.

#### FROM TENEURIN'S FIRST DECADE: KEY INSIGHTS INTO ITS PROTEIN

The Drosophila Teneurin homologs were initially described as ECM molecules (Baumgartner et al., 1994), or as type I transmembrane proteins (Levine et al., 1994). The first recognition that Teneurins were in fact type II single pass transmembrane proteins came for mouse Teneurin4, when it was discovered as DOC4 (Wang et al., 1998). This was validated rigorously when all four mouse genes were sequenced, and when their extracellular portions were imaged by electron microscopy (Oohashi et al., 1999). Several studies established that Teneurins are deployed to the cell membrane as protein dimers, but their full homo- and hetero-dimerization combinatorial repertoire was first methodically shown in mouse (Feng et al., 2002). Among protein – protein interactions proven for Teneurins, perhaps the first was the fly Ten-m RGD motif interaction with integrins (Graner et al., 1998). Broader still, and iconic for Teneurin function, was the discovery of homophilic interactions in chicken (Rubin et al., 2002). While a great deal still needs to be done to nest Teneurins within a complete pathway, their homophilic, and cross-paralog-homophilic, extracellular contacts are at the heart of their signaling role. Proteolytic cleavages at many sites by many proteases are also central to varied aspects of Teneurin protein function, and have been documented since the first works published. They are too numerous to be related here, but two perhaps suggest the most important functional implications. The cleavage and release of intracellular domains, and their freedom to then enter the nucleus to impact transcription was

#### REFERENCES


first described in chicken (Bagutti et al., 2003; Nunes et al., 2005). The cleavage of their extreme carboxy-terminal amino acids to yield independent, biologically active TCAPs (Teneurin C-terminal-associated Peptides) occurs in many important systems (Qian et al., 2004).

Interestingly, the studies on the Teneurin domain structure were recently complemented by two reports showing data based on crystallization and cryo-EM analyses, respectively (Jackson et al., 2018; Li et al., 2018). These in principle confirmed the sequence-based data, however, they revealed that most central domains merged into a large and centrally located 200 kD superfold. They also disclosed new findings, e.g., by highlighting the NHL domain (**Figure 1**) as a particularly well exposed domain where homophilic interactions between teneurins were ascribed (Berns et al., 2018). Moreover, alternative splicing within the NHL domain would allow modulation of this homophilic interaction. Based on sequence comparisons, both Ten-a and Ten-m follow the domain structure that the most-recent crystallization and cryo-EM data defined. Hence, the domain structure as drawn in **Figure 1** likely holds true. Evolutionarily, the 200 kD superfold was adopted as a whole structure from bacteria. This was recognized in these two papers Jackson et al. (2018), Li et al. (2018), and in Ferralli et al. (2018). Teneurin's Latrophilin binding, and its implications, was discovered well after the first decade, and is extensively treated in other articles in this volume.

#### AUTHOR CONTRIBUTIONS

SB and RW collaborated and contributed equally to this paper.

### ACKNOWLEDGMENTS

SB thanks the Swedish Research Council, the Ekhaga, Nilsson-Ehle and Erik Philip-Sörensen Foundations for support.




**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Baumgartner and Wides. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The Natural History of Teneurins: A Billion Years of Evolution in Three Key Steps

#### Ron Wides\*

The Mina and Everard Goodman Faculty of Life Sciences, Bar-Ilan University, Ramat Gan, Israel

The entire evolutionary history of the animal gene family, Teneurin, can be summed up in three key steps, plus three salient footnotes. In a shared ancestor of all bilaterians, the first step began with gene fusions that created a protein with an amino-terminal intracellular domain bridged via a single transmembrane helix to extracellular EGF-like domains. This first step was completed with a further gene fusion: an additional carboxyterminal stretch of about 2000 amino acids (aa) was adopted, as-a-whole, from bacteria. The 2000 aa structure in Teneurin was recently solved in three dimensions. The 2000 aa region appears in a number of bacteria, yet was co-opted solely into Teneurin, and into no other eukaryotic proteins. Outside of bilaterian animals, no Teneurins exist, with a "Monosiga brevicollis caveat" brought below, as 'the third footnote." Subsequent to the "urTeneurin's" genesis-by-fusions, all bilaterians bore a single Teneurin gene, always encoding an extraordinarily conserved Type II transmembrane protein with invariant domain content and order. The second key step was a duplication that led to an exception to singleton Teneurin genomes. A pair of Teneurin paralogs, Ten-a and Tenm, are found in representatives of all four Arthropod sub-phyla, in: insects, crustaceans, myriapods, and chelicerates. In contrast, in every other protostome species' genome, including those of all non-Arthropod ecdysozoan phyla, only a single Teneurin gene occurs. The closest, sister, phylum of arthropods, the Onychophorans (velvet worms), bear a singleton Teneurin. Ten-a and Ten-m therefore arose from a duplication in an urArthropod only after Arthropods split from Onychophorans, but before the splits that led to the four Arthropod sub-phyla. The third key step was a quadruplication of Teneurins at the root of vertebrate radiation. Four Teneurin paralogs (Teneurins 1 through 4) arose first by a duplication of a single chordate gene likely leading to one 1/4– type gene, and one 2/3-type gene: the two copies found in extant jawless vertebrates. Relatively soon thereafter, a second duplication round yielded the −1, −2, −3, and −4 paralog types now found in all jawed vertebrates, from sharks to humans. It is possible to assert that these duplication events correlate well to the Ohno hypothesized 2R (two round) vertebrate whole genome duplication (WGD), as refined in more recent treatments. The quadruplication can therefore be placed at approximately 400 Myr ago. Echinoderms, hemichordates, cephalochordates, and urochordates have only a single copy of Teneurin in their genomes. These deuterostomes and non-vertebrate chordates provide the anchor showing that the quadruplication happened at the root

#### Edited by:

Richard P. Tucker, University of California, Davis, United States

#### Reviewed by:

Timothy Mosca, Thomas Jefferson University, United States Liqun Luo, Stanford University, United States

> \*Correspondence: Ron Wides ronwides@yahoo.com

#### Specialty section:

This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Neuroscience

Received: 02 October 2018 Accepted: 29 January 2019 Published: 15 March 2019

#### Citation:

Wides R (2019) The Natural History of Teneurins: A Billion Years of Evolution in Three Key Steps. Front. Neurosci. 13:109. doi: 10.3389/fnins.2019.00109

of vertebrates. A first footnote must be brought concerning some of the 'invertebrate' relatives of vertebrates, among Deuterostomes. A family of genes that encode 7000 aa proteins was derived from, but is distinct from, the Teneurin family. This distinct family arose early in deuterostomes, yet persists today only in hemichordate and cephalochordate genomes. They are named here TRIPs (Teneurin-related immense proteins). As a second of three 'footnotes': a limited number of species exist with additional Teneurin gene copies. However, these further duplications of Teneurins occur for paralog types (a, m, or 1–4) only in specific lineages within Arthropods or Vertebrates. All examples are paralog duplications that evidently arose in association with lineage specific WGDs. The increased Teneurin paralog numbers correlate with WGDs known and published in bony fish, Xenopus, plus select Chelicerates lineages and Crustaceans. The third footnote, alluded to above, is that a Teneurin occurs in one unicellular species: Monosiga brevicollis. Teneurins are solely a metazoan, bilaterian-specific family, to the exclusion of the Kingdoms of prokaryotes, plants, fungi, and protists. The single exception occurs among the unicellular, opisthokont, closest relatives of metazoans, the choanoflagellates. There is a Teneurin in Monosiga brevicollis, one species of the two fully sequenced choanoflagellate species. In contrast, outside of triploblast-bilaterians, there are no Teneurins in any diploblast genomes, including even sponges – those metazoans closest to choanoflagellates. Perhaps the 'birth' of the original Teneurin occurred in a shared ancestor of M. brevicollis and metazoans, then was lost in M. brevicollis' sister species, and was serially and repeatedly lost in all diploblast metazoans. Alternatively, and as favored above, it first arose in the 'urBilaterian,' then was subsequently acquired from some bilaterian via horizontal transfer by a single choanoflagellate clade. The functional partnership of Teneurins and Latrophilins was discovered in rodents through the LPH1-TENM2 interaction. Recent work extends this to further members of each family. Surveying when the interacting domains of Teneurins and Latrophilins co-exist within different organisms can give an indication of how widespread their functional cooperation might be, across bilaterians. Paralog number for the two families is relatively correlated among bilaterians, and paralog numbers underwent co-increase in the WGDs mentioned above. With co-increasing paralog numbers, the possible combinatorial pairs grow factorially. This should have a significant impact for increasing nervous system complexity. The 3 key events in the 'natural history' of the Teneurins and their Latrophilin partners coincide with the ascendance of particularly successful metazoan clades: bilaterians; arthropods; and vertebrates. Perhaps we can attribute some of this success to the unique Teneurin family, and to its partnership with Latrophilins.

Keywords: Arthropod, vertebrate, urBilaterian, Latrophilin, EGF, Ecdysozoa, chordates, TRIP (Teneurin-related immense protein)

#### INTRODUCTION

Teneurin family genes made their world debut about a billion years (Byr) ago (as argued below), and made their scientific debut 25 years ago in Drosophila melanogaster (Baumgartner and Chiquet-Ehrismann, 1993). Many aspects of Teneurins have been reviewed (Tucker et al., 2012; Ziegler et al., 2012; Leamey and Sawatari, 2014; Mosca, 2015; Woelfle et al., 2016). The family's evolution has been well covered, with a broader understanding emerging from each newly found homolog and each newly completed genome (Tucker et al., 2012).

This present overview of Teneurin evolution benefits from two new game-changing sources of information. First, recent prokaryotic, and eukaryotic (especially those of metazoans), whole genome sequences beneficially fill in many previous evolutionary gaps. Mining those 'gaps' sheds considerable light on key steps in Teneurin evolution. Second, two publications have recently solved the three dimensional structure of a large extracellular portion of Teneurins (Jackson et al., 2018; Li et al., 2018), and another recognized this as a domain block (Ferralli et al., 2018). We can now relate to the domains of these proteins with definitiveness never before possible. In comparing Teneurin homologs and near relatives, we have better focus and clarity for analyzing and dissecting Teneurin relationships.

As never before, this new information streamlines the story of the Teneurin family. I posit that the approximately one billion years of Teneurin history can be described as 3 essential steps: birth of urTeneurin in the urBilaterian; creation of the duplicates Ten-a and Ten-m in the urArthropod; and a quadruplication, in the earliest vertebrates, to create paralogs Tenm-1–Tenm-4. Otherwise, the shape and list of Teneurins is virtually invariant for 1 Byr.

Yet this history has three caveats that serve as tantalizing footnotes to the streamlined story. First, a distinct family of genes encoding 7,000 amino acid (aa) proteins 'spunoff' from Teneurins in early deuterostomes. Second, some further copies of paralogs have emerged, but only in select vertebrates and arthropods, evidently associated with whole genome duplications. Third, a non-bilaterian Teneurin exists in one of the two sequenced choanoflagellates.

#### MATERIALS AND METHODS

Annotated gene, transcript and protein records were collected for Teneurins. To be comprehensive, this was carried out for whole genome projects' annotations. This was complemented with searches of NR to collect data from the directed cloning of genes, genome and transcriptome annotations, and more.

Whole assembled genomes were analyzed directly for evidence of Teneurin gene numbers, for validating annotations, and for detecting missed annotations or partial annotations. This mostly centered on probing genome assemblies with tblastn, searching with the longest protein sequences from closest relevant species. Most effective was the use of: Teneurin protein

fnins-13-00109 March 13, 2019 Time: 18:14 # 3

sequences with the EGF repeats excised; complete-Teneurin protein sequences; and TRIP protein sequences. Isolated EGFlike sequences and different specific Teneurin domains were used for specialized searches.

Unassembled genome traces (sequence reads that are unassembled, for any reason, or "sequence chaff ") were analyzed directly. This mostly centered on employing tblastn using protein sequences from closest relevant species. These searches were performed for evidence of Teneurin gene annotations missed, due to their absence from the genome assemblies. The effort in fact often focused on searches for portions of Teneurins that were missing in the assemblies, and therefore in the annotations.

Sequences were aligned with Clustal X, were compared by PAUP<sup>∗</sup> (Phylogenetic analysis using parsimony), and were then displayed in an unrooted phylogram. In other cases, sequence alignments were done with Clustal Omega, and then were largely shown as rooted cladograms.

#### RESULTS AND DISCUSSION

### The Entire Evolutionary History of the Animal Gene Family Teneurin Can Be Summed Up in Essentially Three Key Steps: Sections I, II, and III

(I) The First Step: The First Teneurin Arose From a Unique Series of Fusions in a Shared Ancestor of All Bilaterians, Leading to Teneurin-Singleton-Genomes Teneurin, the animal gene family, has members widely reported in triploblast-bilaterians (Tucker et al., 2012; Mosca, 2015), but

never in diploblasts. Every sequenced bilaterian genome has at least one Teneurin, but no trace of them can be found in sequenced diploblast genomes to date. The only non-bilaterian Teneurin gene is found in one choanoflagellate, in the phylum considered to be the closest living phylum to that of the animal kingdom (Tucker et al., 2012). That exception is treated below as "footnote three" in section V.

Teneurins encode type II transmembrane proteins composed of four distinct domain component regions: an intracellular domain; a single trans-membrane spanning domain; a extracellular domain with EGF-like repeats and its associates; and carboxy-terminally, a roughly 2000 aa 'multi-domain entity' whose three dimensional structure has been solved (**Figure 1**, Jackson et al., 2018; Li et al., 2018). Prior to these independent, and highly concurring, 3D solutions of this 2000 aa "super-fold" (Jackson et al., 2018), extracellular Teneurin domains were described using varied borders and domain names. The 2000 aa super-fold can be treated as one discrete unit in an evolutionary discussion, in that: it was adopted, all domains en masse, from a precursor gene encoding a bacterial protein; and it is essentially invariant among all Teneurins.

It is therefore most likely that the first Teneurin arose from an ancient small number of gene fusions that brought together the four mentioned component regions. A pair of fusions of a gene encoding a trans-membrane helix: to one encoding EGFlike repeats, C-terminally; and to one encoding an intracellular stretch, N-terminally; led to the type II membrane-orientation now found for every Teneurin (**Figures 1A,B**). The intracellular and transmembrane domains are highly variable and are poorly conserved between Teneurins. Therefore no homology can be found, nor relationships inferred, between these domains and non-Teneurin proteins' domains.

In contrast, the affinities between Teneurin EGF-like repeats and analogous domains in other proteins and protein familiescan be examined. EGF-like repeats are nearly exclusively an animal phenomenon (Wouters et al., 2005). A number of EGF-likedomain bearing proteins have family orthologs that populate and 'straddle' both protostome and deuterostome genomes, so were likely present in the 'urBilaterian.' From among those ancient 'cross-bilaterian' proteins, the EGF-like repeats of Teneurin are most like those of Delta, Serrate/Jagged, Notch, Wif1/Shifted, Eyes-shut, Crumbs, Integrin beta2, and Slit. They are a subset of proteins that contain hEGFs, rather than cEGFs (Wouters et al., 2005). The gene fusion to include these EGF-like repeats into urTeneurin must have taken place in the urBilaterian, most likely by adopting EGF-repeats from one of these proteins. The protein bearing the most similar EGF-like block is most likely the one whose gene Teneurin "adopted from." The eight EGF-like repeat content of Teneurin probably arose from a combination of: the integration of several adopted EGF-like encoding repeats as-ablock, plus accretion of repeats via tandem duplications (and possible repeat loss). Proposing candidate protein families as the most likely donors of Teneurin's EGF-repeats will likely require modeling and testing a combination of these processes. Note: it is formally possible that Teneurins adopted EGFs from some no longer existing family, rather than one among the cluster above.

There is only one set of EGF-like repeats more similar to those of Teneurins, rather than those of the proteins listed above: the Tenascins. Tenascins arose in chordates, and are absent in any other phyla of animals, or life (Tucker and Chiquet-Ehrismann, 2009; Adams et al., 2015). Since Tenascins' EGFs are closest to those of Teneurins, Tenascin's EGFs most likely arose from deuterostome-chordate, Teneurin EGF-like domains. Genes encoding these EGFs must have fused with those encoding fibronectin type-III, and fibrinogen, domains, during Tenascin's genesis. Historically, Teneurins therefore take their name from their 'offspring.' Another conceivable scenario is that Tenascins adopted EGFs from some no longer existing family, or from proteins extant only in lower chordates. There will be further discussion of this below, in section III.

Proceeding carboxy terminally, the remainder of Teneurin is the 2000 aa 'super-fold.' Since it is solved by cryo-EM and X-ray crystallography, the super-fold can now be coherently described using the nomenclature of these structure-solution papers. The terms used now are able to anchor the domains within the super-fold onto clearly-delineated 3D folds. However, the folds are named differently by the two groups: TTR-FN-plug/Iglike; NHL/beta-propeller; YD-shell/Barrel, ABD, and Tox GHH [see **Figure 1B**, as superimposed basically on chicken/human Teneurin-2, (Jackson et al., 2018; Li et al., 2018)]. Subsequent to, or concomitant with, the gene fusions described above, this greater than 2000 aa structure was co-opted as-a-whole from bacteria, and was fused downstream of the EGF-like repeats.

fnins-13-00109 March 13, 2019 Time: 18:14 # 4

The wholesale adoption of this huge block of domains was not recognizable until recent bacterial genomes were sequenced. Some 100 newly sequenced genomes of bacterial species, collected mostly since 2016, encode proteins with this 2000 aa block. In the 100 cases, sequences of Teneurin (for instance of M. brevicollis or Ten-m of D. melanogaster) have about 70% of their length covered at more than 25% aa identity to the prokaryotic proteins. The homology starts just after the Teneurin EGF repeats, and covers nearly the entire post-EGF length of Teneurin. Five of the best hits are presented in **Figure 1C**. One of these hits (**Figure 1C**) was recognized as an analogous prokaryotic 'super-fold' containing protein in Bacillus subtilis strain CW14 in the structure paper (Jackson et al., 2018). The Teneurin extra-cellular regions' structural homology to bacterial Tc-toxins is pivotal in the other structure paper (Li et al., 2018). A recognition that the homology between Teneurin and bacterial proteins extends to a full 2000 aa was also made in Ferralli et al. (2018). The Desulfurivibrio alkaliphilus YD protein discussed there also appears among the best hits shown here in **Figure 1C**. Note that before these recent abundant new bacterial genes were sequenced and recognized, Teneurins at best had RhS and YD-repeat portions with noticeable homology to bacteria. These previous alignments were shorter and localized (Minet and Chiquet-Ehrismann, 2000; Tucker et al., 2012). Those alignments did not suggest a wholesale adoption of the post-EGF portion of Teneurins as a block.

#### **The overall outlook**

A limited number of gene fusions led to the first Teneurin, in an ancestor of both Deuterostomes and Protostomes. Its original domain content and order strongly resist changes, as seen for all family members.

#### (II) The Second Step: A Single Gene Duplication at the Root of Arthropod Radiation Gave Rise to Paralogs Ten-a and Ten-m

The genomes of protostomes contain a single Teneurin gene (Tucker et al., 2012) (**Figure 2**). The marked exception occurs in a single clade, the Arthropods, in all of its species' genomes. The genomes of species in the Arthropod Phylum contain two Teneurin paralogs, Ten-a and Ten-m (**Figures 2**, **3B**). Representatives shown for each sub-phylum (**Figures 2**, **3**) have a Ten-a, and a Ten-m, gene: the fruit fly Drosophila melanogaster for insects in the sub-phylum of Hexapoda; the tick Ixodes scapularis for the class of arachnids, in the subphylum Chelicerata; the water flea Daphnia pulex for the class of branchiopoda, in the sub-phylum Crustacea; and the millipede Strigamia maritima for the class of millipedes, in the

sub-phylum Myriapoda. A single duplication of an ancestral protostome Teneurin gene before divergence of the four subphyla, yielding Ten-a and Ten-m, is consistent with the trees' geometries. The relative phylogenetic distances between Ten-a and Ten-m are appropriate to the accepted evolutionary history of the four sub-phyla (Borner et al., 2014; Chipman et al., 2014; Misof et al., 2014).

All genomes of species from other protostome phyla contain a single Teneurin (**Figures 2**, **3**). This statement includes all lophotrochozoa. More importantly for this discussion, it also includes all non-arthropod genomes examined from among the ecdysozoan cluster phlya of protostomes. Ecdysozoan genomes are generally widely represented, despite the fact that there are still some ecdysozoan phyla with no sequenced genomes: Nematophora (horsehair worms); Kinorhyncha (mud dragons); and Loricifera. A 'Teneurin-singleton' only is found in every sequenced species of non-arthropod ecdysozoans, in: Priapulida; Nematodes; Onychophorans, and Tardigrades (**Figures 3A,B**).

The most diagnostic information for timing the Ten-a/Ten-m duplication comes from the closest, sister, phylum of Arthropods,

Onychophora. It has one species with a sequenced genome, Euperipatoides rowelli (NCBI BioProject 203089, Georg Mayer at the Baylor i5K initiative, as described in Evans et al., 2013). This Onychophoran velvet worm E. rowelli has a single Teneurin gene, roughly equally distant from fly Ten-a and from Ten-m (**Figure 3**). Onychophorans have an estimated divergence date from Arthropods not long before the Arthropod splits to subphyla (Sanders and Lee, 2010). The generation of a Teneurin paralog type occurred once in the history of invertebrates, exclusively in Arthropods, at their inception as a phylum.

#### **To summarize**

Genomes published to date firmly point to a single very ancient Teneurin duplication during all of protostome evolution, in the urArthropod. The four hugely populated extant sub-phyla of the arthropods then inherited the two paralogs from the protourArthropod ancestor. The duplication thus occurred in the proposed brief period between the Onychophoran/Arthropod split and the split of Arthropods to four sub-phyla (or perhaps to more sub-phyla, including trilobites) – in the early Cambrian, in the Ediacaran, or earlier yet. Previous reports that did not detect this deep origin of Ten-a and Ten-m were dependent on many fewer completed arthropod genomes (Tucker et al., 2012).

As described below in IV, the only other instances, whatsoever, of more Teneurin genes in protostomes occur in select Arthropods. These are additional Ten-a and Ten-a copies that arose much later, in limited lineages of chelicerata and crustaceans, as the result of evident whole genome duplications (WGDs). As documented below, appearance of these additional Ten-a's and Ten-m's correlate to known WGDs.

#### **The overall outlook**

Teneurins are unchanging over evolutionary time. An invariant protein architecture, together with only a single lasting duplication over 700 Myr of protostome evolution, attest to a conserved role and unchanging placement in an otherwise dynamic interactome/proteome landscape.

#### (III) The Third Step: Events Early in Chordate-Related Lineages, Plus Later Independent Vertebrate Duplications That Gave Rise to the Vertebrate Paralogs Teneurin 1, 2, 3, and 4

The third milestone in the history of Teneurins, after their appearance in the urBilaterian, and after their duplication in the urArthropod, occurred in deuterostomes around the establishment of the chordate phylum. Most central and relevant to our topic was a quadruplication in vertebrates that led to the four long-recognized Teneurin paralog types in higher vertebrates: 1, 2, 3, and 4. The invariant existence of these four paralog types in all higher vertebrates has already been well reviewed (Tucker et al., 2012; Leamey and Sawatari, 2014; Mosca, 2015).

The window to reconstruct earlier events for deuterostome Teneurins has now opened yet further, due to recently completed genomes. This includes views of proto-chordates: Ambulacraria (Echinoderms plus Hemichordates) and cephalochordates (**Figure 4A**). This also includes Tunicates (or urochordates), and 'lower and higher' – jawed and unjawed – vertebrates. As described below in IIIC of this section, the duplications of bona fide Teneurins only began in jawless vertebrates. However, a Teneurin-derived distinct family, the TRIPs, arose in deuterostomes before the appearance of vertebrates. There are extant TRIPs only among hemi- and cephalo- chordates. This TRIP sidetrack occurred first, and its narrative is treated first, in sections IIIA and IIIB.

#### **(IIIA) The Echinoderm phylum and the Hemichordate phylum. A TRIP sidetrack**

At their divergence from protostomes, deuterostomes also had an evident single Teneurin gene in their genomes. The several sequenced genomes of Echinoderms: including those of starfish, brittle stars, sea urchins and sea cucumbers, all have one Teneurin gene. This strongly supports a 'singleton' content for deuterostomes at their emergence.

In Hemichordates, the sister phylum of Echinodermata, a single Teneurin gene exits. This is the case for the two Hemichordate species with sequenced genomes, the Enteropneusta class acorn worms': Harrimaniidae family's Saccoglossus kowalevskii; and Ptychoderidae family's Ptychodera flava. Each encodes a typical size Teneurin protein of approximately 2800 aa that is roughly equally homologous to arthropods' paralogs Ten-a and Ten-m, or alternatively, to vertebrates' paralogs Tenm-1, 2, 3, and 4 (**Figure 4** and below). The remaining class of Hemichordates, Pterobranchians, have not been sequenced, so their Teneurin content is unknown.

However, there is an additional, novel, occurrence in both sequenced Hemichordate species that is striking. These Hemichordates contain genes encoding approximately 7000 aa polypeptides whose amino-terminal portions are unequivocally related to Teneurins (**Figure 4**). The genes in the two species are clearly homologous. I have named them Teneurin-related (or derived) immense proteins (TRIPs). Following approximately 2600 Teneurin-homologous amino acids, these polypeptides' sequences continue carboxy-terminally with more than 4000 amino acids with no similarity to any eukaryotic proteins. Any similarity to bacterial proteins is in short scattered stretches, at very low homology. TRIPs evidently arose early in deuterostome/proto-chordate evolution from a duplicate of the ancestral singleton Teneurin, which then soon fused with a gene encoding very large protein coding domains. S. kowalevskii contains a single TRIP, and Ptychodera flava contains 3 TRIP paralogs that derive from further very ancient duplications. The S. kowalevskii protein shares more than 35% amino acid identity for over 6000 aa when compared to each of the three P. flava proteins.

Teneurin-related immense proteins' polypeptides are among the largest ever described. They represent a large departure from Teneurins, qualifying them as a distinct protein family. The TRIP Teneurin-like-portions bear a lower level of homology to vertebrate and other deuterostome Teneurins than do the typical 2800 aa, bona fide, Teneurins of these Hemichordates (**Figure 4C**). This large TRIP/Teneurin distance shown in **Figure 4C** is based on the alignment of Teneurins with the first 2700 aa portions of TRIPs. Also rendering them distinct, the TRIPs lack amino acid stretches that are invariant according to

#### FIGURE 4 | Continued

fnins-13-00109 March 13, 2019 Time: 18:14 # 9

One of the two B. belcheri protein annotations had to be constructed 'manually' from adjacent contigs in the genome assembly. The three hemichordate TRIP proteins from the hemichordate acorn worm Ptychodera flava are not included in this alignment, but are fully homologous, and are much closer to the S. kowalevskii sequence. The cephalochordate Asymmetron lucayanum TRIP is not shown, but is more homologous to the other lancelets' TRIPs. (C) Deuterostome Teneurins are aligned together with TRIP proteins. Only the N-terminal 2700 aa of the TRIPs are used for the alignment. TRIPs' sequences that overlap Teneurins strongly partition away from Teneurins. The species used include those in (B), plus the sea star Acanthaster planci, the sea urchin Stronglyocentrotus purpulatus, the tunicate Ciona intestinalis, and the four mouse Teneurins. (D) A Clustal Omega alignment of Teneurins from agnathans (jawless vertebrates), sharks, and fish. The species used were Rhincodon typus (elephant shark), Rhincodon typus (whale shark), Danio rerio (zebrafish), Eptatretus burgeri (agnathan, hagfish), and Petromyzon marinus (agnathan, lamprey). Two Teneurin homologs exist in the two agnathans, and were called "TenmA and TenmB." The TenmB protein of lamprey was too incomplete to use effectively, so it was not included in the alignment. The two Teneurins aligned from hagfish, and the one aligned from lamprey are incomplete protein annotations.

the structures described in the publications establishing Teneurin structure (Li et al., 2018; Jackson et al., 2018). Conjectures about TRIP function appear below, in Section IIIB.

#### **(IIIB) Non-vertebrate chordates: The TRIPs continue and end**

At the root of the chordate phylum, all genomes have a single bona fide Teneurin. Starting with the "proto-chordates," the chordate sub-phylum Cephalochordata has three lancelet species with sequenced genomes: Asymmetron lucayanum; Branchiostoma floridae; and Branchiostoma belcheri. Each has a Teneurin (**Figure 4A**). At the same time, these cephalochordates bear TRIP genes (**Figures 4A–C**). Their phylogenetic relationships show that the cephalochordate TRIPs share their origins with the event that gave rise to Hemichordate TRIP genes, and are their orthologs (see **Figure 4B**, where the entire 7000 aa TRIP protein sequences are used for alignments). Two of these three lancelet species have a single TRIP. The third, Branchiostoma belcheri, has two TRIP proteins that share 98% amino acid identity with each other. This TRIP paralog pair is likely a duplication solely in B. belcheri, occurring after it diverged from B. floridae (**Figure 4B**, Yue et al., 2016).

Like cephalochordates, the chordate Tunicate sub-phylum has genomes bearing a single Teneurin gene (**Figure 4**). Unlike cephalochordates, however, urochordate/Tunicate genomes demonstrate no evidence of TRIP genes whatsoever. This is demonstrated in diverse examples distributed among three distant groupings within the Tunicate sub-phylum (**Figure 4A**). Of the Thaliacia class pelagic swimmers Salps, Salpa thompsoni has a single Teneurin and no TRIP. This genomic content is mirrored in an Appendicularia class pelagic swimmer Larvacean, Oikopleura dioica, in three Ascidian Phlebobranch sessile sea squirts Ciona intestinalis, Ciona savignyi, Phallusia mammillata, and a Stolidobranch ascidian Botryllus schlosseri (Voskoboynik et al., 2013). Unequivocally, TRIPs were lost on the path of the lineage to urochordates, after their divergence from more basal proto-chordates. As described below, this appears to be a momentous loss, as TRIPs also do not appear in any vertebrate (see **Figures 4A–C**, and below).

On TRIP structure: the first assumption is that TRIPs are type II proteins, but their sub-cellular deployment and function must be established and proven with further work. Some TRIP protein sequences have a transmembrane hydrophobic stretch in the position corresponding to that seen in Teneurins. Other TRIPs have no such transmembrane stretches. Missing transmembrane domains might be identified later when TRIP transcripts are isolated, once they are re-examined and better annotated. Likewise, TRIPs do not bear intracellular domains (IC) that correspond to those of Teneurins. Here too, though, better annotation of these genes, transcripts, and proteins might reveal that they contain IC stretches. Given the low sequence conservation of Teneurin ICs, this will take considerable work to investigate and establish. At the same time, however, some of the TRIP protein sequences do have a clear trans-membrane domain followed by EGF-like repeats, which begs the question if they can dimerize with bona fide Teneurins resident in the relevant species. It is also reasonable to ask if the additional extracellular domains, of more than 4000 amino acids, might render them more effective adhesion proteins? An additional question arises: might TRIPs 'retain' the ability to interact with LPHNs, or might there be some 'replacement' LPHN-like alternative for TRIPs.

On Tenascin evolution: it is possible that the EGF-like domains adopted by Tenascins have an alternative source – TRIPs, rather than Teneurins. Tenascins first appeared in lower chordates, so the source of their EGFs that fused with other relevant domains have two alternative closest "contemporaneous" sources: Teneurin EGFs or TRIP EGFs. An analysis of EGF-like domains of Tenascins, Teneurins, and TRIPs indicated that TRIP is the closer and more likely EGF-domain contributor (data not shown). Alignments and homology must be carried out more extensively, however, with care given to EGF-repeats as blocks. In the end, Teneurins might ultimately be named after their 'grandchildren' among their 'offspring.' The evolutionary sequence might well be: Teneurins gave rise to TRIPs, which then contributed EGFs to the 'assembly' of Tenascins. TRIPs were in the right time and phyla to have contributed EGF domains to the nascent family, Tenascins.

#### **To summarize**

The most parsimonious explanation in the early history of deuterostomes is that 'the TRIP sidetrack' began in the shared ancestor of Ambulacraria (Echinoderms plus Hemichordates) and Cephalochordates (**Figure 4A**). TRIPs rapidly became a distinct, non-Teneurin entity. The echinoderm branch evidently then specifically lost them. Subsequently, TRIPs persisted in cephalochordates, but were lost a second time in the entire Tunicate/Vertebrate lineage. Therefore TRIPs are exclusive to Hemicordates and Cephalochordates, (**Figure 4A**). Unlike Teneurins, they don't appear to provide a function essential to general metazoan survival, since they are specific to select Deuterostomes, and proved dispensable on the journey to vertebrates. It is a challenge to consider what shared lifestyle needs constrained hemichordates and cephalochordates from losing TRIPs. Only when the character and expression of those TRIP c-terminal 4300 amino acids are studied can the unique role of TRIPs, as confined to these protochordates, be probed.

#### **The overall outlook**

fnins-13-00109 March 13, 2019 Time: 18:14 # 10

Non-vertebrate chordates did not generate lasting duplicate Teneurin gene paralogs. Instead, a duplication soon followed by fusion to an enormous extracellular domain addition gave rise to a distinct family: TRIPs. The Teneurin-derived sequence of the proto-TRIP drifted significantly, so TRIPs are clearly identifiable as outliers to the homology range seen for Teneurins, and make no contribution to the history of Teneurin lineages.

#### **(IIIC) Vertebrates and the quadruplication of Teneurin to yield the Tenm1 – Tenm4 paralogs**

The earliest vertebrates – the agnathan, jawless, hagfish and lampreys - - each have genomes with two Teneurins, and no TRIPs. A single hagfish, Eptatretus burgeri, from the agnathan Myxiniformes order has a sequenced genome with two Teneurin genes (**Figures 4A,D**). The two sequenced lampreys, Petromyzon marinus and Lethenteron camtschaticum, from the agnathan Petromyzontiformes order, each also have two Teneurin genes (**Figure 4A**). All of these are equally distant to the bona fide Teneurin sequences in lower Deuterostomes, e.g., to the Tunicate C. intestinalis. This can correlate well with the model that these agnathans have undergone a single round of WGD, relative to basal chordates (Dehal and Boore, 2005; Caputo Barucchi et al., 2013; Berthelot et al., 2014). One step more modern in evolution, in jawed vertebrates, all four Teneurin paralog types exist, as has been extensively previously reported: Tenm-1 through Tenm-4 are seen in all species. Examples included here expand this foursome to Chondrichthyes (e.g., sharks and rays) and bony fish (**Figure 4D**). Thus, very early in the history of jawed vertebrates, a second duplication round yielded the −1, −2, −3, and −4 paralog types now found in all vertebrates. The two agnathan genes, named TenmA and TenmB for **Figure 4**, are present in every agnathan genome examined. However, the adequately complete protein annotations chosen for a meaningful multialignment were only P. marinus TenmA plus E. burgeri TenmA and TenmB (**Figure 4D**). The two agnathan paralogs A and B are roughly equally distant to the four higher vertebrate homologs. Agnathan TenmA and TenmB may partition to one 1/4-type and one 2/3 type, once better annotations and alignments are carried out. If so, the four Teneurin vertebrate paralogs (Teneurins 1 through 4) can be modeled as arising first by a duplication that led to the two genes seen in jawless hagfish and lampreys – one 1/4–type gene, and one 2/3-type gene, and thereafter by a second round duplicating each of these two. Whether or not it will be proven that one agnathan gene is a 'more 2/3-type,' and the other is a 'more 1/4 type,' the most parsimonious view is that these two led to the four higher vertebrate paralogs. In any case, a net quadruplication occurred within the short timeframe of the emergence of jawless, then jawed, vertebrates.

The Teneurin quadruplication that occurred at the root of vertebrates is integral to explaining the four extant vertebrate Teneurin paralog types. This net 'double duplication' happened long after the early TRIP 'sidetrack' that occurred in more ancient Deuterostomes that is described above. It can be argued that "true Teneurin" duplication events correlate well with the Ohno hypothesized 2R (2 round) vertebrate whole genome duplication (WGD) (Ohno, 1970). This hypothesis that early vertebrates underwent two rounds of WGD, followed by gene loss for a majority of duplicates, has been refined in many works (McLysaght et al., 2002; Dehal and Boore, 2005; Caputo Barucchi et al., 2013; Berthelot et al., 2014). Again, those works model agnathans as having undergone only the first of the two whole genome duplication rounds. The timing of the quadruplication is carefully modeled at approximately 400 Myr ago (McLysaght et al., 2002).

#### **The overall outlook**

The three steps of Teneurin family evolution harbored only three deep duplications in nearly a billion years of bilaterian evolution: one in the urArthropod; and two in very early vertebrates. The Byr time point is based on dates most ascribed to the divergence of protostomes and Deuterostomes (Nei et al., 2001; Blair and Hedges, 2005). This is a story of a protein and family profoundly resistant to change. Vertebrate duplications only occurred when the whole proteome complement duplicated. In fact, the only further Teneurin duplications observable in sequenced genomes are restricted to specific Vertebrate and Arthropod lineages, and appear to also be associated with Whole Genome Duplications (WGDs). They are isolated events, and are footnotes in comparison to the pivotal duplications described above as "steps 2 and 3" in Teneurin history.

### The Three Key Steps of Teneurin Evolution Are Accompanied by Three Important Footnotes. The 'TRIP Sidetrack' Footnote Is in Sections IIIA and B, Above. The Other Two Are Sections IV and V

#### (IV) Further Duplicates of Teneurins Are WGD Associated. Still an Unusually Unchanging Family

Beyond the two paralogs in arthropods and four paralogs in vertebrates, further Teneurin duplications are rare in the genomes sequenced to date, and only occur in a few very select Vertebrate and Arthropod lineages. The cases are limited to bony fish (teleosts) and amphibians, among vertebrates. None occur in cartilaginous fish, nor in the more modern vertebrates, Amniotes. Among Arthropods, they only occur in certain Chelicerata and Crustaceans. These duplications have arisen significantly more recently than the deep duplications described above, and always yield clearly identifiable paralog types 1, 2, 3, 4, a, or m.

The best known of these rare cases are in fish, modeled to have undergone teleost specific whole genome duplications (WGDs) (Pasquier et al., 2016). Well documented are the six Teneurins of the zebrafish Danio rerio, with a second Teneurin 2 paralog, and a second Teneurin 3 paralog (Howe et al., 2013). These duplicates share 81 and 74% aa identity, respectively, to their sister 2 and 3 genes, versus about 63% aa identity between paralogs Teneurin 2 and Teneurin 3. These extra paralog copies


FIGURE 5 | Additional Teneurin paralog copies have arisen in association with Whole Genome Duplications (WGDs) in specific vertebrates and arthropods. (A) A list of Teneurin genes in the diploid frog Xenopus tropicalis and in the allotetraploid Xenopus laevis. They are all shown with their chromosomal locations. X. laevis has duplicates of every Teneurin, and each one is located in syntenically conserved positions, relative to X. tropicalis. (B) A Clustal Omega alignment showing the duplication of Teneurins in the Chelicerate horseshoe crab Limulus polyphemus, that has two Ten-a and two Ten-m genes. It is compared to Onychophoran Teneurin, and to Teneurins of fly, tick, and water flea. As in Figure 3, the species used are Ixodes scapularis (Ixodes Tick), the fruit fly Drosophila melanogaster (Fly), and the water flea Daphnia pulex (Daphnia-pul).

for Teneurins 2 and 3 are presumably those that persisted after a WGD, with loss of the Teneurin 1 and 4 duplicate-genes having occurred. Teneurin duplications are more extensive in Rainbow trout, whose genome is modeled as having undergone teleost specific, then salmonid lineage specific, WGDs (Berthelot et al., 2014). These two more recent lineage specific WGDs occurred long after the 2R genome quadruplication at the root of vertebrates. Rainbow trout have up to 16 Teneurins, all of identifiable types, Tenm1 – Tenm4. In amphibians, an even clearer case is seen when comparing diploid Xenopus tropicalis to the allotetraploid Xenopus laevis (Riadi et al., 2016). Like most vertebrates, X. tropicalis bears the four Teneurin vertebrate paralogs. X. laevis has clear second copies for each paralog (**Figure 5A**). This is especially compelling, as the eight X. laevis genes are nested in the expected syntenic surroundings, with pairs on the related chromosomes (allotetraploid pairings), as implicit in their chromosome names (**Figure 5A**).

Arthropods have the only other lineages showing duplications beyond those described in "Teneurin history steps two and three." No further duplications are detectable in two of the four arthropod sub-phyla: neither in the one sequenced myriapod Strigamia maritima, nor in the hundreds of sequenced insects. Instead, additional duplications beyond the generation of the original urBilaterian Ten-a/Ten-m genes occur only in specific lineages of Chelicerates and Crustaceans.

In chelicerates, more than one paralog of Ten-m and Tena occur in specific mites, spiders, and horseshoe crabs. Both paralogs appear as multiples in all three sequenced scorpions and false scorpions. This is in contrast to the single Tena and Ten-m paralogs in the deer tick Ixodes scapularis, the

Lyme disease vector known not to have undergone further WGDs. For the mite, spider, and horseshoe crab genomes with multiple Ten-m and Ten-a copies, the two paralog archetypes appear to always have the same duplication timing. This is based on the equivalent divergence tree profiles seen among the Ten-a and Ten-m duplicates, and likely indicates WGDs. An example is shown for the horseshoe crab Limulus polyphemus, that has two Ten-a and two Tenm genes (**Figure 5B**). Compellingly, Limulus polyphemus and the other mentioned species have recently been fully sequenced, and are modeled as having had lineage specific WGDs (Nossa et al., 2014; Kenny et al., 2016; Schwager et al., 2017). The WGDs in horseshoe crabs, that occurred significantly after the original Ten-a/Ten-m split, are themselves still considerably ancient duplications (>135 Myr) (Nossa et al., 2014; Kenny et al., 2016). There is also a modeled deep specific WGD shared by spiders and scorpions (Schwager et al., 2017). These include the common house spider, Parasteatoda tepidariorum, with 6 Teneurins: three of each paralog (Schwager et al., 2017). A list of species examined is found online as **Supplementary Table S1**.

This story repeats itself in Crustaceans, with many distributed species showing more than one Ten-a and more than one Ten-m. It is notable that the class Branchiopoda of crustacea is one clade without any Teneurin duplications. In contrast, there are crustaceans with the two paralogs both duplicated or triplicated, with apparently the same timing. This strongly supports WGD. However, unlike for the case of Chelicerates, no crustacean genome sequenced to date has been modeled as having undergone WGD. This will require further investigation in the coming years to assess the full nature of these crustacean genomes and duplications.

#### **To summarize**

Duplications of Teneurins, after those that generated the longrecognized paralog types, appear restricted to cases of whole genome duplications. Those later duplications have occurred in specific Arthropod and Vertebrate lineages. They are the only metazoan phyla established in the literature to have undergone WGDs, outside of single-Teneurin lophotrochozoan Rotifers (Snell et al., 2015). After the Ten-a/Ten-m duplication, it appears that without exception, new Teneurin paralog generation is driven by whole genome duplication. The Ten-a/Ten-m duplication itself is so ancient in arthropod evolution that no known WGD can be invoked with which it could be associated.

#### **The overall outlook**

There appears to be little tolerance to generate and utilize new paralogs of Teneruins. It was already clear that Teneurin's structure is unchanging. Now, as an extension: Teneurins also only change their 'count' when all other components of the proteome/connectome are duplicated, in WGDs. It is as if additional Teneurins are not suffered unless the entire proteome framework is preserved, via copying the entire endeavor. As such, Teneurin content is rigid relative to the proteome or connectome. It suggests that Teneurins themselves increase only when all their potential interactors increase too.

#### (V) Teneurins Are a Bilaterian-Only Metazoan Family, Except for One Choanoflagellate Clade

Teneurins are a bilaterian-only metazoan gene family, to the exclusion of the Kingdoms of prokaryotes, plants, fungi, and protists. The intriguing only exception comes from among the unicellular, opisthokont, closest relatives of metazoans, the choanoflagellates. Full genome sequences exist for only two choanoflagellates (Hoffmeyer and Burkhardt, 2016). There is a Teneurin in Monosiga brevicollis, as has been reported (Tucker et al., 2012). In contrast, the genome of the closely related Salpingoeca rosetta has no Teneurin.

To put these two contrasting choanoflagellate findings in context, it must be recognized that outside of triploblastbilaterians, no animal Teneurins exist. There are no Teneurins in any diploblast genomes, including even sponges - those metazoans closest to choanoflagellates. This makes M. brevicollis' gene strikingly unique, and all the more intriguing. Perhaps its existence bespeaks horizontal transfer. One of the two most parsimonious explanations for M. brevicollis' gene's existence is: the original Teneurin was 'born' in the shared ancestor of M. brevicollis and metazoans, then was lost in every diploblast metazoan lineage. This is possible, but is not an entirely satisfying explanation, due to the need for serial and multiple losses to "clear-the-board" in every branch. Alternatively, and favored above, is a model where urTeneruin was 'fusion-assembled' in the 'urBilaterian,' then was subsequently acquired by a single choanoflagellate clade, via horizontal transfer. Recently, nineteen additional Choanoflagellates had their transcriptomes extensively sequenced and compared (Richter et al., 2018). Evidence of Teneurins occur in 3 of the 19 species, essentially all clustered in one clade. As a comparison, signatures for domains of Notch and its ligands are found in the 19 transcriptomes, as well as in the two genome-sequenced choanoflagellate species. The conclusion reached is that Notch appears to be 'indigenous' to choanoflagellates, with its creation predating

TABLE 1 | Teneurin and Latrophilin paralog numbers in animal genomes.


the choanoflagellate/metazoan split (Richter et al., 2018). In contrast, Teneurin distribution supports the idea that a metazoan Teneurin entered M. brevicollis and its clade with three other species, by horizontal transfer.

#### **The overall outlook**

fnins-13-00109 March 13, 2019 Time: 18:14 # 13

The evidence is strongly biased toward an urBilaterian origin of Teneurin, followed by horizontal transfer into Monosiga brevicollis' choanoflagellate clade.

#### Teneurins and Latrophilins-Co-prevalence and the Importance of Their Co-existence: VI and VII (VI) LPHN1-TENM2 Are the First Among Functional Partners. Expanded Families Offer Expanded Combinatorials for Interactions

The functional partnership of Teneurins and Latrophilins was discovered in rodents through the LPHN1-TENM2 interaction (Silva et al., 2011). Further work extends this to further family members, with demonstrations that all 3 latrophilins bind all 4 Teneurins, in Mouse (Boucard et al., 2014). A survey of where the interacting domains of Teneurins and Latrophilins co-exist within different organisms can give an indication of how widespread their functional cooperation might be across bilaterians. From protostomes to deuterostomes, how many of each exist?

First, what defines the Latrophilins, and where do they occur? Latrophilins exist as an animal-only gene subfamily within the greater seven-transmembrane GPCR family. They are part of the Adhesion GPCRs, of which there are 33 in humans. Adhesion GPCRs are one of five main groups according to the GRAFS classification (see Hamann et al., 2015). The Adhesion GPCRs are ancient, and are believed to have evolved from the cAMP receptor family, arising approximately 1,275 million years ago, before the split of Unikonts to animals and fungi (Schioth et al., 2010; Nordstrom et al., 2011; Krishnan et al., 2012). Among the adhesion-GPCRs, Latrophilins constitute family I out of 9 (I-IX) (Hamann et al., 2015). Family I LPHNs 1 through 3 are formally known as ADGR1-3, with the less related ELTD1 known as ADGRL4. There are no HRM, Olfaction, or RB-Lectin domains in ELTD1/ADGRL4, making it a distinct outlier. LPHN1 was the first Latrophilin cloned (Lelianova et al., 1997). Again, the interaction with Teneurin 2 was discovered as: LPHN1 to Teneurin2 in 2011, as mentioned above (Silva et al., 2011).

LPHNs are animal only genes that are distributed more broadly than Teneurins. In diploblasts, there is a strong possibility that homologs exist that truly act as LPHNs (Krishnan et al., 2014). Examples of these homologs include sponge Amphimedon queenslandica Aq715659, and anemone Nematostella vectensis Nv24490. Whatever the function of these proteins, it must be Teneurin-independent (**Table 1**).

Protostomes all seem to have one Latrophilin only (**Table 1**). Exceptions are a duplicate in the Ixodes tick, and a possible two for Caenorhabditis elegans, Lat-2, and perhaps also Lat-1, For insects, a singleton clear Latrophilin exists. Therefore a LPHN duplication does not appear to co-occur with the Ten-a plus Tenm generated paralogs in insects. Some insect LPHNs seem to have a more complete domain content than others. Nonetheless, the Drosophila melanogaster homolog Cirl, evidently missing many expected domains, has clearcut functions in flies (Scholz et al., 2015; Scholz et al., 2017).

In non-vertebrate Deuterostomes, there are to one to two LPHNs, although generally only one has all domains to make a firm case for likely function (Krishnan et al., 2013). However, from chordates to vertebrates, a co-increase between Teneurins and Latrophilins can be tracked (**Table 1**). Most intriguingly, the clades of vertebrates and arthropods that underwent WGD and duplication of Teneurin paralogs have a definite trend of equivalent increases for the LPHNs. This can be seen for Xenopus laevis, Zebrafish, Rainbow trout, and Chelicerates (**Table 1**).

The LPHN genes' duplications at the root of vertebrates/chordates could have co-arisen with the vertebrate whole genome 2R quadruplication proposed by Ohno (see Krishnan et al., 2015). WGDs in restricted arthropods and vertebrates appear to co-copy, and co-retain, their balanced collections of Teneurins and Latrophilins.

#### **The overall outlook**

The genome appears to duplicate Teneurins together with LPHNs. A balance of Teneurins to LPHNs, and to the overall content of the proteome and connectome, might need to be maintained. This suggests that the Teneurin content ratio is rigid, relative to gene counts in the genome, and especially to LPHN gene counts.

#### (VII) The Success of Vertebrates and Arthropods, and the Possible Contribution of These Two Families to That Success

There are three key steps to the Teneurin evolution story, the birth-by-fusions, then the very ancient duplication events. Three important footnotes serve either to bolster the character of those steps (e.g., conserved, specific additional duplications), or are not central to the generation of Teneurins as we know them (e.g., the TRIP spinoff, and a choanoflagellate's homolog).

Teneurins are rigidly and extraordinarily conserved, both in their unchanging structure, and in their gene copy number per genome. The lineages where new paralogs arose are in the two most successful phyla on the planet: arthropods, and our own – chordate vertebrates. Is it possible that the addition of Teneurin paralogs to the gene toolkit contributed to the special success of these phyla? Can the overall success of triploblasts-bilaterians, compared to diploblasts, partially be attributed to the presence of Teneurin? Does a relatively fixed ratio of Teneurin to Latrophilin gene copies, conserved even when further arthropod and vertebrate WGDs occur, attest to the special advantage that this 'team' jointly contributes to a metazoan? These are all questions that could be further probed in the future, now that these balances have been uncovered.

The basic outline of this story should now be complete, and is even arguably comprehensive. Given the major gaps now filled in for the tree of animal life, large brushstrokes

expected to change this story in the future seem unlikely. Albeit, some chapters of this story need improvement. A betterdefined indication of the source of EGF-like repeats that were incorporated into Teneurin in the urBilaterian would be informative. An answer on whether Teneurins versus TRIPs provided Tenascins with their EGF-like domains should be more clear-cut with better modeling. Establishing a clearer relationship between agnathans' two Teneurins and higher vertebrates' four Teneurins is attainable with further work. Genome sequences of additional Choanoflagellates, and other unicellular opisthokonts, should shed more light on the significance of the Monosiga brevicollis Teneurin. The question: "just how rigidly conserved is the Teneurin family, relative to other gene families?", should be interesting to model and probe. However, for all of these questions, it can only be expected that new genomes,

#### REFERENCES


and further connections within the omniome, will deliver new surprises.

#### AUTHOR CONTRIBUTIONS

The author confirms being the sole contributor of this work and has approved it for publication.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnins. 2019.00109/full#supplementary-material

an ancient fold for cell-cell interaction. Nat. Commun. 9:1079. doi: 10.1038/ s41467-018-03460-0


analyses reveal the overall hierarchy for most of the G protein-coupled receptor families. Mol. Biol. Evol. 28, 2471–2480. doi: 10.1093/molbev/msr061


**Conflict of Interest Statement:** The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Wides. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

fnins-13-00109 March 13, 2019 Time: 18:14 # 15

# Teneurins: Domain Architecture, Evolutionary Origins, and Patterns of Expression

#### Richard P. Tucker\*

Department of Cell Biology and Human Anatomy, University of California at Davis, Davis, CA, United States

Disruption of teneurin expression results in abnormal neural networks, but just how teneurins support the development of the central nervous system remains an area of active research. This review summarizes some of what we know about the functions of the various domains of teneurins, the possible evolution of teneurins from a bacterial toxin, and the intriguing patterns of teneurin expression. Teneurins are a family of type-2 transmembrane proteins. The N-terminal intracellular domain can be processed and localized to the nucleus, but the significance of this nuclear localization is unknown. The extracellular domain of teneurins is largely composed of tyrosine-aspartic acid repeats that fold into a hollow barrel, and the C-terminal domains of teneurins are stuffed, and least partly, into the barrel. A 6-bladed beta-propeller is found at the other end of the barrel. The same arrangement—6-bladed beta-propeller, tyrosine-aspartic acid repeat barrel, and the C-terminal domain inside the barrel—is seen in toxic proteins from bacteria, and there is evidence that teneurins may have evolved from a gene encoding a prokaryotic toxin via horizontal gene transfer into an ancestral choanoflagellate. Patterns of teneurin expression are often, but not always, complementary. In the central nervous system, where teneurins are best studied, interconnected populations of neurons often express the same teneurin. For example, in the chicken embryo neurons forming the tectofugal pathway express teneurin-1, whereas neurons forming the thalamofugal pathway express teneurin-2. In Drosophila melanogaster, Caenorhabditis elegans, zebrafish and mice, misexpression or knocking out teneurin expression leads to abnormal connections in the neural networks that normally express the relevant teneurin. Teneurins are also expressed in non-neuronal tissue during development, and in at least some regions the patterns of non-neuronal expression are also complementary. The function of teneurins outside the nervous system remains unclear.

Keywords: ABC toxin, brain, development, horizontal gene transfer, odz, teneurin, YD protein

# INTRODUCTION

Teneurins are type-2 transmembrane proteins with a variable N-terminal intracellular domain and a large, phylogenetically conserved extracellular domain. The extracellular domain features epidermal growth factor (EGF)-like domains, a 6-bladed beta-propeller composed of NHL repeats, tyrosine-aspartic acid (YD) repeats, a rearrangement hot spot (RHS) core protein domain and a C-terminal domain related to both GHH toxins and corticotropin-releasing factor (**Figure 1A**).

#### Edited by:

Hubert Vaudry, Université de Rouen, France

#### Reviewed by:

Timothy Mosca, Thomas Jefferson University, United States Robert Hindges, King's College London, United Kingdom E. Joop Van Zoelen, Radboud University Nijmegen, Netherlands

#### \*Correspondence:

Richard P. Tucker rptucker@ucdavis.edu

#### Specialty section:

This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Neuroscience

Received: 26 September 2018 Accepted: 28 November 2018 Published: 11 December 2018

#### Citation:

Tucker RP (2018) Teneurins: Domain Architecture, Evolutionary Origins, and Patterns of Expression. Front. Neurosci. 12:938. doi: 10.3389/fnins.2018.00938

**35**

The genomes of most vertebrates include four related teneurin genes encoding teneurins numbered 1 through 4 (Tucker et al., 2012). In Drosophila melanogaster there are two teneurins, tena and ten-m (Baumgartner et al., 1994; Levine et al., 1994), and in Caenorhabditis elegans there is a single teneurin, ten-1 (Drabikowski et al., 2005). This review will concentrate on what is known about the domain organization of the best studied teneurins, what can be inferred about their evolution from studies of extant genomes, and patterns of teneurin expression.

#### TENEURIN DOMAIN ORGANIZATION

#### The Teneurin Intracellular Domain

The teneurin intracellular domain typically includes one or more proline-rich SH3-binding domain and one (or more) predicted nuclear localization sequence (**Figure 1A**). Yeast two-hybrid screens and co-immunoprecipitation experiments demonstrated that one of the SH3-binding domains from teneurin-1 binds CAP/ponsin (Nunes et al., 2005). CAP/ponsin, also known as sorbin, is a widely expressed adaptor protein involved in the organization of the cytoskeleton and growth factor-mediated signaling (Kioka et al., 2002). The intracellular domain of teneurin-1 also binds to MBD1, a methylated DNA binding protein (Nunes et al., 2005), but the biological significance of this interaction is unknown. When the intracellular domain is overexpressed in tissue culture cells it is found in the nucleus where it co-localizes with PML protein in nuclear bodies (Bagutti et al., 2003; Nunes et al., 2005). In chicken embryos antibodies to the intracellular domain of teneurin-1 often stain the cell nucleus in regions where antibodies to the extracellular domain stain the cell surface (Kenzelmann et al., 2008; Kenzelmann Broz et al., 2010), suggesting that teneurins may be processed so that the intracellular domain can be released for yet-to-be determined function in the nucleus. A likely site for proteolytic cleavage within the intracellular domain is the conserved basic sequence motif RKRK. When fibroblasts are transfected with native teneurin-1, antibodies to the intracellular domain of teneurin-1 stain the nucleus, but they do not stain the nucleus if the cells are transfected with a teneurin-1 following mutation of the basic motif to AAAA (Kenzelmann et al., 2008). The potential for this type of processing has recently been confirmed by others (Vysokov et al., 2016). Finally, there are many alternatively spliced variants of the intracellular domains of teneurins from chicken and human (Tucker et al., 2012), but the biological significance of these variants is unknown.

#### EGF-Like Domains

Most teneurins have eight EGF-like domains starting approximately 200 amino acids C-terminal to their transmembrane domain (Tucker et al., 2012). The Basic Local Alignment Search Tool (BLAST) reveals that these domains, which have the conserved consensus sequence Ex2Cx(D/N)x2Dx(D/E)xDx3DCx3(D/E)CCx4Cx5C (where "x" is any amino acid), are most similar to those found in the tenascin family of extracellular matrix glycoproteins. This explains why teneurins were first identified in a low stringency screen of Drosophila DNA with a probe based on the EGF-like domains of chicken tenascin-C (Baumgartner and Chiquet-Ehrismann, 1993). The names given to the Drosophila teneurins, ten-a and ten-m, reflect this historical connection to tenascins. In turn, the name "teneurin" is a conflation of "ten-a/ten-m" and "neurons," which are a major site of teneurin expression (Minet et al., 1999). Note that teneurins were discovered independently in D. melanogaster and named odd Oz (Levine et al., 1994), which accounts for the alternative name Odz for teneurins in the literature and in some genome search engines.

One well-established function of the teneurin EGF-like domains is to permit dimerization in cis. Most teneurin EGFlike domains have six cysteines that form three pairs of disulfide bonds. However, the second and fifth teneurin EGF-like domains have only five cysteine residues. The odd number allows cysteines in one teneurin to make disulfide bonds with cysteines in a neighboring teneurin, resulting in covalently linked side-by-side dimers (Oohashi et al., 1999; Feng et al., 2002). This explains the distinctive "pair of cherries" appearance of the extracellular domain of teneurins when viewed in the electron microscope after rotary shadowing: the stems are the attached EGF-like domains, and the cherries are the remaining C-terminal part of the extracellular domain (Feng et al., 2002).

#### Beta-Propeller Domain

The central region of the teneurin extracellular domain was first predicted to fold like a beta-propeller (i.e., it contains a series of NHL repeats) in an early study of teneurin domain architecture (Tucker et al., 2012) and later demonstrated conclusively to be a 6-bladed beta-propeller by X-ray crystallography and cryoelectron microscopy (Jackson et al., 2018; Li et al., 2018). Beta-propellers are typically protein–protein interaction domains, and that appears to be the case with teneurins. HT1080 cells expressing the transmembrane and extracellular domains of teneurin-2 clump together in culture, but HT1080 cells expressing the transmembrane domain and a truncated extracellular domain that only includes the EGF-like domains do not (Rubin et al., 2002). The domain used for these teneurin–teneurin interactions was narrowed to the 6-bladed beta-propeller using atomic force microscopy while swapping and deleting the various teneurin extracellular domains that were expressed on the cell surface (Beckmann et al., 2013). This study also showed that the homotypic interactions between the beta-propellers of teneurin-1 were stronger than the heterotypic interactions between the beta-propellers of teneurin-1 and teneurin-2 (Beckmann et al., 2013). The beta-propeller domain of teneurin-1 seems to be critical for its function, as a mutation in this region leads to congenital anosmia in humans (Alkelai et al., 2016).

# YD Repeats and the RHS Core Protein Domain

Almost a third of the huge extracellular domain of teneurins is composed of over two dozen YD repeats. These repeats have the consensus sequence Gx3−9YxYDx2GR(L, I or V)x3−10G,

may represent a proteolytic cleavage site. (C) Most vertebrates have four teneurins numbered 1 through 4. The domain organization of the four teneurins from humans are illustrated. (D) A predicted teneurin is found in the genome of the choanoflagellate Monosiga brevicollis. The extracellular domain of this teneurin is most similar to the extracellular domains of bacterial toxins. Examples of the bacterial toxins are illustrated, as are UniProt ID and GenBank accession numbers.

where "x" is any amino acid (Minet et al., 1999; Minet and Chiquet-Ehrismann, 2000). The presence of YD repeats in teneurins was unexpected: prior to the sequencing of human and chicken teneurins YD repeats had only been identified in prokaryotic proteins. The potential function of the YD repeats became clear following the detailed description of a similar series of repeats found in a toxin from the bacterium Yersinia entomophaga using X-ray crystallography (Busby et al., 2013). The YD repeats in this bacterial toxin form a hollow barrel that is approximately 130 Å long and 50 Å wide [i.e., the approximate size of the "cherry" of the teneurin extracellular domain seen in the electron microscope (Feng et al., 2002)]. The RHS core protein domain forms a plug in the hollow end of the barrel. This bacterial YD repeat-containing protein also has a 6 bladed beta-propeller, and the beta-propeller is exposed to ligand binding at the N-terminal end of the barrel. The high degree of sequence similarity and domain architecture identity between the C-terminal half of the extracellular domains of teneurins and these YD repeat-containing proteins from bacteria strongly suggested that teneurins would fold in a similar way. This was recently confirmed by X-ray crystallography and cryoelectron microscopy with teneurin extracellular domains (Jackson et al., 2018; Li et al., 2018).

#### C-Terminal Domain: A Toxin and a TCAP

Just C-terminal to the RHS core protein domain of all sequenced teneurins, and most predicted teneurins, lies a region with striking amino acid similarity to the C-terminal GHH toxin domain of certain prokaryotic YD repeat-containing proteins (Zhang et al., 2012; Ferralli et al., 2018). GHH toxins are prokaryotic nucleases that are predicted to be encapsulated in a YD repeat barrel (like the toxic C-terminal domain of the YD repeat-containing protein from Y. entomophaga). Though similar, the GHH toxin domain of teneurins lack the key glycine-histidine-histidine motif that is necessary for the bacterial enzyme's nuclease activity (Zhang et al., 2012). However, when the GHH toxin domains of teneurin-1 or teneurin-2 are expressed in HEK 293 cells in culture, or when nanomolar concentrations of the purified GHH toxin domains of chicken teneurin-1 or chicken teneurin-2 are added to the culture medium, the cells rapidly undergo apoptosis (Ferralli et al., 2018). The toxicity may be related to nuclease activity, as purified GHH

toxins from teneurin-1 and teneurin-2 cleave plasmid DNA and completely hydrolyze mitochondrial DNA in vitro (Ferralli et al., 2018).

The C-terminal 40 or 41 amino acids of teneurins is known as TCAP (from "teneurin C-terminal associated peptide"). The TCAP sequence was first identified by researchers who noted its similarity to corticotropin-releasing factor (Qian et al., 2004), and purified TCAP has profound effects on animal behavior when injected into brain ventricles (Tan et al., 2008). For example, TCAP-treated rats behave in acoustic startle, open field and elevated plus maze tests in a manner that is consistent with elevated anxiety. These and other remarkable studies with TCAP have recently been reviewed by others (Woelfle et al., 2016). The TCAP sequence partially overlaps with the GHH toxin domain and extends to the very C-terminus of the protein (see above). Interestingly, teneurins are known to bind to the G-protein coupled receptor latrophilin (Silva et al., 2011), and the teneurin domain responsible for this interaction is the TCAP (Woelfle et al., 2015). This may contribute to the localization of some teneurins, and the C-terminal toxin/TCAP domain, to developing synapses (Li et al., 2018).

# Teneurin Tertiary Organization

The stick diagrams used for describing the domain organization of teneurins can now be refined thanks to the pioneering X-ray crystallography done with a related bacterial protein (Busby et al., 2013) and the elegant X-ray crystallography and cryoelectron microscopy done with the extracellular domains of teneurins themselves (Jackson et al., 2018; Li et al., 2018). We now know that the region found between the EGF-like domains and the 6 bladed beta-propeller folds into a beta-sandwich domain that is reminiscent of either a fibronectin type III (FN3) repeat (Jackson et al., 2018) or an immunoglobulin (Ig)-like domain (Li et al., 2018), and this domain "plugs" the N-terminal end of the hollow YD barrel (**Figure 1B**). This FN3/Ig-like domain has two subregions, one of which is highly conserved across phyla and is rich in cysteines (Tucker et al., 2012), and another which is predicted by standard domain architecture software programs (e.g., Superfamily) to be a carboxypeptidase domain. The latter is particularly interesting because some bacterial YD proteins with a 6-bladed beta-propeller also have a carboxypeptidase domain in this region, and these amino acid sequences align with nearly 50% similarity with the same region in human teneurins (**Figure 2A**). This striking phylogenetic conservation suggests that this unstudied domain may be more than a plug: perhaps it is involved in proteolytic processing of teneurins or teneurinassociated proteins. The RHS core protein fits into the C-terminal end of the teneurin YD barrel, but interestingly, both recent studies of teneurin structure (Jackson et al., 2018; Li et al., 2018) showed the GHH toxin/TCAP domain poking out through the side of the barrel instead of being contained within the barrel like the toxins of bacterial YD proteins. Thus, the conformational changes that are believed to release the toxin from the YD barrel of prokaryotes may not be necessary for the release of the toxin/TCAP domain from teneurins. Moreover, the arrangement of the C-terminal region of teneurins revealed by cryoelectron microscopy means that the TCAP domain is available to bind to latrophilin without any prior processing or the disruption of the YD barrel. Perhaps the toxic nuclease near the C-terminus of teneurins can be released by regulated proteolytic activity after the teneurin reaches the cell membrane. Supporting this hypothesis is the observation that almost all teneurins examined to date have the conserved basic motif RxRR between the RHS core protein and the GHH toxin domain (Tucker et al., 2012), and similar motifs are known to be targeted by proteases that act extracellularly [e.g., by members of the proprotein convertase subtilisin/kexin family of peptidases (Rawlings, 2009)]. Future studies are needed to address this important aspect of teneurin biology.

# Differences Between the Teneurins

The overall domain organization of the four chordate teneurins is identical, but upon closer examination certain distinguishing features can be recognized (**Figure 1C**). For example, the intracellular domains of human teneurin-1 and teneurin-4 have both proline-rich SH3-binding domains and predicted nuclear localization sequences, but the intracellular domain of human teneurin-3 lacks SH3-binding prolines and the intracellular domain of teneurin-2 lacks a nuclear localization sequence. These differences, however, are not conserved across species. In the chicken, where teneurins have been widely studied, the intracellular domains of all four teneurins have predicted nuclear localization sequences, whereas in the mouse only teneurins-1 and -3 have predicted nuclear localization sequences (Tucker et al., 2012). Though the intracellular domains of chordate teneurins and the teneurins found in ecdysozoa share little sequence homology, the intracellular domains of both ten-a and ten-m from D. melanogaster and ten-1 from C. elegans have predicted nuclear localization sequences and SH3-binding domains (Tucker et al., 2012). However, whenever discussing teneurin nuclear localization sequences it is important to remember that the vast majority are only predicted in silico. The only experimental evidence that the intracellular domains of teneurins can be transported to the nucleus come from studies with chicken sequences in cell lines (Bagutti et al., 2003; Nunes et al., 2005), in chicken embryos (Kenzelmann et al., 2008; Kenzelmann Broz et al., 2010) and in C. elegans (Drabikowski et al., 2005).

In chordates a cysteine in the second teneurin EGF-like domain has been replaced by a tyrosine, and the cysteine in the fifth EGF-like domain has been replaced by a tyrosine (teneurin-2 and teneurin-3) or by a phenylalanine or tyrosine (teneurin-1 and teneurin-4). This general arrangement is also found in the teneurins of ecdysozoa, but not teneurins from lophotrochozoa. For example, the predicted teneurin from the blood fluke Schistosoma mansoni has only four EGF-like domains, and all have a complete complement of cysteines (Tucker et al., 2012). Thus, while the dimerization of teneurins via their EGF-like domains is widespread, in some animals teneurins may act as monomers.

Teneurin-2 and teneurin-3 from chordates, as well as the teneurins from almost all invertebrates, have a predicted furin cleavage site between the transmembrane domain and the EGFlike domains. This site was shown to be functional in teneurin-2

(Rubin et al., 1999; Vysokov et al., 2016), and its widespread phylogenetic conservation suggests that it is important for teneurin function. This processing would suggest that the extracellular domain of teneurins is shed from the cell surface. However, the extracellular domain appears to remain anchored to the remaining transmembrane part of the teneurin through noncovalent interactions (Vysokov et al., 2016). Whether or not such interactions are found in other teneurins with the predicted furin cleavage site remains to be determined.

Functional differences in the extracellular domains of different teneurins have also been identified. As mentioned earlier, homotypic interactions between the beta-propellers of teneurins are stronger than heterotypic interactions (Beckmann et al., 2013), suggesting that the beta-propellers have properties that are unique to different teneurin forms. Moreover, the C-terminal regions of different teneurins have different affinities for latrophilins (Boucard et al., 2014). These observations will likely be keys to our understanding of why teneurins have duplicated to become a multigene family independently in arthropods and in chordates (Tucker et al., 2012).

# THE EVOLUTION OF TENEURINS

An examination of sequenced metazoan genomes revealed that teneurins are found in all animals with a central nervous system, but not in sponges, Trichoplax or cnidarians (Tucker et al., 2012). Given the prominent expression of teneurins in the developing central nervous system of flies, worms and chordates, this led, at least temporarily, to the assumption that teneurins evolved together with a complex nervous system. However, when predicted proteins with the teneurin domain organization were searched for in non-metazoan sequences, a teneurin was discovered in the genome of the single-celled choanoflagellate Monosiga brevicollis. The teneurin from this choanoflagellate is remarkable in many ways. First, its domain organization matches

that of chordate teneurins almost perfectly: it has an intracellular domain with a predicted nuclear localization sequence and an extracellular domain with eight EGF-like domains (all with six cysteine residues), a cysteine-rich domain and a carboxypeptidase domain, a beta-propeller, YD repeats and an RHS core domain (**Figure 1D**). It only lacks predicted furin cleavage sites found in most metazoan teneurins and a C-terminal GHH toxin/TCAP domain. Second, it is encoded on only four exons, the third of which contains 6829 residues and encodes almost all of the extracellular domain. Finally, BLAST searches of the sequences encoded on the third exon revealed that the extracellular domain of choanoflagellate teneurin was more similar to the YD proteins of bacteria than to the extracellular domain of metazoan teneurins. This pointed to the possibility that teneurins evolved via horizontal gene transfer from a bacterial prey (with a YD protein gene encoded on a single exon) to a single-celled predator prior to the evolution of metazoa from a choanoflagellate-like ancestor (Tucker et al., 2012).

Horizontal gene transfer into choanoflagellates from their prey (bacteria, algae, and diatoms) is well-documented, and many of these events have contributed genes that are still used in modern choanoflagellates, sometimes replacing similar host genes, and sometimes contributing novel enzymes that can be used by the host to exploit nutrient-deficient niches (Tucker, 2013). However, relatively few metazoan genes appear to have originated from a choanoflagellate gene that was in turn acquired from bacteria or algae. One survey revealed only two: dihydroxy-acid dehydratase and teneurins (Tucker, 2013).

As mentioned earlier, the extracellular domains of teneurins are remarkably similar to many prokaryotic YD proteins, and a protein similar to the modern-day YD proteins of bacteria is the most likely candidate for the ancestral teneurin. Some of these are illustrated schematically in **Figure 1D**. The carboxypeptidase-like domain found near the 6-bladed betapropeller of all teneurins is also found in the YD protein of Desulfurivibrio alkaliphilus, an anaerobic, gram-negative, nonmotile bacterium that lives in the extreme high-saline and high-pH soda lakes of North Africa (Melton et al., 2016). This YD protein has an uncharacterized toxin domain, but otherwise resembles, from the carboxypeptidase-like domain to the RHS core protein domain, the extracellular domain of teneurins. The remarkable similarity of the carboxypeptidaselike domain from the D. alkaliphilus YD protein and the similar domain of various teneurins is shown in **Figure 2A**. This stretch of approximately 70 amino acids is also particularly wellsuited for establishing the phylogenetic relationships between the YD proteins and teneurins as well as teneurins themselves (**Figure 2B**). Consistent with a proposed origin from bacteria, the choanoflagellate teneurin and the bacterial sequence are found in the same clade. In chordates teneurin-1 and teneurin-4 are likely to have evolved through gene duplication, as have teneurin-2 and teneurin-3. These paired relationships are consistent with the organization of the intracellular domains, the predicted extracellular domain furin cleavage sites, and the residues that replace the cysteine residues in the EGF-like domains. Analysis of other domains results in similar phylogenetic trees (Tucker et al., 2012).

From studies of teneurin evolution we can gain insight into teneurin function. For example, what is known about the function of the YD repeat-containing proteins of bacteria? First and foremost, they are toxins (Zhang et al., 2012). The C-terminal toxin domain is encased in a YD repeat barrel, apparently to protect the cell that is expressing the YD repeat protein from the toxin (Busby et al., 2013). One class of bacterial toxins with YD barrels are the ABC toxins (ffrench-Constant and Waterfield, 2005). The B and C parts of the toxin are expressed either from a single gene or on two or more adjacent genes, and they can form a complex containing a beta-propeller, YD repeats, RHS core protein domain and C-terminal toxin domain. Multiple versions of the C gene allow different types of toxins to be deployed (**Figure 1D**). The A protein provides a way for the toxin to get into the cell, either by making a pore or by inserting into the membrane and acting as a receptor for the BC component (Busby et al., 2013). Other YD proteins are, like teneurins, type-2 transmembrane proteins. They appear to be members of the "toxin on a stick" type of bacterial polymorphic toxins (Jamet and Nassif, 2015). Polymorphic toxins are part of self-recognition between bacteria and are used in interbacterial warfare; when identical proteins interact they do not release the C-terminal toxin domain, but when dissimilar proteins interact the toxin domains are released. Given the complementary patterns of expression of many teneurins during the development of the nervous system (see below), one can hypothesize that heterotypic interactions between teneurins may somehow result in the release of the GHH toxin domain. This in turn could lead to programmed cell death or the pruning of dendrites. However, to date this inviting hypothesis is only supported by circumstantial evidence.

#### PATTERNS OF TENEURIN EXPRESSION

# Drosophila melanogaster and Caenorhabditis elegans

The first description of teneurin expression came from Baumgartner and Chiquet-Ehrismann (1993), who used in situ hybridization to determine the sites of expression of ten-a in Drosophila. They found widespread ten-a expression in the early embryo and high levels of expression in the developing ventral nerve cord following germ band retraction. Ten-a transcripts are also observed in muscle apodemes, the clypeolabrum and the antenna-maxillary complex. This work was followed by two independent reports (Baumgartner et al., 1994; Levine et al., 1994) describing the expression of ten-m. Both papers reported ten-m expression in seven stripes during the blastoderm and germ band extension stages of development, and the eventual expression of ten-m in the ventral nerve cord and in cardioblasts. Both papers go on to describe the failure of ventral denticle belts to fuse in P-element insertion mutants (i.e., an "oddless" pair rule phenotype), and one illustrates the disruption of the central nervous system in these mutants (Levine et al., 1994). Higher resolution studies using LacZ expression under the tenm promoter revealed expression in imaginal disks (Levine et al., 1997; Minet et al., 1999). These studies reveal ten-m expression

in sensory mother cells and the R7 photoreceptor in developing ommatidia. The expression of teneurins in cardioblasts was revisited and expanded on by Volk et al. (2014). Ten-a is expressed at the border of cardioblasts and pericardial cells, and ten-m is expressed by both cardioblasts and pericardial cells. However, ten-m and ten-a mutants do not have heart defects (Volk et al., 2014).

The developing olfactory system of Drosophila has proven to be a particularly useful model for studying both the expression of teneurins and their roles in development. In Drosophila, olfactory receptor neurons (ORNs) are the primary neurons that receive olfactory information. ORN axons synapse with the dendrites of projection neurons (PNs) in glomeruli found in the antennal lobe, and the PNs in turn send their axons elsewhere in the central nervous system. ORN/PN pairs have been mapped precisely. For example, Or47b ORNs normally project to the VA 1 lm glomerulus, and Mz19 PN dendrites are found in an adjacent glomerulus. Hong et al. (2012) used this model to screen for genes that might regulate the development of precise neural networks. They observed that overexpression of ten-m in the Mz19 PNs leads to abnormal connections between Mz19 neurons and Or47b ORNs. A similar system was used for a second screen: Or88a ORNs normally project to the VA1d glomerulus where they intermingle with Mz19 PN dendrites. Overexpression of ten-a in Mz19 PNs disrupts the normal intermingling of Mz19 and Or88a dendrites. In a screen of 410 candidate genes, only the ectopic misexpression of ten-a and ten-m caused these disruptions. High levels of ten-a and ten-m are found in antennal lobe glomeruli in mostly non-overlapping patterns, but both are found in low levels in all glomeruli (Hong et al., 2012). In five glomeruli examined in detail, ORNs expressing high levels of ten-m send axons to glomeruli with PNs expressing high levels of ten-m, and the axons of ORNs expressing high levels of tena are found in glomeruli with PN dendrites that also express high levels of ten-a (**Figure 3A**). Genetic and RNAi knockdowns result in shifting patterns of ORN/PN interactions, indicating that homophilic interactions between the teneurins are necessary for proper synaptic patterning.

Drosophila is also a useful model for studying the roles of teneurins in the development of neuromuscular junctions. In this system ten-a is expressed by neurons and is presynaptic, while most of the ten-m is expressed by muscle and is post-synaptic (Mosca et al., 2012). In vivo ten-a and ten-m appear to form a complex, and disruption of the expression of either or both leads to severe defects in the neuromuscular junction. These defects include disorganization of microtubules presynaptically, and disruption of alpha-spectrin post-synaptically. Heterophilic interactions between the ten-a and ten-m expressed at basal levels in antennal lobe glomeruli also appear to occur in neuronal synapses (Mosca and Luo, 2014). Presynaptic ten-a controls the number of ORN synapses that are found in a glomerulus, and tena/ten-m interactions regulate presynaptic active zone number. The possible roles of teneurins in synaptogenesis were reviewed by Mosca (2015).

There is a single teneurin gene in C. elegans, but two transcripts are generated via alternate promoters (Drabikowski et al., 2005). The differences between the variants lie in the size of the intracellular domain: Ten-1L has a longer intracellular domain that includes two predicted nuclear localization sequences, while Ten-1S has a severely truncated intracellular domain. The extracellular domain of the variants is identical. Ten-1L expression was studied using a GFP translational fusion protein (Drabikowski et al., 2005). It is found in neurons in the ventral nerve cord and in the pharyngeal nerve ring. There is also significant non-neuronal expression in vulvar and diagonal muscles, gonadal distal tips cells and in the vas deferens, among other sites (**Table 1**). After injection with Ten-1 RNAi there are neuronal pathfinding defects, abnormal gonad development, and severe morphological defects resulting from abnormal migration of hypodermal cells. Similar defects are seen in Ten-1 null mutants (Drabikowski et al., 2005). In the mutant Ten-1(et5), which has a point mutation leading to a premature stop codon near the end of the EGF-like domains, defects resulting from stalled growth cone migration and abnormal pathways of neurite outgrowth are seen in the pharyngeal nerve ring (Mörck et al., 2010). Unlike Ten-1 null mutants, Ten-1(et5) worms typical live to become reproductive adults, suggesting that the intracellular domain and EGF-like domains can impart some survival benefit. Ten-1 may also play a role in the organization of the extracellular matrix, as basement membrane integrity is compromised in Ten-1 null larvae (Trzebiatowska et al., 2008).

#### Zebrafish

One of the first detailed descriptions of teneurin expression in a vertebrate was reported by Mieda et al. (1999), who cloned and sequenced zebrafish teneurin-3 and teneurin-4 while searching for genes that were regulated by Islet-3. Using in situ hybridization they showed that teneurin-3 is transiently expressed in the notochord, somites, branchial arches, and central nervous system (**Table 1**). Teneurin-4 is expressed faintly during gastrulation, and after that is primarily expressed in the developing brain. Within the central nervous system, teneurin-3 and teneurin-4 are found in largely complementary patterns. For example, at 23 hpf teneurin-4 is found in two lines that wrap around the rostral diencephalon and teneurin-3 is expressed in the region between the lines. The sharp borders between the domains expressing these two teneurins become less clear later in development. The expression of teneurin-4 in narrow bands of cells in the zebrafish central nervous system is remarkably similar to the first report of teneurin-4 expression in the mouse: in E10.5 and E11.5 mouse embryos, teneurin-4 transcripts are found in a sharp line at the boundary between the midbrain and hindbrain (Wang et al., 1998).

Teneurin-3 is also found in the developing zebrafish retina (Antinucci et al., 2013). It is expressed by retinal ganglion cells and amacrine cells, which synapse with each other in the inner plexiform layer. Teneurin-3 is also expressed by the targets of retinal ganglion cell projections in the tectum. Knockdown of teneurin-3 expression with antisense morpholino oligonucleotides leads to both abnormal arborization of retinal ganglion cells in the inner plexiform layer and to abnormal pathfinding in the tectum (Antinucci et al., 2013). Zebrafish larvae normally adapt their level of pigmentation to background lighting and appear lighter in bright light and darker at lower

FIGURE 3 | Teneurins are expressed by interconnected populations of neurons. (A) In Drosophila, olfactory receptor neurons (ORNs) expressing ten-a synapse in antennal lobe glomeruli with projection neurons (PN) expressing ten-a. Ten-m expressing neurons also synapse together in the antennal lobe. (B) In the developing chicken, teneurin-1 is expressed in the tectofugal visual pathway, whereas teneurin-2 is expressed in the thalamofugal visual pathway. dT, dorsal thalamic nuclei; LGN, lateral geniculate nucleus; RGCs, retinal ganglion cells; RN, rotund nucleus; SGC, stratum griseum centrale; SGP, stratum griseum periventriculare. (C) In the mouse hippocampus teneurin-3 is expressed in the CA1 region, the subiculum and the medial entorhinal cortex (MEC). Tracer studies show that these regions are connected to each other. LEC, lateral entorhinal cortex. The data summarized in this figure were published by Rubin et al. (2002); Hong et al. (2012), and Berns et al. (2018), respectively.

levels of illumination. The teneurin-3 morphants are darkly pigmented in bright light, indicating that they probably have severe visual deficits. A teneurin-3 knockout zebrafish was also generated by TALEN genome editing (Antinucci et al., 2016). In the teneurin-3 knockouts amacrine cells that normally express teneurin-3 fail to arborize in the appropriate strata of the inner plexiform layer. The authors of this study go on to show that teneurin-3-expressing neurons form a distinctive circuit in the zebrafish retina that is responsible for orientation selectivity.

#### Patterns of Expression in the Visual Systems of Birds and Mice

The first description of teneurin-1 features low-resolution in situ hybridization images pointing to developing neurons as a primary site of expression (Minet et al., 1999). In the developing chicken diencephalon teneurin-1 transcripts are found in the rotund nucleus, and in the optic tectum teneurin-1 is expressed by the large neurons of the stratum griseum centrale. This study was followed by a paper that compared the expression of teneurin-1 and teneurin-2 in the developing avian central nervous system (Rubin et al., 1999). Teneurin-1 and teneurin-2 mRNAs are both found in the developing thalamus, but in different nuclei, and in the optic tectum teneurin-2 expression was observed in layers that straddled the stratum griseum centrale but is missing from the stratum griseum centrale itself. This led to the conclusion that teneurin-1 and teneurin-2 are expressed in different populations of developing neurons. Similar results were reported in a study of the developing mouse that included other teneurin forms (Zhou et al., 2003), but the complementary patterns are not as obvious in the mouse as in the chicken. More detailed mapping of expression using antibodies specific for teneurin-2 led to the remarkable observation that teneurin-1 and teneurin-2 are each expressed by interconnected populations of neurons (Rubin et al., 2002). This is particularly clear in the developing visual system of the chicken, where teneurin-1 is expressed in the developing tectofugal visual pathway, and teneurin-2 is expressed in the developing thalamofugal visual pathway (**Figure 3B**). The timing of expression typically follows the period of growth cone pioneering and neurite outgrowth and coincides with periods of synaptogenesis, pruning, and apoptosis.

In the mouse, most retinal ganglion cells project to the lateral geniculate nucleus or the superior colliculus. The former projections form a critical map of visual field information prior to further processing in the cortex. Retinal ganglion cells also make a map of the visual field in the superior colliculus, and this map is important for integrating responses to auditory, somatosensory, and visual information (Kandel et al., 2012). Appropriate binocular vision requires that retinal ganglion cells from each retina project to either the ipsilateral or contralateral superior colliculus and lateral geniculate nucleus. Teneurins appear to be critical for the successful development of these visual circuits. Using in situ hybridization, Leamey et al. (2007) found that teneurin-3 is expressed in a gradient in the developing mouse retinal ganglion cell layer, with highest levels of expression in the ventral retina. This was confirmed with quantitative PCR. Teneurin-3 is also expressed in a gradient within the lateral geniculate nucleus, with highest levels of expression found in the dorsal part of the nucleus. As retinal ganglion cells from the ventral retina project to the dorsal part of the lateral geniculate nucleus, the authors next chose to study these projections in teneurin-3 knockout mice. The brains and retinas of the knockout mice appeared normal in standard histological preparations. However, a tracing study with the knockouts reveals abnormal ipsilateral projections that are no longer limited to the dorsal part of the lateral geniculate nucleus, as well as abnormal connections between the lateral geniculate nucleus and the visual cortex (Leamey et al., 2007; Merlin et al., 2013; for review see Leamey and Sawatari, 2014). Behavioral studies are consistent with the TABLE 1 | Non-neuronal expression of teneurins<sup>∗</sup> .

fnins-12-00938 December 7, 2018 Time: 16:19 # 9


(Continued)

#### TABLE 1 | Continued

fnins-12-00938 December 7, 2018 Time: 16:19 # 10


<sup>∗</sup>Expression determined by in situ hybridization, immunohistochemistry, or both. †Similar results are also reported in this paper using human tissues.

hypothesis that teneurin-3 knockout mice lack binocular vision (Leamey et al., 2007). Teneurin-3 mRNA is also expressed in a gradient in the superior colliculus, with highest expression medially and the lowest laterally (Dharmaratne et al., 2012). This also corresponds to the high ventral, low dorsal expression of teneurin-3 in the retina. When the teneurin-3 knockout mice were further analyzed, ipsilateral projections to the superior colliculus are highly abnormal, just as they are in the lateral geniculate nucleus. EphA7 is significantly reduced, and EphB1 is significantly upregulated, in the visual system of teneurin-3 knockout mice. This suggests that teneurins may work together with ephrin/Eph signaling in this system (Glendining et al., 2017). Other teneurins may also be critical for the development of visual pathways. Like teneurin-3, teneurin-2 is expressed by interconnected populations of neurons in the murine retina, lateral geniculate nucleus and superior colliculus (Young et al., 2013), and in teneurin-2 knockout mice there is a reduction in the number of retinal ganglion cells that project to ipsilateral targets. Interestingly, antibodies to teneurin-4 label retinal ganglion cell axons in the nasal, but not temporal, retina of the chicken embryo (Kenzelmann Broz et al., 2010). This may indicate that other teneurins may regulate the development of other circuits in the visual system.

# Expression in the Thalamus, Cortex, and Hippocampus

The thalamus contains dozens of nuclei that generally act as relay stations between sensory inputs and the cerebral cortex. The first studies of teneurin-1 and teneurin-2 noted that these teneurins are prominently expressed in distinct populations of thalamic nuclei (Minet et al., 1999; Rubin et al., 1999), some of which are interconnected parts of the visual system (Rubin et al., 2002). The importance of normal teneurin-2 and teneurin-3 expression in the lateral geniculate nucleus, which is found in the thalamus, was described in the preceding section.

In order to learn more about the repertoire of guidance molecules that are responsible for establishing the complicated set thalamic of circuits, Bibollet-Bahena et al. (2017) performed in situ hybridization with teneurin probes on sections through the rostral and caudal thalamic nuclei of embryonic and newborn mouse brains. Each teneurin has a distinctive pattern of expression. There is significant overlap between the expression patterns of teneurin-2, teneurin-3 and teneurin-4, with teneurin-1 forming a pattern that is largely complementary to the other teneurins. For example, in the rostral thalamus teneurin-1 is expressed in dorsal thalamic nuclei and the reticular nucleus, and these regions show little or no expression of the other teneurins. The other teneurins, but not teneurin-1, are expressed in the laterodorsal nucleus, and both teneurin-2 and teneurin-3 are expressed in the ventral anterior nucleus.

One of the thalamic nuclei, the parafascicular nucleus, projects to the striatum. Teneurin-3 is expressed in a dorsal to ventral gradient in both the parafascicular nucleus and the striatum (Tran et al., 2015). As neurons in the dorsal parafascicular nucleus project to the dorsal striatum, this gradient of teneurin-3 expression matches earlier studies of retinal projections to the lateral geniculate nucleus and superior colliculus (Leamey et al., 2007; Dharmaratne et al., 2012). The size of these regions and the numbers of neurons found in them are similar in both wild type and teneurin-3 knockout mice, but anterograde tracer studies revealed abnormal projections to the striatum as well as the loss of distinctive cluster terminals within the striatum (Tran et al., 2015). Consistent with the known functions of the parafascicular nucleus and striatum in goal-directed learning, teneurin-3 knockout mice exhibit delayed acquisition of motor skills (Tran et al., 2015).

The cerebral cortex is patterned both by intrinsic factors originating in neuronal progenitors and by extrinsic factors that originate in thalamocortical projections. One of the key factors regulating the intrinsic patterning of the neocortex is the homeobox transcriptional regulator EMX2, which is expressed in a high-caudal to low-rostral gradient in the cortical plate. Teneurin-4 was identified in a screen of genes that are differentially regulated in the Emx2(−/−) mouse (Li et al., 2006). In the developing mouse brain, teneurin-4 is normally expressed by cortical neurons and their precursors in a gradient that matches that of EMX2. In the Emx2(−/−) mouse, there is both a reduction in the overall level of expression of teneurin-4 and a loss of the expression gradient (Li et al., 2006). There is also strong evidence that teneurin-1 expression in the cortex is regulated by EMX2 (Beckmann et al., 2011). Finally, teneurin-3 was also identified as a gene that is differentially regulated in the developing mouse cortex (Leamey et al., 2008). It is highly expressed in layer V of the caudal-most cortex, which corresponds well with its prominent roles in the patterning of visual system (see above). Interestingly, overexpression of teneurin-3 in the embryonic cerebral cortex via in utero

electroporation leads to clustering of teneurin-3-expressing cells, suggesting stronger homophilic interactions between these neurons when compared with their neighbors.

The first study of teneurin expression in the mouse using immunohistochemistry described teneurin-1 in the molecular layer of the CA3 region of the adult hippocampus as well as the molecular layer of the cerebellum (Oohashi et al., 1999). This pioneering work was followed with a comparative study showing the expression of all four teneurins in the adult hippocampus: teneurin-1 is expressed in CA 3 and the dentate gyrus, teneurin-2 is expressed most strongly in the CA 1 and CA 2 regions, teneurin-3 is limited to the stratum lacunosum moleculare, and teneurin-4 is most prominently expressed in the molecular layer of the dentate gyrus and in the stratum lacunosum moleculare and stratum oriens of the CA 3 region (Zhou et al., 2003).

A recent study addressed the importance of normal teneurin-3 expression in the developing hippocampus (Berns et al., 2018). Teneurin-3 is expressed by a patch of neurons in the proximal part of the CA1 region of the P10 hippocampus, as well as in the neighboring distal subiculum and the medial entorhinal cortex (MEC). Tracer injected into the MEC labels teneurin-3-positive neurons in the subiculum and the proximal CA1, and tracer injected in the lateral entorhinal cortex labeled both the proximal CA1 and the distal subiculum, demonstrating that the neurons expressing teneurin-3 form a neural network (**Figure 3C**). The model was then exploited experimentally with a teneurin-3 knockout mouse to show the necessity of normal teneurin-3 expression in the development of CA1/subiculum connections (Berns et al., 2018).

# Non-neuronal Patterns of Teneurin Expression in Birds and Mammals

While the name "teneurin" comes from "ten-m" and "neuron" (Minet et al., 1999), it is important to remember that teneurins are also expressed in many non-neuronal tissues. As described above, teneurins are expressed in stripes in Drosophila embryos (Baumgartner et al., 1994; Levine et al., 1994), by motile cells and muscles in C. elegans (Drabikowski et al., 2005) and in somites and branchial arches in zebrafish (Mieda et al., 1999). In early chicken embryos antibodies to teneurin-4 immunostain the mesenchyme in many areas and co-localize with laminin in or near basement membranes (Kenzelmann Broz et al., 2010). Developing limbs also show distinctive and temporally dynamic patterns of teneurin expression. The apical ectodermal ridge (AER) is a prominent site of teneurin-2 expression (Tucker et al., 2001; Kenzelmann Broz et al., 2010). The teneurin-4 is expression pattern is more dynamic. It is initially observed in both the AER and the zone of polarizing activity, but later it is seen in the distal mesenchyme underlying the AER on the anterior part of the limb (Tucker et al., 2000; Kenzelmann Broz et al., 2010). Teneurin-1 expression in the developing limb is particularly interesting. Antibodies to the intracellular domain of teneurin-1 stain the cell surface of ectodermal cells in the dorsal limb, but they stain the cell nucleus in mesenchyme in the ventral limb (Kenzelmann Broz et al., 2010). These patterns are summarized in **Figure 4**. A recent study (Pickering et al., 2017) found that

teneurin expression changes when retinoic acid-soaked beads are applied to the anterior part of the limb bud (teneurin-4 expression decreases, while teneurin-2 expression increases), but the roles of teneurins in limb patterning are unknown. These and other patterns of teneurin expression in non-neuronal tissues are summarized in **Table 1**.

# CONCLUSION

Since their serendipitous discovery 25 years ago considerable progress has been made in our understanding of teneurin organization, evolution and expression. The intracellular domain is more variable than the rest of the protein, and its function remains mostly a mystery. In particular, its processing and localization to the nucleus at some, but not all, sites of expression is an observation in dire need of additional experimental work. More is known about the extracellular domain of teneurins, which apparently evolved from the extracellular domain of a prokaryotic YD protein via horizontal gene transfer. We now know that the YD repeats of both the prokaryotic YD proteins and the teneurins fold into a hollow barrel with a nearby betapropeller that can be used as a protein–protein interaction domain. Remaining work to be done includes studies of the highly conserved carboxypeptidase-like domain and whether or not the C-terminal domain can be released to act as a toxin. And if so, what triggers its release? In the central nervous system of vertebrates and flies teneurins appear to be expressed in largely

non-overlapping patterns than correspond to interconnected populations of neurons. As genetic manipulation of this pattern leads to disruption of the development of these networks, teneurins appear to be key players in brain development. However, just how teneurins accomplish this is unclear. Do they act primarily through differential adhesion, or is the more important interaction the one between TCAP and latrophilins? And is the GHH toxin domain somehow involved in this process? Finally, studies should not neglect the interesting sites of non-neuronal expression of teneurins, such as developing limbs. During the next quarter century, discovering the answers to these questions will present researchers with special challenges.

#### REFERENCES


#### AUTHOR CONTRIBUTIONS

The author confirms being the sole contributor of this work and has approved it for publication.

#### ACKNOWLEDGMENTS

The author is grateful to Matthias Chiquet, Jacqueline Ferralli, and David Lovejoy for their review of the manuscript. They would like to acknowledge the Associate Editor Dr. Roubos for his contribution in handling the review process for this manuscript.



circuits. J. Neurosci. 33, 12490–12509. doi: 10.1523/JNEUROSCI.4708-12. 2013


**Conflict of Interest Statement:** The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Tucker. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Teneurin Structures Are Composed of Ancient Bacterial Protein Domains

Verity A. Jackson<sup>1</sup>† , Jason N. Busby<sup>2</sup>† , Bert J. C. Janssen<sup>3</sup> , J. Shaun Lott<sup>2</sup> \* and Elena Seiradake<sup>4</sup> \*

<sup>1</sup> MRC Laboratory of Molecular Biology, Cambridge, United Kingdom, <sup>2</sup> School of Biological Sciences, The University of Auckland, Auckland, New Zealand, <sup>3</sup> Crystal and Structural Chemistry, Bijvoet Center for Biomolecular Research, Faculty of Science, Utrecht University, Utrecht, Netherlands, <sup>4</sup> Department of Biochemistry, University of Oxford, Oxford, United Kingdom

#### Edited by:

Antony Jr. Boucard, Centro de Investigación y de Estudios Avanzados del Instituto Politécnico Nacional (CINVESTAV), Mexico

#### Reviewed by:

Yuri Ushkaryov, University of Kent, United Kingdom Shuzo Sugita, University Health Network (UHN), Canada

#### \*Correspondence:

J. Shaun Lott s.lott@auckland.ac.nz Elena Seiradake elena.seiradake@bioch.ox.ac.uk

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Neuroscience

Received: 16 December 2018 Accepted: 15 February 2019 Published: 13 March 2019

#### Citation:

Jackson VA, Busby JN, Janssen BJC, Lott JS and Seiradake E (2019) Teneurin Structures Are Composed of Ancient Bacterial Protein Domains. Front. Neurosci. 13:183. doi: 10.3389/fnins.2019.00183 Pioneering bioinformatic analysis using sequence data revealed that teneurins evolved from bacterial tyrosine-aspartate (YD)-repeat protein precursors. Here, we discuss how structures of the C-terminal domain of teneurins, determined using X-ray crystallography and electron microscopy, support the earlier findings on the proteins' ancestry. This chapter describes the structure of the teneurin scaffold with reference to a large family of teneurin-like proteins that are widespread in modern prokaryotes. The central scaffold of modern eukaryotic teneurins is decorated by additional domains typically found in bacteria, which are re-purposed in eukaryotes to generate highly multifunctional receptors. We discuss how alternative splicing contributed to further diversifying teneurin structure and thereby function. This chapter traces the evolution of teneurins from a structural point of view and presents the state-of-the-art of how teneurin function is encoded by its specific structural features.

Keywords: cell adhesion, teneurin, bacterial toxin, evolution, choanoflagellate

# INTRODUCTION

Teneurins are evolutionarily ancient cell surface receptors which have emerged as important regulators of many neurobiological processes in both vertebrates and invertebrates, including synaptic partner matching (Hong et al., 2012), synaptic organization (Mosca et al., 2012), neuronal migration (Drabikowski et al., 2005) and axonal guidance and targeting (Leamey et al., 2007; Berns et al., 2018). This variety of functions has been attributed to both homophilic interactions between teneurins expressed on adjacent cells and also heterophilic interactions with other cell surface receptors, such as the synaptic adhesion G-protein-coupled receptor, latrophilin, which is essential for hippocampal synapse formation in mice (Silva et al., 2011; Boucard et al., 2014; Anderson et al., 2017). The relative importance of heterophilic vs. homophilic interactions of teneurins is not fully understood.

Consistent with their evolutionarily ancient origins, bioinformatic and phylogenetic analyses have revealed that eukaryotic teneurins arose via a horizontal gene transfer event from bacteria early during metazoan evolution (Tucker et al., 2012). However, their relationship to these bacterial proteins has remained unclear due to a lack of structural data for the teneurins. The structural analysis of teneurins has long been hampered by difficulties with recombinant expression and purification of these large, glycosylated and intricately folded type II transmembrane protein receptors. The advent of advanced eukaryotic expression systems, especially those using

mammalian HEK293 cells (Seiradake et al., 2015), and insect cells in suspension (Berger and Poterszman, 2015) have made the production of high quality samples tractable and led to the recent publications of the first teneurin structures (Jackson et al., 2018; Li et al., 2018). These recently published structures agree with each other overall, but there are some discrepancies that cannot be easily explained by differences in the sequences used, and which may arise from the methods employed to determine the structures. Where there are differences, we focus on the results obtained with the most completely modeled and highest resolution structure (solved by X-ray crystallography, up to 2.4 Å resolution), which is that of chicken Teneurin 2 (**Figures 1A,B**; Jackson et al., 2018). This structure is in good agreement with the structure of murine Teneurin 3, solved by cryo-electron microscopy by the same authors (up to 3.8 Å resolution), and with the structure of human Teneurin 2 solved by Li et al. (2018), also using cryo-electron microscopy (up to 3.1 Å resolution). The structure of chicken Teneurin 2 (residues 955-2802) obtained by X-ray crystallography is >99% complete, although less structural rigidity is observed in its peripheral domains. This is reflected by increased thermal motion (Bfactors) in these regions (**Figure 1E**). The two models derived from cryo-electron microscopy reconstructions are both less complete compared to the X-ray structure, in particular in those regions corresponding to those with the highest B-factors in the crystal structure (**Figure 1F**). This suggests that these areas of the protein may be inherently flexible, but what causes the differences in magnitude of the domain flexibility in the different structures is currently not clear. Given the availability of the higher resolution and more complete X-ray structure, we will only discuss the lower resolution cryo-EM structures where the data is most reliable.

Alongside their cryoEM structure of human Teneurin2, Li et al. (2018) also present compelling mutagenesis and functional data. To avoid overlap with other chapters of this review series we focus our discussion on the structural features of teneurins with respect to their evolution from the bacterial protein ancestors.

#### THE TENEURIN SCAFFOLD

Teneurins are typically ∼2800 amino acids long with about half of this sequence encoding a structurally conserved extracellular scaffold (**Figure 1B**). Over half of this scaffold region contains a sequence repeat motif known as a tyrosine-aspartate (YD) repeat [or rearrangement hotspot (Rhs)-repeat] (Zhang et al., 2012). The YD-repeat motif is widespread amongst prokaryotic proteins, and has been structurally characterized in the context of the heterodimeric TcB-TcC subcomplex in bacterial ABC toxin complexes (Busby et al., 2013; Gatsogiannis et al., 2013; Meusch et al., 2014). In both teneurins and TcB-TcC complexes, each YD-repeat forms a pair of β-strands joined by a β-hairpin. Together, multiple repeats form an extended spiraling β-sheet resulting in a large, hollow shell-like structure (the YD-shell) that is sealed at the C-terminal end by spiraling inward to form an "Rhs-associated core domain"-like structure. As well as acting as a central scaffold, the teneurin YD-shell is thought to bind to negatively charged heparin glycans (Minet et al., 1999). A mutation in the YD-shell of human Ten1, P1610L, is linked to congenital anosmia (Alkelai et al., 2016).

Two smaller domains are found N-terminal to the YD-shell and complete the teneurin scaffold. The most N-terminal of these domains is a distinctive fibronectin (FN) type-III domain termed the FN-plug domain (**Figures 1A,D**). Compared to a canonical FN type-III domain, the FN-plug contains an insertion (44 amino acids in chicken Ten2) between β strands 1 and 2 and an extension of its C-terminus (38 amino acids in chicken Ten2). This insertion and extension form a separate subdomain designated the "plug," as it resides inside the YD-shell cavity, forming numerous hydrophobic and hydrogen bonding interactions with the shell interior. The YD-shell and FN-plug combination encloses a space of ∼130 nm<sup>3</sup> .

Between the FN-plug and YD-shell is a six-bladed β-propeller domain (**Figures 1A,B,D**), the NHL domain. The NHL domain has been shown to determine the specificity of teneurin homophilic interactions (Beckmann et al., 2013), which are thought to underlie initial synaptic partner matching in the murine hippocampus (Berns et al., 2018), the vertebrate visual system (Dharmaratne et al., 2012; Antinucci et al., 2013) and in the Drosophila olfactory map (Hong et al., 2012) and neuromuscular junction (Mosca et al., 2012). The NHL domain is positioned perpendicular to the YD-shell and is held in place by the FN-plug domain. The top face (Chen et al., 2011) of the NHL domain is decorated with extended loops, stabilized by five highly conserved disulphide bonds. One of these loops undergoes alternative splicing. Recent in vitro data indicates that inclusion of this splice site in Ten3 enables homophilic interactions between teneurins expressed on adjacent cells "in trans" (Berns et al., 2018). Consistent with these residues being important for homophilic teneurin interactions in trans, teneurin isoforms including this alternatively spliced region have shown cell–cell adhesive properties (e.g., for Ten2 Beckmann et al., 2013), whilst an isoform lacking these residues did not show any adhesion (Silva et al., 2011; Boucard et al., 2014). Recent reports have also indicated that inclusion of this splice site also abrogates the human teneurin-latrophilin interaction (Li et al., 2018), however, the alternatively spliced chicken Ten2 retains its binding capabilities (Jackson et al., 2018). Thus this area requires further investigation to provide meaningful mechanistic insights across species.

# BACTERIAL TENEURIN-LIKE PROTEINS

# Comparison of Teneurin and TcB-TcC

The C-terminal teneurin scaffold shows a strong similarity to the TcB-TcC subcomplex of bacterial ABC insecticidal toxin complexes (Tc), with some significant differences. These ABC toxins are large protein complexes produced by a number of bacterial species against insect targets. These toxin systems are comprised of three major components: TcA forms a pentameric injection apparatus that binds to an insect cell and punctures the cell membrane, injecting the toxic component (Landsberg et al., 2011; Gatsogiannis et al., 2013). Together TcB and TcC form a continuous hollow shell-like structure that contains the

cytotoxic cargo (the C-terminal end of TcC) until delivery (Busby et al., 2013; Meusch et al., 2014). We will compare the C-terminal domain of Teneurin with the combined TcB-TcC complex, as the latter heterodimer acts as a single functional unit. The fact that teneurins are produced as a single large polypeptide, whereas TcB-TcC is a heterodimer of two separate proteins (TcB and TcC), is likely the result of a genetic rearrangement, as both TcB and TcC are required to form a properly folded complex, and the C-terminus of the TcB protein is located directly adjacent to the N-terminus of TcC (Busby et al., 2013; Meusch et al., 2014). The tcB and tcC genes can be fused to produce a single protein that folds correctly (Busby and Lott, unpublished data), and this type of fusion can be found naturally in certain bacterial species such as the TcdB2 protein from Burkholderia pseudomallei.

Overall, the TcB-TcC shell is larger than the equivalent region of teneurin (2177 for the TcB + TcC2-NTD complex from Yersinia entomophaga vs. 1458 amino acids for chicken teneurin 2). This results in the spiral of β-sheet making ∼1 extra turn in TcB-TcC compared to teneurin (**Figures 2A,B**). This difference suggests that the internal cavity formed by the YD-shell can vary in size by simply having a different number of YD-repeats, potentially allowing differently sized protein cargo to be encapsulated. The "cargo" proteins encapsulated by TcB-TcC complexes are typically in the range of 276–295 amino acids. The linker region that is retained inside Ten2 is only 90 amino acids, although the teneurin FN-plug domain also extends into the central cavity.

Where teneurin has a TTR and FN-plug domain upstream of the NHL β-propeller, TcB proteins have an SpvB "domain" that extends the YD-shell. In teneurin these domains cause the β-propeller to be folded out to one side, with the FNplug domain sealing the opening to the shell's internal cavity. However, in TcB proteins this opening is plugged by the β-propeller and SpvB domain. In both cases, the exterior face of the β-propeller is involved in protein-protein interactions: in teneurin, this surface is involved in homophilic interactions (Beckmann et al., 2013), whereas in TcB this is the interaction surface that attaches to the pentameric TcA toxin delivery device. In both teneurins and TcB, the β-propeller forms a separate domain to the YD-shell.

Aside from the N-terminal TTR and FN-plug domains, the basic arrangement of teneurin and TcB-TcC is fundamentally similar. In both structures the β-propeller domain is followed by a series of YD-repeats that form a spiral of β-sheet around a central cavity, capped at the C-terminal end by the Rhs-repeatassociated core domain. This domain serves to both seal the hollow "shell" at the C-terminal end and, in TcC, acts as a selfcleaving aspartic protease to cut loose the "cargo" protein (Busby et al., 2013; Meusch et al., 2014). While there are a number of residues in this domain that are conserved between teneurins and TcC proteins (chicken Ten2 and the TcC protein YenC2 are 26.4% identical and 32.2% similar in amino acid sequence), the highly conserved DxxGx motif present at the self-cleavage site is absent in teneurin, indicating that the autoproteolytic activity has been

lost (**Figure 2C**). Accordingly, there is no evidence of proteolytic cleavage in the crystal structure of chicken Ten2.

The C-terminal regions of TcC proteins contain a highly variable toxin domain encoding a variety of functions, including the ADP-ribosylation of actin and Rho proteins, and a predicted nucleic acid deaminase (Lang et al., 2010, 2011; Marshall et al., 2012). The equivalent region in teneurin instead contains a conserved linker domain that exits the hollow shell through a gap in its side, followed by an antibiotic-binding domain (ABD) and a Tox-GHH domain which are described below.

# Bacterial Teneurin-Like Proteins

Jackson et al. (2018) identified an Rhs-repeat containing protein from Bacillus subtilis that shows greater similarity to teneurin than to the TcB-TcC complexes. Like teneurins, this protein is encoded by a single gene, rather than as a heterodimeric assembly. This protein contains a series of N-terminal bacterial immunoglobulin-like (Ig-like) domains in an equivalent position to the EGF repeats of teneurin. Following this is a similar set of domains to teneurin: a TTR-like domain, FN-plug, NHL β-propeller, YD-repeats, and an Rhs-repeat-associated core domain (**Figure 1C**). Using this sequence to search the Uniprot database reveals a number of similar proteins from diverse bacterial phyla, including Acidobacteria, Alpha, Beta, Gamma and Deltaproteobacteria, and Verrucomicrobia. While these proteins have little or no sequence similarity upstream of the TTR-like domain, they often contain a set of repeats that fold into β-sandwich domains (such as bacterial Ig-like, fibronectin-like, carboxypeptidase regulatory-like, and CARDB domains). Following these repeats, the amino acid sequence conservation increases prior to the TTR domain and continues throughout the FN-plug, β-propeller, YD-shell, and Rhs-repeatassociated core domain. In these proteins, all of the key catalytic residues involved in self-cleavage are conserved, and presumably the Rhs-repeat-associated core domain will cleave the polypeptide backbone, as is the case in TcB-TcC, leaving the C-terminal domain untethered but encapsulated inside the YD-shell. The C-terminal domain itself is highly variable, with little or no sequence conservation between species. This is in line with previous reports suggesting that YD-repeats and their associated C-terminal regions are evolutionarily decoupled in bacteria (Jackson et al., 2009). There is no indication of a linker as is found in teneurins, or that the C-terminus exits the shell. Presumably these bacterial teneurin-like proteins encapsulate their C-terminus in a manner similar to TcB-TcC complexes, rather than displaying it on the outside of the shell as is the case with eukaryotic teneurin. Bacterial proteins containing YD-repeats are widespread (Zhang et al., 2012) and are thought to be involved in toxin delivery into adjacent cells. In the case of ABC toxin complexes, this is the delivery of cytoskeleton-disrupting toxins into a eukaryotic host cell (Waterfield et al., 2001; Ffrench-Constant and Waterfield, 2006; Hurst et al., 2011), whereas other (non-ABC) YD-repeat proteins are thought to be involved in contact-dependent growth inhibition by delivering a protein toxin into other bacterial cells (Koskiniemi et al., 2013). The function of bacterial teneurinlike proteins is currently unknown, but is likely to involve the delivery of the variable C-terminal domain into other cells,

possibly to inhibit the growth of competing microorganisms, or during pathogenesis.

#### TENEURIN DOMAINS DOWNSTREAM OF THE SCAFFOLD

Approximately 500 amino acids link the vertebrate teneurin scaffold region to its single-span transmembrane helix. This region comprises a membrane-proximal region of ∼180 amino acids, eight epidermal growth factor-like (EGF) repeats, a short cysteine-rich motif and a recently identified transthyretin-like (TTR) domain, which is flexibly linked to the central scaffold and packs against the FN-plug domain in the high resolution crystal structure of chicken Ten2, where it is stabilized in a discrete conformation by crystal lattice contacts (Jackson et al., 2018). The TTR domain is not resolved, presumably due to its flexibility, in any of the available structures determined by cryo-electron microscopy (Jackson et al., 2018; Li et al., 2018). The EGF region has not yet been structurally resolved, but bears sequence similarity to the extracellular matrix protein tenascin (Baumgartner et al., 1994; Levine et al., 1994), and has received particular attention due to its ability to dimerise teneurins (Feng et al., 2002). Two of these eight EGF repeats contain only five cysteine residues, compared to the six which would be typical in an EGF repeat. This results in the formation of intermolecular disulphide bridges that presumably form in cis i.e., between teneurin molecules expressed on the same cell. Early rotary shadowing electron microscopy data revealed the overall architecture of these dimers, showing pairs of globular domains (now thought to represent the teneurin scaffold and peripheral TTR, ABD and Tox-GHH), connected by thin elongated rods, presumed to be the EGF repeats (Feng et al., 2002).

Like the NHL domain, the teneurin EGF region also contains an alternatively spliced exon, which lies between the seventh and eighth EGF repeat. Sequencing of mRNA from hippocampal neurons (Berns et al., 2018) revealed three splice variants at this site. Similar to the alternatively spliced site in the NHL domain, inclusion of the alternatively spliced residues of the EGF region enables teneurin homophilic interactions in trans in an in vitro cell aggregation assay (Berns et al., 2018). Future structures of full teneurin ectodomain dimers will likely be valuable in understanding the molecular basis underlying this splicing-dependent homophilic adhesion.

Downstream of the EGF region lies a highly conserved cysteine-rich motif with the consensus sequence ExxCx(D/N) xxDx(D/E)xDxxxDCxxx(D/E)CCxxxxCxxxxxC. The teneurin cysteine-rich motif has not yet been functionally characterized, although bioinformatic analysis has revealed sequence similarity to putative metal-binding motifs in bacterial and algal proteins (Tucker et al., 2012).

#### TENEURIN DOMAINS DOWNSTREAM OF THE SCAFFOLD

As predicted from previously solved structures of YD-repeatcontaining proteins (Busby et al., 2013; Meusch et al., 2014), the residues immediately downstream of the Rhs-associated core (residues 2467–2592 in chicken Ten2 and 2382–2507 in murine Ten3) reside inside the shell cavity. These residues form extended loops and short helices, which pack against the Rhs-associated core domain, FN-plug and shell interior, sealing small gaps in the YD-shell. This region has been designated the "internal linker" and has no detectable structural homologs (**Figure 3A**). The teneurin internal linker is significantly smaller than the proteolytically cleaved C-terminal regions of TcB-TcC complex C-proteins (**Figure 3B**). In contrast to the teneurin internal linker, the cytotoxic C-terminal regions of bacterial C-proteins are thought to be unfolded when encapsulated within the YD-shell, and fold only upon injection into their insect cell targets (Busby et al., 2013; Meusch et al., 2014; Gatsogiannis et al., 2018).

Strikingly, unlike structures of TcB-TcC toxin complexes, where the entire C-terminal toxic domain is enclosed within the YD-shell, the teneurin C-terminal domain leaves the shell interior via a gap in the shell wall (**Figure 3C**). These exposed C-terminal residues (∼200 amino acids) form an intricate fold comprising two subdomains, which pack against the shell exterior (**Figure 3D**). This external C-terminal domain is only fully resolved in the crystal structure of chicken Ten2 and is either completely or partially disordered in both of the teneurin structures solved by single-particle electron microscopy, presumably due to conformational flexibility in this region.

Immediately downstream of the shell exit site, the crystal structure of chicken Ten2 reveals a 120-residue domain (the antibiotic-binding-like domain, ABD), which bears structural homology to a class of bacterial proteins which bind the smallmolecule antibiotics bleomycin and zorbamycin (Maruyama et al., 2001; Rudolf et al., 2015). These bacterial proteins confer resistance to the antibiotics by sequestering them, preventing their activation by oxygen (Rudolf et al., 2015). The bleomycin-binding surface of the bacterial protein has been identified crystallographically (Maruyama et al., 2001). The equivalent surface in chicken Ten2 is solvent-exposed, but there are currently no known small molecule ligands for the teneurin ABD.

The ABD wraps around a long helix that leads into the most C-terminal domain of teneurin – the ToxGHH domain. The ToxGHH domain was originally identified bioinformatically and is prevalent amongst bacterial toxins, where it encodes a putative nuclease activity mediated by a HNH catalytic triad (Zhang et al., 2012). The closest structural homolog of the teneurin Tox-GHH domain (residues 2720–2795 in chicken teneurin 2) is the catalytic nuclease domain of a class of bacteriocins known as group A colicins (**Figure 3E**; Kleanthous et al., 1999). Structural comparison and sequence alignments of teneurin Tox-GHH domains and colicin nuclease domains reveals that teneurins lack both catalytically important histidine residues, and retain only the structurally important asparagine residue (N2790) (**Figure 3F**; Kleanthous et al., 1999). Recent reports have indicated that, when expressed in isolation, the teneurin C-terminal domain (chicken Ten 1 S2346-R2705 or chicken Ten2 F2407-R2802) is capable of cleaving mitochondrial circular nucleic acids in vitro and inducing apoptosis in vivo. However,

in the context of an intact teneurin ectodomain, no nuclease activity was observed (Ferralli et al., 2018). Given the newly available structural data indicating that the teneurin C-terminal domain is solvent-exposed, rather than encapsulated within the YD-shell, future experiments will be necessary to determine how any potential nuclease activity is regulated.

It has previously been shown that the final ∼40 residues of the teneurin Tox-GHH domain bear sequence similarity to pre-prohormones and neuropeptide precursors (Qian et al., 2004). This region corresponds to the teneurin C-terminal associated peptide (TCAP), which is thought to function independently of intact teneurins to stimulate neurite outgrowth, regulate dendritic morphology and modulate anxiety behaviors in rats (Qian et al., 2004; Wang et al., 2005; Al Chawaf et al., 2007; Tan et al., 2008). As well as being transcribed independently (Chand et al., 2013), TCAPs may be generated via proteolytic cleavage from full-length teneurins (Lovejoy et al., 2006). The predicted cleavage site (based on alignment with known neuropeptides) (Lovejoy et al., 2006) is located within the second helix of the Tox-GHH domain, suggesting that the protein may unfold in this region to become accessible to proteases (**Figure 3E**). Recent evidence also suggests that the C-terminus of teneurin is required for binding to the adhesion-G-protein-coupled receptor latrophilin (Lphn) (Silva et al., 2011; Li et al., 2018) as removal of 390 residues from the C-terminus of human Ten2 abolished Lphn binding (Li et al., 2018). The precise nature and stoichiometry of the Lphn-Ten interaction is currently unknown.

# EVOLUTION OF TENEURIN-LIKE PROTEINS

A bioinfomatic approach using a wide range of fully sequenced genomes identified teneurins in chordate genomes (e.g., mouse, chicken and zebrafish), and also in the protochordates Ciona intestinalis and Branchiostoma floridae (Tucker et al., 2012). Teneurin genes were also identified in molluscs, annelids, trematodes, nematodes and arthropods. No teneurin sequences could be identified in cnidarians, ctenophores or sponges, indicating either that they are not present in these deeperbranching eukaryotic clades, or that they have highly divergent sequences (Tucker et al., 2012). However, a single teneurin gene in the genome of the choanoflagellate Monosiga brevicollis was discovered and shown to have only three introns. This gene shows strong sequence similarity to some bacterial teneurin-like proteins (with strongest similarity to the teneurin-like protein from B. subtilis; 69% coverage, 30% identity), albeit having a divergent intracellular domain, providing evidence that the gene could have been acquired by M. brevicollis from its prokaryotic prey via a horizontal gene transfer event (Tucker et al., 2012), as observed for other proteins (Foerstner et al., 2008). As choanoflagellates are considered to be the closest living relatives of metazoans, the authors suggest that teneurins may have played a key role in the evolution of multicellular metazoa from their unicellular choanoflagellate ancestors. Further testing of this hypothesis awaits a careful examination of recently sequenced genomes in the light of the recently determined structures described above.

#### CONCLUDING REMARKS

fnins-13-00183 March 12, 2019 Time: 10:0 # 7

Recent breakthroughs in understanding the first structures of teneurin protein domains have revealed unexpected new insights into this enigmatic family of proteins. First, the structures show that the C-terminal Tox-GHH/TCAP domains are solventexposed and therefore will be accessible to external ligands, proteases and other extracellular factors. Second, teneurin family proteins contain a previously unknown type of fibronectin domain (FN-plug) that specifically acts to seal off the YD-shell. Third, understanding the fold of the central teneurin scaffold led to the discovery of a whole family of teneurin-like proteins in prokaryotes, suggesting evolution from an ancient and widespread fold. Future research on these bacterial proteins, comparing their structure-function relationship with teneurins, will likely reveal more unexpected aspects of teneurin biology. Last but not least, the elaborate structure of the teneurin extracellular C-terminus is unlike that of any other cell adhesion

# REFERENCES


receptor, most of which consist of smaller adhesion domain repeats. The intricate nature of this fold, and the presumably large energetic cost of synthesizing and correctly folding these large proteins, suggests that teneurins may encode functions beyond the role of a canonical cell adhesion receptor that are yet to be explored.

#### AUTHOR CONTRIBUTIONS

All authors contributed to the conception and writing of the manuscript.

#### FUNDING

VJ was funded by the UK Medical Research Council. BJ is supported by a European Research Council Starting Grant (677500). JB was funded by a Marsden Fund Award (UOA1406) from the Royal Society of New Zealand to JL. ES is funded by the Wellcome Trust (202827/Z/16/Z).


an ancient fold for cell-cell interaction. Nat. Commun. 9:1079. doi: 10.1038/ s41467-018-03460-0


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Jackson, Busby, Janssen, Lott and Seiradake. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Teneurin Structure: Splice Variants of a Bacterial Toxin Homolog Specifies Synaptic Connections

#### Demet Araç1,2 \* and Jingxian Li1,2

<sup>1</sup> Department of Biochemistry & Molecular Biology, The University of Chicago, Chicago, IL, United States, <sup>2</sup> Grossman Institute for Neuroscience, Quantitative Biology and Human Behavior, The University of Chicago, Chicago, IL, United States

Teneurins are a conserved family of cell-surface adhesion molecules that mediate cellular communication, and play key roles in embryonic and neural development. Their mechanisms of action remained unclear due in part to their unknown structures. In recent years, the structures of teneurins have been reported at atomic resolutions and revealed a clear homology to bacterial Tc toxins with no similarity to other eukaryotic proteins. Another surprising observation was that alternatively spliced variants of teneurins interact with distinct ligands, and thus specify excitatory vs. inhibitory synapses. In this review, we discuss teneurin structures that together with structureguided biochemical and functional analyses, provide insights for the mechanisms of trans-cellular communication at the synapse and other cell-cell contact sites.

#### Edited by:

Antony Boucard Jr., Center for Research and Advanced Studies of the National Polytechnic Institute (CINVESTAV), Mexico

#### Reviewed by:

Davide Comoletti, Rutgers University, The State University of New Jersey, United States Susanne Ressl, Indiana University Bloomington, United States

> \*Correspondence: Demet Araç arac@uchicago.edu

#### Specialty section:

This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Neuroscience

Received: 24 April 2019 Accepted: 26 July 2019 Published: 07 August 2019

#### Citation:

Araç D and Li J (2019) Teneurin Structure: Splice Variants of a Bacterial Toxin Homolog Specifies Synaptic Connections. Front. Neurosci. 13:838. doi: 10.3389/fnins.2019.00838 Keywords: teneurin, ODZ, adhesion GPCR, latrophilin/ADGRL, synapse, structure, alternative splicing, FLRT

The coupling of extracellular adhesion to intracellular signaling is essential for cells to interpret cues from neighboring cells. With four members in humans, teneurins (TENs) are evolutionarily conserved cell-adhesion proteins that mediate intercellular communication (Baumgartner et al., 1994; Levine et al., 1994; Lossie et al., 2005; Tucker and Chiquet-Ehrismann, 2006; Nakamura et al., 2013; Mosca, 2015). Recent studies show that TENs play central roles in tissue polarity, embryogenesis, heart development, axon guidance, and synapse formation (Levine et al., 1994; Doyle et al., 2006; Silva et al., 2011; Hong et al., 2012; Mosca et al., 2012; Boucard et al., 2014; Mosca and Luo, 2014; Woelfle et al., 2016; Berns et al., 2018). Genetic studies link them to various diseases including neurological disorders, developmental problems, various cancers and congenital general anosmia (Aldahmesh et al., 2012; Ziegler et al., 2012; Hor et al., 2015; Alkelai et al., 2016; Chassaing et al., 2016; Graumann et al., 2017; Talamillo et al., 2017). TENs are type-II transmembrane proteins with large (>2000 amino acids) C-terminal extracellular regions (ECRs) that mediate heterophilic and homophilic trans-cellular interactions that are key for various functions. Most TEN ECRs do not exhibit readily identifiable domains by sequence analysis. Despite the central importance of TENs in multiple physiological functions, the lack of information on the structure of TENs and a molecular understanding of TEN interactions with heterophilic or homophilic ligands has been one of the limiting factors in delineating the mechanisms of TEN action.

# TENEURIN IS HOMOLOGOUS TO BACTERIAL TOXINS

TENs are composed of an N-terminal cytoplasmic tail, a single transmembrane region (TM), and a large ECR (**Figure 1A**; Baumgartner and Chiquet-Ehrismann, 1993; Baumgartner et al., 1994; Levine et al., 1994; Oohashi et al., 1999). The ECR of TENs comprises eight epidermal growth factor (EGF) motifs that are followed by the large unknown region composed of domains identified

as 2, 3, 4 and 5 in **Figure 1A**. TENs form cis-dimers via conserved disulfide bonds formed between their EGF repeats juxtaposed to the transmembrane helix (Oohashi et al., 1999; Feng et al., 2002; Vysokov et al., 2016; **Figure 2**). Recently, important breakthroughs were achieved and the high resolution cryo-electron microscopy structures of the unknown region from human TEN2, mouse TEN3 and the crystal structure of chicken TEN2 have been reported revealing a highly unusual structure (Jackson et al., 2018; Li et al., 2018). The structures agree with each other and show that the TEN YD repeats corresponding to domain 4 has a striking structural similarity to that of bacterial Tc-toxins as previously suspected (Tucker et al., 2012; Busby et al., 2013; Meusch et al., 2014; **Figure 1B**). Surprisingly, the TEN2 ECR has an unusual architecture that has not been observed in any other eukaryotic protein before (**Figures 1B–D**). The structure comprises a large cylindrical β-barrel (blue) sealed at the bottom by an Immunoglobulin (Ig)-like domain (yellow) and a β-propeller (green); and at the top by a C-terminal domain (magenta).

The TEN β-barrel has high homology to bacterial Tc-toxins (**Figures 1B,E**). Bacterial Tc-toxins are secreted proteins and comprise a large barrel that contains a toxic domain (Landsberg et al., 2011; Busby et al., 2013; Meusch et al., 2014; **Figure 1E**). The toxin domain is autocleaved from the rest of the protein,

passes through the barrel tunnel and is injected into the infected cell to execute its toxic effects usually by binding to the infected cell's DNA or other components (**Figure 1E**). The β-barrel of TEN2 structure partially encapsulates a C-terminal toxin-like domain. The toxin-like domain, however, exits the barrel from an opening and is tethered to the outer surface of the barrel (Li et al., 2018; **Figure 1B**). In spite of structural differences, the autoproteolytic site of the bacterial Tc-toxin is conserved in the TEN2 structure (Li et al., 2018; Arac and Li, 2019) (see Li et al., 2018 and Araç and Li, Current Opinion in Structural Biology, 2019 for in-depth structural analysis and comparison of TEN and bacterial Tc toxin structures). These observations raised numerous questions and exciting possibilities about how TENs may function. An immediate question is whether TENs function similar to bacterial toxins. Is the toxin-like domain of TEN autocleaved, released and inserted into the neighbor cells? Is it toxic to neighbor cells? Does it go into the nucleus and bind to DNA or other components of the cells? Does it mediate intracellular signaling or induce synapse formation and developmental changes? Does the previously reported C-terminal peptide of TEN act like a neuropeptide (Wang et al., 2005; Vysokov et al., 2016)? These are very exciting and open-ended questions that need to be answered in the future.

# ALTERNATIVE SPLICING REGULATES TENEURIN INTERACTIONS

Recent studies provided further surprising results about the roles of splice variants of TENs. TENs are alternatively spliced at two sites within the ECR and include nine- and sevenresidue insertions at the EGF repeats and the β-propeller regions, respectively (Tucker et al., 2001). The extracellular region of TENs mediates homophilic and heterophilic interactions that have specific roles in different functions of TENs, however the role of alternative splicing in mediating these interactions remained unknown (Leamey and Sawatari, 2014; Woelfle et al., 2016; Sudhof, 2017). The high-affinity heterophilic interaction of TENs with latrophilins (LPHN1-3), a family of adhesion G Protein Coupled Receptors (GPCRs) that have key roles in synapse development and embryogenesis, regulates synapse formation and organization (Silva et al., 2011; Boucard et al., 2014). On the other hand, trans-homodimerization of TENs is reported to be critical for correct matching of axons with their partner dendrites, a process that is critical for neural circuit-wiring in the nervous system (Hong et al., 2012; Mosca et al., 2012; Mosca, 2015).

Surprisingly, Li et al. (2018) showed that an alternatively spliced seven-residue region (NKEFKHS) within the β-propeller

FIGURE 2 | Different splice variants of TENs regulate distinct synapse specifications. Alternative splicing is a molecular switch to regulate which binding partner TEN2 binds to, and what respective function TEN2 does. On the left side, the -SS isoform of TEN2 (empty star) interacts with LPHN. When it is co-expressed with FLRT, TEN2 isoform that lacks the splice insert induces excitatory postsynaptic differentiation. On the right side, TEN2 that includes the splice insert (full star) cannot interact with LPHN, but it can form trans-homodimers to mediate neural circuit-wiring and induces inhibitory synapses. The left and right sides of the TEN2 dimer represent excitatory and inhibitory synapses, respectively. The membranes, teneurin structure (Li et al., 2018) (PDB ID: 6CMX), and distance between the membranes are drawn to scale. Alternative splice site is shown by red star. Insert: Representative negative stain EM micrographs of TEN2 ECR(1-5) shows a TEN2 -SS cis-dimer. Yellow arrows indicate EGF repeats. Figure modified from Li et al. (2018).

acts as a switch to regulate trans-cellular adhesion of TEN2 to LPHNs (**Figure 2**, red star). The splice variant that lacks the seven amino acids can bind to full-length LPHN presented on the neighbor cells in cell-aggregation assays and this interaction activates trans-cellular signaling in a LPHN-dependent manner (**Figure 2**, left side). The other splice variant that includes the seven amino acids, however, is unable to interact with full-length LPHN in identical experiments (**Figure 2**, right side). Similarly, the same alternatively spliced site that regulates the TEN/LPHN interaction has been reported to also regulate TEN transhomodimerization (Berns et al., 2018). Intriguingly, the splice variant that lacks these seven amino acids cannot mediate transhomodimerization, whereas the other variant can (**Figure 2**, right side). Taken together, these results suggest that TEN splice variants can either interact with LPHNs or mediate transdimerization with itself, but a single variant cannot mediate both interactions.

#### DIFFERENT SPLICE VARIANTS OF TENS PERFORM DISTINCT SYNAPSE SPECIFICATIONS

As different TEN splice variants are involved in different ligand interactions, the next question arose: Do different TEN splice variants perform biologically distinct functions? Subsequent studies by Li et al. (2018) examined the capability of TEN2 + SS and TEN2 -SS to induce artificial synapse formation. In these assays, HEK293 cells expressing TEN variants were co-cultured with primary neurons; and inhibitory and excitatory synapse formation was monitored for pre- and post-synaptic differentiation for both types of synapses (Li et al., 2018). The results showed that TEN2 + SS induced GABAergic (inhibitory) synaptic specializations but failed to induce glutamatergic (excitatory) synaptic specifications (**Figure 2**, right side; Li et al., 2018).

#### REFERENCES


On the other hand, initially, TEN2 -SS failed to recruit both excitatory and inhibitory synaptic markers (Li et al., 2018). However, when fibronectin leucine rich repeat transmembrane protein (FLRT3), another LPHN ligand that on its own is unable to induce pre- or postsynaptic specializations, was coexpressed in HEK293T cells with the TEN2 -SS, these molecules together potently induced excitatory but not inhibitory postsynaptic specializations (Sando et al., 2019; **Figure 2**, left side). The other splice variant, TEN + SS (that cannot bind LPHN and cannot induce excitatory synapses) was still inactive in excitatory synapse formation, even when coexpressed with FLRT (Sando et al., 2019). Thus, TEN -SS and FLRT can stimulate excitatory synapse formation in combination, but not separately. These results are consistent with studies in transgenic LPHN mice in vivo, where coincident binding of both TEN and FLRT to LPHN is required for excitatory synapse formation (Sando et al., 2019).

As we start to understand the functions of TENs, these observations are likely only a glimpse of the complexity of the TEN system. Many questions arise: How can a seven-residue splice site affect ligand binding that is away from the ligand binding site on TEN (**Figure 2**, left side; Li et al., 2018). What are the implications of various TEN variants and isoforms for neural wiring? Do other splice variants have important biological functions? What are the other unknown ligands for TENs? What is the role of TEN/LPHN interaction in other systems such as embryonic development? TEN field is awaiting further exciting and surprising findings.

#### AUTHOR CONTRIBUTIONS

Both authors wrote the manuscript.

#### FUNDING

This work was supported by grants R01 GM120322 and R01 GM134035-01 to DA and the American Heart Association grant #19POST34380439/Jingxian Li/2019-2020.


differentiation and patient survival in ovarian cancer. PLoS One 12:e0177244. doi: 10.1371/journal.pone.0177244


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Araç and Li. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Ancient Function of Teneurins in Tissue Organization and Neuronal Guidance in the Nematode Caenorhabditis elegans

Ulrike Topf and Krzysztof Drabikowski\*

Institute of Biochemistry and Biophysics, Polish Academy of Sciences, Warsaw, Poland

The nematode Caenorhabditis elegans expresses the ten-1 gene that encodes teneurin. TEN-1 protein is expressed throughout the life of C. elegans. The loss of ten-1 function results in embryonic and larval lethality, highlighting its importance for fundamental processes during development. TEN-1 is expressed in the epidermis and neurons. Defects in neuronal pathfinding and epidermal closure are characteristic of ten-1 loss-offunction mutations. The molecular mechanisms of TEN-1 function in neurite outgrowth, neuronal pathfinding, and dendritic morphology in C. elegans are largely unknown. Its genetic redundancy with the extracellular matrix receptors integrin and dystroglycan and genetic interactions with several basement membrane components suggest a role for TEN-1 in the maintenance of basement membrane integrity, which is essential for neuronal guidance. Identification of the lat-1 gene in C. elegans, which encodes latrophilin, as an interaction partner of ten-1 provides further mechanistic insights into TEN-1 function in neuronal development. However, receptor-ligand interactions between LAT-1 and TEN-1 remain to be experimentally proven. The present review discusses the function of teneurin in C. elegans, with a focus on its involvement in the formation of receptor signaling complexes and neuronal networks.

#### Edited by:

Antony Jr. Boucard, Centro de Investigación y de Estudios Avanzados (CINVESTAV), Mexico

#### Reviewed by:

Andrew Chisholm, University of California, San Diego, United States Tobias Langenhan, Leipzig University, Germany

> \*Correspondence: Krzysztof Drabikowski drabikowski@ibb.waw.pl

#### Specialty section:

This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Neuroscience

Received: 22 December 2018 Accepted: 22 February 2019 Published: 08 March 2019

#### Citation:

Topf U and Drabikowski K (2019) Ancient Function of Teneurins in Tissue Organization and Neuronal Guidance in the Nematode Caenorhabditis elegans. Front. Neurosci. 13:205. doi: 10.3389/fnins.2019.00205 Keywords: teneurin, TEN-1, C. elegans, basement membrane, latrophilin, LAT-1, axon guidance

#### INTRODUCTION

Teneurins are large single-pass transmembrane glycoproteins that are conserved in most animals with a nervous system (Tucker, 2018). Teneurins were first discovered in Drosophila (Baumgartner and Chiquet-Ehrismann, 1993; Baumgartner et al., 1994) and later described in Caenorhabditis elegans (Drabikowski et al., 2005) and vertebrate models (Minet et al., 1999). Teneurins are involved in several developmental processes in invertebrate models and expressed most prominently in developing neuronal tissues, contributing to neuronal patterning and axon guidance (Drabikowski et al., 2005; Tucker et al., 2007; Silva et al., 2011; Mosca et al., 2012; Mosca, 2015). The family of teneurin proteins is characterized by a distinct protein domain architecture. Their extracellular domain consists of eight epidermal growth factor (EGF)-like repeats, a region of conserved cysteines, and unique tyrosine and aspartate (YD)-repeats and is highly conserved among vertebrates and invertebrates. The structures of some extracellular domains of chicken

Ten2 and mouse Ten3 were recently solved, revealing a previously unpredicted TTR transthyretin-related domain that plays roles in protein aggregation and lipid recognition in other teneurin-unrelated proteins (Jackson et al., 2018; Li et al., 2018). Unlike the extracellular domain, the composition of the intracellular domain of teneurin proteins, with exception of some predicted phosphorylation sites, is very different between vertebrates and invertebrates. The intracellular domain can be cleaved off and translocate to the nucleus, whereas the extracellular domain can be released into the extracellular milieu. The ability of the intracellular domain to mediate cellular signaling within the nucleus was first observed in a cell culture model of vertebrate teneurin-2, in which overexpressed variants of teneurin-2 colocalized with promyelocytic leukemia protein (PML) bodies (Bagutti et al., 2003). However, the intracellular domain of the endogenous teneurin protein was found in the nucleus only in C. elegans (Drabikowski et al., 2005). Some studies have described a nuclear function of the teneurin intracellular domain that regulates transcription as a transcriptional repressor or activator (Bagutti et al., 2003; Nunes et al., 2005; Scholer et al., 2015; Glendining et al., 2017). However, the mechanism by which the membrane-spanning full-length teneurin protein is released to the intracellular domain from the plasma membrane is mostly unknown. Furincleavage sites between the transmembrane domain and the EGF-like repeats were suggested to be one such processing (Tucker and Chiquet-Ehrismann, 2006; Kenzelmann et al., 2007). This was supported by experiments in which recombinant avian teneurin-2 protein was cleaved by furin protease (Rubin et al., 1999). Similar to the mechanism of processing, signals that trigger the release of the intracellular domain remain to be discovered. Efforts to identify binding partners of the extracellular domain revealed various interactions that contributed to deciphering teneurin function as an organizer of neuronal networks (Mosca, 2015). Vertebrate teneurins form homo- and heterophilic interactions (Feng et al., 2002; Rubin et al., 2002; Beckmann et al., 2013). In Drosophila, teneurins mediate synaptic connections and neuromuscular connections via homophilic interactions (Hong et al., 2012; Mosca et al., 2012). In hippocampal neurons, teneurin-2 acts as a postsynaptic receptor for latrophilin (Silva et al., 2011). However, the ways in which this interaction contributes to synapse formation are unknown. Moreover, teneurin-1 interacts with beta-dystroglycan, resulting in cytoskeletal rearrangements (Chand et al., 2012). Whether teneurin-1 is expressed post- or presynaptically remains unclear. In Drosophila, an interaction between ten-m and integrin in motor neurons and muscles was proposed to be important for normal synaptic function, but the mechanism by which this occurs is unclear (Mosca et al., 2012; Dani et al., 2014). Studies in C. elegans revealed a fundamental role for teneurin in tissue organization and neuronal network development and maintenance (Drabikowski et al., 2005; Trzebiatowska et al., 2008; Morck et al., 2010; Topf and Chiquet-Ehrismann, 2011; Promel et al., 2012).

The present mini review provides an overview of the various identified genetic interactions with the ten-1 gene in C. elegans, providing insights into its ancient function. We focus especially on ten-1-latrophilin connections, which are discussed within the context of recent findings in vertebrate models.

#### TEN-1 EXPRESSION AND LOSS-OF-FUNCTION PHENOTYPE

Most species express several teneurin paralogs (Tucker et al., 2012). Genetic redundancy has impeded investigations of their biological functions because single deletions show minimal phenotypic alterations. In contrast, the C. elegans genome encodes only one teneurin gene, ten-1. This fact, combined with the tremendous genetic tractability of the model organism, makes C. elegans an attractive system to investigate the biological significance of teneurins. Expression of the ten-1 gene is under the control of two promoters that give rise to two transcript versions. These transcripts only differ in the length of the part that codes for the intracellular domain and thus were named short ten-1a and long ten-1b. These two forms of TEN-1 are expressed throughout worm development and in many tissues but have distinct expression patterns. An extensive analysis of TEN-1 expression was performed, in which green fluorescent protein (GFP) was expressed under two different promoters, pten-1a (which controls the expression of TEN-1L) and pten-1b (which controls the expression of TEN-1S). pten-1a was mostly active in the mesoderm, with prominent expression in muscles and the intestine, whereas pten-1b was active in the ectoderm, predominantly in neurons, including the soma and axons (Drabikowski et al., 2005). Using specific antibodies against the N-terminal part of TEN-1L, the intracellular domain was detected in the nucleus (Drabikowski et al., 2005). Expression of the ten-1 transgene that was fused to GFP under the control of pten-1b confirmed epidermal and neuronal expression patterns in the embryonic stage, indicating the potential involvement of TEN-1 in neuronal development (Topf and Chiquet-Ehrismann, 2011).

The depletion of TEN-1 by RNAi results in severe morphological defects. Worms that were injected with RNAi against both transcripts exhibited an increase in embryonic lethality, accompanied by gross defects in hypodermal cell migration. These findings supported the importance of TEN-1 during early worm development. Prominent post-embryonic defects included generally abnormal body morphology, morphological defects in the reproductive system, defects in muscles, and abnormalities in neuronal migration and axonal pathfinding (Drabikowski et al., 2005). A smaller brood size and morphological defects were confirmed in several TEN-1 mutants. Three mutant alleles of ten-1 have been characterized (ok641, tm651, et5; (Drabikowski et al., 2005; Trzebiatowska et al., 2008; Morck et al., 2010). Ten-1(ok641) and ten-1(tm651) are null alleles, and ten-1(et5) is a hypomorphic allele with a weaker post-embryonic phenotype. Neuronal defects in the TEN-1 mutants were not as penetrant as during RNAi depletion but were predominantly observed in mutant worms that exhibited other morphological defects, including epidermal defects (Drabikowski et al., 2005; Morck et al., 2010). Migration defects were observed in some neurons in otherwise healthy-looking

animals, suggesting that TEN-1 function is specifically required for some neurons. However, the migration and pathfinding of neurons also strongly depend on an intact basement membrane. The basement membrane is a specialized extracellular matrix that surrounds most tissues in all Metazoa. TEN-1 is expressed in all major tissues in C. elegans and consists of a large extracellular part with several different structural domains, suggesting that it likely interacts with components of the extracellular milieu.

# GENETIC INTERACTIONS OF TEN-1 IN C. elegans

Several studies have identified multiple genetic interactions with ten-1 (Byrne et al., 2007; Trzebiatowska et al., 2008; Morck et al., 2010; Topf and Chiquet-Ehrismann, 2011; Promel et al., 2012). To date, however, none of these interaction partners have been shown biochemically to interact physically with TEN-1. A high-throughput screen identified glp-1, a receptor of the NOTCH family, as a ten-1 interacting partner (Byrne et al., 2007). Glp-1 is essential for the development of worm gonads as well as TEN-1, and the depletion of glp-1 together with ten-1 is embryonically lethal. Nevertheless, the functional basis of this interaction remains to be determined. Further attempts to investigate the function of ten-1 focused on interactions with genes that encode basement membrane receptors and components and genes that are involved in regulating the cytoskeleton, neuronal guidance, and axon outgrowth (Trzebiatowska et al., 2008; Morck et al., 2010; Topf and Chiquet-Ehrismann, 2011). **Table 1** presents an overview on these genetic interactions with ten-1 (The reader is advised to see original publications for further details on the described phenotypes). Based on phenotypical observations of ten-1 mutant worms that showed the loss of basement membrane integrity, which surrounds the developing gonad in post-embryonic worms, Trzebiatowska et al. (2008) applied a candidate approach and found that ten-1 genetically interacted with the basement membrane receptor α-integrin and dystroglycan and basement

TABLE 1 | Genetic interactions with ten-1 that shape neuronal networks in C. elegans.


<sup>1</sup>Growth temperature of 20◦C. <sup>2</sup>Observed in pharyngeal M2 neuron; percentage in brackets reflects the animals with defects in M2 neuron. Possibly other neuronal defects are here not taken in account. Synthetic lethal, development ends at embryonic or early larval stage. Wnt, wingless/integrated. Ena/VASP, enabled/vasodilatorstimulated phosphoprotein.

membrane components lamin and nidogen (Trzebiatowska et al., 2008). Double mutants of all four genes together with ten-1 resulted in a synthetic lethal or sick phenotype that terminated the development of the double-mutant worms during embryogenesis or at an early larval stage. Previous studies in neuroblastoma cells found that the teneurin-2-dependent induction of filopodia formation was more prominent on lamin substrate (Rubin et al., 1999), and chicken teneurin-2 was shown to colocalize with lamin in basement membranes of the optic cup (Tucker et al., 2001). These findings in C. elegans suggested that teneurin is a receptor that might act redundantly with integrin or dystroglycan in basement membrane function. The ten-1 single mutants display pleiotropic phenotype and those seem likely to show genetic interactions with various genes. However, there is specificity of the interaction between ten-1 and basement membrane components. Mutations in cle-1 (CoLlagen with endostatin domain 1; vertebrate type XV/XVIII collagen homolog in C. elegans) or unc-52 (perlecan) did not enhance embryonic lethality, larval arrest, or sterility of the ten-1(ok641) mutant (Trzebiatowska et al., 2008). Loss-of-function phenotypes of nidogen (nid-1), dystroglycan (dgn-1), and integrin (ina-1) in worms involve defects in the nervous system (Baum and Garriga, 1997; Kang and Kramer, 2000; Kim and Wadsworth, 2000). However, Trzebiatowska et al. (2008) did not investigate neuronal defects in double-mutant worms. Such studies may be difficult because of the early death of such mutant animals. An unbiased genetic screen of ten-1-interacting partners identified phy-1, a prolyl-4-hydroxylase that is important for the modification of procollagens, which are secreted into the extracellular milieu, including basement membranes. The ten-1 also genetically

interacts with collagen IV (let-2 in C. elegans); (Topf and Chiquet-Ehrismann, 2011). Collagen IV is required for the completion of embryonic development, tissue organization, and structural integrity. Collagen IV is produced in muscle cells, and insufficient maturation results in the intracellular retention and aggregation of procollagen. Consequently, the combined loss of phy-1 and ten-1 resulted in deteriorated connections between the epidermis and muscle tissue. Drosophila Ten-a protein also localizes to muscle attachment structures (Kenzelmann-Broz et al., 2010), and the mouse teneurin isoform TEN3 (Odz3) colocalizes with collagens I and II (Murakami et al., 2010). Epidermal defects in ten-1 and phy-1 double-mutant worms were accompanied by neuronal defects. Further evidence that TEN-1 is involved in neuronal guidance was provided by a candidate approach, in which defects in pharyngeal neurons were quantified, with a focus on M2 neurons (Morck et al., 2010). The loss of ten-1 together with genes that are involved in M2 cell body positioning and axon outgrowth resulted in more sever defects in M2 neuron. Among the interacting genes are sax-3 gene and downstream-acting unc-34 gene, which are involved in multiple aspects of sensory, motor, and interneuron axon guidance.

Genetic interaction data have provided strong evidence that teneurin in C. elegans is required for the maintenance of basement membrane integrity. Whether this function is based on structural tasks of teneurin that involve the binding of extracellular proteins or teneurin as a receptor that provides guidance for migrating cells remains to be determined. Nevertheless, this ancient function of TEN-1 may have served to organize and connect tissues, thus providing a foundation for development of the worm's neuronal network.

FIGURE 1 | Teneurin-latrophilin interactions. (A) In vertebrates, teneurin is a part of protein complexes that connect pre- and postsynaptic parts of neurons. Teneurin physically interacts with latrophilin and also with the basement membrane receptor dystroglycan. Latrophilin is connected to the netrin receptor via FLRT. (B) C. elegans TEN-1S is expressed in hypodermal cells and neurons and genetically interacts with dystroglycan Dgn-1 and the netrin receptor Unc-5. The latrophilin LAT-1 is expressed in hypodermal cells. Whether TEN-1S and LAT-1 interact is unknown (arrow with question mark). (C) C. elegans TEN-1L is expressed in many more cells and tissues compared with LAT-1. Thus, cis interactions might be possible but have not yet been proven. A presumed interaction between TEN-1L and LAT-1 could trigger the release of the intracellular domain of TEN-1L, initiating cell signaling pathways. FLRT, (fibronectin leucine-rich repeat transmembrane protein), arrows indicate genetic interaction.

# PHYSICAL INTERACTIONS OF TENEURINS IN OTHER SPECIES

Recent biochemical and structural studies in vertebrate systems showed physical interactions between teneurins and other membrane receptors. Particularly interesting is the interaction between teneurin and latrophilin. Latrophilins (LPHN1-3) belong to the adhesion-type G-protein-coupled receptor (GPCR) family. LPHN1 was identified as a receptor for α-latrotoxin, a black widow spider toxin that triggers massive neurotransmitter release from neurons and neuroendocrine cells. In vertebrates, latrophilins interact with FLRTs (fibronectin leucine-rich repeat transmembrane proteins), UNC5 (netrin receptor), neurexins, and teneurins (Davletov et al., 1996; Krasnoperov et al., 1997). Latrophilin, UNC5, and FLRT form a super complex (Lu et al., 2015; Jackson et al., 2016). In neurons, latrophilin is presynaptic and teneurin is postsynaptic, and both proteins engage in trans interactions (**Figure 1A**). A recently published cryo-electron microscopy structure of human latrophilin 1 with teneurin 2 described this interaction in detail (Li et al., 2018).

# TEN-1 INTERACTION WITH LAT-1/LATROPHILIN IN C. elegans

The C. elegans genome contains two latrophilin paralogs, lat-1 and lat-2 (Willson et al., 2004; Langenhan et al., 2009). The lat-1 is expressed in oocytes, early embryonic blastomeres, and precursors of pharyngeal and hypodermal cells. In larvae and adult worms, lat-1 is expressed in pharynx muscle nerve cells, the gonads, and the vulva. The neuronal and gonadal expression of lat-1 has only been mentioned and not thoroughly described (Langenhan et al., 2009). Lat-2 is expressed in the pharynx and gland cells of the excretory system. The lat-1 deletion is embryonically lethal, and the escapees have a smaller brood size because of defects in sperm development. The lat-2 deletion has no obvious phenotype but enhances the lat-1-null phenotype (Langenhan et al., 2009). LAT-1 in C. elegans has mostly been studied in the context of early embryogenesis, the alignment of mitotic spindle and division planes, and the establishment of anterior-posterior polarity. Comparing the expression of lat-1 and ten-1 is difficult because of insufficient descriptions of lat-1::GFP expression patterns. The expression pattern of lat-1 partially overlaps with ten-1a to a small extent in the developing pharynx. LAT-1 expression in dorsal hypodermis during intercalation partially overlaps with ten-1b promoter expressing Ten1S version of teneurin (Drabikowski et al., 2005).

In C. elegans, lat-1, and ten-1 genetically interact, but the physical interaction has not been demonstrated. In genetic interactions, the alleles displayed non-allelic noncomplementation. The loss of any of the alleles of either gene led to developmental defects, and double-heterozygote worms exhibited strong defects in development and fertility (Promel et al., 2012). The authors showed that the ten-1a promoter is active in the same half of intercalating hypodermal cells as lat-1. Thus, according to these authors, LAT-1 is likely not a receptor

for TEN-1L (**Figure 1B**). Interactions between latrophilin and teneurin in trans but not in cis have been proven biochemically, microscopically, and structurally in vertebrate systems (Davletov et al., 1996; Krasnoperov et al., 1997). Drabikowski et al. (2005) showed that ten-1b promotor-expressing TEN-1S and not ten-1a-expressing TEN-1L is expressed in the intercalating dorsal hypodermis in C. elegans in both the left and right rows of cells. The approach to obtain transgenic animals undertaken by Promel et al. (2012) expressing lat-1::GFP and by Morck et al. (2010) expressing ten-1a::GFP promoter fusions often result in random transgene silencing in a subset of cells. Kelly et al. (1997) have shown that simple, highly repetitive extrachromosomal arrays, as in this case used by Morck et al. (2010); Promel et al. (2012), result in transgene silencing. Thus, these expression patterns might reflect only partial expression pattern of LAT-1 and TEN-1 and conclusions drawn from them should be treated with caution. Regardless of whether lat-1 is expressed in all or only half of intercalating hypodermal cells, TEN-1S appears to be expressed in all intercalating hypodermal cells, thus indicating that in trans interactions between LAT-1 and TEN-1S are possible (**Figure 1C**). TEN-1S protein has a short, 36-amino-acid intracellular domain that does not translocate to the nucleus. Both proteins, TEN-1 and LAT-1, and the processes in which they are involved, are strongly conserved in evolution. Thus, it is highly unlikely that the nature of interactions, in trans or in cis, between these proteins would not be conserved. The elucidation of endogenous expression patterns of both LAT-1 and TEN-1 in C. elegans (e.g., by CRISPR/Cas9 technology) may help resolve these discrepancies.

In C. elegans, possible TEN-1 interactions with LAT-1 that are related to neuronal pathfinding and synapse formation await further investigation. In recent years, teneurin research in vertebrates has focused on neuronal function and interactions with latrophilin. Studies of early expression during mouse and chicken embryogenesis have shown that teneurins function not only in neuronal development but also in non-neuronal tissues during the pattern formation of developing limbs (Tucker et al., 2007), somites, and craniofacial mesenchyme (Tucker et al., 2001). Investigations of teneurins in non-neuronal tissues in vertebrates are still incipient but have already opened new avenues of research on both cancer and congenital diseases. Findings in worms may further guide such research.

# CONCLUSION

Research on teneurin proteins has seen tremendous advances. Teneurins were discovered in 1993 in the labs of Ruth Chiquet-Ehrismann and of Roland Fässler. Since that time, however, the biological role of teneurins in humans has remained elusive. Several excellent studies have been performed in model organisms and cell culture systems, indicating that teneurins play a role as organizers of neuronal networks. Studies of teneurin in C. elegans have demonstrated its importance during development. The elucidation of multiple genetic interactions has shown that teneurin is essential for pattern formation, cell migration, and development of the nervous system. The ancestral

function of teneurin in the nervous system in C. elegans is most pronounced through TEN-1 interactions and the maintenance of basement membrane and tissue integrity. In vertebrates, teneurin function evolved in concert with the multiplication of teneurin genes. Further investigations are required to establish the role of TEN-1 as an organizer of neuronal networks in C. elegans and the involvement of LAT-1 in these processes. State-of-theart genetic tools in worms, combined with detailed descriptions of their development and neuronal connectivity at single-synapse resolution, make this a very promising area of research.

#### REFERENCES


#### AUTHOR CONTRIBUTIONS

UT and KD co-wrote the manuscript and reviewed the references. KD drafted **Figure 1**.

# FUNDING

UT was supported by National Science Centre grant 2015/19/B/NZ1/03444.



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Topf and Drabikowski. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# The Tenets of Teneurin: Conserved Mechanisms Regulate Diverse Developmental Processes in the Drosophila Nervous System

Alison T. DePew† , Michael A. Aimino† and Timothy J. Mosca\*

Department of Neuroscience, Thomas Jefferson University, Philadelphia, PA, United States

#### Edited by:

Antony Jr. Boucard, Centro de Investigación y de Estudios Avanzados (CINVESTAV), Mexico

#### Reviewed by:

Robert Hindges, King's College London, United Kingdom Pei-San Tsai, University of Colorado Boulder, United States

\*Correspondence: Timothy J. Mosca timothy.mosca@jefferson.edu

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Neuroscience

Received: 19 October 2018 Accepted: 11 January 2019 Published: 30 January 2019

#### Citation:

DePew AT, Aimino MA and Mosca TJ (2019) The Tenets of Teneurin: Conserved Mechanisms Regulate Diverse Developmental Processes in the Drosophila Nervous System. Front. Neurosci. 13:27. doi: 10.3389/fnins.2019.00027 To successfully integrate a neuron into a circuit, a myriad of developmental events must occur correctly and in the correct order. Neurons must be born and grow out toward a destination, responding to guidance cues to direct their path. Once arrived, each neuron must segregate to the correct sub-region before sorting through a milieu of incorrect partners to identify the correct partner with which they can connect. Finally, the neuron must make a synaptic connection with their correct partner; a connection that needs to be broadly maintained throughout the life of the animal while remaining responsive to modes of plasticity and pruning. Though many intricate molecular mechanisms have been discovered to regulate each step, recent work showed that a single family of proteins, the Teneurins, regulates a host of these developmental steps in Drosophila – an example of biological adaptive reuse. Teneurins first influence axon guidance during early development. Once neurons arrive in their target regions, Teneurins enable partner matching and synapse formation in both the central and peripheral nervous systems. Despite these diverse processes and systems, the Teneurins use conserved mechanisms to achieve these goals, as defined by three tenets: (1) transsynaptic interactions with each other, (2) membrane stabilization via an interaction with and regulation of the cytoskeleton, and (3) a role for presynaptic Ten-a in regulating synaptic function. These processes are further distinguished by (1) the nature of the transsynaptic interaction – homophilic interactions (between the same Teneurins) to engage partner matching and heterophilic interactions (between different Teneurins) to enable synaptic connectivity and the proper apposition of pre- and postsynaptic sites and (2) the location of cytoskeletal regulation (presynaptic cytoskeletal regulation in the CNS and postsynaptic regulation of the cytoskeleton at the NMJ). Thus, both the roles and the mechanisms governing them are conserved across processes and synapses. Here, we will highlight the contributions of Drosophila synaptic biology to our understanding of the Teneurins, discuss the mechanistic conservation that allows the Teneurins to achieve common neurodevelopmental goals, and present new data in support of these points. Finally, we will posit the next steps for understanding how this remarkably versatile family of proteins functions to control multiple distinct events in the creation of a nervous system.

Keywords: teneurin, Drosophila, synapse formation, partner matching, cytoskeleton, NMJ, olfaction, spectrin

# INTRODUCTION

fnins-13-00027 January 28, 2019 Time: 18:40 # 2

In the nervous system, each neuron undergoes a simultaneously elegant yet complex development. Disparate molecular, cellular, and morphological events are woven together into a united entity, linking cell birth, neuronal differentiation, cell migration, membrane adhesion, synapse formation, and synaptic refinement. These diverse processes, with their distinctive molecular, developmental, and cell biological requirements, are united by a common, broad goal: forming the functional connections essential for life. When the diversity of neuronal subtype in different brain regions, layers, and even systems (peripheral versus central) is added to this already herculean list, it becomes apparent that achieving proper development is no easy task. Each process has its own distinct molecular and physical requirements and challenges, and these processes need to be seamlessly connected both spatially and temporally. If these events occur in the wrong order or in the wrong place, development can go awry, resulting in intellectual disabilities and neurodevelopmental disorders including autism, schizophrenia, and bipolar disorder (Grant, 2012; Gilbert and Man, 2017; Zerbi et al., 2018). Thus, the underlying processes must be finely tuned to ensure fidelity in neurodevelopment.

How are these disparate tasks accomplished? Based on estimations of neuronal diversity (Lodato and Arlotta, 2015), genome sizes (Adams et al., 2000; Cravchik et al., 2001), and synapse number in the brain (Silbereis et al., 2016), it would be impossible to employ a different approach with distinct molecular cues and mechanistic underpinnings for each event. This would require more distinct adhesion and recognition cues than there are actual genes in the genome. There must be some shared use of molecules and processes. Indeed, this is commonly observed throughout development where different classes of neurons use similar molecules and mechanisms to accomplish the goals of axon guidance (Dickson, 2002), synapse formation (Favuzzi and Rico, 2018), and neuronal migration (Geschwind and Rakic, 2013). We see this concept in our cities frequently, in the form of "adaptive reuse": a decommissioned water pumping station becomes a gastropub, a turn-of-thecentury bank becomes a museum, and even a former firehouse becomes a luxury apartment complex. This process of using an "old" molecule or concept for a purpose other than its original intent enables considerable utility. At the molecular level, we see genes originally intended for cell adhesion adaptively reused to form synapses (Giagtzoglou et al., 2009; Sun and Xie, 2012) and cytoskeletal molecules used for movement repurposed for cell migration (Etienne-Manneville, 2013). Backed by this concept, the list of distinct processes needed for neuronal development becomes more manageable, as does its molecular requirements.

Recent years have seen an explosion of research on a family of large cell surface proteins called the Teneurins (Young and Leamey, 2009) that play diverse roles in organismal development (Tucker and Chiquet-Ehrismann, 2006). In the fruit fly, Drosophila melanogaster, the two Teneurin homologs, Ten-m and Ten-a, were originally thought to be involved in body segment patterning (Baumgartner and Chiquet-Ehrismann, 1993; Baumgartner et al., 1994; Levine et al., 1994; Rakovitsky et al., 2007). The last decade, however, has seen the discovery of roles for these cell surface proteins in multiple neurodevelopmental processes including axon guidance, synaptic partner matching, and synapse organization (Zheng et al., 2011; Hong et al., 2012; Mosca et al., 2012; Mosca and Luo, 2014). Drosophila has proven an outstanding model system to assess Teneurin function in that its many genetic tools (Venken et al., 2011), accessible synapses at the NMJ (Harris and Littleton, 2015) and in the olfactory system (Mosca and Luo, 2014) and its stereotyped wiring (Keshishian et al., 1996; Couto et al., 2005) enable detailed molecular and mechanistic study at the singlecell level. In such discovery, a theme of adaptive reuse surfaced for the Teneurins: the same genes controlling multiple steps of neurodevelopment via similar mechanisms. We will focus on two of these processes: synaptic partner matching and synaptic organization to describe recent work highlighting roles for the Drosophila Teneurins in both these processes as well as their shared mechanistic underpinnings.

# MATERIALS AND METHODS

#### Drosophila Genetics

All stocks and crosses were raised on standard cornmeal/dextrose medium at 25◦C in a 12/12 light/dark cycle. Canton S. served as the control strain (Woodard et al., 1989). Df (X) ten-a was used as a ten-a null mutant (Mosca et al., 2012). Mef2-GAL4 was used to drive expression in all muscles (Lilly et al., 1995). SG18.1-GAL4 was used to drive expression in all ORNs (Shyamala and Chopra, 1999). We also used the transgenic strains UAS-Ten-a (Mosca et al., 2012) and UAS-ten-mRNAi−V<sup>51173</sup> (Mosca et al., 2012) for Ten-a expression and ten-m RNAi knockdown, respectively.

# Staining, Spaced Stimulation, and Immunocytochemistry

Spaced stimulation was conducted as previously described (Piccioli and Littleton, 2014). Wandering third instar larvae were processed for immunocytochemistry as previously described (Mosca and Schwarz, 2010). The following primary antibodies were used: mouse anti-Ten-m at 1:500 (Levine et al., 1994), rabbit anti-Dlg at 1:40000 (Koh et al., 1999), rabbit anti-Syt I at 1:4000 (Mackler et al., 2002). Alexa488- and Alexa546 conjugated secondary antibodies were used at 1:250 (Jackson ImmunoResearch and Invitrogen). Cy5-conjugated antibodies to HRP were used at 1:100 (Jackson ImmunoResearch).

#### Olfactory Behavior Trap

Olfactory behavior experiments were conducted and analyzed as previously described (Mosca et al., 2017).

#### Genotypes

**Figure 3**: Control (+; +; +; +); ten-a −/− (Df (X) ten-a; +; +; +); ten-a −/− + ORN Ten-a (Df (X) ten-a; SG18.1-GAL4/UAS-Ten-a; +; +). **Figures 4A,C** (+; +; +; +); **Figure 4B** (+; Mef2- GAL4/+; UAS-ten-mRNAi−V51173/+; +).

### PARTNER MATCHING

fnins-13-00027 January 28, 2019 Time: 18:40 # 3

Before any synapse can be made and organized, the pre- and postsynaptic cells must first identify each other as appropriate partners and begin connecting in a process called partner matching. While there has been extensive research done on many different aspects of synapse formation, the step of partner matching remains poorly understood. In 1963, Roger Sperry proposed that such a process may occur by 'individual identification tags, presumably cytochemical in nature' (Sperry, 1963). While the intervening 56 years have suggested a more complex mechanism including (but not limited to) such tags, no clear-cut cases of "Sperry" partner matching molecules that promote a direct, selective association of individual pre- and postsynaptic neurons had been identified. Due to a need to understand the molecular underpinnings of this process, though, a number of studies in recent years identified the Teneurins as key players in the partner matching step at the Drosophila neuromuscular junction (NMJ) and in the olfactory system (Hong et al., 2012; Mosca et al., 2012). As such, a tempting conclusion is that in both the central and peripheral nervous systems, the Teneurins provide the strongest case to date for a Sperry molecule participating in these events.

In the relatively simple setting of the Drosophila NMJ, a single hemisegment contains 34 motoneurons that must each identify their appropriate muscle target among 31 options (Keshishian et al., 1993; Nose, 2012). What would a partner matching molecule look like in this situation? It would have to be expressed in both the presynaptic motoneuron and the postsynaptic muscle. Further, it would have to be expressed in a limited subset of motoneuron::muscle pairs – widespread expression would suggest it's necessary for all connections, not specific ones. The first hypothesis that the Teneurins could serve this role at the NMJ came from expression studies. Though a basal level of Ten-m expression occurs in all larval muscles, two motoneuron::muscle pairs, those of muscle 3 and muscle 8, specifically express elevated levels of Ten-m (Mosca et al., 2012). To determine whether this expression was related to their function, work focused on altering Ten-m expression to better understand its function (Mosca et al., 2012). Knockdown of tenm expression in muscle 3 and its innervating neuron (where it is normally highly expressed) increases the failure rate of innervation. The same failure occurs whether ten-m was knocked down in only the neuron or the muscle (Mosca et al., 2012), suggesting that both muscle 3 and its motoneuron require ten-m to properly match (i.e., both pre- and postsynaptic partners). This finding supports the idea of a homophilic interaction between pre- and postsynaptic Ten-m (**Figure 1**). This homophilic specificity was further elucidated by experiments showing that ten-a knockdown lacked such defects and Ten-a overexpression could not compensate for loss of Ten-m with respect to matching (Mosca et al., 2012).

Intriguingly, not only is Ten-m required for partner matching at the NMJ, it can also instruct matching between cells that do not normally connect. At muscles 6 and 7 in the developing larva, 40% of the boutons from the same motoneurons form on muscle 7 while the remainder form on muscle 6 (Johansen et al., 1989). However, the misexpression of Ten-m only in muscle 6 (but not 7) and both of their accompanying motoneurons, shifts the balance of connections to predominantly favor muscle 6 (Mosca et al., 2012). This suggests that expression of ten-m in cells that do not normally express it in high levels can direct partner-matching, similar to situations where ten-m expression is reduced. Interestingly, this was not recapitulated with Tena, suggesting not only a homophilic interaction between Ten-m that was instructive, but that there is some level of specificity regarding Ten-m over Ten-a. Therefore, the presynaptic level of Ten-m must be equivalent to the postsynaptic level of Ten-m for partner-matching to occur correctly.

This mechanism between pre- and postsynaptic targets suggested the first tenet of Teneurin function: transsynaptic interaction leading to partner matching, here, a homophilic interaction. But how do the Teneurins mediate this? What functions downstream of teneurin::teneurin interaction? A tantalizing possibility comes from work done to characterize the role of Teneurins in motor axon growth cone guidance (Zheng et al., 2011). Loss of ten-m caused aberrations in fasciculation that resulted in inter-segmental nerves moving to incorrect regions of the NMJ while ectopic overexpression of ten-m in the epidermis also caused axon migration defects. These defects were phenocopied by mutations in cheerio, the Drosophila homolog of the cytoskeletal protein filamin (Zheng et al., 2011). Filamin and Ten-m also interact physically (Zheng et al., 2011), suggesting that the Teneurins may control how neurons move and match with targets through interaction with the cytoskeleton. The defects in fasciculation caused by altering ten-m levels could be the same as the previously discussed defects with partnermatching after Teneurin perturbation. Furthermore, because Ten-m interacts with filamin, and mutations to filamin cause similar defects, partner-matching may very well be mediated by the reorganization of the cytoskeleton by a Ten-m/filamin complex. This would suggest a second tenet of Teneurin function: mediation of downstream function via interaction with the cytoskeleton. As such, additional research should explore this fascinating possibility that Teneurin-related partner matching requires modulation of the cytoskeleton.

Whether these mechanisms were selective for peripheral synapses or if they could also function in the central nervous system remained an open question. In the Drosophila olfactory system, however, there is a similar requirement for partner matching. Neurons in the Drosophila antennal lobe (Jefferis and Hummel, 2006), the first order processing center for olfactory information, must also match presynaptic axons of olfactory receptor neurons (ORNs) to postsynaptic dendrites of a cognate class of projection neurons (PNs). A genetic screen designed to identify potential partner matching molecules at this synapse (Hong et al., 2012) identified Ten-a and Ten-m as regulators of this process. In the olfactory system, all glomeruli have a basal level of both Ten-m and Ten-a, but some classes of neurons have elevated levels of either protein (Hong et al., 2012; Mosca and Luo, 2014). Specifically, certain matching ORN-PN pairs express elevated levels of the same Teneurin (Ten-a or Ten-m), reminiscent of Ten-m expression at the NMJ. Knocking down expression of both teneurin genes in both ORNs and PNs leads

to mismatching between known partners, a phenotype that was also seen when both ten-m and ten-a are reduced in ORNs or PNs. More specific analysis revealed that the levels of Teneurin expression play a role in partner-matching. Knockdown of tena in PNs that normally have high Ten-a levels caused them to mismatch with ORNs that normally have low amounts of Ten-a. However, knockdown of ten-a in PNs that are naturally Ten-a low does not cause ORN mismatching, suggesting that PNs and ORNs must have similar levels of Ten-a to match correctly (**Figure 1**). A similar logic followed for ORN and PN pairs that expressed high levels of Ten-m (Hong et al., 2012). Altogether, these experiments suggested that partner-matching between PNs and ORNs occurs by a homophilic process in which both cells express either Ten-a or Ten-m at the same elevated level. Similarly to the NMJ, this wiring could be mismatched by overexpressing ten-m or ten-a in a specific ORN or PN (**Figure 1**) that normally only has low levels of that teneurin, suggesting that these elevated levels of matching Teneurins can instruct partner matching (Hong et al., 2012). This again supports the notion of a conserved tenet where elevated Teneurin levels control partner matching between select cognate classes of ORNs and PNs.

A number of open questions regarding the Teneurins and partner matching remain. Though some glomeruli follow a "Teneurin code" for expression and matching, others share overlapping expression and difference of phenotypic severity, suggesting partial redundancy between the Teneurins (Hong et al., 2012). The nature of this redundancy is not yet understood. In addition, little is known about other proteins involved in partner matching, as two Teneurins are not sufficient to pattern the entire antennal lobe. Recent work highlighted roles for Toll-6 and Toll-7 receptors (Ward et al., 2015), along with DIP/Dpr proteins (Barish et al., 2018), but a complete understanding remains elusive. The Teneurins may be part of a broader code involving a balance of additional proteins and their expression levels to determine the final correct partner match. Further still, it is unknown whether there is specificity for other cell types such as local interneurons or alternate connection modes such as dendrodendritic connections between PNs, leaving an active area of study. Despite these unknowns, core tenets remain: Teneurins are required pre- and postsynaptically for partner matching and they do so in a homophilic fashion (Hong et al., 2012; Mosca et al., 2012) via elevated levels. This may occur via modulation of the cytoskeleton (Zheng et al., 2011) at both the NMJ and in the CNS, revealing a fascinating instance of mechanistic conservation. Their widespread expression in both invertebrate and vertebrate systems (Feng et al., 2002; Li et al., 2006; Kenzelmann et al., 2008; Mörck et al., 2010; Dharmaratne et al., 2012; Antinucci et al., 2013; Mosca, 2015; Zhang et al., 2018) and conservation of protein structure and mechanistic function (Jackson et al., 2018) suggests they may represent a general and versatile matching mechanism across synapse types and evolutionary taxa.

#### SYNAPSE ORGANIZATION

Once a neuron has identified the correct synaptic partner, it must undergo synaptogenesis to form a functional and lasting connection. This is a complex process that involves multiple steps by which pre- and postsynaptic proteins align, synaptic machinery assembles, and the cytoskeletal components organize. Screens for synaptic molecules at the Drosophila NMJ have identified Teneurins as potential players in this process (Liebl et al., 2006; Kurusu et al., 2008) though their function in synaptic connectivity had not been elucidated until more recently. Work at the NMJ and in the olfactory system identified a conserved function for the Teneurins in synaptogenesis with a mechanism somewhat distinct from, though resembling, that of partner matching. In partner matching, the Teneurins use a homophilic transsynaptic interaction to partner match cells expressing the same, elevated levels of a particular Teneurin (Hong et al., 2012; Mosca et al., 2012). During synaptogenesis, however, Teneurins interact heterophilically and transsynaptically, with Ten-a being found mainly at the presynapse and Ten-m at the postsynapse (Mosca et al., 2012; Mosca and Luo, 2014). Because of this general role in synaptic organization, distinct from partner matching, all olfactory and neuromuscular synapses show a basal level of Teneurin expression, while only select synapses participating in Teneurin-mediated partner matching show high levels of expression (Hong et al., 2012; Mosca et al., 2012). Still using the same family, the concept of Teneurin::Teneurin interaction is conserved, but the use of homo- versus heterophilic interactions allows for the diversification of processes. Further, the mechanism of Teneurin function regulating the cytoskeleton is also conserved in synaptic organization (Mosca et al., 2012; Mosca and Luo, 2014), an echo of the second tenet of Teneurin function suggested by partner matching and axon guidance studies. Here, we will explore the role of Teneurins in establishing synaptic connectivity in both the peripheral and central nervous systems and the evidence supporting these mechanisms.

The Drosophila NMJ is an ideal setting to study synaptic development in that it combines singular simplicity with powerful molecular genetics (Harris and Littleton, 2015). Studying the role of Teneurins at the NMJ opened a window into understanding their trans-synaptic role in synaptogenesis. There, Ten-a is expressed presynaptically, where it colocalizes with the active zone marker Bruchpilot and the periactive zone marker Fasciclin II (Mosca et al., 2012). Ten-m also shows low levels of presynaptic expression, but is predominantly expressed in the postsynaptic muscle, where it colocalizes with postsynaptic markers Dlg (the Drosophila homolog of PSD-95) and the cytoskeletal protein α-spectrin. Pre- or postsynaptic perturbation of Ten-a and Ten-m (respectively) causes similar disruptions in synaptic structure and function, including a reduced number of synaptic boutons, defects in active zone apposition, general disorganization of synaptic components, impaired electrophysiological function, and defective vesicle cycling, many of which are reflected in severe locomotor impairment (Mosca et al., 2012). Taken together, these defects indicate a role for Teneurins in synaptic development. This suggested an extension of the first tenet of partner matching: a Teneurin::Teneurin interaction, but with a distinction that partner matching requires homophilic Teneurin interaction, and synaptic development requires heterophilic interaction of

presynaptic Ten-a with postsynaptic Ten-m. Interestingly, tissuespecific removal of the presynaptic pool of Ten-m also results in a morphological phenotype, suggesting a presynaptic role (Mosca et al., 2012). Furthermore, postsynaptic Ten-m knockdown did not enhance the ten-a mutant phenotype, potentially suggesting presynaptic redundancy, or an additional postsynaptic receptor for presynaptic Ten-m. Overall, these experiments suggest a transsynaptic, heterophilic interaction between motoneuronexpressed, presynaptic Ten-a and muscle-expressed, postsynaptic Ten-m.

But what is the downstream mechanism for how the Teneurins mediate such effects? In addition to general defects in synaptic organization, the interruption of heterophilic Teneurin interaction at the NMJ also causes profound cytoskeletal disorganization. Teneurin perturbation causes a disruption of organized presynaptic microtubule loops and an increase in unbundled Futsch/MAP-1b staining, suggesting a deranged cytoskeleton (Mosca et al., 2012). Additionally, the loss of Teneurin signaling also causes a near complete loss of the postsynaptic spectrin cytoskeleton. As direct cytoskeletal disruption can serve as a common cause for many of the phenotypes observed following Teneurin perturbation, this lead to the hypothesis that Teneurins organize synapses via a link with the cytoskeleton. Indeed, Ten-m colocalizes with and physically interacts with α-spectrin in a complex (Mosca et al., 2012). As spectrin is a molecular scaffold which interacts with actin to form a network along the inside of the plasma membrane, this suggested that Ten-m may represent the link between the synaptic cytoskeleton and the membrane, further strengthening the hypothesis of direct cytoskeletal interaction with the Teneurins. Additionally, loss of postsynaptic spectrin does induce similar synaptic growth defects (Pielage et al., 2006), which is consistent with this hypothesis. Thus, Teneurins are involved in organizing synapses by way of ordering the cytoskeleton, as mediated through a Ten-m link between the synaptic membrane and α-spectrin (Mosca et al., 2012). This further supports the second tenet of Teneurin function: mediating their role in neuronal development via cytoskeletal modulation (**Figure 2**).

Though playing a critical role in synaptic organization, the Teneurins also cooperate with other cell surface proteins to construct a connection. Neurexin and the Neuroligins are transmembrane proteins that instruct synaptic development; phenotypes associated with their disruption include changes in bouton number and disorganization of active zones (Li et al., 2007; Banovic et al., 2010; Sun et al., 2011; Owald et al., 2012; Xing et al., 2014; Zhang et al., 2017). These results have considerable phenotypic overlap with perturbation of ten-a and ten-m, suggesting potential genetic or pathway interaction. The two instead operate in distinct but partially overlapping pathways: Neurexin/Neuroligin 1 largely control active zone

synaptic organization. At the NMJ, presynaptic Ten-a is involved in the formation of stable microtubule loops, promoting synaptic organization. Ten-m in the muscle interacts directly with spectrin, as well as potentially with αPS2 to mediate synaptic organization. In the CNS, however, presynaptic Ten-a functions with spectrin to promote synaptic organization. The role of postsynaptic Ten-m is unknown but may regulate downstream cytoskeletal components.

apposition with minor effects on the cytoskeleton while the Teneurins largely control cytoskeletal organization and cooperate with Neurexin/Neuroligin1 to regulate active zone apposition. This reveals that there is a complex cooperation between cell surface proteins and a division of labor to ensure that synaptic contacts are properly organized.

Teneurins also show remarkable similarities in how they function in the central nervous system, as evidenced by examination of transsynaptic Teneurin signaling in the Drosophila olfactory system (Mosca and Luo, 2014). The olfactory system is valuable for studying synaptic development due to its well defined synaptic connections in the context of a complex circuit (Jefferis and Hummel, 2006). At ORN synapses, perturbations in Teneurin levels (either presynaptic Ten-a in the ORNs or postsynaptic Ten-m in the PNs) also impaired synaptic organization. The number of both presynaptic active zones and postsynaptic acetylcholine receptors are decreased when presynaptic ten-a or postsynaptic ten-m are knocked down. This is a strikingly similar phenotypic result as to the NMJ, suggesting conservation between the CNS and the PNS. The mechanism is also strikingly similar, as Ten-a and Ten-m interact heterophilically, and are found primarily at the preand postsynapse, respectively. Furthermore, the link with spectrin is also conserved between the two systems: Ten-a and spectrin function in the same genetic pathway to control central synapse number, the spectrin cytoskeleton in the antennal lobe is drastically reduced following Teneurin perturbation, and presynaptic spectrin is also important for achieving normal synapse number in ORNs. As in the peripheral nervous system, Teneurins function with the cytoskeleton to allow proper cytoskeletal organization for the formation of a robust synaptic architecture (Mosca and Luo, 2014). Thus, the second tenet of Teneurin function, downstream regulation of the cytoskeleton, is further conserved. Finally, a third conserved aspect links Teneurins with synaptic function. At the NMJ, ten-a mutants show reduced evoked postsynaptic potentials, impaired vesicle cycling, and reduced larval locomotion (Mosca et al., 2012). Restoring Ten-a expression to motoneurons partially rescues the locomotor phenotype, suggesting a presynaptic function for Ten-a in regulating function. In the CNS, olfactory function can be measured by behavioral response: basic function can be assayed by the performance of flies in a modified olfactory trap (Larsson et al., 2004; Potter et al., 2010; Min et al., 2013; Mosca et al., 2017) using apple cider vinegar (ACV) as an attractive odorant source. Control flies are nearly uniformly attracted to ACV (**Figure 3**) – impaired attraction can be indicative of

synaptic defects, as seen when the synaptic organizer LRP4 is removed specifically from ORNs (Mosca et al., 2017). ten-a mutants have significantly impaired ACV attraction (**Figure 3**), suggesting that ten-a is required for normal olfactory function (as it is required for normal neuromuscular function). Studies in the whole-animal mutant, however, do not determine where Ten-a functions to regulate function. To address this, we restored ten-a expression only to adult ORNs and found that this partially rescued the loss of olfactory attraction (**Figure 3**). This suggests that presynaptic Ten-a mediates normal function at olfactory synapses, again similar to the NMJ. Thus, the conservation of a role for Teneurins in promoting normal presynaptic function represents a third tenet of Teneurin mechanisms that span the olfactory and neuromuscular systems. There are, however, some variations on the organizational theme between the CNS and the PNS. Interestingly, the spectrin interaction seems to occur presynaptically in the CNS but postsynaptically at the NMJ. Also, a mild phenotype is present when Ten-m is knocked down at the presynaptic NMJ, indicating a minor presynaptic role, but no such phenotype is observed in the CNS (Mosca et al., 2012; Mosca and Luo, 2014). This indicates perhaps that though the broad mechanisms may be conserved, certain elements differ, perhaps owing to the differing complexity and biological role for each synapse. This offers an interesting way to diversify a conserved mechanism – with mild adjustments to allow for different kinds of synapses. Teneurins may also promote postsynaptic cytoskeletal organization in the CNS, but that interaction has yet to be identified (**Figure 2**).

Beyond the spectrin cytoskeleton, additional work has suggested broader conservation. Teneurins also regulate the cytoskeletal proteins adducin and Wsp (Mosca et al., 2012) and can further interact with integrins via αPS2, a synaptic integrin receptor (Graner et al., 1998). At the Drosophila NMJ, knockout of alpha-N-acetylgalactosaminyltransferases (PGANTs), proteins which regulate integrins, led to decreased levels of αPS2 as well as Ten-m (Dani et al., 2014). Ten-m at the presynapse may be involved in cell adhesion through interaction with αPS2, causing the mild phenotype observed when Ten-m is knocked down only in neurons. Future work on the roles of Teneurins will further determine their effectors and how these factors serve to instruct synaptic connectivity via regulation of the cytoskeleton.

Much like partner matching, the Drosophila Teneurins play a critical role in synaptic organization. Further, their function is conserved in the CNS and PNS and also, in a mechanistic fashion by (1) a transsynaptic interaction and (2) a regulation of the downstream cytoskeleton. However, certain distinctions make the organizational process unique from partner matching. Here, basal levels of Teneurins mediate synaptic organization through a heterophilic transsynaptic interaction: Ten-a is predominantly presynaptic while Ten-m is postsynaptic. Further, the Teneurins are relatively unique among synaptic organizers in that their main role is to mediate cytoskeletal components. The remarkable evolutionary conservation present within these systems indicates the importance of Teneurins in their various roles. Further work is needed to examine the specific functions of Teneurins in regulating synaptic connectivity, but the widely conserved mechanisms already observed in Teneurin function promise the advantage of continued study across systems and synapses (Mosca, 2015).

#### FUTURE DIRECTIONS

The complex series of events that underlie neuronal development have distinct molecular, temporal, and spatial requirements that safeguard their fidelity. To ensure evolutionary economy, these events can be coordinated through reuse of molecular cues and mechanisms. These mechanisms are conserved from peripheral to central synapses in Drosophila; work has also shown similar roles in mammalian nervous systems for wiring and synapse organization (Leamey et al., 2007; Dharmaratne et al., 2012; Mosca, 2015; Berns et al., 2018), suggesting mechanistic conservation across multiple species as well. As such, the Teneurins are an evolutionary constant, working at multiple levels to ensure nervous system development. In our current understanding, however, there is much left to learn about how Teneurins regulate nervous system development. Recent work especially has highlighted the interplay between Teneurins and Latrophilin in mammalian synapse organization (Silva et al., 2011; Boucard et al., 2014; Vysokov et al., 2016; Li et al., 2018) and in behavioral regulation through TCAP, the Teneurin C-terminal Associated Peptide (Woelfle et al., 2015, 2016). In Drosophila, potential interactions between the Teneurins and Latrophilin have not been studied. The Drosophila genome possesses a single Latrophilin homolog, dCirl (Scholz et al., 2015). dCirl is expressed in larval chordotonal neurons and is required for mechanosensation and larval locomotion (Scholz et al., 2015). In these neurons, dCirl functions to reduce cAMP levels in response to mechanical stimulation (Scholz et al., 2017). Whether these functions involve Teneurins remains an open question. There is likely not complete overlap between Teneurins and dCIRL, as dCirl mutants do not phenocopy the synaptic defects associated with ten-a/ten-m perturbation (T. Mosca, unpublished observations). This does not, however, address potential redundancy in the genome with other GPCRs or orphan receptors, so more directed study is needed. As Teneurins are also thought to interact with other cell surface receptors (Mosca et al., 2012) and adhesion molecules (Dani et al., 2014), it is increasingly likely that Teneurins represent a nexus for receptor interaction, suggesting that a number of players are yet to be discovered.

One key unanswered question involves the role of presynaptic Ten-m at the NMJ. Though predominantly postsynaptic at the NMJ (**Figure 4A**), Ten-m also localizes presynaptically in motoneurons (Mosca et al., 2012); this contribution is revealed when Ten-m is removed specifically from the muscle using RNAi (**Figure 4B**). Presynaptic knockdown of Ten-m results in a modest reduction in synaptic bouton number (Mosca et al., 2012). However, as Teneurins and Integrins all promote synaptic maturation (Mosca et al., 2012; Lee et al., 2017), and Ten-m may interact with integrins (Dani et al., 2014), this raises the possibility that ten-m may contribute to activity-dependent synaptic remodeling (Ataman et al., 2008; Piccioli and Littleton, 2014; Xiao et al., 2017). At

the Drosophila NMJ, acute spaced stimulation using high K <sup>+</sup> induces activity-dependent sprouting in the form of "synaptopods" (Ataman et al., 2008). These synaptopods form in as little as 15–20 min and contain Ten-m (**Figure 4C**). This suggests that Ten-m is one of the first components of these nascent neurite branches. Ten-m is present even before synaptic vesicles appear, which are among the earliest components visible in ghost boutons (Ataman et al., 2008). This raises the possibility that Ten-m could promote synaptic maturation and activity-dependent growth. Further experiments will be needed to directly test this hypothesis but could more deeply connect presynaptic Ten-m, neuromuscular growth, and integrins.

Further, our knowledge of downstream Teneurin effectors remains incomplete. At the Drosophila NMJ, neither reduced Neurexin/Neuroligin signaling (Li et al., 2007; Banovic et al., 2010) nor a loss of spectrin (Pielage et al., 2005, 2006) can account for the entire cadre of synaptic phenotypes associated with teneurin perturbation (Mosca et al., 2012). This suggests that additional downstream mechanisms exist to mediate Teneurin function. This could be through additional cytoskeletal proteins, as in C. elegans (Mörck et al., 2010). A more thorough understanding of how Teneurins engage partner matching, either by downstream mechanisms or interaction with other cell surface proteins is also poorly understood. Whether Teneurins interact with axon guidance molecules and cell surface receptors, as in C. elegans (Mörck et al., 2010) is a distinct possibility. Approaches to understand Teneurin-interacting proteins will be essential to understand the different ways they regulate their diverse functions.

Finally, a core question intrinsic to the Teneurins remains. As we understand, Teneurins use multiple interactive mechanisms to enable development: homophilic interactions match and maintain partners while heterophilic interactions organize synaptic connections. This must mean that, at the same connection, both homophilic and heterophilic interactions exist simultaneously. As these distinct pairs have distinct goals, how does a cell interpret which interaction is happening for a particular Teneurin molecule? For example, ORNs that use elevated Ten-a to match their cognate PNs also use basal levels of presynaptic Ten-a to organize their output synapses by interacting with postsynaptic Ten-m. Therefore, these ORNs simultaneously have homophilic and heterophilic interacting Ten-a molecules. How are these distinguished? Are certain downstream interactors only expressed at certain developmental times? This way, the downstream effectors specific to partner matching would only appear during times of neuronal wiring and be downregulated by the time synaptic formation, organization, and maintenance take over as the predominant processes. As partner matching and synapse formation can be separated by as much as 24–48 h in the developing olfactory system (Jefferis and Hummel, 2006) or by as much as 4–6 h at the developing NMJ (Broadie and Bate, 1993), this is a reasonable hypothesis. However, if this is not the case, it could be that the mechanism is more intrinsic to the Teneurin protein. If there was a fundamental difference between a Ten-a::Ten-a and a Tena::Ten-m interaction, this could result in conformational changes that only allowed binding of specific downstream molecules. One hypothesis is that this fundamental difference could come from tension (Mosca, 2015). The NHL domain present in the extracellular domain of Teneurins is thought to mediate interaction in trans (Beckmann et al., 2013). Homophilic NHL domain interactions display stronger adhesive forces than heterophilic (Beckmann et al., 2013): if this tension can be "read out" by the cell, it could recruit different downstream effector molecules depending on the transsynaptic partner of that Teneurin. This could enable a mechanism to distinguish homophilic from heterophilic Teneurin interactions when both may exist in the same small synaptic region. With more recent structural information about the Teneurins

(Li et al., 2018), more directed hypotheses about interaction can now be explored. Beyond an intrinsic tension mechanism, more recent work showed that splice variants of Ten-3 in mouse can regulate cell-to-cell adhesion, potentially affecting neuronal wiring (Berns et al., 2018). Thus, there are multiple options for intrinsic ways that Teneurins could distinguish themselves depending on partners and interactions. Future work will be needed to dissect both the intrinsic and extrinsic mechanisms that enable Teneurins to function so broadly.

Work over the last decade has cemented the Teneurins as essential regulators of neuronal development, functioning via related mechanisms in steps ranging from the initial elements of neurodevelopment in axon guidance to seeing the developmental process through to the end with functions in synaptic organization. Science will take the next bold steps forward from that foundation, venturing out to determine how these core cell surface proteins mediate downstream function, and moving closer to understanding the intricacies and complexities of neuronal development.

#### REFERENCES


### AUTHOR CONTRIBUTIONS

All authors contributed to the writing, research, and revision of this article.

# FUNDING

This work was supported by awards to TM from the National Institutes of Health (R00-DC013059), the Commonwealth of Pennsylvania CURE Fund (SAP 4100077067), and the Alfred P. Sloan Foundation.

#### ACKNOWLEDGMENTS

We would like to thank members of the Mosca Lab for stimulating discussions and camaraderie.


an ancient fold for cell-cell interaction. Nat. Commun. 9:1079. doi: 10.1038/ s41467-018-03460-0


during development. Dev. Biol. 290, 237–245. doi: 10.1016/j.ydbio.2005. 11.038


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 DePew, Aimino and Mosca. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Expression and Roles of Teneurins in Zebrafish

Angela Cheung1,2† , Katherine E. Trevers1,2†‡, Marta Reyes-Corral<sup>1</sup> , Paride Antinucci<sup>1</sup>‡ and Robert Hindges1,2 \*

<sup>1</sup> Centre for Developmental Neurobiology, King's College London, London, United Kingdom, <sup>2</sup> MRC Centre for Neurodevelopmental Disorders, King's College London, London, United Kingdom

#### Edited by:

Antony Jr. Boucard, Centro de Investigación y de Estudios Avanzados (CINVESTAV), Mexico

#### Reviewed by:

Timothy Mosca, Thomas Jefferson University, United States Elena Seiradake, University of Oxford, United Kingdom

> \*Correspondence: Robert Hindges robert.hindges@kcl.ac.uk

†These authors have contributed equally to this work

#### ‡Present address:

Katherine E. Trevers, Department of Cell and Developmental Biology, University College London, London, United Kingdom Paride Antinucci, Department of Neuroscience, Physiology and Pharmacology, University College London, London, United Kingdom

#### Specialty section:

This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Neuroscience

Received: 24 October 2018 Accepted: 12 February 2019 Published: 12 March 2019

#### Citation:

Cheung A, Trevers KE, Reyes-Corral M, Antinucci P and Hindges R (2019) Expression and Roles of Teneurins in Zebrafish. Front. Neurosci. 13:158. doi: 10.3389/fnins.2019.00158 The teneurins, also known as Ten-m/Odz, are highly conserved type II transmembrane glycoproteins widely expressed throughout the nervous system. Functioning as dimers, these large cell-surface adhesion proteins play a key role in regulating neurodevelopmental processes such as axon targeting, synaptogenesis and neuronal wiring. Synaptic specificity is driven by molecular interactions, which can occur either in a trans-homophilic manner between teneurins or through a trans-heterophilic interaction across the synaptic cleft between teneurins and other cell-adhesion molecules, such as latrophilins. The significance of teneurins interactions during development is reflected in the widespread expression pattern of the four existing paralogs across interconnected regions of the nervous system, which we demonstrate here via in situ hybridization and the generation of transgenic BAC reporter lines in zebrafish. Focusing on the visual system, we will also highlight the recent developments that have been made in furthering our understanding of teneurin interactions and their functionality, including the instructive role of teneurin-3 in specifying the functional wiring of distinct amacrine and retinal ganglion cells in the vertebrate visual system underlying a particular functionality. Based on the distinct expression pattern of all teneurins in different retinal cells, it is conceivable that the combination of different teneurins is crucial for the generation of discrete visual circuits. Finally, mutations in all four human teneurin genes have been linked to several types of neurodevelopmental disorders. The opportunity therefore arises that findings about the roles of zebrafish teneurins or their orthologs in other species shed light on the molecular mechanisms in the etiology of such human disorders.

Keywords: teneurin/Odz, retinal ganglion cell, amacrine cell, visual system, synapse adhesion molecule, zebrafish

#### INTRODUCTION

As one of the most complex systems in nature, the functionality of the nervous system is highly dependent on the formation of precise synaptic connections between neurons during development. While progress is still being made in furthering our understanding of these mechanisms, it is becoming increasingly evident that synaptic specificity is a finely attuned process involving a plethora of cell adhesion molecules that act in a combinatorial manner to generate diverse cellular interactions. The teneurins, also known as Ten-m/Odz, are one family of such cell adhesion molecules that has been implicated, among others, in regulating the specificity of synaptic connections.

A phylogenetically conserved family of type II transmembrane glycoproteins first discovered in the early 1990s in Drosophila, the teneurins have been shown to be involved in intercellular signaling during development (Tucker and Chiquet-Ehrismann, 2006; Tucker et al., 2012). Their key role in mediating basic neurodevelopmental processes such as axon guidance and synaptic

partner matching (Drabikowski et al., 2005; Hong et al., 2012) is reflected in the high expression of teneurins in the central nervous system. Across species, C. elegans has a single teneurin (Ten-1), Drosophila have two (Ten-m and Ten-a) and all vertebrates have four paralogs (tenm1–4). Sequence similarity is high between paralogs, with human teneurin paralogs sharing 58–70% sequence identity (Jackson et al., 2018).

The teneurins themselves are large proteins of around 300 kDa with a smaller N-terminal intracellular domain, a single span transmembrane domain and large C-terminal extracellular region (Rubin et al., 1999; Tucker and Chiquet-Ehrismann, 2006; **Figure 1**). While the intracellular domain is able to interact with the cytoskeleton, as has been shown with tenm1 (Nunes et al., 2005), the highly conserved 200 kDa extracellular domain of teneurin, which contains eight epidermal growth factor (EGF)-like repeats and five NHL (NCL-1, HT2A, and Lin-41) repeats, dimerizes in cis to mediate both homo- and heterophilic interactions (Beckmann et al., 2013). In addition, the teneurins can interact trans-synaptically with other cell adhesion molecules such as latrophilins via their NHL domains (Boucard et al., 2014). Tenm2, for example, interacts across the synaptic cleft with presynaptic latrophilin1 to mediate calcium signaling and synapse formation (Vysokov et al., 2016). Recent X-ray crystallography and cryo-EM imaging

FIGURE 1 | Schematic illustration of a teneurin dimer. Teneurins possess a single transmembrane domain and small intracellular domain compared to the relatively large extracellular domain. Depicted is a cis-dimer with identified protein domains indicated in different colors. The extracellular domain consists of the EGF domain, cysteine-rich domain, TTR (transthyretin-related) domain, FN (fibronectin)-plug domain, NHL domain, YD (tyrosine-aspartate)-repeat domain, internal linker domain, ABD (antibiotic-binding domain-like) domain and the Tox-GHH domain. The EGF domains play a key role in regulating cis-interactions between teneurin molecules while the NHL- and YD-repeat domains mediate trans-interactions. Domains are only representative and not to relative scale.

data has demonstrated that the ectodomains of tenm2 and 3 exist in a large β-barrel conformation consisting of an eight sub-domain super-fold structured on a spiraling YD-repeat shell domain (Jackson et al., 2018; Li et al., 2018). Further elucidation of the structural interactions between teneurins and other proteins would provide insight into functionality and factors driving the molecular diversity underpinning synaptic connectivity or even highlight possible new interaction partners.

Indeed, the importance of teneurins in regulating synaptic partner matching and functional connectivity is well demonstrated in the vertebrate visual system where retinal ganglion cells (RGCs) form specific connections with their synaptic partners. In zebrafish, tenm3 is required by RGCs and amacrine cells for acquiring correct structural and functional connectivity in vivo, with tenm3 knockdown or knockout leading to defects in RGC and amacrine cell dendrite stratification in the retina and the disrupted development of orientation selectivity (Antinucci et al., 2013, 2016). Tenm3 has also recently been shown to regulate topographic circuit assembly in the hippocampus of mice (Berns et al., 2018). More broadly, the teneurins are strongly implicated in the establishment of visual mapping in mice (Leamey and Sawatari, 2014). Anterograde tracing of RGC axons in tenm2 and tenm3 knockout mice showed notable aberrant changes in the mapping of ipsilateral projections from the retina to the dorsal lateral geniculate nucleus (dLGN) and the superior colliculus (Leamey et al., 2007; Young et al., 2013).

Although the fundamental role of teneurins in establishing synaptic connectivity is becoming increasingly apparent, data on the spatio-temporal expression pattern of teneurins across the central nervous system, and its functional and physiological significance, is largely lacking. Through in situ hybridization and the generation of transgenic bacterial artificial chromosome (BAC) reporter lines in zebrafish, we present the expression patterns of teneurin 1–4 across the central nervous system and identify some of the teneurin-positive cell types, focusing particularly on the visual system.

### MATERIALS AND METHODS

### Zebrafish Husbandry

Zebrafish (Danio rerio) adults and embryos were maintained in accordance with the Animals (Scientific Procedures) Act 1986 under license from the United Kingdom Home Office (PPL70/9036).

Pairwise zebrafish spawnings were set up using Ekkwill wild type adults (Ekkwill Breeders, Florida). Larvae were maintained between 0 and 5 days post fertilization (dpf) at 28.5◦C on a 14 h ON/10 h OFF light cycle in 1 × Danieau solution (58 mM NaCl, 0.7 mM KCl, 0.6 mM Ca(NO3)2, 0.4 mM MgSO4, 5 mM HEPES) supplemented with PTU (Sigma) at a final concentration of 200 µM. Where necessary, chorions were removed using forceps. Larvae were anesthetized with MS222 and fixed overnight with 4% paraformaldehyde (PFA) in PBS at 4◦C, then dehydrated in 100% methanol and stored at −20◦C.

#### RNA Probe Synthesis

fnins-13-00158 March 11, 2019 Time: 17:18 # 3

cDNAs spanning several exons of each tenm were blunt cloned into the pSC-B-Amp plasmid using the following primers: tenm1 exons 19–22 Fwd: gatctcagcaggaatgtggagg, Rev: cagcatcccggcgttactgatg; tenm2 exons 27–29 Fwd: gcatgtgttcaaccactcca, Rev: gcgccatttaacaccagaac; tenm3 exons 26– 28 Fwd: gggactatgacattcaagcaggtc, Rev: cattgttggcactgtcggccag; tenm4 exons 21–25 Fwd: catttcccagcagcctccagtc, Rev: ctttcgcgtagccgtcgtcg. Plasmids were linearized and transcribed using NotI and T3 (for tenm1, tenm3, tenm4) or EcoRV and T7 (for tenm2) to generate DIG-labeled anti-sense riboprobes. Probes were purified by LiCl precipitation and suspended in ultrapure water. Working probes were diluted to 250 ng/ml in hybridization mix (HM) containing 50% formamide, 5 × SSC pH 6.0, 0.1% Tween-20, 50 µg/ml Heparin and 500 µg/ml tRNA.

#### Wholemount in situ Hybridization

In situ hybridization was performed according to Thisse and Thisse (2008). In brief, samples were rehydrated through a series of 75, 50, and 25% methanol, into PBS containing 0.1% Tween-20 (PBTw) and permeabilized by digestion with 10 µg/ml Proteinase K in PBTw at room temperature. Digestion times varied according to developmental stage (10 min for 1 dpf, 20 min for 2 dpf and 30 min for 3–5 dpf). Digestion was stopped by post-fixing the embryos in 4% PFA in PBS for 20 min at room temperature. Residual fix was removed by washing 2 × 10 min in PBTw before embryos were prehybridized in HM at 70◦C for 3 h. Probes were hybridized overnight at 70◦C and the embryos were subsequently washed 1 × 10 min through 75, 50, and 25% HM, into 2 × SSC followed by 2 × 30 min washes with 0.2 × SSC, all at 70◦C. Then, at room temperature, the larvae were washed through a series of 75, 50, and 25% 0.2 × SSC, into PBTw before non-specific binding was blocked for 3 h at room temperature using 2 mg/ml BSA and 2% sheep serum in PBTw. Samples were incubated overnight at 4◦C in blocking buffer containing anti-DIG antibody (Roche) diluted 1:10,000. Excess antibody was removed by washing 4 × 1 h in PBTw. Finally, samples were washed 2 × 15 min in staining buffer containing 10 mM Tris–HCl pH 9.5, 100 mM NaCl and 0.1% Tween-20 and color staining was developed in the dark at room temperature using NBT/BCIP stock solution (Roche) diluted in staining buffer. The color reaction was checked frequently and stopped by washing in PBTw. Images were captured using a Leica M165 FC stereomicroscope, QImaging Retiga camera and Volocity software.

#### Paraffin Sections

Larvae were post-fixed overnight in 4% PFA at room temperature and dehydrated for 30 min in 100% methanol and 30 min in isopropanol. Samples were cleared with tetrahydronaphthalene (THN) for 30 min. Molten Paraplast wax (Sigma) was added to give a 1:1 mix of THN:wax and incubated at 60◦C for 30 min. This was replaced 3 × 1 h with fresh molten wax at 60◦C before samples were oriented, embedded and left to solidify. 12 µm transverse sections were cut on a Microm HM315 microtome and mounted on slides treated with glycerin albumin. Slides were dried overnight at 37◦C and dewaxed by 2 × 4 h washes in Histoclear II. Coverslips were mounted using a 1:3 mixture of Histoclear II: Canada Balsam. Images were captured using an Olympus Vanox-T with 40 objectives and a QImaging Retiga 2000R camera with QCapture Pro software.

#### BAC Transgenesis

The BAC clones containing the gene sequences for tenm 1, 2.14 and 4 were identified by using the ENSEMBL database of the Wellcome Trust Sanger Institute (Daniokey library reference: tenm1, DKEY-275B22; tenm2.14, DKEY-47H20; tenm4, DKEY-84I3). Zebrafish have two genes that encode for tenm2, tenm2.14 and tenm2.21, present on chromosomes 14 and 21, respectively (**Supplementary Figure S1**). The tenm2.14 paralog was chosen for further investigation, based on its higher resemblance to other tenm2 orthologs. BAC clones were transformed first with a pRed-Flp4 plasmid before a recombineering step was used to insert a membrane-localized reporter gene (mCitrine) together with a kanamycin resistance cassette at the beginning of the tenm gene. Transient transgenic zebrafish lines were created by microinjection of teneurin:mCitrine BAC constructs into onecell stage embryos, before embryos were allowed to develop at 28.5◦C with PTU until 3–5 dpf. Embryos were mounted laterally on glass slides using 1% low-melting point agarose in Danieau's solution and imaged on an LSM 710 Zeiss confocal equipped with a spectral detection scan head and a 20×/1.0 NA waterimmersion objective.

# RESULTS

#### Tenm1–4 Are Widely Expressed Across the Central Nervous System

The expression patterns of the four vertebrate paralogs of teneurin were investigated via wholemount in situ hybridization on zebrafish embryos over 1 to 5 dpf. Tenm1 is first detected in a discrete spot in the ventral midbrain at 1 dpf (**Figure 2A**; arrow). Later expression is more widespread but localized anteriorly with tenm1 observed in multiple regions of the forebrain, midbrain and hindbrain at 2–5 dpf (**Figure 2**). Similarly, significant tenm2 expression is first detected anteriorly at 2 dpf in the olfactory bulbs and in multiple clusters of neurons in the midbrain and hindbrain, persisting over 3–5 dpf (**Figure 3**).

In contrast to tenm1 and 2, which were either weak or absent at 1 dpf, strong tenm3 expression is already detected early at 1 dpf in the forebrain and midbrain, as well as the developing retina (**Figures 4A,F,K**). By 2 dpf, high expression is specifically localized to the optic tectum, ventral retina, medulla oblongata and tips of the fin buds, persisting over 3–5 dpf (**Figure 4**). Similarly, strong tenm4 expression is also detected early at 1 dpf in the forming retina, midbrain and hindbrain, and becomes localized to the inner layers of the retina, olfactory bulbs, optic tectum, subset of hindbrain neurons, and along the midhindbrain boundary at 2–3 dpf. This expression persists over 4–5 dpf (**Figure 5**).

# The Teneurins Are Expressed Across Interconnected Regions of the Zebrafish Nervous System During Development

Transverse sections across the zebrafish central nervous system were collected from wholemount in situ hybridization samples in order to gain more detailed insight into the expression pattern of teneurins from 1 to 4 dpf.

At 1 dpf, tenm1 is only expressed in a discrete population of cells in the ventral midbrain (**Figure 6AB**; arrow), while tenm2 expression is absent. In contrast, tenm3 and tenm4 are expressed strongly in broad domains of the forming retina and midbrain

(**Figures 6CA, DA, CB, DB**). Tenm4 is expressed throughout the rostral and caudal hindbrain (**Figures 6DC, DD**), whereas tenm3 is restricted laterally (**Figures 6CC, CD**). While tenm4 is strongly expressed in the neural tube (**Figure 6DE**), tenm3 expression is much weaker (**Figure 6CE**).

By 2 dpf, tenm3 and tenm4 are expressed strongly in the forebrain pallium (**Figures 7CA, DA**), while tenm1 transcripts are only faintly detected (**Figure 7AA**). In the retina, tenm3 is restricted to retinal progenitors in the inner ventral domain (**Figure 7CB**), while tenm4 is expressed more broadly across

FIGURE 4 | Wholemount expression of tenm3 during zebrafish development. Tenm3 expression during 1–5 dpf from lateral (A–E), dorsal (F–J) and frontal (K–O) perspectives. A, Anterior; D, Dorsal; F, Fin; FB, Forebrain; HB, Hindbrain; MB, Midbrain; MO, Medulla Oblongata; OB, Olfactory Bulb; OT, Optic Tectum; P, Posterior; Re, Retina; V, Ventral; vRe, ventral Retina. Scale bar in all panels = 250 µm.

the retina and at the ciliary marginal zone (**Figure 7DB**). Neither tenm1 nor tenm2 are detected in the retina at these stages (**Figures 7AB, BB**). Tectal cells express tenm4 broadly (**Figure 7DC**), whereas tenm3 is restricted dorsally (**Figure 7CC**). A weak expression of tenm1 can also be observed in the lateral tectum (**Figure 7AC**), and overlapping tenm2, 3 and 4 expression (**Figures 7BD, CD, DD**) is detected in the cerebellum of the rostral hindbrain, but only tenm3 and tenm4 are observed at the rhombic lip and medulla oblongata of the caudal hindbrain (**Figures 7CE, DE**). No teneurin expression is detected in the spinal cord at 2 dpf.

Tenm1 and tenm2 are not detected in the forebrain, retina or midbrain at 3 dpf. Tenm3 and tenm4 can be detected in the pallium and subpallium, but the latter only weakly (**Figures 8CA, DA**). Tenm3 and tenm4 are also detected in amacrine cells, retinal ganglion cells (**Figures 8CB, DB**) and in tectal neurons (**Figures 8CC, DC**). All teneurins are expressed in the rostral and caudal hindbrain (**Figures 8AD, AE, BD, BE, CD, CE, DD, DE**) but in different layers of the cerebellum and medulla oblongata, with tenm1 and tenm2 expression weak in this area. Tenm3 and tenm4 are also present at the rhombic lip, while again, no teneurin expression is observed along the spinal cord.

Finally, at 4 dpf, tenm3 and tenm4 expression persists in the pallium (**Figures 9CA, DA**), while in the retina, tenm3 mRNA is no longer detected in amacrine cells and remains only in the ventral population of RGCs (**Figure 9CB**). Tenm4 expression in these cells is also reduced compared to 3 dpf (**Figure 9DB**), while both tenm3 and tenm4 are present in tectal cells (**Figures 9CC, DC**), as well as in the cerebellum, rhombic lip and medulla oblongata (**Figures 9CD, CE, DD, DE**). Very weak tenm1 expression overlaps in the cerebellum (**Figure 9AD**). Tenm2 expression is too weak to detect in sections at 4 dpf.

Although teneurin 1–4 were found to be widely expressed in overlapping regions of the nervous system, we cannot infer from our studies whether different teneurin paralogs are co-expressed within individual cells.

#### The Teneurins Are Expressed in Multiple Cell Types of the Retina

To further investigate the types of retinal cells that are positive for individual teneurins, we generated BAC constructs for tenm1, 2.14 and 4, inserting the coding sequence for the fluorescent reporter mCitrine at the place of the ATG in the genomic locus (**Supplementary Figure S1**).

We have previously described the expression of tenm3 in the zebrafish visual system in detail, including the identification

of tenm3-positive retinal cells using a BAC transgenic strategy (Antinucci et al., 2013, 2016). Here we extend these analyses to describe retinal and tectal cells in zebrafish positive for other members of the teneurin family. Transient transgenic zebrafish embryos expressing teneurin:mCitrine BAC constructs were imaged at 3–5 dpf to investigate the morphology and distribution of tenm1-, tenm2.14- and tenm4-positive cells in the developing visual system, where teneurins have been shown to play a key role in establishing functional neural circuitry. All three BAC constructs tested labeled single cells of the visual system, including RGCs, amacrine cells and tectal cells (**Figure 10A**).

The tenm1:mCitrine BAC construct consistently labeled RGCs (**Figures 10B–F**) and a low number of amacrine cells in the retina, as well as cells in the optic tectum. All tenm1-positive tectal cells shared a characteristic elongated morphology with the axon projecting from the dendritic arbor instead of from the cell soma (**Figures 10G,H**). Citrine-expressing RGCs in the retina were classified morphologically into three main groups according to their arborization patterns (**Figure 10I**); monostratified, bistratified or diffuse. Monostratified RGCs were further subdivided into ON, OFF and asymmetric dendritic arbors according to the distance of the dendritic field to the cell body. Although all RGC types had similar dendritic field diameters, the monostratified asymmetric and diffuse asymmetric RGCs had the largest dendritic field coverage/area (**Table 1**).

In direct contrast to tenm1:mCitrine, the tenm2.14:mCitrine BAC construct labeled predominantly amacrine cells (**Figures 11A–J**) along with a very low number of RGCs. Tectal cells were also labeled (**Figure 11K**), and similar to when the tenm1:mCitrine construct was used, the number of labeled cells was higher in the optic tectum than in the retina. For Tenm2.14:mCitrine, the fluorescently labeled cells could be classified into narrow-field, medium-field and wide-field amacrine cells according to their dendritic arborization and morphology, with further subdivision in accordance to their stratification within the inner plexiform layer (IPL); ON or OFF (**Figure 11L**). A total number of 9 amacrine cell types were defined: 6 narrow-field (2 monostratified and 4 diffuse), 2 medium-field (1 monostratified and 1 diffuse) and one wide-field (monostratified), with the narrow-field amacrine cells being the most abundant. The average dendritic field for narrow-field amacrine cells was 4 times and 14 times smaller than the average

Scale bar = 250 µm in A–D; 30 µm in all other panels.

dendritic field of medium-field amacrine cells and wide-field amacrine cells, respectively (**Table 2**).

While the tenm1:mCitrine and tenm2.14:mCitrine BAC constructs strongly labeled a range of distinct cell types, the tenm4:mCitrine BAC only lead to weak reporter expression in cells and further detailed morphological analyses could not be carried out.

#### DISCUSSION

We describe here the expression pattern of different members of the teneurin family across the central nervous system in zebrafish during development, focusing particularly on the visual system where they have been shown to play a pivotal role in establishing connectivity, via in situ hybridization and BAC transgenesis.

Although the teneurins are also expressed in non-neuronal tissues, such as at sites of cell migration and at muscle attachment points, they are primarily concentrated in the central nervous system and at sites of pattern formation during development (Baumgartner and Chiquet-Ehrismann, 1993; Baumgartner et al., 1994; Fascetti and Baumgartner, 2002; Drabikowski et al., 2005). This localization is observed across all teneurin expressing species and supports the role of teneurins in neurodevelopmental processes such as neurite outgrowth, axon guidance and synapse formation (Tucker and Chiquet-Ehrismann, 2006; Leamey and Sawatari, 2014). Consistent with that, we observed an early widespread expression of teneurins across the central nervous system with tenm1, 3 and 4 expressions already present in embryos at 1 dpf. Indeed, the early strong expression of tenm3 and 4 in the developing retina supports the role of teneurins, particularly of tenm3, in directing the functional connectivity of RGCs and amacrine cells in the developing visual system (Antinucci et al., 2013). Both genes remain strongly expressed in various structures of the developing central nervous system well into 4 dpf. Tenm1 and 2 expression, on the other hand, is not observed at all in the retina during development and overall expression is weak, especially by 4 dpf.

All four teneurins are also expressed at various levels in different layers of the cerebellum at 3 dpf. Interestingly,

Optic Tectum; Pa, Pallium; P, Posterior; Re, Retina; RGC, Retinal Ganglion Cells; RL, Rhombic Lip; V, Ventral. Scale bar = 250 µm in A–D; 30 µm in all other panels.

mutations in tenm3 have been implicated as a plausible candidate for a new dominantly inherited form of pure adult-onset cerebellar ataxia in humans (Storey et al., 2009). Essentially, all four teneurin paralogs are expressed in an overlapping manner across interconnected areas of the central nervous system of the zebrafish embryo, such as the retina and the optic tectum. The strong co-expression of tenm3 and 4 across many areas may be suggestive of a possible functional redundancy between the two teneurins, although this is still to be further investigated. Mieda et al. (1999), also performed in situ hybridization studies on tenm3 and 4 in the developing zebrafish embryo and showed that the expression patterns were complementary to each other early on in the developing embryo (<26 hpf) but becomes ambiguous at later stages.

While we are one of the first to present an in-depth study of the expression pattern of all four teneurin paralogs across the zebrafish embryo during development, comprehensive expression data is already available for other species such as the mouse. For example, Zhou et al. (2003) showed that the expression of the four teneurin genes in the developing and adult central nervous system of mice was distinct but partially overlapping. Similar to what we found in the zebrafish, the teneurins were expressed widely across the central nervous system and significant levels of tenm3 and 4 were detected early on during development (Zhou et al., 2003). All four teneurins show a graded expression across the cortex and striatum during embryonic and early postnatal stages (Li et al., 2006; Bibollet-Bahena et al., 2017), and intriguingly, tenm4 is also expressed in the subventricular zone, suggesting a possible role in determining cortical progenitor cell fate (Li et al., 2006).

While tenm3 is expressed in interconnected areas of the zebrafish visual system such as the retina and optic tectum, similarly, in the mouse visual system, tenm3 is expressed across interconnected regions of the retina, dLGN, superior colliculus (SC) and visual cortex (Leamey et al., 2007, 2008). Tenm2 and tenm4 in mice have also been shown to be expressed strongly in the interconnected visual cortex and dLGN (Li et al., 2006). Mouse hippocampal tenm expression also matches the topographic connectivity between the entorhinal cortex, CA1 and subiculum (Berns et al., 2018).

Zhou et al. (2003) also noted a differential distribution of mRNA transcripts and protein, which has also been previously observed in Drosophila, and which they explain as possible axonal transport into target regions (Baumgartner et al., 1994; Zhou

TABLE 1 | Mean dendritic field diameter and area of mCitrine-labeled retinal ganglion cells in tenm1-BAC-injected fish.


The dendritic field diameter was measured as the width of the dendritic arbor, while the dendritic field area was measured from the top views of the RGCs. Measurements were made with ImageJ. n represents the number of cells analyzed per type. SD is shown for all cell types where n > 1.

et al., 2003). It would be insightful to investigate this further once reliable antibodies for detecting teneurin protein expression in zebrafish are available.

In the developing chick visual system, tenm1 and 2 have been found to be predominantly expressed in non-overlapping populations of neurons during the time of axonal growth in embryos (Rubin et al., 1999). Tenm1 is expressed in the IPL, the ganglion cell layer (GCL) and the tectum and displays a complementary pattern of expression with tenm2 in different sublaminae of the IPL in chick (Kenzelmann et al., 2008). To further investigate teneurin expression in the developing visual system of zebrafish, different teneurin:mCitrine BAC constructs were injected into zebrafish embryos in an attempt to create transient transgenic lines expressing tenm1, 2.14 and 4. Although tenm1 and 2 transcript expression was not observed in the retina with in situ hybridization techniques, BAC transgenesis enabled a more sensitive detection of lower levels of teneurin expression.

In addition to validating these BAC constructs, which consistently labeled both amacrine cells and RGCs, a preliminary classification of cells expressing teneurins in the inner retina and optic tectum was accomplished. While the tenm1:mCitrine BAC construct primarily labeled five subtypes of RGCs, the tenm2.14:mCitrine BAC labeled nine amacrine cell subclasses. Tenm4:mCitrine BAC transgenesis was less efficient and exhibited low fluorescence levels so a detailed morphological classification of tenm4 expressing cells could not be established. Tenm3, on the other hand, has been shown to be expressed by RGCs, amacrine cells and also tectal neurons in zebrafish embryos via in situ hybridization studies (Antinucci et al., 2013), while a BAC transgenic line with Gal4FF under the

transcriptional control of regulatory elements upstream and downstream of the tenm3 start codon, Tg(tenm3:Gal4), was combined with a Tg(UAS:tagRFP-CAAX) responder line and labeled subsets of amacrine and tectal cells (Antinucci et al., 2016). The tenm3-positive amacrine cells consistently stratified their neurites in three IPL strata but did not stratify correctly in tenm3 knockout mutants (Antinucci et al., 2016). Indeed, tenm3 expression is critical for the proper development of orientation selectivity in the retina (Antinucci et al., 2016; Antinucci and Hindges, 2018). Further studies into the classification of different teneurin positive cells, will allow us to potentially relate distinct RGC and amacrine cell types to visual functionalities based on their morphology and stratification, with BAC transgenesis being a viable method for accomplishing this.

While the functional dimerization of teneurin into covalent, disulphide-linked homodimers mediates many of its physiological effects via homophilic interactions, all four teneurin paralogs may also participate in the formation of heterodimers (Feng et al., 2002). If multiple teneurins are expressed in a single cell type, it would be prudent to suggest that different hetero- and homodimer combinations could be utilized to establish a range of hetero- and homophilic interactions at the

TABLE 2 | Mean dendritic field diameter and area of mCitrine-labeled amacrine cells in tenm2.14-BAC-injected fish.


The dendritic field diameter was measured as the width of the dendritic arbor, while the dendritic field area was measured from the top views of the amacrine cells. Measurements were made with ImageJ. n represents the number of cells analyzed per type. SD is shown for all cell types where n > 1.

cell membrane. Combined with alternative splicing events and heterophilic interactions with other cell adhesion molecules, the teneurins may form part of a wider molecular code that functions to increase the diversity of cellular interactions from a limited set of genes. In this fashion, different cell adhesion molecules that are specific for a certain cell type may interact in a combinatorial fashion to drive synaptic specificity.

Much of the focus on teneurin functionality in circuit assembly has been through studying tenm3 (Antinucci et al., 2016; Berns et al., 2018). The functional involvement of the other three paralogs, if any, is less well established and more research is needed to shed light on whether these other teneurins may also act in a similar way to regulate the precise connectivity of the nervous system during development. With genetic variations in the human teneurin gene loci implicated as a significant susceptibility factor in many neurological disabilities such as bipolar disorder (Sklar et al., 2011), intellectual disability (Tucker and Chiquet-Ehrismann, 2006) and autism spectrum disorders (Nava et al., 2012), the involvement of the teneurins in conferring precise neural circuitry is paramount. Future studies will build up on the expression data we have presented here, relating the sites of expression to teneurin functionality and synaptic connectivity. Furthering our understanding of this enigmatic process will help bring us a step closer to understanding how the intricate connections of the nervous system are established.

#### DATA AVAILABILITY

All datasets generated for this study are included in the manuscript and/or the **Supplementary Files**.

#### REFERENCES


#### ETHICS STATEMENT

This work was approved by the local Animal Welfare and Ethical Review Body (King's College London), and was carried out in accordance with the Animals (Scientific Procedures) Act 1986, under license from the United Kingdom Home Office to RH.

#### AUTHOR CONTRIBUTIONS

RH designed the study. KT, MR-C, and PA carried out the experiments. KT, MR-C, PA, and AC analyzed the data. AC and RH wrote the manuscript with input from the other authors.

# FUNDING

This work was supported by grants to RH from the Biotechnology and Biological Sciences Research Council (BB/M000664/1), the Leverhulme Trust (RPG-2017-168) and the Medical Research Council (MR/N026063/1). Costs for open access publication fees were covered by institutional block grant to King's College London.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnins. 2019.00158/full#supplementary-material


pathway and is required for binocular vision. PLoS Biol. 5:e241. doi: 10.1371/ journal.pbio.0050241


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Cheung, Trevers, Reyes-Corral, Antinucci and Hindges. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Teneurins and Teneurin C-Terminal Associated Peptide (TCAP) in Metabolism: What's Known in Fish?

Ross M. Reid, Khalid W. Freij, Joel C. Maples and Peggy R. Biga\*

Department of Biology, University of Alabama at Birmingham, Birmingham, AL, United States

Teneurins have well established roles in function and maintenance of the central nervous systems of vertebrates. In addition, teneurin c-terminal associated peptide (TCAP), a bioactive peptide found on the c-terminal portion of teneurins, has been shown to regulate glucose metabolism. Although, the majority of research conducted on the actions of teneurins and TCAPs has strictly focused on neurological systems in rodents, TCAP was first identified in rainbow trout after screening trout hypothalamic cDNA. This suggests a conserved functional role of TCAP across vertebrates, however, the current depth of literature on teneurins and TCAPs in fish is limited. In addition, the overall function of TCAP in regulating metabolism is unclear. This review will highlight work that has been conducted specifically in fish species in relation to the teneurin system and metabolism in order to identify areas of research that are needed for future work.

#### Edited by:

Richard P. Tucker, University of California, Davis, United States

#### Reviewed by:

Hélène Volkoff, Memorial University of Newfoundland, Canada Kouhei Matsuda, University of Toyama, Japan

> \*Correspondence: Peggy R. Biga pegbiga@uab.edu

#### Specialty section:

This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Neuroscience

Received: 01 October 2018 Accepted: 14 February 2019 Published: 05 March 2019

#### Citation:

Reid RM, Freij KW, Maples JC and Biga PR (2019) Teneurins and Teneurin C-Terminal Associated Peptide (TCAP) in Metabolism: What's Known in Fish? Front. Neurosci. 13:177. doi: 10.3389/fnins.2019.00177 Keywords: teneurin, teneurin C-terminal associated peptide, teneurin C-terminal associated peptides, fish, metabolism

#### INTRODUCTION

Teneurins, a family of highly conserved proteins, are large signaling molecules that act as type II transmembrane receptors at the cell surface, and intracellularly as transcriptional regulators when the intracellular domain is released (Tucker and Chiquet-Ehrismann, 2006; Tucker et al., 2007; Scholer et al., 2015). Originally identified in the fruit fly, Drosophila melanogaster, as ten-m and ten-a (Baumgartner and Chiquet-Ehrismann, 1993; Baumgartner et al., 1994; Levine et al., 1994), additional teneurin genes have been described in multiple species including chicken (Gallus gallus) (Minet et al., 1999; Rubin et al., 1999; Tucker et al., 2000, 2012), mouse (Mus musculus) (Oohashi et al., 1999; Tucker et al., 2012), rats (Rattus norvegicus) (Otaki and Firestein, 1999), human (Homo sapiens) (Minet and Chiquet-Ehrismann, 2000; Tucker et al., 2012), zebrafish (Danio rerio) (Mieda et al., 1999; Tucker et al., 2012), roundworm (Caenorhabditis elegans) (Drabikowski et al., 1999), and more recently, in the vase tunicate (Ciona intestinalis) (Colacci et al., 2005; Tucker et al., 2012). Four vertebrate teneurins and single C. elegans and C. intestinalis homologs have been identified (see review Tucker et al., 2007).

The teneurins contain a single transmembrane domain and large extracellular C-termini that contain domains important in protein-carbohydrate (YD-repeats) and protein-protein (EGFrepeats) interactions (**Figure 1A**; Tucker and Chiquet-Ehrismann, 2006). Recent evidence in rodents supports transcriptional regulation activity from the N-terminus (Scholer et al., 2015), and some teneurins contain Ca2+-dependent binding domains and other functional domains (Kawasaki and Kretsinger, 1994; Rubin et al., 1999). Much of teneurin biology and the role of

the numerous functional domains and potential interactions across vertebrates remain unknown. Teneurins can homo- or hetero-dimerize (**Figure 1B**) and subsequently interact with other cells through hemophilic binding or teneurins can directly interact with the latrophilin receptor to elicit cellular responses (Li et al., 2018). Additionally, a small peptide released from the C-terminus, known as teneurin C-terminal associated peptide (TCAP; **Figure 1A**), can be cleaved and is known to exert action on several cellular functions independently of teneurin action (Wang et al., 2005).

Overall, the function of teneurins as signaling molecules is highly conserved, and consistent with their ancient origin, teneurins have essential mechanisms of action during development, and more specifically during the ontogeny of the nervous system. In C. elegans and D. melanogaster, teneurins have been shown to be required for fundamental developmental processes, like cell migration and axon pathfinding (Topf and Chiquet-Ehrismann, 2011; Beckmann et al., 2013). In fruit flies, teneurin-a is a dimeric receptor present in late stages of neuronal development and regulates eye and nervous system development and muscle attachment (Baumgartner and Chiquet-Ehrismann, 1993; Fascetti and Baumgartner, 2002); while teneurin-m/Odz pair rule genes regulate segmentation and body organization (Baumgartner et al., 1994; Levine et al., 1994). The overall orientation and formation of basement membranes has been shown to be regulated by teneurin-1 in C. elegans (Trzebiatowska et al., 2008). Additionally, teneurin-m/Odz is a homolog of mammalian teneurin-4 (Oohashi et al., 1999; Fascetti and Baumgartner, 2002), which is essential for gastrulation and the epithelial to mesenchymal transition in mice (Lossie et al., 2005). It is well understood that teneurins play a vital role in neuronal wiring, and recent evidence supports involvement in regulating synaptic connections and trans-synaptic signaling (see review Mosca, 2015).

Teneurin-1 occurs in a heterodimeric form via the EGFlike repeats, and is expressed in most tissues in the developing rats (Oohashi et al., 1999). Both teneurin-1 and teneurin-2 direct signaling pathways as their intracellular domains can be translocated to the nucleus (Nunes et al., 2005). Teneurin-1 can be found in the full length or alternatively spliced forms (with medium length or intracellular domain). Most commonly, the intracellular domain of teneurin-1 is found interacting with CAP/ponsin complex to function in cell adhesion to both the matrix and other cells. Additionally, teneurin-1 intracellular domain interacts with methyl-CpG-binding protein (MBD1) to possibly regulate gene transcription of cells within the nervous system (Nunes et al., 2005).

Outside of nervous system development and function, Ishii and co-workers recently demonstrated that teneurin-4 regulates postnatal muscle growth in mice (Ishii et al., 2015). Specifically, removal of teneurin-4 resulted in stunted postnatal growth accompanied by fewer satellite cells, suggesting a role of teneurin-4 in satellite cell proliferation. Interestingly, despite fewer satellite cells, muscle repair and renewal capacity was not affected by the absence of teneurin-4 (Ishii et al., 2015). Much of the work related to teneurin biology has focused heavily on nervous system development in rodent species, however, recent evidence suggests important roles of teneurins outside of this niche.

# TENEURINS IN TELEOST FISH

Teneurins were discovered in fish in 1999, when 2 homologs of tenm/odz were isolated while looking for factors regulated by LIM/homeodomain transcription factor Islet-3 in zebrafish (Mieda et al., 1999). Tenm/odz sequence alignments and identity comparisons confirmed the presence of teneurin homologs in zebrafish: ten-m3 (teneurin-3) and ten-m4 (teneurin-4) (Mieda et al., 1999). The expression profiles of ten-m3 and ten-m4 were shown to be consistent with reported profiles from rodent species, where expression is high during development in zebrafish embryos, particularly in the developing central nervous system (Mieda et al., 1999). More specifically, ten-m3 is expressed developmentally in somites, pharyngeal arches, notochord, and the brain, whereas ten-m4 appears to only be expressed in the developing brain of zebrafish embryos (Mieda et al., 1999).

In 2012, additional teneurin genes were identified in zebrafish and stickleback (Gasterosteus aculeatus), including a complete teneurin-1 sequence and two teneurin-2 paralogues named teneurin-2a and teneurin-2b (Tucker et al., 2012). Further analysis of the teneurin sequences revealed that the teneurin sequences were mostly conserved, except for the potential absence of key processing sites, such as a furin cleavage site in teneurin-1 and nuclear localization sequences (NLS) in teneurin-1 and teneurin-2a (Tucker et al., 2012). Additionally, there is a predicted proline-rich SH3 binding domain in the N-terminal intracellular domain of teneurin-3 and a potential additional furin cleavage site on teneurin-2a. Five teneurin genes were also identified in stickleback, where stickleback have retained two teneurin-3 paralogues (teneurin-3a and teneurin-3b) and only a single teneurin-2 gene (Tucker et al., 2012). Additionally, the stickleback teneurin-1 gene contains a potential furin cleavage site and NLS in the N-terminal intracellular domain (Tucker et al., 2012). More recently, teneurin-2 expression was reported outside the developing fish embryo in the mature clownfish brain (Baraban et al., 2007) Thus, the discovery of teneurin genes in a few fish species has been fundamental in aiding in the understanding of teneurin conservation and evolution; however, little research has been performed to elucidate the roles of teneurins in fish.

Only a few publications have explored the roles of teneurins in fish (**Figure 1C**). A recent study demonstrated that teneurin-3 is expressed in retinal ganglion cells (RGCs), amacrine cells (ACs), and tectal neurons in zebrafish embryos, and evidence supports a role of teneurin-3 in shaping the morphological and functional connectivity of RGCs in developing zebrafish embryos (Antinucci et al., 2013). Additionally, teneurin-3 was shown to specify the correct development of functionally and morphologically defined subsets of ACs and RGCs, which is responsible for the formation of a circuit underlying retinal orientation selectivity. Outside of retinal developmental, teneurin-4 appears to be essential in regulating tremor disorders by modifying the intensity of myelin in the brain and the number of small-diameter


FIGURE 1 | (A) Teneurin peptides are highly conserved and contain an intracellular domain that contains polyproline cites and EH-hand-like Ca2<sup>+</sup> binding cites, a transmembrane domain, and an extracellular domain containing 8 epidermal growth factor-like repeats, 17 cystein residues, 26 tyrosine-aspartic acid repeats, and a teneurin c-terminal associated peptide (TCAP). Figure rendered from Tucker and Chiquet-Ehrismann (2006) and Woelfle et al. (2015). (B) Teneurins can dimerize via EGF-like repeats (2 and 5), causing conformational changes that can lead to homophillic binding with neighboring cells. Following dimerization, the intracellular domain can anchor to the cytoskeleton (1) or can be cleaved (2) and translocate to the nucleus where it can interact with transcriptional regulators (Woelfle et al., 2015). (C) Outline of teneurins and TCAPs gene expression and functional analyses from teleost species.

neuronal axons along the notochord in developing zebrafish (Hor et al., 2015). This study also demonstrated that teneurin-4 regulates motor axon pathfinding, outgrowth, and branching in zebrafish embryos (Hor et al., 2015; Antinucci et al., 2016). Furthermore, reducing or removing functional teneurin-3 has been recently shown to play roles in seizure activity, as knockdown models appear to have enhanced seizure resistance (Baraban et al., 2007; Hortopan et al., 2010). Therefore, evidence supports a fundamental role of both teneurin-3 and teneurin-4 in the development and functionality of nervous system tissue in zebrafish.

#### TENEURINS IN METABOLISM

It is well established that teneurins have conserved roles in neuronal development across taxa, but recent studies suggest that teneurins can affect the development and functions of nonneuronal tissues in fish. Recent work in clownfish (Amphiprion ocellaris) suggests that teneurin-2 might be important in maturing gonads, as teneurin-2 expression was upregulated in mature gonad tissue of both male and female clownfish (Casas et al., 2016). In fact, teneurin-2 (tenm-2) seems to play a role in the induction of adipocyte markers in humans, as tenm-2 expression is much higher in white adipose tissue compared to brown adipose tissue (Tews et al., 2014). After adipocyte differentiation occurs in humans, expression of tenm-2 drops sharply, however, loss-of-functions studies demonstrated that tenm-2 expression levels alone do not regulate differentiation of adipocytes (Tews et al., 2017). However, tenm-2 loss-offunction in human fat cells led to the expression of brown adipocyte markers, such as UCP1, within white adipocytes, which was corroborated by corresponding mitochondrial respiration rate increases (Tews et al., 2017), suggesting a functional role of teneurin-2 in regulating adipocyte metabolism. Further, removing tenm-2 function resulted in increased basal and cAMP-stimulated leak respiration leading to improved overall oxidative metabolism (Tews et al., 2017). Together, these data suggest that teneurins are likely to play regulatory roles in non-developing tissues and might aid in regulating metabolic functions across taxa. Furthermore, additional processing of teneurins can lead to biological activity of the TCAP that can also regulate metabolism. Little is known about the role teneurins might play in fish metabolism, leading to a wide open area awaiting investigation.

#### TENEURIN C-TERMINAL ASSOCIATED PEPTIDES (TCAP)

Around the same time that tenm-2 metabolic functions were being elucidated, another group demonstrated that the distal peptide portion of teneurins can also regulate metabolism. This distal portion, termed TCAP, was first discovered in 2004 while searching for corticotrophin-releasing factor (CRF) paralogs (Qian et al., 2004). The TCAP region was identified on the final 3 0 exon of the teneurin-3 protein in rainbow trout hypothalamic cDNA and showed characteristics of a bioactive peptide (Qian et al., 2004). TCAP can be independently transcribed or be cleaved from teneurin as a peptide, suggesting it has functional independence from its pro-protein teneurin (Qian et al., 2004; Chand et al., 2013a). Several TCAP peptides homologs (TCAP-1-4) have been identified at the extracellular end of teneurins in a number of species (Qian et al., 2004; Colacci et al., 2005; Wang et al., 2005; Lovejoy et al., 2006; Tan et al., 2011; Chand et al., 2013a, 2014; D'Aquila et al., 2017). TCAP peptides are well conserved across animal taxa (Lovejoy et al., 2009), including teleost fishes (**Figures 2A,B**). Thus far, single TCAP-1 and - 4 orthologs have been identified in zebrafish, stickleback, and pufferfish; while stickleback, pufferfish, and medaka each have known TCAP-3 paralogs (TCAP-3a and TCAP-3b) (**Figure 2C**). Zebrafish has maintained two TCAP-2 paralogs (TCAP-2a and TCAP-2b) (**Figure 2C**). Phylogenetic analysis suggests that TCAP-1 and TCAP-4 shara a common ancestor, as do TCAP-2 and TCAP-3 (**Figure 2C**), which is consistent with previously reported relationships for full-length teneurin proteins (Tucker et al., 2012). The significance and expression profiles of these fish TCAP homologs is currently not known.

Previous work has shown that TCAP shares ∼22% sequence similarity with the CRF superfamily and thus initial studies focused on targeting the roles of TCAP in relation to the stress axis (Chand et al., 2013b; Chen et al., 2013; Erb et al., 2014). Work has shown that TCAP-1 decreases stress-related behaviors in rodents and can block CRF stress-inducing effects (Al Chawaf et al., 2007a,b; Tan et al., 2008, 2009, 2011; Kupferschmidt et al., 2011). These studies established TCAP-1 as a potent anxiolytic in vivo, as rats treated with TCAP plus CRF displayed reduced stress-related behaviors, consistent with decreased anxiety (Al Chawaf et al., 2007b). Additionally, several studies have shown that TCAP-1 treatment can cause cytoskeletal reorganization of neurons in mammals (Bernstein and Bamburg, 2003; Al Chawaf et al., 2007a; Chand et al., 2013a), suggesting at least partial functional conservation in nervous system development and function between teneurins and TCAPs. Thus, further studies applied exogenous TCAP to elucidate its functions outside of stress and have since shown TCAP to be a novel growth and metabolic regulator. For example, exogenous rainbow trout TCAP-3 (rtTCAP-3) stimulated cellular proliferation and cAMP levels in Gn11 neuronal cells in vitro (Qian et al., 2004).

More recently, TCAP-1 was shown to exhibit metabolic effects in rats (Hogg et al., 2018), as TCAP-1 decreased blood glucose 40% in both Wistar rats and in the type II diabetic insulin-insensitive pathological model, Goto-Kakizaki rats (Hogg et al., 2018). Furthermore, TCAP-1 decreased insulin and increased serum glucagon levels, suggesting an effect on glucose metabolism systemically (Hogg et al., 2018). Similarly, TCAP-1 increases glucose uptake, similar to, but independent of, insulin in mouse mHypoE-38 hypothalamic cells, in vitro, with an accompanied decrease in cytosolic calcium levels and increased plasma membrane expression of GLUT3, the primary glucose transporter for neurons (Hogg et al., 2018). More importantly, TCAP-1 increased intracellular ATP concentrations in a dose dependent manner while simultaneously decreasing the levels of pyruvate and lactate, in vitro, suggesting that TCAP can enhance

Omega alignment of TCAP paralogs from the zebrafish (Danio rerio). (B) Alignment of TCAP orthologs from selected teleost species: zebrafish, D. rerio; stickleback, Gasterosteus aculeatus; spotted pufferfish, Tetraodon nigroviridis; clownfish, Amphiprion ocellaris; medaka, Oryzias latipes; rainbow trout, Oncorhynchus mykiss. (C) Predicted TCAP peptide sequences used to generate a Clustal Omega alignment and build this unrooted phylogenetic tree to predict evolutionary relationship of fish TCAP peptides.

neuronal metabolism via oxidative energy production processes (Hogg et al., 2018).

Furthermore, evolutionary analysis of the TCAP peptide revealed that it predates insulin, which suggests its metabolic functions are highly conserved (Hogg et al., 2018). D'Aquila et al. (2017) showed that TCAP-1 treatment increased contractile behaviors in C. intestinalis, which is an energy dependent behavior. This suggests that TCAP can increase the energy production and metabolism in primitive species, further supporting the notion that TCAP has conserved functions in metabolism. Taken together, these recent studies provide critical insights for the conserved roles of teneurins and TCAPs in metabolism. Unfortunately, there is no data published on the effects of TCAP on metabolism in any fish species. However, it is expected that TCAP peptides likely regulate glucose uptake and enhance energetic efficiency.

# FUTURE DIRECTIONS NEEDED TO FURTHER ELUCIDATE THE ROLE OF TENEURINS AND TCAPs IN REGULATION GROWTH AND METABOLISM

It is clear that teneurins play a regulatory role in the developing nervous system, and recent evidence suggests regulatory roles of teneurins in metabolism. Additionally, when liberated, the 40- 41-residue, TCAP, can regulate nervous system remodeling (Tan et al., 2012), the stress response of CRF signaling (Tan et al., 2012; Chen et al., 2013), and cellular metabolism. Recent evidence also suggests that TCAP likely regulates energy efficiency in cells allowing for physiological changes outside of the nervous system, such as in skeletal muscle and adipose tissue. The major focus of functional work related to teneurins and TCAPs has been conducted in rodent species and more specifically in nervous system development in rodents. However, as outlined here, more recent work has focused on the high conservation of teneurins and TCAPs functions in development, growth, and metabolism in non-rodent species, including several teleost fish species (**Figure 1C**). Teneurins, and presumably TCAPs, appears to function in nervous system development and cell morphology and migration in teleosts.

Teleosts are the largest vertebrate group and are dominant in freshwater and marine environments. Due to their worldwide distribution, they have amassed a vast amount of diversity in morphology, ecology, and behavior (Nelson et al., 2016). However, teleosts possess physiological features common to all vertebrates, as well as high genomic conservation, making them attractive models for the study of many biological questions, including the evolution of development, growth, and metabolic efficiency regulation. With this review, we wish to expand the interest in studying the functions of teneurins and TCAPs in relation to functions highlighted here, as well as related metabolic dysfunctions that are common to many human diseases. Areas of specific focus might include calcium transport, ATP production, mitochondrial function, and cell growth regulation.

#### AUTHOR CONTRIBUTIONS

fnins-13-00177 March 1, 2019 Time: 19:55 # 6

PB conceived the concept and designed the outline for this mini review, as well as contributed to the writing, reviewing, and editing, and provided final editing and approval of mini review. RR contributed by assisted in designing the outline of the mini

#### REFERENCES


review and provided critical intellectual content. JM and KF provided critical intellectual content.

#### FUNDING

Research reported in this mini review was supported by the National Institute on Aging of the National Institutes of Health under award number P30 AG050886. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.



produce enduring changes in behavioral responses to corticotropin-releasing factor (CRF) in rat models of anxiety. Behav. Brain Res. 188, 195–200. doi: 10.1016/j.bbr.2007.10.032


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Reid, Freij, Maples and Biga. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Catching Latrophilin With Lasso: A Universal Mechanism for Axonal Attraction and Synapse Formation

Yuri A. Ushkaryov<sup>1</sup> \*, Vera Lelianova<sup>1</sup> and Nickolai V. Vysokov<sup>2</sup>

<sup>1</sup> Medway School of Pharmacy, University of Kent, Chatham, United Kingdom, <sup>2</sup> BrainPatch Ltd., London, United Kingdom

Latrophilin-1 (LPHN1) was isolated as the main high-affinity receptor for α-latrotoxin from black widow spider venom, a powerful presynaptic secretagogue. As an adhesion G-protein-coupled receptor, LPHN1 is cleaved into two fragments, which can behave independently on the cell surface, but re-associate upon binding the toxin. This triggers intracellular signaling that involves the Gαq/phospholipase C/inositol 1,4,5 trisphosphate cascade and an increase in cytosolic Ca2+, leading to vesicular exocytosis. Using affinity chromatography on LPHN1, we isolated its endogenous ligand, teneurin-2/Lasso. Both LPHN1 and Ten2/Lasso are expressed early in development and are enriched in neurons. LPHN1 primarily resides in axons, growth cones and presynaptic terminals, while Lasso largely localizes on dendrites. LPHN1 and Ten2/Lasso form a trans-synaptic receptor pair that has both structural and signaling functions. However, Lasso is proteolytically cleaved at multiple sites and its extracellular domain is partially released into the intercellular space, especially during neuronal development, suggesting that soluble Lasso has additional functions. We discovered that the soluble fragment of Lasso can diffuse away and bind to LPHN1 on axonal growth cones, triggering its redistribution on the cell surface and intracellular signaling which leads to local exocytosis. This causes axons to turn in the direction of spatiotemporal Lasso gradients, while LPHN1 knockout blocks this effect. These results suggest that the LPHN1-Ten2/Lasso pair can participate in long- and short-distance axonal guidance and synapse formation.

# of Munich, Germany

\*Correspondence: Yuri A. Ushkaryov y.ushkaryov@kent.ac.uk

Daegu Gyeongbuk Institute of Science and Technology (DGIST),

Ludwig Maximilian University

#### Specialty section:

Edited by: Antony Jr. Boucard,

> Reviewed by: Jaewon Ko,

South Korea Ozgun Gokce,

Centro de Investigación y de Estudios Avanzados (CINVESTAV), Mexico

> This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Neuroscience

Received: 22 December 2018 Accepted: 05 March 2019 Published: 22 March 2019

#### Citation:

Ushkaryov YA, Lelianova V and Vysokov NV (2019) Catching Latrophilin With Lasso: A Universal Mechanism for Axonal Attraction and Synapse Formation. Front. Neurosci. 13:257. doi: 10.3389/fnins.2019.00257 Keywords: teneurin, latrophilin, lasso, axonal attraction, cell adhesion

#### ISOLATION AND ARCHITECTURE OF LATROPHILIN

This story began in the early 1970s, when it was found that the venom from the black widow spider, Latrodectus mactans, causes massive release of neurotransmitters from vertebrate synapses (Longenecker et al., 1970). The neurotoxin purified from this venom, α-latrotoxin (αLTX), was shown to form Ca2+-permeable pores in artificial membranes (Finkelstein et al., 1976). However, it acted only after binding a high-affinity presynaptic receptor/s in neuronal cells. Even more intriguingly, αLTX could act in the absence of extracellular Ca2<sup>+</sup> (Longenecker et al., 1970). These findings suggested that the toxin receptor had a potential to stimulate the presynaptic neurotransmitter release machinery directly, bypassing the requirement for Ca2<sup>+</sup> in vesicular exocytosis.

Fascinated by these characteristics, several groups began their quest for the Ca2+-independent αLTX receptor, using the toxin as an affinity adsorbent (Scheer and Meldolesi, 1985; Ushkarev and Grishin, 1986; Petrenko et al., 1990). The first receptor preparation contained several proteins (Petrenko et al., 1990), of which the largest was termed neurexin Iα (Ushkaryov et al., 1992). However, as neurexin required Ca2<sup>+</sup> to bind αLTX and did not display clear signaling capabilities, the search for the Ca2+-independent receptor continued. Eventually, two laboratories simultaneously isolated this protein using αLTX affinity columns and called it latrophilin 1 (LPHN1) (Davletov et al., 1996) or Ca2+-independent receptor for αLTX 1 (CIRL1) (Krasnoperov et al., 1996). Its amino acid sequence (Krasnoperov et al., 1997; Lelianova et al., 1997) showed homology to G proteincoupled receptors (GPCRs) of the secretin group.

However, the toxin receptor was clearly different (**Figure 1A**): (1) it had a very long N-terminal extracellular domain (ECD) containing regions of homology to extracellular proteins (lectin and olfactomedin), (2) it was proteolytically cleaved upstream of the first transmembrane domain (TMD), (3) this constitutive cleavage occurred inside the cell and did not lead to signaling (Krasnoperov et al., 2002; Volynski et al., 2004), (4) the resulting N-terminal fragment (NTF) remained largely associated with the 7TMD C-terminal fragment (CTF) (Krasnoperov et al., 1997), but (5) the fragments could dissociate and behave as independent membrane proteins (Volynski et al., 2004; Silva et al., 2009a).

A number of similarly, built receptors was soon identified either biochemically or genetically. Based on their common features, they were isolated into a separate family, "Adhesion GPCRs" (aGPCRs) (Fredriksson, 2003). According to the modern nomenclature recommended by the International Union of Basic and Clinical Pharmacology, the group is now called ADhesion G protein-coupled Receptors (ADGRs), of which LPHN1 represents the Latrophilin subfamily, ADGRL (Hamann et al., 2015).

It is now established that aGPCRs are a large and ancient family of GPCRs (Hamann et al., 2015). They all contain similar 7TMD domains, which also resemble GPCRs from other families, but these are connected to variable C-terminal tails and to a surprisingly vast array of long N-terminal ectodomains. This diversity of the extracellular domain, featuring homology to various protein classes involved in protein-protein interactions and cell-adhesion, combined with a conserved signaling domain, has led to this group being dubbed "chimerical receptors" (e.g., Kwakkenbos et al., 2006), which probably reflects the way they appeared in evolution. In all aGPCRs (except GPR123 with a very short ectodomain) the ectodomains are connected to the 7TMDs by a conserved "GPCR autoproteolysis-inducing" (GAIN) domain (Araç et al., 2012), previously known as a "GPCR proteolysis site" (GPS) (Krasnoperov et al., 1997). The GAIN domain in almost all aGPCRs undergoes internal proteolysis and is then unequally divided between the NTF and CTF: the larger portion of the GAIN domain remains part of the NTF and can bind the smaller C-terminal portion, which forms the very N-terminus of the CTF. This interaction mediates noncovalent association of the fragments (**Figures 1A,B**), but the two parts of the GAIN domain can dissociate, leading to important changes in receptor functions. This dynamic structure may be key to understanding the physiological functions of aGPCRs. In full agreement with their name, many aGPCRs have been shown to bind large ligands on the surface of other cells or in the extracellular matrix, thus enabling the conversion of extracellular interactions into intracellular signals. Many family members have been demonstrated to signal via G proteins, as proper GPCRs, while others can signal independently of G proteins, however, the signaling capabilities of aGPCRs are only beginning to be understood (Hamann et al., 2015), and LPHN1 is one of the few aGPCRs for which G protein coupling has been unequivocally demonstrated.

# SIGNALING

LPHN1 signaling has been extensively studied using LTXN4C, a mutant αLTX that acts as an exogenous ligand of this receptor but fails to form tetramers and membrane pores (Ichtchenko et al., 1998; Volynski et al., 2003, 2004), which are characteristic of the wild-type αLTX (Orlova et al., 2000). LTXN4C binds to the GAIN domain within the NTF (Krasnoperov et al., 1999; Lin et al., 2004; Araç et al., 2012) with high affinity (∼1 nM) (Ichtchenko et al., 1998; Volynski et al., 2003) and causes a strong and sustained increase in "spontaneous" neurotransmitter release (Ashton et al., 2001; Capogna et al., 2003; Volynski et al., 2004; Lelyanova et al., 2009; Déak et al., 2009). This effect is purely presynaptic, as only the frequency of miniature events is affected, but not their amplitude or duration (Capogna et al., 2003). Unable to make transmembrane pores, LTXN4C can only exert its action via receptor-mediated signaling, and receptor knockout or mutagenesis (leading to a loss of signal transduction) obliterates the toxin-evoked signal (Tobaben et al., 2000; Volynski et al., 2004).

Binding of LTXN4C to the NTF induces its re-association with the CTF and subsequent signaling (Volynski et al., 2004; Silva et al., 2009a; Vysokov et al., 2016, 2018). A very similar behavior was reported also for EMR2 (Huang et al., 2012) and may be a universal feature of all aGPCRs. However, it is not clear whether the NTF-CTF complex has the same structure before the separation of its fragments and after their re-association.

Similar to many other GPCRs, LPHN1 probably activates multiple signaling mechanisms, but at least one that leads to increased neurotransmitter release has been studied in detail (**Figure 1C**). LTXN4C-induced association of the NTF and CTF causes Gαq-mediated (Rahman et al., 1999) activation of phospholipase C (PLC) (Davletov et al., 1998; Capogna et al., 2003; Volynski et al., 2004), which cleaves phosphatidylinositol 4,5-bisphosphate (PIP2), producing inositol 1,4,5-trisphosphate (IP3) (Lelianova et al., 1997; Ichtchenko et al., 1998) and diacylglycerol. Both physical and functional interaction of CTF with Gαq was demonstrated by NTF-mediated pull-down experiments, where the CTF–Gαq complex persisted in the presence of GDP, but was lost when GDP was replaced with GTP (Rahman et al., 1999). Furthermore, the overexpression of LPHN1 in COS7 cells itself substantially decreased the resting concentration of IP<sup>3</sup> (due to non-productive binding

of the bulk of cellular Gαq by the inactive overexpressed receptor); reciprocally, activation of LPHN1 upregulated IP<sup>3</sup> (Lelianova et al., 1997) The specific involvement of Gαq and PLC in LPHN1-mediated effects was experimentally demonstrated in synaptosomes, organotypic neuronal cultures, LPHN1-transfected NB2a cells, LTX-sensitive MIN6 β-cell line, and neuromuscular junctions (Davletov et al., 1998; Capogna et al., 2003; Volynski et al., 2004; Lajus et al., 2006; Lelyanova et al., 2009). The IP3-induced increase in cytosolic Ca2<sup>+</sup> can be inhibited by intracellular Ca2<sup>+</sup> chelators, intracellular store depletion using thapsigargin, or by inhibition of the IP<sup>3</sup> receptor using xestospongin C or 2-APB (Davletov et al., 1998; Capogna et al., 2003; Lajus et al., 2006). This demonstrates the strict dependence of LPHN1-mediated effect on intact intracellular Ca2<sup>+</sup> stores, IP<sup>3</sup> receptor activity, and ultimately on an increase in cytosolic Ca2<sup>+</sup> concentration. Calcium released by LTXN4C from the stores is not, however, sufficient to stimulate substantial exocytosis, at least in large synapses, such as neuromuscular junctions (Lelyanova et al., 2009), and extracellular 0.2–1 mM Ca2<sup>+</sup> is required to support the effect of LTXN4C-evoked LPHN1 signaling on neurotransmitter exocytosis (Davletov et al., 1998; Ashton et al., 2001; Capogna et al., 2003; Volynski et al., 2003; Lajus et al., 2006; Lelyanova et al., 2009). This is most likely due to the signaling-induced opening of store-operated Ca2<sup>+</sup> channels and influx of extracellular Ca2+, as hypothesized previously (Ushkaryov et al., 2008). Interestingly, presynaptic Ca2<sup>+</sup> stores

and, more specifically, store-operated Ca2<sup>+</sup> entry into nerve terminals has been recently shown to play a critical role in the control of neurotransmitter release (de Juan-Sanz et al., 2017).

The endogenous ligand of LPHN1 teneurin-2 (Ten2), or Lasso, (see below) causes a similar NTF-CTF reassociation and rise in cytosolic Ca2<sup>+</sup> which then stimulates rapid store-operated Ca2<sup>+</sup> entry (Silva et al., 2011; Vysokov et al., 2018), although the duration of the Ten2/Lasso effect is relatively short (Vysokov et al., 2018). The ligand-bound NTF thus appears to serve as an agonist of the CTF (Volynski et al., 2004; Huang et al., 2012), although the CTF may have its own ligands.

In fact, at least some signaling by free CTF may be induced by the small piece of the ECD that remains at the N-terminus of the CTF after the cleavage of NTF (**Figure 1C**). This hydrophobic peptide, called 7 amino acids (Volynski et al., 2004), stalk (Kishore et al., 2016) or Stachel peptide (Liebscher et al., 2014), can act as a "tethered ligand" (Liebscher et al., 2014; Stoveken et al., 2015). Normally, Stachel mediates the interaction between the CTF and NTF. It is thought that conformational changes induced by ligand binding to the NTF (or its complete removal) free up Stachel peptide, allowing it to interact with the 7TMD and trigger signaling (Liebscher et al., 2014; Stoveken et al., 2015; Nazarko et al., 2018). Micromolar concentrations of exogenous Stachel can activate signaling even without ligand binding to, or removal of, the NTF (Liebscher et al., 2014; Nazarko et al., 2018; **Figure 1B**). However, in LPHN1 Stachel-induced signaling appears to be different from that produced by the binding of NTF ligands. Thus, exogenous Stachel peptide caused a pertussis toxin-sensitive decrease in cAMP levels (Nazarko et al., 2018). By contrast, NTF ligands usually increase cAMP levels (**Figure 1C**, left) [e.g., after activation of LAT-1 by its endogenous ligand in Caenorhabditis elegans (Winkler and Prömel, 2016) or activation of rat LPHN1 expressed in COS7 cells by LTX (Lelianova et al., 1997)]. Also, the NTF-CTF complex did not bind Gαi in pull-down experiments (while Gαs was not tested) (Rahman et al., 1999).

These data indicate that LPHN1 might send different intracellular signals depending on (1) the interaction between the NTF and CTF, (2) the agonist involved and (3) the state of cell's signaling and protein modification machinery.

### ISOLATION OF LASSO

Several features of LPHN1 – (1) the ability of its NTF (in complex with its ligand/s) to activate the CTF (Volynski et al., 2004; Silva et al., 2009a) and send an exocytotic signal; (2) the size of the NTF, which is sufficient to span half of the synaptic cleft; and (3) the presynaptic localization of LPHN1 (Silva et al., 2011; Vysokov et al., 2016) – led us to hypothesize that the NTF could bind a postsynaptic ligand. Not only would then the NTF, being held at the active zone by trans-synaptic interactions with a postsynaptic protein, always localize close to presynaptic vesicle release sites, but it would also provide presynaptic docking sites for the independently recycling CTF and potentially enable retrograde signaling (Volynski et al., 2004). These ideas prompted us to start looking for an LPHN1 ligand, operationally called "LPHN1-associated synaptic surface organizer" (Lasso) (Silva et al., 2009b).

When designing a soluble LPHN1 construct to make an affinity column (**Figure 1D**), we relied on our knowledge of the NTF-CTF relationship. Thus, although the NTF-CTF complex has a high affinity for αLTX/LTXN4C, it can also dissociate (Silva et al., 2009a), possibly upon binding an antagonist, so anchoring the NTF-CTF complex via CTF could be inefficient. On the other hand, if the NTF is synthesized without Stachel or if the NTF-CTF cleavage is blocked (e.g., due to a mutation), the NTF assumes a conformation that does not bind αLTX (Silva et al., 2011) but could bind non-specific ligands. Thus we anchored the full ECD (containing the NTF and Stachel peptide) on the column via an N-terminal V5 epitope (**Figure 1D**).

Affinity chromatography of solubilized rat brain on this adsorbent at moderate stringency (0.5 M NaCl), resulted in the isolation of the long-sought Lasso, a protein of ∼270 kDa (Silva et al., 2011). We did not observe even minute amounts of FLRT3 or neurexin, the other proposed ligands of LPHN1 (Boucard et al., 2014; O'Sullivan et al., 2014). This indicates that the chromatography conditions were too stringent for their binding to LPHN1 and that Lasso is the strongest ligand of LPHN1. Subsequent sequencing of highly purified Lasso (Silva et al., 2011) indicated that it was identical to Ten2 (Oohashi et al., 1999).

#### INTERACTION BETWEEN LPHN1 AND LASSO

Ten2/Lasso has a high affinity for LPHN1: the Kd of this complex is 0.47–1.7 nM (Silva et al., 2011; Boucard et al., 2014). The interaction between LPHN1 and Ten2/Lasso is mainly mediated by the lectin-like domain in the NTF of LPHN1 and the C-terminus of Ten2 (Boucard et al., 2014). More narrowly, it involves a short portion of the toxin-like domain of Ten2 that protrudes from the globule (Li et al., 2018). However, we found that this minimal interaction is relatively weak (Silva et al., 2011), but becomes much stronger when other parts of both ECDs are present, especially when Ten2/Lasso constructs are able to dimerize (Silva et al., 2011; Vysokov et al., 2016). Indeed, our observations suggest that dimeric Ten2/Lasso can clasp LPHN1. This could explain why a splice site (SS) in the Ten2 β-propeller domain, which is located far from the toxin-like domain, affects cell-surface interactions between Ten2 and LPHN1 (Li et al., 2018): the small SS insert could change the relative positions of the Ten2 monomers in the dimer, rendering them unable to clasp the LPHN1 molecule (**Figure 2A**).

The length of the NTF of LPHN1 (as indicated by the crystal or NMR structure of its domains, **Figure 1B**) is 10–15 nm, while Ten2 is longer than 12 nm (Li et al., 2018), which is sufficient for the two proteins to interact across the synaptic cleft (about 20 nm).

As mentioned, Ten2/Lasso binding to LPHN1 stimulates Ca2<sup>+</sup> signaling (Silva et al., 2011; Vysokov et al., 2016, 2018; **Figure 1C**). This is true of the whole soluble ECD of Lasso (Vysokov et al., 2016, 2018) or even its C-terminal toxin-like fragment, when used at higher concentrations (Silva et al., 2011).

Furthermore, when Lasso is allowed to interact with LPHN1 prior to LTXN4C, it substantially decreases the delay that normally precedes toxin's action (Vysokov et al., 2018), thought to be required for NTF and CTF rearrangement on the cell surface prior to signaling (Volynski et al., 2004).

#### LOCALIZATION OF LPHN1 AND LASSO IN THE BRAIN

Both LPHN1 and Ten2/Lasso are expressed early in development (Vysokov et al., 2018) and are highly enriched in the CNS, but there seems to be some disagreement regarding the localization of LPHN1 in the synapse. Although LPHN1-mediated effects of α-LTX are irrefutably presynaptic, there have been suggestions that LPHN1 is expressed on the postsynaptic membrane (Meza-Aguilar and Boucard, 2014). This assumption is based on proteome analysis of postsynaptic densities (Collins et al., 2006) and on LPHN1 interaction with a postsynaptic protein Shank3 (Tobaben et al., 2000).

However, these indirect findings did not indicate that LPHN1 was located in the postsynaptic membrane. First, the proteomic study (Collins et al., 2006) only isolated synaptic densities and made no attempt to separate them from presynaptic components tightly associated with postsynaptic components by trans-synaptic complexes and scaffold proteins (Dresbach et al., 2001). As a result, such presynaptic/vesicular proteins as synapsin-1, Munc-13, NSF, bassoon, synaptotagmin-1, and SNAP-25 co-purified with postsynaptic densities even to a greater extent than LPHN1. In contrast, postsynaptic neuroligin appeared to be equally "presynaptic" as its presynaptic ligand neurexin. In addition, it is important to note that the NTF of LPHN1 is non-covalently anchored in the presynaptic membrane and, being strongly bound to Ten2/Lasso on the postsynaptic membrane (Silva et al., 2011), it could ectopically co-purify with postsynaptic membrane. Finally, although the CTF of LPHN1 can interact with Shank3 (Ponna et al., 2018), Shank3 is not exclusively postsynaptic and is also present in presynaptic nerve terminals (Halbedl et al., 2016).

On the other hand, the presynaptic localization of LPHN1 is supported by several findings: during neuronal development LPHN1 concentrates at the leading edge of axonal growth cones (Vysokov et al., 2018) and subsequently becomes enriched in mature nerve terminals (Silva et al., 2011). Furthermore, comparative distribution of Ten2 and LPHN1 in the cerebellum leads to unequivocal conclusions.

Thus, Ten2/Lasso protein is most abundant in the molecular layer of the cerebellum (Zhou et al., 2003). In this layer, the bulk of presynaptic components are provided by granule cell axons (parallel fibers), while the majority of postsynaptic components is located on the dendritic trees of Purkinje and basket cells. Interestingly, Ten2 mRNA is highly expressed in Purkinje, basket and stellate cells, but is almost absent from granule cells (Zhou et al., 2003). LPHN1 protein is also highly enriched in the molecular layer, as evidenced by Ca2+ independent α-LTX binding (Davletov et al., 1998). In contrast to Ten2, LPHN1 mRNA is predominantly found in granule cells, but not in Purkinje cells (Lein et al., 2007) and so can only be delivered to the molecular layer with parallel fibers. This complementary expression of the two proteins in the cerebellum strongly indicates that LPHN1 is presynaptic and Ten2/Lasso is postsynaptic, and that they interact across the synaptic cleft. Moreover, this arrangement holds for the bulk of central synapses, as was shown by denaturing synaptic cleft complexes with urea and dithiothreitol and separating preand postsynaptic components using differential centrifugation (Berninghausen et al., 2007). After this procedure, 88 ± 8% of the NTF of LPHN1 were clearly presynaptic, while only 12 ± 4% of it might be actually present in the postsynaptic membrane (Silva et al., 2011).

#### CLEAVAGE AND SHEDDING OF LASSO

Soon after the discovery of Ten2, it was shown to be cleaved at an extracellular furin site between the TMD and EGF repeats (Oohashi et al., 1999; Rubin et al., 1999). This led to suggestions that teneurins can act both as cell-surface receptors and as diffusible signaling molecules (Rubin et al., 1999; Tucker et al., 2001). Furin-induced cleavage was thought to release the ECD into the medium, but it was unclear whether this shedding was

constitutive or signaling-induced. Unexpectedly, our recent work showed that furin-mediated proteolysis of Ten2/Lasso occurs constitutively inside the cell (**Figure 2B**). When this fully cleaved protein is delivered to the cell surface, its ECD remains tethered to the membrane by non-covalent interactions with the fragment containing the TMD (Vysokov et al., 2016).

The shedding of Ten2/Lasso occurs as a result of further, regulated proteolysis at another, near-membrane site, which releases the whole ECD into the medium (**Figure 2B**). Given that Ten2/Lasso shedding begins early in neuronal cultures (Vysokov et al., 2018), when it is not yet involved in transsynaptic interactions, and because this shedding slows down dramatically at the end of synaptogenesis (Vysokov et al., 2016, 2018), we thought that Ten2/Lasso cleavage had a role in synapse formation.

What could be the target of released Ten2/Lasso? Homophilic interaction between Ten dimers was previously proposed (Oohashi et al., 1999), and homophilic adhesion between cells expressing exogenous Ten2 was reported (Rubin et al., 2002; Beckmann et al., 2013), but not confirmed by other researchers (Silva et al., 2011; Boucard et al., 2014; Li et al., 2018; Vysokov et al., 2018). On the other hand, we observed a reliable and strong binding of shed Ten2/Lasso to LPHN1 on the surface of cultured cells and axonal growth cones (Vysokov et al., 2016, 2018). This led us to hypothesize (Vysokov et al., 2016) that during neuronal development released ECD of Ten2 could act as a soluble ligand of LPHN1, leading to changes in growth cone behavior.

# LASSO AND LATROPHILIN IN AXONAL ATTRACTION

As we began exploring the role of LPHN1—Ten2 (-SS)/Lasso interaction in brain development and neurotransmitter release, a series of studies was published describing the role of teneurins in axon guidance (Kenzelmann et al., 2007; Young and Leamey, 2009). This was further confirmed when experiments with Ten3 and Ten2 knockouts in mice demonstrated a profound deficit in at least the visual circuitry (Leamey et al., 2007; Young et al., 2013). However, axon guidance was unlikely to be mediated by the proposed homophilic interactions of Ten2, as they had been shown to inhibit, rather than promote, neurite outgrowth (Beckmann et al., 2013; Young et al., 2013). In addition, symmetric homophilic interactions between teneurins were unlikely to determine the distinct behaviors of axons and dendrites. Therefore, when we discovered that Ten2/Lasso ECD binds LPHN1 (Silva et al., 2011; Vysokov et al., 2016), this suggested fundamentally novel functions for both proteins.

First evidence to support the role of LPHN1—Ten2/Lasso interaction in axon guidance came from our finding that, in contrast to Lasso, LPHN1 is expressed on axonal growth cones (Vysokov et al., 2018). Additionally, LPHN1 activation by exogenous ligands was known to induce exocytosis via IP3 induced Ca2<sup>+</sup> release (Capogna et al., 2003; Volynski et al., 2003), a mechanism common for many axonal attractants (Tojima et al., 2011). Therefore, it was reasonable to hypothesize that the

released fragment of Ten2/Lasso could mediate axonal attraction via LPHN1. We then used microfluidic devices to create spatiotemporal gradients of soluble Ten2/Lasso ECD and demonstrated that attracted rat hippocampal axons, without increasing their general length (Vysokov et al., 2018; **Figure 3**). Importantly, this steering effect was mediated by LPHN1, because it was not detected in LPHN1 knockout mice.

We also demonstrated a possible mechanism for this attraction, whereby released ECD of Ten2/Lasso, similar to LTXN4C, was able to bind LPHN1 on transfected cells and growth cones, causing an association of LPHN1 fragments, induction of Ca2<sup>+</sup> release and an increase in the rate of exocytosis. Again, LPHN1 knockout experiments indicated that LPHN1 is required for such a mechanism (Vysokov et al., 2018).

This mechanism could mediate axonal attraction throughout the CNS, but may not be limited to it. Given that Ten2 is expressed in chicken embryo both in the CNS, but also in dorsomedial edges of somites, craniofacial mesenchyme and developing limb buds (Tucker et al., 2001), it is tempting to speculate that Ten2/Lasso released by peripheral tissues could also serve as a diffusible factor attracting motor and sensory axons to grow toward their peripheral targets.

#### REFERENCES


Taken together, these results indicate that the shed ECD of Lasso/Ten2 can act as a soluble guidance molecule through its interaction with LPHN1. This work has provided a plausible first explanation of teneurins' role in brain development and discovered a universal mechanism that uses the same protein-protein interactions both for long-distance axonal attraction and for cell contacts during synapse formation (as summarized in **Figure 3**).

#### AUTHOR CONTRIBUTIONS

YU conceived and coordinated the work, wrote the manuscript. VL and NV analyzed the literature and wrote parts of the manuscript. All authors contributed to the conception and/or writing of the manuscript.

# FUNDING

This work was supported by a Wellcome Trust Project Grant WT083199MF, a Biotechnology and Biological Science Research Council Core Support Grant BBF0083091, and core funding from the University of Kent School of Pharmacy (to YU).


neuronal development. Cell. Mol. Life Sci. 64, 1452–1456. doi: 10.1007/s00018- 007-7108-9



**Conflict of Interest Statement:** NV is affiliated with BrainPatch Ltd.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Ushkaryov, Lelianova and Vysokov. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Latrophilins: A Neuro-Centric View of an Evolutionary Conserved Adhesion G Protein-Coupled Receptor Subfamily

Ana L. Moreno-Salinas<sup>1</sup> , Monserrat Avila-Zozaya<sup>1</sup> , Paul Ugalde-Silva<sup>1</sup> , David A. Hernández-Guzmán<sup>2</sup> , Fanis Missirlis<sup>2</sup> and Antony A. Boucard<sup>1</sup> \*

<sup>1</sup> Department of Cell Biology, Centro de Investigación y de Estudios Avanzados del Instituto Politécnico Nacional (CINVESTAV-IPN), Mexico City, Mexico, <sup>2</sup> Department of Physiology, Biophysics and Neurosciences, Centro de Investigación y de Estudios Avanzados del Instituto Politécnico Nacional (CINVESTAV-IPN), Mexico City, Mexico

#### Edited by:

Stanko S. Stojilkovic, National Institutes of Health (NIH), United States

#### Reviewed by:

Joao Carlos dos Reis Cardoso, University of Algarve, Portugal Taka-aki Koshimizu, Jichi Medical University, Japan

#### \*Correspondence:

Antony A. Boucard antonyboucardjr@cell.cinvestav.mx

#### Specialty section:

This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Neuroscience

Received: 12 April 2019 Accepted: 20 June 2019 Published: 09 July 2019

#### Citation:

Moreno-Salinas AL, Avila-Zozaya M, Ugalde-Silva P, Hernández-Guzmán DA, Missirlis F and Boucard AA (2019) Latrophilins: A Neuro-Centric View of an Evolutionary Conserved Adhesion G Protein-Coupled Receptor Subfamily. Front. Neurosci. 13:700. doi: 10.3389/fnins.2019.00700 The adhesion G protein-coupled receptors latrophilins have been in the limelight for more than 20 years since their discovery as calcium-independent receptors for α-latrotoxin, a spider venom toxin with potent activity directed at neurotransmitter release from a variety of synapse types. Latrophilins are highly expressed in the nervous system. Although a substantial amount of studies has been conducted to describe the role of latrophilins in the toxin-mediated action, the recent identification of endogenous ligands for these receptors helped confirm their function as mediators of adhesion events. Here we hypothesize a role for latrophilins in inter-neuronal contacts and the formation of neuronal networks and we review the most recent information on their role in neurons. We explore molecular, cellular and behavioral aspects related to latrophilin adhesion function in mice, zebrafish, Drosophila melanogaster and Caenorhabditis elegans, in physiological and pathophysiological conditions, including autism spectrum, bipolar, attention deficit and hyperactivity and substance use disorders.

Keywords: latrophilin, teneurin, adhesion G protein-coupled receptors, cell adhesion molecules, neuronal synapse, alternative splicing, actin cytoskeleton, psychiatric disorders

# LATROPHILINS, 22 YEARS AFTER THEIR DISCOVERY

As the field of research on Adhesion G Protein-Coupled Receptors (aGPCR) is rapidly expanding, so is the interest for many of its subfamilies given their involvement in various physiological and pathophysiological events relevant to human health. A prototypical aGPCR subfamily named the latrophilins has attracted attention for more than 20 years since their discovery as part of an effort to identify the biological target mediating the calcium-independent effects of α-latrotoxin, a potent neurotoxin from the black widow venom (Krasnoperov et al., 1997; Lelianova et al., 1997; Sugita et al., 1998). Latrophilins qualify as prototypical because the study of these proteins provided many landmark discoveries that have later paved the way for understanding aGPCRs structure and function in general. Fast-forward 22 years later what do we know about latrophilins? Here, after reviewing many studies, we can only start formulating the broad realm of their function: the widespread expression of latrophilin receptors in many tissues uncovers just the tip of the iceberg; their role in neuronal tissues places latrophilins at the crown of prototypical aGPCRs.

# LATROPHILIN DOMAIN ORGANIZATION

Latrophilins are composed of the following domains which are schematized in **Figure 1**: two adhesion modules, the Lectin and Olfactomedin domains (the latter being absent in invertebrates); followed by a Hormone Binding Region adjacent to a GPCR autoproteolytic inducing domain (GAIN) which encompasses a cleavage site (GPS); and a GPCR region characterized by seven transmembrane helices with interconnecting loops and a C-terminal tail. The autoproteolytic event generates a bipartite protein composed of an extracellular N-terminal fragment (NTF) and a C-terminal fragment (CTF), with both fragments noncovalently linked to each other at the cell membrane (Arac et al., 2012). Latrophilins are among the most conserved aGPCRs with a presence that spans a wide spectrum of the evolutionary tree, suggesting that they may contribute to important functions in neuronal physiology (Krishnan et al., 2016).

# ENDOGENOUS LIGANDS FOR LATROPHILINS

Latrophilins are expressed as three isoforms in mammals (Krishnan et al., 2016) and are receptors for a variety of ligands. While some ligands appear to be common to all three isoforms, others are rather restricted to some isoforms. The list of ligands has been growing with the discovery of teneurin-2 (also known as Lasso, latrophilin associated protein; splice variant) (Silva et al., 2011), followed by neurexins (Boucard et al., 2012), FLRT (O'Sullivan et al., 2012) and finally contactin-6 (Zuko et al., 2016).

#### Teneurins

The first family of endogenously expressed extracellular ligands described for latrophilins comprise members of a four-isoforms group in mammals: teneurin-1, -2, -3, and -4 (Silva et al., 2011). Out of these high molecular weight proteins, teneurin-2 or Lasso (teneurin-2 splice variant), was the first to be identified as a highaffinity partner for latrophilins although the remaining isoforms were subsequently included as part of the potential interactors (Silva et al., 2011; O'Sullivan et al., 2012, 2014; Boucard et al., 2014). As type-II membrane proteins teneurins project their c-terminal adhesion domains toward the extracellular media to consolidate their interaction with latrophilins (**Figure 1A**). Such interaction mainly occurs between the extreme c-terminal region of teneurin and the Lectin-like domain of latrophilins but requires the additional contribution of the Olfactomedin domain in order to reconstitute a high-affinity binding site (**Figure 1A**; Boucard et al., 2014). As is the case for other adhesion molecule families, alternative splicing modifies the quaternary structure of teneurins, a mechanism that has recently been reported to generate homophilic adhesion complexes stabilizing cell-cell contacts (Berns et al., 2018). The teneurin-latrophilin pair has also been the first fully functional complex to be characterized, as their interaction not only stabilizes intercellular adhesion but also generates an intracellular signal involved in modulating calcium levels and/or cAMP related pathways (Muller et al., 2015; Li et al., 2018; Vysokov et al., 2018). It is noteworthy that in addition to the presence of alternative splicing, teneurin proteins can also generate c-terminally cleaved products known as teneurin C-terminal Associated Proteins or TCAP, which are capable of regulating events as diverse as metabolism and reproduction but also neuronal morphology (Al Chawaf et al., 2007; Colacci et al., 2015; Hogg et al., 2018). The evidence that TCAP sequences overlap with the proposed latrophilin bindingdomain makes them likely candidates as latrophilin ligands and recent studies suggest that TCAP-mediated effects require a functional interaction with latrophilins (Silva et al., 2011; Husic et al., 2019).

#### Neurexins

This family of type I proteins is expressed by three genes in mammals, each producing two main isoforms: the large isoforms, α-neurexins, and the short isoforms known as β-neurexins (**Figure 1B**). As a consequence of extensive alternative splicing, these molecules present a highly polymorphic profile with the potential to interact with different sets of partners/ligands (Treutlein et al., 2014). The binding of neurexins to latrophilins is strictly regulated by alternative splicing of the former (Boucard et al., 2012). However, despite the fact that all three latrophilin isoforms possess the highly homologous Olfactomedin domain, only latrophilin-1 was shown to establish heterophilic contact with neurexins through that domain to stabilize intercellular adhesion while attempts to demonstrate similar binding for latrophilin-2 and latrophilin-3 have failed (**Figure 1B**; Boucard et al., 2012; O'Sullivan et al., 2014; Zuko et al., 2016). Interestingly, both neurexins and latrophilins have been described as neuronal receptors for α-latrotoxin, a potent component of the black widow spider venom which acts on the presynaptic compartment in order to induce massive neurotransmitter release. In particular, neurexin-1α and latrophilin-1 were thought to account for the majority of binding sites targeted by the neurotoxin in neuronal tissues (Tobaben et al., 2002). It is still unclear how contact between both latrophilin and neurexin leads to neuronal synapse formation but their genetic interdependence in mice brains suggests a yet unknown functional mechanism that warrants further investigation (Tobaben et al., 2002).

# FLRT

Previously known for their role in cell migration, the Fibronectin and Leucine-Rich Transmembrane proteins or FLRT were identified as high-affinity ligands for latrophilins in brain tissues (O'Sullivan et al., 2012). Ubiquitously expressed as 3 isoforms in vertebrates (FLRT1, 2, and 3), most of FLRT functions in neurons have been attributed to adhesion events mediated by homophilic contacts or repulsion events through their heterophilic interaction with Unc5 family of membrane receptors that respond to guidance cues (Yamagishi et al., 2011; Seiradake et al., 2014). Thus, the Unc5-FLRT pair forms a chemorepellent complex while FLRT-FLRT interactions recapitulate an adhesive complex (Karaulanov et al., 2006; Seiradake et al., 2014). However, FLRT-mediated adhesion would prove to not only rely on homophilic binding but also on heterophilic interactions that came into light after latrophilins were identified as potential

FIGURE 1 | Latrophilin-ligands pairings at the mammalian synapse. Representation of a mammalian mature synaptic formation with pre- and post-synaptic compartments schematized. (A–D) Molecular complexes are shown between latrophilins and indicated ligands in dedicated zoomed-in boxes. (E,F) Components of excitatory synapses are shown such as N-methyl-D-aspartate receptor (NMDAR), α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR), PSD95, SHANK, Cortactin, MINT, CASK, voltage-dependent calcium channel α. (D–F) Indicated domains are the following: LNS, laminin, neurexin and sex-hormone binding; EGF, epidermal growth factor; TM, transmembrane; PDZ B.M., PSD95, Dlg, Zona occludens binding domain; Lec, Lectin; Olf, Olfactomedin; S/T, serine-threonine rich; Horm, hormone binding; GAIN, GPCR autoproteolysis inducing; Tox, Toxin; FN/FnIII, fibronectin type III; LRR/LR, Leucine-rich repeats; SH3, Src homology 3; GUK, guanylate kinase; CamK, Ca2+-Calmodulin kinase; PRO, proline rich; SAM, SH3 and multiple ankyrin repeat; PTB, phosphotyrosine binding; ATD, amino-terminal domain; VFT, Venus fly trap; CRD, cysteine rich domain.

partners for FLRT (**Figure 1A**; O'Sullivan et al., 2012). Further characterization of FLRT structure would provide unexpected findings on this newly described interaction. The characterization of both Unc5-FLRT and FLRT-FLRT binding determinants denoted that FLRT leucine-rich repeats constituted the only domain necessary for maintaining the interaction and stabilizing intercellular adhesion or repulsion (**Figure 1A**; Karaulanov et al., 2006; Seiradake et al., 2014). Along the same line, the determinants establishing latrophilin-FLRT interaction were circumscribed by the exact same region which could potentially create a competitive interaction pattern between latrophilin, FLRT and Unc5, a situation that would allow the segregation between two contrasting functions of FLRT because repulsive (FLRT-Unc5) and adhesive (FLRT-latrophilin) functions would compete in order to be mutually exclusive as one would expect (Jackson et al., 2015, 2016; Lu et al., 2015). However, this scenario received counter-evidences following the elucidation of the latrophilin-FLRT-Unc5 crystal structure (Lu et al., 2015; Jackson et al., 2016). Indeed, the three molecules were observed as forming part of the same molecular complex facilitated by determinants of the "arc shaped" FLRT leucine-rich repeats (LRR) region: latrophilin interacted with LRR convex side while Unc5 interacted with LRR concave side, which is consistent with both molecules maintaining a non-competitive binding dynamic due to non-overlapping binding interfaces (Lu et al., 2015; Jackson et al., 2016). In the future, it will be interesting to elucidate which cellular functions are supported by the formation of these super-complexes and how these interactions are regulated in synaptogenesis events.

#### Contactins

As part of the immunoglobulin cell adhesion molecules, contactins extracellular region consists of immunoglobulinlike and fibronectin repeats. Although contactins lack a transmembrane domain, they are linked to the cell surface via a GPI anchor, a feature that allows them to restrict their cellular localization to the cell membrane but that impedes them from autonomously initiating intracellular signaling. Contactins have been shown to form molecular complexes with various transmembrane proteins therefore allowing them to provide a complement to their signaling function due to the ability of the newly formed complexes to interact with cytoplasmic signaling cascades. Out of the six isoforms of contactins found in vertebrates (contactin-1,-2,-3,-4,-5, and -6), only contactin-6 was identified as a latrophilin ligand (**Figure 1C**). In contrast to previously described ligands, contactin-6 was unable to mediate trans-cellular adhesion through its contact with latrophilin-1 (Zuko et al., 2016). Instead of a contact in trans (between two separate cell membranes) a cis-configuration was the preferred description for this molecule pairing at the cell membrane. Indeed, not only were contactin-6 and latrophilin-1 expressed in the same neuronal cells (cortical neurons) but the effect of the molecular complex on apoptosis pathways and neuronal morphology was only appreciable in a cell-autonomous fashion (Zuko et al., 2016). While the protein domains involved in the stabilization of the contactin-6/latrophilin-1 complex are unknown, the question regarding their function in the adhesive properties of the cell remains open.

# LATROPHILINS: ESTABLISHING NEURONAL CONNECTIONS

# Growth Cones Formation

In a developing neuronal network, the purpose of neuronal migration is presumably to help find the adequate partnering cell. This migration is facilitated by the elongation of axonal structures driven by extending microtubules onto which actin structures contribute to increasing the surface contact with the surroundings (**Figures 2A–C**). Such neuronal specializations that are represented by the formation of growth cones provide polarity and directionality to this active exploration/migration process (Rich and Terman, 2018). The dynamic nature of growth cones and their ability to respond quickly to ever changing environmental cues allows for a highly accurate target recognition process to occur, thus leading to precise interneuronal contacts. Molecular determinants that guide the formation and migration of growth cones have been identified, thus providing the initial description of how environmental cues can instruct migration patterns by engaging the cytoskeleton to induce movement (Lowery and Van Vactor, 2009). Among such molecules, the Ephrins and their receptors Ephs are probably the best described. A complex network of membrane-attached Ephrins and Ephs establish a chemical gradient throughout which neurons appendices will physically progress until the right target is found (Xu and Henkemeyer, 2012). Thus, the notion that migration cues should establish a molecular gradient in order for the neuronal protrusion to follow its course has permeated our knowledge of the molecular basis of growth cones formation and function. However, the vast diversity of neuron types suggests the existence of an equally diverse set of guidance cues and receptors.

Proteins that assist the formation of growth cones have been described using proteomic assays aimed at detecting proteins that are differentially distributed along growth cones versus the ones present in axons. Such assays revealed an enrichment of latrophilin-3 (Lphn3) at the tip of migration. Concurrently, neurexin1 followed the same expression pattern as for Lphn3, whereas teneurin-2 was equally distributed along both structures (Nozumi et al., 2009). Enrichment of Lphn3 at these growth cones was accompanied by actin remodeling proteins such as cofilin, and proteins from neurotransmitter vesicles release machinery such as Munc18, Snap25 or Synaptotagmin, thus reinforcing the subcellular localization of latrophilins as presynaptic proteins.

Cementing the role of latrophilin in growth cone migration, a study conducted by Vysokov et al. (2018) evidenced the importance of the right Lphn/ligand pairing for providing the instructional signals to achieve axonal elongation and directionality (**Figure 2C**; Vysokov et al., 2018). Indeed, the group showed that hippocampal neurons responded to a gradient of a secreted splice variant of teneurin-2 by sending a higher number of axons toward the established gradient than in control conditions not exposed to soluble teneurin-2. This effect was greatly dependent on Lphn1 expression as Lphn1-deficient

FIGURE 2 | Signaling pathways underlying the potential involvement of latrophilins in growth cone and actin structures formation. Representation of growth cone structures and associated signaling. (A) Lamellipodia Formation: GPCR activation leads to their coupling with G proteins provoking subunits dissociation . Activation of the Rac pathway by Gβγ subunits results in the recruitment of WAVE and ARP2/3 complexes at the front of migration and generates actin polymerization (Lowery and Van Vactor, 2009; Chia et al., 2013). The reported interaction between Lphn1 and SHANK could presumptively couple the aGPCR to the actin cytoskeleton (Tobaben et al., 2000). (B) Filipodia Formation: Activation of G proteins by GPCR stimulation at the migration front leads to dissociation of α subunits from βγ subunits, which will in turn activate small GTPase Cdc42 recruiting n-WASP and ARP2/3 complexes and provoking actin polymerization (Lowery and Van Vactor, 2009). (C) Axonic Cone Formation: Proteolytically cleaved teneurin-2 activates latrophilin leading to Gα<sup>q</sup> protein induction of PLC followed by an increase in Ca2<sup>+</sup> release from the endoplasmic reticulum through IP<sup>3</sup> receptors (Vysokov et al., 2018). Alternatively, cAMP levels can be modulated by activation of Gαi or Gαs proteins (yellow) (Muller et al., 2015; Nazarko et al., 2018). In parallel, G protein activation can also lead to the stimulation RhoA/ROCK pathway supporting filopodial formation through actin stabilization (green) (Siehler, 2009).

neurons failed to respond to such teneurin-2 gradient, thus suggesting that latrophilin and teneurin heterophilic contacts support growth cone formation (Vysokov et al., 2018). In support of these observations, a functional teneurin-1 deficiency in C. elegans revealed neuronal pathfinding defects in pharyngeal neurons development, a process that seems to be engaging components of both the extracellular matrix and of the actin cytoskeleton which constitute important elements of axonguidance events (Drabikowski et al., 2005; Morck et al., 2010). However, studies of hippocampal neurons from teneurin-3 deficient mice have provided contrasting evidences that rather point to the importance of a splicing-dependent homophilic teneurin-teneurin contact in instructing neuronal wiring (Berns et al., 2018). It will be interesting to witness how these paradigms will be reconciled in the future as they might reveal unknown mechanisms of action for this cooperating pair of adhesion molecules.

**Latrophilins and Modulation of Actin Cytoskeleton Elements** Whether cell migration events require Lphn/teneurin or teneurin/teneurin interactions, evidences highlight the possible involvement of these molecules in reshaping the cell cytoskeleton. This remodeling is essential for allowing the formation or retraction of contact structures such as filopodia and lamellipodia, actin-rich protrusions that increase the surface contact with the supporting matrix to yield a more efficient exploration pattern. Latrophilins have been reported to interact with intracellular scaffolding proteins known to be associated with the actin cytoskeleton but the functionality of such interaction remained elusive (**Figures 2A–C**). Recent data from our lab monitoring the formation of actin-rich structures evidenced an active role for latrophilins in regulating the formation of filopodia and lamellipodia (**Figures 2A,B**; Cruz-Ortega and Boucard, 2019). While all isoforms of latrophilins led to a constitutive activation of cofilin, which is an important modulator of actin rearrangement, isoform-specific functions were detected in the genesis of cell protrusion in response to teneurin binding (Cruz-Ortega and Boucard, 2019). Importantly, teneurin signals removed Lphn-induced inhibition on cell protrusions formation leading to an increase in filopodia. Interestingly, teneurin C-terminal peptides have been shown to activate small GTPases that contribute directly to the formation of actin structures and to act through latrophilins to modify actin dynamics (Chand et al., 2012; Husic et al., 2019). Thus, we hypothesize that latrophilins provide a framework for the establishment of adhesion structures by interacting with the actin cytoskeleton machinery (**Figures 2A,B**), a role that might precede the formation of adhesion complexes at the synapse. Molecular adhesion events via latrophilin-teneurin interactions would therefore act as permissive signals allowing intercellular contacts to establish a given adhesive structure.

# Latrophilins and Ligands: Molecular Aspects Involved in Synapse Formation and Function

The immense diversity of neuronal connections begs for a molecular code that can sustain such a high level of heterogeneity.

Because adhesive properties of neurons are mainly embodied by cell adhesion molecules (CAMs), the biological support for heterogeneity should reflect an array of adhesion profiles supported by distinct sets of CAMs. However, it becomes clear that the shear number of adhesion molecule genes cannot by itself explain the diversity seen in neuronal connections. Thus, it is conceivable that the spatio-temporal patterns of established neuronal circuitry would be sculpted by the three following factors: (a) the types and forms of adhesion molecules expressed, (b) the net synaptic content of adhesion molecules, and (c) the pairing pattern of these adhesion molecules across synapses, in a given time frame throughout development. Thus, the genetic framework of neurons would have to provide the required information to produce a diverse array of proteins which can then carry the system's heterogeneity on their shoulders. We will discuss how neuronal networks benefit from latrophilin and teneurin isoforms heterogeneity to generate adhesion complexes that are both diverse and hierarchical in nature.

#### Domain Modularity

Latrophilin has adjacent Lectin-like and Olfactomedin-like extracellular adhesion motifs separated by a short linker sequence which can be found inserted or absent from the translated protein as a result of alternative splicing of the corresponding transcribed mRNA. Thus, both these motifs are physically independent from each other and consequently, each domain is available to function independently with respect to their adhesion function. Indeed, the Lectin-like domain has been identified as the main interaction motif with teneurins, as its presence is absolutely necessary for latrophilin-1 to establish intermolecular complexes with teneurins (Boucard et al., 2014). On the other hand, it is completely dispensable when it comes to latrophilin-1 interacting with Neurexins or all latrophilins interacting with FLRT proteins. Conversely, the Olfactomedin-like domain represents the main interaction site for Neurexin and FLRT but is not essential for teneurins' contact with latrophilin-1 as it serves modulatory purposes in this case by increasing affinity for the ligandreceptor pair. Important insights on ligand-receptor interaction were obtained from the first crystallographic determinations of a Lphn-ligand complex structure (Jackson et al., 2015, 2016; Lu et al., 2015). The Lphn-FLRT crystallographic complex revealed that the Lphn Olfactomedin-like domain forms a "rosette-like" structure of which the open face is engulfed within the concave face of FLRT LRR (Leucine Rich Repeats) horseshoe domain. Importantly, this domain is sufficient and necessary to form the required interaction with FLRT thus indicating that it can act in a modular fashion by restricting ligand binding to this region alone, leaving the other adhesion domain free to establish additional contacts. This interaction model is further supported by a recent study evidencing that latrophilins can form simultaneous complexes with FLRT and teneurins to generate different synaptic functions (Sando et al., 2019).

#### Alternative Splicing

As a strategy to generate a high order of multiplicity in interneuronal contacts, adhesion molecules expressed in the nervous system display multiple variants originating from alternative splicing. The physiological importance of splicing events for neuronal functions such as synapse identity or maturation is best exemplified by Neurexins, Down Syndrome Cell Adhesion Molecule (DSCAM) and protocadherins which can produce from 3,000 to 40,000 variants, each with potentially different binding/adhesion functions (Wu et al., 2012; Treutlein et al., 2014; Kostadinov and Sanes, 2015). Latrophilin mRNAs are prone to multiple events of alternative splicing that show a certain level of heterogeneity between isoforms (Sugita et al., 1998; Matsushita et al., 1999). These splicing events create receptor variants that differ in their extracellular and/or their intracellular regions (**Figure 3**).

#### **Extracellular splice inserts**

The splicing pattern of latrophilins paints a complex portrait depicting isoforms with alternative initiation sites and intraexonic splicing events (**Figure 3**). To this date, the impact on latrophilins of all extracellular splicing events are unknown except for one designated as splice site A (SSA). SSA is located in a region corresponding to the N-terminal extracellular portion of latrophilins, is common to all mammalian latrophilin isoforms and introduces or removes a 4–5 amino acid sequence between the Lectin and Olfactomedin domains (Sugita et al., 1998; Boucard et al., 2014). Lphn2 SSA variants display a slight variation from Lphn1 and Lphn3 SSA since they can include two variations of this insert, one that is identical to the other isoforms SSAs and another shorter form that differ in its N-terminal residues (Boucard et al., 2014). Splicing at SSA has been shown to modulate Lphn1-teneurin2/Lphn1-teneurin4 interactions such that the presence of this insert decreases their binding affinity (Boucard et al., 2014). Interestingly, this splicing-dependent modulation of affinity is specific to the Lphn1-teneurin pairs as Lphn1-Neurexin/FLRT pairings did not display a significant change to their binding properties whether the SSA insert was present or not. The function of the additional latrophilin-2 and -3 extracellular splicing events are unknown but it is likely that they may also contribute to ligand selection or receptor activation paradigms by stabilizing different conformations (**Figure 3**).

#### **Intracellular splice inserts**

The splicing events affecting the intracellular portions of latrophilins describes a rather complex pattern. In contrast to SSA splicing, a partial overlap has been observed between Lphn2 and Lphn3 variants while most Lphn1 splicing variants are unique to this isoform. The Lphn1 intracellular splicing site B, SSB, inserts or removes a 45 amino acid domain in the C-terminal tail of mouse Lphn1 but has not been detected in human Lphn1 (**Figure 3A**); Lphn2 and 3 are modified in their third intracellular loop and in their C-terminal tail at a site different from Lphn1 SSB and which displays a tandem splicing pattern (**Figures 3A– C**; Sugita et al., 1998). The function of these splicing events is still elusive; however, a recent study suggests that they could have a role in modulating intracellular signaling pathways such as functional coupling to G proteins (Rothe et al., 2019).

#### **Regulation of alternative splicing**

How these splicing events are regulated is not understood to this date. It is likely that various splicing factors are involved in this

FIGURE 3 | Alternative splicing events for Lphn1, 2 and 3 in Homo sapiens and Mus musculus. The potential alternative splicing events for Lphn1 (A), Lphn2 (B) and Lphn 3 (C) are shown for H. sapiens and M. musculus. Each red box represents a possible splicing event. Green-red double colored boxes indicate splicing events which are only used to introduce a translation start site and thus cannot be combined within the same isoform, while the added asterisk indicates that these splicing events can alternatively be included as a continuous protein sequence of larger isoforms and can be combined with others of the same color code. The yellow box represents splicing events that are exclusively only present in the short isoforms lacking the Lectin domain. The pink box does not indicate a splicing event but represents the site of a potential alternative promoter sequence within a universally used exon. The Open Reading Frame Change (ORFc) indicates a splicing event present in the Lphn1 of H. Sapiens that alters the reading frame to generate a translation start site. Blue-red double colored boxes represent mutually exclusive splicing events (in the case of Lphn2 and Lphn3) two splicing sites with the same start sequence but which are carried out in different isoforms and are mutually excluding. The dotted lines indicate the common events between isoforms and species. The legend at the bottom indicates the possible consequences of the splicing events. Lec, Lectin domain; Olf, Olfactomedin domain; Horm, Hormone binding domain; GAIN, GPCR Autoproteolytic Inducing domain; GPS, GPCR proteolytic site; 7TM, Seven transmembrane domain; PDZ BD, PDZ binding domain; ICL1, ICL2, ICL3: Intracellular loop 1,2,3. ADGRL1,2,3, Adhesion G Protein-Coupled Receptor Latrophilin-1,2,3. Data were extracted from the NCBI database (www.ncbi.nlm.nih.gov) as well as the Ensembl database (www.ensembl.org). Primary assemblies for Homo sapiens and Mus musculus: GRCh38.p12 and GRCm38.p4. ADGRL1 Homo sapiens (Gene ID: 22859; NC\_000019.10 and ENSRNOG00000072071); ADGRL1 Mus musculus (Gene ID: 330814; NC\_000074.6 and ENSRNOG00000072071); ADGRL2 Homo sapiens (gene ID: 23266; NC\_000001.11 and ENSG00000117114); ADGRL2 Mus musculus (Gene ID: 99633; NC\_000069.6 and ENSMUSG00000028184); ADGRL3 Homo sapiens (Gene ID:23284; NC\_000004.12 and ENSG00000150471); ADGRL3 Mus musculus (Gene ID: 319387; NC\_000071.6 and ENSMUSG00000037605) (National Center for Biotechnology Information, 1988; Zerbino et al., 2018).

process given the complexity of the events observed. No splicing factors have been identified so far but evidences suggest that these events of alternative splicing are highly regulated. Indeed, while the relative expression of receptor variants resulting from mRNA splicing does not seem to vary according to different development stages in a given tissue, it differs between tissues at a given developmental stage. For example, the insert in SSA that is unique to Lphn2 is inserted in 60% of transcripts from the brain but is almost inexistent in transcripts from heart tissues of adult mice (Boucard et al., 2014). Moreover, the respective proportion of receptor variants in a given tissue varies between isoforms: the main Lphn3 isoform in brain does not contain SSA insert while Lphn1 SSA is present in approximately 50% of brain transcripts (Boucard et al., 2014). Thus, cell environments might be the dominating factor in determining which splicing variants of latrophilin will be generated.

#### **Alternatively spliced latrophilin ligands**

The molecular counterparts of latrophilins also exhibit alternative splicing that affects specificity of interaction with these adhesion GPCRs.

− Teneurins can be spliced in two extracellular sites: one within the EGF repeats region and another in the β-propeller region. The teneurin splice variant containing an insert in the β-propeller site loses its ability to form intercellular adhesion complexes through latrophilins, an effect that could be attributed

to structural rearrangements rather than direct perturbation at the binding interface because this site is remote from where the binding occurs with latrophilins (Silva et al., 2011; Li et al., 2018). Interestingly, this variant of teneurin which is deficient in latrophilin binding loses the ability to induce post-synaptic excitatory specializations, instead it stabilizes inhibitory synapses, which suggests that latrophilins might not be the exclusive binding partners of teneurins. Indeed, Berns et al. (2018) described a strict homophilic interaction between teneurin-3 splice variants which specifically involve the variant which cannot bind to latrophilins (includes the splice insert in β-propeller site).

− Neurexins form a family of highly polymorphic proteins due to alternative splicing involving 5 sites for α-neurexin (SS1– SS5) and two for β-neurexins (SS4–SS5) (Treutlein et al., 2014). The binding of neurexins to their canonical ligand neuroligin is partly regulated by neurexin splicing at SS4 such that presence of an insert in this site decreases its affinity for neuroligin (Boucard et al., 2005). A similar pattern of interaction was observed for neurexin binding to latrophilin-1 although resulting in a complete abrogation when SS4 was present (Boucard et al., 2012). Importantly, both latrophilin-1 and neuroligin compete for the same binding pocket on SS4-deficient neurexin which makes them mutually exclusive in eventual molecular complex formation centered in Neurexins. This binding characteristic might explain why we and others were not successful in our attempts to directly isolate latrophilin-Neurexin complexes from brain extracts which contain high amounts of neuroligins (Boucard et al., 2005, 2014; Silva et al., 2011).

# CIS- VERSUS TRANS-INTERACTIONS: RELEVANCE FOR LIGAND-DEPENDENT LATROPHILINS' FUNCTION

Inter-neuronal adhesion functions of latrophilins are primarily thought to occur via interactions in trans, i.e., latrophilins from one neuron interact with ligands expressed in another neuron. This interaction model infers that latrophilins would be restricted to one synaptic compartment in order to link another synaptic compartment displaying its ligands at the cell surface, thus fulfilling their role as de facto adhesion pairs. However, is it possible for latrophilins to participate in adhesion if they form complexes in cis, i.e., with adhesion molecules expressed in the same cells. We address three questions: is latrophilin pre- or postsynaptic? on which synaptic compartment are the latrophilin ligands present? which of these two types of interaction (cis versus trans) is functional?

The answer to the first question as to if latrophilins are pre- or post- synaptic relies on the following evidence: (1) The presynaptic neurotransmitter release machinery is activated when α-latrotoxin acts through latrophilin; (2) Electron microscopy with immunodetection of latrophilin's extracellular domain detected an enrichment in the pre-synaptic membrane (Silva et al., 2011); (3) Growth cones, which can be conceptually seen as immature presynaptic structures, respond to latrophilin ligand teneurin to follow their course and acquire directionality (Vysokov et al., 2018); (4) A presynaptic phenotype was observed when knocking down Lphn isoforms (O'Sullivan et al., 2012). On the other hand, there is also evidence for post-synaptic localization of latrophilin: (1) Latrophilins can form a complex with proteins from the SHANK family predominantly expressed in the postsynaptic compartments of excitatory synapses (Kreienkamp et al., 2000; Tobaben et al., 2000); (2) Conditionally expressed Lphn2 and 3 fusion proteins colocalize with postsynaptic markers in the hippocampus (Sando et al., 2019); (3) Mouse models of Lphn2 and 3 deficiency display postsynaptic phenotypes in neurons of the hippocampus (Anderson et al., 2017; Sando et al., 2019). It is unclear whether Lphn are enriched in a given synaptic compartment, however, it appears conceivable that these receptors would be present in both, perhaps depending on the given developmental stages or neuronal types.

The localization of latrophilin ligands (question 2) can be observed in both synaptic compartments. In mammals, teneurins are thought to participate in homophilic binding from both sides of the synapse, they are present in growth cones and shape neuronal circuits through axon guidance mechanisms (Nozumi et al., 2009; Young et al., 2013; Tran et al., 2015; Berns et al., 2018). In C. elegans, the neuromuscular expression of latrophilin (lat-1) and teneurin (ten-1) revealed a pattern of partly overlapping and complementary labeling suggesting both cis and trans configurations in the pharyngeal system with muscle/neuron lat-1/ten-1 complementarity along with neuron/neuron lat-1/ten-1 overlap (**Figure 4**). Non-neuronal systems from C. elegans suggest a trans configuration as exemplified by the epidermoblast stage. In this organism the establishment of an anterior (a) posterior (p) axis is indispensable for cell polarity and future cell divisions. At the fourth division the Ca and Cp cells are generated according to their position in the anteriorposterior axis, respectively. In accordance with the above, the eigth division gives rises to Cpaaaa cells expressing ten-1 which are surrounded by Caaa lineage cells expressing lat-1, thus depicting a trans configuration. However, at this cell division stage, a cis configuration is also supported giving the concomitant expression of both proteins in Caaa lineages (**Figure 4D**; Langenhan et al., 2009; Promel et al., 2012). In Drosophilia, teneurin orthologs participate in axonal pathfinding thus evidencing their role in both sides of contacting membranes (Hong et al., 2012; Mosca et al., 2012). In an attempt to probe the Drosophila teneurin (Ten-m) expression pattern, we conducted experiments using a promoter enhancer trap that allowed for the visualization of enhanced yellow fluorescent protein (eYFP) driven from the Ten-m promoter, relying on the Gal4-UAS system (**Figure 5**; Hacker et al., 2003; Horn et al., 2003). We observed a complementary expression pattern with the reported Drosophila latrophilin (dCirl) expression at the larval stage in the chordotonal organ (**Figure 5**). Indeed, while dCirl has been reported to be expressed in chordotonal neurons (Scholz et al., 2015), our analysis revealed that Ten-m was expressed in the adjacent scolopale cells, thus suggesting a trans configuration (**Figure 5**). Moreover, the expression of Ten-m in the optic lobe of the adult Drosophila brain was detected in photoreceptor neurons that project to the medulla where dCirl-expressing neurons have been identified (**Figure 5**; Gehring, 2014). Although

the Ten-m/dCirl interaction in trans has yet to be reported in Drosophila, our data point to a complementary pattern of expression that is suggestive of a trans configuration in this organism's sensory organs.

As the possibility of many configurations of interactions seems more than likely, we are left with the third issue, which is bound to capture further attention: can the functionality of cis interactions rival the one from trans interactions? Since latrophilins and their ligands are transmembrane proteins (except for contactin-6), each can elicit an intracellular signal independently of their respective partnership. On one hand, the hypothesis of trans configuration calls for a system that segregates anterograde and retrograde signaling in different cells, thus generating a "one-ligand-one-signal" environment in regards to individual contacting cells. On the other hand, the cis configuration has the potential to create intersecting signaling pathways. Evaluating the functionality of both configurations, Li et al. (2018) observed that teneurin-2 was capable of inducing a similar decrease in cAMP accumulation in cells expressing Lphn1 and Lphn3 whether expressed in the same cell or on different cells. Because latrophilins are known to couple to G proteins (Li et al., 2018; Rothe et al., 2019) from which the β/γ subunits, once dissociated from the α subunit, can in turn activate the MAPK pathway, it remains to be seen if the teneurin-dependent activation of the FAK pathway would affect MAPK signaling differently in a cis versus a trans configuration (Suzuki et al., 2012, 2014). The functional impact of these configurations will have to be assessed in future studies in order to grasp the full understanding of their physiological or pathophysiological implications and to test whether the functional considerations we propose as a hypothesis, are valid.

# LATROPHILINS AND SYSTEMIC FUNCTIONS IN MODEL ORGANISMS

The physiological functions of latrophilins have been investigated in multiple organisms. The accumulating data point to an evolutionary conserved role while simultaneously demonstrating divergences. We will detail observations emanating from the study of latrophilin deficient animals in order to highlight overlapping as well as non-overlapping latrophilin functions.

#### Latrophilins in the Nematode Worms

The two known orthologs of latrophilins in Caenorabditis elegans were named lat-1 and lat-2 (Mee et al., 2004; Willson et al., 2004). These proteins lack an olfactomedin domain at their amino terminal end, which differentiates them from their mammalian orthologs (Mee et al., 2004; Willson et al., 2004). During the early stages of the nematode's life cycle, lat-1 is expressed in the gonads during oogenesis, in blastomeres (especially in those derived from the AB lineage) and pharyngeal and epidermal precursors (**Figure 4**; Langenhan et al., 2009). During the larval and adult stage, its expression has been reported in the vulva, plasma membrane of pharyngeal cells, in neurons of the terminal bulb and the corpus (with projections inside the isthmus) and in neurons of the nervous ring (Willson et al., 2004; Langenhan et al., 2009; Promel et al., 2012). On the other hand, the expression of lat-2 overlaps with lat-1 in the pharyngeal primordium during the early stages and is limited to cells of the excretory and pharyngeal system in the larval and adult stages (**Figure 4**; Langenhan et al., 2009). lat-1, but not lat-2, is required for proper development during the early stages of embryogenesis, specifically for regulating the alignment of the anterior and posterior planes during the fourth round of cell division through its coupling with Gα<sup>S</sup> proteins. This coupling fostered the activation of adenylate cyclase, increasing the intracellular levels of cAMP in wild-type embryos, which were decreased in lat-1 knockout embryos (Langenhan et al., 2009; Promel et al., 2012; Muller et al., 2015).

The C. elegans pharynx is a neuromuscular feeding organ that is related to the transport of food from the mouth to the intestine through pharyngeal pumps and isthmus peristalsis (Albertson and Thomson, 1976; Trojanowski et al., 2016). These relaxationcontraction cycles are regulated in part by neurotransmitters such as acetylcholine, serotonin and glutamate from neurons of the pharyngeal and extra pharyngeal nervous system, and by myogenic activity (Bhatla et al., 2015; Trojanowski et al., 2016). The main motoneurons that regulate pumping are the cholinergic neurons MCs and glutamatergic M3 neurons, which connect with pm4 muscle cells of the metacorpus (**Figure 4**). When food is present in the environment, neurosecretory motoneurons (NSM) start secreting serotonin to activate MCs and M3 which in turn release acetylcholine and glutamate, respectively, on cells of the pharyngeal muscle, thus regulating the duration of the food intake circuit. Ablation of MCs and M3 neurons led to a decrease in the number of pharyngeal contractions and interestingly, so did the lat-1 knockout and knockdown models (Avery and Horvitz, 1989; Willson et al., 2004). In order to investigate how lat-1 expression in pharyngeal cells and nearby neurons relate to the neural network that regulates pumping during food intake, lat-1 knockout worms were treated with the serotonin reuptake inhibitor imipramine or the anthelmintic emodepside acting at the neuromuscular junction, observing a resistance to their effect compared to wild-type worms (Mee et al., 2004; Willson et al., 2004). Thus, serotonin and acetylcholine may mediate lat-1 function in worms. However, another question arises: could lat-1 adhesion function from the pharyngeal muscle be completed by another adhesion molecule located in pharyngeal neurons? Teneurins come to mind as potential candidates giving their similarities to their mammalian orthologs. In C. elegans, a single gene has been reported that transcribes two isoforms of teneurin: teneurin 1-L (large) and 1-S (small, because it lacks the intracellular domain). While both isoforms share overlapping expression profiles in the nervous system from embryonic stages to the adult stage, teneurin 1-L is also expressed in intestinal cells and the pharynx (**Figure 4**). Because neurons that express teneurin 1-L are part of the circuit that regulates the pharyngeal pumping (M1–M4, I3, and NSM), it is tempting to speculate that a lat-1/ten1 complex could be mediating neuromuscular functions related to pharyngeal pumping.

pharyngeal nervous system (1, Corpus; 2, Extra-pharyngeal neurons; 3, Neuron in the terminal bulb; 4, Intestine; 5, Pharyngeal neurons; 6, Neuron in the corpus with projections into the isthmus; 7, Isthmus; 8, Terminal bulb; 9, Pharyngeal-intestinal valve) (Willson et al., 2004; Langenhan et al., 2009); (C) Expression of lat-1 in muscle cell membrane of the pharynx (Promel et al., 2012); (D) Representation of epidermoblast during dorsal intercalation. ten-1 expressed in Cpaaaa and Caaa lineages, while lat-1 expression is restricted to Caaa lineages (Promel et al., 2012); (E) Expression of lat-1 in excretory cells and neurons of the nerve ring (1, Sensory dentrites; 2, Nerve ring; 3, Dorsal nerve cord; 4, Excretory cells; 5, Ventral nerve cord) (Willson et al., 2004; Langenhan et al., 2009; Promel et al., 2012); (F) representation of the expression of lat-1 in gonads of an adult hermaphrodite (Langenhan et al., 2009; Promel et al., 2012). Pm: Pharyngeal muscle, NSMs: neurosecretory motor sensory. Images were adapted with the permission of Wormatlas (www.wormatlas.org).

# Latrophilin in Flies

The dipteran fly Drosophila melanogaster is an accessible model organism for scientific research, arguably the multicellular organism understood in most detail. Genetic, cellular and molecular tools have been developed in over a century of continuous genetic and biological research in this model (Yamaguchi and Yoshida, 2018). Its genome is comparatively small, however, many genes, as well as principles and mechanisms of development, are evolutionarily conserved in vertebrates (Adams et al., 2000). Drosophila neurons and glia are no exception and share many molecular and functional characteristics with the related cell types in mammals (Venkatasubramanian and Mann, 2019; Yildirim et al., 2019). Neuronal axons have all the machinery necessary to transmit nerve impulses in a similar way to how action potentials are generated in mammals leading to neurotransmitter release at the synapses (Jekely et al., 2018; Kasture et al., 2018; Rich and Terman, 2018; Sugie et al., 2018). Hence, Drosophila offers a good model to study neuronal proteins (Monnier et al., 2018; Rosas-Arellano et al., 2018). Drosophila melanogaster only has a single homolog of latrophilins (Scholz et al., 2015). The single homolog of latrophilins in this species is known as dCirl, expressed during the larval stage in peripheral sensory neurons, including those of the pentascolopidial chordotonal organs (lch5), and in the ventral nerve cord (**Figure 5**; Scholz et al., 2015). dCirl shows a strong expression pattern in the dendritic membrane and the single cilium of chordotonal (ChO) neurons of lch5. The mature lch5s are composed of multicellular units called scolopidia; each unit consists of three bipolar neurons and support cells. The distal segment of each dendrite in these neurons ends in a cilium that is protected by the supporting scolopale cell. The lch5 is in charge of mediating the mechanosensation process by means of which the mechanical stimuli such as touch, hearing and mechanical deformation of the larval body during the locomotion induce the movement of the ciliated dendrites causing the opening of cationic channels and an inrush of K<sup>+</sup> leading to a neuronal depolarization that is translated into neuronal impulses (Prahlad et al., 2017).

A null dCirl mutation demonstrated that this gene is not essential for development and viability. The mutant organisms did, however, manifest obvious alterations in their sensory organs: the structures most affected were the lch5s (Scholz et al., 2015). The dCirl knockout larvae showed abnormal behaviors, such as a conspicuous crawling pattern and traveling less distance than control larvae. These results suggested that dCirl participates in shaping locomotion. In addition, the dCirl mutants showed diminished touch sensitivity, as well as a reduction in mechanosensory responses after mechanical stimuli. All these alterations were corrected after the re-expression of dCirl in the ChO neurons, which showed that the observed effects were due specifically to the loss of dCirl. On the other hand, the morphology of ChO neurons was not altered after the removal

of dCirl, suggesting that dCirl plays a functional rather than a structural role in these neurons, being required for adequate sensitivity gentle touch, sound and proprioceptive feedback during larval locomotion (Scholz et al., 2015). The mechanism by which dCirl is thought to relay mechanotransduction was investigated through genetic interaction assays which revealed that two subunits from Transient Receptor Potential (TRP) channel, TRPN1/NompC and TRPV/Nanchung (Kim et al., 2003; Cheng et al., 2010), could mediate its function possibly by providing the ion flux necessary for the decoding of the mechanical strain by generating a receptor potential in mechanosensory neurons (Scholz et al., 2015).

An interesting observation was made when intracellular signaling of dCirl was investigated using a FRET- based cAMP sensor. Mechanostimulation of dCirl decreased the concentration of cAMP in mechanosensory neurons (Scholz et al., 2017). dCirl-deficient flies did not display a reduction in cAMP upon mechanostimulation and consequently experienced a quenching of neuronal activity. These observations suggest that dCirl modulates neuronal activity by suppressing cAMP production, a signaling feature that reveals a stark contrast with lat-1 signaling in C. elegans (see section "Latrophilin in the Nematode Worms"). This difference could be accounted for if dCirl and lat-1 are coupled to distinct Gα subunits. Conversely, the deficiency observed in dCirl-deficient flies could be rescued by pharmacological inhibition of adenylate cyclase (Scholz et al., 2017). The role of dCirl in the lch5 of Drosophila as a mediator of mechanosensation represents a novel function for this family of receptors and highlights the importance of conformational changes for its ability to trigger intracellular signaling cascades, a feature resembling canonical activation mechanisms of members of the GPCR family (Oldham and Hamm, 2008).

In the adult brain dCirl expression was observed in the medulla of the optic system and in the mushroom bodies, the latter of great importance for olfactory learning and memory in Drosophila (**Figure 5**; Gehring, 2014; Guven-Ozkan and Davis, 2014; Hige, 2018). The latrophilins have been associated with various neuropsychiatric diseases, among those Attention-Deficit Hyperactivity Disorder (ADHD) has received particular attention, since multiple studies associate the Lphn3 gene with the etiology of the disease (discussed below) (Arcos-Burgos et al., 2010; Domene et al., 2011; Ribases et al., 2011; Jain et al., 2012; Labbe et al., 2012; Fallgatter et al., 2013; Acosta et al., 2016; Kappel et al., 2017; Huang et al., 2018). A conditional dCirl knockdown model based on RNA interference was generated in Drosophila (van der Voet et al., 2016). Neuronal-specific decrease in dCirl expression induced hyperactivity and reduced average sleep time during the night (dark) phase providing evidence that the Dopamine-related paradigm for latrophilin function is also conserved in Drosophila.

As mentioned, a potential endogenous ligand for Drosophila latrophilin is the teneurin Ten-m, a transmembrane protein with a documented role in synapse formation in this organism (Mosca and Luo, 2014). Using a reporter line, we assessed the expression of Ten-m in the Drosophila larva (**Figure 5**). Expression was observed in the bolwig organ (the larval eye) neuron, projecting into the optic lobes of the brain and in the lch5 organs, but, in contrast to dCirl expression in the neurons, Ten-m is observed in the surrounding scolopate cells (**Figure 5** and **Supplementary Figure S1**). In the adult fly, we detected Ten-m promoter expression in the brain optic lobes in a pattern that stopped at

the medulla right where the expression of dCirl was reported (**Figure 5**). These results suggest the possibility of a trans-synaptic contact between dCirl and Ten-m in Drosophila, which remains to be tested experimentally.

# Latrophilin in Zebrafish

Danio rerio, commonly known as zebrafish, is one of the model organisms for vertebrates used in scientific research, particularly in studies that address different aspects of neurogenesis. Among the advantages offered by this organism are that they have an external development accessible to experimental manipulation, as well as a rapid development of the larval nervous system, which is established within 4 days of development. Additionally, the zebrafish has also been used for behavioral studies, since it is a diurnal and naturally sociable animal that shows preferences for community life. Currently a large amount of genetic and anatomical information of zebrafish is available in databases, which facilitates studies using this organism as a model (Kuwada, 1995; Norton and Bally-Cuif, 2010; Schmidt et al., 2013).

Due to its characteristics the zebrafish was used to study the development and function of latrophilins. The zebrafish has two orthologs for the isoform 3 of the latrophilins, which are called Lphn3.1 and Lphn3.2. Both orthologs present a similar expression profile during development. As the larval maturation progresses, Lphn3.1 and Lphn3.2 display a shared expression pattern becoming more prominent in the ventral part of the telencephalon and diencephalon, in the posterior brain and in the ventral area of the spine (**Figure 6**). In the brain of adult zebrafish the expression of Lphn3.1 is detected along the telencephalic midline, as well as in lower levels in the telencephalic parenchyma, the anterior thalamus, the periacal ductal gray matter, the superior nucleus of raphe, the periventricular nucleus of the inferior hypothalamus, the cerebellum and the nucleus of the medial longitudinal fascicle. Because Lphn3.1 expression profile coincided with the expression of its murine ortholog, this receptor<sup>0</sup> s function in zebrafish was further investigated (Lange et al., 2012). Lphn3.1 knockdown morphants increased their swimming distances and displayed hyperactivity, a phenotype that has been associated with the dysfunction of Lphn3 gene in humans affected by ADHD. However, careful considerations should be taken when comparing different organisms because of existing differences in neuroanatomy and circuit formation (Lange et al., 2012; Akutagava-Martins et al., 2016). Such variations are well exemplified in the dopaminergic system of zebrafish. In this model organism, the major dopamine (DA) regions are olfactory bulb, preoptic region, pretectum, posterior tuberculum and hypothalamus. This pattern differs with mammals mainly because no DA neurons are found in the mesencephalon of the zebrafish (Schweitzer and Driever, 2009). Among other things, DA helps to regulate movement, which was altered in Lphn3.1 morphants, leading to the hypothesis that reduction of Lphn3.1 expression influences the dopaminergic system in some way. However, when the concentrations of DA and its metabolite 3,4-dihydroxyphenylacetic acid were evaluated in whole larvae, no significant difference was found compared to the controls (Lange et al., 2012).

One of the structures involved in the control of locomotion in zebrafish is the posterior tuberculum, a structure that corresponds to one of the regions with the highest number of DA neurons in the brain of zebrafish. Interestingly, disorganization, as well as a reduction in the overall number of neurons in the posterior tuberculum was observed in Lphn3.1 morphants. These organisms also showed a reduction in the total number of DA neurons in this structure (Schweitzer and Driever, 2009; Tay et al., 2011; Lange et al., 2012).

The characterization of Lphn3.2 remains elusive and it is tempting to speculate on their level of redundancy in zebrafish physiology. However, the studies conducted in zebrafish show the importance of Lphn3.1 in the control of movement and provide a clue of its relationship with the dopaminergic system.

Teneurin homologs reported in zebrafish brain include Tenm3 and Ten-m4, as potential Lphn3 ligands. In the forebrain and the midbrain, Ten-m3 and Ten-m4 have a complementary expression: Ten-m3 is expressed in the optic vesicles, the region covering the caudal diencephalon and the mesencephalon showing strongest expression at its most anterior part, while Ten-m4 is expressed in the rostral diencephalon with the least expression in the optic vesicles, and a region covering the mesencephalon and the midbrain/hindbrain boundary (Mieda et al., 1999). Unlike their mammalian counterparts, there are no reports of an interaction between orthologs of latrophilins and homologs of teneurins in zebrafish (Boucard et al., 2014).

#### Latrophilins in Mice

Mice genomes express three isoforms of latrophilins: Lphn1, Lphn2 and Lphn3. While enriched in neurons these receptors can also be found expressed in non-neuronal tissues such as kidney, lung and heart.

#### Synaptic Phenotypes

Lphn1- Latrophilin-1 is the most abundant isoform expressed in the adult mouse brain (**Figure 7**; Sugita et al., 1998; Matsushita et al., 1999; Boucard et al., 2014). Despite this early observation very few studies report on the role of this isoform in central synapses. An indirect assessment of its synaptic role obtained through the use of α-Latrotoxin on isolated synaptosomes of mice lacking Lphn1, revealed that this receptor isoform participated in glutamate release (Tobaben et al., 2002).

Lphn2- This isoform appears to be widely expressed, thus showing a widespread presence in many neuronal cell types (**Figure 7**; Kreienkamp et al., 2000; Anderson et al., 2017). Despite latrophilin-2 being potentially present in many types of synapses, its predominant function was observed to contribute to the development of specific synaptic sites. A study conducted by Anderson et al. (2017) aiming to characterize the role of latrophilin-2 in the synaptic physiology of the hippocampus found that this presumptive receptor for α-latrotoxin played a post-synaptic role rather than a pre-synaptic one, at least in the system surveyed. In the hippocampus neuronal network, neurons from the entorhinal cortex send projections to the CA1-region pyramidal neurons of the stratum lacunosum-moleculare (SLM), thus representing the pre- versus post- synaptic configurations, respectively. Latrophilin-2 expression was found to be enriched in the SLM dendritic spines where its deletion led to post-synaptic defects of excitatory synapses linked to spine development and function whereas the properties and characteristics of inhibitory synapses where kept unchanged. Moreover, because the number of excitatory synapses was selectively reduced following a genetic deletion of Lphn2 in hippocampal neurons, the authors attributed this deficiency to the consequent alteration of the target recognition abilities of neurons lacking this receptor. Giving that the other Lphn isoforms are also expressed in the same neuronal network of the hippocampus, these data would suggest that the function of Lphn2 is not redundant in this network. This study provided unsuspected data pertaining to Lphn2 localization and function at synapses that raised the following questions amongst others: how is Lphn2 trafficked to both pre- and post- synaptic compartments? Which is the presynaptic ligand responsible for Lphn2 role in target recognition? To which extent can the function of Lphn2 be dissociated from the function of other Lphn isoforms? While the role of Lphn2 in neuronal physiology and function remain intriguing, one observation remains clear, Lphn2 is an essential element amongst the molecular determinants that support synaptogenesis in mammals.

Lphn3- As referenced in Section "The Role of Latrophilins in Human Neuropathophysiology," this latrophilin isoform differs from the other isoforms because it amounts for most of the genetic associations made with human neurological disorders so far. While an assumption can be made for the role of Lphn3 in neuronal functions, its role at the synapse is far from being elucidated. Indeed, loss-of-function studies resulting in genetic deletion of Lphn3 in mice (Mus musculus) revealed

that dopaminergic neurons as well as molecular determinants of the dopamine pathway were altered in genetically modified animals (Lange et al., 2012; Wallis et al., 2012; Orsini et al., 2016). The first piece of evidence indicating that Lphn3 might play a role at the synapse was provided by an RNA interference approach in which a reduction in Lphn3 mRNA levels in mice hippocampal neurons led to a defect in presynaptic function of excitatory neurons, an effect consistent to the receptor's presumptive localization (O'Sullivan et al., 2012). The same RNA interference approach applied to cortical neurons in vivo provided a second line of evidence linking Lphn3 mRNA levels with the formation and function of specific synaptic contacts of the cortical circuitry (O'Sullivan et al., 2014). In this paradigm, the functional excitatory connection between cortical neurons from Layer 2/3 to Layer 5 neurons exhibited a postsynaptic defect with no measurable presynaptic defect while morphological determinants of this connection revealed both pre- and post-synaptic defects (O'Sullivan et al., 2014). The third line of evidence came from the study of Lphn3 conditional knockout mice which located Lphn3 expression to excitatory synapses of pyramidal CA1 neurons particularly enriched in stratum oriens (SO) and stratum radiatum (SR). These mice displayed defects in dendritic spines formation within SO and SR as well as a selective loss of excitatory synaptic inputs from Schaffer collateral projections, most recently many of these findings were replicated in a rat model (Regan et al., 2019; Sando et al., 2019). Thus, elucidating the function of Lphn3 will prove instrumental to understanding its role in physiological and pathophysiological brain functions.

#### Behavioral Phenotypes

Lphn1—Out of all latrophilin isoforms in mice, latrophilin-1 displays the higher expression in the central nervous system (Boucard et al., 2014). Despite this widely reported observation, no striking phenotype has been described for mice expressing a loss-of-function genotype for latrophilin-1. Other than a perceived lack of maternal instinct, Lphn1 deficient mice do not display severe behavioral defects and these mice are seemingly both viable and fertile in a manner that is indistinguishable from their wild-type counterparts (Tobaben et al., 2002).

Lphn2— This widely expressed latrophilin isoform appears to be essential for the proper development of mice. Indeed, constitutive Lphn2 deletion is embryonically lethal as litters from heterozygous crossing do not yield homozygous pups, thus hinting at a role that is most likely not neuron-specific but rather would refer to its importance in crucial developmental check points (Anderson et al., 2017). On the other hand, when Lphn2 is deleted in neurons only, mice suffer behavioral impairments that are linked to learning paradigms. Mice that lack Lphn2 in neurons possess less flexibility in the way they apply their learning abilities as they are unable to adapt to new learning paradigms that require a temporal change in a sequence of events (Anderson et al., 2017). These findings are particularly interesting in the context that these mice can learn tasks at a rate similar to their wild type counterparts because it suggests that Lphn2 would be required to allow generalized learning which supports the notions of abstraction or generalization, concepts that describe how a learning experience acquired in a particular context can then be applied when the context later changes by retaining core elements of learning.

Lphn3— The brain-enriched expression of this latrophilin isoform emphasizes its potential role in cognitive functions. Genetic manipulations leading to deletion of Lphn3 in the full organism causes marked alterations in behavior of engineered mice. A stark impediment in reward-seeking behavior can be observed in Lphn3-deficient mice as exemplified by a higher food consumption and a higher locomotor response to cocaine administration than their wild-type littermates (Wallis et al., 2012; Orsini et al., 2016). Additionally, these mice expressed a hyperactive phenotype measured in both horizontal and vertical activity with a concomitant higher level of stereotypy (Wallis et al., 2012). These behavioral phenotypes are reminiscent of traits elicited in addiction paradigms, thus suggesting that Lphn3 is required for regulating reward pathways.

# THE ROLE OF LATROPHILINS IN HUMAN NEUROPATHOPHYSIOLOGY

According to the preferred expression of latrophilins in the brain, this family of receptors seems to be having an important role in this central decision making and executive organ. Thus, it is conceivable that modifications to the function of these GPCRs and/or their ligands will have repercussion in human health. Here we summarize a few neuronal disorders with which latrophilin genes defects have been associated such as: ADHD, substance use disorder (SUD), autism spectrum disorder (ASD), bipolar disorder (BD), schizophrenia (SCZ), epilepsy and microcephaly (MCP) (**Figure 8**).

# Attention Deficit and Hyperactivity Disorder (ADHD)

Attention deficit and hyperactivity disorder is a neurodevelopmental disorder that affects the brains cognitive functions and is characterized by a deficit of attention, hyperactivity and impulsivity. Its high level of heritability of approximately 75% suggests the involvement of a strong genetic components although some environmental factors are also suspected to influence the etiology of ADHD (Akutagava-Martins et al., 2016; Bonvicini et al., 2016). Due to the nature of the disease and its symptoms, the first genes to be studied associated with ADHD were part of the dopaminergic and serotonergic pathways, given that the neurotransmitters dopamine and serotonin are involved in attention, learning and motor control. Additionally, patients with ADHD are treated with medication that affect the transport of dopamine to the synapse or its retention or recapture by synaptic components but the mechanism by which these drugs act is not entirely clear (Faraone et al., 2005; Mick and Faraone, 2008; Genro et al., 2010).

In an effort to identify the genetic risk factors that contribute to the etiology of ADHD, a multigenerational study was carried out in an isolated population from Colombia with a high prevalence of ADHD. In this study, a significant link between

disorder; BD, Bipolar disorder; SUD, Substance use disorder; SCZ, Schizophrenia; MCP, microcephaly; RES, rhombencephalosynapsis.

ADHD and a region of chromosome 4q13.2 was reported and later circumscribed to the latrophilin-3 gene (ADGRL3) (Arcos-Burgos et al., 2010). Moreover, the presence of ADGRL3 single nucleotide polymorphisms (SNPs) were confirmed in other populations samples. A thorough analysis of ADGRL3 variants by gene sequencing led to the identification of polymorphisms in both exonic and intronic regions (Domene et al., 2011). Very few studies have addressed the ADHD-related ADGRL3 variations at the molecular level. One such study combining model organism genetics and in vitro assays identified an evolutionary conserved region located in a potential regulatory sequence within the minimal critical region attributed to ADGRL3. This region contained a three-variant ADHD risk haplotype (rs17226398, rs56038622, and rs2271338) that reduced the enhancer activity by 40%. One risk allele (rs2271338) was associated with a reduced expression of Lphn3 in the thalamus and the same risk allele was found to disrupt binding to the YY1 transcription factor, an important regulator of development of the central nervous system (Martinez et al., 2016).

Reinforcing the role of ADGRL3 in the etiology of ADHD, variants or haplotypes of this gene have been linked to the effectiveness of stimulant medication. However, the results obtained were controversial. On the one hand Arcos-Burgos et al. (2010) observed that the G allele carriers within ADHDassociated SNP rs6551665, presented a better response to medication regarding inattention whereas Labbe et al. (2012) reported that carriers of the same pathogenic allele displayed a lower response to treatment with respect to hyperactivity (Arcos-Burgos et al., 2010; Labbe et al., 2012). On the other hand, another study suggested that the homozygous carriers of the CGC haplotype (rs6813183, rs1355368, and rs734644) expressed a faster response to symptoms' improvements following methylphenidate (MPH) treatment, a psychostimulant medication prescribed to alleviate symptoms of ADHD and thought to block dopamine reuptake (Volkow et al., 2002; Genro et al., 2010; Bruxel et al., 2015). However, a meta-analysis study of existing literature revealed that variant rs6551665 was not significantly associated with MPH response in children (Myer et al., 2018). The apparent discrepancies as to if ADGRL3 haplotypes represent clinically relevant predictors of treatment response could be attributed to differences in the ethnicity of the populations studied.

Little is known about the effects that environmental factors exert on the development of ADHD. Among known environmental factors, maternal smoking and stress during pregnancy are thought to increase the risk for developing ADHD. A significant association was detected between previously described ADGRL3 SNPs (rs6551665, rs1947274, rs6858066, and rs2345039) and MPH treatment after ADHD diagnosis under these environmental factors such that the patients which mothers experienced less stress had a better response outcome (Choudhry et al., 2012). However, another environmental factor associated with the use of acetaminophen during pregnancy did not yield a significant increase in ADHD-related symptoms in a model organism (Brandlistuen et al., 2013; Liew et al., 2014; Thompson et al., 2014; Reuter et al., 2016).

The identification of ADGRL3 provided a disease-relevant target because: (a) it is expressed in brain areas related to attention and activity in human such as the prefrontal cortex, cerebellum, amygdala and temporal lobes (Krain and Castellanos, 2006; Plessen et al., 2006; Arcos-Burgos et al., 2010); (b) ADGRL3-deficient animal models display phenotypes linked to ADHD such as hyperactivity, deficiencies in dopamine and serotonin molecular pathways, but also show a response to MPH treatment in alleviating symptoms (Lange et al., 2012; Wallis et al., 2012; Orsini et al., 2016; van der Voet et al., 2016).

#### Autism Spectrum Disorder

Autism spectrum disorder (ASD) is a neuropsychiatric disorder characterized mainly by deficits in social communication and interactions, restricted or repeated actions referred as stereotypic behaviors (Levy et al., 2009). According to different studies the heritability contributes from 54 to 95% of its etiology (Gaugler et al., 2014; Sandin et al., 2014; Colvert et al., 2015). Hence, most studies aiming to elucidate the causes of ASD focused on identifying genetic factors. Interestingly, comorbidity with other neurological diseases often arises when a diagnosis of ASD is given such as the one existing with ADHD. Both disorders present neurological alterations and many of the genes that have been related to their etiology encode synaptic proteins, suggesting that the disorders present dysfunction at the

synaptic level (Rowlandson and Smith, 2009; Matson et al., 2010; Guang et al., 2018).

The temporal and frontal lobes are the main brain areas affected in patients with ASD, highlighting the role of the amygdala by its association with aggressive and social behaviors. These areas also contain an important proportion of neurons producing dopamine, a neurotransmitter reinforcing pleasant behaviors through the reward pathway. The mesolimbic pathway which regulates reward processing connects the ventral tegmental area (VTA) of the midbrain and the nucleus accumbens (NAc) of the striatum via white matter tracts (Haber and Knutson, 2010; O'Connell and Hofmann, 2011). Patients with ASD display reduced structural connectivity between the VTA and NAc in the mesolimbic pathway and weaker connectivity in this area translates to more severe social deficits (Supekar et al., 2018). These findings support the hypothesis that patients with ASD find social stimuli less rewarding than their neurotypical peers, which is reflected in their social skills (Sah et al., 2003; Chevallier et al., 2012). Thus, this condition might be related to defects in dopaminergic signaling, a phenotype that is reminiscent of ADHD neuronal deficiencies.

Among the genes whose variants have been related to autism are the genes encoding for synaptic proteins Lphn3, neurexins, neuroligins and SHANK (Wang et al., 2016; Chen et al., 2017; Stessman et al., 2017). It is worth mentioning that both neurexins and the neuroligins form a complex within the synapses which is responsible for recruiting synaptic components such as neurotransmitter receptors and scaffolding proteins to promote an assembly of the synapse as well as its maturation and differentiation (Krueger et al., 2012; Sudhof, 2017). Copy-number variations (CNVs) within NRXN1 have been associated with ASD, however, they are extremely rare and have low penetrance in the general population while NLGN mutations related to ASD exhibited defects in synaptic properties and ASD-like behavioral changes when studied in mouse models (Jamain et al., 2003; Tabuchi et al., 2007; Jaramillo et al., 2014; Todarello et al., 2014). The SHANK family of PDZ domain proteins function as molecular scaffolds at excitatory synapses and through their multiple domains are able to interact with more than 30 synaptic proteins, which confers them an essential role in the formation of synapses (Monteiro and Feng, 2017). Impairments in cognitive function were detected in mice heterozygous for Shank3 with the PDZ domain deleted (Mei et al., 2016). Interestingly, Lphn1 is able to interact with Neurexins to form adhesion complexes (Boucard et al., 2012) while Shank proteins are also able to interact with Lphns PDZ binding domain (Kreienkamp et al., 2000; Tobaben et al., 2000). This network of interaction hints to a common biological pathway underlying the etiology of ASD (Mosca et al., 2017).

#### Bipolar Disorder

Bipolar disorder (BD) is a severe chronic mood disorder whose symptoms are episodes ranging from mania, hypomania to severe depression (Vieta and Phillips, 2007). Within the cerebral regions affected in the disease, the hippocampus stands out. In addition to having a critical role in cognitive functions, the hippocampus is also involved in emotion and other functions that are altered in BD such as motivational behaviors and response to stress (Surget et al., 2011; Rive et al., 2013). Several studies have reported hippocampal subfield-level volume reductions in BD, particularly in the right cornu Ammonis 1 (CA1), the granule cell layer (GCL), and the whole hippocampus compared with healthy controls (Haukvik et al., 2015; Han et al., 2019).

Genetic factors play an important role in the disease. The heritability of BD according to twin studies has been estimated to range between 60 and 80%, while lower rates of risk have been found in intergenerational family studies in large population cohorts (Smoller and Finn, 2003; Wray and Gottesman, 2012). Like for many psychiatric disorders, BD presents comorbidity with other mood disorders such as ADHD and/or alcoholism, while displaying pathophysiological defects suggestive of a common etiology based in a monoaminergic imbalance, more specifically alterations of the dopaminergic system, similar to what has been reported for ADHD and substance use disorder, as discussed above (Martinowich et al., 2009; Lydall et al., 2011; Vaughan and Foster, 2013).

Several risk alleles for BD have been identified in genomewide association studies involving diagnosed patients, among these the following genes encoding Lphn ligands:TENM2, NRXN1, NRXN3 and FLRT2 (Rouillard et al., 2016). Within the chromosomal regions identified as part of the risk loci for BD lies the gene TENM4 encoding teneurin-4 (Craddock and Sklar, 2013; Muhleisen et al., 2014). Interestingly, the teneurin family are known ligands for latrophilins, forming trans-synaptic interactions that are suggested to participate in the formation and maintenance of neuronal synapses (Boucard et al., 2014). Although a direct participation of latrophilins in BD has not been reported so far, there could be a pathway associated with the disease in which the Lphns are involved. This hypothesis would be supported by the comorbidity between diseases BD (associated with Ten4) and ADHD (associated with Lphn3) in addition to the alteration of the dopaminergic system observed in latrophilin3-deficient animal models.

#### Substance Use Disorder (SUD)

Substance use disorder is an important health problem at a global level, with high economic costs and which is expected to continue to grow over time. This disorder is characterized by a prolonged use of legal or illegal drugs as well as medications, which triggers a loss of self-control (American Psychiatric Association [APA], 2013). Like for other neuropsychiatric diseases, SUD has a strong genetic component. Studies report that the parental background of alcoholism or family history of diagnosis of SUD considerably increase the chances of developing alcohol problems, however, environmental factors also play an important role, which has been shown in studies with twins (Chassin et al., 1991; Agrawal et al., 2010; Huizink et al., 2010).

The areas of the brain that seem to be mostly involved in initial drug reward/saliency are mid-brain dopamine neurons projecting into the prefrontal cortex as well as the dorsal and ventral striatum, these data are supported by imaging studies that show that drug use increases striatal dopamine proportionally to self-reported euphoria (Drevets et al., 2001; Sharma and Brody, 2009; Volkow et al., 2009).

ADHD is a disease that frequently presents comorbidity with SUD. Children diagnosed with ADHD and who were followed up toward adolescence exhibit higher rates of alcohol, tobacco, and psychoactive drug use, as well as greater professional, social and personal impairment than non-ADHD subjects (Molina and Pelham, 2003; Molina et al., 2013; Nogueira et al., 2014). Interestingly, a study investigating whether ADHD risk variants at the ADGRL3 locus interact with clinical, demographic, and environmental variables associated with SUD revealed that the presence of SUD in patients with ADHD can be predicted efficiently, thus identifying ADGRL3 as a risk gene for SUD (Arcos-Burgos and Velez, 2019). In agreement with these results, treatment for ADHD was associated with lower concurrent risk of SUD (Quinn et al., 2017).

#### Microcephaly

Genetic alterations during the development of the nervous system are one of the main causes that lead to malformations of cortical development (MCP). Microcephaly is a type of MCP that is characterized by a reduction in the circumference of the head of a human, where infectious, environmental and genetic factors are considered as the causative agents in this condition (Parrini et al., 2016). During the development of the cerebral cortex, cell proliferation, neuronal migration or postmigrational cortical and connectivity are key stages for a successful development and any defect in the regulation of these cellular processes can lead to different types of MCP. Particularly in microcephaly there is a deregulation in DNA replication that leads to the decrease of cell proliferation, and therefore to the aforementioned phenotype (Barkovich et al., 2012; Kalogeropoulou et al., 2019). It has been reported that patients with defects in a single gene display comorbidity between microcephaly and other MCPs, such as lissencephaly (which is associated with deficiencies in neural migration) and agenesis of the corpus callosum (ACC) and more recently with rhombencephalosynapsis (RES)(Parrini et al., 2016). RES is an extremely rare malformation in which there is no anatomical differentiation of the cerebral hemispheres (Aldinger et al., 2018). A new variant in the LPHN2 gene was detected in a sample from a human fetus which presented severe microcephaly, severely reduced sulcation and RES. This nonsense variant resulted in the change of a leucine to a histidine at position 1262 of its intracellular domain, which affected its functionality in the mobilization of calcium through its coupling to G proteins and the organization of the cytoskeleton, promoting an increase in the cell adhesion and decrease in cell migration, processes which are of crucial importance in cortical development (Vezain et al., 2018). These findings highlight the regulatory capacity of Lphn2 in determining the etiology of neurodevelopmental disorders.

#### Schizophrenia

Schizophrenia (SCZ) is a neurological disorder affecting approximately 1% in the world population. Patients with this disorder usually present psychotic symptoms (hallucinations), social withdrawal and deficits in attention and working memory. Schizophrenia usually displays late adolescence onset or early adulthood onset and is considered multifactorial (Kellendonk et al., 2009). However, genetic factors contribute approximately 60–80% in its etiology. At the molecular level, alterations in the synthesis and release of dopamine were detected in the striatum and in the dorsolateral prefrontal cortex of affected individuals (Howes et al., 2012; Conio et al., 2019). Recently, a single nucleotide variation in an intronic sequence of the ADGRL2 gene was reported in patients diagnosed with schizophrenia who were prescribed clozapine, an antipsychotic usually recommended for the treatment of schizophrenia (Legge et al., 2018). Deletions or polymorphisms in the genes that encode Lphn ligands neurexins and teneurins, have also been associated with this condition (Kirov et al., 2008; Gauthier et al., 2011; Ivorra et al., 2014). Notably, association studies with SCZ identified a single nucleotide modification in NRXN1 gene which resulted in poor synaptic differentiation and loss of interaction with its canonical ligand, neuroligin, in neuronal co-cultures; in addition, variants located in the YD repeat domain of teneurin-4 were also identified in samples from SCZ patients some of which presented a comorbidity with bipolar disorder (Yamada et al., 2004; Kirov et al., 2008; Gauthier et al., 2011; Xue et al., 2018). Although the contribution of latrophilins to the etiology of schizophrenia is unknown, their role in synapse formation and their association with the regulation of dopaminergic signaling constitute key features that warrant a closer look at the pathophysiological functions of these molecules for this psychiatric disorder.

#### Epilepsy

Epilepsy is a disorder that is characterized by the presence of seizures presumably because of an alteration in the balance between excitatory and inhibitory impulses in the brain. This condition usually presents comorbidity with other disorders such as depression, SCZ and MCP. There are different types of epilepsy according to the type of convulsion, the affected brain area, age of the patient, and etiological factors. Its heritability is high but there are also sporadic cases where its condition is related to environmental factors (Stafstrom and Carmant, 2015; Hauser et al., 2018). Mutations in genes that code for sodium (SCN1A) and potassium (KCNA2) channels and N-methyl-Daspartate (NMDA) receptors have been highly associated with their condition (Perucca and Perucca, 2019). However, there is a significant interest in the study of new variants related to epilepsy among them neurexins and contactin-6, ligands that interact with latrophilins. A female infant with Early infantile epileptic encephalopathy presented variations of a single nucleotide in the NRXN1 and 2 genes generating a missense mutation in the corresponding proteins (Rochtus et al., 2019). In another study of patients with generalized epilepsies (IGEs) of European ancestry, exon disrupting deletions were reported in the promoter region of NRXN1 (Moller et al., 2013). On the other hand, the relationship between epilepsy and variations in the contactin-6 gene (CNTN6) remains scarce, but a deletion of exons 21 and 22 has been highly associated with the presence of schizophrenia and seizures (which is the hallmark symptom of epilepsy) (Juan-Perez et al., 2018). Thus, of the two latrophilin ligands, neurexins associations with epilepsy retain the most interest. As for latrophilins, the evidences are scarce and can be summed up to a study in patients diagnosed with partial epilepsy of European ancestry which reported five variants in different

intronic regions of the ADGRL3 gene; however, its relationship with the disease did not reach significance (Kasperaviciute et al., 2010). This result does not nullify the possibility of its relationship with the disease, because due to its multifactorial etiology, ADGRL3 could be related to other types of epilepsy. Lphn3 role in modulating the formation of specific excitatory synaptic contacts (Sando et al., 2019) suggest this molecule as a potential epilepsy risk factor giving that some of its variants could lead to an imbalance at the excitatory level that generates symptoms related with this disorder.

#### CONCLUDING REMARKS AND HYPOTHESIS

Latrophilins are bound to affect proper neuronal functions given their conserved expression in this cell type, from mechanosensation in Drosophila, to pharyngeal pumping in C. elegans or learning in M. musculus. While some specific roles of latrophilins across organisms may vary, the underlying basic mechanisms that rely on their domain structure are likely conserved. Their adhesion function in mammals requires heterophilic interactions with teneurins, but this binding profile has not been replicated in invertebrate animal models. Latrophilins in these organisms are likely to form complexes with teneurins given that (a) the Lectin-like domain of latrophilin that binds teneurins in mammals is conserved in invertebrates, (b) the expression of both adhesion molecules overlaps with one another in certain tissues or are directly adjacent.

Latrophilins' association with numerous psychiatric disorders hints to their importance in the modulation of cognitive functions in humans. Animal models deficient in latrophilin-3 orthologs display behavioral phenotypes that relate to the human condition of ADHD and respond to clinically relevant medication, thus suggesting an interspecies role of this receptor in regulating dopaminergic pathways. However, more needs to be done to understand the underlying biological role of latrophilins. Our theory is that latrophilins, by transducing adhesion events into G protein-dependent and G proteinindependent cell signaling cascades are relevant for neuronal development and brain functions. Furthermore, latrophilins

#### REFERENCES


mediate synaptogenesis and therefore the plastic behavior of the nervous system. We propose that latrophilins can act both in cis and trans configurations with their ligands to produce signaling complexes that can elicit configuration-dependent signaling schemes. Despite significant progress in various models, more work is required to identify the specific contexts in which these receptors function.

#### DATA AVAILABILITY

All datasets generated for this study are included in the manuscript and/or the **Supplementary Files**.

#### AUTHOR CONTRIBUTIONS

AM-S, MA-Z, and AB reviewed the literature. AM-S, MA-Z, FM, and AB drafted and designed the manuscript. PU-S electronically drew and assembled all figures. AB, AM-S, MA-Z, PU-S, and FM provided the concept for the figures and discussed figure details. AM-S, MA-Z, and AB conducted and analyzed the alignment data. DH-G, FM, and AB provided experimental data and analysis of microscopy images. All authors provided intellectual inputs, revised, and approved the final version of the manuscript.

#### FUNDING

This project was provided by the Consejo Nacional de Ciencia y Tecnología (CONACyT) to AB (Ciencia Básica #221568) and FM (Ciencia Básica #179835) with doctoral degree scholarships to AM-S (#587784), MA-Z (#295903), PU-S (#233278), and master degree scholarship to DH-G (#298900).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnins. 2019.00700/full#supplementary-material


recognition in entorhinal-hippocampal synapse assembly. J. Cell Biol. 216, 3831–3846. doi: 10.1083/jcb.201703042






**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Moreno-Salinas, Avila-Zozaya, Ugalde-Silva, Hernández-Guzmán, Missirlis and Boucard. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Latrophilins and Teneurins in Invertebrates: No Love for Each Other?

Torsten Schöneberg and Simone Prömel\*

Medical Faculty, Rudolf Schönheimer Institute of Biochemistry, Leipzig University, Leipzig, Germany

Transsynaptic connections enabling cell–cell adhesion and cellular communication are a vital part of synapse formation, maintenance and function. A recently discovered interaction between the Adhesion GPCRs Latrophilins and the type II single transmembrane proteins Teneurins at mammalian synapses is vital for synapse formation and dendrite branching. While the understanding of the effects and the molecular interplay of this Latrophilin-Teneurin partnership is not entirely understood, its significance is highlighted by behavioral and neurological phenotypes in various animal models. As both groups of molecules, Latrophilins and Teneurins, are generally highly conserved, have overlapping expression and often similar functions across phyla, it can be speculated that this interaction, which has been proven essential in mammalian systems, also occurs in invertebrates to control shaping of synapses. Knowledge of the generality of this interaction is especially of interest due to its possible involvement in neuropathologies. Further, several invertebrates serve as model organisms for addressing various neurobiological research questions. So far, an interaction of Latrophilins and Teneurins has not been observed in invertebrates, but our knowledge on both groups of molecules is by far not complete. In this review, we give an overview on existing experimental evidence arguing for as well as against a potential Latrophilin-Teneurin interaction beyond mammals. By combining these insights with evolutionary aspects on each of the interaction partners we provide and discuss a comprehensive picture on the functions of both molecules in invertebrates and the likeliness of an evolutionary conservation of their interaction.

Keywords: adhesion GPCRs, Latrophilins, Teneurins, invertebrates, interaction

# LATOPHILINS AND TENEURINS FORM A TRANSSYNAPTIC COMPLEX IN MAMMALS

The formation of synapses is one of the key steps in warranting the development of a functioning neuronal network. This highly complex process is not fully understood, but it involves various interactions of molecules with adhesive and transmembrane signaling properties. A pair of proteins which has recently taken the stage to be essential for synaptic organization in many vertebrates are Latrophilins and Teneurins. Both have already been separately recognized as synaptic cell surface proteins several decades ago.

#### Edited by:

Antony Jr Boucard, Centro de Investigación y de Estudios Avanzados (CINVESTAV), Mexico

#### Reviewed by:

Theodoros Tsetsenis, University of Pennsylvania, United States Konark Mukherjee, Fralin Biomedical Research Institute (FBRI), United States

#### \*Correspondence:

Simone Prömel simone.proemel@medizin.unileipzig.de

#### Specialty section:

This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Neuroscience

Received: 08 January 2019 Accepted: 11 February 2019 Published: 12 March 2019

#### Citation:

Schöneberg T and Prömel S (2019) Latrophilins and Teneurins in Invertebrates: No Love for Each Other? Front. Neurosci. 13:154. doi: 10.3389/fnins.2019.00154

**132**

Teneurins are large type II one-transmembrane domain proteins with a cytoplasmic N-terminus and an extracellularly located C-terminus containing tyrosine-aspartate (YD) repeats and numerous epidermal growth factor (EGF) domains (Oohashi et al., 1999; Tucker and Chiquet-Ehrismann, 2006; **Figure 1A**). They have various neuronal functions, for example in mediating interneuronal connections, promoting synapse formation and shaping dendritic morphology in diverse types of neurons in vertebrates and invertebrates (Hong et al., 2012; Mosca et al., 2012; Antinucci et al., 2013; Berns et al., 2018). Consistently, the four vertebrate homologs (TEN1–4) are widely expressed in the developing and the adult brain, for instance in the hippocampus, the cerebellum and the visual cortex (Oohashi et al., 1999; Tucker et al., 2000; Rubin et al., 2002; Zhou et al., 2003; Kenzelmann et al., 2008). Studies on animal models further reveal the essential impact of Teneurins on neuronal circuits. For example, mice knockout for Ten3 display neurological defects, in particular deficits in visually mediated behavior (Leamey et al., 2007). Similarly, in zebrafish, knockdown of Ten-3 leads to retinal ganglion cell stratification defects (Antinucci et al., 2013).

The molecular details underlying Teneurin function involve the formation of homotypic or heterotypic dimers depending on the synapse type [summarized in Mosca (2015)]. Most details on Teneurin function, however, have not been collected in vertebrates, but using the fruit fly Drosophila melanogaster as a model (section "Latrophilins and Teneurins in D. melanogaster– No Evidence for Interaction").

The functions of Latrophilins have by far not been as well characterized as the ones of Teneurins. Latrophilins belong to the class of Adhesion G protein-coupled receptors (Adhesion GPCRs, aGPCRs). The three mammalian homologs (LPHN1–3/ADGRL1-3) comprise an intracellular C-terminus, a seven transmembrane region (7TM) and an extracellular N-terminus containing a rhamnose-binding lectin (RBL), an olfactomedin (OLF), a hormone binding (HRM) and a GPCR autoproteolysis-inducing (GAIN) domain, which harbors the GPCR proteolytic site (GPS) (**Figure 1B**). Latrophilins first came into the focus of science as targets of α-Latrotoxin, a component of the Black widow spider's toxin (Krasnoperov et al., 1996; Lelianova et al., 1997; Sugita et al., 1998). Specifically, Latrophilin-1 (LPHN1/ADGRL1) has subsequently been characterized to be expressed in various neurons of the murine central nervous system and evidence exists that LPHN1 is localized presynaptically (Silva et al., 2011; Vysokov et al., 2018) as well as on the post-synapse (Tobaben et al., 2000; Anderson et al., 2017). The impact of this localization on both sides of the synapse has not been clarified to date. Studies on LPHN3 in mouse and zebrafish models suggest a role for the receptor in the dopaminergic system and an association of variants in the receptor gene with the pathogenesis of attention-deficient hyperactive disorder (ADHD) (Arcos-Burgos et al., 2010; Lange et al., 2012; Wallis et al., 2012).

Teneurins and Latrophilins are both found enriched in neuronal growth cones (Nozumi et al., 2009). Recently, strong evidence has been provided that mammalian Teneurins and Latrophilins form heterophilic dimers at the synapse. This interaction, which occurs between LPHN1 and the Teneurin homolog TEN2 [also termed Lasso (Silva et al., 2011)], is transsynaptic and mediates cell adhesion (Silva et al., 2011; Boucard et al., 2014; Li et al., 2018). As a consequence, it induces synapse formation in murine hippocampal neurons and neuronal cultures (Silva et al., 2011; **Figure 2A**). Another study has shown that besides TEN2 also TEN4, but not TEN1 is able to bind LPHN1 (Boucard et al., 2014). It needs to be noted that not for all Teneurin functions in neurons, interaction with Latrophilin is essential [reviewed in Mosca (2015)]. It has only proven vital for cell adhesion and synapse formation so far. For its other roles homophilic interactions or different heterophilic partners have been shown.

Although the molecular details of the interaction between Latrophilins and Teneurins have not been clarified yet, the regions within both molecules taking part in the intermolecular interaction have been roughly identified (**Figure 2A**) using binding assays and mutation analyses. For TEN2, the interaction is mediated via its C-terminal portion, mainly by a sequence within the Tox-GHH domain, the so-called Teneurin C-terminal-associated peptide (TCAP). This sequence can act as a bioactive peptide upon cleavage and shapes dendritic morphology, stimulates neurite outgrowth and mediates anxiety behavior (Wang et al., 2005; Al Chawaf et al., 2007a,b; Tan et al., 2011). Interestingly, besides this core sequence within the Tox-GHH domain, a 7-amino acid-long region located in a β-propeller close to the NHL (NCL-1/HT2A/Lin-41) repeats also regulates binding (Li et al., 2018). The same seems to be true for the interaction site within Latrophilins. While the presence of the RBL domain is mainly responsible for binding Teneurins (Boucard et al., 2014), an alternative exon encoding a region between RBL and OLF domains modulates binding affinity to TEN2 (Boucard et al., 2014). It needs to be noted that currently existing data on the partnership of Latrophilins and Teneurins does not exclude the possibility that the interaction occurs in the context of a larger complex involving other molecules. This scenario has been already proposed (Woelfle et al., 2015, 2016) based on the findings that Teneurin also interact with dystroglycans (Chand et al., 2012) and Latrophilins bind to Neurexins (Boucard et al., 2012) or (in a complex) to fibronectin leucine-rich transmembrane (FLRT) proteins (O'Sullivan et al., 2012, 2014; Jackson et al., 2015; Lu et al., 2015). These interaction partners are all expressed by neurons.

As both, Latrophilins and Teneurins, can act as ligand, it is conceivable that each of them functions as receptor transducing signals into their host cell. It has not been determined beyond doubt to date which of them is the ligand and which the signalreceiving molecule or if both of them signal. However, some studies show that Teneurins are cleaved at several distinct sites rendering liberated fragments (Wang et al., 2005), which are involved in different functions in the brain such as neurite outgrowth (Al Chawaf et al., 2007a; Erb et al., 2014). It has been suggested that one of these TEN2 fragments, generated by regulated proteolysis, is soluble and can still bind LPHN1 and trigger signaling (Silva et al., 2011; Vysokov et al., 2016,

scale. Domains were annotated using InterPro (EMBL-EBI) and SMART (Letunic et al., 2009).

2018) indicating that LPHN1 is the receptor transducing information into the cell.

#### LATROPHILINS AND TENEURINS IN INVERTEBRATES HAVE SIMILAR FUNCTIONS

Due to the obvious relevance of the Latrophilin-Teneurin interaction in mammals the question of the generality of this partnership and thus, its conservation, arises. This question is especially of interest as invertebrate models are often used for elucidation of neurobiological aspects and understanding of association with pathologies. The described interaction between Latrophilins and Teneurins is so far limited to vertebrates, it has not been shown in invertebrate systems to date. However, Teneurins and Latrophilins are both highly conserved groups of molecules. First discovered in the fruit fly Drosophila melanogaster (Baumgartner and Chiquet-Ehrismann, 1993; Levine et al., 1994), Teneurins are evolutionarily as old as the unicellular choanoflagellates and are present in all metazoa investigated so far (Tucker and Chiquet-Ehrismann, 2006; Tucker et al., 2012). Similarly, Latrophilins belong to the evolutionarily oldest groups of Adhesion GPCRs being present in vertebrates and in invertebrates. Functionally, Teneurins also seem to be highly conserved, not only in respect to their neuronal roles (sections "Latrophilins and Teneurins in D. melanogaster– No Evidence for Interaction" and "Latrophilins and Teneurins in C. elegans Development Do Not Function as Ligand-Receptor Pair"). The receptors also have functions beyond synapse formation. It has been shown in mice that TEN4 is required for mesoderm induction and gastrulation (Lossie et al., 2005; Nakamura et al., 2013). Consistently, non-neuronal expression of mammalian Teneurins is found during embryonic development. This pattern is similar to the one of the Caenorhabditis elegans ortholog, suggesting conserved non-neuronal functions (section "Latrophilins and Teneurins in C. elegans Development Do Not Function as Ligand-Receptor Pair"). In contrast, for all that

FIGURE 2 | Interaction of Teneurins and Latrophilins. (A) In mammals, the interaction between LPHN1 on the pre-synapse of hippocampal neurons with postsynaptic TEN2 contributes to the control of synapse formation. The interaction interfaces are roughly known. TEN2 binds to LPHN1 via the C-terminal portion of the Tox-GHH domain, the Teneurin C-terminal-associated peptide (TCAP). A short amino acid sequence further N-terminal is involved in regulation of the binding. On the LPHN1 side the rhamnose-binding lectin (RBL) is required for binding as well as a sequence between the RBL and olfactomedin (OLF) domains. (B) In the fruit fly, the Teneurins Ten-a and Ten-m interact heterophilically at the neuromuscular junction to ensure synapse formation. Further, a homophilic interaction between Teneurins controls partner matching for instance in the olfactory system. The only Latrophilin homolog in Drosophila, dCirl, is located on neurons of the chordotonal organs and are involved in mechanosensation. (C) The C. elegans homologs of Latrophilins, LAT-1, and Teneurin, TEN-1, are present on the same embryonic blastomeres, excluding the possibility of a classical ligand-receptor pair. Rather, they are acting in parallel. Note that it is rather likely that for any of the interactions shown additional molecules or dimerization are required which are not depicted here.

is known to date, the functional conservation of Latrophilins throughout phyla has not been shown beyond doubt (sections "Latrophilins and Teneurins in D. melanogaster–No Evidence for Interaction" and "Latrophilins and Teneurins in C. elegans Development Do Not Function as Ligand-Receptor Pair").

Due to the overall similar conservation of the two molecules it has been postulated that their interaction and its physiological impact are also evolutionarily old and conserved (Chand et al., 2013; Woelfle et al., 2015). Although experimental proof is lacking that in invertebrates Latrophilins and Teneurins interact, a body of functional proof in the invertebrate model organisms D. melanogaster and C. elegans exists suggesting that an interaction of the two is conceivable. However, there is also some information arguing against this assumption which will be discussed below.

# Latrophilins and Teneurins in D. melanogaster – No Evidence for Interaction

Teneurins were first discovered in the fruit fly D. melanogaster as pair-rule genes tenascin-like molecule accessory (Ten-a) (Baumgartner and Chiquet-Ehrismann, 1993) and tenascin-like molecule major (Ten-m) (Baumgartner et al., 1994), which was also named odd oz (Odz) (Levine et al., 1994). The structural (**Figure 1A**) and functional conservation between mammalian and Drosophila Teneurins is evident. In Drosophila, Teneurins are widely expressed in neurons of the central and peripheral nervous system (Minet et al., 1999; Fascetti and Baumgartner, 2002) and several studies show their involvement in two different aspects at the synapse. Firstly, screens have revealed that they contribute to synapse formation of the neuromuscular junction (Liebl et al., 2006; Kurusu et al., 2008; **Figure 2B**). Further, Teneurins have implications in partner matching between presynaptic motoneurons and postsynaptic muscles as well as pre- and postsynaptic olfactory neurons and pre-synaptic motoneurons with postsynaptic muscles (**Figure 2B**; Hong et al., 2012; Mosca et al., 2012). Both functions can also be discriminated based on the connections that are formed by Teneurins. While in synaptogenesis Teneurins interact heterophilically (with another molecule or another Teneurin), they form homophilic interactions (with the same Teneurin) during partner matching. The heterophilic interaction partners described consist of presynaptic Ten-a and postsynaptic Ten-m and deletion of each of the molecules yields dysfunctional synapses and less synaptic boutons (Mosca et al., 2012). It is conceivable that Latrophilin might be another partner for a heterophilic interaction of Teneurins in this context. However, for Latrophilins, the functional conservation between mammals and Drosophila is not that evident, which is partly due the lack of knowledge about the receptor in the fruit fly. The one Latrophilin homolog the Drosophila genome carries, dCirl, has only been recently characterized. It is located on the neuronal dendrites and cilia of chordotonal organs in the fly and mediates sensitivity to touch (Scholz et al., 2015). This Adhesion GPCR is involved in mechanosenation, specifically shaping mechanically gated receptor currents by decreasing intracellular cyclic AMP levels, possibly by activating G<sup>i</sup> proteins (Scholz et al., 2017). The details of this function remain elusive and thus, no interaction with one of the Teneurins has been described so far. Such

interaction can be debated because Ten-m/Odz and dCirl both are located in neurons of the chordotonal organs (Levine et al., 1997), but it seems that they are on the same cell rather than on opposing neurons. However, a partnership might still be likely as we are only just beginning to understand the functions of dCirl.

# Latrophilins and Teneurins in C. elegans Development Do Not Function as Ligand-Receptor Pair

In the roundworm C. elegans, Teneurins and Latrophilins appear to have very similar functions and at first sight, it can be speculated that they form a classical interaction as described in mammals. However, a closer look prohibits this conclusion as of yet. In contrast to vertebrates or Drosophila, the nematode only has one Teneurin gene, ten-1. However, several transcript variants exist. The two most prominent ones are generated by two different transcription start sites: one variant with a longer (280 amino acids) intracellular domain and one with a short (36 amino acids) N-terminus (Drabikowski et al., 2005; **Figure 1A**). Both of these TEN-1 variants are present in distinct subsets of neurons (Drabikowski et al., 2005). Consistently, a role for TEN-1 in neuronal pathfinding has been postulated (Drabikowski et al., 2005). Although lat-1, one of the two Latrophilin homologs in C. elegans (**Figure 1B**), is also expressed in neurons (Langenhan et al., 2009), so far no neuronal function of LAT-1 has been described leading to the question whether the classical Latrophilin-Teneurin interaction plays a role in C. elegans. LAT-1 is functionally highly diverse. It has roles in fertility and cell polarity during development (Langenhan et al., 2009; Prömel et al., 2012), where it elicits a G<sup>s</sup> protein-mediated signal raising intracellular cyclic AMP levels (Müller et al., 2015), but a role in synaptogenesis similar to the one in mammals, has not been described yet, precluding a final assessment.

However, similar to mammalian and Drosophila Teneurins, expression of ten-1 is not limited to neuronal tissues but is also found in hypodermal cells (long TEN-1 variant), in cells of the gut, the somatic gonad, distal tip cell, and in few muscle cells (short TEN-1 variant) (Drabikowski et al., 2005). Interestingly, the expression pattern in non-neuronal cells is almost identical to the one of lat-1, which is mainly confined to cells of the somatic gonad and the distal tip cell (Langenhan et al., 2009), suggesting that LAT-1 and TEN-1 might have similar functions in a non-neuronal context. Indeed, not only the expression pattern of lat-1 and ten-1 is highly similar, but also the phenotype that respective knockout mutants display. Both mutants lat-1(ok1465) and ten-1(ok641) exhibit morphogenesis defects (Drabikowski et al., 2005; Langenhan et al., 2009). However, genetic analyses revealed that both genes act in parallel during development implying a synergistic rather than linear interaction between lat-1 and ten-1 (Langenhan et al., 2009; **Figure 2C**). In line with these findings, expression data show localization of TEN-1 and LAT-1 on the same embryonic blastomeres rather than on opposing cells, indicating that the two receptors do not form the classical ligand-receptor pair on two different cells in C. elegans (Prömel et al., 2012). However, since it is conceivable that Teneurins have multiple functions beyond their role in neurons, it cannot be fully excluded that for some other function, a classical interaction with Latrophilins is required. Further, the second Latrophilin homolog in C. elegans, lat-2, has not been functionally characterized yet and thus, might also be a candidate for a partnership with TEN-1.

# AN EVOLUTIONARY VIEW ON LATROPHILINS AND TENEURINS POINTS TOWARD A YOUNG INTERACTION

Due to their similar expression and function in vertebrates and invertebrates and their high conservation it has been speculated that the Latrophilin-Teneurin interaction also exists in invertebrates (Woelfle et al., 2015). Indeed, a high general sequence conservation of Teneurins from choanoflagellates to vertebrates has been found (Tucker and Chiquet-Ehrismann, 2006) together with structural conservation of core folds and several domains (Jackson et al., 2018; Li et al., 2018). Parts of the Teneurin N-terminus are probably derived from an evolutionarily ancient YD-repeat shell domain that is widespread across the bacterial kingdom by horizontal gene transfer into an early metazoan genome (Jackson et al., 2018). The EGF domains of the Teneurin N-terminus appear first in multicellular animals. Further, comparison of the gene organization among human Ten1, Drosophila Ten-a and Ten-m and the C. elegans ten-1 revealed the presence of both, conserved intron locations and exon sequences (Minet and Chiquet-Ehrismann, 2000; Tucker et al., 2012), suggesting that Teneurins arose from a single ancestral gene. This high structural and sequence conservation points toward comparable functions of Teneurins in similar molecular contexts in different species. However, a closer look at the evolution of Latrophilins can cast doubt on the hypothesis that the interaction of Teneurins with Latrophilins is old.

# Invertebrate Latrophilins Are Not One-to-One Orthologs of Mammalian Latrophilins

The class of Adhesion GPCRs belongs to the oldest GPCRs and their sequence signatures in the 7TM part appear first in unicellular organisms such as Dictyostelium discoideum and fungi (Krishnan et al., 2012). It needs to be noted that the appearance of genes in unicellular organisms should be taken with caution in the analysis of evolutionary history of gene families due to the possibility of horizontal gene transfer. However, in evolutionarily basal animals such as placozoa (Trichoplax adhaerens) and choanoflagellates (Salpingoeca rosetta and Monosiga brevicollis) there is already a number of Adhesion GPCR-encoding genes indicating their stable integration into animal genomes. Due to high sequence distances it is hard to assign them to Latrophilins or to another of the eight distinct groups of vertebrate Adhesion GPCRs (Nordstrom et al., 2011). Furthermore, none of these evolutionarily old Adhesion GPCRs have been found to present themselves with an RBL-, an OLF-, or an HRM domain (Krishnan et al., 2012), which have been suggested to interact with Teneurins (Woelfle et al., 2015). Therefore, it is rather unlikely that

a functional paring of Adhesion GPCRs and Teneurin-like proteins, as described in vertebrates, was already established at this early evolutionary stage, although Teneurin-like proteins are present in placozoa and choanoflagellates.

In the genomes of the roundworm C. elegans and the fruit fly D. melanogaster, two and one Latrophilin genes, respectively, have been assigned based on sequence similarities in the N-terminus and the 7TM domain. Re-evaluation of the already described phylogenetic relationship of these invertebrate and vertebrate Adhesion GPCRs (Schioth et al., 2010) revealed a more complex picture placing the 7TM domains of the C. elegans Latrophilins LAT-1 and LAT-2 basal to both, the vertebrate Latrophilin (ADGRL) and EMR (ADGRE) groups (**Figure 3**). In tunicates and evolutionarily old chordates such as lancelet (Branchiostoma belcheri) there are obviously no orthologs or paralogs of the ADGRE group, which contains EMR1- 4 (ADGRE1-4) and CD97 (ADGRE5) (**Figure 3**). However, as these can be found in fishes, one can assume that the ADGRE group evolved from the ADGRL group [containing besides LPHN1-3 also ELTD1 (ADGRL4)] in early vertebrate evolution or, alternatively, but more unlikely, was eliminated from all invertebrates. Therefore, the 7TM of LAT-1 and LAT-2 from C. elegans and other invertebrates are not in oneto-one orthology to vertebrate Latrophilins but rather share

FIGURE 4 | Domain assembly of invertebrate Latrophilins. The N-terminus domain composition of invertebrate Latrophilin-like sequences are shown. Putative conserved domains have been detected with the algorism implemented in NCBI BLAST (Marchler-Bauer et al., 2017). Domain names are given in the box. Note that most of the Latrophilin-like sequences are predicted from genome assemblies, which may contain errors, and are not supported by mRNA data. Species are: sk, Saccoglossus kowalevskii (Hemichordata); ci, Ciona intestinalis (Tunicata); ac, Acanthaster planc (Echinodermata); sp, Strongylocentrotus purpuratus (Echinodermata); my, Mizuhopecten yessoensis (Mollusca); cg, Crassostrea gigas (Mollusca); ob, Octopus bimaculoides (Mollusca); dm, Drosophila melanogaster (Insecta); la, Lingula anatina (Brachiopoda); ct, Capitella teleta (Annelida); ce, Caenorhabditis elegans (Nematoda); hd, Hypsibius dujardini (Tardigrada); rv, Ramazzottius varieornatus (Tardigrada); of, Orbicella faveolata (Cnidaria); pd, Pocillopora damicornis (Cnidaria); aq, Amphimedon queenslandica (Parazoa).

phylogenetic relation to all members of both groups including Latrophilins, ELTD, EMRs, and CD97. Further, based on substitution rates, C. elegans lat-1 and lat-2 and most invertebrate Latrophilin-like sequences are even more distantly related to the vertebrate Adhesion GPCR groups ADGRL and ADGRE than the C. elegans muscarinic acetylcholine receptors gar-1/-2/-3 to their vertebrate orthologs/paralogs (**Figure 3**). Most interestingly, the fruit fly Latrophilin dCirl is even more distantly related to the ADGRL group being placed closer to the Latrophilin-like sequences of Cnidaria and Parazoa and other Adhesion GPCR groups (**Figure 3**). Phylogenetic relation built on the basis of the 7TM sequences provides only weak support considering dCirl a member of the Latrophilin group at all. Even if the extracellular N-terminus and its modular composition presents with some structural features of the Latrophilin group, the very distant relation of the 7TM domain may explain differences in their G protein-mediated signal transduction in different species (Lelianova et al., 1997; Müller et al., 2015; Scholz et al., 2017; Nazarko et al., 2018). It has to be noted that already the five vertebrate muscarinic acetylcholine receptors (represented in the lilac triangle in **Figure 3**) differ in their signaling properties by coupling to Gq/<sup>11</sup> (mAChR-1, -3, -5) and Gi/<sup>o</sup> (mAChR-2, -4).

### The Postulated Teneurin-Latrophilin Interaction Sites Are Not Evolutionarily Old

Although the phylogenetic analyses on Latrophilins argue at least against the receptor binding to Teneurin and eliciting a conserved signal into the cells, it is still conceivable that an interaction between invertebrate Latrophilins and Teneurins occurs with Latrophilins acting as ligands for Teneurins. The interaction of Latrophilins and Teneurins is mediated by their N-termini and, taking this thought further, one can hypothesize that the 7TM is only modularly attached mediating the appropriated intracellular signal in the different species. As already seen in **Figure 1**, the worm LAT-1/LAT-2 and the fruit fly dCirl N-termini do not contain an OLF domain and additionally, the HRM domain is missing in dCirl. Detailed analysis of the Latrophilin N-termini in currently available genomes revealed that the ensemble of RBL-, OLF-, and HRM domains in the N-termini of Adhesion GPCRs is found in tunicates (e.g., Ciona intestinalis) (**Figure 4**), in lancelet (Branchiostoma belcheri), and Chondrichthyes (Callorhinchus milii). In Hemichordata, Echinodermata, Mollusca, Nematoda, Arthropoda, Tardigrada, and Brachiopoda only the GAIN, RBL, and HRM domain (sometimes degenerated or absent) are mostly present (**Figure 4**), but none of these sequences contains an OLF domain. Interestingly, several invertebrate Latrophilins contain domains (e.g., EGF, Ig, LamG, and FN3) not seen in vertebrate Latrophilins (**Figure 4**), indicating a modular structure of these Adhesion GPCRs.

Analyses on Latrophilin-Teneurin interactions provide strong evidence that the main site of interaction is the RBL domain with contribution of a short sequence C-terminal of the domain (**Figure 2A**; Silva et al., 2011; Boucard et al., 2014). Although protein domain identification tools constantly assign RBL and HRM domains in Latrophilins, the amino acid sequence conservation is low (**Figure 5**). The

Frontiers in Neuroscience | www.frontiersin.org

matching templates pdb: c5afbA and pdb: c4dlqA, respectively. Again, the conserved cysteine (yellow) and other (red) residues are highlighted.

domain assignment is mainly based on conserved disulfide bond-forming cysteine residues keeping constant folds of the domains. The few other conserved residues mainly surround the conserved disulfide bonds (**Figure 5**). This suggests that these backbone structures provide the threedimensional scaffold of the RBL and HRM domains. The remaining amino acid residues most probably participate in specific functions of the two domains. One can speculate that these domains mediate low affinity interactions to proteins or compound or that the sequence variability is the result of a co-evolutionary process with an also variable interaction partner. Although it cannot be fully excluded that invertebrate Latrophilins interact with Teneurins, it does not appear to be likely based on the re-evaluation of existing data above.

#### CONCLUSION AND FUTURE PERSPECTIVES

Synapse formation is a highly complex and tightly regulated process and although several aspects have been already well understood, many details are still obscure. Latrophilins and Teneurins are both transmembrane proteins which have been described to have implications in synaptogenesis and synapse function. While for Teneurins this has been shown in vertebrate and invertebrate systems, a lot of information is still lacking for Latrophilins. However, a transsynaptic interaction of the two is essential for adhesion and synapse formation in mammals. The question of whether this interaction represents a common principle in the generation of synapses throughout phyla remains unanswered, mainly due to lacking experimental evidence, but is highly intriguing. Similarities in expression, seemingly functional redundancy in Drosophila and a general evolutionary conservation makes it tempting to conclude that this transsynaptic interaction is old and also meaningful in invertebrate species. However, a closer look at phylogenetic evidence and existing data sheds light on a different picture.

Our phylogenetic analyses indicate that, although basal metazoans already contain Adhesion GPCR, clearly Latrophilin-7TM-related sequences only appeared at the level of Eumetazoa and thus, later in evolution than Teneurins. The connected N-termini contain RBL- and HRM-like domains but not constantly. Further, the OLF domain only appears in the N-termini of Latrophilins in early chordate evolution. Although the conserved cysteine bonds and a few other conserved positions allow for assignments as RBL and HRM domains, most of the remaining sequence is highly variable in these domains. This indicates that the RBL and HRM domains in Latrophilins may have specific functions in the different species and/or underwent co-evolution with interaction partners rather than mediating evolutionarily conserved protein-protein interactions. This analysis yields some evidence that a conserved interaction of Latrophilins and Teneurins in invertebrates might not be likely. It cannot be excluded that additional, not yet identified interaction sites in Latrophilins exist, which represent highly conserved sequences. Further, the role of other proteins or molecules aiding or promoting the interaction cannot be evaluated. For instance, dystroglycans have been discussed to be part of a larger complex (Woelfle et al., 2015). However, if a physical interaction may occur, a potential signal elicited by the Adhesion GPCR is not comparable to signals transduced by mammalian Latrophilins as invertebrate Latrophilins, in particular the homolog in Drosophila, are not one-to-one homologs of mammalian Latrophilins, but also bear resemblance to other Adhesion GPCRs. This argument is further supported by experimental data highlighting distinct signaling cascades activated by Latrophilin homologs of different species: While mammalian LPHN1 can signal via G<sup>s</sup> or G<sup>i</sup> proteins (Müller et al., 2015; Nazarko et al., 2018), Drosophila dCirl activates G<sup>i</sup> proteins and C. elegans LAT-1 G<sup>s</sup> proteins. A functional evaluation of these different cascades will shed light on the impact of these cascades.

We cannot exclude an interaction between Latrophilins and Teneurins in invertebrates, however, the mode of interaction might be realized differently from their mammalian counterparts. While both groups of proteins have essential functions in invertebrates and the ones of Teneurins in particular are highly conserved roles across phyla, they might not realize this role via the help of Latrophilins. Invertebrates have less complex regulatory circuits and hence, different requirements for synapse formation and function. Thus, it would not be surprising that they utilize different mechanisms to establish and maintain synapses and their function.

Future analyses need to focus on gaining a better understanding of the physiological functions mediated by both, Latrophilins and Teneurins, in mammals and invertebrates. These will help understand similarities as well as differences in the function of each receptor in different contexts and aid the understanding of the molecular mechanisms underlying synaptogenesis and neuronal wiring in vertebrates and invertebrates. It will be highly interesting to gain information on the existence and composition of potential synaptic complexes involving Latrophilins and/or Teneurins. Further, identifying interaction interfaces of mammalian Latrophilins with Teneurins can be highly informative for the prediction and characterization of a potential interaction in other species.

#### AUTHOR CONTRIBUTIONS

TS and SP researched and wrote the manuscript.

# FUNDING

This work was supported by the European Social Fund and the German Research Foundation (DFG: FOR 2149/P02 and P04, SFB 1052/B6).

#### REFERENCES


an ancient fold for cell-cell interaction. Nat. Commun. 9:1079. doi: 10.1038/ s41467-018-03460-0


pair-rule gene ten-m, is a neuronal protein with a novel type of heparin-binding domain. J. Cell Sci. 112(Pt 12), 2019–2032.


ubiquitous G-protein-linked receptors. G-protein coupling not required for triggering exocytosis. J. Biol. Chem. 273, 32715–32724. doi: 10.1074/jbc.273.49. 32715


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Schöneberg and Prömel. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Latrophilin's Social Protein Network

#### J. Peter H. Burbach<sup>1</sup> and Dimphna H. Meijer<sup>2</sup> \*

<sup>1</sup> Department of Translational Neuroscience, UMCU Brain Center, University Medical Center Utrecht, Utrecht, Netherlands, <sup>2</sup> Department of Bionanoscience, Kavli Institute of Nanoscience Delft, Delft University of Technology, Delft, Netherlands

Latrophilins (LPHNs) are adhesion GPCRs that are originally discovered as spider's toxin receptors, but are now known to be involved in brain development and linked to several neuronal and non-neuronal disorders. Latrophilins act in conjunction with other cell adhesion molecules and may play a leading role in its network organization. Here, we focus on the main protein partners of latrophilins, namely teneurins, FLRTs and contactins and summarize their respective temporal and spatial expression patterns, links to neurodevelopmental disorders as well as their structural characteristics. We discuss how more recent insights into the separate cell biological functions of these proteins shed light on the central role of latrophilins in this network. We postulate that latrophilins control the refinement of synaptic properties of specific subtypes of neurons, requiring discrete combinations of proteins.

Keywords: latrophilin, synapse biology, developmental neuroscience, neurodevelopmental disorders, interaction networks

#### Edited by:

David Lovejoy, University of Toronto, Canada

#### Reviewed by:

Yuri Ushkaryov, University of Kent, United Kingdom Arturo Ortega, Center for Research and Advanced Studies (CINVESTAV), Mexico

#### \*Correspondence:

Dimphna H. Meijer d.h.m.meijer@tudelft.nl

#### Specialty section:

This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Neuroscience

Received: 21 March 2019 Accepted: 05 June 2019 Published: 26 June 2019

#### Citation:

Burbach JPH and Meijer DH (2019) Latrophilin's Social Protein Network. Front. Neurosci. 13:643. doi: 10.3389/fnins.2019.00643

#### INTRODUCTION

Brain circuits function by virtue of precise connections between nerve cells. These connections are the ultimate result of coordinated developmental processes, involving direct interactions between cells for correct positioning and organization of cell layers in the brain, guidance of outgrowing axons, and the formation and shaping of synaptic contacts between them. Central to these processes are cell adhesion molecules which serve the communication and interaction between cells and thus have a key role in creating and tuning precisely-wired neural circuits.

During evolution, cell adhesion molecules have been instrumental in organizing multicellularity, thereby undergoing extreme diversifications (Abedin and King, 2008). These diversifications have been established through extensive variation of a limited number of structural amino acid motifs and protein domains. Based on structural characteristics, cell adhesion molecules have been accordingly classified in vast superfamilies such as cadherins and Ig domain cell adhesion molecules (IgCAMs). Besides these, families with fewer members and atypical adhesion domains have also been recognized, often serving more refined functions in specifying precise connections between nerve cells in specialized circuits, latrophilins (LPHNs) being one of them.

In cell adhesion, specificity is based on the nature of at least two partnering cell adhesion molecules. There is an extensive repertoire of interactions between cell adhesion molecules that forms the basis of interaction networks. Cell adhesion molecules can reside on contacting cells and interact in trans, or form a complex in cis on a single cell before partnering in trans. Furthermore, cell adhesion molecules can act in combination with an identical partner (homophilic complex), a different partner (heterophilic complex) or with multiple partners (multiprotein complex). While some cell adhesion molecules display strict specificity toward partners, others are more promiscuous. These properties together with the extensive diversity of cell adhesion molecules

**143**

provide a shear endless combinatorial potential. It has been postulated by Thomas Südhof that the diverse and multifold protein-protein interactions constitute molecular codes that drive formation, stability and dynamics of synaptic contacts which are required for the precision of neural circuitry formation (Sudhof, 2017, 2018).

In this context, we will explore principles of neuronal cell adhesion by focusing on the family of latrophilins that display multimodal interactions. The family of latrophilins itself has already been reviewed extensively (Meza-Aguilar and Boucard, 2014). Instead, we zoom in on well-established partners of the latrophilins, particularly teneurins, fibronectin leucinerich repeat transmembrane proteins (FLRTs) and contactins. We synopsize temporal and spatial expression profiles in combination with structural characteristics that together allow these interactions, and we discuss their functional consequences.

#### INTRODUCING LATROPHILINS

Latrophilin has initially been discovered as the Ca2++ independent receptor for alpha-latrotoxin, which is one of the toxic substances in the widow spiders' venom (Krasnoperov et al., 1997; Lelianova et al., 1997; Sugita et al., 1998). Fast-forwarding to two decades later, the latrophilin family is now known to contain three family members (LPHN1-3), of which all three are classified as adhesion G-protein coupled receptors (GPCRs) and linked to neuronal and non-neuronal disorders including ADHD and cancer (reviewed in Meza-Aguilar and Boucard, 2014). Furthermore, it has been shown that LPHN2 and LPHN3 are highly expressed in specific brain areas, whereas LPHN1 is detected at lower levels but more ubiquitously distributed throughout the brain (Sugita et al., 1998; Kreienkamp et al., 2000). Interestingly, in rodent brain LPHN1 levels are low during early postnatal development and increase with age, whereas LPHN2 shows the opposite pattern (Kreienkamp et al., 2000; Boucard et al., 2014). Recent data for LPHN3 show that protein expression peaks at approximately P12, when synaptogenesis is taking place (Sando et al., 2019). In contrast, peaks in LPHN3 mRNA levels were seen very early during rat postnatal development (Kreienkamp et al., 2000) as well as at later stages in the developing mouse brain (Boucard et al., 2014). Finally, the repertoire of endogenous ligands/interacting partners of latrophilins has been expanded to four different families, namely neurexins, teneurins, FLRTs and contactins (see **Figure 2A**). In this review we will focus on well-described interactions with teneurins and FLRTs, and its most recently discovered interacting partner Contactin6 (CNTN6). Neurexins are not considered here, since their interaction with latrophilins has been questioned and downplayed (O'Sullivan et al., 2014; Sudhof, 2018).

#### INTERACTION WITH TENEURINS

In the search for ligands of latrophilins, members of the type II transmembrane teneurin family of cell adhesion molecules were the first latrophilin-interacting proteins to be identified (Silva et al., 2011). Teneurins are non-classical cell adhesion molecules that may well have functions beyond simple cell adhesion.

#### Expression

The teneurin transmembrane proteins (TENM) family members display specific developmental and topographical expression patterns in the mammalian brain. During embryonic development of the mouse central nervous system (CNS), TENM3 and TENM4 are expressed as early as E7.5, followed by expression of TENM2 around E10.5. TENM1 expression starts later, at E15.5 (Zhou et al., 2003). At that embryonic timepoint, all four teneurins are expressed in the telencephalon and diencephalon with partial overlapping expression (Bibollet-Bahena et al., 2017). Later during embryonic development, TENM2 is additionally expressed in the midbrain, as well as in the nasal cavity and TENM3 shows prominent expression in the developing whisker pad (Zhou et al., 2003; Young et al., 2013). In the adult mouse brain, this diverging – but partially overlapping – expression pattern is maintained. For instance, all four teneurins are highly expressed in the CA1 region of the hippocampus, but the CA2 region expresses TENM2, TENM3, and TENM4 at very low levels, while CA3 expresses only TENM2 and TENM4 at appreciable levels, and the dentate gyrus (DG) expresses TENM1 and TENM2, as based on single-cell transcriptomics (see **Figure 1**; Habib et al., 2016). Earlier papers have reported variations on this pattern (Ben-Zur et al., 2000; Zhou et al., 2003; Berns et al., 2018).

### Function

Early observations in fly already demonstrated that teneurins play a functional role in circuitry formation, specifically between olfactory receptors neurons and projection neurons in the olfactory circuitry, and also in formation of the neuromuscular junction (Hong et al., 2012; Mosca et al., 2012). Intriguingly, a mutation in TENM1 has now indeed been associated with the neurological disorder congenital general anosmia, characterized by the loss of olfaction (Alkelai et al., 2016). A more detailed understanding of teneurin functions in the developing and adult CNS is steadily emerging with the overarching concept that teneurins are required for specific targeted projections in multiple brain circuits. Currently, no functional studies have been published on the role of TENM1 in the CNS, and only one group has reported functional experiments on TENM4 (Suzuki et al., 2012). This study demonstrates that oligodendrocyte differentiation is stalled in the absence of TENM4, which results in a tremor-like phenotype in TENM4-null mice. Notably, Hor et al. identified three missense mutations in the human TENM4 gene (also called ODZ4) that are associated with patient families displaying Essential Tremor movement disorder (see **Table 1**; Hor et al., 2015).

Considerably more functional work has been published on the role of TENM2 and TENM3 in the striatum, the visual cortex, and the hippocampus. TENM3-null mice have been reported to show defects in the thalamostriatal pathway, the retinal ganglial cell (RGC) to superior colliculus (SC) connections and retina to dorsal lateral geniculate nucleus (dLGN) connections

(Leamey et al., 2007; Dharmaratne et al., 2012; Tran et al., 2015). Similar abnormalities were noted in the TENM2-null animals, where a reduced number of RGCs project to the SC and dLGN (Young et al., 2013). Antinucci et al. have demonstrated that TENM3 is also essential for connections between RGCs and the optic tectum (homologous to the mammalian SC) in fish (Antinucci et al., 2013). In fact, in absence of TENM3, the animals were less able to detect shapes or position stimuli, known as orientation-selectivity. A function for TENM3 in wiring the visual system is substantiated by human genetics research in microphthalmia disease. Patients with microphthalmia have abnormally small eyes that are functionally impaired. Thus far, two patients have been identified with homozygous mutations in the TENM3 gene. These mutations result in a premature stop codon such that TENM3 is only partially translated (T695Nfs<sup>∗</sup> 5 and V990Cfs<sup>∗</sup> 13, see **Table 1**; Aldahmesh et al., 2012; Chassaing et al., 2013; Singh et al., 2019).

follow-up studies, involving for instance directed qPCR, are required when pursuing such hypotheses.

Most recently, an important role for TENM3 in the hippocampus has been reported (Berns et al., 2018). Berns and coworkers showed that TENM3 expression in the CA1 region and in the distal subiculum is required for connectivity between these two hippocampal regions. Using advanced mouse genetics they showed that axonal as well as dendritic teneurin is required in the connecting synapse to establish correctly-wired hippocampal circuitry. It should be noted that the much broader expression of teneurin proteins in the embryonic and adult CNS, and association with a variety of disorders (see **Table 1**), warrants additional functionality in other brain areas yet to be discovered.

#### Structure

The amino acid sequence of the extracellular region of teneurins is 59–71% identical between teneurins, and also the predicted domain organization is highly comparable. Structures of human and chick TENM2 and mouse TENM3 show that the extracellular region is folded into a large barrel-shaped structure, termed YD-shell, adorned with a beta-propeller perpendicular to the YDshell (see **Figure 2B**; Jackson et al., 2018; Li et al., 2018). The barrel is sealed by a so-called fibronectin plug domain and capped by its own inward spiraling C-terminal. This C-terminal end aligns with the barrel wall and threads out through a gap in the barrel to form two additional domains, the ABD and Tox-GHH domains. So far, only the beta-propeller and the C-terminal domains have been implicated in protein-protein interactions. The barrel itself shows striking similarities with the bacterial toxin system TcB, TcC of Y. enteromophaga and P. luminscencens, and teneurinlike protein-coding genes have been identified in several other bacteria as well (Tucker et al., 2012; Jackson et al., 2018). In these bacterial systems, the barrel-containing protein is part of a much larger protein complex important for toxin injection. Although the similarity to bacterial toxin systems might lead to tempting speculations, the practical implications of the structurally similar YD-shell in mammalian teneurins remain unknown. A notable difference between the bacterial and mammalian teneurins is that in the case of mammalian teneurins, covalent dimerization is induced by a non-traditional EGF-repeat domain, that has not been observed in bacterial teneurins.

# Molecular Mechanisms: Teneurin – Latrophilin Interactions

Latrophilins are adhesion GPCRs and consist of a small intracellular domain, seven-pass transmembrane helices and a larger extracellular domain (ECD) with multiple protein motifs. The extracellular domain can be cleaved by autoproteolysis, possibly resulting in a conformational change (Hamann et al., 2015; Arac et al., 2016). The extracellular region contains the proteolytic GAIN domain and a hormone-binding (HRM) domain, followed by a glycosylated linker region and the olfactomedin-like domain as well as a rhamnose-binding lectin domain (see **Figures 2B,C**; Vakonakis et al., 2008;

transmembrane domain; EGF, epidermal growth factor-like; FN, fibronectin; NHL, NCL-1, HT2A, and Lin-41 repeat; YD, tyrosine and aspartate-rich repeat; ABD, Antibiotic-binding domain; Ig, Immunoglobulin; LRR, leucine-rich repeat; LNS, laminin, neurexin, sex-hormone binding globulin domain.

Arac et al., 2012; O'Sullivan et al., 2014). The lectin domain specifically interacts with the extracellular domain of teneurin. Although this domain is sufficient and necessary for binding, the full length ECD of latrophilin has a higher binding affinity for teneurin than lectin alone (Silva et al., 2011; Boucard et al., 2012).

Which domain on teneurin is required for the formation of this complex? Silva et al. demonstrated that the C-terminal fragment of teneurin containing only the ABD and Tox-GHH domains was able to bind full-length latrophilin. Furthermore, a deletion construct of TENM2 that is missing the ABD and Tox-GHH domains (referred to as Tox-like domain in Li et al., 2018) abrogated its capability to interact with latrophilin. Thus, the latrophilin – teneurin interaction might be mediated by the lectin and ABD with Tox-GHH domains, respectively.

Latrophilin is somewhat promiscuous in its teneurin partner choice. Whereas LPHN1 binds TENM2 as its highest affinity ligand (Silva et al., 2011; Boucard et al., 2012; Vysokov et al., 2016; Li et al., 2018), and vice versa (Silva et al., 2011), cellular binding assays reveal additional interactions between LPHN1 and TENM4 (Boucard). Furthermore, LPHN2 interacts with TENM2 and TENM4 (Boucard et al., 2014; Jackson et al., 2018), and LPHN3 can interact with all members of the TENM family (O'Sullivan et al., 2012; Boucard et al., 2014; Berns et al., 2018; Li et al., 2018; Sando et al., 2019). Notably, a splice insert in all three LPHNs (for mouse LPHN1, KVEQK – following Y131) as well as two splice inserts in TENM2 and TENM3 (for mouse TENM3, AHYLDKIVK following I,740 and RNKDFRH, following L1218) might both decrease binding affinity of one for another (Boucard et al., 2014; Berns et al., 2018; Li et al., 2018). The latrophilin splice insert does not affect binding to FLRT3 (this interaction is discussed in more detail below) (Boucard et al., 2014). Detailed structural information derived from X-ray crystallography and cryo-electron microscopy (cryo-EM) has provided some insight into the structural consequences of these splicing events. The splice insert in latrophilin is situated between the lectin and olfactmedin domain (Boucard et al., 2014; Jackson et al., 2015). In teneurins, the second splice insert is located in between the first and the second blade of the NHL domain, a 6 bladed beta-propeller. This insert promotes homophilic TENM2 and TENM3 interactions, with the splice insert itself forming a potential dimerization interface (Jackson et al., 2018; Li et al., 2018). Conversely, the

TABLE 1 | Genetic mutations impinging upon the structure of latrophilins, teneurins, FLRTs and CNTN6 associated with human disorders.


Only those mutations with structural consequences (stop, missense, nonsense) are listed, excluding copy number variations and polymorphisms. In case of CNTN6 and ASD, only mutations that are identified in more than 1 patient are listed. X, nonsense mutation; +, compound heterozygous patients; ICD, intracellular domain; TM, transmembrane domain; ADHD, attention deficit hyperactivity disorder; ASD, autism spectrum disorder. References in round brackets demonstrate lack of association between gene and disorder.

absence of this splice insert may increase the affinity of teneurin for latrophilin.

What about the functional consequences of latrophilin – teneurin complex formation? As noted previously, both latrophilin and teneurin contain transmembrane segments and complex formation has indeed been shown to occur on the membrane in NB2A cells and in HEK293T cells (Silva et al., 2011; Beckmann et al., 2013; Vysokov et al., 2016; Li et al., 2018). Furthermore, when the full-length proteins are expressed in two different cell populations, mixing and aggregation of these populations is induced, indicating that this interaction occurs in trans (Silva et al., 2011; Boucard et al., 2014; Berns et al., 2018; Li et al., 2018). In neurons, teneurin and latrophilin family members are both localized to the synaptic compartments. Their pre- and/or postsynaptic localization remains a topic of discussion (see **Figure 3A**). For instance, LPHN1 is mostly documented as pre-synaptic (Silva et al., 2011; Vysokov et al., 2016, 2018), however it is also part of the postsynaptic proteome of murine CNS (Collins et al., 2006). In addition, LPHN2 has been identified as a postsynaptic protein (Kreienkamp et al., 2000; Tobaben et al., 2000; Anderson et al., 2017), whereas LPHN3 has been observed in both compartments (Collins et al., 2006; Nozumi et al., 2009; O'Sullivan et al., 2012; Sando et al., 2019). Likewise, TENM2 and TENM3 have both been identified in presynaptic terminals (Nozumi et al., 2009; Berns et al., 2018; Li et al., 2018) as well as postsynapticcaly

(Silva et al., 2011; Berns et al., 2018). Trans-synaptic interactions have already been demonstrated for the high-affinity pair Latrophilin1 and Teneurin2 (Silva et al., 2011), as well as for homophilic Teneurin3 interactions (with the splice inserts) (Berns et al., 2018). Such interactions might be instrumental for neuronal outgrowth and synapse formation, as well as synapse maintenance.

Beside the well-documented intercellular effects on cell adhesion, complex formation has also been suggested to induce intracellular signaling events. Cells that express latrophilin respond to its natural ligand alpha-latrotoxin with an increase in intracellular Ca2<sup>+</sup> levels. Silva et al. have demonstrated that this response is enhanced when latrophilin-expressing cells are incubated with the ABD and Tox-GHH region of TENM2, using neuroblastoma cells (Silva et al., 2011). Also, in primary hippocampal neurons this C-terminal fragment of TENM2 induced calcium signaling (Silva et al., 2011). The relative contribution of the latrophilin – teneurin complex compared to TENM2 by itself remains to be tested. Furthermore, Li et al. (2018) show that the co-expression of LPHN1 and TENM2 tempers the levels of another second messenger, namely cAMP, both in experiments that mimic a cis-interaction as well as in a trans configuration set-up, using HEK293 cells. In these experiments, latrophilin1 by itself reduced cAMP levels slightly, whereas teneurin2 alone did not affect cAMP levels at all.

Most recently, in vivo evidence for the functional relevance of the LPHN3-TENM2 interaction has been provided by Sando et al. (2019). These authors demonstrate that a LPHN3 mutant that cannot bind TENM2, is unable to rescue a reduction in Schaffer collateral synaptic strength induced by LPHN3 deficient CA1 neurons in the hippocampus (Sando et al., 2019). Notably, a FLRT3-binding mutant of LPHN3 is also unable to rescue this phenotype (see also section Molecular Mechanisms: FLRT – Latrophilin Interactions and Discussion), indicating that TENM2 and FLRT3 binding are together required for LPHN3 function (Sando et al., 2019).

#### LATROPHILIN – FLRT INTERACTION

Members of the fibronectin leucine-rich transmembrane (FLRT) family of cell adhesion proteins are a second class of binding partners of latrophilins. This subfamily consists of three members, sharing the fibronectin III domain, and are part of a much larger group of proteins that all have the leucinerich repeat, including AMIGOs, LINGOs, LLRTMs, NLLRs, and others (Chen et al., 2006).

# Neurobiological and Developmental Functions

The original identification of the FLRTs stems from a screen for extracellular matrix proteins expressed in muscle cells (Lacy et al., 1999). Early expression and essential defects in mouse embryos indicated the important role of these proteins (Maretto et al., 2008). A defined neural role was readily assigned to FLRT2 and FLRT3. These proteins were found to be shed and to act in soluble form as repulsive cues for Unc5D-positive neurons in the mouse cortex (Yamagishi et al., 2011). Specifically, FLRT2-Unc5D interactions were further shown to direct radial migration of cortical neurons, whereas the homophilic FLRT3-FLRT3 interaction controlled tangential migration of these neurons (Seiradake et al., 2014). This dual role in corticogenesis may be a key mechanism of FLRTs in cortical folding. This is supported by the findings that accessory sulci are formed in the cortex of FLRT1/3-null mice, where also cortical migration defects were observed (Del Toro et al., 2017). Interestingly, species with low cortical expression of FLRT2 and FLRT3, such as man and ferret, develop folded cortices, further supporting this hypothesis. The finding that FLRTs are ligands of latrophilin and that they together instruct the development of excitatory synapses added a new dimension to the insight in the functions of FLRTs (O'Sullivan et al., 2012).

### Characteristics and Expression Profiles

Examining FLRT expression profiles in the mouse hippocampus through single-cell transcriptomics and anatomical localization databases shows that all FLRTs are expressed in the hippocampus, but with different characteristics (see **Figure 1**; Habib et al., 2016). FLRT1 is expressed highest in the CA1 region, less in the CA3 region and not in the CA2 and DG, while FLRT2 is expressed in the CA1 and CA2 region as well as in GABAergic interneurons (see **Figure 1**; Schroeder et al., 2018). FLRT3 has the most restricted hippocampal expression as it is predominantly expressed by DG and CA3 neurons (O'Sullivan et al., 2012). These patterns show that all neurons in the hippocampus express at least one type of FLRT protein. Moreover, examination of single-cell transcriptomics based on data from Habib et al. shows that co-expression of different combinations of FLRTs occurs in specific neuronal cell types (Habib et al., 2016). In the mouse cerebral cortex, the expression of FLRTs is low compared to the hippocampus. In fact, singlecell RNA seq studies indicate that FLRT1 and FLRT2 mRNAs are virtually absent in the cortex, while expression of FLRT3 is particularly expressed in a subset of GABAergic interneurons (Tasic et al., 2016).

#### Structure

The amino acid sequences of all three FLRTs is quite similar, with FLRT1 and FLRT3 being the most divergent with 59% overall identity. All three FLRTs are type I single pass transmembrane proteins with a ∼100 amino acid long intracellular region and an extracellular region

that compasses 10 leucine rich repeats (LRR) followed by a single fibronectin (FN) type III domain (see **Figures 2B,C**). The LRR domain is folded into an elongated, incurvated structure with inward facing beta strands and outwardlyextending loops (Seiradake et al., 2014). Links between FLRTs and diseases are virtually lacking as yet; there is one genomewide association study for Kallman Syndrom (with anosmia as its most distinguishing feature) identifying three patients with mutations in FLRT3, all located in the LRR domain (see **Table 1**).

#### Molecular Mechanisms: FLRT – Latrophilin Interactions

FLRT3 was first identified as the postsynaptic interaction partners of presynaptic LPHN3 using affinity chromatography followed by mass spectrometry (O'Sullivan et al., 2012). However, in light of the debated latrophilin localization (see section Molecular Mechanisms: Teneurin – Latrophilin Interactions), there is also some uncertainty about the trans-orientation of this protein interaction pair (Lu et al., 2015; see **Figure 3A**). On a structural level, a number of crystal structures of the FLRT – latrophilin complex reveal how the β-propeller-shaped olfactomedin domain of latrophilin is tightly bound to the incurvated surface of the LRR domain of FLRT (Lu et al., 2015; Ranaivoson et al., 2015; Jackson et al., 2016). Interestingly, in the tertiary FLRT2 – LPHN3 – Unc5D complex, a direct interaction is observed between the lectin domain of LPHN2 and Unc5D, mediated by a salt bridge between residue E105 in LPHN2 and R156 in Unc5D (Jackson et al., 2016). FLRTs interact with LPHNs with some specificity for certain family members: FLRT1 and FLRT3 interact with all three LPHNs, while FLRT2 interacts with LPHN3 (O'Sullivan et al., 2012; Jackson et al., 2015, 2016).

Insight into the neurobiological significance of the LPHN-FLRT interaction has been limited so far to one particular high-affinity partnership, namely FLRT3-LPHN3 (O'Sullivan et al., 2012). In neuronal cultures, reduction in FLRT or LPHN3 expression, or interference with the FLRT-LPHN interactions resulted in a decrease of the density of glutamatergic synapses (O'Sullivan et al., 2012). In a similar fashion, reduction of FLRT expression in vivo reduced the number of perforant-path synapses and the strength of glutamatergic transmission (O'Sullivan et al., 2012). Moreover, Sando and coworkers have demonstrated that a reduced number of Schaffer collateral synapses in LPHN3 transgenic mice is not rescued by FLRT3 or TENM2 binding-mutants of LPHN (Sando et al., 2019). Instead, they postulate that LPHN3 requires simultaneous FLRT3 and TENM2 interactions for its synaptogenic functions (Sando et al., 2019). Together, these findings may be prototypical for the neurobiological potential of FLRT interactions, however, more extended studies on these functions have not been published yet. An argument to suspect that FLRT interactions play a more generic role in shaping morphology and function of brain circuits comes from the work of de Wit group, showing that an array of cell adhesion proteins of the Leucine-rich repeat (LRR) family, including FLRT2, regulates synapse structure and function of CA1 pyramidal neurons (Schroeder et al., 2018).

# LATROPHILIN – CONTACTIN6 INTERACTION

The most recent protein found to interact with LPHNs is Contactin6 (CNTN6). CNTN6 is classified as an IgCAM and belongs together with its five paralogs to the mammalian contactin IgCAM subfamily. In CNTN6 pull-down experiments, latrophilin-family member LPHN1 appeared as one of the most prominent proteins bound to CNTN6 (Zuko, 2015, 2016a). This interaction was demonstrated to occur in cis, using cell-binding and cell-aggregation assays. In view of the identified CNTN6 – LPHN1 interaction, we here focus on CNTN6 and only discuss other CNTN members in that context.

#### Expression

CNTN6 expression is strongly regulated during mouse development with peak expression in early postnatal stages, as revealed by expression of a LacZ gene inserted in the mouse CNTN6 locus (Takeda et al., 2003). Brain regions with strong X-Gal staining were the accessory olfactory bulb, the anterodorsal thalamus, layer V of the cerebral cortex, inferior colliculus and the cerebellum. Further analyses based on in situ hybridization of CNTN6 mRNA confirmed CNTN6 expression in these areas, but also indicated a wider expression involving other areas, in particular the hippocampus and multiple cortical layers. Regarding the hippocampal area, CNTN6 expression is found in the CA1 and the hilus of the dentate gyrus (Zuko et al., 2016b). Sakurai et al. also observed CNTN6-immunoreactivity in the subiculum, the stratum lacunosum–moleculare of the CA1 region and confirmed the expression of CNTN6 in the hilus of the dentate gyrus (Sakurai et al., 2010). Examining data from single-cell RNA sequencing partly confirms these observations and indicates that the CNTN6-positive cells in the hilus may be interneurons (see **Figure 1**; Habib et al., 2016). Analyses of the mouse cortex showed presence of CNTN6 mRNA and protein in layers II/III and confirmed expression in layer V (Zuko et al., 2016b), with single-cell RNA sequencing revealing CNTN6 expression in multiple cell types (Tasic et al., 2016). Specifically, highest levels of CNTN6 transcripts were found in layer V pyramidal neurons, and vasointestinal peptide (VIP) and somatostatin (Sst)-expressing interneurons in the motor cortex (Tasic et al., 2016).

CNTN6 is highly expressed in the cerebellum and displays differential expression over lobules in the adult brain. For instance, CNTN6 is highly expressed in subpopulations of granule cells and in the molecular layer of lobule 1 to the rostral half of lobule 9, but expression in the distal region of lobules 9 and 10 is weak (Takeda et al., 2003). During the development of the cerebellum, the CNTN6 gene is first expressed in the Purkinje cells of lobules 9 and 10 and is followed by expression in the internal granule cells of all lobules. At P5 and thereafter, CNTN6

immunoreactivity was observed in the developing molecular layer and granule cell layer but not in Purkinje cells in lobule 23.

### Function

CNTN1 and CNTN2 are the prototypical members of the CNTN family. For over 20 years these proteins are known as important components in neuron-glia interactions and formation of the nodes of Ranvier (Peles and Salzer, 2000; Ascano et al., 2012). CNTN1 and CNTN2 have been demonstrated to regulate neuronal migration, axon guidance and the organization of specific subdomains in the nodes of Ranvier through cisand trans-interactions with distinct cell adhesion molecules (Mohebiany et al., 2014). CNTN1 and CNTN2 have been taken as examples of the principal functions and mechanisms of action for the other members of this family. However, the other members lack the essential functions in neuron-glia interactions in myelination. Although much less well characterized, these CNTN family members have prominently come forward in genetic studies on neuropsychiatric developmental disorders (see **Table 1**; Guo et al., 2012; Nava et al., 2014; Huang et al., 2017; Oguro-Ando et al., 2017). Further, phenotypes in null mutant mice have indicated functions of these CNTNs in the developing and mature brain (Shimoda and Watanabe, 2009).

Specifically CNTN6 has been recognized as a potential player in a number of neurobiological processes, mostly through human genetics, loss-of-function studies in mice and gain-of-function studies in vitro. Human genetics has shown the association of CNTN6 variants, mostly copy number variations, with neuropsychiatric conditions including autism spectrum disorder, hyperacusis, anorexia nervosa and Tourette syndrome (Huang et al., 2017; Oguro-Ando et al., 2017). In animal models, there is only a single study available at the behavioral level, reporting mild phenotypes in CNTN6-null mice (Takeda et al., 2003). These mice exhibited impaired motor coordination indicating cerebellar deficits. This finding may well relate to the neuroanatomical observations of developmental cerebellar expression of CNTN6 (see section Expression).

Several studies have reported neuroanatomical phenotypes of CNTN6-null mice. Sakurai and coworkers showed that in the hippocampus of CNTN6-null mice CNTN6 appears to affect glutamatergic but not GABAergic synapses based on reduced expression of VGLUT1 and VGLUT2 and unaltered expression of VGAT (Sakurai et al., 2010). Similarly, CNTN6 is involved in the development of glutamatergic neurons in the cerebellum. In particular, CNTN6 was shown to colocalize with presynaptic marker VGLUT1 in parallel fibers that synapse on Purkinje cells (Sakurai et al., 2009). In the cortex of the same mouse strain, a modest shift in the numbers of subtype-specific projection neurons and interneurons in the visual cortex was observed (Zuko et al., 2016b). Furthermore, Ye et al. noted misorientation of the apical dendrite of pyramidal neurons in the visual cortex, particularly in layer V, in CNTN6-null mice (Ye et al., 2008). Combined loss-of-function of CNTN6 and one of its interaction partners CHL1, a neural IgCAM, dramatically aggrevated this dendritic phenotype. Interestingly, it was demonstrated that both CNTN6 and CHL1 interacted with protein tyrosine phosphatase α (PTPα, PTPRA), which is highly abundant in the brain (Kaplan et al., 1990; Ye et al., 2008). The authors proposed a signaling complex in which PTPα is downstream of CHL1 and CNTN6 and which regulates apical dendrite projections in the developing cortex (Ye et al., 2008).

Additional data have supported a role of CNTN6 in neuronal outgrowth and survival. For instance, the formation and terminal branching of the corticospinal tract is delayed in CNTN6-null mice, and neurite growth and neuronal survival is impaired in CNTN6-null mice with cerebral ischemia, aggrevating ischemic damage (Huang et al., 2011, 2012). Furthermore, a neuronal outgrowth role was also revealed in another condition of neurotrauma, namely spinal cord injury. In these mice, the regrowth of corticospinal axons was stimulated in the absence of CNTN6 protein or when CNTN6 was downregulated by shRNA (Huang et al., 2016).

From a molecular perspective, CNTN6 protein has also been characterized as a ligand for the receptor protein NOTCH. CNTN6 binds to NOTCH1, induces the cleavage and nuclear translocation of the NOTCH intracellular domain and subsequently, drives the expression of NOTCH1 target genes such as HES1 (Cui et al., 2004). CNTN6-mediated Notch activation was proposed to serve the differentiation of oligodendrocytes (Cui et al., 2004).

### Structure

All CNTNs are composed of six Ig domains followed by four fibronectin-III (FnIII) domains, anchored to the membrane via a GPI-linker. Several of these domains have now been solved structurally using X-ray crystallography, but no fulllength structure of the extracellular segment is yet available. The structure of the first four Ig domains has been determined for chicken CNTN2, human CNTN2 and mouse CNTN4. In all three cases, the Ig domains fold into a typical horsehoelike configuration, in which the first Ig domain contacts the fourth Ig domain, and the second Ig domain interacts with the third Ig domain (see **Figures 2B,C**; Freigang et al., 2000; Mortl et al., 2007; Bouyain and Watkins, 2010). More recently, a structure of CNTN3 spanning the fifth Ig domain until the second FnIII domain, as well as the structure of the first three FnIII domains of all six contactins, was determined (Nikolaienko et al., 2016). Together, these structures reveal how a sharp bend between the second and third FnIII domain might induce a parallel orientation of the extended Contactin structure toward the cell surface. Mutations in the third and fourth FNIII domains as well as in the Ig3-Ig4 and Ig6 domains have been found in patients with ASD and hyperacusis, supporting the notion that all these structural domains contribute essentially to the functional properties of CNTN6 protein (see **Table 1**; Mercati et al., 2017). While CNTN6 has no transmembrane or intracellular regions, CNTN6 can still participate in synaptic signaling through multiple protein interactions. From in vivo and in vitro studies, a model for cis interactions of CNTN6 proteins has been put forward (Ye et al., 2008; Ye et al., 2011). In this model, CNTN6 is part of an axonal complex with CHL1 and PTPσ (PTPRS), a protein related to PTPα. The latter interaction is supported by crystallographic studies providing the structural basis of the interaction of CNTN6 with PTPγ (PTPRG), another member

of the PTP family (Bouyain and Watkins, 2010; Nikolaienko et al., 2016). In addition, CNTNs might also form homodimers that could result in interactions in trans (Huang et al., 2016). Thus, involvement of CNTN6 in neurobiological processes might require multimodal cis and trans interactions, that we are now only starting to unravel.

#### Molecular Mechanisms: Contactin – Latrophilin Interactions

How can CNTN6 complex with latrophilins? Using cellular aggregation assays, a cis-LPHN1-CNTN6 complex is more strongly supported than a complex in trans (Zuko et al., 2016a). Thus far, no experiments have been performed to map the interacting domains or residues. As such, it is difficult to predict the architecture of the complex. For a more distant family member of the contactins, namely Neurofascin, it has been show that its FN domain interacts with gliomedin, containing multiple olfactomedin domains (Labasque et al., 2011). Thus, by comparison we could speculate that the FN domains of CNTNs are likely to interact with the olfactomedin domain of LPHN1. On the other hand, FN domains can also interact with lectins, which suggests the possibility of an interaction between the FN domains of CNTNs with the lectin domain of LPHN1 (Praetorius et al., 2001). In addition, the FN domains of CNTN5 have been assigned as interaction sites with amyloid precursor-like protein-1 (APLP1) (Shimoda et al., 2012). Clearly, additional data are needed to understand the structural basis of this complex.

Functionally, it has been shown that CNTN6 and LPHN1 indeed modulate each other's activity. In neuronal cultures, LPHN1 overexpression resulted in an increase of apoptosis, which was blocked by co-expression of CNTN6 (Zuko et al., 2016a). Notably, overexpression of CNTN6 by itself had no effect on neuronal morphology or survival. In contrast, in cultured neurons, as well as in cortical tissue derived from CNTN6-null mice, enhanced apoptosis was observed (Zuko et al., 2016a). This was counteracted by shRNA-mediated LPHN1 knockdown. These results indicate a context-dependent functional interaction between CNTN6 and LPHN1. Future work is needed to resolve in greater detail how this interactions controls apoptosis in neuronal cultures, as well as in the in vivo setting.

Future research might reveal functional interactions in additional brain areas, since directed in situ hybridization experiments with LPHN1 and CNTN6 revealed co-expression in the thalamic nuclei, cortical layer V, hippocampal area CA1, and in the granular cell axons of the molecular layer in the cerebellum, pointing to the possibility that in these regions functional interactions can occur (Malgaroli et al., 1989; Zuko et al., 2016a).

# DISCUSSION

Trans-synaptic interactions between cell adhesion molecules have been identified as essential elements for synapse formation and plasticity. These processes are ruled by combinatorial codes of cell adhesion molecules, which is illustrated by the complexity and multifold interactions of proteins encoded by the neurexin genes (Sudhof, 2017). The mechanisms uncovered for neurexins are pivotal when considering functions of multimodal interactions of latrophilin as reviewed here.

The neurexin family of cell-adhesion proteins consists of thousands of isoforms of transmembrane proteins encoded by three separate genes. Neurexins are expressed by neurons all over the nervous system and their expression is already initiated during brain development before synaptogenesis occurs. Neurexins have a presynaptic localization and have been extensively characterized for their central organizing roles in synapse formation, maintenance and plasticity (Sudhof, 2017, 2018).

We postulate here an analogous organizing role for latrophilins, although the extent of this role is more limited than that of neurexins in view of the more restricted expression of LPHN2 and LPHN3. Furthermore, co-expression of latrophilins with established partners suggests that specific combinations exist in small subsets of neurons only, rather than in global neuronal populations as is the case for neurexins. Some of these combinations may be more widely occurring, like interactions with teneurins, than with others. This argues against a general role of latrophilin interaction networks, but rather points toward a role in refining synaptic properties of specific subtypes of neurons, requiring specific combinations of proteins. The current data start to reveal what this refinement may imply and what neuronal subtypes employ latrophilin interaction networks.

Here, we have reviewed temporal and spatial expression patterns of its protein partners teneurins, FLRTs and CNTN6 (see **Figure 3B**). Discrete expression in time together with cell-type specificity determines which interacting partners are available for complex formation. For instance, during late embryonic brain development, interactions with LPHN2 are less likely to play an important role due to very low to absent protein expression. Beyond temporal and spatial availability, specificity of interactions is also generated by splicing events, exemplified by the interaction between latrophilin and teneurin. Furthermore, although no data are available yet on posttranslational modifications (PTMs) in latrophilins or its partners, PTMs in general are well-known to determine binding specificity and affinity of interactions. A type of PTM that is especially of importance for extracellular interactions is modification by glycans, also known as glycosylation. In fact, N-linked glycans are now known to be highly abundant and particularly variable in synaptic proteins (Trinidad et al., 2013). For instance, latrophilins are predicted to be decorated with as many as 7 N-linked glycans and harbor an O-linked sugar-rich region in between the HRM and OLF domains (O'Sullivan et al., 2014). Future studies are needed to test their importance in protein-protein interactions. The exact location and identity of these glycans can be mapped using a combination of mass spectrometry and structural biology techniques. An elegant example of how N-glycans impact cell adhesion complexes is the presence of three glycosylated residues in SynCAM that regulate adhesion (Fogel et al., 2010).

What about the functional consequences of latrophilin interactions? Evidence for a functional role of latrophilincentered protein networks comes from an in vivo transgenic study, as well as cellular assays on the LPHN3 – FLRT3 – TENM2 network (Sando et al., 2019). Neither of the latrophilin mutants

that are incapable of FLRT3 or TENM2 binding could rescue the latrophilin-induced decrease in Schaffer collateral synaptic strength in vivo (Sando et al., 2019). The requirement of both binding sites for latrophilin function indicates that multimodality might be essential in latrophilin-instructed synaptogenesis. In addition, the finding that LPHN3 mutations associate with ADHD indicates an important functional role in humans (see **Table 1**). Also for teneurins, FLRTs and CNTN6 human genetic data indicate specific defects to be associated with these interactors. It will be essential to determine which of these relate to partnering to latrophilins, and what other, still unknown partners are involved in these phenotypes.

# OUTLOOK

For integration of spatial and temporal expression patterns, splicing events and PTMs of latrophilin and its protein partners, high-resolution imaging while maintaining temporal and spatial information is desired. Whereas previous insights mostly involved freeze substitution electron microscopic tomography, techniques such as cryo-electron microscopy and cryo-electron tomography are now expected to produce high resolution structures in cellular contexts (Lucic et al., 2005; Zuber

#### REFERENCES


et al., 2005; High et al., 2015; Perez de Arce et al., 2015). The follow-up, understanding of the precise function of such protein networks will still be an enormous endeavor, but we may expect that on the way we will be able to recognize novel neurobiological mechanisms that are inherent to the latrophilins.

# AUTHOR CONTRIBUTIONS

Both authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

# FUNDING

DM was supported by a Netherlands Organization for Scientific Research Veni grant (722.016.004).

# ACKNOWLEDGMENTS

We gratefully acknowledge helpful conversations with Dr. C. P. Frias (Technical University Delft, the Netherlands).


and multiprotein complexes in the postsynaptic proteome. J. Neurochem. 97(Suppl. 1), 16–23. doi: 10.1111/j.1471-4159.2005.03507.x


mode of homophilic interaction. Protein Sci. 16, 2174–2183. doi: 10.1110/ps. 072802707


orientation in the neocortex via PTP alpha. EMBO J. 27, 188–200. doi: 10.1038/ sj.emboj.7601939


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Burbach and Meijer. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Teneurins: Mediators of Complex Neural Circuit Assembly in Mammals

#### Catherine A. Leamey\* and Atomu Sawatari

Discipline of Physiology, School of Medical Sciences and Bosch Institute, Faculty of Medicine and Health, The University of Sydney, Sydney, NSW, Australia

The teneurins (Ten-m/Odz) are a family of evolutionarily ancient transmembrane molecules whose complex and multi-faceted roles in the generation of mammalian neural circuits are only beginning to be appreciated. In mammals there are four family members (Ten-m1-4). Initial expression studies in vertebrates revealed intriguing expression patterns in interconnected populations of neurons. These observations, together with biochemical and over-expression studies, led to the hypothesis that homophilic interactions between teneurins on afferent and target cells may help to guide the assembly of neural circuits. This review will focus on insights gained on teneurin function in vivo in mammals using mouse knockout models. These studies provide support for the hypothesis that homophilic interactions between teneurin molecules can guide the formation of neural connections with largely consistent results obtained in hippocampal and striatal circuits. Mapping changes obtained in the mouse visual pathway, however, suggest additional roles for these glycoproteins in the formation and specification of circuits which subserve binocular vision.

#### Edited by:

Richard P. Tucker, University of California, Davis, United States

#### Reviewed by:

Robert Hindges, King's College London, United Kingdom Timothy Mosca, Thomas Jefferson University, United States

#### \*Correspondence:

Catherine A. Leamey catherine.leamey@sydney.edu.au

#### Specialty section:

This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Neuroscience

Received: 22 March 2019 Accepted: 22 May 2019 Published: 05 June 2019

#### Citation:

Leamey CA and Sawatari A (2019) Teneurins: Mediators of Complex Neural Circuit Assembly in Mammals. Front. Neurosci. 13:580. doi: 10.3389/fnins.2019.00580 Keywords: Ten-m/Odz/teneurin, visual pathway, chemoaffinity, development, hippocampus, striatum, neural circuits

# INTRODUCTION

The idea that groups of afferent and target neurons positioned at locations remote from each other could set up precise, ordered patterns of connectivity due to the affinity of chemicals expressed on or by these cells was postulated formally by Roger Sperry in his chemoaffinity hypothesis (Sperry, 1963). Over the last few decades, a few families of molecules that exhibit expression patterns which fit largely with his predictions have been identified, with notable examples including the Ephs/ephrins, cadherin and immunoglobulin superfamilies (McLaughlin and O'Leary, 2005; Zipursky and Sanes, 2010). One of the more recent entrants to this stage is the teneurins. In a range of different species and brain circuits, these molecules were found to exhibit distributions across afferent and target fields which pointed to the idea that they may indeed help to determine patterns of neural connectivity (e.g., Rubin et al., 1999, 2002; Zhou et al., 2003; Li et al., 2006; Leamey et al., 2008; Carr et al., 2013, 2014; Bibollet-Bahena et al., 2017; Cheung et al., 2019). Over recent years, genetically modified mice have been generated which have enabled these ideas to be tested in vivo. The focus of this article is to review what has been learnt from these studies. As will be discussed below, they show key roles for teneurin molecules in regulating the patterns of connectivity in multiple neural circuits, including visual, hippocampal, and striatal networks. Compelling evidence that homophilic interactions between teneurins on axons and targets help to specify precise patterns of connectivity will be described. Evidence that teneurins also play other important roles in mediating appropriate wiring and synaptic

efficacy, including interactions with, and regulation of the expression of other molecules will also be presented.

### TOPOGRAPHICALLY CORRESPONDING GRADIENTS MEDIATE PRECISE MATCHING OF NEURAL CONNECTIONS VIA HOMOPHILIC INTERACTIONS

Teneurins exhibit differential expression patterns within neural circuits. In the chick visual system, for example, Ten-m1 and Tenm2 were found to be differentially expressed by the tectofugal and thalamofugal pathways, respectively (Rubin et al., 1999, 2002). Dynamic and differential, but partially overlapping, expression patterns have also been observed in the nervous system of zebrafish, with particularly strong expression of Ten-m3 and Ten-m4 (Mieda et al., 1999; Cheung et al., 2019). In this species, expression of Ten-m3 in the amacrine and ganglion cells of the developing retina is important for the formation of intraretinal circuitry (Antinucci et al., 2013, 2016). In addition to different Ten-ms being selectively expressed by specific pathways, topographically corresponding gradients of expression have also been observed at multiple levels within given circuits, suggesting a role in generating precise patterns of connectivity between remotely located afferent and target fields. The most notable examples of this are the expression patterns of Ten-m3 in the developing visual, hippocampal, and striatal circuits in mice.

Initial descriptions of the expression patterns of teneurins in the cortex of the mouse described high levels of Ten-m2, Ten-m3, and Ten-m4 in caudal regions of cortex, with Ten-m1 expressed in more rostral areas (Li et al., 2006; Leamey et al., 2008). Expression in the caudal domain included the primary visual cortex, multiple subregions of the hippocampus and associated cortical areas, as well as intriguing expression patterns in the thalamus and striatum (Zhou et al., 2003; Li et al., 2006; Leamey et al., 2008; Tran et al., 2015; Bibollet-Bahena et al., 2017).

While both Ten-m2 and Ten-m4 displayed fairly uniform expression across given subregions of the hippocampus, Ten-m3 showed evidence of differential expression within these areas (Zhou et al., 2003; Li et al., 2006; Leamey et al., 2008). Recent work has confirmed the presence of a gradient of Ten-m3 across three interconnected regions: CA1, subiculum, and entorhinal cortex (Berns et al., 2018). Further, the gradients of Ten-m3 are topographically aligned across these regions: medial entorhinal cortex, proximal CA1, and distal subiculum, all express high levels of Ten-m3 (**Figure 1A**), and are connected to each other. In contrast, the lateral entorhinal cortex, distal CA1, and proximal subiculum circuit are similarly interconnected and all exhibit low levels of Ten-m3 expression (Berns et al., 2018). Further, this paper showed that multiple other interconnected regions of the hippocampal circuit including the mammillary bodies, anteroventral thalamic nucleus, and pre- as well as parasubiculum also display gradients of Ten-m3 (Bibollet-Bahena et al., 2017; Berns et al., 2018).

A topographic correspondence in the expression patterns of Ten-m3 between connected areas has also been found for the thalamostriatal pathway (Tran et al., 2015). In the striatum, Tenm3 expression is patchy but distributed in an overall high dorsal to low ventral gradient within the matrix. A topographically corresponding high-dorsal to low-ventral gradient of Ten-m3 expression pattern is found in the parafascicular thalamic nucleus, a major source of input to the striatal matrix (**Figure 1B**). Interestingly, thalamostriatal terminals have a patchy distribution that overlaps with Ten-m3-positive regions (Tran et al., 2015).

The observation of a high caudal to low rostral expression gradient across the visual cortex (Leamey et al., 2008) sparked an investigation of other areas within this sensory pathway. The presence of a high ventral to low dorsal gradient of Ten-m3 across the retina has been revealed, including in retinal ganglion cells (RGCs), which provide output to central visual structures (Leamey et al., 2007). Even more compellingly, the two main topographically organized primary targets of RGC axons, the dorsal lateral geniculate nucleus (dLGN) and superior colliculus (SC), also exhibit graded expression patterns that are high in the areas which received input from ventral retina (dorsal dLGN and medial SC, respectively) and low in regions that are driven by dorsal retina (ventral dLGN and lateral SC) (Leamey et al., 2007; Dharmaratne et al., 2012; **Figures 1C**, **2A**). Interestingly, similar gradients were found in the marsupial wallaby, suggesting a conservation of Ten-m3 function across mammalian species widely separated by evolution (Carr et al., 2013, 2014).

The remarkable consistency in these patterns within a range of neural circuits across Mammalia pointed strongly to the idea that Ten-m3 may function as a classic chemoaffinity molecule, promoting connectivity between afferent axons and target cells with corresponding levels of expression. While in vitro studies have provided general support for this idea, showing that teneurins promote cellular adhesion in vitro (Rubin et al., 2002; Berns et al., 2018), proof of a role in mediating connectivity requires in vivo manipulations.

Analysis of the Ten-m3 knockout mouse, generated by deletion of exon 4 (Leamey et al., 2007), has demonstrated functionally important roles for this molecule in the formation of appropriate connectivity. Global Ten-m3 removal results in a loss of precision in thalamostriatal connectivity (**Figure 1E**). Consistent with the pattern of Ten-m3 expression, deletion of the active gene results in changes in the overall topography of the pathway, as well as inducing the normally tight clusters of thalamostriatal terminals observed in wild types (WTs) to become more diffuse (Tran et al., 2015). Subtle changes in the accuracy of contralateral retinocollicular projections have also been observed, with terminal zones exhibiting a narrowing across the mediolateral, as well as an elongation along the rostrocaudal axes of the SC (Dharmaratne et al., 2012; **Figure 1F**). In the hippocampus, the targeting of proximal CA1 to distal subiculum is less precise in global Ten-m3 KOs than in WTs (Berns et al., 2018; **Figure 1D**). This work provides good evidence that Ten-m3 acts homophilically to promote accurate connectivity between areas that express similar levels of the protein in a variety of

gray), with contralateral projections tending to terminate in slightly rostral and medial areas compared to ipsilateral termination zones in the opposite hemisphere. (D–F) Removal of Ten-m3 leads to axonal miswiring. (D) In the hippocampus, deletion of Ten-m3 results in CA1 projections exhibiting a greater spread of termination, targeting distal, as well as more proximal areas within subiculum. (E) Dorsolateral PFN projections terminate in more ventral striatal areas in Ten-m3 KOs. In addition, there is a loss of the patchy distribution of thalamostriatal terminals. Both hippocampal, as well as thalamostriatal connectivity exhibit changes consistent with the removal of a homophilic signal. (F) Ipsilateral retinal projections targeting the SC are also miswired, with terminals detected in more lateral, as well as posterior locations. Contralateral retinal termination patterns show only subtle alterations with terminal zones narrowed mediolaterally and elongated along the anterior–posterior axis. Thus, for the retinocollicular pathway, the change in wiring can only partially be explained by the removal of a homophilic gradient, suggesting other downstream factors are contributing to proper topographic mapping of this pathway (based on Dharmaratne et al., 2012; Tran et al., 2015; Berns et al., 2018).

circuits, consistent with the notion that it serves as a classic chemoaffinity molecule.

Several questions remain unanswered, however. Most notably, is Ten-m3 expression required in afferent axons, target cells, or both? The recent development of a conditional Ten-m3 KO mouse has enabled this issue to be elegantly addressed in the hippocampal circuit. Deletion of Ten-m3 in either the afferent or the target cells alone is sufficient to disrupt the usual pattern of connectivity between CA1 and subiculum (Berns et al., 2018) and reinforces the suggestion that homophilic interactions are important for Ten-m3 function. Of interest, the phenotype observed when Ten-m3 is deleted only from a subregion of subiculum suggests that Ten-m3-positive axons will avoid terminating in areas they would normally innervate to target Ten-m3-positive targets. Curiously, while there are multiple, differentially spliced variants of Ten-m3 expressed in the circuit, all except one is able to mediate homophilic interactions between cells, as well as cellular aggregation

(Berns et al., 2018). Interestingly, this exception was still able to promote adhesion with cells that express latrophilin-3, a molecule that acts by forming heterophilic bonds with Teneurins across synapses (Boucard et al., 2014; Sando et al., 2019).

# BEYOND HOMOPHILIC ADHESION: TENEURINS IN THE FORMATION OF BINOCULAR VISUAL CIRCUITS

The studies described above provide compelling evidence that the graded expression pattern of Ten-m3 is fundamentally important in promoting precise patterns of connectivity within neural circuits, supporting its role as a homophilic adhesion molecule. Evidence suggests, however, that this may be only one component of Ten-m3's function, at least for the formation of binocular visual circuits.

As noted above, topographically connected regions of the early visual pathway show similar expression levels of Ten-m3. Evidence of altered mapping is apparent in the contralateral retinocollicular pathway of Ten-m3 KOs (Dharmaratne et al., 2012). It should be pointed out, however, that these changes are quite subtle when compared to the much more dramatic miswiring observed in the mapping of both the ipsilateral retinocollicular (Dharmaratne et al., 2012) and retinogeniculate pathways in these mice (Leamey et al., 2007). The ipsilaterally projecting RGC population originates from a subset of cells in the peripheral ventrotemporal crescent (VTC) of the retina (Drager and Olsen, 1980), and usually projects exclusively to a patch in the dorsomedial region of the thalamic nucleus (**Figure 2A**). In Ten-m3 KO mice, it has been found that while the ipsilateral pathway originates from the same region of the retina as in WTs, its terminals form an elongated strip that extend from dorsomedial to far ventrolateral dLGN (Leamey et al., 2007; **Figure 2C**). Ipsilateral retinocollicular projections in KOs also exhibit highly aberrant wiring, with terminals normally confined to rostral and medial areas targeting more caudolateral locations (Dharmaratne et al., 2012; **Figure 1F**). The degree of change observed indicates a profound difference in the effect of

Ten-m3 on the targeting of ipsilateral and contralateral retinal projections.

Since the ipsilateral projection arises from, and projects to, regions associated with high levels of Ten-m3 expression, the changes observed in KOs are broadly consistent with the idea that this molecule promotes the formation of synaptic contacts between areas with similar expression levels, and thus helps to set up topographical alignment within the visual pathway. A critical look, however, suggests a more complex role. Notably, the expression of Ten-m3 has a broad ventrodorsal retinal gradient, but shows no difference between temporal and nasal regions (Dharmaratne et al., 2012). Further, Ten-m3 expression is clearly not restricted to the ipsilateral population, which comprises only a fraction of the RGCs within the VTC (Drager and Olsen, 1980). The much more pronounced effect of Ten-m3 deletion on ipsilateral mapping is therefore not well-correlated with its expression pattern. Further, although a loss of precision in the pattern of connectivity of ipsilateral projections is observed along the axis of Ten-m3 expression, much more pronounced changes are apparent along the axis that is orthogonal to the Ten-m3 gradient (Dharmaratne et al., 2012). These observations suggest that Ten-m3 may have additional mechanisms of action in the targeting of ipsilateral projections.

An investigation into the development of the retinogeniculate projection revealed that the mistargeting of ipsilateral RGC terminals in the dLGN in Ten-m3 KOs is preceded by an abnormally early exit of retinal axons from the optic tract into the nucleus (Glendining et al., 2017). In WT mice, ipsilateral axons remain largely confined to the optic tract until they reach the dorsal half of the dLGN. In Ten-m3 KO mice, however, the ipsilateral axons leave the optic tract to enter the nucleus near its ventral border. This change is difficult to explain as a direct consequence of the deletion of an attractive adhesion molecule expressed in the dorsal part of the dLGN. Even the demonstration that cleavage products of teneurins can form soluble proteins that impact axon guidance (Vysokov et al., 2018) does not really help here, as the attractive molecule has been deleted, yet the axons enter a region from which they would usually be repelled. The avoidance of ventrolateral dLGN by ipsilateral retinal axons has been shown to involve repellent interactions between EphA receptors and their ligands (Pfeiffenberger et al., 2005), which are also expressed in gradients in the retina, SC, and dLGN (Feldheim et al., 1998, 2000; Frisen et al., 1998). Intriguingly, the expression gradients of the EphA/ephrinA families are orthogonal to the Ten-m3 gradient, so an interaction with this pathway would help to explain both the axis and direction of change observed in the KO phenotype. While no changes in expression levels of most of the relevant EphA/eprhinA family members were detected, a significant reduction in the expression of the EphA7 receptor has been revealed in Ten-m3 KO mice on the day of birth (Glendining et al., 2017) which may help to account for the observed changes in ipsilateral termination. In vitro studies have shown that the intracellular domain of Ten-m2 may be cleaved and translocate to the nucleus where it interacts with transcription factors such as Zic1 (Bagutti et al., 2003). Since the intracellular domain of Ten-m3 contains both a potential

cleavage site and a nuclear localization signal (Tucker et al., 2012; Leamey and Sawatari, 2014), interactions with transcription factors seem likely. Indeed, a pull-down assay has demonstrated that the intracellular domain of Ten-m3 interacts with Zic2 (Glendining et al., 2017), a transcription factor which is a key determinant of ipsilateral identity (Herrera et al., 2003; Garcia-Frigola et al., 2008; Lee et al., 2008). Moreover, mRNA for both Zic2, and its downstream mediator of ipsilateral axonal guidance at the chiasm, EphB1 (Williams et al., 2003) are upregulated in Ten-m3 KOs (Glendining et al., 2017). Thus, a key component of the role of Ten-m3 in the formation of binocular visual circuits is likely to arise via the interaction with, and regulation of, other signaling molecules.

Interestingly, since Zic2 and EphB1 promote ipsilateral identity and retinal axon guidance, respectively (Herrera et al., 2003; Williams et al., 2003; Lee et al., 2008), the upregulation of these molecules seen in Ten-m3 KOs at around the day of birth (Glendining et al., 2017) could be expected to result in an increase in the size of the ipsilateral projection in Ten-m3 KOs. Retrograde tracing showed, however, no change in either the number or distribution of ipsilaterally projecting RGCs in adult mice lacking Ten-m3 (Leamey et al., 2007). It is possible that the upregulation of Zic2 in Ten-m3 KOs occurs too late to induce a change in the laterality of projections, although since Zic2 expression peaks at embryonic day 16.5 (Herrera et al., 2003) and Ten-m3 is usually expressed at high levels in the retina by this time (Dharmaratne et al., 2012), this possibility seems unlikely. Alternatively, rather than more cells expressing Zic2 and EphB1, the level of expression of these molecules within individual RGCs may be increased. If this was the case, given the known role EphB molecules in retinal mapping (Hindges et al., 2002; McLaughlin et al., 2003), an increase in EphB1 would be expected to cause ipsilaterally projecting retinal axons to map more laterally in the SC in Ten-m3 KOs than in WTs. Interestingly, this fits with what has been observed (**Figure 1F**; Dharmaratne et al., 2012).

As noted above, two other members of the Teneurin family, Ten-m2 and Ten-m4, are also highly expressed in the mouse visual cortex at around the time of birth (Li et al., 2006; Leamey et al., 2008). Analysis of Ten-m2 KOs has revealed a key role for this family member in the formation of binocular visual circuits which complements that of Ten-m3. Similar to what is observed in the hippocampal circuit, Ten-m2 does not display an obvious differential expression within the visual pathway that would suggest a role in topography. Rather, the molecule appears to be uniformly distributed across the RGC layer and within the SC, dLGN, and V1 in perinatal mice (Young et al., 2013). Despite this fairly uniform distribution pattern within the visual pathway (**Figure 2B**), analyses of the Ten-m2 KO yielded evidence of a highly specific defect which again impacted the formation of the binocular visual circuit. In Ten-m2 KO mice, the ipsilateral visual pathway is found to be reduced in size. Further, the loss of ipsilaterally projecting RGCs is only observed in the ventral part of the VTC (Young et al., 2013; **Figure 2D**). This highly specific role of Ten-m2, which does not correlate easily with its expression pattern, suggests that, as with Ten-m3, its role in the formation

of binocular visual circuits is likely to involve interactions with other molecules. Given its association with the ipsilateral pathway, potential changes in Zic2 and EphB1 have been investigated. While no difference in Zic2 is observed, a downregulation of EphB1 is seen specifically in the ventral part of the VTC, suggesting that Ten-m2 may work downstream of, or in parallel with, Zic2.

#### FUNCTIONAL IMPACTS FROM LOSS OF TENEURIN FUNCTION

The impact of the loss of teneurin function on behavior in KO mice has been best characterized for the visual pathway. In Ten-m3 KOs, the mistargeting of ipsilateral retinal axons is associated with profound visual deficits. Assessment of behavior that requires patterned vision reveals performance at chance levels, although the mice show an ability to distinguish between dark and light (Leamey et al., 2007). Interestingly, acute inactivation of neural activity in one eye significantly improves performance on tasks requiring patterned vision. This suggested that the misalignment of ipsilateral and contralateral visual inputs to one hemisphere may lead to a suppression of activity in V1. A subsequent investigation supported this by demonstrating that binocular, but not monocular, drive to V1 is significantly reduced in Ten-m3 KOs (Merlin et al., 2013). Similar mechanisms may, at least in part, contribute to the visual disorders associated with Ten-m3 mutations in humans (Aldahmesh et al., 2012). Moreover, the Ten-m3 KO is not the only teneurin model that exhibits defects of visual function. The loss of ipsilateral projections from the ventral part of the VTC in Ten-m2 KOs is also associated with a reduced ability to discriminate visual stimuli presented to dorsal visual field (Young et al., 2013).

While the Ten-ms clearly play an important role in the formation of binocular visual circuits in mice, it should be pointed out that they are also expressed in the visual pathway of zebrafish and chicks (Rubin et al., 1999, 2002; Antinucci et al., 2013; Cheung et al., 2019), species which have little or no ipsilateral retinal projection. In zebrafish, which have entirely crossed retinal projections and lack binocular overlap, Ten-m3 has been shown to contribute to the specification of RGC connectivity and function (Antinucci et al., 2013, 2016), consistent with its role as a homophilic adhesion molecule. The manner in which Ten-ms help to regulate the formation of binocular circuits of mice may be an evolutionary "addon", critical to the alignment and function binocular visual circuits in mammalian species. More information regarding the expression of Ten-ms in mammals with varying degrees of binocularity would be helpful as a first step in assessing this possibility.

The impaired thalamostriatal targeting in Ten-m3 KOs is also associated with functional changes. Notably, while there is no difference in initial and post-acquisition performance levels of a simple motor task, the rate of learning is negatively affected in Ten-m3 KOs (Tran et al., 2015). Although changes in spatial learning might also be expected following the miswiring in hippocampal connectivity (Berns et al., 2018), this has yet to be reported in Ten-m3 KOs.

Ten-m1 is highly expressed in the olfactory bulb and cortex (Allen Brain Atlas). Deletion of Ten-m1 has been shown to affect the KOs ability to detect appetitive and aversive odors (Alkelai et al., 2016). Although less well-characterized in mice compared to the other Ten-ms, this finding correlates with the identification of Ten-m1 in patients with congenital anosmia, as well as an important role for teneurins in the establishment of olfactory circuits in Drosophila (Hong et al., 2012).

While a thorough behavioral characterization has yet to be conducted on Ten-m4 KOs, this teneurin has been linked to bipolar disorder and schizophrenia in humans (Heinrich et al., 2013; Ivorra et al., 2014). The insertion of a transgene which disrupts Ten-m4 expression in mice has been shown to impede oligodendrocyte differentiation. This is associated with reduced myelination and tremors (Suzuki et al., 2012). Mapping studies of humans with essential tremor has revealed a mutation in the Ten-m4 gene which correlates well with what has been observed in this model (Hor et al., 2015). Analysis of zebrafish morpholinos for Ten-m4 also showed changes in myelination as well as defects in motor axon pathfinding (Hor et al., 2015). The role of Ten-m4 in regulating the formation of visual or cortical circuits has yet to be reported.

#### CONCLUDING REMARKS

These studies reviewed above demonstrate multiple, complex and important roles for teneurins in the formation and function of neural circuits. Their ability to mediate homophilic interactions is clearly crucial for the formation of precisely mapped connections between afferent and target fields in these circuits. Each of the teneurins, however, contains multiple domains and cleavage sites that may allow these molecules to also undergo heterophilic interactions with other key signaling molecules, such as the latrophilins (Boucard et al., 2014; Vysokov et al., 2016, 2018; Berns et al., 2018; Sando et al., 2019) as well as transcription factors such as Zic1 and Zic2 (Bagutti et al., 2003; Glendining et al., 2017). Thus, while expression patterns can help formulate hypotheses regarding function, other factors must also be taken into account when considering the roles of these highly complex molecules. Further information regarding the roles for different regions of these glycoproteins, the circumstances under which they are cleaved, and how this relates to their homophilic and heterophilic interactions is critical to a more comprehensive understanding of their function. The presence of multiple splice variants with differing binding properties is likely to add further to this complexity (Berns et al., 2018). The development of more refined tools such as conditional KOs will help to further reveal the manner in which this fascinating family of molecules interacts both at a circuit and cellular level to promote the proper wiring and function of critical sensory and learning networks. The demonstration that two members of the teneurin family play complimentary roles in enabling the generation of functional visual circuits

is particularly intriguing, and together with evidence of conserved patterning across widely separated mammalian species tempts the speculation that they may have played crucial roles in the evolution of binocular vision in mammals.

#### REFERENCES


#### AUTHOR CONTRIBUTIONS

Both authors contributed to the writing and editing of this manuscript.



**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Leamey and Sawatari. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Nomenclature and Comparative Morphology of the Teneurin/TCAP/ ADGRL Protein Families

Luciane V. Sita<sup>1</sup>† , Giovanne B. Diniz<sup>1</sup>† , José A. C. Horta-Junior<sup>2</sup> , Claudio A. Casatti<sup>3</sup> and Jackson C. Bittencourt1,4 \*

<sup>1</sup> Laboratory of Chemical Neuroanatomy, Department of Anatomy, Institute of Biomedical Sciences, University of São Paulo, São Paulo, Brazil, <sup>2</sup> Department of Anatomy, Institute of Biosciences, São Paulo State University, São Paulo, Brazil, <sup>3</sup> Department of Basic Sciences, São Paulo State University, São Paulo, Brazil, <sup>4</sup> Center for Neuroscience and Behavior, Department of Experimental Psychology, Institute of Psychology, University of São Paulo, São Paulo, Brazil

The teneurins are a family of glycosylated type II transmembrane proteins synthesized in several tissue from both vertebrate and invertebrate species. These proteins interact with the latrophilins, a group of adhesion G protein-coupled receptors. Both teneurins and latrophilins may have been acquired by choanoflagellates through horizontal gene transfer from a toxin-target system present in prokaryotes. Teneurins are highly conserved in eukaryotes, with four paralogs (TEN1, TEN2, TEN3, and TEN4) in most vertebrates playing a role in the normal neural development, axonal guiding, synapse formation and synaptic maintenance. In this review, we summarize the main findings concerning the distribution and morphology of the teneurins and latrophilins, both during development and in adult animals. We also briefly discuss the current knowledge in the distribution of the teneurin C-terminal associated protein (TCAP), a peptidergic sequence at the terminal portion of teneurins that may be independently processed and secreted. Through the analysis of anatomical data, we draw parallels to the evolution of those proteins and the increasing complexity of this system, which mirrors the increase in metazoan sensory complexity. This review underscores the need for further studies investigating the distribution of teneurins and latrophilins and the use of different animal models.

Keywords: TEN, Odz, ADGRL, latrophilin, teneurin C-terminal associated peptide

# INTRODUCTION AND NOMENCLATURE

The teneurins are a family of glycosylated type II transmembrane proteins synthesized in several tissue from both vertebrate and invertebrate species (Tucker et al., 2012). The nervous system has been conserved as the main site of teneurins synthesis in a variety of species, where teneurins are prevalent in neuronal projections (Oohashi et al., 1999; Lovejoy et al., 2006). As a type II transmembrane protein, the teneurins have a simple amino terminus located on the cytoplasmic side of the cell, while a carboxy terminus rich in binding motifs is located outside the cell, including an epidermal growth factor (EGF)-like domain, which contains a region of conserved cysteine residues and a stretch of tyrosine-aspartate-repeats (Rubin et al., 1999; Tucker and Chiquet-Ehrismann, 2006; Young and Leamey, 2009). The carboxy-terminal sequence of teneurins, spanning approximately 40 residues and located in the last exon, is flanked by a dibasic cleaving motif and an amidation motif on the carboxy terminal, suggesting this sequence can be cleaved

#### Edited by:

Antony Jr Boucard, Centro de Investigación y de Estudios Avanzados (CINVESTAV), Mexico

#### Reviewed by:

Dalia Barsyte-Lovejoy, University of Toronto, Canada Hervé Tostivint, Muséum National d'Histoire Naturelle, France

> \*Correspondence: Jackson C. Bittencourt jcbitten@icb.usp.br

†These authors have contributed equally to this work

#### Specialty section:

This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Neuroscience

Received: 21 December 2018 Accepted: 15 April 2019 Published: 03 May 2019

#### Citation:

Sita LV, Diniz GB, Horta-Junior JAC, Casatti CA and Bittencourt JC (2019) Nomenclature and Comparative Morphology of the Teneurin/TCAP/ ADGRL Protein Families. Front. Neurosci. 13:425. doi: 10.3389/fnins.2019.00425

**164**

and amidated, generating a small biologically active peptide. This processed peptide has been called teneurin C-terminal associated peptide (TCAP) (Qian et al., 2004; Lovejoy et al., 2006). Mounting evidence suggests that some TCAP paralogs can be processed independently of the main peptide, reinforcing the idea that these small molecules may act as neuronal messengers (Chand et al., 2013). Both teneurins and TCAPs are believed to interact with latrophilins, a family of G-protein coupled receptors (GPCRs) of the adhesion family (Lelianova et al., 1997; Husic et al., 2019 ´ ). This transsynaptic complex appears to participate in axonal guiding, synapse formation and synaptic maintenance. To do that, teneurins, TCAPs and latrophilins also interact with other membrane proteins, such as fibronectin leucine-rich transmembrane (FLRT), neurexins and dystroglycans (as reviewed in Woelfle et al., 2015). In this review, we will evaluate the available morphological information about the teneurins and latrophilins, highlighting the scarcity of available information about these proteins.

# Family History

Recent works suggest teneurin homologs first originated from a prokaryotic transmembrane polymorphic proteinaceous toxin gene, which was incorporated to the genome of the choanoflagellate Monosiga brevicollis (Tucker et al., 2012; Tucker, 2013). This theory is supported by the presence of proteins with a similar structure to the extracellular domain of the Monosiga teneurin in aquatic bacteria while other nonmetazoan opisthokonts lack similar proteins, and by the frequent occurrence of horizontal gene transfer in this species (Tucker, 2013). It is believed that this prokaryotic toxin gene then merged with an EGF-like domain repeats-rich gene, forming a protein similar in structure to the modern teneurin homologs. Among the structural similarities between the Monosiga and the metazoan teneurins are eight EGF repeats, a cysteine rich domain, and a partially similar (YD)-repeat motif (Tucker, 2013). The general structure of teneurins is illustrated in **Figure 1**. Since choanoflagellates are the closest relatives of metazoans (King et al., 2008), it is likely that this event of horizontal gene transfer was then conserved upon the emergence of the metazoan due to the teneurins actions to promote normal development (Tucker et al., 2007; Trzebiatowska et al., 2008).

Complete or partial teneurin sequences have been found in trematodes (Schistosoma mansoni), nematodes (Caenorhabditis elegans), annelids (Capitella teleta), mollusks (Lottia gigantea), arthropods (Apis mellifera, Tribolium castaneum, Aedes aegypti, Culex quinquefasciatus, Drosophila melanogaster, Daphnia pulex, and Ixodes scapularis) and the purple sea urchin (Strongylocentrotus purpuratus). There is no homologous gene found in the sequences from sponges, placozoans, ctenophores, cnidarians, fungi, ichthyospores or nucleariids (Tucker et al., 2012). As a whole, these results suggest the acquisition of the teneurin gene in the choanoflagellate, its loss in some clades (Porifera, Ctenophora, Placozoa, and Cnidaria) and its retention in bilaterians, what is illustrated in **Figure 2**.

Most non-chordate bilaterians have a single teneurin gene (Ten), including the well-studied animal model C. elegans. Two genes are found in D. melanogaster, as well as in several other insect species (honey bee, flour beetle, mosquitoes). The same, however, is not observed for other arthropods, such as the crustacean D. pulex and the arachnid I. scapularis, suggesting that a gene duplication occurred after the divergence of the ectognatha clade (Tucker et al., 2012). These observations, coupled to the fact that the protochordates Ciona intestinalis and Branchiostoma floridae have a single Ten-coding gene in their genome, while the elasmobranch Rhincodon typus (whale shark) has four predicted Ten paralogs (Ten1, Ten2, Ten3, and Ten4), suggests that the two rounds (2R) of whole genome duplication that likely occurred in the early vertebrate lineage (Dehal and Boore, 2005; Ohno, 2013; Sacerdot et al., 2018) account for the diversification of teneurin paralogs.

The teleost fish Danio rerio has five different Ten genes, identified as Ten1, Ten2A, Ten2B, Ten3, and Ten4 (Tucker et al., 2012). Another teleost, the stickleback Gasterosteus aculeatus, also has five Ten paralogs, but with a distinct set: Ten1, Ten2, Ten3A, Ten3B, and Ten4 (Tucker et al., 2012). This observation fits well with the 3R theory (Meyer and Van de Peer, 2005), with another whole genome duplication round taking place in the actinopterygii lineage, resulting in eight Ten genes, followed by the fast loss of some of these genes in individual teleost lineages. Finally, chicken (Gallus gallus), mice (Mus musculus), brown rats (Rattus norvegicus), and humans (Homo sapiens) have each four Ten paralogs, suggesting those early duplications were largely maintained during vertebrate evolution (Tucker et al., 2012).

# The Name of the Game

As it has occurred with most bioactive substances, the discovery and the description of teneurins did not follow any kind of phylogenetic reasoning. This led to a somewhat convoluted nomenclature. The first description of teneurins was made in D. melanogaster by two different groups. Baumgartner and Chiquet-Ehrismann (1993), searching for invertebrate homologs of the vertebrate extracellular matrix glycoprotein tenascin, identified a gene that coded for a protein with a partially similar structure, which they called the tenascin-like molecule accessory, or ten<sup>a</sup> /Ten<sup>a</sup> . Looking for other tenascin-like sequences in the Drosophila genome, Baumgartner et al. (1994) found a second sequence, which they called the tenascin-like molecule major, tenm/Tenm. This second protein was located to odd segments during Drosophila development, with mutants exhibiting a pairruled phenotype. In the same year, Levine et al. (1994) discovered a gene (and its protein) that was also expressed in odd-segments of the Drosophila embryo, was not a transcription factor, and had EGF-like repeats that linked this protein to the vertebrate tenascin. They called this gene odd Oz (odz), and its protein Odz. We now know that Ten<sup>m</sup> and Odz are the same protein, and despite the significant similarity between Ten<sup>m</sup> EGF-like repeats and that of the tenascins, they are structurally and functionally distinct, forming their own family of proteins.

The first vertebrate homolog of the tenascin-like proteins was discovered 4 years later, by Wang et al. (1998). These researchers were searching for genes that were differentially expressed upon stress response in the endoplasmic reticulum of mice. One of the identified genes, called DOC4 (after downstream of CHOP 4), coded for a protein that showed 31% identity and 50%

FIGURE 1 | The overall structure of teneurins and their molecular partners. (A) Teneurins are composed of seven major domains: intracellular domain (teal), transmembrane domain (black), 8 EGF repeats (red), an IgG-like domain (yellow), a β propeller domain (green), a β barrel domain (blue) and a toxin domain (purple). Among the EGF repeats are two modified copies, repeats 2 and 5 (darker red). The EGF stem is important for the dimerization of teneurins in the presynaptic membrane. The IgG-like and β propeller domains help seal the β barrel domain from underneath. The β barrel domain partially surrounds the N-terminal portion of the toxin domain, but a gap in its wall allows some of the toxin domain to extend outwards. At the C-terminal portion of the toxin domain is the processable sequence called the teneurin C-terminal associated protein, or TCAP. Three major cleavage sites are found in the teneurin structure. One cleavage site is located in the intracellular domain and allows this portion to be processed and translocated to the nucleus. Exclusively in teneurin 2, a cleavage site (marked with an asterisk) allows the whole extracellular portion of the protein to be released in the extracellular space. Finally, a cleavage site inside the toxin domain allows the release of TCAP. (B1) Extracellularly, the teneurins are able to interact with ADGRLs, a class of adhesion GPCRs. The binding of teneurin and latrophilins depends mainly on the interaction between the outer portion of the toxin domain and the lectin domain of ADGRLs. (B2) Teneurins are also able to interact with other teneurins. Essential for that interaction is the β propeller domain, who strongly binds similar pairs of teneurins. (C1) Intracellularly, teneurins can bind to actin filaments of the cytoskeleton through linker proteins. (C2) In certain conditions, the intracellular domain of teneurins can be cleaved and translocated to the nucleus to act as a transcription modulator.

similarity to Drosophila Tenm/Odz, and its expression pattern in the developing mouse embryos resembled that described for the invertebrate counterpart. These authors also described the existence of a homologous gene to DOC4 in the human genome, the first human teneurin ortholog described.

The discovery of DOC4 was followed by the description of several vertebrate orthologs. Oohashi et al. (1999) made an important advance in our understanding of teneurins describing four mouse genes that are similar to the Drosophila tenm. These genes and their protein were then termed ten-m1/Ten-m1, ten-m2/Ten-m2, ten-m3/Ten-m3 and ten-m4/Ten-m4. These authors also showed that ten-m4 was identical to the previously described DOC4. Concomitantly, Ben-Zur and Wides (1999) and Ben-Zur et al. (2000) also described four mouse orthologs of tenm, using the Odz1 through Odz4 nomenclature. In that same year, Minet et al. (1999) and Rubin et al. (1999) described the first avian ten<sup>m</sup> homologs, which they called teneurin-1 and teneurin-2, referencing both the phylogenetic history of this protein and the main site of expression of this gene in the chicken. Mieda et al. (1999) identified two ten<sup>m</sup> homologs in zebrafish, which they called ten-m3 and ten-m4 due to their correspondence to the recently described mouse genes.

While investigating the second extracellular loops of odorant receptors, Otaki and Firestein (1999) identified a rat homolog of tenm, which they called neurestin. Meanwhile, Minet and Chiquet-Ehrismann (2000) described the four human teneurins. In this work, we will employ a nomenclature that mostly follows that proposed by Minet and Chiquet-Ehrismann (2000), with a minor change: Drosophila genes will be noted as Ten-a and Ten-m, following the rules for Drosophila genes and the entry of these genes on the available databanks. **Figure 3** summarizes the available knowledge about teneurin orthologs/paralogs in terms of their distribution during development and maturity in different organisms. **Supplementary Tables S1** and **S2** compile the different antibodies and probes used in the literature to investigate the distribution of members of the Teneurin family of proteins. **Table 1** summarizes the nomenclature employed in this work when referring to the teneurin system.

### The Sorcerer's Hat

A striking feature of the teneurins is the presence of hallmarks associated to bioactive substances in the 40 or 41 C-terminal residues. This sequence was discovered when Qian et al. (2004) screened a hypothalamic cDNA library of the rainbow trout (Oncorhynchus mykiss) for paralogous genes of the CRF family. Upon their search, one of the clones identified was the trout ortholog of ten3. The predicted C-terminal region encoded in the last exon of ten3 had a neuropeptide-like structure with low sequence similarity to the CRF family of peptides, hence why this gene was cloned during their search. Furthermore, a predicted cleavage motif in the N-terminal of that putative peptide and an amidated carboxy terminal also contribute to the idea that this peptide could be synthesized, processed and release

occurred after the divergence of the Insecta clade, with several dipterans reported to contain two copies of Ten, identified as Ten-a and Ten-m in the most studied animal model of this clade, D. melanogaster. (5) The final major phylogenetic event concerning the teneurin-latrophilin system occurred at the emergence of the vertebrata clade. While tunicates have a single copy of Ten and Adgrl, vertebrates have four paralogs of Ten (Ten-1 through Ten-4) and three paralogs of Adgrl (Adgrl1 through Adgrl3). The double duplication of teneurins at the root of vertebrates fits well with the 2R hypothesis of a double whole genome duplication occurring at this timepoint. A subsequent loss of one of the newly formed Adgrl paralogs must have occurred to result in the current observable three paralogs of this gene in vertebrates. Exclusively in the Actinopterygii lineage, an additional whole genome duplication occurred, followed by successive gene losses, resulting in 5 paralogs of teneurins in modern day teleosts.

as a bioactive substance. The authors named it the teneurin C-terminal associated peptide (TCAP)-3. Despite the similarity between TCAP and CRF, there is no clear evidence that these two peptides share a common phylogenetic history. It is likely that they were introduced at different times in the metazoan lineage, and the three-dimensional structure of TCAP is highly dissimilar to that of CRF (Chen et al., 2013; Li et al., 2018).

Since a cleavage signal can be found in all paralogs of vertebrate teneurins, the identification of TCAP-3 by Qian et al. (2004) allowed the inference that TCAP-1, TCAP-2, and TCAP-4 could


FIGURE 3 | Visual representation of the available information about the morphology of teneurin system in different species. Empty half-circles represent complete lack of information, gray half-circles represent partial or incomplete information, and full half-circles represent complete and abundant description. The left half-circle represents developmental information, and the right side represents adult information. Several gaps in our knowledge about the morphology of the teneurin system can be seen in the graph. Besides the invertebrate metazoans, only the lab rodent Mus musculus has what can be described as a complete anatomical description. Another popular lab rodent, Rattus norvegicus, has partial information available during its development and next to nothing in the adult. Since mice and rats diverged over 10 million years ago, comparisons between these two species could be informative in terms of plasticity of the teneurin system. Detailed mappings performed in rats are still necessary to allow these comparisons, however. The marsupial Macropus eugenii has been used solely for descriptions of Ten3 in the visual system and could represent an interesting animal model for the study of the teneurins. In a similar vein, the information available for teleost species is scarce and should be obtained with high priority. This information could help us understand the emergent complexity of paralogs originated at the vertebrate lineage and the specific duplications of the Ten gene that occurred in this clade could inform us about the retention of newly generated proteins. The amount of information for primates can also be improved, as only punctual data is available for teeth and ovaries from human origin. This lack of works in humans can be attributed to the difficulties in obtaining and working with these tissues, but such barriers can be overcome through the use of other primate models, such as the Rhesus or the Sapajus monkeys. Finally, more information about TCAP and its distribution must be obtained for virtually every species known.

all be potentially synthesized and released as bioactive substances in vertebrates. TCAP orthologs were soon identified in mice and humans by Wang et al. (2005). Since both Ten-m and Ten-a have related peptide sequences to TCAP in their C-terminal portions (Lovejoy et al., 2006), it is likely that TCAP is a ubiquitous peptide for animals that have teneurin genes and may represent the modern equivalent of the ancient toxin payload of prokaryotes. Notably, all studies so far have used a synthetic formulation of TCAP based on the predicted sequence and structure of mouse TCAP1 (Wang et al., 2005). The isolation and purification of TCAP, therefore, could represent an interesting source of information about TCAP and its mechanisms of action.

A major advance in the understanding of the TCAP system came with the discovery that TCAP-1 may be synthesized independent of Ten1, with several lines of evidence supporting this idea. First, a short transcript of 600 base pairs corresponding to the last exon of Ten1 (exon 31) can be identified in whole mouse brain extracts by northern blot, in addition to the whole Ten1 transcript of approximately 8,000 base pairs. This short transcript could either be the result of independent gene expression or a highly specialized alternative splicing, where only the last exon of the protein is transcribed. Probes directed to other exons result only in the detection of the full-length transcript. Western blot analysis corroborate the presence of short proteins with molecular weights compatible with the independent expression of a segment of exon 31 corresponding to TCAP. In addition, TCAP-1 and Ten1 occupy largely different subcellular compartments, with both TCAP-1 and Ten1 found in the surface of cells in culture, while only TCAP-1 is found in the cytosol. Finally, the distribution of Ten1 mRNA and the mRNA corresponding to the TCAP-1 portion are expressed in a distinct pattern in some regions of the brain, while overlapping in others (Zhou et al., 2003; Wang et al., 2005; Chand et al., 2013). This potentially decouples TCAP-1 and Ten1 in the rodent CNS. A similar mechanism may take place with TCAP-3, but evidence is lacking (Woelfle et al., 2015). On the other hand, TCAP-2 and TCAP-4 are believed to have their synthesis coupled to their respective teneurins (Woelfle et al., 2015). Therefore, the distribution of Ten2, Ten3, and Ten4 will be used as proxy to determine the distribution of TCAP-2, TCAP-3, and TCAP-4, while the distribution of Ten1 and TCAP-1 will be examined separately, when data is available.

#### Toxic Origins

The adhesion G protein-coupled receptor L (ADGRL) family, also known as latrophilins, is a family of G Protein-Coupled Receptors (GPCR) belonging to the adhesion family of GPCRs (Silva et al., 2011). These receptors were initially identified as ligands for the black widow spider toxin component, α-latrotoxin (Lelianova et al., 1997). This is evocative of the teneurin origin, as they are thought to be originated from a prokaryotic toxin gene, making the ancient ADGRL orthologs the possible targets for that toxin. This idea is further strengthened by the indication that ADGRLs may also have evolved by lateral gene transmission from prokaryotes to metazoan ancestors (Zhang et al., 2012).

There are three paralogs of ADGRLs, termed ADGRL1, ADGRL2, and ADGRL3. As is the case with the teneurins, the ADGRLs also received multiple names depending on the group, species and publication. In the original discovery of ADGRLs, by Davletov et al. (1996) and Lelianova et al. (1997), the rat receptor to α-latrotoxin was termed latrophilin 1, and its abbreviation LPH1. In parallel, an independent group also identified α-latrotoxin ligands that were independent of calcium, which they called CIRL, for Calcium-Independent Receptor of

#### TABLE 1 | Nomenclature of the teneurin system.

fnins-13-00425 May 3, 2019 Time: 17:43 # 6


The table presents a collection of synonymous presented in the literature for the teneurin orthologs/paralogs found in different species and the corresponding protein and gene abbreviations.

α-Latrotoxin (Krasnoperov et al., 1996, 1997). Following, the protein was isolated from mouse synaptosomes (Ichtchenko et al., 1998) and called CIRL/latrophilin (CL1). Soon, the same group described the protein paralogs, which received the names CIRL/latrophilin (CL1-3) (Sugita et al., 1998), latrophilin-2 (LPH-2) and latrophilin (LPH3) and CIRL-2 and CIRL-3 (Ichtchenko et al., 1999; Matsushita et al., 1999).

In 2004, orthologs of the ADGRLs were discovered in C. elegans, where they were called lat-1/LAT-1 and lat-2/LAT-2 (Mee et al., 2004; Willson et al., 2004; Langenhan et al., 2009). Some authors employ the abbreviation Lphn1–Lphn3 for the protein, and Adgrl1-Adgrl3 for the corresponding gene (Lu et al., 2015; Anderson et al., 2017). Others also use the abbreviation ADGRL for the proteins (Leon et al., 2017). In this review, we opted to follow the recommended name by Hamann et al. (2015) for adhesion GPCRs and the entry in diverse databanks: Adgrl1/ADGRL1, Adgrl2/ADGRL2 and Adgrl3/ADGRL3. The exception to that rule will be the C. elegans orthologs of ADGRLs, which will be abbreviated as lat-1/LAT-1 and lat-2/LAT-2 as those are the forms still employed in the adequate gene/protein repositories. **Table 2** summarizes the nomenclature employed in this work regarding the latrophilins/ADGRLs.

The ADGRLs are composed of multidomain regions possessing a rhamnose-binding lectin-like domain, an olfactomedin-like domain, and a hormone-binding domain that is similar to those of the CRF family of receptors (Matsushita et al., 1999). TEN2 binds and activates ADGRL1 with high affinity, while TCAP-2 binds to ADGRL2 inducing calcium release from intracellular stores (Silva et al., 2011). It was later shown that TEN4 also binds to ADGRL1, with a slightly lower affinity than TEN2, suggesting ADGRL1 is the cognate receptor of both TEN2 and TEN4 (Boucard et al., 2014). While the lectin-like domain of ADGRLs appear to be the most significant domain for teneurin binding and activation, it has been


The table presents a collection of synonymous presented in the literature for the latrophilins/ADGRLs found in different species and the corresponding protein and gene abbreviations.

suggested that TCAP may interact with the hormone-binding domain (Woelfle et al., 2015).

# THE DISTRIBUTION OF TENEURINS/ TCAP IN INVERTEBRATE SPECIES

### Caenorhabditis elegans

fnins-13-00425 May 3, 2019 Time: 17:43 # 8

Although there is a single teneurin gene in C. elegans homologous to the two D. melanogaster Ten genes, this gene is controlled by two different promoters, what results in two different isoforms of TEN: a short version, TEN-S, and a long version, TEN-L. Promoter activity related to the TEN-L isoform was detected, during development, primarily on the derived cells of the EMS and C lineages, which give rise, respectively, to somatic tissues (endoderm, mesoderm, and stomodeum) and hypodermic and neural tissues (Drabikowski et al., 2005). Mörck et al. (2010) described a similar pattern at 150 min after fertilization, which will give rise to hypodermal cells by the end of gastrulation. In the somatic gonad, TEN-L is also active during gonadogenesis, what is reflected by its expression in z1 and z4 cells in the embryo. As early as 350 min, and consolidating at the 1.5-fold stage, expression of the long promoter can be detected in the pharynx, gut, and somatic gonad precursor cells, as well as in the aforementioned z1 and z4 cells (Drabikowski et al., 2005; Mörck et al., 2010).

During postembryonic development, a similar pattern was preserved. In the L1 larval stage, the pharynx, gut cells, hyp11 and somatic gonad precursors show TEN-L promoter activity, as well as z1 and z4 cells. At stage L4, fluorescence can be detected in some neurons of the nerve ring, as well as the anchor cell, vulva muscles and distal tip cells. In the adult hermaphrodite, staining is seen in the pharynx, selective nerve ring neurons, the vulva muscles, the distal tip cells and in coelomocytes. In the adult male, the same head structures are labeled in addition to the vas deferens, diagonal muscles and spicule sheath cells (Drabikowski et al., 2005). Mörck et al. (2010) also report strong expression of the long form of the transcript that persists in pharyngeal and intestinal cells and in several head neurons, including eight pharyngeal cells: the three marginal cells mc1, the three marginal cells mc3, and the neurons M2L and M2R.

The promoter associated to the TEN-S isoform, on the other hand, follows a distinct pattern of expression. During embryonic development, it is initially detected at 150 min after fertilization in anterior cells, although less abundantly than TEN-L. This signal can be traced to the descendants of lineage ABp. By 300 min after fertilization, the presence of TEN-S in posterior hypodermal cells becomes clear, in addition to ventral leading cells. Mörck et al. (2010) reports that the hypodermal expression then gradually fades away, with only head neurons being labeled by the end of embryogenesis. TEN-S post-embryonic activity was detected in specialized epithelial cells, such as the arcade cells of the anterior end and the excretory duct. The short promoter activity was also observed in a subset of neurons, consisting of CAN and HSN neurons, as well as lumbar and retrovesicular ganglion motorneurons and some nerve ring interneurons. In the adult, staining of TEN-S is limited to nerve ring, the ventral cord and a few cells in the tail, in addition to R8 and R9 ray neurons in the adult male (Drabikowski et al., 2005; Trzebiatowska et al., 2008).

To investigate the subcellular localization of different TEN isoforms in the C. elegans, Drabikowski et al. (2005) raised antibodies directed to the N- and C-terminus of TEN, with the C-terminal-raised antibody labeling only TEN-L, while the N-terminal antibody labels both isoforms in embryos. The C-terminal antibody labeled exclusively the plasma membrane, while the N-terminal antibody resulted in labeling of both the plasma membrane and a punctuate pattern of staining inside the nucleus. The authors suggested such staining pattern indicates the translocation of the N-terminal sequence of TEN as part of its signaling process in C. elegans.

# Drosophila melanogaster

As stated before, two genes are found in Drosophila (Ten-a and Ten-m), encoding the transcripts TEN-a and TEN-m.

#### Ten-m

The expression of Ten-m starts around the blastoderm stage, 2 h after fertilization. Transcripts are found in the central area, outside the anterior and posterior poles, ubiquitously expressed on most of the embryo with exception of the dorsal side, where fewer transcripts are found. At the second half of germband elongation (5.5–7.5 h-old embryos), the diffuse expression of Ten-m is replaced by a quickly emerging pattern of stripes in mesodermal and ectodermal cells, that ends in a sharply defined pattern of 14 bands as germband elongation reaches its conclusion. During germ band retraction, the expression of Ten-m becomes more prominent at the dorsal margin of the germ band, and by the end of this stage transcripts are found in cardiac cells, lymph gland, posterior spiracles and in the tracheae. At this stage, expression also becomes clear in the nervous system, including the ventral cord and the supra-esophageal ganglia, with reminiscent patterned expression near the segmental furrows. At the time of hatching, transcripts are mainly confined to the ventral cord and to the brain (Baumgartner et al., 1994; Levine et al., 1994).

The synthesis of TEN-m follows a partially similar progression to its coding mRNA. Staining is located to the periphery of the cell, suggesting a plasma membrane localization that is common for several species. Immunoreactivity is detectable around the same time the transcripts are detected, encompassing the blastoderm stage. Its distribution overlaps with the mRNA, covering most of the central part of the embryo, but it's absent of the anterior and posterior poles. A distinct small group of immunoreactive cells is found close to the anterior pole, which was identified as the anterior domain by Baumgartner et al. (1994). During gastrulation, a pattern of seven immunoreactive bands can be readily identified. This patterning occurs considerably sooner than that of the mRNA, suggesting TEN-m may be released and the bind to receptors that are already patterned in preparation of germ band elongation. As the stage progresses, the seven-band pattern becomes a fourteen-band pattern, and then immunoreactivity becomes almost exclusive to the mesoderm after a steep decline in ectodermal staining. As elongation advances, the banded pattern is reduced as the

bands appear to fuse, and neuroblasts become stained at the anterior domain. At the time of germ band retraction, staining is visible on the tracheal system and in neurons along the midline. By the end of germ retraction, the pattern of immunoreactivity resembles once more that of the RNA, with staining on the ventral cord, cardiac cells and the lymph gland. During head involution, staining on axons of the ventral cord become prevalent, with a strict temporal pattern of synthesis (Baumgartner et al., 1994; Levine et al., 1994; Zheng et al., 2011).

During the larval stage, TEN-m immunoreactivity is found on axons and imaginal discs (Baumgartner et al., 1994). Novel synthesis is particularly evident on the Drosophila eye disc. TEN-m immunoreactivity in third instar larvae is evident at undifferentiated imaginal disc cells of the morphogenic furrow, a single cell in each maturing ommatidium, and a cluster of non-epithelial cells deep at the center of the eye disc. The appearance of one sharp staining centered point per ommatidia suggests its relation to the development of a photoreceptor cell with R7 identity. The third imaginal site of TEN-m synthesis can be described as a cluster of round adepithelial cells under the epithelial monolayer of the eye disc, which migrated from mesoderm germ into eye disc (Levine et al., 1997). Ten-m gene activity can also be detected on the morphogenetic furrow of the eye disc and in the brain optic lobes (Minet et al., 1999). It is likely, therefore, that the expression of Ten-m in both the developing eye and the optic lobe of the brain, and the presence of TEN-m immunoreactivity in the optic stalk, are indicative of a TEN-m function on the correct mapping between ommatidia and the visual lobe. As we will see, this function likely remained and was expanded in vertebrates. A distinct cluster of strong TEN-m staining cells are seen in the antennal, wing and leg discs, representing progenitors of a column of glial cells in which Ten-m is expressed. TEN-m immunoreactivity is very strong in the maxillary palps and rostral membrane that will give rise to the dorsal head capsule. Wing-disc staining of TEN-m is predominant in the wing blade, wing hinge and thoracic epidermis. Leg discs present rings of immunoreactivity, also characterized by a cluster of cells at the central core of the disc (Levine et al., 1997).

Information about the adult expression of Ten-m is limited to subjects immediately after eclosion. Staining is coherent with the disc staining, including strong staining in the three antennal segments, the maxillary palpus, the rostral membrane of the head capsule and derivatives of the clypeolabral and labial discs (Levine et al., 1997). Signals were detected in the brain and eyes but could not be explored in detail by the authors. It is noteworthy that the staining reported may be from residual beta-galactosidase expressed during pupation (Levine et al., 1997), so more studies in adult flies are necessary to confirm the expression of Ten-m in the adult animal.

#### Ten-a

The expression of Ten-a transcripts in the developing D. melanogaster is partially similar to that of Ten-m. Expression of the two start around the blastoderm stage, with widespread distribution. While Ten-m is not present on the anterior and posterior poles, Ten-a is uniformly distributed over the entire embryo. This uniform distribution persists through gastrulation, with a slightly more pronounced signal on the furrows. At the beginning of germ band elongation, signal becomes restricted to the ectoderm and mesoderm, when a pattern starts to appear at around 5 h of development. Although largely similar in timing, Ten-a transcription is better localized to the ectoderm, with only faint staining on the mesoderm, the opposite pattern of Ten-m. Clear signals can also be detected at the procephalic neuroblast region. As the germ band elongation progresses, the band pattern of Ten-a expression becomes clearer. By the time germ band retraction starts, it is possible to localize Ten-a to the ventral cord and the supraesophageal ganglion. At this stage, small labeled cells are seen near the segmental furrows, possibly representing sites of muscle attachment. During head involution, both the brain and the ventral cord show strong labeling, that will remain at the end of embryonic development and the three larval stages. Ten-a mRNA cannot be detected in the adult fly (Baumgartner and Chiquet-Ehrismann, 1993).

The described pattern of TEN-a immunoreactivity, however, is drastically different to that of TEN-m for the early period of embryogenesis. Fascetti and Baumgartner (2002) first detected TEN-a during germ band retraction, at stage 12, in neurons. Protein could be located to some cell bodies and on pioneering axons. By the end of germ band retraction, clear staining can be seen on the commissures of the ventral cord, especially on the posterior commissures. The hindgut is also labeled at this stage. During head involution, the pattern of TEN-a immunoreactivity becomes quite similar to TEN-m, including the brain, the ventral cord and structures of the antennomaxillary complex as the main sites of labeling. Low immunoreactivity is found in the CNS after differentiation (Fascetti and Baumgartner, 2002). Fortyeight hours after puparium formation, TEN-a can be detected in specific glomeruli. The subset of glomeruli synthesizing elevated TEN-a was distinct but partially overlapping with that synthesizing elevated TEN-m (Hong et al., 2012).

It is remarkable that Ten-a mRNA expression starts hours before the protein can be detected. A possible explanation for this observation is that, in early stages, Ten-a mRNA translation is silenced, or TEN-a is readily degraded after synthesis. This silencing is then selectively shut off in the neuronal lineage, allowing the protein to acquire a predominantly neuronal phenotype. Comparing the patterns of Ten-a and Ten-m expression and protein synthesis in D. melanogaster to that of ten-1 in C. elegans, it is apparent that D. melanogaster Ten-m echoes the ten-1 expression pattern of the common ancestor of euarthropoda and nematoda, while Ten-a differentiated its pattern of expression, becoming a late, predominantly neuronal-driven molecule in the insect lineage.

#### Ciona intestinalis

No information regarding the teneurin distribution is available for the developing or adult C. intestinalis. The distribution of TCAP-1 has been investigated in the adult animal. Such scarcity of information certainly derives from the almost nonexistent anatomical information about C. intestinalis. Since this information has been provided by Colacci et al. (2015), works

describing the distribution of teneurin in these animals will be extremely informative.

#### Teneurin C-Terminal Associated Peptide (TCAP)

Immunoreactivity to TCAP in the adult C. intestinalis is found in the intestinal and sexual areas. In the testis, staining is found in putative Sertoli cells, outside of the tubules and in the epithelium of the sperm duct cells. In the ovary, labeling occurred in granulosa cell homologs and periovulatory cells, but not in the ovum. The subcellular localization of TCAP showed complex patterns depending on the region. In the intestine, labeling was detected both in the cytosol as in the periphery and along the plasma membrane. In the testis, most labeling in Sertoli cells was pericellular, with less staining inside those cells. In the ovary, on the other hand, labeling occurred primarily within the cytosol (Colacci et al., 2015). The distribution of TCAP mRNA was coherent, if broader, with that of the protein. RT-PCR results showed transcripts in the buccal siphon, central ganglion, branchial basket, testes, ovary, and stomach (D'Aquila et al., 2017).

### THE DISTRIBUTION OF TENEURINS/ TCAP IN VERTEBRATE SPECIES

#### Danio rerio

No morphological information is available for D. rerio Ten1, Ten2A, or Ten2B (Tucker et al., 2012).

#### ten3

Detection of ten3 mRNA starts on the notochord and somites at tailbud stage, approximately 10 h after fertilization, what is coincident with the completion of the neural tube's basic plan dorsal to the notochord. During segmentation, expression of ten3 mRNA is not exclusively mesodermal anymore, as transcripts can be found in the developing nervous system. We will describe the mesodermal and ectodermal expression of ten3 separately in this session (Mieda et al., 1999).

At 14 h after fertilization, ten3 mRNA is strongly detected in the caudal forebrain, corresponding to the diencephalic area, while medium expression is found in the optic vesicles and midbrain. At this stage, a segmental expression of ten3 mRNA is found on the rhombencephalon, with transcripts expressed in low levels in rhombomeres 3 and 5. By the end of the segmentation stage, additional expression of ten3 mRNA is seen in the midbrain–hindbrain boundary. At 23 h post fertilization, ten3 mRNA ceases to be detected in the hindbrain, it's weakly detected in the anterior part of the midbrain–hindbrain boundary and becomes strongly expressed in the dorsal part of the tectal primordium, ventral part of the mesencephalon and caudal part of the diencephalon. At this stage, weak expression is found in the optic vesicles. By 36 h postfertilization, ten3 mRNA expression is concentrated on the forebrain (Mieda et al., 1999). Further ten3 mRNA expression has been described by Antinucci et al. (2013) in retinal cells. At 48 h postfertilization, expression is predominantly found in the ventral retina and medial portion of the stratum periventricular (tectal cells), while at 3 and 5 days postfertilization, ten3 mRNA is diffusely expressed.

Regarding ten3 mRNA mesodermic expression, at 14 hpf strong expression is found on the developing somites, while only low expression is seen on the notochord. At 17 hpf, a mediolateral gradient of expression develops, with the medial parts presenting weaker expression, and by 20 h ten3 mRNA is detectable in the pharyngeal arches. By the end of segmentation, ten3 mRNA is not detectable in the medial somites, and expression fully vanishes on somites by 36 hpf. At this stage, expression of ten3 mRNA is observed in the pectoral fin buds and on pharyngeal arches (Mieda et al., 1999). No information is available about the distribution of ten3 in the adult D. rerio.

#### ten4

In contrast to ten3 mRNA, the expression of ten4 mRNA is exclusively ectodermic in nature. By 10 hpf, ten4 mRNA is expressed along the anterior margin of the neural plate, and by 14 h postfertilization the brain is the main site of expression. At this stage, the pattern of ten4 mRNA is partially complimentary to ten3 mRNA, with transcripts found on the forebrain, including the optic vesicles (albeit in lower levels than ten3 mRNA) and the rostral diencephalon, the mesencephalon, and in rhombomere 5 and 6 of the rhombencephalon. On 20 hpf, the segmental pattern of ten4 mRNA becomes stronger, with expression found in the rostral diencephalon, the mesencephalon, the mesencephalonhindbrain boundary and rhombomere 2 (previously clear of ten4 mRNA expression). Expression in rhombomeres 5 and 6 persist and increase in intensity. Finally, transcripts are also found in the anterior spinal cord, with additional weak expression of ten4 mRNA is found in individual neurons of the caudal spinal cord. At 23 hpf the distribution of ten4 mRNA strongly diverges of that of ten3 mRNA, with dorsal and ventral bands of expression in the rostral diencephalon, weak expression on the caudal diencephalon, mesencephalic expression on the ventral part of the tectal primordium, strong expression in the caudal mesencephalic-hindbrain boundary, very strong expression on rhombomeres 2, 5, and 6 and in the anterior spinal cord. Widespread expression is found in the brain at 36 hpf (Mieda et al., 1999). No information is available about the distribution of ten-4 in the adult D. rerio.

Comparing the expression of teneurins in D. rerio to the distribution of C. elegans ten-1 and D. melanogaster Ten-a and Ten-m, a pattern can be distinguished. D. rerio ten3 has a similar pattern of expression to the C. elegans teneurin and the D. melanogaster Ten-m, as it is found in both mesodermal and ectodermal tissue early during embryogenesis and then becomes prevalent in neural cells during axiogenesis and connection formation. On the other hand, D. rerio ten4 mRNA carries semblance to D. melanogaster Ten-a, being utilized later during embryogenesis by the nervous system in a pattern partially complementary to that of ten-3 mRNA. As commented before, however, it is likely that a duplication event happened specifically on the insect lineage to generate Ten-a and Ten-m, while the four vertebrate ten genes result from two other events of duplication specific to the early vertebrate lineage. The similarities between ten3 and Ten-m and ten4 and Ten-a, therefore, must result

from evolutionary convergence, rather than homology itself. This is underscored by the fact that the mechanism that regulates the nervous system-specific paralog in each species is different: while in D. melanogaster Ten-a is expressed early but the protein only appears late, be it by RNA silencing or protein degradation, in D. rerio ten4 mRNA will only be detectable once the nervous system begins to differentiate. We cannot exclude, however, the possibility that the presence of two different promoters for the teneurin gene early in evolution may have facilitated the differentiation of functions once the gene was duplicated, what may have contributed to the retention of multiple paralogs and may have guided how the new patterns of expression/synthesis emerged.

Although ten3 and ten4 are equidistant to the C. intestinalis teneurin (Tucker et al., 2012), it is likely that, from a morphofunctional perspective, ten3 represents a more conserved expression pattern when compared to early forms of teneurin. More studies in D. rerio are necessary to establish if the complementary pattern of expression between ten3 and ten4 developed before or after the genic events that resulted in the creation of ten1, ten2A, and ten2B.

#### Gallus gallus TEN1

Unfortunately, expression analyses of TEN1 mRNA are not available for the early embryogenesis of chicken. By day 5, transcripts are detected on the embryo head, but not the trunk. On day 7, a hybridization signal can be found only in the developing nervous system. By days 14 and 17, strong signals are found in the tectofugal elements of the visual system, such as retinal ganglion cells, the stratum griseum centrale of the optic tectum, and the rotund nucleus of diencephalon. Further signals are also detected in the inner nuclear layer of the retina and in other layers of the optic tectum. Areas linked to olfactory sensing and processing are also stained, such as the mitral cells of the olfactory bulb and neurons from the hippocampus and piriform cortex. In the hindbrain, transcripts were found in the nucleus laminaris, nucleus magnocellularis and throughout the cerebellum. TEN1 immunoreactivity is largely compatible with the aforementioned distribution, but extends to some regions connected with TEN1 mRNA synthesizing areas, such as the glomerular layer of the olfactory bulb (which accommodates the dendrites of mitral cells) and the outer portion of the inner nuclear layer of the retina (Minet et al., 1999; Rubin et al., 1999; Kenzelmann et al., 2008). Information about TEN1 mRNA in the adult chicken is limited. Minet et al. (1999) detected signals for TEN1 mRNA in northern blot experiments of adult chicken brain extracts. No signal was detected in adult kidney, heart or liver.

#### TEN2

The expression of TEN2 and TEN2 immunoreactivity are both found in neuroectodermal and mesodermal tissues during development, and that is why these two sites of expression/synthesis will be addressed separately.

Trunk expression of TEN2 mRNA can be detected as early as 3 days after incubation, during late somitogenesis stage. At this stage, transcripts can be found in branchial arches, heart, somites, craniofacial mesenchyme and the apical ectodermal ridge of developing buds. After 5 days of incubation, transcripts are seen in the wing and hindlimb, the trunk, maxillary and mandibular processes and the head mesenchyme. At 7 days of development, the signals disappear (Tucker et al., 2001). Immunohistochemistry for TEN2 resulted in agreeable results to those of in situ hybridization, including the cranial mesenchyme, branchial arches, the developing somites, and the apical ectodermal ridge. The only exception was the notochord, which was stained with the antibody but did not show signs of TEN2 mRNA expression. The investigation of TEN2 immunoreactivity outside the nervous system also gave clues about the subcellular localization of TEN2 in the avian model, as the punctuate outlining-pattern of staining resulting from immunohistochemistry is expected from a membrane-anchored protein (Tucker et al., 2001).

In the nervous system, TEN2 mRNA is first detected at 4 days after incubation. In 7-days old embryos, transcripts are found in the retina, telencephalon, diencephalon and the optic tectum, in a pattern that mostly overlaps with that of TEN1 mRNA, except in the diencephalon, where TEN2 mRNA is found on the anterior thalamus, while TEN1 mRNA is prevalent on the dorsal thalamus. As was the case of TEN1, most of TEN2 expression was concentrated on members of the tectofugal system. In day 10 after incubation, strong hybridization signals are found in the forebrain, particularly on the hippocampus and in the visual Wulst, and by day 12 the retinal cell ganglion is clearly labeled. By 14 days after incubation, TEN2 mRNA expression concentrates on the stratum griseum periventriculare of the optic tectum, mostly separated from TEN1 mRNA in the stratum griseum centrale. Additional signals are found on the lateral geniculate nucleus (Rubin et al., 1999; Rubin et al., 2002).

Immunoreactivity to TEN2 was similar to gene expression, with some additional areas of staining that were not previously found. On day 7 post incubation, labeling can be seen in the retinal nerve fiber layer. Labeling has expanded to other elements of the visual system by day 11, including the inner plexiform layer, the optic nerve and the optic tectum. As was the case with the messenger RNA, by day 17/18 post incubation the synthesis of TEN2 becomes more widespread. On the visual system, immunoreactivity is found in the inner plexiform layer of the retina, visual Wulst, the ventral geniculate nucleus, pretectal nuclei, stratum griseum periventriculare and centrale. Some of these areas show remarkable separation between TEN1 and TEN2, such as the retina (TEN2 is found in laminae 1 and 3, while TEN1 is found in laminae 2 and 5) and the optic tectum (while TEN1 appears to be actively synthesized by stratum griseum centrale cells, TEN2 is found in a punctiform manner that suggests its synthesis by cells in other areas that are synapsing at the stratum griseum centrale). Other areas that include TEN2 immunoreactivity are the olfactory bulb, piriform cortex, hippocampus, septal nuclei, and cerebellum. In several cases, TEN2 immunoreactivity was found associated to the presence of a basement membrane (Tucker et al., 2001; Kenzelmann et al., 2008). In the adult, TEN2 mRNA transcripts

are detected by northern blot and RT-PCR in the adult brain, but no signals are found in the heart or liver (Rubin et al., 1999).

Looking at the distribution of TEN1 and TEN2 in the Gallus brain, we see several of the morphofunctional aspects exhibited as early as in C. elegans, as well as the more complex patterned expression between TEN1 and TEN2 that will be characteristic of integrative areas of the amniote brain. In several ways, the expression and synthesis of TEN2 is evocative of the C. elegans teneurin, the D. melanogaster Ten-m, and the D. rerio ten3, including its expression in both non-neural and neural tissues, the timing of expression, and its association to the basement membrane. On the other hand, TEN1 is evocative of the Drosophila Ten-a and the D. rerio ten4, with a predominantly neural expression that occurs in tandem with that of the other paralog in regions of great connectivity, such as the visual pathway and olfactory areas.

#### TEN3

No information is available for the G. gallus TEN3 expression or TEN3 synthesis.

#### TEN4

The largest body of evidence available points to a peripheral expression of TEN4 mRNA in non-neuronal tissue of the developing chicken. Transcripts are first detected 3.5 days after incubation in zones of polarizing activity (ZPA), branchial arches, cells lining the intersomitic clefts and in the cranial mesenchyme. As development progresses, increased signals are detected in the mesenchyme, including the mesenchyme dorsal to the dorsomedial lip of the somites, the cranial mesenchyme and the mesenchyme of the first, second and third brachial arches. By 5 days of development, however, most of the signal is found in the developing limbs, including additional patches of expression (outside the zones of polarizing activity) in the leg and wing buds (Tucker et al., 2000). Unfortunately, little has been described about the distribution of TEN4 mRNA in the developing nervous system of the chicken, except for the observations of Tucker et al. (2000) that transcripts can be seen in the midbrain–hindbrain junction and in the diencephalon.

#### Macropus eugenii TEN3

Carr et al. (2013, 2014) investigated the post-natal development of ipsilateral retinogeniculate projections in the wallaby marsupial (Macropus eugenii). In these animals, TEN3 mRNA can be found in the retina, with labeling concentrated on the retinal ganglion cell layer and in the superior colliculus, where labeling is found in the superficial retinorecipient layers at all ages examined, from P12 to P99, and in the adult. Additional labeling was found in the dorsal part of the lateral geniculate nucleus of P12–P71 animals, but older animals were not examined. The main difference between developing and adult animals was the existence of dorsoventral (retina) and mediolateral (superior colliculus) gradients in the young subjects, which disappeared in the adult.

# Mus musculus

#### Ten1

The earliest reported expression of Ten1 mRNA in the developing mouse is at E13.5. By E15.5, transcripts can be detected in subplate and cortical plate in a rostral-low/caudal-high, dorsomedial-low/ventrolateral-high gradient. Ten1 mRNA is also detected in the dorsal thalamus at this stage, including the ventroposterior nucleus, posterior complex and lateral geniculate nucleus, dorsal part. By E18.5 the gradient is reversed, with a rostral-high/caudal-low pattern. At P2, the overall pattern of expression stays the same, with Ten1 mRNA found in layer 4 and subplate in the cortex, while low signals are found in layers 5 and 6. Low expression also remains in the thalamic ventroposterior nucleus, and strong expression is found in the thalamic reticular nucleus. By P7, additional expression is seen on CA1, CA3 and dentate gyrus of the hippocampal formation (Li et al., 2006). The distribution of Ten1 mRNA is likely to be more widespread than that, as Zhou et al. (2003) describe hybridization signals for Ten1 mRNA in the midbrain, hindbrain, spinal cord, and trigeminal ganglion. Furthermore, despite not described by the authors, it is clear in the work of Li et al. (2006) that other areas of the brain are stained in the developing brain, including midline nuclei of the thalamus, the amygdaloid complex and discreet nuclei of the hypothalamus. No data is available on non-neuronal expression of Ten1 mRNA in the mouse.

In the adult mouse, Oohashi et al. (1999) report the presence of Ten1 mRNA transcripts in the adult brain, in addition to faint signals in the kidney, testis, and thymus. In 6-week old animals, prominent Ten1 mRNA hybridization signals are found in the stratum pyramidale of the CA2 subfield of the hippocampus proper and in the granular layer of the dentate gyrus, while weak expression is found in the CA1 and CA3 subfields. TEN1 immunoreactivity in the hippocampus at this stage was only partially similar, with stronger staining in the stratum lucidum of the CA3 region and weaker staining in CA1 and dentate gyrus. This mismatch between expression and immunoreactivity likely reflects the intra-hippocampal connectivity, with immunoreactivity found in the axons of neurons that express Ten1 mRNA (Oohashi et al., 1999; Zhou et al., 2003). The cerebellum also contained Ten1 mRNA transcripts, which were found in the granular layer. The protein was found in the molecular and granular layers, in addition to cerebellar nuclei (Oohashi et al., 1999; Zhou et al., 2003). Additional hybridization/protein is found in the cerebral cortex (layers 2–6) and thalamus, and protein has been found in the brainstem and retina (Oohashi et al., 1999; Zhou et al., 2003). Outside the brain, TEN1 is found in the lung, kidney and in the testes. In the latter, immunofluorescence signals are detected in the tunica propria of the seminiferous tubule and the surrounding interstitial cells (Oohashi et al., 1999; Chand et al., 2014).

#### TCAP-1

The presence of TCAP-1 immunoreactivity was examined in adult BALB/c mice using an antibody directed to its sequence. TCAP-1 immunoreactivity is observed in the pyramidal layer and stratum oriens of the CA1 subfield; pyramidal layer, stratum lucidum and stratum radiatum of the CA2/CA3 subfields, and

additional weaker staining in the granular layer of the dentate gyrus (Chand et al., 2013). Outside the nervous system, TCAP-1 immunoreactivity is found in germ cells and spermatocytes adjacent to the basement membrane (Chand et al., 2014).

#### Ten2

Expression of Ten2 mRNA in the developing mouse starts around E10.5, when hybridization signals are positive in the forebrain, rostral and central midbrain, the outer linings of the optic cup and in the auditory vesicle. At E12.5, expression becomes more widespread and can be detected in the caudal forebrain, very strongly on the midbrain, strongly in the hindbrain and weakly on the spinal cord. By E15.5, transcripts are more clearly located to the roof of the midbrain, in addition to the hindbrain and the nasal cavity. In the telencephalon, Ten2 mRNA transcripts are distributed in a rostral-low/caudal-high within the cortical plate. Expression is also found in the diencephalon, including the centrolateral nucleus, dorsal part of the laternal geniculate nucleus posterior complex and ventral part of the medial geniculate nucleus, ventral thalamus, and discreet nuclei of the hypothalamus. During development, the cortical gradient of Ten2 mRNA remains the same. By P2, Ten2 is widely expressed in the cortex but the strongest signal is found in layer 5. A new site of expression after birth is the hippocampal formation, with high expression in CA1 when compared to the weaker signals of CA3 and DG. By P7, the different levels of expression in the hippocampus equalize (Zhou et al., 2003; Li et al., 2006).

Immunoreactivity data for TEN2 during mouse development is available in the investigation of Young et al. (2013) of the developing visual pathway. At E14, immunoreactivity is found in the retinal ganglion cell layer within the central retina, while by E16 the dorsoventral axis has become uniform. Synthesis of TEN2 is ample on axons of the retinofugal pathway, including the optic nerve, optic tract, and optic chiasm. Targets of the retina are also immunoreactive to TEN2, including the superior colliculus and the dorsal part of the lateral geniculate nucleus, and such synthesis was observed from E16 until P7. By P7, immunostaining can be identified on layers 4 and 5 of the primary visual cortex (Young et al., 2013).

In 6-week old adults, Ten2 mRNA transcripts are found in the brain, kidney, and testis (Oohashi et al., 1999; Ben-Zur et al., 2000). In the brain, Ten2 mRNA was thoroughly expressed in the pyramidal and granular layers, with immunoreactivity found in the stratum oriens, stratum radiatum, and stratum lacunosum moleculare of the CA1/CA2 subfields (Zhou et al., 2003). In the cerebellum, Ten2 mRNA is found in the molecular layer and Purkinje's cell layer. Immunoreactivity, on the other hand, is found in the molecular and granular layers. Finally, expression and synthesis of Ten2/TEN2 are found on layers 2–6 in the cerebral cortex (Zhou et al., 2003).

#### Ten3

As more information is available about Ten3 mRNA expression in non-neuronal tissues, this distribution will be discussed separately from that in neuronal tissue. At E7.5, high levels of Ten3 mRNA are found in the notochord. Two days later, Ten3 mRNA is detected on anterior somites and limb buds. By E10.5, expression starts to be detected in the first, second, and third branchial arches. At E12.5, expression of Ten3 mRNA is found in the facial mesenchyme and in the head meninges, and at E16.5 transcripts are seen in the mesentery of the gut and the urogenital system. Finally, at E18.5, expression is seen in the dermis, developing limbs and in the outer layers of the periosteum and muscle epimysium (Ben-Zur et al., 2000; Zhou et al., 2003).

In neural tissue, Ten3 mRNA is detected as early as E7.5 in the neural plate, with higher expression levels in the neural folds. At E8.5, transcripts are found in the caudal forebrain, dorsal midbrain region and in the otic vesicle. A day later, Ten3 mRNA expression can be seen in the telencephalon, diencephalon, dorsal midbrain and in the otic vesicle (Zhou et al., 2003). At E10, expression has also been reported in the Rathke's pouch (Ben-Zur et al., 2000). By E12.5, the optic tectum becomes the most preeminent site of Ten3 mRNA expression. Hybridization signals are also seen in all layers of the neocortex, in the hippocampus, and in the thalamus. At this developmental stage, high levels of transcription are seen in the optic recess of the diencephalon, in the optic stalk, in lens cells and in the corneal ectoderm. Restricted signals are detectable in the pons and the rostral medulla at this stage, as well as in the dorsal horn of the spinal cord. At E15.5, cortical plate expression of Ten3 mRNA follows a low-rostral/high-caudal gradient. In the dorsal thalamus, transcripts are detected in the centrolateral nucleus, dorsal part of the lateral geniculate nucleus, ventral part of the medial geniculate nucleus and posterior complex. At E16.5, the retinal expression of Ten3 mRNA shifts to the inner neuroblastic layer, with a high-ventral/low-dorsal gradient that will remain into the first postnatal week. In the superior colliculus, a highmedial/low-lateral gradient of Ten3 mRNA is observed in the superficial retinorecipient layers. Outside the central nervous system, low levels of expression were found in dorsal root ganglia while higher levels were found in trigeminal ganglia (Ben-Zur et al., 2000; Zhou et al., 2003; Li et al., 2006; Leamey et al., 2007a; Dharmaratne et al., 2012).

Shortly after birth, on P0, Ten3 mRNA transcripts are found on layer 5 of the developing virtual cortex, while the protein is found in the same region as well as in axons projecting from this area. P2 expression of Ten3 mRNA in the cortex is detectable in layers 5 and 6, with the somatosensory cortex also displaying labeling in layer 4. In the hippocampus, staining is seen in CA1 exclusively. Thalamic expression has increased and includes the centrolateral nucleus, ventral posterior nucleus, lateral posterior nucleus, posterior complex, lateral geniculate nucleus and reticular thalamic nucleus. At P3, Ten3 expression can also be detected in the striatum, with a stronger labeling in the dorsal striatum following the characteristic striatal organization in patches. This pattern is conserved by P7 (Li et al., 2006; Leamey et al., 2007b; Tran et al., 2015). At P10, TEN3 immunolabeling is seen in the proximal CA1, distal subiculum and medial entorhinal cortex. TEN3 was most prominent in synaptic layers, including the stratum lacunosummoleculare of CA1 and the molecular layer of the subiculum, what is consistent with TEN3 being present in the synaptic cleft. TEN3 was also present in axons, dendrites and cell bodies (Berns et al., 2018).

In 6-week old adults, Ten3 mRNA transcripts are found in the brain, liver and testes (Oohashi et al., 1999; Ben-Zur et al., 2000). In the hippocampus, the expression of Ten3 was relatively lower when compared to other teneurin paralogs, with transcripts found in the CA2 subfield and weakly in CA1. TEN3 immunoreactivity was broadly found in the cerebellum, including the molecular, granular and Purkinje's cells layers and the white matter. Hybridization signals for Ten3 were found in layers 2–6 of the cerebral cortex (Zhou et al., 2003).

#### Ten4

Outside the nervous system, Ten4 mRNA transcripts are first detected in the mesentery of the gut, as early as E7.5 (Ben-Zur et al., 2000). Expression is then detected in the posterior somites and the tail bud at E9.5, and by E10.5 Ten4 mRNA was also observed in the periocular area and in the first, second and third branchial arches (Zhou et al., 2003). At E12.5, ample expression of Ten4 mRNA is detected in the facial mesenchyme, nasal epithelium, trachea, mesentery of the gut and urogenital system. By E18.5, Ten4 mRNA is expressed by the epidermis of the skin and the developing joints between bones in the limbs and in adipose tissue, but this expression subsides as birth approaches (Ben-Zur et al., 2000). It is noteworthy that Lossie et al. (2005) found a radically different pattern of Ten4 expression in the developing mouse, with transcripts found in the epiblast by E6.5, in the mesoderm of the developing embryo by E7.5, and expression exclusive to the tail bud and limbs in E11.5 embryos. It is likely that the probe used by Lossie et al. (2005) detected a splicing variant of Ten4 that was different from that of Ben-Zur et al. (2000) and Zhou et al. (2003). The existence of multiple splicing variants with specific patterns and timings of appearance in the developing embryo considerably increases the complexity of studying the morphology of the teneurins and represents a challenge to be overcome in future studies.

In the nervous tissue, Ten4 mRNA is detected at E7.5 in the neural plate. At E8.5, transcripts are detected in the caudal forebrain and the rostral midbrain region. Considerable expansion in expression occurs in E9.5, with Ten4 mRNA found in the alar and basal regions of the caudal diencephalon and in the midbrain–hindbrain boundary, including the caudal alar mesencephalon and the basal rostral rhombencephalon. At this stage, weak expression of Ten4 mRNA starts to be detected in the cortex. At 12.5E, low levels of expression are found in all layers of the cerebral cortex, with higher levels found in the mantle layer, and in the hippocampus. The diencephalic staining at this stage becomes sharper, with signals localized to the medial thalamus, the mammillary bodies, and the optic recess. The inferior colliculus and the optic tectum are also labeled at this stage, as well as the saccule. At E15.5, Ten1 and Ten4 show an overlapping expression in the cortex, with Ten4 mRNA found in a low-rostral/high-caudal gradient in both differentiating cells in the cortical plate and proliferating cells in the ventricular zone. Thalamic expression can be pinpointed to the dorsal lateral geniculate nucleus, ventral medial geniculate nucleus and posterior complex. Finally, at E16.5, Ten4 mRNA expression is seen in the inner neuroblastic layer of the retina. In the periphery, Ten4 mRNA is found in high levels in the dorsal root ganglia and in lower levels in the trigeminal ganglia. After birth, the low-rostral/high-caudal gradient of Ten4 mRNA expression in the cortex is maintained at least in P2 and P7, as well as in the thalamus and in the CA1 field of the hippocampus (Wang et al., 1998; Ben-Zur et al., 2000; Zhou et al., 2003; Li et al., 2006).

In 6-week old adults, Ten4 mRNA transcripts are found in the brain, liver and testes (Oohashi et al., 1999; Ben-Zur et al., 2000). In the hippocampus, expression of Ten4 is weakly found in the granular layer of the dentate gyrus and in the stratum lacunosum moleculare, as well as in the entire stratum pyramidale. Immunoreactivity to TEN4, on the other hand, was prominent in both molecular layers of the dentate gyrus, stratum lacunosum pyramidale and stratum oriens of the CA3 subfield. In the cerebellum, hybridization signals for Ten4 mRNA were found in the Purkinje's cells zone and in the white matter, while immunoreactivity was strongly found in the granular layer and only weakly observed in the molecular layer and white matter. Hybridization signals for Ten4 mRNA were found throughout layers 2–6 of the cerebral cortex and in the thalamus (Zhou et al., 2003).

#### Rattus norvegicus Ten1

The expression of Ten1 mRNA in the nervous system starts weakly at E16 and increases in intensity from E17 forward. In the E17 cerebral cortex, Ten1 mRNA is found in a low-rostral/highcaudal gradient, with signals mostly located to pyramidal cells in the cortical plate and subplate. In the thalamus, the anterior and intermediate parts were more strongly labeled than the posterior area. The midbrain, the hypothalamus, cerebellum, pons, medulla and spinal cord were also positive for signals. Proliferating zones were free of staining. The main olfactory bulb shows transcripts in the external, middle and internal tufted cells, and the accessory olfactory bulb is staining in its caudal-most area. By E19, the output cell layer of the accessory olfactory bulb is strongly stained, and the main olfactory bulb has uniform expression throughout the external plexiform layer. At this stage, staining intensity increases in the thalamus and the septum, and signals start to be detected in the subicular area, hippocampus, and dentate gyrus. By E21, signal is strong in the cortex, thalamus, septum and midbrain. At this stage, weak signals are detected in the rhinencephalon. After birth, several sites of Ten1 mRNA waned, including the cortex, the thalamus, the septum, the midbrain, the hypothalamus and the rhinencephalon. Staining in tufted cells was significantly decreased by P1 and vanished completely by P5, but granule cells started expressing Ten1 mRNA at P3. In the accessory olfactory bulb, Ten1 mRNA expression increased by P3 but vanished by P5. By P30, the expression of Ten1 mRNA increased in the granule cells of the dentate gyrus and in pyramidal cells of the hippocampus proper (Otaki and Firestein, 1999).

#### TCAP-1

The distribution of the C-terminal exon of Ten1 was investigated as a proxy for TCAP-1 expression in the adult rat brain by Wang et al. (2005). Several brain areas were positive for TCAP-1-corresponding transcripts, including allocortical areas

(olfactory bulb, piriform cortex, hippocampus proper and dentate gyrus), the central and basolateral nuclei of the amygdala, the ventromedial nucleus of the hypothalamus, the subthalamic nucleus, the vagus and hypoglossal nuclei, and the Purkinje cell layer of the cerebellum. **Figure 4** shows the main prosencephalic regions presenting TCAP-1 immunoreactivity.

#### Ten2

The only available data about TEN2 synthesis and Ten2 expression in the rat is in the developing teeth. Ectomesenchymal cells of the dental papilla subjacent to the internal enamel epithelium layer are immunoreactive to TEN2 starting at bell stage, by E20. Later at this stage there is an increase in labeling intensity in the odontoblast cell layer adjacent to pre-ameloblasts. In postpartum animals (P0-P7), TEN2 immunoreactivity is still detected in odontoblasts during crown formation. In mature teeth, TEN2 is diffusely distributed in the cell bodies and processes of odontoblasts of the coronal and radicular pulps. The expression patterns of Ten2 mRNA and that of exon 28 of Ten2 (corresponding to the TCAP-2 region of the protein) were coherent with that of the protein (Torres-da-Silva et al., 2017).

#### Homo sapiens TEN2

Torres-da-Silva et al. (2017) investigated immunoreactivity to TEN2 in the dental pulp of adult humans. Coronal dental pulp fragments evidenced a uniform distribution of TEN2 immunoreactivity only in odontoblasts. The authors report that, sometimes, the initial segment of the odontoblastic process was preserved, showing TEN2-immunoreactivity, similar to rat odontoblasts. RT-PCR analysis confirmed expression of TEN2 and that of exon 29 of TEN2 (corresponding to the TCAP-2 region of the protein) in human coronal pulp samples. Tews et al. (2014, 2017) detected the expression of TEN2 mRNA in adipocyte precursor cells, with higher expression on white cell precursors when compared to brown cell precursors.

#### TEN4

Graumann et al. (2017) investigated TEN4 expression in the ovary of normal and tumoral patients. The expression of TEN4 mRNA was detected by PCR in the normal ovary.

# THE DISTRIBUTION OF ADGRLs

#### Caenorhabditis elegans

The expression of lat-1 reporter begins on the zygote and is more evident in the AB lineage. In early development, lat-1 is broadly expressed in a stripped pattern visible in epidermal and pharyngeal precursors during dorsal intercalation. In larval and adult stages, lat-1 is expressed in the pharynx, the nervous system, the gonad and the vulva (Langenhan et al., 2009). The expression of lat-2 is generally imbricated with lat-1 in the pharyngeal primordium and it is restricted to the pharynx and the excretory cell in later stages (Langenhan et al., 2009).

FIGURE 4 | The telencephalic expression of TCAP-coding mRNA in the rat. Darkfield photomicrographs of coronal male rat brain slices that underwent in situ hybridization for the localization of TCAP-coding mRNA. It is noteworthy that TCAP-coding mRNA staining pattern is diffuse and seldom aggregate in cells, as seen for other markers. (A) The piriform cortex has one of the highest concentrations of mRNA, particularly in layer 2 and, to a lesser extent, in layer 3. (B) The pyramidal layer of the dentate gyrus also contains a high density of in situ hybridization labeling, while some degree of staining can also be seen in CA3 of the hippocampus proper. (C) The somatosensorial cortex has ample expression of TCAP-coding mRNA, with low expression in layer 1 and high, uniform expression on layers 5 through 6. CA3, cornus ammonis 3 of the hippocampus proper; Ctx, somatosensorial cortex, layer 1; MoDG, molecular layer of the dentate gyrus; PoDG, polymorphic layer of the dentate gyrus; Pir1, piriform cortex, layer 1; Pir2, piriform cortex, layer 2; Pir3, piriform cortex, layer 3. Scale bar: 100 µm. Based on the data published by Wang et al. (2005).

#### Drosophila melanogaster

fnins-13-00425 May 3, 2019 Time: 17:43 # 16

The main sites of Cirl expression in D. melanogaster are the larval CNS and the brain. Other enriched areas with Cirl mRNA expression are the thoracoabdominal ganglion, the salivary gland, the head and the eyes<sup>1</sup> (van der Voet et al., 2016). Cirl is expressed in several peripheral sensory neurons of the D. melanogaster larva, including type I and type II neurons. Expression of this gene was most prominent in larval pentascolopidial chordotonal organs (Scholz et al., 2015).

#### Gallus gallus

Doyle et al. (2006) investigated the presence of ADGRL2 on heart formation in stage 5, 16 and 21 of chicken embryos. By using in situ hybridization, the author observed ADGRL2 expression in the mesoderm leaving the primitive streak of the stage 5 embryo and also a weak labeling in the epiblast. In the stage 16, ADGRL2 expression was detected in forming somites, notochord, nephric ducts and the cardiac endothelium. At stage 21, immunostaining was observed in both the endothelium and the mesenchyme of the cardiac cushions as well as the muscle.

#### Mus musculus

In situ hybridization of P7 mouse brain slices revealed Adgrl1 expression in layer 5 of the cerebral cortex, the anterodorsal and anteroventral nuclei of the thalamus and the internal and external glanular layers of the cerebellum. Signals were also detected in the subiculum and the CA1 region of the hippocampus. Immunostaining in P14 mice confirmed the presence of ADGRL1 in layer 5 of the cerebral cortex and in the anterodorsal and anteroventral nuclei of the thalamus (Zuko et al., 2016).

#### Rattus norvegicus

Northern blot analysis of different adult rat tissues analyzed revealed the presence of Adgrl1 exclusively in the brain, and not in liver, heart, lung, kidney, spleen, skeletal muscle, and duodenum. In the brain, Western Blot of different brain areas revealed that ADGRL1 was most preeminent in the striatum, somewhat lower in cortex and hippocampus, and much less in the cerebellum (Krasnoperov et al., 1997).

Matsushita et al. (1999) investigated the distribution of Adgrl1 mRNA in northern blots of RNA isolated from different adult rat tissues. Adgrl1 mRNA was almost exclusively brain-specific, with very low levels of Adgrl1 detected in kidney, lung and spleen. The concentration of Adgrl1 mRNA in the brain is leastways 50-fold higher in the brain than in any other tissue. No ADGRL1 was detected in liver samples by Western Blotting.

Matsushita et al. (1999) found that Adgrl2 mRNA expression was prevalent in liver, lung and brain tissues, but were found to variable extent in all tissue tested (heart, spleen, muscle, and kidney). Disagreeing with the RNA expression, however, Western Blot experiments performed by Matsushita et al. (1999) could not find the ADGRL2 protein in the brain. Matsushita et al. (1999) also investigated the distribution of Adgrl3 mRNA in different adult rat tissues. As is the case of Adgrl1, Adgrl3 mRNA was found mostly in the brain and in lower amounts in the lung and spleen.

<sup>1</sup>www.flyatlas.org

#### Homo sapiens

According to Sugita et al. (1998), ADGRL1 mRNA was largely enriched in human brain samples, but longer exposure times revealed ADGRL1 expression in several other tissues, including the placenta, lung, kidney and pancreas. PCR confirmed that ADGRL1 is expressed in fibroblasts. The authors suggest that the failure to observe ADGRL1 expression outside the CNS in other studies was probably because of the short exposure times used.

Compared to ADGRL1 mRNA, ADGRL2 mRNA presented a substantial different organization because ADGRL2 mRNA expression is ubiquitous and uniformly distributed in all tissues. The highest expression of ADGRL2 was observed in placenta and lung, and the lowest were observed in brain and liver (Sugita et al., 1998). According to Ichtchenko et al. (1999), employing northern blot on adult human tissues, ADGRL2 mRNA expression is detected almost in all tissues tested. The highest expression was detected in placenta, heart, lung, kidney, pancreas, spleen, and ovary. Moderate expression was seen in brain, liver, and testis. Weak expression was observed in the skeletal muscle and thymus and peripheral blood leukocytes did not presented ADGRL2 mRNA expression.

According to Sugita et al. (1998), ADGRL3 mRNA expression was only observed in the human brain. Subsequently, Ichtchenko et al. (1999) described that ADGRL3 mRNA was expressed mainly in the brain. Weak expression was reported in heart, placenta, pancreas, kidney, and testis. Northern blot analysis of different human brain tissue samples performed by Arcos-Burgos et al. (2010) showed significant expression of ADGRL3 mRNA in amygdala, caudate nucleus, cerebellum and cerebral cortex. Lower expression was found in corpus callosum, hippocampus, whole brain extract, occipital pole, frontal lobe, temporal lobe, and putamen. No expression was detected in thalamus, medulla and spinal cord. In situ hybridization in human brain of different ages revealed strong cytoplasmatic signals in the amygdala, caudate nucleus, pontine nucleus and in Purkinje cells of the cerebellum at all ages tested. Weak signals were detected in cingulate gyrus neurons in the 2- and 5-year old, but not in the 8- and 30-year old, and in indusium griseum neurons in the 2-year old. Areas of the brain that were labeled by in situ hybridization also were labeled by immunohistochemistry.

# CONCLUSION

The teneurin-latrophilin system is a remarkable model for the study of protein-receptor interactions. The study of this system gives us a window into the fascinating exchanges that occurred between prokaryotes and basal eukaryotes, the repurposing of bioactive molecules once they are acquired, and the increasing complexity of neuropeptidergic systems as metazoans evolved. In particular, the teneurin system is remarkable in the sense that it is possible to build a phylogenetic tree based on the distribution and temporal dynamics of Ten expression that will result in a similar tree based on sequence similarity. Ten1 and Ten4 genes, in the available models, share the early and widespread

expression of the C. elegans teneurin that results in the long isoform. On the other hand, Ten2 and Ten3 are expressed late during development and became more pronounced in the nervous system and, in particular, in axonal guidance, resembling the C. elegans teneurin that results in the short isoform. The appearance of alternative regulation of teneurin expression after the divergence of Cnidaria/Ctenophora may have fundamentally impacted the way the teneurins acquired new functions and new corresponding distributions as nervous system complexity increased. In a similar vein, the appearance of isoforms before the nematode lineage likely facilitated the Insecta lineage to develop two teneurin paralogs that resemble the Ten1/Ten4 and the Ten2/Ten3 in terms of morphology, despite no direct homology between these proteins. This is a case of convergent evolution facilitated by a molecular event common to the species involved. It is interesting to question, however, how much causation there is between the increase in teneurin complexity in metazoans and the increase in complexity in sensory systems and in the nervous system as a whole. If certainly not the only promoter in that increasing complexity, it is hard to imagine that more complex sensory systems could have evolved without a system in place to ensure the correct connectivity, and the teneurins-latrophilins must have contributed in that process. If this is the case, teneurinlatrophilin interactions can be predicted to play an essential role in human physiology. As reviewed in this work, however, the study of those interactions is undermined by insufficient morphological data and a limited number of animal models

#### REFERENCES


employed. New studies are urged to fill the gaps and facilitate our understanding of this system.

#### AUTHOR CONTRIBUTIONS

LS wrote the text and reviewed the literature. GD designed and wrote the text and reviewed the literature. JH-J reviewed the literature. CC reviewed the literature. JB designed and wrote the text.

# FUNDING

This article was supported by Fundação de Amparo à Pesquisa do Estado de São Paulo (São Paulo Research Foundation – FAPESP) grants numbers 2016/02224-1, 2016/02748-0, and 2019/07844-6. We also would like to thank Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (Agency for the Advancement of Higher Education–CAPES). JB is a CNPq (Conselho Nacional de Desenvolvimento Científico e Tecnológico) Investigator.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnins. 2019.00425/full#supplementary-material


homologous to the Drosophila pair-rule gene product Ten-m. Dev. Biol. 216, 195–209. doi: 10.1006/dbio.1999.9503


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Sita, Diniz, Horta-Junior, Casatti and Bittencourt. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

fnins-13-00425 May 3, 2019 Time: 17:43 # 19

# Activity of the Carboxy-Terminal Peptide Region of the Teneurins and Its Role in Neuronal Function and Behavior in Mammals

David W. Hogg, Mia Husic, David Wosnick, Thomas Dodsworth, Andrea L. D'Aquila and ´ David A. Lovejoy\*

Department of Cell and Systems Biology, University of Toronto, Toronto, ON, Canada

Teneurin C-terminal associated peptides (TCAPs) are an evolutionarily ancient family of 40- to 41-residue bioactive peptides located on the extracellular end of each of the four teneurin transmembrane proteins. TCAP-1 may exist as a tethered peptide at the teneurin-1 carboxy end or as an independent peptide that is either released via post-transcriptional cleavage from its teneurin-1 pro-protein or independently expressed as its own mRNA. In neurons, soluble TCAP-1 acts as a paracrine factor to regulate cellular activity and neuroplastic interactions. In vitro studies indicate that, by itself, synthetic TCAP-1 promotes neuron growth and protects cells from chemical insult. In vivo, TCAP-1 increases hippocampal neuron spine density, reduces stressinduced behavior and ablates cocaine-seeking behaviors. Together, these studies suggest that the physiological effects of TCAP-1 are a result of an inhibition of corticotropin-releasing factor (CRF) activity leading to increased energy production. This hypothesis is supported by in vivo functional positron emissions tomography studies, which demonstrate that TCAP-1 significantly increases glucose uptake in rat brain. Complimentary in vitro studies show that enhanced glucose uptake is the result of TCAP-1-induced insertion of the glucose transporter into the neuronal plasma membrane, leading to increased glucose uptake and ATP production. Interestingly, TCAP-1-mediated glucose uptake occurs through a novel insulin-independent pathway. This review will focus on examining the role of TCAP on neuronal energy metabolism in the central nervous system.

Keywords: TCAP, teneurin, metabolism, glucose, stress, peptide evolution

# INTRODUCTION

The teneurins are a family of type-II transmembrane glycoproteins that are widely expressed in the central nervous system (Baumgartner et al., 1994; Levine et al., 1994). They are involved in a number of cellular processes including the regulation of synaptic adhesion and maintenance of synaptic structures (Hong et al., 2012; Mosca et al., 2012; Boucard et al., 2014; Mosca and Luo, 2014; Li et al., 2018). Initially discovered in Drosophila, the teneurins were first thought to be related to the tenascin proteins. Baumgartner and colleagues were in search of Drosophila

Edited by:

Jae Young Seong, Korea University, South Korea

#### Reviewed by:

Gina Leinninger, Michigan State University, United States Tamas Kozicz, Mayo Clinic, United States

\*Correspondence: David A. Lovejoy david.lovejoy@utoronto.ca

#### Specialty section:

This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Neuroscience

Received: 19 February 2019 Accepted: 22 May 2019 Published: 31 July 2019

#### Citation:

Hogg DW, Husic M, Wosnick D, ´ Dodsworth T, D'Aquila AL and Lovejoy DA (2019) Activity of the Carboxy-Terminal Peptide Region of the Teneurins and Its Role in Neuronal Function and Behavior in Mammals. Front. Neurosci. 13:581. doi: 10.3389/fnins.2019.00581

**183**

orthologs of the vertebrate tenascins when they identified two proteins and subsequently named them tenascin-like protein major (ten-m) and tenascin-like protein accessory (ten-a) (Baumgartner and Chiquet-Ehrismann, 1993; Baumgartner et al., 1994). At the same time, Levine et al. (1994) conducted a screen for novel phosphotyrosine-containing proteins when they identified a protein with homology to the tenascin family and named it odd Oz (Odz), as the odz mutant embryos did not contain odd-numbered body segments. Shortly after, a tenm homolog was identified in chicken and termed teneurin-1, due to its robust expression pattern in the nervous system (Minet et al., 1999). Since then, these proteins were found to be structurally and functionally distinct from the tenascins, and the name teneurin was adopted as the standard nomenclature to reflect their initial discovery and their neuronal expression patterns in various organisms (Tucker and Chiquet-Ehrismann, 2006). Subsequent homology studies revealed that the teneurins are present across most metazoans. Four homologs have been identified in vertebrates, whereas invertebrates contain only one copy, with the exception of insects, where two teneurin paralogs have been discovered (Baumgartner and Chiquet-Ehrismann, 1993; Baumgartner et al., 1994; Lovejoy et al., 2006; Chand et al., 2013a).

Teneurin genes encode proteins approximately 2800 amino acids in length that contain an intracellular amino terminal, a single transmembrane region and a large conserved extracellular carboxy terminal domain (Tucker et al., 2012). The four teneurin proteins (teneurin-1-4) exhibit a high degree of structural similarity to each other, as indicated by conservation of eight tenascin-type epidermal growth factor-like repeats, a cysteine rich region, five NHL domains and 26 tyrosine-aspartic acid repeats (Minet and Chiquet-Ehrismann, 2000; Kenzelmann et al., 2008; Tucker et al., 2012; Beckmann et al., 2013). Toward the distal end of the teneurin C-terminus is a short bioactive peptide sequence, termed teneurin C-terminal associated peptide (TCAP). The structure of the teneurins suggests that they evolved from a horizontal gene transfer event, which was likely mediated by a genomic internalization of a polymorphic proteinaceous toxin payload from a prokaryote donor to a choanoflagellate. This is corroborated by the widespread, yet highly conserved expression pattern of teneurins across various cell types (Drabikowski et al., 2005; Zhang et al., 2012; Chand et al., 2013b; Ferralli et al., 2018). Teneurins are most highly expressed within the central nervous system, where they form hetero- or homo-dimers to facilitate downstream signaling (Baumgartner et al., 1994; Kenzelmann et al., 2008), though their prevalence in other non-neuronal tissue types has been well-established (Baumgartner and Chiquet-Ehrismann, 1993; Baumgartner et al., 1994).

The TCAP region of the teneurins exhibits several characteristics of an independent peptide. It was first discovered during a screening for corticotropin-releasing factor (CRF) related homologs in a rainbow trout hypothalamic cDNA library using a hamster urocortin probe (Qian et al., 2004). A clone of the C-terminal region of rainbow trout teneurin-3 was isolated and termed TCAP-3. Subsequently TCAP-1, -2 and -4 were identified, totaling four highly conserved paralogs (Wang et al., 2005). The TCAP sequence is 40–41 amino acids in length and shares a structural similarity with CRF, calcitonin and other members of the Secretin superfamily of peptides (Lovejoy et al., 2006). TCAP shows widespread genetic expression in various brain regions, with in situ hybridization analyses showing TCAP mRNA expression in the murine olfactory bulb, cerebellum and brainstem (Wang et al., 2005). Though TCAP and teneurin share considerable overlap in expression patterns, the TCAPs are discretely expressed in some cortical regions that lack may teneurin expression, such as regions involved in the limbic system (Zhou et al., 2003; Wang et al., 2005; Chand et al., 2013a). These studies demonstrate that TCAP and teneurin may be differentially expressed, and ultimately suggest that TCAP processing may be independent from that of the teneurins.

### EVIDENCE OF TCAP AS A SMALL SOLUBLE PEPTIDE

To determine if TCAP exists as a separate gene that produces a soluble peptide independent of teneurin, the existence of TCAPspecific mRNA was investigated by Northern blot (Chand et al., 2013a). Antisense probes to the terminal exon of all four mouse teneurins labeled the four full-length teneurin transcripts, as expected. However, in the case of teneurin-1 and -3, shorter mRNA species less than 800 bases in length were readily identified. Sequence analysis of the teneurin-1 cDNA after 5<sup>0</sup> RACE PCR-based cloning revealed a transcript of 485 bases, corresponding to the mRNA sequence encompassing two-thirds of the terminal teneurin-1 exon oriented toward the 3<sup>0</sup> end (Chand et al., 2013a). The TCAP-1 mRNA is characterized by a short 5<sup>0</sup> untranslated region, a peptide-encoding region of 357 bases corresponding to 118 translated residues and a 3<sup>0</sup> untranslated region of about 74 or more bases. Within the peptide-encoding region additional postulated furin and basic residue cleavage sites were identified using the criteria suggested by Seidah and Chrétien (1997). These sites could potentially liberate smaller peptides of 107 and 41 residues. A 13 kDa band was identified by Western blot, which could correspond to either a furin cleavage product or the full-length translation product of the TCAP-1 mRNA (Chand et al., 2013a). A smaller 5 kDa band was shown in protein extracts from the vase tunicate, Ciona intestinalis, which could represent a soluble form of its 39-amino acid TCAP (Colacci et al., 2015; D'Aquila et al., 2017).

There is no evidence of a signal peptide in the putative translation product of the TCAP-1 mRNA, or in the equivalent TCAP regions of teneurin-2, -3, and -4 (Wang et al., 2005; Lovejoy et al., 2006; Chand et al., 2013a). Numerous bioactive peptide hormones and paracrine factors do not possess signal peptides. However, those belonging to the Secretin superfamily of ligands typically do contain such structures. Signal peptides facilitate entry into the vesicles of the secretory pathway, and peptides without this region typically remain in the cytosol. The full-length teneurins, however, do possess the hydrophobic transmembrane region that allow them to be inserted into the plasma membrane via fusion with secretory vesicles (Baumgartner et al., 1994; Levine et al., 1994). This

distinction in cellular localization is apparent in the differential expression of immunoreactive teneurin-1, which is primarily found at the plasma membrane, and immunoreactive TCAP-1, which is confined to the cytosol (Chand et al., 2013a).

Taken together, these findings indicate that a soluble TCAP peptide could be liberated by direct cleavage from the teneurins, or in the case of TCAP-1 and possibly TCAP-3, transcribed as a smaller, independent mRNA that specifically encodes the TCAP region. The mechanism by which TCAP could be cleaved directly from the teneurins is not clear, though furin or prohormone convertases associated with secretory vesicles could be responsible for this (Seidah and Chrétien, 1997). However, this supposition remains mostly theoretical, as it has not been possible to confirm the existence of an endogenous soluble 40- or 41 mer TCAP.

#### IDENTIFICATION OF THE TENEURIN/TCAP RECEPTOR

Although both teneurin and TCAP possess clear cellular action, the receptor mechanisms by which these effects occur were poorly understood until recently (Baumgartner et al., 1994; Levine et al., 1994; Qian et al., 2004). Early studies on the structure of TCAP showed that it contains several amino acid motifs found in peptides of comparable sizes belonging to the CRF and calcitonin families, suggesting a phylogenetic relationship between them (Lovejoy and Jahan, 2006; Lovejoy et al., 2006). Further evidence indicated that TCAP may be related to the Secretin peptide superfamily, and thus, its receptor may be part of the Secretin superfamily of GPCRs (Qian et al., 2004; Wang et al., 2005; Lovejoy and de Lannoy, 2013). However, receptor binding and activation studies showed that this is not the case, as Secretin GPCR family members had no significant interaction with TCAP-1 (Lovejoy and Barsyte-Lovejoy, unpublished observations).

In an attempt to further elucidate a putative receptor for TCAP-1, gene array studies investigating changes in gene expression of immortalized murine neuronal cells upon TCAP-1 treatment were performed. These studies showed that TCAP-1-treated cells had higher expression of dystroglycan genes than vehicle-treated cells did (Chand et al., 2012), suggesting a functional relationship between the two proteins. The dystroglycans are transmembrane proteins associated with several receptor systems, particularly those of certain growth factors, integrins and other intercellular adhesion factors (Peng et al., 2008; Lombardi et al., 2017). Interestingly, no association between dystroglycans and Secretin GPCRs has been established to date. Fluorescence studies using a FITC-labeled variant of TCAP-1 and immunolabeled dystroglycan revealed that the two are, indeed, proximal to each other on the cell membrane. Moreover, TCAP-1 treatment induces activation of a MEK-ERK signal transduction system associated with the dystroglycans (Chand et al., 2012). Despite this, no direct evidence of binding between the two has been observed to date, indicating that they likely do not form a receptor-ligand pair, and may simply be parts of a larger intercellular complex.

Further research into teneurin binding partners has revealed a putative receptor for both teneurin and TCAP. Although full-length teneurins were previously shown to homo- and heterodimerize, leading to activation of downstream signaling cascades (Kenzelmann et al., 2007), the teneurins also bind to the GPCR Latrophilin-1 (LPHN1). Together, they form a trans-synaptic complex with both adhesion and cell signaling properties (Silva et al., 2011; Boucard et al., 2014; Vysokov et al., 2016). The LPHNs comprise a group of three GPCRs (LPHN1- 3) that were first discovered in search for a calcium-independent receptor of α-latrotoxin, the primary vertebrate toxin in black widow spider (genus Lactrodectus) venom (Davletov et al., 1996). Upon their discovery, the LPHNs were initially classified as Secretin GPCRs based on the high sequence similarities of their putative hormone-binding domain (HBD) with the signature HBDs of the Secretin GPCRs (Lelianova et al., 1997). They have since been reclassified as members of the Adhesion GPCR family due to their newly discovered adhesion functions and their long extracellular domains which contain several adhesion motifs (Fredriksson et al., 2003; Nordström et al., 2009). Recent phylogenetic analyses indicate that the Adhesion GPCR family is ancestral to the Secretin GPCR family, suggesting that Secretin GPCRs inherited their HBDs from their Adhesion GPCR ancestors (Nordström et al., 2009). If this is the case, other Adhesion GPCRs may also possess additional peptide ligands that have yet to be discovered. In this respect, TCAP may act as a model system to understand peptide-receptor interactions amongst the Adhesion GPCRs.

Further to these phylogenetic studies, several recent binding studies provide clear evidence that the TCAP region is required for teneurin-LPHN interaction. LPHN1 binds LPHN1-associated synaptic surface organizer (Lasso), a splice variant of teneurin-2 comprised of the protein's distal C-terminus, including the TCAP-2 region (Silva et al., 2011). Full-length teneurin and Lasso both exhibit a high affinity with LPHN1, suggesting that TCAP itself might likewise bind with the LPHN family as an independent ligand. Deletion of the teneurin C-terminus reduces binding of teneurin with the extracellular domains of LPHN1 (Silva et al., 2011). Additionally, recent structural studies indicate that the teneurin extracellular domain forms a barrellike structure from which the TCAP-containing C-terminus protrudes (Jackson et al., 2018; Li et al., 2018). This conformation may make the TCAP portion of teneurin accessible to potential interacting partners, such as LPHN1, on adjacent cells and allow it to interact with said partners as either an active site of the full teneurin protein or independently as a cleavable peptide. Moreover, an interaction may occur directly between TCAP and LPHN1 at the LPHN1 HBD, as this is the region of peptide binding in Secretin GPCRs. It is also the most highly conserved region among the three LPHN isoforms and is involved in binding with other LPHN ligands such as α-latrotoxin (Holz and Habener, 1998; Krasnoperov et al., 1999). Recent studies showed that a transgenically expressed TCAP-1 construct can be immunopreciptated with a transgenically expressed HBDcontaining fragment of LPHN1, indicating that an interaction does occur between these two proteins (Husic et al., 2018 ´ ). In addition, over-expression of LPHN1 in HEK293 cells results in increased uptake of TCAP-1 and subsequent cytoskeletal reorganization consistent with what has been observed in other cell types upon TCAP-1 treatment, further indicating a functional interaction between these two proteins (Al Chawaf et al., 2007a; Chand et al., 2012; Husic et al., 2018 ´ ).

# BIOACTIVITY OF TCAP-1

fnins-13-00581 July 31, 2019 Time: 17:13 # 4

Numerous studies have indicated potent bioactivity for synthetic TCAP-1. Treatment of Gn11 cells with multiple doses of TCAP-1 showed that a low dose (1 nM) leads to an increase in intracellular cAMP levels whereas a high dose (100 nM) decreases cAMP (Wang et al., 2005). Immortalized murine hypothalamic cells also display higher survivability under stress conditions when treated with TCAP-1, where TCAP-1 treatment led to increased total cell number and decreased necrotic cell number after 48 h in cells grown under high pH conditions (Trubiani et al., 2007). Likewise, cells exposed to peroxide exhibited decreased cell death when treated with TCAP-1, pointing to a role for the peptide in neuroprotection. In addition to this, TCAP-1 may also have actions in neuroplasticity, as it has potent effects on cytoskeletal dynamics. Treatment with TCAP-1 increases β-actin and β-tubulin in murine immortalized hypothalamic cells and increases neurite outgrowth in a hippocampal cell line (Al Chawaf et al., 2007a; Tan et al., 2012). It also induces filamentous actin polymerization in a variety of cell lines through activation of a dystroglycan-associated MEK-ERK downstream signaling cascade (Chand et al., 2012; Husic et al., 2018 ´ ).

These effects were further elucidated in an in vivo study utilizing intracerebroventricular injection of TCAP-1 under unstressed and restraint conditions. Under unstressed conditions, TCAP-1 decreased dendritic branching while under stressed conditions it increased dendritic branching (Al Chawaf, 2008). Although the mechanism is still unknown, these results provide evidence that TCAP-1 plays a role in neuroplasticity.

Immunohistochemical analysis has revealed that immunoreactive TCAP-1 is present within the limbic system, particularly in the areas associated with regulation of the behavioral stress response, such as the pyramidal layer of the hypothalamus and the basolateral nucleus of the amygdala (Tan et al., 2012). Neurons in these regions are morphologically plastic and can change in response to stimuli such as stress and learning (Tan et al., 2009, 2011; Chen et al., 2013). The effects of TCAP-1 on stress-related behavior have been observed using several methodologies including the acoustic startle response (ASR), an indicator of anxiety and a test of reflexive fear response (Rotzinger et al., 2010). Through these studies, two major bioactive attributes of TCAP-1 have been identified: its neuromodulatory effects and its regulation of the CRF-induced stress response. In one such study, rats were separated based on their baseline ASR response prior to any treatment, and subsequently treated intracerebroventricularly with TCAP-1 (Wang et al., 2005). Rats with a low baseline ASR response exhibited an increase of in response upon TCAP-1 treatment. In contrast, rats with a high baseline ASR showed an attenuated response after treatment with TCAP-1. TCAP-1 was also shown to induce long-term attenuation of the stress response, as rats had a 50% reduction in ASR up to 15 days after TCAP-1 treatment (Wang et al., 2005). Moreover, pretreatment with TCAP-1 is able to modulate CRF-induced stress responses in several behavioral paradigms including ASR, elevated plus maze and open field tests (Al Chawaf et al., 2007b; Tan et al., 2008; Rotzinger et al., 2010). CRF-induced cocaine reinstatement is also reduced in rats given TCAP-1 pre-treatment (Kupferschmidt et al., 2011; Erb et al., 2014). These studies highlight the role TCAP-1 has in regulating CRF-associated behaviors related to anxiety and depression in rodent models (**Figure 1**).

# ROLE OF ENERGY METABOLISM BY TCAP

In both animal and cell models, synthetic TCAP-1 activates several processes that necessitate increased energy production. This includes protection of neurons against alkaline chemical insults and cell death (Trubiani et al., 2007), stimulation of neurite outgrowth, reorganization of cytoskeletal elements in neurons (Al Chawaf et al., 2007a; Tan et al., 2011; Chand et al., 2012), and modulation of stress-related behaviors (Wang et al., 2005; Al Chawaf et al., 2007b; Tan et al., 2008). These actions are energetically costly, indicating that TCAP-1 may also stimulate energy production to maintain them, as cellular supply of ATP must meet cellular energy demand.

Glucose is the preferred energy substrate in brain, and a steady supply is critical for neuronal function; however, neurons have a limited capacity to store glucose intracellularly. Therefore, a TCAP-1-mediated increase in intracellular glucose would indicate that TCAP-1 can also stimulate cellular energy metabolism. In a study using functional positron emission

tomography, a single subcutaneous injection of TCAP-1 induced a significant increase in <sup>18</sup>F-deoxyglucose uptake into the brain 3 days post-treatment (Hogg et al., 2018). Glucose uptake was highest in the frontal cortex and subcortical regions, although it occurred throughout the cortical regions. In addition, a single injection of TCAP-1 decreased whole animal blood glucose by 35–40%, with a concomitant decrease in serum insulin and an increase in serum glucagon. This pattern mimics the effect of insulin on blood glucose and glucagon, demonstrating that TCAP-1 alters whole animal glucose metabolism. Similar results were obtained in Goto-Kakizaki rats, a type-II diabetic insulin-insensitive pathological model, suggesting that TCAP-1-stimulated glucose uptake is independent of the insulin system.

In confirmation of these studies, TCAP-1 also increases glucose uptake in a hypothalamic neuron cell model. In these cells, TCAP-1 increased uptake of [3H-]2-deoxyglucose, a non-hydrolysable form of glucose, by 50% following 60 and 90 min of treatment. This profile differed from that of insulin, which induced uptake at 30 min post-treatment, further indicating that TCAP-1 regulates glucose uptake independently of insulin. This was confirmed when insulin- and TCAP-1 mediated glucose uptake were assessed in the presence and absence of a depolarizing stimulus. Insulin-mediated glucose uptake requires membrane depolarization for glucose transporter (GLUT) insertion into the plasma membrane (Uemura and Greenlee, 2006). Unlike insulin, TCAP-1-mediated glucose uptake occurrs in the absence of membrane depolarization events (Hogg et al., 2018). The timeline and depolarization-independent nature of TCAP-1-induced glucose uptake indicates that TCAP-1 activates a signaling mechanism distinct from that of insulinmediated glucose uptake.

Glucose uptake into neurons occurs by faciltated diffusion through GLUTs, and is dependent on the plasma membrane expression of GLUTs and the diffusion gradient of glucose into the cell. The actions of TCAP-1 are consistent with this mechanism. TCAP-1 increases GLUT3 transport to the plasma membrane of a hypothalamic neuronal model by 37.5% within 1 h, where the increase is maintained for up to 3 h post-treatment (Hogg et al., 2018). In addition, TCAP-1 increases GLUT3 immunoreactivity by ∼250% in the growth cones of extending neurites 1–2 h post-treatment. Moreover, TCAP-1 does not significantly increase membrane expression of GLUT1 or GLUT4, indicating that TCAP-1 induced glucose uptake likely occurs specifically through a GLUT3-mediated process. GLUT3 is the primary glucose transporter in brain, whereas the GLUT1 transporters are typically located in endothelial cells of the blood-brain barrier, and the GLUT4 transporters are highly expressed in skeletal muscle and adipose tissue and known to be insulin-dependent (Shepherd et al., 1992; Haber et al., 1993; Choeiri et al., 2002; Airley and Mobasheri, 2007).

Increases in intracellular glucose importation stimulate cellular energy metabolism. TCAP-1 significantly increases intracellular ATP turnover in a dose-dependent manner in an immortalized neuronal cell line, demonstrating that TCAP-1-mediated glucose uptake does, indeed, stimulate neuronal energy metabolism (Hogg et al., 2018). This increase in intracellular ATP turnover is likely the result of stimulation of aerobic energy-producing pathways, as TCAP-1 also decreases cellular lactate concentrations. Cellular pyruvate concentrations likewise decreased, indicating the metabolic pathway favors producing pyruvate for oxidative energy production. Previous studies have demonstrated that TCAP-1 can significantly increase catalase, superoxide dismutase and the superoxide dismutase copper chaperone (Trubiani et al., 2007), reducing intracellular reactive oxygen species. If TCAP-1 is increasing mitochondrial activity, there would be a consequential increase of reactive oxygen species production, thus the TCAP-1 system may have evolved a mechanism to compensate this. Taken together, these data demonstrate that TCAP-1 is a functional component of the teneurin protein that regulates glucose uptake and neuronal energetics in the rodent brain.

# CONCLUSION

Teneurin C-terminal associated peptide represents a bioactive region of the teneurin proteins that may act as a tethered ligand or a distinct peptide that is either cleaved from the full-length teneurin protein or expressed independently (Chand et al., 2013a). TCAP-1 has several energetically favorable functions, and increases uptake of glucose into the rodent brain (Hogg et al., 2018). This indicates that it can act to increase energy availability, allowing for its other implicated functions to take place. The exact mechanism by which TCAP-1 acts is yet to be elucidated; however, TCAP-1-mediated glucose uptake appears to be independent of insulin. Although further studies are required to tease out the precise signaling cascade that TCAP-1 induces to facilitate increased neuronal glucose uptake via GLUT3, the current studies presented in this review reveal a novel and essential function of this peptide family, further supporting TCAP as a critical stressresponse regulator.

#### AUTHOR CONTRIBUTIONS

DH: analysis of energy metabolism and description of key studies. MH: interpretation of receptor interaction and associated experiments, review and editing of the manuscript. DW: overview of aerobic studies on TCAP and associated literature. TD: molecular biology of TCAP receptor interactions and associated literature. AD'A: development of key experiments associated with TCAP action. DL: supervision of research program and overview of the manuscript.

#### FUNDING

The studies described in the manuscript have been funded by the Natural Sciences and Engineering Research Council (NSERC) Canada, and Protagenic Therapeutics, Inc., United States.

#### REFERENCES

fnins-13-00581 July 31, 2019 Time: 17:13 # 6


sympathetic neurons of dystrophic mdx mice. Mol. Cell. Neurosci. 80, 1–17. doi: 10.1016/j.mcn.2017.01.006


**Conflict of Interest Statement:** DL is a co-founder of Protagenic Therapeutics, Inc., a commercial entity that may have interests in the findings of these studies.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Hogg, Husi´c, Wosnick, Dodsworth, D'Aquila and Lovejoy. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# A Putative Role of Teneurin-2 and Its Related Proteins in Astrocytes

Gestter W. L. Tessarin1,2, Ola M. Michalec<sup>3</sup> , Kelly R. Torres-da-Silva1,2, André V. Da Silva2,4 , Roelf J. Cruz-Rizzolo<sup>1</sup> , Alaide Gonçalves <sup>1</sup> , Daniele C. Gasparini <sup>1</sup> , José A. C. Horta-Júnior <sup>2</sup> , Edilson Ervolino<sup>1</sup> , Jackson C. Bittencourt <sup>5</sup> , David A. Lovejoy <sup>3</sup> and Cláudio A. Casatti 1,2 \*

<sup>1</sup> Department of Basic Sciences, School of Dentistry of Araçatuba, São Paulo State University (UNESP), Araçatuba, Brazil, <sup>2</sup> Department of Anatomy, Institute of Biosciences of Botucatu, São Paulo State University (UNESP), Botucatu, Brazil, <sup>3</sup> Department of Cell and Systems Biology, University of Toronto, Toronto, ON, Canada, <sup>4</sup> School of Medicine, Federal University of Mato Grosso do Sul (UFMS), Três Lagoas, Brazil, <sup>5</sup> Department of Anatomy, Institute of Biomedical Sciences, São Paulo University (USP), São Paulo, Brazil

#### Edited by:

James A. Carr, Texas Tech University, United States

#### Reviewed by:

David Vaudry, Institut National de la Santé et de la Recherche Médicale (INSERM), France Timothy Mosca, Thomas Jefferson University, United States

> \*Correspondence: Cláudio A. Casatti claudio.casatti@unesp.br

#### Specialty section:

This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Neuroscience

Received: 23 November 2018 Accepted: 07 June 2019 Published: 27 June 2019

#### Citation:

Tessarin GWL, Michalec OM, Torres-da-Silva KR, Da Silva AV, Cruz-Rizzolo RJ, Gonçalves A, Gasparini DC, Horta-Júnior JAC, Ervolino E, Bittencourt JC, Lovejoy DA and Casatti CA (2019) A Putative Role of Teneurin-2 and Its Related Proteins in Astrocytes. Front. Neurosci. 13:655. doi: 10.3389/fnins.2019.00655 Teneurins are type II transmembrane proteins comprised of four phylogenetically conserved homologs (Ten-1-4) that are highly expressed during neurogenesis. An additional bioactive peptide named teneurin C-terminal-associated peptide (TCAP-1-4) is present at the carboxyl terminal of teneurins. The possible correlation between the Ten/TCAP system and brain injuries has not been explored yet. Thus, this study examined the expression of these proteins in the cerebral cortex after mechanical brain injury. Adult rats were subjected to cerebral cortex injury by needle-insertion lesion and sacrificed at various time points. This was followed by analysis of the lesion area by immunohistochemistry and conventional RT-PCR techniques. Control animals (no brain injury) showed only discrete Ten-2-like immunoreactive pyramidal neurons in the cerebral cortex. In contrast, Ten-2 immunoreactivity was significantly up-regulated in the reactive astrocytes in all brain-injured groups (p < 0.0001) when compared to the control group. Interestingly, reactive astrocytes also showed intense immunoreactivity to LPHN-1, an endogenous receptor for the Ten-2 splice variant named Lasso. Semi-quantitative analysis of Ten-2 and TCAP-2 expression revealed significant increases of both at 48 h, 3 days and 5 days (p < 0.0001) after brain injury compared to the remaining groups. Immortalized cerebellar astrocytes were also evaluated for Ten/TCAP expression and intracellular calcium signaling by fluorescence microscopy after TCAP-1 treatment. Immortalized astrocytes expressed additional Ten/TCAP homologs and exhibited significant increases in intracellular calcium concentrations after TCAP-1 treatment. This study is the first to demonstrate that Ten-2/TCAP-2 and LPHN-1 are upregulated in reactive astrocytes after a mechanical brain injury. Immortalized cerebellar astrocytes expressed Ten/TCAP homologs and TCAP-1 treatment stimulated intracellular calcium signaling. These findings disclose a new functional role of the Ten/TCAP system in astrocytes during tissue repair of the CNS.

Keywords: teneurin, teneurin c-terminal associated peptide, latrophilin, mechanical brain injury, cerebral cortex, reactive astrocytes, adult rat

# INTRODUCTION

Teneurins are type II transmembrane glycoproteins composed of four paralogues (Ten-1-4), mainly expressed during central nervous system (CNS) development (Baumgartner and Chiquet-Ehrismann, 1993; Baumgartner et al., 1994; Levine et al., 1994; Rubin et al., 1999; Tucker and Chiquet-Ehrismann, 2006). Teneurins encompass around 2800 amino acids, with the Nterminal intracellular domain consisting of approximately 300– 375 amino acids, which can be cleaved and translocated to the nucleus acting as a transcription factor or can mediate cytoskeletal interactions (Bagutti et al., 2003; Nunes et al., 2005; Tucker and Chiquet-Ehrismann, 2006). The transmembrane and C-terminal extracellular domains comprise 34 and 2400 amino acids, respectively (Bagutti et al., 2003; Nunes et al., 2005; Tucker and Chiquet-Ehrismann, 2006). The extracellular domain contains several sites for homophilic or heterophilic interactions and additional potential cleavage sites that can generate soluble signaling molecules (Bagutti et al., 2003; Nunes et al., 2005; Lovejoy et al., 2006; Tucker and Chiquet-Ehrismann, 2006; Tucker et al., 2012; Mosca, 2015). Splice variants of the teneurins are also found in vertebrates, transcribing some proteins such as a Ten-2 related protein named Lasso (latrophilin1-associated synaptic surface organizer), an endogenous ligand for the Gprotein-coupled receptor named latrophilin (LPHN) (Silva et al., 2011; Tucker et al., 2012; Boucard et al., 2014). Latrophilins are constituted by three isoforms (LPHN-1-3), also known as Adhesion G Protein-Coupled Receptor, subfamily L (ADGRL1- 3) (Meza-Aguilar and Boucard, 2014). Lasso can also undergo intracellular cleavage resulting in soluble molecules, which are secreted and may interact with LPHN in other cells (Vysokov et al., 2016). Recently, a study demonstrated similarities of the Ten-2 carboxyl terminal to bacterial Tc-toxins (Li et al., 2018). This region shows complex arrangements, permitting an heterophilic interaction with LPHN, which controls intracellular cyclic AMP (cAMP) (Li et al., 2018).

Teneurins are primarily involved with neuronal migration, axonal guidance, as well as formation, differentiation and maintenance of synapses in the CNS (Mosca, 2015; Antinucci et al., 2016). Previous studies have also correlated these proteins with mental disorders, congenital diseases and some types of tumors (Vinatzer et al., 2008; Ziegler et al., 2012; Boeva et al., 2013; Heinrich et al., 2013; Nakaya et al., 2013; Ivorra et al., 2014; Zhang et al., 2014; Bastías-Candia et al., 2015; Hor et al., 2015; Lovejoy and Pavlovic, 2015; Schöler et al., 2015; ´ Vater et al., 2015; Alkelai et al., 2016; Cheng et al., 2017; Graumann et al., 2017; Talamillo et al., 2017). Interestingly, teneurins contain a cleavage site in their carboxyl terminal that originates fragments of 40 to 41 amino acids, named teneurin Cterminal-associated peptides (TCAP-1-4), which show structural similarities to corticotrophin-releasing factor (CRF) (Qian et al., 2004; Wang et al., 2005; Lovejoy et al., 2006). In vitro studies using immortalized neurons have demonstrated that TCAP-1 stimulates neurite outgrowth, regulates brain-derived neurotrophic factor (BDNF) and acts as a neuroprotective agent (Al Chawaf et al., 2007a; Trubiani et al., 2007; Ng et al., 2012). In vivo studies in rats established that TCAP-1 treatment modulates dendritic morphology in hippocampal neurons and reduces FOS induction in neurons in limbic regions, stimulated by CRF intracerebral administration (Al Chawaf et al., 2007b; Tan et al., 2009, 2011). TCAP-1 also reduces stress-related behaviors and eliminates cocaine-seeking reinstatement in adult rats (Wang et al., 2005; Al Chawaf et al., 2007b; Kupferschmidt et al., 2011; Tan et al., 2011; Erb et al., 2014). A novel finding reported significantly increased glucose uptake in the rat brain 3 days after a single subcutaneous TCAP-1 injection, as well as decreased blood glucose 1 week later (Hogg et al., 2018). In vitro data corroborated the TCAP-1 action in the glucose metabolism in neurons, indicating that TCAP-1 can represent a peptide signaling substance that regulates glucose uptake, regardless of insulin-mediated glucose regulation (Hogg et al., 2018).

Preliminary unpublished screening assays performed in our laboratory indicated that Ten-2/TCAP-2 showed substantial changes in experimental brain disorders induced in adult rats (Tessarin and Casatti, unpublished data). Thus, we focused mainly on the correlations between Ten-2/TCAP-2 and reactive astrocytes. For this, controlled mechanical brain injury was induced by a metal needle insertion lesion in the cerebral cortex of adult rats, followed by immunohistochemistry and conventional RT-PCR analysis. Additionally, LPHN immunoreactivity was qualitatively analyzed by immunohistochemistry. In order to adopt an in vitro model for further studies, immortalized mouse cerebellar astrocytes were also characterized for Ten/TCAP homolog

**Abbreviations:** Ab1, primary antibody; Ab1(LPHN-1), LPHN-1 primary antibody omission; Ab1(Ten-2) Ten-2 primary antibody omission; aCSF, artificial cerebrospinal fluid; ADS, adsorption; AM, fluo-4 acetomethyl; ANOVA, analysis of variance; BBB, blood-brain barrier; BDNF, brain-derived neurotrophic factor; Ca2+, unbound calcium; CaCl22H2O, calcium chloride dihydrate; CCD, camera-coupled device; cDNA, complementary DNA; CEUA, Institutional Committee of Animal Welfare; CNS, central nervous system; CRF, corticotrophin-releasing factor; Cy<sup>3</sup> , cyanine; DAB, diaminobenzidine; DAPI, 4',6-diamidino-2-phenylindole; DGC, dystrophin-dystroglycan complex; ddH2O, double-distilled water; DEPC, diethyl pyrocarbonate; DMEM, Dulbecco's modified eagle's medium; DMSO,dimethyl sulfoxide; DNA, deoxyribonucleic acid; DNase, deoxyribonuclease; dNTP, deoxynucleotides; DOC2, double C2-like domain-containing protein; DTT, DL-Dithiothreitol; EtOH, ethanol; FBS, fetal bovine serum; FGF8, fibroblast growth factor; FITC, fluorescein isothiocyanate; GFAP-LI, glial fibrillary acidic protein-like immunoreactive; HEPES, 4-(2 hydroxyethyl)-1-piperazineethanesulfonic acid; KCl, potassium chloride; Kg, kilograms; LPHN, latrophilin; -LPHN-1, absence of LPHN-1 antibody/ LPHN-1-LI, latrophilin-1-like immunoreactive; LPHN-2-LI, latrophilin-2-like immunoreactive; LPHN-3-LI, latrophilin-3-like immunoreactive; µg, microgram; µL, microliter; MAPK, mitogen-activated protein kinase; mL, milliliter; mg, milligram; MgCl2, magnesium chloride; MgCl26H2O, magnesium chloride hexahydrate; Mm, millimol; mRNA, messenger RNA; n, number; NaCl, sodium chloride; NFL, neurofilament light; nM, nanomol; PBS, phosphate-buffered saline; PBS-T, phosphate-buffered saline and triton X-100; PCR, polymerase chain reaction; RCF, relative centrifugal force; RNA, ribonucleic acid; RNase, ribonuclease; rpm, revolutions per minute; RT-PCR, reverse transcription polymerase chain reaction; S100B, S100 calcium-binding protein B; SEM, standard error of the mean; TCAP, teneurin C-terminal-associated peptides; TCAP-1, teneurin C-terminal-associated peptide 1; TCAP-2, teneurin C-terminalassociated peptide 2; TCAP-3, teneurin C-terminal-associated peptide 3; TCAP-4, teneurin C-terminal-associated peptide 4;Ten-1, teneurin-1; Ten-2, teneurin-2; Ten-3, teneurin-3; Ten-4, teneurin-4; Ten-2-LI, teneurin-2-like immunoreactive; Uv, ultraviolet.

expressions and evaluated concerning calcium-signaling modulation after TCAP-1 treatment.

#### MATERIALS AND METHODS

#### In vivo Study Animals

Forty-five adult male Wistar rats (280–300 g) were obtained from the central animal facility at the School of Dentistry of Araçatuba (UNESP, Araçatuba, SP, Brazil) and maintained in the experimental room of the Morphology division of the Department of Basic Sciences for 15 days for environmental adaptation. The rats were kept in a 12/12 dark-light cycle (lights on at 7:00 am) under controlled temperature (23 ± 1 ◦C) and humidity (50–65%), as well as water and rat chow ad libitum. The experimental protocols for animal handling and care were approved by the Institutional Committee of Animal Welfare (CEUA, process number 2015-00318). All efforts were made to reduce the number of animals and to minimize suffering.

#### Surgery Procedures

Animals were anesthetized by intramuscular injection of ketamine (80 mg/ kg; Virbac, SP, Brazil) and xylazine (10 mg/ kg; Bayer, RS, Brazil), and then positioned in stereotaxic apparatus, where the scalp was incised along the midline using a scalpel blade. The brain was exposed after drilling, using a spherical bur coupled to a high-speed rotation handpiece. The dura-mater was discretely incisioned to expose the cerebral cortex. For mechanical brain injury, a needle-insertion injury was created through vertical insertion of a sterile metal needle (0.8 mm diameter) maintained for 5 s in the cerebral cortex. The coordinates for the cerebral cortex lesion were 6.63 mm (rostrocaudal axis), 1.5 mm (mediolateral) and 4 mm (dorsoventral, from the cortical surface) (Paxinos and Watson, 1998). After surgery, animals were kept in individual cages and divided into groups of 24 h, 48 h, 3 and 5 days of postoperative periods (n = 9 per experimental time, considering n = 5 for immunohistochemistry and n = 4 for conventional RT-PCR analysis). Rats from the control group (n = 9, considering n = 5 for immunohistochemistry and n = 4 for conventional RT-PCR analysis) were subjected to brain exposure with no dura-mater or brain injury (sham-surgery: control group).

## Immunohistochemistry

#### Tissue Preparation

Animals were anesthetized as previously described and transcardially perfused with heparinized saline solution at room temperature (RT) (100–150 mL) followed by cold fixative solution (1,000 mL) containing 4% formaldehyde (obtained from paraformaldehyde heated to 65◦C, #P6148, Sigma-Aldrich, MO, USA) diluted in phosphate buffer saline 0.1 M (PBS; pH 7.4). The brains were dissected and post-fixed in the same fixative solution for 4 h at 4◦C. Subsequently, the brains were cryoprotected in PBS with 30% sucrose (#1894-1, Dinâmica, SP, Brazil) overnight at 4◦C. Coronal 30 µm-thick histological sections were obtained in a freezing microtome (SM 2000R, Leica, HE, Germany) and stored in 12-well culture plates with anti-freezing solution (PB 0.025 M, NaCl 0.225%, sucrose 15% and ethylene glycol 35%) at −20◦C for further processing.

#### Immunoperoxidase Staining

Histological sections were submitted to immunoperoxidase staining method to obtain a detailed morphology of the neural cells exhibiting immunoreactivity to Ten-2. Initially, one series of histological sections (360µm intersection interval) was washed (3 × 10 min) in PBS and submitted to peroxidase endogenous inhibition using 0.3% hydrogen peroxide (#H3410, Sigma-Aldrich, MO, USA) in PBS, for 30 min at RT. Next, the sections were washed (3 × 10 min) in PBS and blocked for non-specific bindings using 5% non-fat milk in PBS with 0.03% triton X-100 (PBS-T, #100882547, X-100, Sigma-Aldrich, MO, USA), followed by 3% bovine serum albumin (A9647, Sigma-Aldrich, MO, USA) in PBS-T, for 1 h at RT for each blocking. Additional blocking was performed using 2% normal donkey serum (#017-000- 121, Jackson Immunoresearch, PA, USA) in PBS-T, overnight at 4◦C. Sections were initially incubated in primary polyclonal antibody anti-Ten-1-4 (Ten-1, 1:250, H00010178-A01, lot # 07310, ABNOVA, Taipei, Taiwan; Ten-2, 1:100, Lot # K1910, sc-165674, Santa Cruz Biotechnology, CA, USA; Ten-3, 1:500, Lot # B0910, sc 136918, Santa Cruz Biotechnology; Ten-4, 1:1000, Lot # B2610, sc-134883, Santa Cruz Biotechnology), diluted in PBS-T and 2% normal donkey serum, for 48 h at 4◦C. Then, the sections were incubated in biotinylated secondary antibody (for Ten-1, 1;800, Lot # X0623, BA-9200, Vector Laboratories Inc., CA, USA; for Ten-2, 1:800, lot # G0815, sc-2042, Santa Cruz Biotechnology; for Ten-3 and Ten-4, 1:800, Lot # E2213, sc-2089, Santa Cruz Biotechnology), followed by avidin-biotin complex (1:500, PK-6100, Vector Laboratories Inc., CA, USA) in PBS-T, for 1 h at RT each step. The immunoreaction was visualized by developing the sections in 0.05% diaminobenzidine as a chromogen (DAB, #32741, Sigma-Aldrich, MO, USA) with nickel ammonium sulfate (#N48-500, Fisher Chemical, NJ, USA) and 0.3% hydrogen peroxidase, under light microscope analysis for reaction control. After that, the sections were mounted on gelatin-coated slides and maintained for approximately 24 h at 55–56◦C in an oven. Finally, they were dehydrated in alcohol, cleared in xylenes and cover-slipped with DPX as a mounting medium (#06522, Sigma-Aldrich, MO, USA).

#### Double Immunofluorescence

This method was performed for detection and counting of nuclear astrocyte profiles (DAPI; glial fibrillary acidic protein— GFAP) exhibiting immunoreactivity to Ten-2 or to qualitatively analyse latrophilin immunoreactivity in the experimental groups. For this, one series of histological sections was washed (3 × 10 min) in PBS and submitted to the same blocking steps for elimination of possible unspecific antibody interaction, as previously mentioned. Sections were incubated in primary polyclonal antibody anti-Ten-2 or LPHN1-3 (Ten-2, 1:100, Lot # K1910, sc-165674, Santa Cruz Biotechnology, CA, USA; LPHN-1, 1:200, D-20, Lot # I2909, sc-34484; LPHN-2, 1:200, A-14, Lot # H0608, sc-47091; LPHN-3, 1:200, P-17, Lot # A0907, sc-47095, Santa Cruz Biotechnology, CA, USA) diluted in PBS-T and 2% normal donkey serum, for 48 h at 4 ◦C. Subsequently, sections were incubated in species-specific biotinylated secondary antibody, followed by Cy<sup>3</sup> -streptavidin (1:500, #016-160-084, Jackson Immunoresearch, PA, USA). The sections were then washed (3 × 10 min) in PBS and incubated in primary polyclonal antibody anti-GFAP (1:250, Lot # 2145934, AB5804, Millipore, MA, USA), overnight at 4◦C. After that, they were incubated using FITC-conjugated secondary antibody (1:200, Lot # I1213, sc-2090, Santa Cruz Biotechonology) and counterstained with DAPI (TR-100-FJ, Biosensis, SA, Australia). Finally, the histological sections were mounted onto gelatincoated slides, and coverslipped with buffered glycerol as a mounting medium.

#### Immunohistochemistry Control Reactions

Control reactions for Ten-2 immunohistochemistry were performed by primary and/or secondary antibody omissions. Additionally, an adsorption test was done using Ten-2 primary polyclonal antibody and the control peptide (Ten-2, Lot # E1011, sc-165674P, N-13, 100 µg/0.5 mL, Santa Cruz Biotechnology, CA, USA). For this, Ten-2 antibody (1:100) was incubated with different concentrations of control peptide (1:1; 1:0.1; 1:0.01; 1:0.005) during 24 h in PBS. After that, these solutions were used to incubate histological rat brain sections during 24 h at 4◦C, following the procedures adopted in the immunoperoxidase staining method described previously. Similar procedures were performed in a previous study (Torres-da-Silva et al., 2017). All immunolabeled cells noticed in the present study were considered to be "like-immunoreactive" neurons or reactive astrocytes.

# Microscopy and Data Analysis

Histological sections submitted to indirect immunoperoxidase staining method were qualitatively analyzed to identify Ten-2-LI cells using a light microscope (Axiolab A1, Carl Zeiss, BW, Germany) coupled to a digital camera (AxioCam MRc5, Carl Zeiss, BW, Germany). The selected areas were captured using imaging software (Zen2, Carl Zeiss, BW, Germany). When necessary, brightness, contrast and intensity were adjusted in the digital images, without changing the immunolabeling pattern, using Corel Draw software (Corel Corporation, ON, Canada).

Histological sections submitted to double indirect immunofluorescence were quantitatively (GFAP-LI/Ten-2- LI) or qualitatively (GFAP-LI/LPHN-LI) analyzed. For this, the sections were analyzed under a 40 × objective lens and images were captured by a confocal laser scanning microscope (TCS-SP5 AOBS Tandem Scanner, LEICA, HE, Germany) coupled to an inverted optical microscope (Leica DMI 6000CS) from Electron Microscopy Center of the Institute of Biosciences of Botucatu (IBB-UNESP, Botucatu, SP, Brazil). The confocal microscope is equipped with Diode, Helium-Neon and Argon lasers enabling the excitation wavelength lines of 405–633 nm. Fluorochromes were detected sequentially, and we carefully used fluorophores situated far apart in the fluorescence emission spectrum, to avoid a false positive colocalization result. The background marking was controlled in real time through the voltage photomultiplier and adjusted to obtain the best compromise between sensitivity and non-specificity. Planapochromat objectives of 20 ×, 40 × and 63 × (numerical aperture 1.30) with oil immersion were used, which allowed a resolution of up to ∼150 nm in axes x, y and ∼300 nm in the z axis (pinhole of 1 Airy unit). Nuclear astrocyte profiles (DAPI/GFAP-positive), as well as those exhibiting Ten-2-LI were manually quantified using ImageJ software (Rasband, W.S., ImageJ, U. S. National Institutes of Health, Bethesda, Maryland, USA, https://imagej. nih.gov/ij/, 1997-2018.). Five serial histological sections (360 µm equidistant), encompassing the brain lesion or control areas were analyzed in all experimental groups. Next, four microscopic fields (each measuring 15 × 10<sup>4</sup> µm<sup>2</sup> ) flanking the track lesion from each histological section were captured for counting. A blind examiner was previously calibrated for cell counting parameters. The morphology had normal distribution and the data were submitted to ordinary one-way ANOVA, followed by Dunnett's multiple comparison post-hoc tests, considering p < 0.05 as significant (GraphPad Prism 6, GraphPad Software, Inc., CA, USA).

For 3D-cell reconstructions, twelve reactive astrocytes (Ten-2-LI/GFAP) were analyzed in confocal laser scanning microscope (TCS-SP5 AOBS Tandem Scanner, LEICA, HE, Germany) under a 63 × objective lens and 4 × zoom. The cells were scanned 30– 40 times at intervals of 0.3µm and reconstructed using TCS-SP5 AOBS software (Tandem Scanner, LEICA, HE, Germany).

# RNA Extraction and Conventional RT-PCR

The RT-PCR method was used to check possible Ten-2/TCAP-2 gene expression changes during tissue injury of the cerebral cortex. For this, animals from the experimental groups were anesthetized as previously mentioned, positioned on a stereotaxic apparatus, and the brain surface was again exposed. The cerebral cortex around the lesion was collected under surgical stereomicroscopy (Model MC A-199, DF Vasconcellos, RJ, Brazil), transferred to DNase and RNase free ice-cold saline, and further trimmed to eliminate normal cortex around the lesion to the maximum. This fragment was transferred to appropriate centrifuge tubes with 1.0 mL trizol (#15596026, Life Technologies, CA, USA), immediately homogenized (#985370EUR-04, Tissue-Tearor, Biospec Products, OK, USA) at 30,000 rpm for 40 s and incubated on ice for 5 min. Next, 200 µL of chloroform (#0757, Biochemicals Life Science Research Products, OH, USA) was added, vortexed for 15 s and incubated on ice during 10 min. The tube was submitted to refrigerated (4◦C) centrifugation (Mikro 220R, Hettech Zentrifugen, BW, Germany) at 12,000 ×g for 15 min and the upper phase containing the total RNA was transferred to a new tube. After that, 0.5 mL of isopropyl alcohol (#I9030, Sigma-Aldrich, MO, USA) was added to the total RNA solution, incubated at RT for 10 min and centrifuged again at 12,000 ×g, at 4◦C for 10 min. The supernatant was discarded and 1 mL of 75 % alcohol (#E7023, Sigma-Aldrich, MO, USA) was added, followed by centrifugation at 7,500 ×g for 5 min at 4 ◦C. Subsequently, the supernatant was discarded and the total RNA pellet was dehydrated at RT for 5–10 min. Total RNA was resuspended in 100 µL of sterilized nuclease-free water (#W4502, Sigma-Aldrich, MO, USA) and heated in dry block (MD-01N, Major Science, CA, USA) for 15 min. To ensure

that the total RNA preparation was not contaminated by DNA, the sample was submitted to treatment with Turbo DNA-free Kit (#AM1907, lot 00353291, Life Technologies, CA, USA) following the manufacturer's instructions. The quality and quantity of total RNA of the samples were measured using a spectrophotometer (Optizen POP, Mecasys, South Korea) and submitted to electrophoresis of nucleotides on denaturing gel with 1 % agarose (#N605, Amresco, OH, USA) and 0.0005 % ethidium bromide (#X328, Sigma Aldrich, MO, USA). The remainder of total RNA was stored at −80◦C in ultralow freezer (Scientific 923, Forma Scientific, OH, USA).

Previously, all gene expressions were evaluated in order to establish the optimum cycle number in the exponential phase, permiting an accurate semi-quantitative analysis, using RNA samples from control animals. Conventional RT-PCR was used to amplify Ten-2, TCAP-2, GADPH (used for normalization) and neurofilament light (NFL, used for relative data expression) genes (**Table 1**, in vivo analysis). RT-PCR reactions were performed using a commercial kit (#210212, lot 154021850, Qiagen, CA, USA). The total mix (50 µL) of RT-PCR reaction was prepared, containing 10 µL OneStep RT-PCR buffer, 2 µL dNTP mix, 1 µL of each primer (10 nM) (**Table 1**), 2 µL OneStep RT-PCR enzyme mix, 1–4 µL of total RNA and 26–30 µL RNase-free water. The reaction tubes were placed in a thermal cycler (Mastercycler ProS Eppendorf, GmbH, Germany) for initial reverse transcription for 30 min at 50◦C, initial denaturing and PCR activation for 15 min at 95◦C and subsequently for 27 cycles (Ten-2 or TCAP-2), 27 cycles (GADPH) or 29 cycles (NFL) of denaturing for 1 min at 94◦C, annealing for 1 min at 53◦C (Ten-2 or TCAP-2) or 50◦C (GADPH; NFL) and elongation for 1 min at 72◦C, then for the final elongation cycle for 10 min at 72◦C. The DNA samples were stored at 4◦C until gel electrophoresis was performed and they were run through a 1.5% agarose gel.

#### Control Reactions

Control reactions were performed without RNA addition for RT-PCR or with RNA addition for PCR assays (#C1141, GoTaq Flexi DNA Polymerase, Promega, WI, USA), using at least 30–35 cycles in both assays.

# Data Analysis

RT-PCR bands were captured and digitalized under UV light using ImageQuant LAS 500 (GE Healthcare Bio-Sciences AB, Uppsala, Sweden). The optical density of the bands for Ten-2, TCAP-2, GADPH and NFL gene expressions were analyzed by densitometry using ImageQuant LAS 500 (GE Healthcare Bio-Sciences AB, Uppsala, Sweden) software. Absolute data were normalized with GADPH and expressed in relation to NFL gene expression. Ten-2 data were expressed in relation to NFL gene expression, as Ten-2 is present in cortical neurons in control animals and in neurons and astrocytes in animals with cerebral cortex injury. This analysis is more confident because we collected samples more restricted to the injury area, reducing the healthy tissue around it to the maximum, where NFL expression is more elevated due to presence of more neurons. Thus, we can infer that if there is some increase in the Ten-2 gene expression in samples from animals with cerebral cortex injury, this expression increase most likely comes from Ten-2 reactive astrocyte gene expression than from neurons, since NFL expression is low in these animals group.

The gene expression data had normal distribution and were submitted to ordinary one-way ANOVA, followed by Dunnett's multiple comparison post-hoc tests, considering p < 0.05 as significant (GraphPad Prism 6, GraphPad Software, Inc., CA, USA).

#### In vitro Study Cell Culture

We searched for an astrocyte cell lineage that expresses teneurin gene expression in order to adopt it in future in vitro assays. For this, C8D1A mouse cerebellar immortalized astrocytes (#CRL-2541, ATCC, VA, USA) were used. In addition, based on the fact that TCAP-1 is a bioactive peptide in neurons supported by in vitro and in vivo studies, we tested whether TCAP-1 is able to change calcium signaling in this cell lineage.

This astrocytes cell lineage was cultured in Dulbecco's Modified Eagle's Medium (DMEM) containing 4,500 mg/ L glucose content, 4 mM L-glutamine, 1 mM sodium pyruvate, 1,500 mg/ L sodium bicarbonate (#30-2002, ATCC, VA, USA) with 10% fetal bovine serum (FBS, #12483020, Thermo Fisher Scientific, MA, USA), 100 µg/ mL penicillin and 100 µg/ mL streptomycin (#15140-122, Thermo Fisher Scientific, VA, USA) added to the medium. Astrocytes were incubated at 37◦C, 5% CO<sup>2</sup> and cells were maintained at 70–80% confluency.

# RNA Extraction, Reverse-Transcription, and Polymerase Chain Reaction (PCR)

Once at 70–80% confluency in a 6-well plate, C8D1A astrocytes were serum-starved for 3 h. Then, 1 mL trizol reagent was added to each well to extract RNA from the cells after which they were incubated for 2 min at RT. Lysates were transferred and 200 µL of chloroform (#C3300, ACP Chemicals, QC, Canada) was added. The solution was thoroughly mixed and incubated at RT for 3 min. Subsequently, the samples were centrifuged at 12,000 ×g for 15 min at 4◦C. The RNA-containing supernatant was transferred to a new tube and 500 µL of isopropanol (#A426, Thermo Fisher Scientific, MA, USA) was added. The solution was incubated at RT for 10 min and then centrifuged at 14,000 ×g for 10 min and 4◦C. After centrifugation, the supernatant was discarded and the pellet was washed using 75% ethanol (EtOH) (#PO16EA95, Commercial Alcohols, Canada) and centrifuged at 7,400 ×g for 5 min. After two rounds of EtOH washes and centrifugation, the EtOH was removed and the pellet was re-suspended in 20 µL diethyl pyrocarbonate (DEPC)-water (#W4502, Sigma-Aldrich, MO, USA). RNA sample absorbance was determined using the NanoDrop 2000 spectrophotometer at 260 nm and 280 nm wavelengths (Thermo fisher scientific, MA, USA, v.1.4.2).

The RNA extracted from C8D1A immortalized astrocytes was reverse-transcribed to complementary DNA (cDNA). RNA-free H2O, random primers (#SO142, Thermo Fisher Scientific, MA, USA) and dNTPs (#RO192, Thermo Fisher Scientific, MA, USA) were added to the sample RNA and the sample was heated to 65◦C for 5 min. The sample was left on ice for 1 min, after which

#### TABLE 1 | Primer sets used in RT-PCR and PCR assays.


5X first-strand buffer (#YO2321, Invitrogen, CA, USA) and 0.1 M DTT (#YO0147, Invitrogen, CA, USA) were added, mixed and left at RT for 2 min. Subsequently, Superscript II RT (#100004925, Invitrogen, CA, USA) was added, after which the sample was left at RT for 10 min, heated to 42◦C for 50 min and then heated to 70◦C for 15 min.

To perform PCR, mastermix containing ddH2O, Taq buffer (#B33, Thermo Fisher Scientific, MA, USA), MgCl<sup>2</sup> (#B34, Thermo Fisher Scientific, MA, USA) and dNTPs was prepared and appropriate primer pairs were added (**Table 1**, in vitro analysis) along with the cDNA template and Taq polymerase (#EP0402, Thermo Fisher Scientific, MA, USA). The reaction tubes were then placed in the thermal cycler (#6331000025, Eppendorf, Germany) for the first cycle of denaturing for 7 min at 95◦C, subsequently for 35 repeated cycles of denaturing for 1 min at 95◦C, annealing for 1 min 30 s at 60–65◦C and elongation for 35 s at 72◦C and then for one final cycle of elongation for 5 min at 72◦C. The DNA samples were stored at 4◦C until gel electrophoresis was performed and they were run through a 3% agarose gel (#9012-36-6, BioShop, Canada) and imaged using Image-Lab software (Bio-Rad, CA, USA, v.4.1).

#### Intracellular Calcium Fluorescence Microscopy

Cells were grown on coverslips and once at 60% confluency in a 6-well plate, they were incubated for 30 min at 37◦C in 3.6µM Fluo-4 acetomethyl (AM) ester (1 µg/µL dissolved in DMSO; #F14201, Thermo Fisher Scientific, MA, USA) in culture medium. Subsequently, the cells were washed with artificial cerebrospinal fluid (aCSF) (135 mM NaCl, 4.5 mM KCl, 2 mM MgCl26H2O mM, 10 mM HEPES, 10 mM glucose, 2 mM CaCl22H2O) and the coverslip remained in aCSF until it was mounted to the stage of an inverted fluorescence microscope (Zeiss Axio Observer.Z1, Carl Zeiss Microimaging, BW, Germany) for perfusion. The microscope was equipped with a 40 × oil immersion lens (1.4 NA, Carl Zeiss Microimaging, BW, Germany) and a digital CCD camera (C4742-80, Hamamatsu Photonics, Hamamatsu, Japan) to capture fluorescent images. Velocity cellular imaging software (Improvision, version 4.3.2) was used to obtain fluorescent images. Solutions were administered using a perfusion system and teneurin C-terminal associated peptide (TCAP-1) was added to aCSF for treatment at a 100 nM concentration.

# Data Analysis

Data from calcium assays were normalized to the fluorescence ratio baseline and presented as standard error of the mean (SEM). Data exhibited normal distribution and were submitted to comparisons of multiple conditions using two-way ANOVA with a Bonferroni post hoc test, considering p < 0.05 as significant (v.5 GraphPad Prism 6, GraphPad Software, Inc., CA, USA).

# RESULTS

# In vivo Study

#### Immunoperoxidase Staining

The immunoperoxidase staining method was important to detail the morphological characteristics of neural cells that exhibited immunoreactivity to Ten-2. Ten-2-like immunoreactive (Ten-2-LI) cells in the cerebral cortex from control animals were represented mainly by pyramidal neurons situated in layer V. However, this immunolabeling was discrete and exhibited two patterns, one associated with the cell membrane delineating the cell body (perikarya) and the other one homogenously distributed in the cell cytoplasm of the cell body and apical dendrite (**Figures 1A-C**). In animals with mechanical brain injury, Ten-2-LI cells significantly changed in the cerebral cortex with strong immunolabeling in reactive astrocytes in all postoperative groups (**Figures 1D–H**). Occasionally, rare Ten-2- LI neurons scattered among reactive astrocytes were observed. The normal cortical areas around the brain lesion maintained the same immunolabeling pattern in neurons.

Frequently, reactive astrocytes were nearly absent in the area closest to the needle track or haemorragic area. Close to that area, only Ten-2-LI palisading reactive astrocytes sending cell processes to haemorragic area were observed, mainly in later postoperative periods (**Figures 1E,F**). Adjacent to that area, Ten-2-LI reactive astrocytes showed significant hypertrophy, decreasing toward deeper cerebral cortex layers and white matter (**Figures 1E,G,H**, **2**). At times, Ten-2-LI reactive astrocytes showed clear cell extensions projecting to blood vessels (**Figure 1H**).

Control reactions for Ten-2 immunohistochemistry by pre-adsorption Ten-2 epitope resulted in complete absence of immunolabeling in neurons and reactive astrocytes in all peptide concentrations (**Figure S1**). In addition, no immunolabeling was noticed after primary and/or secondary antibody omissions (**Figure S1**).

For comparative analysis, additional data concerning indirect immunoperoxidase staining for detection of Ten-1, Ten-3, or Ten-4 in the cerebral cortex of animals with mechanical brain injury of the cerebral cortex are also presented (**Figure S2**). There is no immunolabeling for these proteins in reactive astrocytes. Only Ten-1-LI cortical neurons were evident, but it was quite similar in relation to the control groups (**Figure S2**).

#### Double Immunofluorescence

Since the main cells that exhibited immunoreactivity to Ten-2 were reactive astrocytes during tissue injury, the double immunofluorescence method was used in order to count nuclear astrocyte profiles (nuclear DAPI staining/GFAP-LI astrocytes) exhibiting immunoreactivity to Ten-2 in the experimental groups. GFAP-LI astrocytes showed homogenous distribution in the cerebral cortex of the control group. Immunoreactivity was present in the cytosol of the cell body and in its thin cell extensions (**Figures 1I-I"**; **Figure S3**; **Video S1**). Astrocytes rarely exhibited detectable immunolabeling to Ten-2 in control group animals (**Figures 2A–A"**). On the other hand, animals with brain injury from all postoperative periods showed a clear and strong Ten-2 immunolabeling in reactive astrocyte cell profiles (**Figures 2B–E"**). These cells exhibited large cell bodies, besides arborized and elongated cell extensions (**Figures 2B-E"**; **Figure S3**; **Video S1**). Interestingly, the immunolabeling pattern to Ten-2 was punctiform and distributed in the cytosol of reactive astrocytes (**Figures 1I-I"**; **Figure S3**; **Video S1**).

Quantitative analysis demonstrated that the number of nuclear astrocyte cell profiles DAPI/GFAP-LI) did not show statistical difference among all experimental groups (**Figure 2F**). However, the number of Ten-2-LI nuclear astrocyte profiles (DAPI/GFAP-LI/Ten-2-LI) was significantly increased in animals submitted to mechanical brain injury in all postoperative periods (p < 0.0001), compared to the control group (**Figure 2F**).

In order to investigate with more detail whether the immunoreactivity to Ten-2 was present in the cell membrane and/or in inner parts of the reactive astrocytes, some histological sections with Ten-2-LI reactive astrocytes (DAPI/GFAP-LI/Ten-2-LI) were used for 3D reconstruction in confocal microscope (**Figure S3**; **Video S1**). The cells clearly showed that the immunolabeling was present in the cytoplasm with granular arrangement or sparce punctiform labeling linked to the cell membrane (**Figure S3**; **Video S1**).

Based on the fact that latrophilins are involved in heterophilic interactions with teneurins, we submitted histological sections for double immunofluorescence to qualitatively evaluate whether immunoreactivity to LPHN is evident in reactive astrocytes (GFAP-LI). Animals with mechanical brain injury exhibited strong immunoreactivity to LPHN-1 in reactive astrocytes, moderate to LPHN-3 and discreet to LPHN-2 (**Figure 3**). Sections from control animals did not show any LPHN-LI astrocytes. Control reactions with primary antibody omission for latrophilins resulted in absence of immunolabeling in reactive astrocytes (**Figure 3**).

# Conventional RT-PCR

RT-PCR analysis was used in order to confirm Ten-2 and TCAP-2 expression as well as to check possible changes during mechanical cerebral cortex injury. RT-PCR analysis showed both Ten-2 and TCAP-2 mRNA expression in all experimental groups (**Figure 4**). Mechanical brain-injured groups (48 h, 3 and 5 days) revealed a significant increase in Ten-2 and TCAP-2 mRNA expressions, compared with control (p < 0.0001) and 24 h (p < 0.0001) groups (**Figure 4**). Control reactions in PCR using RNA samples and Ten-2, TCAP-2 or NFL primers did not show any bands (**Figure S4**). Similarly, RT-PCR without addition of RNA samples showed no bands.

section of adult rat cerebral cortex. (A–C), control group, observe neurons in cerebral cortex layer V exhibiting discrete immunoreactivity to Ten-2, homogenously distributed in cell cytosol (large arrow) or associated with cell membrane (small arrows). In (D–H), mechanical brain lesion in cerebral cortex from 48 h postoperative period group, showing strong immunoreactivity to Ten-2 in reactive astrocytes. In (E,F), note Ten-2-LI reactive astrocytes with palisading aspect sending some cell extensions (arrows) to haemorraghic area (F, black asterisk). Ten-2-LI hipertrophic reactive astrocytes are densely grouped immediately below (G) and decreasing toward deeper layers of the cerebral cortex (H). Reactive astrocytes exhibited large cell bodies with numerous long arborized cell extensions (G,H), sometimes encircling the blood vessel (H, arrow). In (I–I") immunolabeling pattern of Ten-2 in reactive astrocytes (GFAP-LI) analyzed by confocal microscopy. Observe that immunolabeling to Ten-2 is distributed in the cytosol of the cell body and cell extensions (arrows), exhibiting a punctiform pattern and sometimes associated with plasmatic membrane. GFAP-LI, glial fibrillary acidic protein-like immunoreactive; Ten-2-LI, teneurin-2-like immunoreactive; V, cerebral cortex layer V.

#### In vitro Study Conventional RT-PCR

We searched for teneurin and TCAP gene expression in the C8D1A mouse cerebellar immortalized astrocytes, in order to evaluate whether these proteins are expressed in cell lineage astrocytes which can be used in future assays. RT-PCR analysis showed immortalized cerebellar astrocytes expressing Ten-1, Ten-3, and Ten-4, but not Ten-2 (**Figure 5A**). Gene expression of TCAP-1-4 was also present in the immortalized astrocytes (**Figure 5A**).

# Intracellular Calcium Fluorescence Microscopy

Based on the fact that TCAP-1 is a bioactive peptide in neurons supported by in vitro and in vivo studies (Al Chawaf et al., 2007a,b; Trubiani et al., 2007; Tan et al.,

FIGURE 2 | 24 h (B–B") and nearly all of them exhibit immunoreactivity to Ten-2 in all experimental groups. (F) quantitative analysis of absolute values (mean ± SEM) of astrocyte cell profiles exhibiting immunoreactivity to Ten-2 (GFAP-LI/Ten-2-LI) in all experimental groups (n = 5, five animals per group). Note that reactive astrocyte cell profiles (GFAP-LI) showed no statistical difference among all groups, indicating that there is no cell proliferation. However, note that reactive astrocyte cell profiles (GFAP-LI/Ten-2-LI) significantly increased (p < 0.0001) in all experimental groups with mechanical brain injury compared to control group. Mean (± SEM) values from each group were submitted to one-way ANOVA and Dunnett's multiple comparisons post hoc test, considering p < 0.05 as significant. \* Statistically significant difference when compared with control group (p < 0.0001). GFAP-LI, glial fibrillary acidic protein-like immunoreactive; Ten-2-LI, teneurin-2-like immunoreactive.

2009, 2011; Chand et al., 2012), we tested whether TCAP-1 is able to change calcium signaling in immortalized astrocytes. Thus, immortalized cerebellar astrocytes were treated with synthetic TCAP-1. These cells exhibited substantial increases in intracellular Ca2+, as quantified by the emitted fluorescence of Fluo-4-AM ester binding to Ca2+, after TCAP-1 treatment (**Figures 5B–D**). The level of intracellular Fluo-4-AM fluorescence increased significantly at 3 (p < 0.001) and 6 (p < 0.01) minutes after TCAP-1 treatment (100 nM), in relation to vehicle treatment (**Figure 5E**).

# DISCUSSION

The present study is the first to show that Ten-2 and TCAP-2 are upregulated in reactive astrocytes after mechanical brain injury induced by needle-insertion lesion in the cerebral cortex of adult rats. LPHN-1, the main endogenous receptor for a Ten-2 splice variant named Lasso, was also evidenced in reactive astrocytes by the immunohistochemistry method. In vitro analysis showed that immortalized cerebellar astrocytes also express additional Ten/TCAP homologs and increase calcium uptake after TCAP-1 treatment. These findings disclose a new functional role of the Ten/TCAP system in glia during CNS tissue repair.

Several studies have reported that reactive astrocytes secrete several anti-inflammatory substances, as well as act as a barrier for harmful substances that diffuse from blood vessels (Sofroniew, 2015). Considering gene expression analysis in brain injuries, a previous study showed that reactive astrocytes collected from different regions of the mouse brain with middle cerebral artery occlusion (MCAO) had significant gene upregulations, where Ten-2 gene was one of the 50 most upregulated genes (Zamanian et al., 2012). Furthermore, the Ten-2 gene was up-regulated in the human frontal cortex from patients with amyotrophic lateral sclerosis (Andrés-Benito et al., 2017), as well as in patients with chronic traumatic encephalopathy (Seo et al., 2017). There is a rare astrocyte disease caused by dominant gain-of-function mutations in the GFAP gene named Alexander disease (Messing et al., 2012). Astrocytes differentiated from induced pluripotent stem cell (iPSCs) samples collected from patients developing Alexander disease revealed that ODZ2 (Ten-2) is one of the most up-regulated genes in comparison with astrocytes differentiated from healthy subjects (Kondo et al., 2016). These experimental and clinical data support a possible participation of Ten-2 in brain disorders, mainly in astrocytes under certain CNS diseases. Our in vivo immunohistochemistry data strongly suggest that reactive astrocytes are synthetizing Ten-2 and its up-regulation was confirmed by significant increase of Ten-2 mRNA in the cerebral cortex of animals with mechanical brain injury. Supporting this statement, we did not notice an increase of immunoreactivity to Ten-2 in neurons that produce Ten-2, or in microglia, or oligodendrocyte in the cerebral cortex with mechanical brain injury. Moreover, it is important to mention that reactive astrocytes are not phagocytes like microglia, which could endocyte possible Ten-2 released from neurons. Based on these considerations, it can be concluded that immunoreactivity to Ten-2 evidented in reactive astrocytes is from the Ten-2 gene expression upregulation in these cells.

It is worth mentioning that we adopted some controls for our immunohistochemistry analysis. Omission of the primary antibodies for Ten-2 and LPHN and/or secondary antibody resulted in absence of immunolabeling. Pre-adsorption test for Ten-2 eliminated any immunolabeling of neurons and reactive astrocytes suggesting that polyclonal antibody used in the present study was specific to Ten-2 epitope. However, we did not perform immunohistochemistry assay controls using our primary antibody lots or aliquotes in histological brain sections or in cell lineages generated by knockout procedures for teneurins, latrophilins or GFAP. It is known that one of the best controls to confirm antibody specificity is through immunohistcohemical assays using samples from knockout models (Saper and Sawchenko, 2003; Burry, 2011). Based on these considerations and limitations of the present study, we have adopted the term "like-immunoreactive" for all immunolabeling detected in the neurons and/or reactive astrocytes.

Brain injury has been used as a model to study CNS repair. The needle-insertion lesion model is a focal and controlled injury of the rat cerebral cortex, which exhibits two different lesion areas (Purushothuman et al., 2013; Purushothuman and Stone, 2015). One area is along the track of the needle insertion, where there is mechanical tissue disruption with neuronal and synapse degeneration, associated with haemorrhagic and several extracellular modifications (Purushothuman et al., 2013). The other area is tissue flanking the track, where there is nonsignificant neuronal degeneration and intracellular transient effects, up-regulating different mechanisms for self-protection to minimize the effects of injury (Purushothuman et al., 2013; Purushothuman and Stone, 2015), similar to the tissue surrounding intracerebral haemorrhagic area. In the present study, only cell extensions from Ten-2-LI reactive astrocytes were noticed close to the haemorraghic area induced by needle insertion; while in the flanking area, nearly all GFAP astrocyte profiles exhibited immunoreactivity to Ten-2. Thus, Ten-2/TCAP-2 reactive astrocytes are mainly located in an area that expresses neuroprotective molecules. Interestingly, TCAP-1 acts as a neuroprotective molecule in immortalized hypothalamic neurons, as it induced superoxide dismutase, superoxide dismutase-1 copper chaperone and catalase enzymes (Trubiani et al., 2007).

CNS injuries or diseases implicate several neural and non-neural cell types to protect the brain and try to recover its function (Burda and Sofroniew, 2014). Astrocytes are a pivotal cell type involved in brain damage, including stroke, tumors, infections, neurodegeneration, traumatic brain injury, chemical toxicity and epilepsy (Sofroniew and Vinters, 2010; Kang and Hébert, 2011; Liu and Chopp, 2016). After the occurrence of brain injury, the astrocytes become active, undergo hypertrophy and hyperplasia processes (astrogliosis), and consequently up regulate GFAP, vimentin and other mediators (Eng and Ghirnikar, 1994; Sofroniew and Vinters, 2010; Burda and Sofroniew, 2014; Liu et al., 2014; Sofroniew, 2015; Burda et al., 2016; Liu and Chopp, 2016). In the present study, it was observed that reactive astrocytes exhibited significant immunoreactivity to Ten-2 up to 5 days after injury. After that, the expression decreased significantly up to 30 days postoperatively (data not shown). Thus, the presence of Ten-2/TCAP-2 in astrocytes seems significant only on the first days after cortical injury, indicating its involvement during the cascate of events to minimize the inflammatory mechanism, as evidented in astrocytes of subtype A2 (Liddelow and Barres, 2017).

The in vitro analysis using immortalized mouse cerebellar astrocytes was useful; as it confirmed that astrocytes also express Ten/TCAP homolog mRNAs, supporting their application in further studies to detail the possible functions of these proteins and their signaling mechanism. In contrast with our in vivo data, immortalized cerebellar astrocytes expressed additional Ten/TCAP homologs. However, this differential gene expression can be due to the in vitro environment, which does not simulate all in vivo conditions. Moreover, this cell lineage is derived

FIGURE 5 | In vitro characterization of Ten/TCAP gene expressions and TCAP-1 induced activation in mouse cerebellar astrocytes. Gene expression was determined using RT-PCR with RNA extracted from C8D1A mouse cerebellar astrocytes. (A) teneurins 1, 3 and 4 were expressed (n = 4) and teneurin 2 was not expressed (n = 4). TCAPs 1-4 were expressed (TCAP-1, n = 3; TCAP-2, n = 5; TCAP-3, n = 3; TCAP-4, n = 3). β-actin served as positive control (n = 5). (B) fluo-4 fluorescence shown prior to 100 nM TCAP-1 administration (non-stimulated). (C) fluo-4 fluorescence shown 3 min after 100 nM TCAP-1 administration (TCAP-1 treatment). (D) differential interference contrast image (DIC) C8D1A astrocytes. Arrows are showing discreet band. (E) normalized fluo-4 fluorescence comparing vehicle (aCSF) and 100 nM TCAP-1 treated cells (n = 3 for each treatment, where each n is an average of five cells per coverslip). Mean (± SEM) values from each group were submitted to two-way ANOVA and Bonferroni post hoc test. \*\*p < 0.01; \*\*\*p < 0.001.

from a newborn mouse cerebellum (postnatal day 8), which has particular astrocytes, such as Bergmann glia and velate astrocytes. It is interesting to mention that there are a few studies analyzing Ten-2 distribution in the cerebellum during development or in adult rodent brain and they did not mention Ten-2 presence in astrocytes (Otaki and Firestein, 1999; Zhou et al., 2003). One study only pointed out noticeable Ten-2 presence in neurons situated in molecular and Purkinge layer cells and less pronounced presence in the granular layer in the mouse cerebellum (Zhou et al., 2003); while Ten-2 presence was not noticed in the cerebellum in adult Sprague-Dawley rats (Otaki and Firestein, 1999).

Teneurins are a family of proteins mainly expressed in the CNS during development and recent studies have shown that they stablish homophilic or heterophilic interactions (Silva et al., 2011; Mosca, 2015; Woelfle et al., 2015). Heterophilic interactions between teneurins with integrins, dystroglycans and latrophilins have been suggested by several studies (Trzebiatowska et al., 2008; Silva et al., 2011; Topf and Chiquet-Ehrismann, 2011; Boucard et al., 2014; Mosca, 2015; Woelfle et al., 2015; Vysokov et al., 2016; Li et al., 2018). Integrins and dystroglycans are present in astrocytes and play a role of maintaining bloodbrain barrier (BBB) homeostasis (Guadagno and Moukhles, 2004; del Zoppo and Milner, 2006; Wolburg-Buchholz et al., 2009; Sofroniew and Vinters, 2010). A possible interaction between integrins and/or dystroglycans with Ten-2/TCAP-2 in some reactive astrocytes related to BBB components can be considered, since our immunohistochemistry analysis revealed several Ten-2-LI reactive astrocytes projecting cell extensions to the vicinity of the blood vessels. Supporting this assumption, there are some studies indicating teneurin interaction with integrins and distroglycan in neurons as well as in other tissues like connective tissue and testis in other tissues (Löer et al., 2008; Trzebiatowska et al., 2008; Topf and Chiquet-Ehrismann, 2011; Chand et al., 2012, 2014).

LPHN-1 is the main endogenous receptor for a Ten-2 splice variant named Lasso. Remarkably, substantial immunoreactivity to LPHN-1, moderate to LPHN-3 and modest to LPHN-2 was observed in reactive astrocytes in the present study. Possibly, reactive astrocytes display Ten-2 and its receptor LPHN-1 during brain injury. This information indicates that Ten-2 and/or its related proteins can be released by reactive astrocytes, inducing self-stimulation or self-inhibition by coupling in the LPHN-1 autoreceptor and modulating the intracellular signaling mechanism. In addition, the possibility that reactive astrocytes simultaneously exhibit Ten-2 and LPHN-1 can also indicate that these cells stablish an intercellular interaction by heterophilic coupling involving these proteins. A previous study showed that LPHN-1 and Ten-2 are sited in distinct parts of the synapse, where the former is in the presynaptic and the latter in the postsynaptic membrane of neurons (Boucard et al., 2014). Lasso interacts with LPHN-1 in neurons, inducing calcium signaling in these cells (Silva et al., 2011; Boucard et al., 2014; Mosca, 2015; Woelfle et al., 2015; Vysokov et al., 2016). It is known that calcium signaling in astrocytes results in gliotransmitter release, modulating neurons through tripartite synapses (Araque et al., 2014; Gundersen et al., 2015; Mitterauer, 2015; Covelo and Araque, 2016). Tripartite synapses are formed by pre- and post-synaptic neuronal membranes, besides cell extensions from nearby astrocytes, permitting cross-interactions and modulation between astrocytes and neurons (Araque et al., 2014; Gundersen et al., 2015; Covelo and Araque, 2016). In addition, astrocytes establish ionic coupling through gap junctions, particularly involving calcium, which allows for the synchronization of their activity; thus, enabling several astrocytes to act as a syncytium (Araque et al., 2014; Gundersen et al., 2015; Covelo and Araque, 2016). This characteristic enables a small number of activated astrocytes to substantially expand their ability to modulate neuronal activity in larger areas of the CNS (Covelo and Araque, 2016). Therefore, a possible interaction among Ten-2-related proteins and LPHN in reactive astrocytes could induce calcium uptake and modulate gliotransmitter release by these cells or eventually modulate their vascular functions. In line with this assumption, our in vitro assays using immortalized astrocytes demonstrated that TCAP-1 treatment was able to induce significant calcium signaling in these cells. Such data support the idea that the carboxy terminus of Ten-2 plays a fundamental role in calcium signaling in neurons (Silva et al., 2011; Boucard et al., 2014; Mosca, 2015; Woelfle et al., 2015; Vysokov et al., 2016) and TCAP is possibly part of this mechanism in neurons and astrocytes.

Teneurins may be cleaved at the extracellular domain, specifically in the C-terminal region, resulting in TCAP, a bioactive peptide (Qian et al., 2004; Wang et al., 2005; Lovejoy et al., 2006). TCAP-1 can also be separately encoded by the last exon of Ten-1 (Chand et al., 2013). TCAP acts in stress modulation, neuroprotection, CRF-induced cocaine addiction reinstatement, among other functions (Qian et al., 2004; Wang et al., 2005; Lovejoy et al., 2006; Al Chawaf et al., 2007a,b; Trubiani et al., 2007; Kupferschmidt et al., 2011; Tan et al., 2011; Chen et al., 2013; Erb et al., 2014; Colacci et al., 2015). In the present study, it was observed that TCAP-2 expression was increased in the injured cerebral cortex; however, there is no available antibody to analyze TCAP-2 immunoreactivity. Thus, it can only be hypothesized that TCAP-2 is present in reactive astrocytes, corroborated by TCAP-2 gene expression data from in vitro analysis.

Finally, reactive astrocytes showed immunolabeling to Ten-2 present in the cytoplasm with granular arrangement or linked to the cell membrane with disperse punctiform distribution. These data can suggest that Ten-2/TCAP-2 proteins in reactive astrocytes may not work only as cell interaction molecules, since the immunolabeling did not show a typical plasmatic membrane presence in these cells. The possibility that Ten-2/TCAP-2 proteins are released by a secretory regulated pathway, generating soluble proteins exerting other functional roles has been suggested in previous studies (Chand et al., 2013; Vysokov et al., 2016). Moreover, the Ten-2 punctiform immunolabeling visible in the cell membrane from reactive astrocytes may characterize a specific membrane domain enriched with Ten-2 transmembrane proteins or some cell membrane regions with docked secretory granules filled with Ten-2/TCAP-2 proteins. Further studies analyzing Ten-2-LI reactive astrocytes by immunohistochemistry combined with electron microscopy could help elucidate this question. Unfortunately, there are no commercially available antibodies raised for different parts of the Ten-2 that could clarify whether Ten-2 upregulation has the role to produce Ten-2 transmembrane protein and/or secretory related proteins for autocrine or paracrine actions.

# CONCLUSIONS

Based on the limitations of the present study, a significant increase in Ten-2-LI reactive astrocytes was demonstrated for the first time after mechanical injury of the adult rat cerebral cortex. Reactive astrocytes also exhibited immunoreactivity to LPHN-1, the main endogenous receptor of Ten-2 splice variant named Lasso. Ten-2/TCAP-2 gene expression was also up-regulated in the cerebral cortex with mechanical brain injury. In vitro analysis using immortalized cerebellar astrocytes confirmed that these neural cells are potentially able to express additional Ten/TCAP homologs and that TCAP-1 treatment significantly increased calcium signaling in this cell line. Further studies are necessary to evaluate the role of Ten-2/TCAP-2 in reactive astrocytes, as well as to investigate the potential maneuver of these proteins as adjuvant therapies in CNS injury repair.

#### ETHICS STATEMENT

The experimental protocols for animal handling and care were approved by the Institutional Committee of Animal Welfare (CEUA, process number 2015-00318) from School of Dentistry of Araçatuba (UNESP, Araçatuba, SP, Brazil).

# AUTHOR CONTRIBUTIONS

GT: in vivo study: surgery procedures, immunohistochemistry and analysis, RT-PCR and analysis, manuscript drafting. OM: in vitro study: culture cells, RT-PCR and analysis, calcium signaling and analysis, manuscript drafting. KT: in vivo study: animal handling, immunohistochemistry and analysis, RT-PCR, manuscript drafting. AD: in vivo study: immunohistochemistry and analysis, manuscript drafting. RC-R: in vivo study: critical review analysis, manuscript drafting. AG: in vivo study: RTPCR and analysis, manuscript drafting. DG: in vivo study: RTPCR, manuscript drafting. JH-J: research project delineation, immunohistochemistry analysis, manuscript final review. EE: in vivo study: critical review analysis, manuscript drafiting. JB: research project delineation; manuscript final review. DL: in vitro study, research project delineation, manuscript final review. CC: research project delineation, study coordinator, manuscript drafting, manuscript final review.

#### ACKNOWLEDGMENTS

We thank FAPESP (Sao Paulo State Research Foundation, SP, Brazil) for research grants (Project EMU 2009/54141-9 to the Electron Microscopy Center of the Institute of Biosciences of Botucatu IBB—UNESP; 2013/26779-4 to EE, 2012/03067- 6 to CC, 2010/52068-0 to JB and 2008/02771-6 to JH-J) and fellowship grants (KT, FAPESP 2012/08833-9) and the Canadian Natural Sciences and Engineering Research Council for a Discovery Grant (DL) and post-graduate scholarship (OM). Part of this study was developed during the Master Thesis Project (GT) supported by a fellowship grant from Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (Finance Code 001, CAPES, Brazil). We would like to thank Mr. Andre Luis Mattos Piedade (Basic Sciences Department, School of Dentistry

#### REFERENCES


of Araçatuba, UNESP) and Mr. Robson Varlei Ranieri (Pathology and Propaedeutic Department, School of Dentistry of Araçatuba, UNESP) for technical support. We thank Ms. Shelly Favorito de Carvalho and Dr. Elton Luiz Scudeler for helping with image acquisition in laser confocal microscopy at the Electron Microscopy Center of the Institute of Biosciences of Botucatu (IBB-UNESP). We would like to thank Mrs. Karina Vieira Casatti for the time and effort taken to review the manuscript. JB is an investigator with the Conselho Nacional de Desenvolvimento Cientifico e Tecnológico (CNPq, National Council for Scientific and Technological Development).

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fnins. 2019.00655/full#supplementary-material

Figure S1 | Immunoperoxidase staining control in histological sections of adult rat cerebral cortex with mechanical brain injury from 48 h postoperative period group. (A–B'), Pre-adsorption with Ten-2 primary polyclonal antibody and Ten-2 epitope (A–A', 1:1; B–B', 1:0.01). Observe absence of immunoreaction in neurons and reactive astrocytes in the area with mechanical injury of the cerebral cortex. (C–C') Ten-2 primary polyclonal antibody omission, resulting in absence of immunoreaction in neurons and reactive astrocytes. Ab1 (Ten-2), Ten-2 primary antibody omission; ADS, adsorption test; <sup>∗</sup> , cortical lesion area.

Figure S2 | Immunoperoxidase staining to identify Ten-1, Ten-3 or Ten-4 in histological sections of adult rat cerebral cortex with mechanical brain injury from 48 h postoperative period group (A-C). Observe only neurons (arrow) exhibiting immunoreactivity to Ten-1 (A–A"). Ten-3-LI (B–B") or Ten-4-LI (C–C") cells are not observed in the cerebral cortex. Ten-1-LI, Ten-1-like immunoreactive; Ten-3-LI, Ten-3-like immunoreactive; Ten-4-LI, Ten-4-like immunoreactive; <sup>∗</sup> , cortical lesion area.

Figure S3 | Reactive astrocyte from adult rat cerebral cortex with mechanical brain injury (48 h after lesion) analyzed by confocal microscopy showing nucleus DAPI staining (A, blue fluorescence), immunoreactivity to GFAP (B, DTAF—green fluorescence), immunoreactivity to Ten-2 (C, Cy<sup>3</sup> immunostaining—red fluorescence) and simultaneous labeling (D, merge). Ten-2-LI reactive astrocyte exhibited a punctiform pattern, mainly distributed in the cytosol and occasionaly associated with plasmatic membrane.

Figure S4 | Control reactions using conventional PCR, total RNA extracted from cerebral cortex of all experimental groups (n = 4, four animals per experimental group) and Ten-2, TCAP-2 and neurofilament light (NFL) primers, visualized in 1.5% agarose gel stained with bromide ethidium. No bands were observed to Ten-2 (A), TCAP-2 (B) and NFL (C) indicating no significant DNA contamination of the RNA samples.

Video S1 | 3D reconstruction movie in confocal microscopy of a reactive astrocyte from adult rat cerebral cortex with mechanical brain injury (48 h after lesion) shown in Figure S3.

C-terminal associated peptide-1 (TCAP-1). Peptides 28, 1406–1415. doi: 10.1016/j.peptides.2007.05.014


calcitonin families of peptides. Gen. Comp. Endocrinol. 148, 299–305. doi: 10.1016/j.ygcen.2006.01.012


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Tessarin, Michalec, Torres-da-Silva, Da Silva, Cruz-Rizzolo, Gonçalves, Gasparini, Horta-Júnior, Ervolino, Bittencourt, Lovejoy and Casatti. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Teneurins: An Integrative Molecular, Functional, and Biomedical Overview of Their Role in Cancer

#### Boris Rebolledo-Jaramillo and Annemarie Ziegler\*

Center for Genetics and Genomics, Facultad de Medicina, Clínica Alemana Universidad del Desarrollo, Santiago, Chile

#### Edited by:

Richard P. Tucker, University of California, Davis, United States

#### Reviewed by:

Jong-Ik Hwang, Korea University, South Korea Wei-jiang Zhao, Shantou University Medical College, China

> \*Correspondence: Annemarie Ziegler aziegler@udd.cl

#### Specialty section:

This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Neuroscience

Received: 28 September 2018 Accepted: 28 November 2018 Published: 11 December 2018

#### Citation:

Rebolledo-Jaramillo B and Ziegler A (2018) Teneurins: An Integrative Molecular, Functional, and Biomedical Overview of Their Role in Cancer. Front. Neurosci. 12:937. doi: 10.3389/fnins.2018.00937 Teneurins are large transmembrane proteins originally identified in Drosophila. Their essential role in development of the central nervous system is conserved throughout species, and evidence supports their involvement in organogenesis of additional tissues. Homophilic and heterophilic interactions between Teneurin paralogues mediate cellular adhesion in crucial processes such as neuronal pathfinding and synaptic organization. At the molecular level, Teneurins are proteolytically processed into distinct subdomains that have been implicated in extracellular and intracellular signaling, and in transcriptional regulation. Phylogenetic studies have shown a high degree of intra- and interspecies conservation of Teneurin genes. Accordingly, the occurrence of genetic variants has been associated with functional and phenotypic alterations in experimental systems, and with some inherited or sporadic conditions. Recently, tumor-related variations in Teneurin gene expression have been associated with patient survival in different cancers. Although these findings were incidental and molecular mechanisms were not addressed, they suggested a potential utility of Teneurin transcript levels as biomarkers for disease prognosis. Mutations and chromosomal alterations affecting Teneurin genes have been found occasionally in tumors, but literature remains scarce. The analysis of open-access molecular and clinical datasets derived from large oncologic cohorts provides an invaluable resource for the identification of additional somatic mutations. However, Teneurin variants have not been classified in terms of pathogenic risk and their phenotypic impact remains unknown. On this basis, is it plausible to hypothesize that Teneurins play a role in carcinogenesis? Does current evidence support a tumor suppressive or rather oncogenic function for these proteins? Here, we comprehensively discuss available literature with integration of molecular evidence retrieved from open-access databases. We show that Teneurins undergo somatic changes comparable to those of well-established cancer genes, and discuss their involvement in cancer-related signaling pathways. Current data strongly suggest a functional contribution of Teneurins to human carcinogenesis.

Keywords: Teneurin/ODZ, somatic aberrations, cancer signaling pathways, data mining, tumorigenesis

# INTRODUCTION

Teneurins compose a family of large transmembrane proteins identified over two decades ago in Drosophila (Baumgartner et al., 1994; Levine et al., 1994). Their high degree of interspecies conservation was evidenced through early phylogenetic analyses, and in the human genome, four highly related gene paralogues (TENM1 through TENM4) were predicted based on DNA sequence homologies and partial expression data (Minet and Chiquet-Ehrismann, 2000; Tucker et al., 2012). In vertebrates as in other species, expression of Teneurins has consistently been allocated to the developing nervous system, where they guide formation of neuronal networks through homophilic and heterophilic interactions across synaptic spaces (Hong et al., 2012; Beckmann et al., 2013; Boucard et al., 2014; Berns et al., 2018). Avian and invertebrate Teneurins have further been related to the development of non-neural tissues including the gonads, heart, pharynx, limbs and the gut, among others (Tucker et al., 2001; Drabikowski et al., 2005; Trzebiatowska et al., 2008). In terms of gene organization, Teneurins have shown to be large and complex, displaying multiple alternatively spliced forms in all species analyzed (Tucker et al., 2001, 2012; Lossie et al., 2005; Berns et al., 2018). Accordingly, Teneurin functional analysis is not straightforward and requires experimental consideration of multiple protein subspecies, which might present distinct spatial and temporal expression patterns during development. This inherent complexity challenges the study of newly discovered Teneurin structural and genetic variants. However, the prominent technological and bioinformatics development, together with the ease of access to molecular information gathered into large cancer databases, facilitates oncology research, and can be applied to understanding the significance of Teneurins in carcinogenesis. To this end, the current review integrates tumor-derived genomic and transcriptomic data available through the literature and open-access repositories. The main subjects have been organized according to the historic timeline of their appearance, with the goal to construct a role of Teneurins in cancer starting form a structural point of view into the function and biomedical associations. Based on current knowledge, we propose a plausible role for Teneurins in tumor cell signaling and as recurrent targets of somatic alterations. The function of Teneurins in neural development will not be explicitly addressed, as it will be discussed in detail in other contributions within the current issue.

# TENEURIN GENES: TARGETS FOR STRUCTURAL ALTERATIONS IN CELL LINES AND TUMORS

# Chromosomal Rearrangements in Reports and Data Repositories

Since their discovery over two decades ago, Teneurin research has predominantly addressed their importance as morphogens and determinants of neural connectivity during embryonal development. Intrinsically, this has determined the use of suitable animal models such as C. elegans, Drosophila, chicken and mice, that allowed functional and structural characterization of Teneurins and analysis of gene expression patterns (Baumgartner et al., 1994; Oohashi et al., 1999; Rubin et al., 1999; Fascetti and Baumgartner, 2002; Drabikowski et al., 2005). Conversely, studies assessing the role of human Teneurins were less frequent and their corresponding gene organization was largely predicted through homology-based sequence alignments and partial cloning strategies (Minet et al., 1999; Ben-Zur et al., 2000; Beckmann et al., 2011; Tucker et al., 2012). Discoveries comprising the human orthologs have thus been belated. This includes the identification of tumor-associated somatic changes involving the Teneurins, which has occurred incidentally and is less evident in the literature. In this context, one of the first traceable publications described a chromosomal translocation involving the TENM4 and NRG1 (Neuregulin-1) genes in a breast cancer cell line (Liu et al., 1999; Wang et al., 1999). Importantly, this rearrangement generated a biologically active fusion protein (γ-heregulin) that acted as a secreted autocrine and paracrine growth factor for MDA-MB-175 and MCF-7 breast cancer cells, respectively (Schaefer et al., 1997). Some 10 years later, a second report described recurrent translocations affecting the IGH (Immunoglobulin Heavy Locus) and TENM2 genes in mucosa associated lymphoid tissue (MALT) lymphomas of the skin and the ocular adnexa (Vinatzer et al., 2008). This seems consistent with the role of Teneurin-2 in development of binocular circuits in the visual system (Rubin et al., 2002; Young et al., 2013), while the C. elegans ortholog (Ten-1) is essential for hypodermal development (Mörck et al., 2010). A third translocation involving C11orf73 (current gene name is HIKESHI) and TENM1 was detected in an advanced B3 thymoma (Petrini et al., 2013).

Although scarce, the above findings pose the question whether rearrangement of Teneurin genes is a frequent and biologically relevant event in tumorigenesis. The advent of large, open-access data repositories gathering cancer-related data from thousands of patients, can now partly overcome the lack of explicit reports. Upon a search for chromosomal alterations in two curated sites (ChimerDB 3.0, http://203.255. 191.229:8080/chimerdbv31/mindex.cdb, and "Atlas of Genetics and Cytogenetics in Oncology and Hematology", http:// atlasgeneticsoncology.org/) (Huret et al., 2013; Lee et al., 2017), we were able to retrieve additional chromosomal rearrangements involving the TENM1, TENM2, and particularly TENM4 genes (**Table 1**). Some preliminary observations can be drawn from this overview. First, most reported translocations are derived from a few large and rather recent studies. Earlier data not submitted to a repository might have been missed, as exemplified by the IGH/TENM2 and C11orf73/TENM1 translocations in MALT and B3 thymoma, respectively. The frequency of Teneurin rearrangements might thus be underrepresented. Second, most translocations occur with distinct fusion partners and as unique events, both within each tumor type and among different tumors. Third, rearrangements involving TENM4 clearly predominate (24/32, 75%), followed by TENM2 (5/32, 15.6%) and TENM1 (3/32, 9.4%). No translocations have apparently been reported for TENM3. Finally, 18/32 (56%) of translocations occurred in breast cancer, and most were intrachromosomal (22/32, 69%).

TABLE 1 | Reported tumor cytogenetic rearrangements involving Teneurin genes.


\*c11orf30 current consensus gene name is EMSY; \*\*c11orf67 current consensus gene name is AAMDC; \*\*\*present in DNA only, no expression of fusion transcript. Only rearrangements with information on tumor origin were included.

# Mechanistic Base For Chromosomal Alterations

Under the assumption that all Teneurin genes had comparable sequence coverage in the above genome-wide approaches, these data would suggest preferential chromosome breakage at the TENM4 locus. The presence of common or rare fragile sites, as defined by their susceptibility to form gaps and breaks in metaphase chromosomes exposed to replicative stress, has been associated with structural DNA variation, including chromosomal rearrangements, and genomic instability in cancer (Glover et al., 2017). No explicit fragile sites encompassing Teneurin genes have been described in the literature, which is consistent with the lack of findings when we performed Teneurin gene-based searches of a database covering 118 reported fragile sites in human chromosomes HumCFS, http://webs.iiitd.edu.in/raghava/humcfs/, (Kumar et al., 2017). However, an alternative mechanism proposed that transcription of large genes is associated with both copy number variation (CNV) and the formation of common fragile sites in the genome, through unrepaired lesions derived from failure at the replication fork (Smith et al., 2006; Wilson et al., 2015). Further, differential isoform expression of large genes has been associated with CNVs and common fragile sites that were cell-line specific. Based on this, we might predict that rearrangement of Teneurins could occur in a cell type-specific pattern, depending on the intrinsic transcriptional activity of each Teneurin gene and/or isoform expression in a particular tissue. This seems consistent with our previous detection of Teneurin-4 and Teneurin-2 expression in ovarian tumors and breast cancer cell lines (Graumann et al., 2017), as both genes were preferentially targeted by rearrangements in breast tumors and in one ovarian cancer sample (**Table 1**). Conversely, structural alterations encompassing other Teneurins might be expected in tumors less represented in current databases, such as those affecting the nervous system where Teneurin-3 expression could be more prevalent. In fact, structural alterations of TENM3, including one case of homozygous inactivation, occurred in 5/87 pediatric neuroblastomas and were thus considered recurrent events (Molenaar et al., 2012). The same study detected an interchromosomal rearrangement involving CSMD2 (CUB and Sushi Multiple Domains 2) and TENM3, and one case of massive TENM2 rearrangement caused by chromothripsis, a catastrophic event leading to focal chromosome shredding. Expression of a hybrid transcript encompassing XRCC3 (X-Ray Repair Cross Complementing 3) and TENM4 sequences was detected in a further analysis of neuroblastoma (Boeva et al., 2013). Chromothripsis associated to massive rearrangements of genes between chromosomes 3 and 11, including TENM4, was also described in two small cell lung cancers (SCLCs) (George et al., 2015). Together, this evidence suggests that Teneurin genes are targets of chromosomal rearrangements in multiple solid tumors. This is consistent with the proposed fragility of large transcribed genes, and the observed tissue-specificity of Teneurin rearrangements might possibly relate to their normal, intrinsic expression patterns.

Assuming the model that predicts an overlap of common fragile sites and CNVs in large genes (Wilson et al., 2015) is pertinent, additional signs of genetic instability at Teneurin gene loci should be expected in human tumors. Among these, compelling evidence suggests that oncogenic viruses recurrently integrate at genome fragile sites throughout different cancer types (Feitelson and Lee, 2007; Dall et al., 2008). Indeed, disruption of the TENM2 gene through insertion of hepatitis B virus (HBV) DNA was reported in a liver sample affected by chronic hepatitis (Minami et al., 2005). As this condition can precede hepatic cancer, the presence of early genetic alterations might be considered as initiating events implicated in tumorigenesis. This notion is supported by an additional finding of HBV integration upstream of TENM2 in a hepatocellular carcinoma and its corresponding normal adjacent tissue, which might share common premalignant genetic changes (Jiang et al., 2012). HBV integration in proximity to TENM1 was detected in a fourth adjacent normal specimen, while intragenic insertion at the TENM1 locus occurred in another tumor. HBV integration in hepatocellular carcinoma was also found to target TENM4 (Zhao et al., 2016). In addition to HBV, the 4q35.1 locus encompassing part of the TENM3 gene was identified among a list of recurrent integration sites for human papilloma virus (HPV) DNA in cervix cancer cells (Jang et al., 2014). Although this study did not find fragile sites within TENM3, integration occurred in close proximity to a DNA region interacting with the chromatinbinding BRD4 (Bromodomain Containing 4) and viral E2 proteins. A genome-wide assessment of such sites revealed that they are frequently affected by deletions and that they act to nucleate viral replication foci, whereby viral integration can occur. This mechanism, involving the generation of deletions, could also overlap with a predicted presence of CNVs in the form of deletions/duplications, described as a third hallmark of large gene fragility (Wilson et al., 2015). Accordingly, TENM3 was classified as a large gene recurrently affected by deletions in low grade glioma in a study that analyzed genomic data derived from 30 tumor types (Glover et al., 2017). Nearby focal deletions at 4q34.3 were also recurrent in ovarian, endometrial and adrenocortical carcinoma. TENM3 instability in the form of DNA duplication has also been reported by two independent neuroblastoma studies (Molenaar et al., 2012; Pugh et al., 2013). Importantly, a statistical assessment of gene size distributions suggested that TENM3 structural alterations did not accumulate solely based on gene length, but were the result of active selection during tumorigenesis (Molenaar et al., 2012).

Considering the previous evidence, there appears to be sufficient tumor data to document Teneurin genes as recurrent sites of targeted disruption by chromosome rearrangements, through mechanisms that include translocations, CNVs, chromothripsis, and viral genome integration. Teneurins could thus fulfill criteria to be classified as large transcribed units prone to genetic instability. Additional structural determinants related to this fragility should be examined in more detail in future studies. These include sequence-based DNA motifs and epigenetic parameters such as chromosome organization (Debatisse et al., 2012; Canela et al., 2017). Further, a metaanalysis of virus integration sites confirmed the preference of HPV and HBV for transcriptionally active regions in accessible chromatin, and revealed a consistent mark of DNA methylation and specific histone modifications at these sites (Doolittle-Hall et al., 2015). We previously predicted the presence of several CpG-rich islands at the TENM2 and TENM4 gene regions (Graumann et al., 2017). Although we found no evidence of methylation-based transcriptional regulation of Teneurins in tumor cells, DNA hypomethylation might influence susceptibility to viral integration and translocation events by rendering DNA more accessible, as shown for some lymphoid malignancies (Cui et al., 2013; Lu et al., 2015). Finally, the occurrence of prevalent tandem rearrangements, defined by the involvement of genes that reside on the same strand and chromosome, has been related to an increased length of introns surrounding the fusion break points of both partner genes (Greger et al., 2014). This might explain the predominance of intrachromosomal rearrangements (22/32, 69%) observed for Teneurin genes, that harbor conserved and particularly large introns between the first predicted exon sequences (Minet and Chiquet-Ehrismann, 2000) (**Figure 1A**). Of 21 break points included in **Table 1**, 12 (57%) occurred within intron sequences.

# Are Chromosomal Alterations Functionally Relevant?

RNA-level expression data was available for most Teneurin gene rearrangements reported in tumors, but concomitant protein expression was not assessed. A potential functional contribution to tumor formation can thus only be inferred. To our knowledge, the only exception relates to γ-heregulin, a secreted TENM4/NRG1 fusion protein that displayed growthpromoting biological activity (Schaefer et al., 1997). Consistent with the finding that most Teneurin translocations were unique and involved distinct fusion partners (**Table 1**), γ-heregulin failed to be detected in additional breast cancer cell lines and tumor specimens (Wang et al., 1999; Sánchez-Valdivieso et al., 2002). However, the affected break region maps to a recurrent rearrangement site involving chromosomes 8 and 11 in breast and pancreatic tumors (Adélaïde et al., 2000). Further, in a subset of early-onset pancreas cancers, NRG1 rearrangement correlated with wild-type KRAS and susceptibility to pharmacologic ERBB inhibition (Heining et al., 2018). This matches the biologic behavior of MDA-MB-175 cells expressing γ-heregulin, which do not carry KRAS mutations (Hollestelle et al., 2007) and are sensitive to pertuzumab-mediated inhibition of HER3 (ERBB3) signaling in a mouse xenograft model (Lee-Hoeflich et al., 2008). Together, these data suggest that retention of TENM4 N-terminal sequences in γ-heregulin, which include the entire intracellular and transmembrane domains (Schaefer et al., 1997), generates a functional gene product with oncogenic activity comparable to that of other NRG1 fusion partners. Conservation of the 5′ -Nterminal TENM4 domains is recurrent to other translocations (**Figure 1B**).

For chromosomal alterations that lacked functional assessment, some general points should be considered. First, different outcomes would be expected for gene disrupting rearrangements, as opposed to those generating chimeric proteins through gene fusion. In the former category, gene disruption through recurrent deletion was only described for TENM3 in gliomas (Glover et al., 2017), and homozygous loss of TENM3 was also found in an embryonal rhabdomyosarcoma associated with high risk clinical parameters (Walther et al., 2016). No comparable deletion hotspots have been reported at the TENM1, TENM2, and TENM4 loci (Hazan et al., 2016). However, about one third of chromosomal translocations (**Table 1**) resulted in out-of-frame gene fusions. Interruption of the reading frame is typically associated with a premature translation halt leading to degradation of faulty products, through mechanisms such as nonsense-mediated mRNA decay (He and Jacobson, 2015). Rarely, translation of functional truncated proteins from such rearrangements has been documented (Mertens et al., 2015; Rodriguez-Perales et al., 2016). As Teneurin genes were in a 5′ position in 5/9 out-of-frame rearrangements, expression of a truncated form is unlikely, but theoretically possible. A third mechanism associated to gene disruption occurred through viral insertional mutagenesis, though the effect on Teneurin gene expression was not assessed (Minami et al., 2005; Jiang et al., 2012; Jang et al., 2014; Zhao et al., 2016). In fact, HBV appears to recurrently integrate at Teneurin gene loci. Integration of this virus was found to preferentially target cancer genes and actively transcribed genome sites, consistent with a potential gene disruptive outcome (Doolittle-Hall et al., 2015). Finally, Teneurin genes suffered massive, disruptive rearrangements through chromothripsis (Molenaar et al., 2012; George et al., 2015). When considered together, these mechanisms suggest that a subset of structural alterations might lead to inactivation of Teneurin gene expression in some tumors. The heterozygous loss of function, potentially leading to a haploinsufficient phenotype, could be consistent with a tumor suppressive role in such cases.

In contrast, a different outcome might be expected for rearrangements generating hybrid transcripts. As shown in **Table 1**, expression of fusion-derived RNA was asserted for 20/31 (64.5%) translocations. Of all predicted fusion products, 13/32 (40.6%) were in-frame and at least 2 placed novel 5′ - UTR sequences ahead of a Teneurin gene, potentially enabling chimeric protein expression or transcriptional control through a foreign regulatory region, respectively. Interestingly, 8/9 (89%) out-of-frame rearrangements were also detected at the RNA level, supporting their transcriptional expression. However, such forms might not be active in terms of protein production, as suggested by the low translation index of diverse out-offrame chimeras transcribed in breast cancer cells (Inaki et al., 2011), and as discussed above. With regard to translocations substituting 5′ -UTR regulatory sequences, placement of a strong IGH promoter upstream of TENM2 was associated with at least 3-fold higher Teneurin-2 transcript levels in MALT lymphomas, as compared to tumors not bearing the translocation (Vinatzer et al., 2008). This finding is highly reminiscent of IGH promoter-driven overexpression of MYC, CCND1 (Cyclin D1), BCL2, and BCL6 proto-oncogenes in B-cell malignancies (Zheng, 2013). Importantly, gene translocations have been recognized as early clonal events in hematologic malignancies, and have been associated with causative roles in oncogenic transformation (Mitelman et al., 2007). Since the IGH/TENM2 rearrangement was recurrent in 3 tumors, its phenotypic impact should be investigated. The consequence of other 5′ -UTR substitutions is less evident. Leukemias carrying the KSR1/TENM1 rearrangement showed a hybrid transcript ratio of 1.0 (Yoshihara et al., 2015), indicating lack of endogenous Teneurin-1 expression with sequencing reads solely derived from the fusion transcript. Since basal expression of KSR1 (Kinase Suppressor of Ras 1) was reported in HL60 acute promyelocytic leukemia cells (Wang et al., 2006), a transcriptionally active KSR1 promoter might be expected in leukemia and drive de novo expression of TENM1, with potential functional consequences. No equivalent expression data was specified for the KCTD21/TENM4 translocation detected in breast adenocarcinoma. In ERBB2-amplified breast cancer, expression of KCTD21 (Potassium Channel Tetramerization Domain Containing 21) was related to genomic copy number

alterations (Sircoulomb et al., 2010). The corresponding 5′ -UTR could thus be active in a subset of breast cancers and drive TENM4 expression in the above case. Although validation is missing, this evidence suggests that Teneurin (over)expression could be driven through foreign upstream regulatory sequences in some tumors, suggestive of an oncogenic rather than tumor suppressive role for this mechanism.

Third, almost 40% of translocations (**Table 1**) were predicted to generate in-frame gene fusions, which expressed a corresponding hybrid transcript in all cases where RNA was assessed. This frequency agrees with findings derived from 13 cancer types, where 36% of gene fusions were in-frame (Yoshihara et al., 2015). In the current set, 10/13 in-frame rearrangements occurred in breast tumors. The presence of chromosomal rearrangements in solid tumors has increasingly been reported in recent years and is a matter of current investigations (Mertens et al., 2015). Interestingly, mutation of relevant cancer genes was significantly reduced in breast cancers and other tumors carrying in-frame fusion transcripts, suggesting the latter could function as driver events (Yoshihara et al., 2015). In light of missing functional data to support an analogous role for Teneurin rearrangements, an indirect appraisal might be gained through the analysis of the corresponding fusion partners. Not surprisingly, most of these genes (**Table 1**) have been implicated in the context of malignant diseases. Among the in-frame fusion partners, EMSY (c11orf30), a BRCA2-interacting transcriptional repressor involved in DNA repair, is frequently amplified in breast, ovarian and lung cancers, where it was proposed to exert oncogenic functions (Hughes-Davies et al., 2003; Baykara et al., 2015). For the hepatitis virus A receptor gene, HVCR1, overexpression has been well-documented in clear cell renal cancer and shown to induce growth and angiogenesis-promoting factors (Cuadros et al., 2014). KIF13B, a member of the kinesin gene family, is involved in trafficking of the Vascular Endothelial Receptor 2 (VEGFR2), and disrupting this interaction has demonstrated anti-angiogenic potential in a lung tumor xenograft model (Yamada et al., 2017). Of note, a KIF13B/NRG1 fusion was identified in a liver metastasis and associated with tumor progression (Xia et al., 2017). Both KIF13B and NRG1genes can thus act as Teneurin fusion partners. For all remaining in-frame fusion partners with available cancer-related studies, equivalent results supported their oncogenic roles through over-expression or gene amplification. Only two genes showed opposite behaviors consistent with tumor suppressive functions. Hence, decreased expression of the mitochondrial transporter MTCH2, a pro-apoptotic gene, was associated with enhanced invasiveness and tumor progression in various tumor types (Yu et al., 2008; Arigoni et al., 2013). Similarly, RBM4 was shown to suppress tumor progression and to counter a colorectal metastatic cascade through modulation of alternative splicing (Wang et al., 2014; Lin et al., 2018), and its expression was decreased in various tumor types. For CDCD3 (Desmin), only one cancer related publication could be retrieved that reported a MS4D7/CDCD3 translocation with unknown function in oropharyngeal cancer (Wang et al., 2013). For out-of-frame fusion partners, evidence for both oncogenic (ANO1, TMPRSS2) (Ko et al., 2015; Wang et al., 2017) and tumor suppressive roles (DLEC1, WWC1) (Knight et al., 2018; Li et al., 2018b) could be identified. Of note, all Teneurin fusion partner genes in the current set have been involved in additional translocations (documented at the "Atlas of Genetics and Cytogenetics in Oncology and Hematology," http://atlasgeneticsoncology.org/), albeit at varying frequencies. This suggests that, besides their documented roles in cancer, these genes are frequent targets of structural rearrangements in tumors with probable functional consequences. Thus, Teneurin rearrangements are not incidental and preferentially involve genes related to cancer.

Finally, additional evidence pertinent to tumor gene rearrangements should briefly be considered. Of particular relevance to Teneurins, large transcriptional units prone to breakage in cancer cells show a frequent involvement in neurological development (Smith et al., 2006). This would agree with alterations in Teneurins and other genes that regulate neuritogenesis, which were associated with an aggressive phenotype in neuroblastoma (Molenaar et al., 2012). A role in neurological development and/or disorders is also known for some Teneurin fusion partners (e.g., NRG1, NARS2, SHANK2, GRIK4, TMTC2, and KIF13B), as revealed by generelated information available at Gene Cards (https://www. genecards.org/). Further, large genes often display a high degree of evolutionary conservation, which extends to intronic sequences and fragile sites acting as preferential targets for chromosomal breakage (Greger et al., 2014; Glover et al., 2017). Not surprisingly, genome caretaker, and tumor suppressor functions have been postulated for frequently altered large genes, whose loss might provide a selective advantage to affected cells (Hazan et al., 2016; Karras et al., 2016). Gene conservation is thus considered a possible hallmark of essential cellular functions, which include the DNA damage and stress induced responses (Smith et al., 2006; Hazan et al., 2016). These features might well fit the highly conserved Teneurin genes (Tucker et al., 2012). In the particular case of Teneurin-4, gene expression further responds to conditions of endoplasmic reticulum stress, although a direct biological consequence has not been established (Wang et al., 1998). Finally, high-throughput RNA sequencing revealed that chimeric transcripts resulting from chromosomal rearrangements are expressed in a prominently tissue specific manner (Frenkel-Morgenstern et al., 2012). It was observed that chimeras often retained signal peptides and transmembrane regions, which might redirect potential hybrid proteins to unusual subcellular compartments. As shown in **Figure 1B**, all rearrangements retaining Teneurin intracellular domains preserved the transmembrane region, providing potential membrane anchorage to the 3′ fusion partner. In summary, distinct types of genomic rearrangements can lead to a range of potential outcomes for a same gene, including both oncogenic and tumor suppressive contributions to tumor development (De Braekeleer et al., 2012). Based on current evidence, a similar scenario seems highly probable for Teneurin genes.

# TRANSCRIPTOMIC EVIDENCE OF DYSREGULATED TENEURIN GENE EXPRESSION

Depending on the tumor context, cancer genes can be targeted by different activating and/or inactivating mechanisms, both genetic and epigenetic. Consistent with this, altered expression of Teneurins has been reported for various tumors, in addition to structural alterations discussed above. Previously, we had reviewed transcriptomic data and reported on evidence of decreased Teneurin expression in cancers of the liver, esophagus, and kidney, while increased expression was found for brain tumors and lymphomas (Ziegler et al., 2012). With exception of increased Teneurin-2 levels related to the IGH/TENM2 translocation, accompanying data to explain the biological basis of these changes was not available. However, gene disruption through HBV insertion might be associated to reduced Teneurin-2 levels in hepatocellular carcinoma and was suggested to occur early in tumor development (Minami et al., 2005). For Teneurin-2, further data supported an early time point for expression changes in carcinogenesis of the breast (Lee et al., 2007), while in cervical cancer, RNA levels decreased in advanced stages with nodal compromise or the presence of metastasis (Noordhuis et al., 2011). Based on this data, a first appraisal suggested that both Teneurin up- and downregulation could occur in tumors, with a potential involvement in cancer initiating events as well as in tumor progression.

# Is Teneurin Gene Expression Regulated by Epigenetic Mechanisms?

Since epigenetic modifications are a frequent cause of aberrant gene expression in tumors, we had previously analyzed the effect of a demethylating agent (5-azacytidine) on the expression of Teneurin-2 and Teneurin-4 in cancer cells (Ziegler et al., 2012; Graumann et al., 2017). Although the presence of CpGrich regions within several Teneurin genes was predicted by us and others (Beckmann et al., 2011), we found no evidence for an effect of DNA demethylation on Teneurin-2 and Teneurin-4 expression in breast and ovarian cancer cell lines. However, analysis of the TENM3 promoter region revealed increased methylation in early breast lesions as compared to normal breast tissue (Tommasi et al., 2009), and in immortalized keratinocytes, Teneurin-4 gene expression could be downregulated by overexpression of DNMT3B, an enzyme involved in de novo methylation of DNA (Peralta-Arrieta et al., 2017). This evidence would favor a DNA methylation-based downregulation of Teneurin expression in a manner analogous to that of other tumor suppressor genes, perhaps in a tissue-specific manner. A main drawback of these studies, however, is that concurrent analysis of DNA methylation and transcript expression were not performed, precluding conclusions on a causal relationship between both processes. To date, a single report accomplished such assessment and could demonstrate that, in a subset of glioblastoma cells, methylation of TENM1 upstream sequences was indeed inversely correlated with transcript expression (Talamillo et al., 2017). For Teneurin-1, a further role in modulating the epigenetic landscape has been proposed based on the interaction of its cleaved intracellular domain with MBD1, a nuclear CpG-binding transcriptional repressor (Nunes et al., 2005). MBD1 participates in mechanisms that regulate heterochromatin formation, and in tumors, it has been associated with silencing of tumor suppressor genes, promotion of oncogenic attributes, and a reduced response to cisplatin and radiation (Li et al., 2015; p. 1). Interestingly, MBD1 is highly expressed in neural stem cells and its defects can impair neuron differentiation (Li et al., 2015; p. 1), underscoring its close functional link to Teneurins in non-tumor tissues. These data suggest that Teneurin-1 might be part of an epigenetic regulatory circuit, both as a modulator of gene expression and as a target of methylationbased gene silencing. However, it should be kept in mind that, owing to its cytogenetic localization, Teneurin-1 might be subjected to distinct epigenetic control mechanisms related to gender-dependent X-chromosome inactivation. Whether other Teneurin genes are regulated by equivalent methylation-based mechanisms remains to be demonstrated. As an example, differential methylation of TENM2, TENM3, and TENM4 was reported in a genome-wide analysis of neuroblastoma, but no findings for TENM1 were registered (Gómez et al., 2015). Further, methylation occurred outside of CpG islands most commonly associated with methylation-based transcriptional control, and methylation changes were not correlated with altered Teneurin gene expression. This data suggests that, although TENM1, TENM2, and TENM3 genes might show tumor-related differences in methylation patterns, per se this does not imply a concomitant change in gene expression. These aspects demand comprehensive analyses that remain to be addressed. At present, epigenetic control has been sufficiently documented only for TENM1.

### Uncovering Teneurin Gene Regulatory Pathways in Tumors: The NOTCH Connection

As discussed above, DNA methylation seems not sufficient to explain tumor-related changes in Teneurin gene expression. A reasonable alternative should consider Teneurins as downstream targets of cancer-specific cellular signaling, which might impact at the level of gene expression, proteolytic protein processing, or the regulation of specific domain activity by posttranslational modifications, among others. This would be consistent with the fact that Teneurin translocation partners were enriched in cancer-related genes, which delineates a well-described oncogenic mechanism. With regard to signaling pathways, vast structural and functional parallels between Teneurins and Notch proteins have recurrently been noted, which include their analogous transmembrane localization, their capability to dimerize upon ligand interaction, the presence of multiple epidermal growth factor-like (EGF) repeats in their extracellular domains, and their processing into multiple domains through proteolytic cleavage (Tucker and Chiquet-Ehrismann, 2006; Schöler et al., 2015; Vysokov et al., 2016). These associations were recently shown to reach deeper, as overexpression of the NOTCH1 intracellular domain, which functions in a nuclear transcription activator complex, revealed that TENM4 is a NOTCH-responsive gene (George et al., 2015). The authors proposed that NOTCH signaling acts as a key regulator of neuroendocrine differentiation in small cell lung cancer. This seems consistent with a potential effector role of Teneurins, considering their essential participation in neural development (Tucker et al., 2007). TENM4 responsiveness to NOTCH signaling was also noted in muscle satellite cells, where both were required for maintenance of cell quiescence and inhibition of myogenic differentiation (Bröhl et al., 2012; Ishii et al., 2015). In small cell lung cancer, NOTCH signaling was ascribed a tumor suppressive role, and as a NOTCHregulated gene, an equivalent function might be expected for TENM4. Notably, the similarities between both gene families go well-beyond the observations outlined above, and could define a highly probable setting governing Teneurins' involvement in cancer. As recently reviewed in great detail (Nowell and Radtke, 2017), the contribution of NOTCH-mediated signaling is particularly dependent on the cellular and tumor context, with both oncogenic and tumor suppressive outcomes. Oncogenic activation can occur through chromosome translocations or activating NOTCH mutations in leukemia, while inactivating mutations occur in tumors where it exerts a tumor suppressive role, such as in small cell lung cancer (George et al., 2015). Further, NOTCH4 is a target for mouse mammary tumor virus (MMTV) integration in mice, and NOTCH gene expression can be altered by some HPV viral proteins, unveiling a role for viral-dependent mechanisms that shows remarkable similarities with Teneurins. The phenotypic contribution of NOTCH dysregulation could relate to its ability to negatively and positively regulate differentiation and stem cell fate, in analogy to the role of Teneurins in modulating cellular differentiation in different cell types (Suzuki et al., 2012, 2014a; Ishii et al., 2015; Tews et al., 2017). Not surprisingly, expression of NOTCH family genes has been associated with clinical parameters and patient outcome in cancer. Based on the vast implications of NOTCH signaling in malignant diseases, this pathway is a current target for development of directed therapeutic interventions.

# Teneurin Expression Is Associated With Biological and Clinical Parameters

Assuming that Teneurins are active players in tumorigenesis through mechanisms analogous to NOTCH, similar biological and clinical findings would be expected for this gene family. Not surprisingly, Teneurin expression has been related to tumor behavior and patient survival in several cancer types. In invasive and aggressive-invasive prolactin pituitary tumors, upregulation of Teneurin-1 mRNA was associated with tumor progression (Zhang et al., 2014), and a similar observation was made for papillary thyroid cancer (Cheng et al., 2016). In the latter case, Teneurin-1 expression was further associated with extra-thyroidal invasion, an advanced disease stage, the risk of recurrence, and the presence of BRAF V600E, an actionable mutation with a known prognostic significance in this cancer. For Teneurin-3, decreased levels predicted a worse survival in neuroblastoma patients (Molenaar et al., 2012), while expression was upregulated in breast tissue of nulliparous women, known to be at higher risk of developing breast cancer (Balogh et al., 2006). Interestingly, differentiation of breast tissue is not fully accomplished in these women and maintains a high proportion of stem cells, suggesting that upregulation of Teneurin-3 might relate to an altered differentiation process, as discussed above. In the same line, we found a decrease in Teneurin-4 expression in high grade serous ovarian tumors that undergo dedifferentiation, and reduced Teneurin-2 levels predicted a poor survival in these patients (Graumann et al., 2017). To retrieve additional prognostic associations, we performed a gene-based query at the Human Protein Atlas (https://www.proteinatlas. org/pathology), which provided statistical correlations of patient survival based on quantitative analysis of transcriptomic data. **Table 2** summarizes Log-rank p-values based on either optimal or median separation of low and high-expressing tumor groups, providing evidence for significant prognostic implications of Teneurin expression in a range of different cancers. Interestingly, all four Teneurins were associated with patient survival in cancers of the endometrium, kidney and stomach, and with the exception of Teneurin-3 in renal cancer, survival was improved for patients with low Teneurin levels, possibly hinting to a common underlying mechanism. These findings raise the question whether all Teneurins were concomitantly expressed in these tumors. At least in some cell lines, we could demonstrate a simultaneous expression of Teneurin-2 and Teneurin-4 (Graumann et al., 2017), and coexpression of different Teneurins was also noted in one neuroblastoma cell line (Suzuki et al., 2014b), although the functional implications remain unknown. In contrast, other tumors showed an apparent gene-specific association, such as melanoma and colorectal cancer (Teneurin-2 only) and cervical cancer (Teneurin-3 only) (**Table 2**). About one third of cancers showed a better survival upon increased Teneurin expression, suggesting a distinct biological behavior for this group. Hence, results based on transcriptomic analysis of large TCGA patient cohorts strongly support an involvement of Teneurins in tumor biology, evidenced through their prognostic association with patient survival and tumor differentiation. It is highly probable that, as additional cancer types will be analyzed in sufficiently large numbers, new findings will emerge in a near future. Based on current data, it is conceivable that Teneurins' role in tumorigenesis could relate to their ability to modulate cell differentiation, in addition to other processes discussed below.

### Cancer Pathways: Interaction Points of Teneurins and WNT Signaling

Considering the above associations with clinicopathological parameters, it seems reasonable that cancer-related biological mechanisms might regulate—or be regulated by—Teneurins. Besides the dazing relation between Teneurins and NOTCH, several studies are consistent with this prediction. For instance, in node-positive, poor prognosis cervical cancers, Teneurin-2 levels were increased together with CTNND1 (Noordhuis et al., 2011), a member of the catenin gene family that stabilizes E-Cadherin at epithelial adherens junctions and mediates noncanonical WNT signaling (Schackmann et al., 2013). Upon E-Cadherin loss, CTNND1 mislocalizes to the cytoplasm and aberrantly regulates Rho-mediated signals, leading to the induction of a migratory and invasive phenotype through epithelial-mesenchymal transition (EMT). Interactions between TABLE 2 | Significant associations between Teneurin transcript levels and patient survival.


\*"High" and "Low" refer to Teneurin transcript expression level groups according to applied threshold, defined by either Optimal separation or Median expression levels.

Teneurins and WNT signaling are also relevant to normal embryonal development. Hence, Teneurin-4 was a main binding partner of Olfactomedin-1 (OLFM1), a secreted glycoprotein involved in regulation of proliferation and differentiation of neural progenitor cells in the brain, and proposed to regulate small GTPase RhoA activity and WNT signaling (Nakaya et al., 2013). In avian limb development, a potential interaction between the WNT7A ligand and Teneurin-1 or Teneurin-3 was suggested based on their expression in common cellular compartments (Bagutti et al., 2003). A similar association might be deduced from the concomitant decrease in WNT7A and Teneurin-4 expression observed in neurons of schizophrenia patients obtained by in vitro differentiation (Brennand et al., 2011). Further, defects in WNT7A impair female genital tract development and have been associated with Müllerian duct anomalies in mice (Choussein et al., 2017), while in C. elegans, disruption of Ten-1 impairs development of somatic and germline gonadal cells (Drabikowski et al., 2005). These data suggest that during development, Teneurin expression might respond to WNT7A-mediated signals in some common compartments, thereby altering biological processes such as cell adhesion and migration. However, these associations do not always hold and especially in tumors, the role of WNT7A is more complex as it can exert opposite oncogenic as well as tumor suppressive functions (Stewart, 2014; Huang et al., 2018). In fact, Teneurin-1 and EMX2 levels were highly increased and HOXA10 levels were reduced in women with Müllerian defects leading to a partially separate uterus (Zhu et al., 2016). Activity of the homeobox transcription factor HOXA10, a direct negative regulator of EMX2 gene expression, is dependent on an intact WNT7A function (Miller and Sassoon, 1998). In turn, EMX2 is a direct activator of TENM1 transcription (Beckmann et al., 2011). This would define an opposite signaling cascade where WNT7A positively regulates HOXA10 function, leading to repression of EMX2 and its downstream targets, including Teneurin-1. Since EMX2 has an antiproliferative function in endometrial cells (Taylor and Fei, 2005), its elevated expression seems consistent with the impaired cell growth associated with Müllerian duct anomalies, where Teneurin-1 could act as growth restricting downstream effector. Equally, Kaplan-Meier estimates available at The Human Protein Atlas (https://www.proteinatlas. org/pathology), register improved survivals for patients with endometrial cancer expressing high EMX2 (log-rank test P = 0.0054) and low WNT7A (P = 0.00003) levels, indicating an advantage of persistent antiproliferative signals. However, the same does not hold for Teneurin-1, whose low expression was associated with improved survival in these patients (**Table 2**). This result further contrasts findings in C. elegans, where germline tumors resulted upon deletion-mediated loss of expression of the Ten-1 ortholog (Drabikowski et al., 2005). Thus, prognostic conclusions can not necessarily be extrapolated from predicted pathway interactions and must consider additional factors, as exemplified by the finding of both positive and inverse associations between Teneurins and WNT7A in different cellular contexts.

Recently, a deeper insight has been provided into molecular mechanisms underlying the role of Teneurin-1 in cancer. The authors showed that in glioblastoma cells, loss of Teneurin-1 expression through chromosomal deletion or epigenetic silencing was associated with resistance to serum-induced differentiation (Talamillo et al., 2017). Although exogenous Teneurin-1 expression restored a differentiated phenotype, it provided cells with an enhanced migratory and invasive potential, suggesting a fine equilibrium between Teneurin-1-mediated regulation of differentiation fate and migratory capacity. Further, increased Teneurin-1 levels were predictive of a poor outcome in glioblastoma patients and xenograft models, consistent with transcript-based survival estimates for glioma patients (**Table 2**). Strikingly, the cleaved intracellular (ICD) but not the extracellular Teneurin-1 domain was capable of eliciting the migratory and invasive properties, as well as a rearrangement of the actin cytoskeleton, the expression of mesenchymal markers, and an increased resistance to toxicity mediated by the alkylating agent temozolomide. In functional terms, the Teneurin-1 ICD acted in the nucleus through direct interaction with the MYC oncoprotein, inducing transcriptional activation of the small GTPase RHOA gene. These findings place Teneurin-1 as an executor of MYC-RHOA-induced responses, which are associated with oncogenic signaling through WNT pathways in glioblastoma, although additional components of this pathway were not analyzed (Lee et al., 2016). This study strengthens the above evidence of a functional link between Teneurins and WNT signaling, and points to the relevance of proteolytic processing in the generation of Teneurin domains with distinct functional attributes. Similar findings had been described earlier for the Teneurin-2 and Teneurin-3 ICDs, which were implicated as negative regulators of ZIC1 function and ZIC2 expression, respectively (Bagutti et al., 2003; Glendining et al., 2017). The ZIC transcription factors appear to accomplish mainly tumor suppressive roles, but overexpression occurs in some cancers, suggesting an alternative oncogenic function (Houtmeyers et al., 2018). At least for the Teneurin-3 ICD, a direct interaction with ZIC2 could be demonstrated in vitro (Glendining et al., 2017). Interestingly, the activity of ZIC transcription factors can be inhibited in C. elegans neural progenitors by WNT downstream effectors ß-catenin and TCF (Murgan et al., 2015), raising the possibility that Teneurin ICDs might act as part of this protein complex to modulate transcription factor activity. In addition, Teneurin-3 knock-out experiments showed that its ICD can act as a positive and negative regulator of Ephrin receptors EPHA7 and EPHB1 expression, respectively (Glendining et al., 2017). The concomitant decrease in ZIC2 and EPHB1 is consistent with the positive regulatory role that ZIC2 exerts on EPHB1 transcription (García-Frigola et al., 2008). Ephrin receptors activate signaling cascades through RHOA, AKT and ERK, to regulate cell growth, migration, and EMT, and their expression is altered in numerous tumor types (Kou and Kandpal, 2018). Hence, modulation of Ephrin-mediated processes provides an additional explanation of Teneurins contribution to tumor development. Further, the distinct activities of cleaved Teneurin ICDs imply that their preservation in chromosomal translocation products (**Figure 1B**) might actively contribute to oncogenic functions, through ICD-mediated transcriptional modulation of relevant target genes (Bagutti et al., 2003). Since proteolytic release of the Teneurin-2 ICD can be promoted by homophilic interactions (Bagutti et al., 2003), it should be examined if alternative ligand-dependent mechanisms operate in tumor cells, perhaps as the result of Teneurin interactions with components of the extracellular matrix (ECM). Thus, current evidence supports a role of Teneurins as mediators or effectors of WNT signaling, which might be associated with tumor suppressive as well as oncogenic outcomes. We had previously proposed a model to describe such potential interactions (Ziegler et al., 2012).

# Teneurins and Modulation of Neuregulin-ErbB-Mediated Signaling

In cancer cells, different signaling pathways often converge at common cross-points to establish complex regulatory networks. It is thus not surprising that Teneurins might appear in different molecular contexts directed at a common functional endpoint. As noted above, a hybrid NRG1/TENM4 fusion product (γ-heregulin) displayed growth promoting activity on different tumor cell lines (Schaefer et al., 1997). Although these experiments focused on growth stimulation associated with the secreted Neuregulin-1 ligand, it should be recalled that proteolytic cleavage of γ-heregulin would simultaneously generate a membrane anchored Teneurin-4 ICD, which might exert additional roles to support oncogenic transformation. Interestingly, NRG signaling through ERBB receptors accomplishes crucial functions in nervous system development that are highly reminiscent of Teneurins (Mei and Nave, 2014). Moreover, single nucleotide variants (SNVs) in NRG and ERBB genes have been associated with an increased risk of psychiatric conditions such as schizophrenia and bipolar disorder (Mei and Nave, 2014), analogous to findings reported for Teneurin-4 (Psychiatric GWAS Consortium Bipolar Disorder Working Group, 2011; Ivorra et al., 2014). Similarly, neurons from schizophrenia patients differentiated in vitro showed misexpression of various genes including TENM4, NRG1 and ERBB4, suggestive of a functional connection between these proteins. A similar association might be inferred from experiments with overexpression of the Teneurin-1 ICD in glioblastoma cells, which lead to an increased transcriptional activation of melanogenesis-associated transcription factor (MITF) target genes, including ERBB3 (Schöler et al., 2015). Teneurins might thus support this oncogenic pathway by promoting the expression of ERBB receptors to provide sufficient binding sites for NRG ligands. In line with this notion, a recent study using CRISPR-CAS9 to knock-out ZEB1, a homeobox transcription factor that mediates epithelial to mesenchymal transition (EMT) in response to TGF-ß, found that Teneurin-2 transcription is directly repressed by ZEB1 in triple-negative breast cancer cells (Maturi et al., 2018). Interestingly, ERBB4 expression increased over 5-fold upon ZEB1 knockout, with a concomitant 16-fold increase in Tenerin-2, which agrees with the association proposed for these gene families. In the case of breast cancer, it is interesting to speculate that the positive prognostic impact of Teneurin-1 overexpression (**Table 2**) might be associated with a simultaneous expression of ERBB receptors, which provide an actionable target for directed therapies able to positively impact on patient survival (Hynes, 2016). This interaction illustrates how additional factors can modify an expected prognostic impact predicted by molecular parameters. As a further example, ZEB1 knock-out cells showed diminished invasiveness and a delayed migratory behavior, which might predict a positive prognostic impact of Teneurin-2 reexpression. However, the opposite was true in a small group of patients with triple-negative breast cancer, where increased Teneurin-2 was associated with shorter metastasis-free survival (Maturi et al., 2018). A possible explanation might relate to the poor prognostic impact of ERBB4 expression in triple negative breast cancer patients receiving standard, non-targeted treatment regimens (Kim et al., 2016). In the former study, the therapeutic modalities were not specified and preclude assessment of this variable, which might modify the prognostic impact of Teneurins in a treatment-dependent manner (Maturi et al., 2018).

If the proposed association holds, Teneurins would be linked to additional cancer signaling pathways, as NRG1/ERBBdependent phosphorylation triggers PI3K/AKT and MEK/ERKmediated responses (Roskoski, 2014). This interaction should be analyzed in depth in normal and tumor cells, as it involves essential mechanisms in both contexts. To analyze if coexpression of Teneurin and ERBB genes occurs frequently in tumors, transcriptomic analyses of larger sample groups are required and should consider the therapeutic modalities received by patients as key determinants of patient outcome. The molecular role of Teneurin ICDs in mediating transcriptional activation of ERBB genes should also be assessed.

#### Additional Functional Considerations

Another aspect to be considered is the well-documented interaction of Teneurins with cytoskeletal and extracellular matrix (ECM) components, which could bear potential relevance to cancer. Hence, a role for Teneurins in cell migration was recognized early in Drosophila, as heterophilic interaction of Ten-m with PS2 integrins was shown to promote cell spreading (Graner et al., 1998). Further, disruption of Ten-m or the actincrosslinking protein Filamin resulted in a comparable phenotype characterized by altered cell migration and routing of motor neurons (Zheng et al., 2011), suggesting a connection between Teneurins and the actin-based cytoskeleton in cell motility. The authors could also demonstrate that Filamin and Tenm physically interact in epidermal cells. Interestingly, actin cytoskeleton dynamics appears to be regulated by both the Teneurin-1 ICD (Talamillo et al., 2017) and the Teneurin Cterminal associated peptide (TCAP-1) (Chand et al., 2012). These constitute two structurally unrelated domains located on opposed Teneurin protein termini, derived either through proteolytic cleavage or by transcription from alternative intronic promoters. In mouse hypocampal cells, cellular effects of the secreted TCAP1 domain were mediated by interaction with a dystroglycan receptor complex. This elicited the activation of MEK and ERK-dependent signaling, leading to filamin phosphorylation and actin polymerization required for outgrowth of cellular protrusions. In contrast, the Teneurin-1 ICD acted through its proposed nuclear role in transcriptional regulation by inducing RHOA gene expression, with resulting activation of Rho-dependent kinase (ROCK) signaling and remodeling of the actin cytoskeleton. In this case, glioblastoma cells acquired features of increased tumor cell aggressiveness, as discussed above. No evidence is currently available to support the concomitant activity of TCAPs and Teneurin ICDs in tumors or normal tissues. TCAPs are bioactive peptides capable of eliciting intracellular signals that could be relevant to tumorigenesis (Wang et al., 2005). These issues deserve further clarification and illustrate the complex mechanisms that underlie the activity of Teneurins.

With regard to interaction of Teneurins with components of the extracellular matrix (ECM), the integrity of basal membrane structures in C. elegans was dependent on the interaction of Ten-1 at the surface of epidermal cells with collagen IV in the extracellular space (Topf and Chiquet-Ehrismann, 2011). Conversely, Teneurin-4 was shown to negatively regulate expression of collagens type II and X through mechanisms involving ERK-dependent signaling, which was associated with suppression of chondrogenic differentiation and preservation of a mesenchymal phenotype (Suzuki et al., 2014a). Thus, Teneurins can alter determinants of cell adhesion and migration, which are essential targets in tumor cells, by direct interaction and modulation of components of the ECM. Interestingly, a proteomic study revealed a prominent increase of membrane-associated Teneurin-1 upon activation of platelets with a collagen-related peptide (Wright et al., 2011). Platelets are important promoters of tumor development through mechanisms that include the release of proangiogenic and growth promoting factors (Plantureux et al., 2018). Further, platelets adhere to tumor cells and protect them against mechanical forces and immune surveillance in the bloodstream, providing an essential contribution to metastatic spreading. The adhesion of tumor cells to platelets is mediated by various cell surface molecules such as integrins, P-selectin, and podoplanin. Since Teneurins engage in homophilic and heterophilic interactions that can mediate intercellular adhesion (Rubin et al., 2002; Boucard et al., 2014), they might facilitate the contact of platelets with tumor cells and promote tumor metastasis. The reported interaction of Teneurins with integrins seems consistent with such mechanism (Graner et al., 1998; Trzebiatowska et al., 2008), as specific integrin subtypes are expressed in tumor cells and on the platelet surface (Wright et al., 2011). A crossing point of Teneurins and integrins is supported by additional data. Hence, focal adhesion kinase (FAK)-dependent signaling was induced by Teneurin-4 in neuroblastoma cells, and phosphorylated FAK colocalized with Teneurin-4 at sites of neurite protrusion formation, together with the Rho GTPases cdc2 and Rac1 (Suzuki et al., 2014b). Integrin-mediated signaling commonly activates the FAK pathway, and inhibition of oncogenic FAK activity bears therapeutic importance (Kolev et al., 2017). Further, Rho GTPases are key regulators of cell migration (Sadok and Marshall, 2014) and have been implicated in several contexts as mediators of Teneurin functions (Nakaya et al., 2013; Glendining et al., 2017; Talamillo et al., 2017), as discussed above. In addition, the Teneurin-1 ICD was shown to bind to CAP/Ponsin (Nunes et al., 2005), a cytoskeleton adaptor molecule that interacts with FAK to regulate focal adhesion and cytoskeleton dynamics, thus impacting on cell adhesion and migration (Tomasovic et al., 2012). Finally, Teneurin-4 and Laminin, the common ligand for integrins, showed partly overlapping localization patterns in the developing avian gut (Kenzelmann-Broz et al., 2010), and in C. elegans, evidence suggested that laminin Epi-1 might act as a Ten-1 ligand (Trzebiatowska et al., 2008). Together, these findings suggest that Teneurins modulate and interact with components of integrin mediated signaling to modify crucial components required for cell migration and invasion. The localization of Teneurins at essential sites of cytoskeletal anchorage, focal adhesion, and attachment to the extracellular matrix, is consistent with this role. Teneurinmediated adhesion is actively involved in cell signaling through well-characterized cancer-related pathways. Such signals might be initiated by context-dependent, hemophilic, or heterophilic intercellular contacts between Tenerins and/or integrins, or by Teneurin-mediated signaling derived from interaction with ECM components such as collagens.

The role of Teneurins as modulators of cytoskeleton dynamics might also be relevant to drug-resistance, where microtubules play an important part. As showed in Drospohila, Teneurin disruption lead to disorganized microtubule and α-spectrindependent cytoskeletal structures that impaired transsynaptic organization (Mosca et al., 2012), and in mouse hippocampal cells, TCAP-1 increased levels of tubulins alfa and beta at cellular protrusion sites (Chand et al., 2012). Further, the microtubuleactin cross-linking factor 1 (MACF1) protein was identified as a Teneurin-1-ICD binding protein (Schöler et al., 2015). Interestingly, interaction with MACF1 would strengthen the link of Teneurins with ERBB-mediated signaling, as in breast cancer cells, ß-heregulin could induce ERBB2-dependent protrusions that were enriched in microtubules (Zaoui et al., 2010). It could be shown that MACF1 acted as a downstream effector of ERBB2 signaling that mediated microtubule capture at the cell membrane. MACF1, which is highly expressed in the developing nervous system, is further required for WNT-signaling (Chen et al., 2006), whose involvement with Teneurins was discussed above. MACF1 expression was also prominent at advanced stages in brain tumors, including glioblastoma (Afghani et al., 2017). The authors showed that MACF1 knock-out could reduce proliferation and migration of glioblastoma cells, which was accompanied by a reduction in WNT signaling effectors. Further, downregulation of MACF1 increased sensitivity of glioblastoma cells to the DNA alkylating agent temozolomide. Since overexpression of the Teneurin-1-ICD increased resistance to this drug in glioblastoma (Talamillo et al., 2017), interaction of Teneurin-1 and MACF1 might contribute to this phenotype, possibly involving the stabilization of actin and microtubule cytoskeletal structures and WNT signaling activity. A role for Teneurins in drug resistance is also supported by the massive overexpression (> 200-fold) of Teneurin-2 in an ovarian cancer cell line resistant to vincristine, a microtubule-targeting vinca alkaloid (Buys et al., 2007). In addition to increases in transporter genes, Fibronectin-1 was also massively augmented, which lead the authors to propose a cell adhesion mediated drug resistant (CAM-DR) phenotype, in which anchorage to the ECM appears essential for cell survival in the presence of antineoplasic drugs. Interestingly, Lathrophilin-3 and several collagens were also upregulated, suggesting that Teneurins could interact in a common pathway with these adhesion molecules to mediate drug resistance related to adhesion. Consistent with this prediction, CAM-DR involves signaling through MEK/ERK and FAK pathways (Dickreuter and Cordes, 2017) and in breast cancer, upregulation of Teneurin-related Tenascin-C was implicated (Jansen et al., 2005). Also, the Ephrin receptor-A4 (EPHA4) was required for CAM-DR in multiple myeloma (Ding et al., 2017), and Teneurins are known regulators of Ephrin receptor expression in structures of the visual system (Young et al., 2013; Glendining et al., 2017). In this context, Teneurin-2 knock-out caused concomitant reductions in EPHB1 and CFOS, and the latter was also reduced upon Teneurin-3 knock-out (Merlin et al., 2013; Young et al., 2013). This suggests that expression of the oncogenic transcription factor c-Fos, which is a target of ERK-mediated phosphorylation, might be under transcriptional regulation of Teneurins. These findings again highlight the role of MEK-ERK signaling as a central component of Teneurin-mediated functions in tumors. The transcriptional regulation of Ephrin receptors, whose expression is frequently augmented in tumors, might provide a further mechanism of CAM-DR promotion through Teneurins.

In summary, altered expression of Teneurins has been demonstrated in numerous tumors and can be an early event in cell transformation. Recent data have provided new functional insights demonstrating that Teneurins respond-to and can orchestrate signaling pathways with known roles in carcinogenesis, invasion and drug resistance. Not surprisingly, Teneurin expression is associated to patient outcome and could bear prognostic implications. In future studies, distinct and domain-specific functions must be considered that involve Teneurin-mediated transcriptional regulation, extracellular signaling, and adhesion between cells and with components of the extracellular matrix. Consistent with this complex scenario, both up- and downregulation of Teneurins can be expected in tumors, and their dual tumor suppressive or oncogenic function might parallel that of other proteins implicated in cancer.

# TENEURINS AS TARGETS OF SOMATIC MUTATION

As evidenced in the above sections, tumor-related Teneurin alterations conform to mechanisms expected for typical cancer genes, which include changes in gene structure and expression levels suggestive of both oncogenic and tumor suppressive functions. If these predictions hold, it should be possible to identify additional mechanisms common to cancer genes, such as sequence changes through somatic mutation. The analysis of such findings presents some difficulties. First, Teneurins are encoded by very large genes. In probabilistic terms, this means that large DNA sequence stretches might be expected to show a proportionally high number of variants throughout the entire gene length, affecting different domains of the encoded protein. Second, functional studies to address the impact of every such change do not exist or are limited to different contexts, as will be discussed. Similarly, there is insufficient evidence to catalog single nucleotide variants as benign causes of genetic variation, or instead, as pathogenic somatic changes. In spite of these limitations, records of somatic alterations can be retrieved from large databases and subjected to a comparative analysis, as we expose below.

#### The Impact of Single Nucleotide Variants

Considering the unknown functional impact of Teneurin SNVs in tumors, supportive evidence can be inferred from the analysis of variants in other pathogenic conditions. As mentioned, Teneurin genes are strongly conserved throughout species (Minet and Chiquet-Ehrismann, 2000; Tucker et al., 2012) and sequence variation might thus impact on protein function. Accordingly, a novel variant in TENM3 was associated with developmental defects of the visual system leading to small eyes (microphtalmia) and impaired vision (Aldahmesh et al., 2012). The variant (p.T695Nfs<sup>∗</sup> 5) introduced a frameshift mutation resulting in a premature stop codon at the extracellular side of the Teneurin-3 transmembrane domain. It was shown to completely abolish TENM3 gene expression, although the intracellular and transmembrane domains were correctly encoded. Since the variant was homozygously inherited in two affected siblings, it resulted in a null phenotype with complete absence of Teneurin-3. This phenotype is consistent with impairments in the visual system observed in Teneurin-3 knock-out mice (Leamey et al., 2007; Glendining et al., 2017). In terms of frequency, the mutation could not be found in healthy controls or in a database with variant coverage derived from >10.000 chromosomes, which supported it was an uncommon, pathogenic change. Some years later, 12 novel missense variants in TENM4 were discovered in Spanish patients affected with essential tremor (Hor et al., 2015), a phenotype also observed in mice upon Teneurin-4 knock-out (Suzuki et al., 2012). Unlike microphtalmia, this condition was inherited in an autosomal dominant fashion, and eleven variants were predicted to be damaging by various in silico algorithms. Three variants were characterized further and showed a trend to altered Teneurin-4 protein clustering at cell membranes. Further, they increased the number of smalldiameter neuronal axons and induced errors in branching and pathfinding in zebrafish, consistent with previous findings in Teneurin-4 knock-outs. The authors suggested a dominantnegative effect possibly related to an impaired ability of Teneurin-4 to engage in homophilic interactions, as two of the variants (p.T1367N, p.A1442T) altered essential NHL-repeat/β propeller motifs in the extracellular domain. In contrast, TENM4 variants could not be associated with essential tremor in Canadian patients (Houle et al., 2017). In this cohort, missense variants were more frequent in cases (25%) than controls (14%), but could not be statistically associated with disease. One shortcoming was that variant segregation in cases and their relatives was not assessed. With few exceptions, missense changes were present in 1 or 2 individuals only, and nonsense variants were extremely rare with only one identified case. This suggests that deleterious loss of function changes are not well-tolerated in the germline and might be subjected to negative selection (Martincorena et al., 2017), while missense variants were more frequent but occurred at low allelic frequencies. The nonsense mutation generated a premature stop codon close to the Teneurin-4 Cterminus, which should preserve the largest portion of the protein intact. This bears similarity to a missense variant affecting a conserved residue within the globular C-terminal domain of murine Teneurin-4, which lead to embryonic failure due to severe gastrulation defects (Lossie et al., 2005). An additional missense variant within the Teneurin-3 C-terminal TCAP-domain was recently found to segregate with developmental dysplasia of the hip in a large pedigree with severe manifestations (Feldman et al., 2017). Interestingly, the cytogenetic localization of TENM3 maps to a locus previously associated with this dysplastic condition, where at least one additional gene has been implicated (Zhao et al., 2017). These data suggest that, although a major part of the protein might not be affected, alterations at Teneurin C-terminal domains should not be dismissed in terms of potential phenotypic impact. Besides TCAPs, an apoptogenic C-terminal domain was recently described for Teneurin-2 (Ferralli et al., 2018). The physiologic stimuli triggering its release remain unknown, but the domain was proposed to model neural networks through selective induction of apoptosis at intersynaptic spaces. Finally, a rare missense variant in TENM1 was associated with a familial case of congenital general anosmia (Alkelai et al., 2016). This variant (p.P1610L) had not been reported in the population and was categorized as damaging by eight predictive algorithms. According to current gene annotations, this change would localize amidst conserved tyrosine-aspartic acid (YD) repeat motifs. These repeats, best known from bacterial proteins, appear to be glycosylated and to mediate cellular aggregation (Feng et al., 2002; Rubin et al., 2002). Taken together, these reports demonstrate that SNVs in Teneurin genes can result in pathogenic outcomes, although the mechanisms involved have not been studied in depth. A potential functional redundancy between different Teneurins should also be considered, which might constrain the phenotypic severity derived from single pathogenic variants (Leamey et al., 2007; Trzebiatowska et al., 2008).

#### Recurrent Somatic Changes in Lymphoma

The above examples of microphtalmia and essential tremor represent homozygous loss of function as opposed to heterozygous, dominant function gain, respectively. In cancer, such changes would match mechanisms of tumor suppressor loss and oncogene activation, triggered through acquisition of somatic mutations. Two studies addressing the mutational landscape in lymphomas have listed somatic variants in Teneurins. The first performed whole genome sequencing of diffuse large B-cell lymphoma, and could identify mutations in all Teneurin genes (Morin et al., 2013). Interestingly, TENM2 was validated as frequently mutated in this cancer by analyzing data from additional cohorts. Further, TENM2 mutations showed evidence for positive selection, which favored their driver function in tumorigenesis. Two of these mutations were nonsense and predicted the expression of a truncated protein or its degradation, in analogy to the variant observed in microphtalmia (Aldahmesh et al., 2012). However, it should be noted that most Teneurin variants in this report were accompanied by annotation errors and their precise location cannot be deduced. A second study performed whole-exome sequencing of primary lymphomas of the central nervous system (Vater et al., 2015). One missense mutation was found in TENM3, while TENM4 was among the 4 most frequently mutated genes (missense mutations in 4 of 9 tumors). Three mutations mapped to the extracellular domain, of which two lied between YD repeats and one was predicted to affect an EGF-like motif. A fourth mutation affected a potential phosphorylation site within the ICD. These data suggest a potential functional impact for Teneurin mutations, which might contribute to tumorigenesis. The authors further noted that mutations in TENM4 and PIM1 were mutually exclusive. Interestingly, the oncogenic PIM1 serine-threonine kinase is essential for MYC-mediated tumorigenesis in triple-negative breast cancer cells, where knock-out of both genes was synthetic lethal (Horiuchi et al., 2016). Since Teneurin-1 did mediate MYC-RHOA-induced responses in glioblastoma (Talamillo et al., 2017), additional Teneurins might operate through a similar mechanism. Assuming that PIM1 and Teneurins both act through a MYC-dependent pathway, co-selection of mutations might not provide an additional advantage to tumor cells, explaining the lack of concurrent variants in these genes.

# The Mutation Profile of Teneurin Genes

Based on the above data, Teneurins appear to be a frequent target for somatic mutation in lymphomas, and the phenotypic consequence of singe nucleotide changes is evidenced from genetic conditions associated with germline changes. To assess if similar findings apply to other tumors, we performed gene-based searches for all Teneurins in The Cancer Genome Atlas (TCGA, https://cancergenome.nih.gov/). Teneurin mutations recorded for the entire TCGA cohort are represented in **Figure 2A**, sorted according to their subtype. The mutation spectrum of known oncogenes (AKT1, PIK3CA, BRAF) and tumor suppressor genes (TP53, BRCA1, CDKN2A) is included for comparison. Over 750 variants could be retrieved for each Teneurin gene. As evidenced by the graphs, Teneurin missense changes clearly predominate and their mutation spectra do not resemble those of TP53 or CDKN2A. Nonsense variants were infrequent, in analogy to germline findings in essential tremor patients (Houle et al., 2017). When the mutation subtype frequency was represented in a Euclidean-distance cladogram (**Figure 2B**), Teneurin genes clustered together and their mutation profiles were closer to that of AKT1. However, the next association level was with BRAF and BRCA1, an oncogene and tumor suppressor gene, respectively. According to this distribution, the mutation spectrum of Teneurins places them closer to oncogenes, but a less frequent tumor suppressive pattern cannot be excluded. Functional assessment will be required to assign a phenotypic impact to each variant, in addition to predictive algorithms of pathogenic potential as those applied for heritable conditions. An open question concerns the presence of frequent synonymous variants in the germline (Houle et al., 2017) and in tumors, assumed to represent genetic polymorphisms with restrained phenotypic effects. However, a recent bioinformatics analysis suggested that oncogenes, but not tumor suppressors, are affected by an excess of synonymous somatic mutations in tumors (Supek et al., 2014). Further, selection of synonymous mutations was gene and tumor-specific and targeted evolutionary conserved residues. The data suggested that synonymous variants could

numbers were 97 for AKT1, 212 for BRAF, 385 for BRCA1, 346 for CDKN2A, 404 for PIK3CA, and 1000 for TP53. (A) Waffle charts describing the proportion of each mutation category. Each square represents a frequency of 1%. Waffle charts at the bottom were calculated from TCGA through the same procedure and are included for comparison. (B) Euclidean-distance cladogram grouping genes based on their mutation type frequency distribution was constructed using the R package hclust.

alter exonic splicing elements, leading to differential exon usage. As reported recurrently, Teneurins are subjected to diverse alternative splicing patterns throughout species (Tucker et al., 2001, 2012; Lossie et al., 2005; Silva et al., 2011; Graumann et al., 2017; Berns et al., 2018; Jackson et al., 2018; Li et al., 2018a), and the sequences that define the splicing system might be altered by missense but also synonymous changes. If applied to Teneurins, the reported results appear to favor an oncogenic role driven by missense and perhaps synonymous single nucleotide mutations.

# CONCLUDING REMARKS

In recent investigations, exciting discoveries have revealed the amazing level of complexity that surrounds the function of Teneurins. These large proteins present different domains enabling a range of biological functions from adhesion to cell signaling. As could be predicted form their evolutionary conservation, genetic changes are demonstrating phenotypic consequences that underlie different heritable conditions. Reviewing the available literature and retrieving information from cancer databases, we could now gather evidence that strongly supports a role of Teneurins in tumorigenesis. Clearly, these genes are affected by tumor-specific changes through mechanisms expected for validated cancer genes, including chromosomal alterations, somatic mutations, and aberrant expression patterns. The inherent function of Teneurins as signaling molecules consistently involves well-established cancer pathways, such as NOTCH, WNT, and the NRG/ERBB axis. Further, their essential function as adherence molecules can alter processes related to cell adhesion, migration and invasion, and impact on cell plasticity by modulating cell differentiation and transition between epithelial and mesenchymal phenotypes. Accordingly, a prognostic association of Teneurin expression with patient survival has been demonstrated for various cancer types. Based on current molecular evidence, it seems highly probable that Teneurins might exhibit an oncogenic contribution to tumor initiation, growth and progression, although a dual function including tumor suppressive roles cannot be excluded at present. A major issue will be the validation of the pathogenic impact of Teneurin somatic changes, which will be challenging considering Teneurin gene lengths, their complex pattern of alternatively spliced species, and their tissue-specific relevance. The number of somatic variants in tumors is large and most seem to occur at low allele frequencies. The rising volume of omics data available through patient-based repositories will indubitably provide an important means to address some of these challenges, and aid to the validation of Teneurins as representative cancer genes.

# AUTHOR CONTRIBUTIONS

AZ performed literature and data searches, retrieved data from repositories, created tables, and wrote the manuscript. BR-J retrieved and analyzed data from repositories, created figures, and critically reviewed the manuscript.

# FUNDING

This work was supported by grant CONICYT FONDEQUIP EQM150093.

# REFERENCES


identifies a new susceptibility locus near ODZ4. Nat. Genet. 43, 977–983. doi: 10.1038/ng.943


discovery of novel translocations in head and neck squamous cell carcinoma. Head Neck 35, 831–835. doi: 10.1002/hed.23037


metastasis and survival in multiple solid cancers. PLoS Genet 4:e1000129. doi: 10.1371/journal.pgen.1000129


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2018 Rebolledo-Jaramillo and Ziegler. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Teneurin C-Terminal Associated Peptide (TCAP)-1 and Latrophilin Interaction in HEK293 Cells: Evidence for Modulation of Intercellular Adhesion

Mia Husic´ 1 , Dalia Barsyte-Lovejoy <sup>2</sup> and David A. Lovejoy <sup>1</sup> \*

<sup>1</sup> Department of Cell and Systems Biology, University of Toronto, Toronto, ON, Canada, <sup>2</sup> Structural Genomics Consortium, University of Toronto, Toronto, ON, Canada

#### Edited by:

Vance L. Trudeau, University of Ottawa, Canada

#### Reviewed by:

Timothy Mosca, Thomas Jefferson University, United States Nibaldo C. Inestrosa, Pontificia Universidad Católica de Chile, Chile

> \*Correspondence: David A. Lovejoy david.lovejoy@utoronto.ca

#### Specialty section:

This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Endocrinology

Received: 13 August 2018 Accepted: 14 January 2019 Published: 01 February 2019

#### Citation:

Husic M, Barsyte-Lovejoy D and ´ Lovejoy DA (2019) Teneurin C-Terminal Associated Peptide (TCAP)-1 and Latrophilin Interaction in HEK293 Cells: Evidence for Modulation of Intercellular Adhesion. Front. Endocrinol. 10:22. doi: 10.3389/fendo.2019.00022 The teneurins are a family of four transmembrane proteins essential to intercellular adhesion processes, and are required for the development and maintenance of tissues. The Adhesion G protein-coupled receptor (GPCR) subclass latrophilins (ADGRL), or simply the latrophilins (LPHN), are putative receptors of the teneurins and act, in part, to mediate intercellular adhesion via binding with the teneurin extracellular region. At the distal tip of the extracellular region of each teneurin lies a peptide sequence termed the teneurin C-terminal associated peptide (TCAP). TCAP-1, associated with teneurin-1, is itself bioactive, suggesting that TCAP is a critical functional region of teneurin. However, the role of TCAP-1 has not been established with respect to its ability to interact with LPHN to induce downstream effects. To establish that TCAP-1 binds to LPHN1, a FLAGtagged hormone binding domain (HBD) of LPHN1 and a GFP-tagged TCAP-1 peptide were co-expressed in HEK293 cells. Both immunoreactive epitopes were co-localized as a single band after immunoprecipitation, indicating an association between the two proteins. Moreover, fluorescent co-labeling occurred at the plasma membrane of LPHN1 over-expressing cells when treated with a FITC-tagged TCAP-1 variant. Expression of LPHN1 and treatment with TCAP-1 modulated the actin-based cytoskeleton in these cells in a manner consistent with previously reported actions of TCAP-1 and affected the overall morphology and aggregation of the cells. This study indicates that TCAP-1 may associate directly with LPHN1 and could play a role in the modulation of cytoskeletal organization and intercellular adhesion and aggregation via this interaction.

Keywords: TCAP, teneurin, latrophilin, LPHN, GPCR, receptor-ligand interaction, peptides, adhesion

# INTRODUCTION

The teneurins are a family of type II transmembrane proteins critical for the development and maintenance of the central nervous system in both vertebrates and invertebrates. Vertebrates contain four paralogous teneurins (teneurin-1 through -4), each of which are 2,500–2,800 residues in length and are comprised of numerous multifunctional domains involved in adhesion, cytoskeletal binding, and other protein-protein interactions (1–5). In both vertebrates and invertebrates, the teneurins have been implicated in the formation of filopodia and outgrowth of neurites, as well as neuronal mapping, axonal path-finding, and increased cell-cell adhesion (6–15).

At the distal end of its extracellular carboxy terminus, each of the teneurins contains a conserved peptide sequence named the teneurin C-terminal associated peptide (TCAP) (16, 17). The four vertebrate TCAP paralogues have notable primary structure similarity to that of corticotropin-releasing factor (CRF), calcitonin and most other Secretin G protein-coupled receptor (GPCR) ligands (16–18). TCAP-1, the most studied of the vertebrate TCAP paralogues to date, is known to be expressed as an independent mRNA that yields a 15 kDa pro-TCAP-1 peptide which may then be processed into the mature 4.7 kDa TCAP-1 (19). The mature TCAP-1 peptide has a number of biological actions independent from teneurin-1. TCAP-1-treated murine immortalized hippocampal and hypothalamic cells show a marked increase in neurite and filopodia production, which is associated with an increased expression of the cytoskeletal components β-actin and β-tubulin. TCAP-1 treatment also increases neurite sprouting, axon fasciculation, and modifies dendrite arborization (19, 20). Similar observations have been made in vivo, with CA1 hippocampal neurons exhibiting greater dendritic spine density upon treatment with TCAP-1 (21). TCAP-1 regulates these cytoskeletal changes through activation of the MEK/ERK-1/2 signaling pathway, ultimately leading to modulation of microtubule formation and actin polymerization (22). Additionally, TCAP-1 administration in various rodent models significantly alters anxiety- and stressrelated behaviors in acoustic startle response, elevated plus maze, and cocaine-reinstatement studies, further cementing its neuromodulatory roles (21, 23–26). Yet despite the high efficacy TCAP-1 shows both in vitro and in vivo, the precise mechanism by which these actions occur is not well-understood.

Recent studies indicate that the teneurins are endogenous ligands of Adhesion GPCR subfamily L/latrophilin (ADGRL), or simply, latrophilin (LPHN) (8, 27, 28). The LPHNs are a family of three Adhesion GPCRs found in both vertebrates and invertebrates, and, until the discovery of their interaction with teneurin, were considered orphan receptors, as their only prior known ligand was the exogenous α-latrotoxin, the toxic component of black widow spider venom (29). The binding between teneurin and LPHN1 involves the teneurin C-terminal region and at least the lectin-like domain, olfactomedin-like domain, and the serine-threonine rich region of the LPHN1 extracellular tail, which come together to form a trans-synaptic complex that mediates neuronal cell adhesion and signaling (8, 27, 28, 30). In rat hippocampal cell isolates, LPHN1 and teneurin-2 co-occur at synapses, with LPHN1 primarily being located on the presynaptic membrane, whereas teneurin-2 is primarily found post-synaptically (8). LPHN1-expressing Nb2a neuroblastoma cells preferentially aggregate with those expressing teneurin-2, with the proteins co-localizing at points of cell contact to interact specifically across cell-cell junctions (27). Moreover, co-cultures of HEK293 cells expressing LPHN1 with those expressing either teneurin-2 or teneurin-4 show increased cell aggregate formation, indicating greater adhesion between adjacent cells (8).

Direct interaction between LPHN1 and TCAP-1 has not yet been ascertained; however, both structural and functional evidence suggests that this interaction is likely. Although the extracellular region of teneurin contains a β-barrel formation that partially encapsulates the teneurin C-terminus, the TCAP sequence-containing tip of this region emerges from the barrel and is exposed to the extracellular environment (30), placing TCAP in a favorable position to interact with LPHN1. Furthermore, the three vertebrate LPHN paralogues each contain an extracellular hormone binding domain (HBD) with high sequence similarity to the peptide-binding region of many Secretin GPCRs, such as CRF receptors 1 (CRFR1) and 2 (CRFR2), and are thought to be involved in LPHN ligand binding (31, 32). As the four TCAP paralogues contain sequence similarity to CRF, it is possible that an interaction between TCAP-1 and LPHN1 may occur through this LPHN domain. Therefore, in this study we examined whether TCAP-1 and LPHN1 can interact directly at the LPHN1 HBD, and if TCAP-1 co-localizes with a labeled variant of LPHN1 in HEK293 cells. The results of these studies suggest that TCAP-1 can interact directly with LPHN1 to modify cell-to-cell adhesion and cytoskeletal organization.

# MATERIALS AND METHODS

#### Cell Culture

HEK293 cells were grown on 10-cm culture plates in 12 ml McCoy's 5A medium containing L-glutamine (Gibco) supplemented with 10% heat-inactivated fetal bovine serum (FBS; Invitrogen), 100 U/ml penicillin and 100µg/mL streptomycin. Cells were maintained at 70–90% confluency at 37◦C in a humidified CO<sup>2</sup> incubator during growth. To passage, the cells were rinsed with phosphate buffered saline (PBS) prior to treatment with 3 mL trypsin for 1–2 min. Four mL of fresh medium was then added to inactivate the trypsin and the cells were centrifuged at 16,000 rpm for 3 min. The supernatant was aspirated, and the cell pellet re-suspended in 5 mL of fresh medium. Cells were re-seeded at 100,000 cells per 10 cm plate in 12 mL of fresh culture medium and allowed to grow for 2–3 days. Prior to all experimentation, cells were grown to 50–80% confluency in 6-well culture plates and serum-starved for 3 h in 2 mL of culture medium without FBS but with penicillin and streptomycin.

# Sequence Alignments

The HBD amino acid sequences of the murine LPHN1-3 (acc#: NP\_851382.2, NP\_001074767.1, NP\_941991.1, respectively) and Secretin GPCRs CRFR1 (acc#: NP\_031788.1), CRFR2 (acc#: NP001275547.1), calcitonin receptor (CALCR; acc#: NP\_031614.2) and calcitonin gene-related peptide receptor (CGRPR; acc#: NP\_061252.2) were obtained from the National Center for Biotechnology Information protein database. A multiple sequence alignment of the HBDs was then performed using the multiple sequence comparison by log-expectation (MUSCLE) alignment tool ver. 3.8 (33).

#### Constructs and Transfection

To assess the interaction between LPHN1 and TCAP-1, a LPHN1 construct based on a splice variant of murine LPHN1 (acc#: XM\_006531122.2) was expressed in HEK293 cells via lentiviral transfection (**Figure 1A**). LPHN1 cDNA from the Mammalian Gene Collection clone BC085138 was PCR-amplified using the forward primer GATCACCGGTGCCACCATGGCCCGCT TGGCTGCA and the reverse primer GATCGTCGACTCAGG AGTCACCCCAAGGGA containing AgeI and SalI restriction endonuclease sites, respectively. The PCR product was isolated via gel electrophoresis, purified, and digested using AgeI and SalI. This was subsequently sub-cloned into a pRRL vector (original vector from Addgene plasmid 12252, modified to contain a CMV promoter, multiple cloning sites XbaI, BamHI, AgeI, SalI, and an IRES-puromycin cassette). A FLAG-tag sequence was inserted after the LPHN1 signal peptide (SP) sequence (MARLAAALWSLCVTTVLVTSATQGL) using the Q5 Site-Directed Mutagenesis Kit (New England BioLabs). The pRRL SP-FLAG-LPHN1-IRES-puromycin vector was then co-transfected with pMD2.G (Addgene, 12259), pRSV.REV (Addgene, 12253), and pMDLg/pRRE (Addgene, 12251) plasmids into HEK293 cells. Virus particles were harvested after 72 h and used to transfect HEK293 cells (HEK-LPHN1-S). Wild-type HEK293 cells (HEK-WT) were used as a control cell line, while HEK293 cells transfected with a vector containing puromycin only (HEK-Puro) were used as a transfection control. HEK293 cells were then puromycin-selected and the resulting cell populations were verified for LPHN1 expression using immunocytochemistry and western blotting analysis.

To assess the binding of TCAP-1 with the LPHN1 HBD, two constructs encompassing the LPHN1 HBD region with an added N-terminal FLAG tag were designed (**Figure 1A**, HBD constructs). The constructs spanned LPHN1 residues V444 to either C579 or E634, and, in both cases, included part of the GPCR autoproteolysis-inducing (GAIN) domain of LPHN1 (**Figure 1B**). The constructs were transiently co-expressed in HEK293 cells along with either a green fluorescence protein (GFP)-tagged mouse pro-TCAP-1 construct (GFP-pro-mTCAP-1) or a GFP-tagged mature TCAP-1 construct (GFP-mTCAP-1), the amino acid sequences of which were determined based on the TCAP-1 mRNA transcript identified by Chand et al. (19). AntiFlag (Sigma) antibody was used for immunoprecipitation. All transient transfections were performed using the X-treme Gene system (Roche).

# LPHN1 HBD-TCAP-1 Immunoprecipitation Assay

Immunoprecipitation assays were performed to assess the interaction of LPHN1 HBD and TCAP-1. Constructs of the LPHN1 HBD region with an added FLAG tag (**Figure 1**) were designed and transiently expressed in HEK293 cells along with either GFP-mTCAP-1 or GFP-pro-mTCAP-1. The degree of expression of the LPHN1 HBD constructs was confirmed using western blot (data not shown). To determine if TCAP-1 interacts with the HBD, first the HBD construct proteins were isolated using an anti-FLAG antibody. The HBD constructs were then precipitated through a series of wash and centrifugation steps and eluted. The eluate was then resolved via western blot and probed for presence of the GFP tag on TCAP-1 to see if TCAP-1 interacts with either HBD construct.

# Western Blotting

For cell lysate collection, cells were washed once with ice cold PBS and then lysed on ice for 5 min using radio-immunoprecipitation assay buffer (RIPA; Cell Signaling Technology) with added phenylmethylsulfonyl fluoride (PMSF) protease inhibitor. The lysates were then harvested and centrifuged at 14,000 rpm and 4 ◦C for 20 min to remove any debris. The resulting supernatant was collected for further use in western blot analysis.

To determine the protein concentrations of collected cell lysate samples, a Pierce bicinchoninic acid (BCA) protein assay (Thermo Scientific) was performed according to the manufacturer's instructions. Briefly, standards containing known concentrations of bovine serum albumin (BSA) ranging from 0 to 2,000µg/ml were prepared. 25µL of the standards and previously collected cell lysates were added to individual wells of a 96 well-plate. 200µL of working reagent was then added to each well, and the plate was put on a shaker for 30 s to allow the samples and reagent to sufficiently mix. The plate was then incubated at 37◦C for 30 min. Absorbance levels of the standards and samples were measured at 562 nm using a Spectramax Plus Microplate Reader (Molecular Devices). Absorbance values of the standards were then used to create a standard curve from which protein concentrations of cell lysates could be interpolated. Once lysate protein concentrations were determined, all lysates were normalized to provide an equal protein concentration across all samples prior to proceeding with Western blot analysis. Lysates were stored at −20◦C.

The expression of LPHN1 in HEK293 cells was determined by western blot. 15µL of sample were combined with Tricine sample loading buffer (Bio-Rad) containing 2% βmercaptoethanol. All samples were resolved via 12% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) at 100 V for 1 h. The peptides were then electro-transferred onto a Hybond ECL nitrocellulose blotting membranes (Amersham) at 100 V for 75 min. The membranes were washed 3x for 10 min with PBS and blocked in a 5% BSA-PSBT solution (5% BSA w/v dissolved in PBS with 0.2% Tween <sup>R</sup> 20) on a shaker for 1 h at RT. They were then incubated in 5% BSA-PBST with goat polyclonal LPHN1 primary antibody (Santa Cruz) at a 1:1,000 dilution overnight at 4◦C with gentle agitation. Immunoprecipitation immunoblotting and GFP detection was performed using mouse anti-GFP (1:5,000, Clontech). The following day, the membranes were given 3x 10 min washes with PBST and incubated for 1 h at RT in 1% milk PBST (1% dehydrated milk w/v dissolved in PBST) containing horseradish peroxidase-linked donkey antigoat secondary antibody (Santa Cruz) at a 1:7,500 dilution or IR800 conjugated anti-mouse secondary antibody (LiCor). The membranes were then washed 3x for 10 min with PBST prior to a 1 min incubation in chemiluminescence reagent (ECL, Amersham). For protein detection, the membranes were exposed onto ECL Hyperfilm (VWR) for 0.5–6 min. IR800 signal was visualized on Odyssey scanner (LiCor). Western blot gel images

peptide (SP), a lectin domain (LEC), an olfactomedin-like domain (OLF), a hormone binding domain (HBD) and a GPCR autoproteolysis inducing (GAIN) domain, a GPCR proteolytic site (GPS), followed by a 7-transmembrane region (TMR) and an intracellular (IC) tail. The LPHN1 construct expressed in HEK293 cells, LPHN1-S, contains a FLAG tag downstream of the SP and a truncated intracellular (IC) tail. Constructs used for the co-immunoprecipitation assay of the LPHN1 HBD and TCAP-1 and their constituent domains are also indicated (HBD constructs). Each construct contains an N-terminal FLAG tag. The range of amino acid residues surrounding the HBD domain in each construct is indicated in the construct names. (B) Sequence of the LPHN1 region used for co-immunoprecipitation of TCAP-1. Construct V444-Q579 is composed of Valine 444 (\*) to Glutamine 579 (\*\*); construct V444-E634 is composed of residues Valine 444 (\*) to Glutamic acid 634 (\*\*\*). Gray highlight: HBD; Underline: GAIN domain.

were quantified using Fiji software (34), and statistical analysis was performed using GraphPad Prism 7.

#### Peptide Synthesis

Mouse TCAP-1 (mTCAP-1) was synthesized at 95% purity using f-moc-based solid phase synthesis. mTCAP-1 with an arginine (R) to lysine (K) substitution at position 37 (K37 mTCAP-1) was synthesized as previously described (17). K37-mTCAP-1 was further tagged with Fluorescein (FITC) (Thermo Scientific) using N-hydroxysuccinimide according to the manufacturer's instructions. Briefly, the K37-mTCAP-1 was solubilized in borate buffer while the Fluorescein dye was dissolved in dymethylformamide (DMF). A 20 fold molar excess of the dye was added to the peptide and the solution was incubated at RT for 1 h in the dark. The solution was then passed through polyacrylamide desalting columns (Thermo Scientific) for purification and 8 fractions were collected. Protein absorbance of the fractions was measured at 280 and 495 nm using a Spectramax Plus Microplate Reader (Molecular Devices) to determine which fractions contained the highest protein content. The fractions with the highest absorbance readings, indicating the highest FITC-tagged K37-mTCAP-1 (FITC-K37-mTCAP-1) content, were then combined and stored as aliquots at −20◦C.

#### Immunocytochemistry

For all immunocytochemical (ICC) analysis, HEK-WT, and HEK-LPHN1-S cells were first grown on poly-D-lysine coated cover slips to 50–80% confluency. To confirm successful LPHN1 transfection, cells were first fixed onto the cover slips via treatment with 4% paraformaldehyde (PFA) for 20 min. They were then washed 3x with PBS, permeabilized with 0.3% Triton X-100 (Sigma Aldrich) and washed again 3x with PBS prior to blocking with PBS containing 10% v/v normal goat serum (NGS) for 1 h. The cells were incubated for 1 h with Cy3-tagged Flag antibody (Sigma Aldrich), given 3x PBS washes and mounted onto microscope slides using VECTASHIELD Mounting Medium with DAPI (Vector Laboratories).

LPHN1 and FITC-K37-mTCAP-1 co-localization studies were done by first incubating HEK-WT and HEK-LPHN1-S cells in culture medium containing 20 nM HEPES buffer (pH 7.4), 0.1% BSA and FITC-K37-mTCAP-1 at a 1:400 dilution for 12 h at 4◦C. The cells were then given 3 washes with cold culture medium, fixed with 4% PFA and washed 3x with ice cold PBS. Subsequently, the cells were blocked for 1 h at RT with PBS containing 3% w/v BSA and incubated with Cy3-tagged FLAG antibody (Sigma Aldrich) for 1 h. Following this, they were washed 3x with 3% BSA blocking solution and once with water, and then finally mounted onto microscope slides as described above.

For morphology studies involving cell membrane staining with wheat germ agglutinin (WGA), cells were first fixed for 10 min using 4% PFA, and then washed 3x with Hank's Balanced Salt Solution (HBSS; Gibco) prior to incubation with HBSS containing WGA (Invitrogen) at a 1:1,000 dilution for 10 min. The cells were washed 2x with HBSS, permeabilized using 0.2% Triton X-100 for 10 min, washed 3x with PBS and mounted onto microscope slides as described above.

To examine cytoskeletal morphology, cells were either first treated with 100 nM TCAP-1 or vehicle for 1 h or immediately washed 3x with PBS and incubated in a solution containing 1 mL of 4% PFA, 10 µL Triton X-100, and 15 µL of the filamentous actin (f-actin) probe Alexa Fluor 594 Phalloidin (Life Technologies) for 10 min. The cells were then washed 3x with PBS and mounted onto microscope slides as described above.

All cell imaging was done using confocal microscopy (TCS SP8, Leica Microsystems) with 40x, 63x, or 100x oil immersion objectives. Image acquisition settings were calibrated to control cell groups (non-TCAP-1-treated HEK-WT cells). After acquisition, the images were converted into JPEG format for further analysis. HEK293 nuclear height was measured via Zstacking from the base of the nucleus to just beyond the top of the nucleus, to acquire a measurement of the full organelle.

# Digital Image Analysis

Immunofluorescence intensity of all digital images of cells was analyzed using Fiji software (34). For whole cell size analysis, each cell that was completely visible within a merged image displaying WGA and DAPI staining as well as the differential interference contrast (DIC) was digitally analyzed. Every cell that was clearly visible in an ICC image was individually isolated using the Fiji freehand selection tool, and the area and perimeter of the cells were obtained and averaged. To analyze nuclear size, DAPI nuclear stain images were used. First, a color threshold was set to create multiple regions of interest (ROIs) based on blue pixel intensity. This allowed for simultaneous isolation of multiple nuclei within a single image. Any single ROIs consisting of more than one nucleus or of cells in the process of mitosis (as indicated by anaphase-like chromosomal arrangement) were discarded and the remaining ROIs were measured for their perimeter and area. Nuclei that were not captured by this method were individually isolated using the Fiji freehand selection tool and their area and perimeter were measured using the same method as for whole cell measurements.

FITC-K37-mTCAP-1 uptake by cells was quantified by immunofluorescence intensity, specifically by examining green pixel intensity histograms of confocal microscopy images for HEK-WT, HEK-Puro, and HEK-LPHN1-S cells. Analysis was done by discarding the first 20 intensity values on the histogram, as these corresponded to black pixels, indicating no green FITC tag signal. For each image examined, the total number of cells per image was counted, and the total number of green pixels with an intensity of 20–255 was obtained. This number of pixels was then divided by the number of cells in that picture as a way to account for differences in cell count per image. These values were used for further statistical analysis by GraphPad Prism 7. Cytoskeletal differences between HEK-WT and HEK-LPHN1-S cells were quantified in the same manner using Phalloidin stain images for each cell group and red pixel intensity histograms.

### Statistical Analysis

All results are represented as a mean ± standard error of the mean (SEM). An a priori hypothesis of p < 0.05 was utilized for all analyses. The data was analyzed with GraphPad Prism 7 using either a two-tailed t-test or one-way or two-way analyses of variance (ANOVA) with a Tukey's post hoc test. Mean values were obtained from a minimum of 3 independent repeats of an experiment, where a single repeat refers to cells grown in a single well of a 6-well plate. For digital analysis of ICC images, representative photos of each repeat were analyzed. Cell height measurements were taken from 4 distinct regions of each slide cells were mounted onto, where 4 cells per region were measured for a total of 16 measurements per slide (one repeat). Data was considered statistically significant if p < 0.05 (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001).

# RESULTS

# Comparison of LPHN and Secretin GPCR HBD Amino Acid Sequences

The putative HBD region of LPHN1 showed about 30% identity at the amino acid level with the HBD regions of the calcitonin and CRF receptors (**Figure 2A**), confirming the homology of this domain within this receptor group. This was also reflected by conserved residues at LPHN1 positions 475 (C), 485 (W), 492 (G), 499 (C), 500 (P), 511 (C), 516 (G), and 518 (W). With respect to LPHN, the CRF receptors showed a slightly higher degree of identity than the calcitonin receptors, noted by the conservation of residues at LPHN1 positions 598 (P), 526 (S), and 528 (C). Furthermore, at least 50% identity was observed between the 64 residue HBD sequences of the three LPHN paralogues themselves (**Figure 2B**).

# TCAP-1 Interaction With a LPHN1 HBD Cassette

To determine if TCAP-1 interacts directly with the LPHN1 HBD, FLAG-tagged LPHN1 HBD constructs V444-Q579 and V444- E634 (**Figure 1**) were transiently expressed in HEK293 cells along with GFP-pro-mTCAP-1 and GFP-mTCAP-1 peptides. The HBD constructs were then used as bait proteins in a co-immunoprecipitation (co-IP) assay to determine if either the pro-TCAP-1 or the mature TCAP-1 peptide interacts with the LPHN1 HBD (**Figure 3**). First, the expression of both GFP-pro-mTCAP-1 and GFP-mTCAP-1 in HEK293 cells were determined (**Figure 3**, inputs). Western blot bands, at ∼40 and 30 kDa, corresponding to the sizes of GFP-pro-mTCAP-1 and GFP-mTCAP-1, respectively, were observed, indicating strong expression of these peptides in their respective cell lines. The results of the co-IP assay (**Figure 3**, IPs) showed no bands at 40 kDa, corresponding to GFP-pro-mTCAP-1, when either the V444-Q579 or the V444-E634 construct was used as a bait protein. However, bands as 25 and 50 kDa were observed with both constructs (IgG light and heavy chains; data not shown). In contrast to these findings, a band at 30 kDa, corresponding to GFP-mTCAP-1, was observed when the V444-E634 construct was used as bait (**Figure 3**, IPs). A fainter 30 kDa band could also be seen when the V444-Q579 construct was used. Again, additional bands at 25 and 50 kDa were observed. These results suggest that a stronger affinity of the TCAP-1 construct occurred when a larger proportion of the GAIN domain was included. Control experiments in which no anti-FLAG antibodies were used to precipitate the HBD constructs were also performed where the eluates showed no detectable bands for either GFPpro-mTCAP-1 or GFP-mTCAP-1 (**Figure 3**, IPs, "no Ab" lanes).

# Over-Expression of LPHN1 Constructs in HEK293 Cells

The LPHN1-S construct, containing an N-terminal FLAG tag and a truncated intracellular tail, was expressed in HEK293 cells via lentiviral infection to create a cell line in which the interaction between TCAP-1 and LPHN1 could be examined. Western blot analysis using anti-LPHN1 antibodies was performed to confirm successful over-expression of the LPHN1-S construct in HEK-LPHN1-S cells, and to determine the endogenous degree of LPHN1 expression in HEK-WT cells (**Figures 4A,B**). A band at 120 kDa was detected in HEK-LPHN1-S cells, but not in HEK-WT cells. It should be noted that endogenous expression of the LPHN2 and LPHN3 isoforms in HEK-WT and HEK-LPHN1-S cells was not examined in this study.

To observe the pattern of LPHN1 expression in these cells, Cy3-tagged anti-FLAG antibodies were used to label the LPHN1- S protein for confocal microscopy imaging (**Figure 4**). HEK-WT cells showed no detectable FLAG signal (**Figure 4E**). HEK-Puro vector control cells expressing just a puromycin resistance gene without LPHN1-S also showed no detectable FLAG signal (**Figure 4I**). The HEK-LPHN1-S cells, however, showed a strong FLAG signal that was localized primarily to the cell membrane (**Figures 4M,N**, white arrow).

# Binding of TCAP-1 to LPHN1 in HEK293 Cells

Immunocytochemistry analysis was performed using HEK-WT, HEK-Puro, and HEK-LPHN1-S cells treated with FITC-K37-mTCAP-1 to observe the degree of TCAP-1 uptake in each cell type and to determine if TCAP-1 co-localizes with LPHN1 (**Figure 5**). FITC-K37-mTCAP-1 and LPHN1 were found exclusively at the cell membrane, with largely overlapping localization patterns (**Figures 5I,L–O** arrows); however, regions of the cell membrane with just FITC-K37-mTCAP-1 or just LPHN1 fluorescence were also observed (**Figure 5II**). HEK-WT and HEK-Puro cells had little or low FITC-K37-mTCAP-1 uptake. In contrast, considerable FITC-K37-mTCAP-1 uptake was present in HEK-LPHN1-S cells. The degree of FITC-K37 mTCAP-1 uptake was quantified as a function of green pixels per cell (**Figures 5I,P**). Treatment with FITC-K37-mTCAP-1 yielded a significant signal increase (green pixels per cell) in HEK-LPHN1-S cells compared to HEK-WT and HEK-Puro cells (WT: 16.491 ± 0.942 pixels/cell; Puro: 12.790 ± 2.536 pixels/cell; LPHN1-S: 49.498 ± 3.042 pixels/cell; p < 0.0001).

# Changes in Cell Size Upon LPHN1 Over-Expression

To examine the effects of over-expressing LPHN1 on HEK293 cell morphology, HEK-WT and HEK-LPHN1-S cells were labeled with the cell membrane marker WGA and imaged using confocal microscopy (**Figures 6A–H**). The HEK-LPHN1-S cells appeared to be smaller in diameter and clustered closer together than the HEK-WT cells. However, this observation was due to the initial 2-dimensional nature of this analysis. Subsequently, the morphology of these cells was characterized by quantifying the whole cell and nuclear area and perimeter (**Figures 6I–L**). HEK-WT cells had an average whole cell area of 293.6 ± 6.2 µm<sup>2</sup> , whereas HEK-LPHN1-S cells had a whole cell area that was 27.2% smaller at 213.7 ± 16.4 µm<sup>2</sup> (p < 0.05; **Figure 6I**). Similarly, the HEK-LPHN1-S whole cell perimeter of 59.6 ± 2.3µm was 17% smaller than the HEK-WT cell perimeter of 71.8 ± 0.9µm (p < 0.01; **Figure 6J**). The HEK-LPHN1-S cell nuclear area of 131.8 ± 3.0 µm<sup>2</sup> was 9.6% smaller than that of the HEK-WT

are indicated in pink, and homologous replacements are indicated in yellow.

cells at 146.0 ± 1.4 µm<sup>2</sup> (p < 0.05; **Figure 6K**). Finally, the HEK-LPHN1-S nuclear perimeter of 44.1 ± 0.3µm was 4.6% smaller than HEK-WT nuclear perimeter of 46.3 ± 0.4µm (p < 0.05; **Figure 6L**).

### Changes in Cytoskeletal Organization Upon LPHN1 Over-Expression

Next, the f-actin content of HEK-WT and HEK-LPHN1-S cells was assessed as a biomarker to further characterize the morphological differences between these cell types. The factin cytoskeleton was fluorescently stained using Phalloidin (**Figure 7**). HEK-WT cells showed a strong degree of Phalloidin labeling, indicating a large amount of f-actin. HEK-WT cells were much larger than HEK-LPHN1-S cells and had more f-actin projections extending from them compared to HEK-LPHN1-S cells, which had little to no projections of the same morphology (**Figures 7E,J**, white arrows). Similarly, the HEK-WT cells appeared to be more spread out, with more f-actin between individual cells compared to the HEK-LPHN1-S cells, which had a more clustered appearance with little f-actin between individual cells.

The amount of f-actin present in each cell type was further quantified as a function of red pixels per cell, in which the number of red pixels corresponds to the amount of Phalloidin-bound f-actin present per cell (**Figure 7K**). HEK-LPHN1-S cells had 52% fewer red pixels per cell than HEK-WT cells (WT: 3818.327 ± 144.874 pixels/cell; LPHN1-S: 1834.192 ± 166.365 pixels/cell; p < 0.01), indicating a significantly lower level of f-actin.

FIGURE 3 | The mature TCAP-1 peptide interacts with the HBD and a partial GAIN domain of LPHN1. (HBD constructs) LPHN1 HBD constructs were successfully expressed in HEK293 cells. (Inputs) Input lanes indicate strong presence of GFP-pro-mTCAP-1 or mature TCAP-1 in cell lysates prior to immunoprecipitation with HBD constructs V444-Q579 and V444-E634. Expected band size of GFP-pro-mTCAP-1 and GFP-mTCAP-1 are 40 and 30 kDa, respectively (black arrows). (IPs) Immunoprecipitation lanes show western blot resolution of corresponding eluates from input lanes. "No Ab" indicates that no anti-FLAG antibody was used to for immunoprecipitation of the HBD construct, serving as a negative control. No bands corresponding to GFP-pro-mTCAP-1 were isolated with either of the HBD constructs used for IP. A faint band at ∼30 kDa, corresponding to GFP-mTCAP-1, is present when HBD V444-Q579 was used for IP. A stronger band of the same size is seen when IP was performed using HBD V444-E634 (red arrow). No bands corresponding to the GFP-pro-mTCAP-1 peptide were observed in the IP.

FIGURE 4 | HEK-LPHN1-S cells strongly express LPHN1-S, whereas HEK-WT and HEK-puro cells do not. (A) Western blot of LPHN1 expression in HEK-WT and HEK-LPHN1-S cells. A band at ∼120 kDa is present in the HEK-LPHN1-S cells, but not in the HEK-WT cells (black arrow). Expected size is 116 kDa. (B) Quantification of LPHN1 protein expression in HEK-WT and HEK-LPHN1 cells. (Mean ± SEM; n = 4; \*\*\*\*p < 0.0001; two-tailed t-test) (C–F) Confocal images of HEK-WT cells. (G–J) Confocal images of HEK-Puro cells. (K–N) Confocal images of HEK-LPHN1-S cells. (C,G,K) DIC image. (D,H,L) DAPI staining of nuclear proteins. (E,I,M) FLAG-tagged LPHN1-S. (F,J,N) Merged images of A–C, G–I, and K–M, respectively. White arrow in M,N indicates FLAG-tagged LPHN1-S expression in HEK-LPHN1-S cells. A strong FLAG signal is seen at the cell membrane of HEK-LPHN1-S cells only, indicating strong expression of the LPHN1-S construct in this region (white arrow). HEK-WT and HEK-Puro cells showed no FLAG signal. Scale bar in all images is 25µm.

all images is 25µm. (P) Average number of green pixels per cell with an intensity ranging from 20 to 255 in HEK-WT, HEK-Puro, and HEK-LPHN1-S cells treated with FITC-K37-mTCAP-1. Pixel intensity range is indicated above graph. HEK-LPHN1-S cells have a significantly higher number of green pixels with intensities of 20–255 per cell, indicating a higher degree of FITC-K37-mTCAP-1 uptake compared to HEK-WT and HEK-Puro cells. No significant differences between HEK-WT and HEK-Puro cells were observed. (Mean ± SEM; n = 5; \*\*\*\*p < 0.0001; one-way ANOVA) II: (A–J) Confocal images of HEK-LPHN1-S cells. (A,F) DIC. (B,G) FITC-K37-mTCAP-1. (C,H) LPHN1-S. (D,I) Merged images of B–C and G–H, respectively. (E,J) Merged images of A–C and F–H, respectively. Scale bar in A–E is 20µm and in F–J is 6µm. Regions of FITC-K37-mTCAP-1 and LPHN1-S co-localization on the cell membrane were predominant (arrows labeled 1); however, areas of only FITC-K37-mTCAP-1 localization (arrows labeled 2) or of only LPHN1-S localization (arrows labeled 3) were also seen present in HEK293-LPHN1-S cells.

# Changes in Cytoskeletal Organization Upon Treatment With TCAP-1

TCAP-1 acts on the MEK-ERK1/2 pathway to induce polymerization of f-actin via activation of p90RSK and filamin A, leading to changes in neuronal cell cytoskeletal organization and morphology (22). If TCAP-1 is a ligand of LPHN1, then it is possible that its actions on this pathway occur through LPHN1 and its associated G proteins. To examine whether TCAP-1 induces cytoskeletal changes through an interaction with LPHN1, HEK-WT, and HEK-LPHN1-S cells were treated with either 100 nM mTCAP-1 or vehicle (control) for 60 min, and the cytoskeletal profile of the cells was observed (**Figure 8**). Vehicle-treated HEK-WT cells showed a strong f-actin signal at the cell perimeter as well as between adjacent cells (**Figure 8C**).

This pattern was not seen in vehicle-treated HEK-LPHN1-S cells, which had a much weaker f-actin signal at the cell perimeter and little to no signal between adjacent cells (**Figure 8M**). mTCAP-1-treated HEK-WT cells showed little differences in f-actin labeling compared to vehicle-treated HEK-WT cells and had very few f-actin projections (**Figure 8H**). In contrast, mTCAP-1 treatment had a strong effect on f-actin expression in HEK-LPHN1-S cells (**Figure 8R**). These cells showed a much greater degree of f-actin labeling compared to those treated with vehicle, with a high amount of f-actin present throughout the cytosol of individual cells as well as between clustering cells. mTCAP-1-treated HEK-LPHN1-S cells also had more f-actin projections extending from them than vehicle-treated cells (**Figure 8T**, white arrows).

The expression of f-actin in these cells was again quantified as a function of the number of red pixels per cell, indicative of the degree of f-actin staining by Phalloidin (**Figure 8U**). No significant difference in red fluorescence was observed between HEK-WT cells treated with mTCAP-1 or vehicle, indicating no difference in their f-actin expression (WT + Veh: 986.706 ± 65.626 pixels/cell; WT + mTCAP-1: 823.586 ± 78.778 pixels/cell; p > 0.05). In contrast, mTCAP-1-treated HEK-LPHN1-S cells showed a 343% increase in red pixels per cell, indicating an increase in f-actin compared to vehicle-treated HEK-LPHN1-S cells (LPHN1-S + Veh: 175.314 ± 22.488 pixels/cell; LPHN1- S + mTCAP-1: 777.063 ± 49.511 pixels/cell; p < 0.0001). Furthermore, vehicle-treated HEK-LPHN1-S cells again showed a significantly decreased expression of f-actin compared to vehicle-treated HEK-WT cells, having ∼82% less red pixels per cell than their wild-type counterpart (WT + Veh: 986.706 ± 65.626 pixels/cell; LPHN1-S + Veh: 175.314 ± 22.488 pixels/cell).

n = 5, \*\*p < 0.01, two-tailed t-test).

# Changes in Cell Morphology Upon Treatment With TCAP-1

Previous studies have indicated that LPHN1 and teneurin form an adhesion complex between adjacent cells, leading to increased formation of cell-cell contacts and aggregates (8). If their total volume is not affected, as cells cluster closer together due to an increase in cell-to-cell points of contact between LPHN1 and teneurin, changes in cell dimensions can be expected to occur, such as decreases in width and length and an increase in height. Similarly, if TCAP-1 is acting on the LPHN1-teneurin adhesion complex, a reversion to wild-type cell dimensions upon mTCAP-1 treatment may be expected. To determine if this is the case in HEK293 cells, the heights of HEK-WT and HEK-LPHN1-S cells upon treatment with 100 nM mTCAP-1 or vehicle for 60 min were measured (**Figure 8V**). For HEK-WT cells, no significant differences in cell height were observed between treatment with mTCAP-1 and vehicle (HEK-WT + Veh: 5.708 ± 0.180µm; HEK-WT + mTCAP-1: 5.976 ± 0.180µm). However, vehicletreated HEK-LPHN1-S cells had a height of 8.113 ± 0.298µm, which was significantly larger than that of HEK-WT cells treated with either vehicle or mTCAP-1 (p < 0.0001). Treatment of HEK-LPHN1-S cells with mTCAP-1 resulted in a 39% decrease in height to 4.968 ± 0.199µm (p < 0.0001). There was no significant difference in height between vehicle-treated HEK-WT cells and mTCAP-1-treated HEK-LPHN1-S cells (HEK-WT + Veh: 5.708 ± 0.180µm; HEK-LPHN1-S + mTCAP-1: 4.968 ± 0.199µm; p > 0.05).

# DISCUSSION

The data presented in this study provides novel evidence that the TCAP-1 region of teneurin-1 associates directly with LPHN1, and, as a diffusible peptide, can modulate cell-to-cell adhesion and cytoskeletal dynamics. Specifically, a GFP-tagged TCAP-1 construct co-immunoprecipitated with a portion of the LPHN1 extracellular domain containing the LPHN1 HBD, indicating affinity between the two. This is the first study to show that an Adhesion-type GPCR binding region has the potential to bind a ligand related to the Secretin family of peptides. When treated with FITC-K37-mTCAP-1, HEK-LPHN1-S cells had a higher level of co-localization of the peptide with LPHN1 than HEK-WT cells. Morphologically, over-expression of LPHN1 modulated cell size and decreased f-actin expression in HEK-LPHN1-S cells relative to the HEK-WT cells. mTCAP-1 treatment had little effect on the morphology of HEK-WT cells, whereas treatment of HEK-LPHN1-S cells resulted in changes to the cytoskeletal organization consistent with previous observations of TCAP-1 function. Together, these studies link the actions of synthetic TCAP-1 described in previous studies with the actions of the teneurin-LPHN complex as well as results reported in recent studies on the structure of teneurin and the LPHNs.

# TCAP-1 Interaction With LPHN1 at the Receptor Hormone Binding Domain

Since their discovery, the TCAP peptides were established to have major sequence identity initially with CRF and calcitonin, and subsequently, to a lesser but still compelling degree, with other members of the Secretin peptide family (16–18). The phylogenetic rationale for this relationship is not clear, currently. However, Secretin peptides are the ligands of Secretin GPCRs, and interact with their respective receptors at the receptor HBD. CRF itself, for example, binds to the HBD of its cognate receptors, CRFR1 and CRFR2, both of which belong to the Secretin family of GPCRs (35–38). Upon their discovery, the LPHNs were initially classified as members of the Secretin GPCR family due to their sequence similarity to the CRFR HBD and transmembrane region, but were later reclassified to the much

FIGURE 8 | T indicate f-actin projections. mTCAP-1 treatment does not cause an increase in f-actin in HEK-WT cells. In HEK-LPHN1-S cells, however, a significant upregulation of f-actin is seen upon treatment with mTCAP-1. Scale bar in A–D, F–I, K–N, P–S is 25µm and in E,J,O,T is 10µm. (U) Average number of red pixels (f-actin-bound Phalloidin signal) per cell with an intensity of 20–225 (intensity range indicated above) in HEK-WT and HEK-LPHN1 cells treated with either vehicle (gray) or 100 nM mTCAP-1 (black) for 60 min. Compared to vehicle treatment, mTCAP-1 increased f-actin in HEK-LPHN1-S cells only. (Mean ± SEM; n = 5, \*\*\*\*p < 0.0001; two-way ANOVA and a Tukey's post hoc test) (V) Height measurements of HEK-LPHN1-S and HEK-WT cells treated with either vehicle (gray) or 100 nM mTCAP-1 (black) for 60 min. No differences were observed between mean cell heights of vehicle and mTCAP-1 treated HEK-WT cells. Vehicle-treated HEK-LPHN1-S cells were larger than either HEK-WT group. mTCAP-1 treatment of HEK-LPHN1-S cells decreased cell height. (Mean ± SEM; n = 3, 4 measurements per n; \*\*\*\*p < 0.0001; two-way ANOVA with a Tukey's post hoc test).

more ancient Adhesion GPCR family (31, 35, 36, 39). The Adhesion GPCRs are themselves ancestral to the Secretin GPCRs, which suggests that the Secretin GPCRs inherited their ligandbinding HBDs from their Adhesion GPCR ancestors (36). Thus, given the similarity between the TCAP-1 and CRF peptides, and the structural similarities and phylogenetic histories of their respective receptors, we postulated that TCAP-1 interacts and with LPHN1 at its HBD.

To further characterize the similarities between the HBDs of LPHN and the Secretin family GPCRs, the LPHN1-3 HBD amino acid sequences were compared to those of the CRF receptors CRFR1 and CRFR2 and the calcitonin receptors CALCR and CGRPR (**Figure 2A**). These receptors were chosen as they are the most ancient of the Secretin GPCRs and the most closely related to the Adhesion GPCR family (36, 40). The comparison showed key sequence similarities at several HBD sites known to be critical for Secretin GPCR ligand binding (41–44). For example, in LPHN1, the cysteine residues C475, C499, and C511 are conserved in CRFR1 as C44, C68, and C87. Mutation of these residues in CRFR1 ablates CRF binding to the receptor (43). Studies using double-mutation of C68 and C87 in CRFR1 suggest that these residues form a disulfide bridge with each other, likely shaping the structure of the CRFR1 binding pocket (43). Moreover, the side chain of CRFR1 G64/CRFR2 G90 takes part in ligand interaction (42), and is conserved in all three LPHN paralogues. The strong conservation of these residues suggests their functional significance may also be conserved, potentially acting to form a LPHN1 HBD ligand-binding pocket. Interestingly, studies by Krasnoperov et al. (32) showed that multiple LPHN1 mutants lacking the HBD were unable to bind α-latrotoxin, further indicating the importance of this domain with respect to LPHN1 ligand interaction.

The potential ability of TCAP to interact with LPHN is consistent with previous studies, but provides a novel understanding of this relationship. The C-terminal region of teneurin-2, containing the TCAP-2 sequence, does bind to LPHN1 (27, 28); however, evidence of a direct interaction between LPHN and TCAP itself was not yet established in initial studies reporting LPHN1-teneurin-2 binding. Given these results and the conservation of the ligand and HBD structure described above, a co-IP assay was performed in which HEK293 cells were co-transfected with GFP-pro-mTCAP-1 or GFP-mTCAP-1, as well as constructs encompassing different portions of the LPHN1 HBD region (**Figure 1**). The TCAP-based constructs were designed according to the expected full-length TCAP-1 mRNA established from a previous study (19). The GFP-pro-mTCAP-1, based on the full-length TCAP-1 mRNA, was not detected in any eluates, suggesting that no interaction occurs between the TCAP-1 pro-peptide and LPHN1 HBD. In contrast, the GFP-mTCAP-1 construct, based on the putative 41-mer TCAP-1 region, was present as a band at approximately 30 kDa in the V444-E634 eluate and, to a lesser degree, in the V443-Q579 eluate (**Figure 3**, IPs, red arrow). This corresponds to the 30 kDa band observed for GFP-mTCAP-1 in the eluate inputs (**Figure 3**, inputs), and is consistent with a protein composed of the 25 kDa GFP and the 4.7 kDa mature TCAP-1 peptide. These results indicate that the mature TCAP-1 peptide may interact with LPHN1, most likely with a segment encompassing HBD residues V444 to E634. The presence of a weaker GFP-mTCAP-1 band in the V444- Q579 eluate suggests that this specific region of LPHN1 may be able to bind the mature TCAP-1 peptide at a lower affinity. Thus, although LPHN1 residues V444-to Q579 participate in ligand binding, they may represent only a partial binding pocket, whereas residues V444 to E634 may provide a more complete binding domain with which LPHN1 ligands can interact. It is important to note that both HBD constructs also contained a portion of the LPHN1 GAIN domain (**Figure 1**). This domain is unique to the Adhesion family of GPCRs and is thought to have a role in the transduction of conformational changes to the receptor transmembrane region, ultimately leading to induction of intracellular responses upon ligand-receptor binding (45). As the V444-E634 construct included a greater proportion of the GAIN domain than V444-Q597, it is likely that elements of the LPHN1 GAIN domain do play a role in peptide-binding. This may occur either through a direct contribution to the binding pocket, or by indirect stabilization of the tertiary HBD structure. Further studies will be required to ascertain the domains involved in the formation of the LPHN1 peptide-binding pocket.

# TCAP-1 Co-localizes With LPHN1 at the Cell Membrane

Having established that TCAP-1 can interact with the HBD/GAIN region of LPHN1, the next step was to determine if TCAP-1 could co-localize with LPHN1-overexpressing HEK293 cells. To confirm successful expression of the LPHN1-S vector, HEK-WT, and HEK-LPHN1-S cells were analyzed using western blot, which showed a strong band of about 120 kDa for HEK-LPHN1-S cells only (**Figure 4A**). This is consistent with a band at around 120 kDa found by Davletov et al. (29) in their study describing the initial discovery of LPHN1 as a receptor for α-latrotoxin. To corroborate this, and to observe the expression pattern of the LPHN1-S construct, an ICC analysis was performed using fluorescently-tagged antibodies targeting the FLAG tag at the construct N-terminal (**Figure 4C**). A strong signal was observed at the cell membrane in HEK-LPHN1-S cells, confirming successful transfection and high expression of the construct. No such signal was evident in the HEK-WT cells or the transfection control HEK-puro cells. Thus, HEK-WT and HEK-LPHN1-S cells were used to further investigate the binding of TCAP-1 and LPHN1, and the resultant downstream signaling effects.

HEK-WT and HEK-LPHN1-S cells were treated with FITC-K37-mTCAP-1 followed by a fluorescent antibody targeting the LPHN1-S construct, and the degree of TCAP-1 uptake and TCAP-1/LPHN1 co-localization in each cell type was observed (**Figure 5**). HEK-WT cells showed minimal uptake of FITC-K37-mTCAP-1, whereas HEK-LPHN1-S cells had a significant FITC-K37-mTCAP-1 signal at the plasma membrane. A marked overlap of FITC-K37-mTCAP-1 and LPHN1 at the cell membrane was also observed, indicating that the two are proximal to each other. As the endogenous presence of other LPHN isoforms was not assessed in the two cell lines used in this study, it is possible that TCAP-1 may be binding with these proteins as well. However, the significant difference in TCAP-1 uptake between HEK-WT and HEK-LPHN1-S cells and the strong overlap of the LPHN1-S and TCAP-1 fluorescence signals observed here indicate that TCAP-1 is most likely primarily interacting with LPHN1 in HEK-LPHN1-S cells. Despite these findings, our model renders an incomplete picture with respect to interaction between teneurin/TCAP and LPHN. Because there are four forms of teneurins/TCAPs and at least three LPHN paralogues in the vertebrate genome, it has been a challenge to find the perfect model to understand the interactions among the teneurins and LPHNs. We cannot discount the possibility that TCAP interacts with other LPHNs or indeed other receptor systems. However, it is important to note that this is the first study to report such an interaction between LPHN1 and the TCAP-1 region of teneurin-1. Recently, new studies of TCAP-1 in skeletal cells indicate that siRNA and CRISPR knockdowns of the LPHN1 receptor ablate TCAP-1 secondary messenger activity (D'Aquila et al., manuscript in preparation). Together, these studies indicate that TCAP-1 may interact with LPHNs.

It is important to note that particular regions of the cell membrane showed combined LPHN1 and TCAP-1 signals, whereas others showed only a LPHN1 signal or only a FITC-K37-mTCAP-1 signal (**Figure 5II**). This is the first study to show a potential interaction of the putative TCAP-1 peptide with LPHN1; however, our findings indicate that although TCAP-1 can interact with LPHN1, it may do so under only certain structural orientations. Our study differs from previous studies investigating the interaction between LPHN and teneurin (8, 27, 28) in that the soluble FITC-K37-mTCAP-1 peptide used here was introduced into an environment where intercellular interactions between LPHN1 and its binding partners, such as teneurin, may have already been established. It is possible that the TCAP-1 peptide possesses less affinity for the receptor to compete with existing teneurin-LPHN1 interactions and thus preferentially binds to LPHN1 receptors that are not occupied by the full-length teneurin proteins. If this is the case, regions with TCAP-1-LPHN1 interaction would be co-labeled with the fluorescent tags for both, whereas regions of teneurin-LPHN1 interaction would present only LPHN1 labeling (**Figures 5II,C,H**), accounting for many of the fluorescence patterns observed here. Similarly, regions of the cell membrane with only a FITC-K37-mTCAP-1 signal were observed (**Figures 5II,B,G**). As endogenous expression of the three LPHN isoforms in HEK-WT and HEK-LPHN1-S cells was not examined, it is possible that TCAP-1 may also be interacting with another LPHN isoform, or another protein at the cell membrane. It is not uncommon for peptides to bind multiple isoforms of their receptors; CRF is able to bind both CRFR1 and CRFR2 (37, 38), and α-latrotoxin binding has been observed for both LPHN1 and LPHN2 (29, 46). Interaction with other endogenously expressed LPHN isoforms would also account for the slight degree of FITC-K37-mTCAP-1 fluorescence observed in HEK-WT cells. In addition, a recent study by Li et al. (30) on the structure of the teneurin/TCAP region indicates that, in vivo, the TCAP region of the teneurins may be partially hidden by the teneurin protein. If so, such an arrangement may act to reduce the immunoreactive TCAP-1 signal.

# Morphological Effects of LPHN1 Expression and TCAP-1 Treatment in HEK293 Cells

To assess the effects of LPHN1 over-expression in HEK293 cells and the effects of TCAP-1 treatment on the HEK-WT and HEK-LPHN1-S cell lines, several components of cell morphology were examined. Initial observations indicated that HEK-LPHN1-S were significantly smaller compared to HEK-WT cells, as shown by measurements of nuclear and whole cell area and perimeter (**Figures 6I–L**). However, these changes were quantified on a two-dimensional basis and further investigation showed that vehicle-treated HEK-LPHN1-S cells are taller than vehicletreated HEK-WT cells. This suggests that the observed changes in cell size are likely to be purely morphological, with total cell volume being unaffected.

To date, a role in pathways that influence mammalian cell size, such as those associated with mTOR and P13K (47), has not been reported for LPHN1; however adhesion roles have been individually established for both LPHN1 and its binding partners, and the formation of the teneurin-LPHN transsynaptic complex can increase adhesion between cells (8, 13, 48). Increased expression of LPHN1 could lead to increased formation of intercellular adhesion complexes, leading to cells clustering together more tightly. Within a confined space, this would result in a shift from a spherical or cubic cell shape to one that is more columnar, suggesting that the morphological differences observed between HEK-WT and HEK-LPHN1-S cells are simply due to increased adhesion between HEK-LPHN1-S cells. These cells also had a significantly reduced expression of f-actin and had fewer f-actin projections compared to HEK-WT cells (**Figure 7**). As HEK-LPHN1-S cells cluster closer together, a reduction in cytoskeletal elements between cells is to be expected. To date, a role for LPHN in cytoskeletal modulation has not been established; however, knock-down of LPHN2 in chicken cardiac tissue results in differential expression of 37 cytoskeletal genes (49). Together, these data suggest a potential role for LPHN1 in the regulation of f-actin polymerization.

As TCAP-1 induces cytoskeletal changes in neurons by modulating f-actin polymerization (19–21), the next step was to investigate the effects of TCAP-1 treatment on the morphology and cytoskeletal profiles of HEK-WT and HEK-LPHN1-S cells. Treatment with mTCAP-1 had no significant effects on HEK-WT actin polymerization, whereas it induced a significant factin increase in HEK-LPHN1-S cells. HEK-LPHN1-S cells also had more f-actin projections than their wild-type counterparts. This is consistent with the actions of TCAP-1 on the cytoskeletal arrangement of neuronal cells. TCAP-1-treated murine hypothalamic neurons show elongated neurites and increased expression of cytoskeletal components (20), whereas primary hippocampal neurons have greater neurite number, larger axon bundles, and changes in their dendritic arborization (21). These cytoskeletal changes are due to ERK1/2-induced polymerization of f-actin and re-organization of microtubules (19). TCAP-1-activated ERK1/2 phosphorylates p90RSK, which can in turn phosphorylate filamin A, causing it to induce cross-linking and stabilize actin filaments in neuronal cells. The marked increase in the f-actin expression of the HEK-LPHN1- S cells but not of the HEK-WT cells suggests that this action of TCAP-1 on the cytoskeleton may occur via LPHN1. It is possible that over-expression of the LPHN1-S isoform affects the health and functioning of HEK293 cells in such a way that impacts their cytoskeletal components and cellular morphology and that treatment with TCAP-1 simply acts to sequester LPHN1 and thus reduce those effects. However, similar studies to this one regarding LPHN1-teneurin binding in HEK293 cells have previously been conducted with no evidence of such an effect (8), and cytoskeletal modulation is a well-documented effect of TCAP-1 (19–21), making this an unlikely interpretation of the data presented here. Thus, taken together, these results are the first to show that TCAP-1 induces its effects through an interaction with LPHN1, indicating that TCAP-1 and LPHN1 form an endogenous ligand-receptor pair. This is particularly important, as it is also the first time that a role in cytoskeletal modulation has been reported for LPHN1.

#### TCAP and LPHN as an Evolutionarily Ancient Receptor-Ligand Pair

Teneurin/TCAP and LPHN comprise the only known trans-synaptic pair to be conserved in both vertebrates and invertebrates (50), and they appear to have a shared evolutionary history. TCAP is an ancient peptide related to CRF and other members of the Secretin peptide family (18, 51, 52), whereas LPHN belongs to the Adhesion GPCRs, from which the Secretin GPCRs evolved (35, 36). Furthermore, both are found in the

#### REFERENCES

1. Baumgartner S, Chiquet-Ehrismann R. Tena, a Drosophila gene related to tenascin shows selective transcript localization. Mech Dev. (1993) 40:165–76. doi: 10.1016/0925-4773(93)90074-8

choanoflagellate, a single-celled ancestor of the metazoans (30, 53–55), where TCAP is hypothesized to have been acquired from a prokaryote genome via a horizontal gene transfer of an ancestral teneurin-like gene (56). Interestingly, the teneurins are structurally similar to bacterial polymorphic proteinaceous toxins, which possess a soluble toxin payload at their C-terminus that can be released into target cells (30, 53, 56). This payload is highly conserved and corresponds to the TCAP portion of teneurin, further indicating the similarities between these proteins and highlighting the extended evolutionary history of the teneurins.

# CONCLUSION AND FINAL REMARKS

In summary, TCAP-1 is a highly bioactive peptide with actions both in vitro and in vivo. In vitro, it is associated with multiple signal transduction systems, such as the MEK-ERK1/2 pathway, and can modulate the cytoskeleton (20–22). In vivo, TCAP-1 affects anxiety- and stress-related behaviors in a manner that is dependent, in part, on the baseline emotionality of animals, with different responses to treatment being observed between high baseline and low baseline animals (17). Previous studies suggest that the teneurins and LPHN represent a conserved trans-synaptic ligand-receptor pair with a number of intercellular actions (8, 27, 28). This study indicates that the TCAP region of teneurin, as a soluble peptide, also plays a role in this interaction and that its roles in cytoskeletal remodeling occur in part via LPHN1. This sets the stage for future research to further elucidate the actions of TCAP-1 at both cellular and behavioral levels.

#### AUTHOR CONTRIBUTIONS

MH contributed to experimental design, collection, analysis, and interpretation of the data, writing of the manuscript and subsequent revisions of the manuscript. DB-L contributed to experimental design, performed cell transfections, and collected and aided in the analysis and interpretation of the data. DL supervised the project, contributed to experimental design, interpretation of the data, and writing and revision of the manuscript. All authors have reviewed and approved of the final draft of the manuscript.

#### ACKNOWLEDGMENTS

This work was carried out with funding from the Natural Sciences and Engineering Research Council (NSERC) and Protagenic Therapeutics Inc. to DL. MH is a recipient of an NSERC Post-Graduate Scholarship.


rule gene. Cell (1994) 77:587–98. doi: 10.1016/0092-8674(94)90 220-8


primary cultures by teneurin C-terminal associated peptide-1. Neuroscience (2007) 144:1241–54. doi: 10.1016/j.neuroscience.2006.09.062


characterization of a cDNA expressed in heart. Proc Natl Acad Sci USA. (1995) 92:2969–73. doi: 10.1073/pnas.92.7.2969


**Conflict of Interest Statement:** DL is a co-founder and chief scientific advisor of Protagenic Therapeutics Inc.

The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Husi´c, Barsyte-Lovejoy and Lovejoy. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

# Wnt Signaling Upregulates Teneurin-3 Expression via Canonical and Non-canonical Wnt Pathway Crosstalk

Sussy Bastías-Candia1,2, Milka Martínez<sup>1</sup> , Juan M. Zolezzi1,2 and Nibaldo C. Inestrosa1,2,3 \*

<sup>1</sup> Basal Center for Aging and Regeneration, Facultad de Ciencias Biológicas, Pontificia Universidad Católica de Chile, Santiago, Chile, <sup>2</sup> Center of Excellence of Biomedicine of Magallanes, Universidad de Magallanes, Punta Arenas, Chile, <sup>3</sup> School of Psychiatry, Centre for Healthy Brain Ageing, Faculty of Medicine, University of New South Wales, Sydney, NSW, Australia

#### Edited by:

Antony Jr. Boucard, Centro de Investigación y de Estudios Avanzados (CINVESTAV), Mexico

#### Reviewed by:

Nils Lambrecht, University of California, Irvine, United States May Ann Lee, Experimental Therapeutics Centre (A∗STAR), Singapore

> \*Correspondence: Nibaldo C. Inestrosa ninestrosa@bio.puc.cl

#### Specialty section:

This article was submitted to Neuroendocrine Science, a section of the journal Frontiers in Neuroscience

Received: 28 January 2019 Accepted: 02 May 2019 Published: 17 May 2019

#### Citation:

Bastías-Candia S, Martínez M, Zolezzi JM and Inestrosa NC (2019) Wnt Signaling Upregulates Teneurin-3 Expression via Canonical and Non-canonical Wnt Pathway Crosstalk. Front. Neurosci. 13:505. doi: 10.3389/fnins.2019.00505 Teneurins (Tens) are a highly conserved family of proteins necessary for cell-cell adhesion. Tens can be cleaved, and some of their proteolytic products, such as the teneurin c-terminal associated-peptide (TCAP) and the intracellular domain (ICD), have been demonstrated to be biologically active. Although Tens are considered critical for central nervous system development, they have also been demonstrated to play important roles in adult tissues, suggesting a potential link between their deregulation and various pathological processes, including neurodegeneration and cancer. However, knowledge regarding how Ten expression is modulated is almost absent. Relevantly, the functions of Tens resemble several of the effects of canonical and non-canonical Wnt pathway activation, including the effects of the Wnt pathways on neuronal development and function as well as their pivotal roles during carcinogenesis. Accordingly, in this initial study, we decided to evaluate whether Wnt signaling can modulate the expression of Tens. Remarkably, in the present work, we used a specific inhibitor of porcupine, the key enzyme for Wnt ligand secretion, to not only demonstrate the involvement of Wnt signaling in regulating Ten-3 expression for the first time but also reveal that Wnt3a, a canonical Wnt ligand, increases the expression of Ten-3 through a mechanism dependent on the secretion and activity of the non-canonical ligand Wnt5a. Although our work raises several new questions, our findings seem to demonstrate the upregulation of Ten-3 by Wnt signaling and also suggest that Ten-3 modulation is possible because of crosstalk between the canonical and non-canonical Wnt pathways.

Keywords: teneurin-3, Wnt signaling, Wnt3a, Wnt5a, neuronal development, C59

# INTRODUCTION

Teneurins (Tens; e.g., Ten-m/ODZ) are part of a conserved family of type II transmembrane proteins that are highly relevant during embryogenesis and have functions related to the proper development of the central nervous system (CNS), specifically neuronal matching and neuronal circuitry patterning (Baumgartner and Chiquet-Ehrismann, 1993; Baumgartner et al., 1994; Levine et al., 1994; Kenzelmann et al., 2007; Tucker et al., 2007; Young and Leamey, 2009; Hong et al., 2012). Ten proteins, which in vertebrates include four members (1 to 4), exhibit

functions in several processes, including cell adhesion, cytoskeleton interaction, and calcium binding (Tucker and Chiquet-Ehrismann, 2006; Tucker et al., 2007). Cell adhesion and neuronal matching are dependent on Ten dimerization at cysteine residues located in the extracellular domain (Feng et al., 2002; Rubin et al., 2002; Beckmann et al., 2013). On the other hand, filopodium formation, synaptogenesis, and axonal growth and guidance also depend on the interaction of Tens with CAP/Ponsin (Nunes et al., 2005; Mosca et al., 2012), a key regulatory protein in actin polymerization (Zhang et al., 2013). Additionally, Tens can be cleaved, leading to the release of the intracellular domain (ICD), which can translocate to the nucleus and act as a transcriptional regulator (Bagutti et al., 2003; Nunes et al., 2005). Similarly, the extracellular domain of Tens can also be cleaved, leading to the release of the teneurin c-terminal-associated peptide (TCAP), which has been found to demonstrate interesting neuroactive properties (Wang et al., 2005; Tan et al., 2011). Accordingly, deregulated Tens expression during embryogenesis leads to severe alterations, including impaired binocular vision (Leamey et al., 2007; Dharmaratne et al., 2012; Young et al., 2013), microphthalmia, visual defects (Aldahmesh et al., 2012) and impaired hippocampal neuronal networking (Berns et al., 2018).

Although these observations have been restricted to the CNS, Tens are expressed in several adult tissues, suggesting that these proteins might be related to normal physiology as well as to chronic degenerative processes observed in fully developed tissues. Although information about this latter issue is scarce, altered expression of Tens has been reported in several types of cancer, including breast, ovarian, liver, and nervous system cancers (Molenaar et al., 2012; Ziegler et al., 2012a,b). Furthermore, it has recently been proposed that the expression levels of Ten-2, Ten-3, and Ten-4 might have interesting prognostic value in some types of cancer, such as ovarian cancer and neuroblastoma (Molenaar et al., 2012). On the other hand, the critical roles demonstrated by Tens in the establishment of the neuronal circuitry, especially in the hippocampal region, and the neuroactive properties of the TCAP make Tens highly interesting candidates for evaluation in the context of neurodegenerative disorders such as Alzheimer's disease, where synaptic loss, neuronal cell death and hippocampal circuitry failure are part of the pathophysiological process (Selkoe and Hardy, 2016).

Although Ten expression has been suggested to be tightly regulated during embryogenesis by a self-regulated mechanism that depends on the overlap of Ten functions and on interactions with additional cell adhesion molecules, such as neurexin and neuroligin (Ben-Zur et al., 2000; Zhou et al., 2003; Kinel-Tahan et al., 2007; Mosca, 2015), our knowledge of the mechanisms able to modulate Ten expression, including crosstalk with known signaling pathways, is limited. Such limitation compromises our understanding of the involvement of these proteins in different cellular processes and the potential roles of these proteins as pharmacological targets for various pathological conditions.

Importantly, the activities of Tens share relevant similarities with the effects of the Wnt signaling pathway, a critical molecular pathway for CNS development and function (Bastías-Candia et al., 2015; Mosca, 2015). Moreover, several of the known functions of Tens and Wnt seem to suggest direct crosstalk between these two elements. In this regard, Wnt signaling, which can be divided into the canonical Wnt/β-catenin pathway and the non-canonical Wnt/planar cell polarity (PCP) and Wnt/Ca++ pathways, has been demonstrated to be critical for dendritic arborization, axonal elongation, maintenance of the synaptic architecture, and neurogenesis in the adult brain (Inestrosa and Arenas, 2010; Varela-Nallar and Inestrosa, 2013). However, the Wnt signaling pathway also constitutes a fundamental growth control pathway, and its alteration has been linked with several pathological conditions ranging from abnormal development/function of different biological systems to cancer development (Koo et al., 2015; Nusse and Clevers, 2017). Indeed, the Wnt pathway plays a critical role in carcinogenesis and is considered one of the most significant molecular pathways associated with the malignant transformation leading to tumorigenesis (Polakis, 2012).

Accordingly, given the Tens-Zic2 relationship, the presence of calcium-sensitive motifs within the structures of Tens and the already well-defined functions of Tens and the Wnt pathway, we hypothesized that there is direct communication between these pathways, which could be of significance in the context of chronic degenerative processes (Bastías-Candia et al., 2015). Thus, in the present work, we tested this hypothesis and evaluated whether Wnt signaling can modulate the expression of Ten-3, a representative Ten whose expression has been demonstrated to be necessary for CNS and hippocampal network development as well as for neuroblastoma tumorigenesis. After performing in silico analysis to corroborate the presence of binding motifs for TCF/Lef (the conserved canonical Wnt signaling transcription factor) in the TEN-3 gene promoter, we used C59, a highly specific porcupine inhibitor, to ablate the Wnt signal (Hofmann, 2000; Herr et al., 2012; Ho and Keller, 2015; Koo et al., 2015; Nusse and Clevers, 2017). Interestingly, we observed that both the Wnt3a (canonical) and Wnt5a (non-canonical) ligands were able to increase basal expression of Ten-3 mRNA by up to 81 and 247%, respectively. Moreover, we observed that the Wnt3a-mediated increase in Ten-3 was dependent on the release of the ligand Wnt5a, suggesting a central role for the noncanonical pathway in Ten-3 expression. Altogether, our findings not only support Wnt-mediated modulation of Ten-3 expression but also suggest a more complex mechanism of regulation involving direct and necessary crosstalk between the canonical and non-canonical Wnt pathways. Although preliminary, our work constitutes the very first report of Wnt-Ten crosstalk and will initiate a very interesting field of research given the potential implications of such communication in the context of cell physiology and certain pathophysiological processes, including cancer and neurodegenerative disorders.

# MATERIALS AND METHODS

# Identification of TCF/Lef Consensus Binding Sites on Genes of Interest

To identify potential TCF/Lef binding sites, an in silico analysis was carried out. Genomic sequences of the human TEN-3

and WNT5A gene promoters were screened for putative DNA binding motifs using the JASPAR database with an 80% relative score threshold.

# Cell Culture and Treatments

The neuroblastoma cell line SH-SY5Y was purchased from Sigma-Aldrich (St. Louis, MO, United States) and was handled as recommended by the supplier. Briefly, cells were grown in DMEM supplemented with 10% fetal bovine serum and a 1% penicillin and streptomycin solution. The cells were allowed to reach 70% confluence prior to subculture. The cells were passaged at least eight times and were seeded onto 96-well and 12-well plates to carry out all the experiments. Additionally, some cells were seeded onto 12 mm coverslips for immunofluorescence assessment of Ten-3 expression. For experimentation, the SH-SY5Y cells were treated for 24 h with recombinant Wnt3a (150 ng/ml), Wnt5a (150 ng/ml) and Wnt7a (150 ng/ml) (R&D systems, Minneapolis, MN, United States) alone or in the presence of C59 (Tocris, Minneapolis, MN, United States).

# C59 Cytotoxicity Assay and Wnt Secretion Blockade

SH-SY5Y cells were seeded onto 96-well plates and treated with 1, 10, 100, or 200 µM C59. After 24 h in culture, cytotoxicity was evaluated using an MTT assay. Briefly, the cells were treated with different concentrations of C59 for 24 h. At the end of treatment, the cells were washed with 1× PBS, and 100 µl of fresh 1× PBS was added to each well. Then, 10 µl of thiazolyl blue tetrazolium bromide (MTT, 0.45 mg/ml) was added to each well and incubated for 3 h at 37◦C. After the incubation with MTT, DMSO was added as a solubilization solution to dissolve the formazan crystals. The absorbance was measured in an iMark microplate reader (Bio-Rad, Hercules, CA, United States) at 570 nm.

Once we established the non-cytotoxic concentrations of C59, we evaluated the inhibition of Wnt ligand secretion. To do this, we used the trichloroacetic acid (TCA) protein precipitation method to assess Wnt3a levels in the extracellular medium. Briefly, after cells were treated for 24 h with 1 and 10 µM C59, the culture medium was replaced, and 100% TCA solution was added in a 1:4 (TCA:sample) ratio. The samples were incubated at 4◦C for 10 min and centrifuged at 14,000 rpm for 5 min. The resulting pellet was washed twice in cold acetone and dried at 95◦C for 5 min. Total protein was loaded and resolved by SDS-PAGE using a 10% polyacrylamide gel, and the separated proteins were transferred to a PVDF membrane. The membrane was incubated with a rabbit anti-Wnt3a antibody (overnight, 4◦C, 1:1000, ab28472, Abcam, Cambridge, MA, United States) and an HRP-conjugated secondary antibody (1 h, room temperature, 1:5000, cat. no. 31460, Thermo Scientific, Waltham, MA, United States), and Clarity Western ECL Substrate (Bio-Rad) was used for the chemiluminescence reaction. Chemiluminescence was detected using a ChemiDoc-It 515 Imager (UVP, Upland, CA, United States).

#### Immunofluorescence Staining

After treatment, SH-SY5Y cells seeded on coverslips were fixed with a 4% paraformaldehyde/sucrose solution, permeabilized with a 0.1% PBS/TWEEN 20 solution and blocked using a 1% PBS/bovine serum albumin (BSA) solution. Then, the cells were incubated with a Ten-3 (ICD) primary antibody (1:50, ab205507, Abcam) overnight. Secondary antibody (Alexa Fluor 488, 1:1000, Abcam) incubation was performed for 1 h at 37◦C. Phalloidin (Alexa Fluor 568 Phalloidin, 1:50, Thermo Fisher Scientific) and TO-PRO-3 (Alexa Fluor 647 TO-PRO-3, 1:50, Thermo Fisher Scientific) antibodies were used to stain actin and nuclei, respectively. Images were captured using a Zeiss LSM5 Pascal confocal microscope (Zeiss, Oberkochen, Germany). Image analysis was carried out using ImageJ software.

# Quantitative PCR (qPCR)

RNA extraction was performed using the TRIzol (Invitrogen, Carlsbad, CA, United States) method according to the manufacturer's instructions. The extracted RNA was quantified in a microplate spectrophotometer (BioTek, Winooski, VT, United States), and 500 ng of RNA was used for reverse transcription with Superscript IV (Invitrogen). Primers for qPCR were custom designed and synthesized by Integrated DNA Technologies (IDT, Skokie, IL, United States). The mRNA levels of Ten-3, Wnt5a and Cyclin D1 were determined using the <sup>11</sup>Ct method with GAPDH as the housekeeping control. The primer sequences used to perform qPCR were as follows: GAPDH, 5 0 -AGACAGCCGCATCTTCTTGT-3<sup>0</sup> (forward) and 5<sup>0</sup> -CTTGC CGTGGGTAGAGTCAT-3<sup>0</sup> (reverse); Ten-3, 5<sup>0</sup> -CGGGTACCCA CACAGAAGTC-3<sup>0</sup> (forward) and 5<sup>0</sup> -GCCTTAGGGTAAAATT CTGTCCTTG-3<sup>0</sup> (reverse); Wnt5a, 5<sup>0</sup> -AACTGGCGGGACTTT CTCAA-3<sup>0</sup> (forward) and 5<sup>0</sup> -GTCTCTCGGCTGCCTATTTG-3<sup>0</sup> (reverse); and Cyclin D1, 5<sup>0</sup> -GACCCCGCACGATTTCATTG-3<sup>0</sup> (forward) and 5<sup>0</sup> - AAGTTGTTGGGGCTCCTCAG-3<sup>0</sup> (reverse).

#### Statistical Analysis

The data are presented as the mean ± SEM. The data were transferred to Excel spreadsheets after collection. Statistical analysis was carried out using Prism 6 software v.6.0h (GraphPad, La Jolla, CA, United States). The experiments were conducted in triplicate, and one-way ANOVA followed by Bonferroni's post hoc test was applied to identify statistically significant differences. Significance was set at p < 0.05 (p <sup>∗</sup> = 0.05; p ∗∗ = 0.01; p ∗∗∗ = 0.001).

# RESULTS

#### The Human Ten-3 Promoter Region Possesses Several TCF/Lef Binding Motifs, Suggesting Canonical Wnt-Dependent Modulation

Using an in silico approach, we assessed the potential regulation of Ten-3 expression by canonical Wnt signaling. A total of 14 TCF/Lef binding motifs were found up to 2 kb upstream of the transcription start site of Ten-3. These motifs were located homogeneously throughout the 2 kb region, with three motifs in the most proximal 500 bp region, four in the following 500 bp region, and seven in the last 1 kb region (**Figure 1**). This initial

observation strongly suggested that the canonical Wnt/β-catenin pathway might regulate the expression of Ten-3. Accordingly, we tested in vitro whether Wnt3a, a well-known canonical Wnt ligand, could induce the expression of Ten-3.

C59, a Specific Porcupine Inhibitor, Blocks Wnt3a Secretion in SH-SY5Y Cells

Prior to evaluating the effects of Wnt3a on Ten-3 expression, we used the compound C59, a well-known and highly selective porcupine inhibitor, to ablate the basal Wnt signal. To do so, we treated SH-SY5Y cells with different concentrations of C59, and we evaluated both the cytotoxicity of C59 and the inhibition of the secretion of the ligand Wnt3a. After 24 h of treatment, it was evident that C59 affected cell survival when a dose greater than 100 µM was administered, inducing up to 40% mortality at 200 µM (0.6 ± 0.011, p ∗∗∗) (**Figure 2A**). Based on this result, we evaluated whether 1 µM and 10 µM C59 effectively prevented Wnt3a secretion. As expected, both concentrations almost completely abolished Wnt3a in the extracellular medium, reducing the levels of the ligand by up to 85% (1 µM: 14.49 ± 9.18, p ∗∗∗; 10 µM: 14.92 ± 8.17, p ∗∗∗) (**Figure 2B**).

#### Wnt3a Increases Ten-3 Signal and mRNA Levels in SH-SY5Y Cells

After determining the concentration of C59 to be used, we evaluated the effects of the ligand Wnt3a on the expression of Ten-3. After 24 h of treatment with 150 ng/ml recombinant Wnt3a, the treated cells showed 1.5-fold higher Ten-3 (ICD) signal levels than the control cells (2.52 ± 0.064, p ∗∗∗) (**Figures 3A,B**). On the other hand, the cells treated with 10 µM C59 exhibited a small reduction in the Ten-3 (ICD) signal compared to control cells (**Figures 3A,B**). However, co-incubation of cells with Wnt3a and C59 completely prevented the Wnt3a-induced increase in the Ten-3 (ICD) signal (**Figures 3A,B**). Interestingly, through analysis of mRNA levels, we confirmed that Wnt3a increases the expression of Ten-3 mRNA (1.81 ± 0.07, p ∗∗∗) but that this increase is abolished in the presence of C59 (**Figure 3C**). To eliminate the possibility of an issue with the effectiveness of the Wnt3a stimulus, we determined the mRNA expression levels of Cyclin D1, a well-established canonical Wnt target gene, in the same samples. C59-treated cells exhibited a significant decrease in cyclin D1 mRNA compared with control cells (0.66 ± 0.032, p ∗ ); however, the cyclin D1 levels recovered and slightly increased when Wnt3a was added (Wnt3a+C59 compared with C59; 1.16 ± 0.032, p ∗∗) (**Figure 3D**). Considering the specificity of

FIGURE 2 | C59 cytotoxicity and inhibitory concentration for the secretion of Wnt ligands. (A) Cytotoxicity was evaluated through assessment of mitochondrial functionality with an MTT assay. Treatment with C59 for 24 h at concentrations over 100 µM reduced SH-SY5Y vitality, causing a mortality rate of up to 40% at the 200 µM concentration (0.6 ± 0.011, ∗∗∗p < 0.001). Although a slight decrease in SH-SY5Y vitality was observed at 100 µM, this difference was not significant. (B) The inhibitory concentration of C59 was estimated based on the results of the cytotoxicity assay. Thus, 1 µM and 10 µM C59 were tested. After 24 h of C59 treatment, the levels of the ligand Wnt3a, a representative Wnt ligand, were assessed in the culture medium using the trichloroacetic acid precipitation method. The concentration of Wnt3a was decreased by up to 85% under treatment conditions compared with control conditions (1 µM: 14.49 ± 9.18, ∗∗∗p < 0.001; 10 µM: 14.92 ± 8.17, ∗∗∗p < 0.001).

after treatment with the ligand Wnt3a (2.52 ± 0.064, ∗∗∗p < 0.001). The effect was completely lost when Wnt3a was used in combination with C59. (C) Quantification of mRNA levels showed that the levels of Ten-3 mRNA increased when SH-SY5Y cells were treated with Wnt3a (1.81 ± 0.07, ∗∗∗p < 0.001 compared to control levels). As observed in the IF results, the combination of Wnt3a and C59 abolished the effect of Wnt3a on Ten-3 mRNA levels. (D) C59 treatment significantly decreased the levels of cyclin D1 mRNA (0.66 ± 0.032, <sup>∗</sup>p < 0.05 compared to control levels). However, exogenous Wnt3a prevented this decrease and recovered cyclin D1 mRNA to the levels in control cells, which were significantly different than those in C59-treated cells (1.16 ± 0.032, ∗∗p < 0.01).

the inhibitor C59, these results suggest that Wnt3a modulates Ten-3 expression by affecting the secretion of a secondary Wnt ligand. Moreover, considering that the canonical ligand was not able to counterbalance the effects of C59 on Ten-3 mRNA levels, we inferred that non-canonical Wnt signaling might be part of the molecular mechanism involved in the upregulation of Ten-3 expression.

## Wnt3a Increases the mRNA Levels of the Ligand Wnt5a in SH-SY5Y Cells

Accordingly, we conducted a second in silico evaluation to screen for TCF/Lef binding sites, this time in the promoter region of the WNT5A gene, a representative non-canonical Wnt ligand. Interestingly, the in silico analysis showed that even though Wnt5a exists in two isoforms and thus has two promoter regions, TCF/Lef binding sites are present in both regions (**Figure 4A**). After verifying the sites and using the same samples to evaluate the Ten-3 mRNA levels, we assessed the expression of the noncanonical ligand Wnt5a. Interestingly, Wnt5a mRNA levels were significantly higher in Wnt3a- and Wnt3a+C59-treated cells compared to control cells (1.69 ± 0.077, p ∗∗, and 1.44 ± 0.094, p ∗ , respectively) (**Figure 4B**). This finding further suggests that the canonical ligand Wnt3a might induce the expression of Ten-3 but in a manner linked to the secretion and activity of the non-canonical ligand Wnt5a.

# Wnt5a Dramatically Increases Ten-3 mRNA Levels in the Presence of the Inhibitor C59

To confirm the above finding, we proceeded to evaluate cells treated with Wnt3a+C59, this time adding Wnt5a or Wnt7a recombinant ligands. In this case, Wnt7a was included as an additional control for the Wnt canonical pathway. Remarkably, when Wnt5a was added to the Wnt3a+C59 group, the expression level of Ten-3 mRNA increased to 247% of the level under control conditions (3.47 ± 0.199, p ∗∗∗). In contrast, the Wnt7a ligand was not able to replicate this result, indicating that the effect of Wnt3a on Ten-3 mRNA levels was Wnt5a-dependent (**Figure 4C**).

# DISCUSSION

In recent years, the role of Tens in the developing nervous system has been well documented, demonstrating that appropriate expression of this family of proteins is mandatory for neuronal development and neuronal network formation. Recent data have also pointed out a significant role for Tens in the context of adult tissue biology (Ziegler et al., 2012b). Indeed, Tens and its proteolytic products, such as TCAP and ICD, have been linked with important effects on the adult nervous system, including the management of addiction and anxiety (Kupferschmidt et al., 2011; Tan et al., 2011; Erb et al., 2014). Similarly, other studies have shown that aberrant Ten expression is associated with tumor development and malignancy, suggesting that specific Tens can be used as valuable prognostic cancer biomarkers (Molenaar et al., 2012; Ziegler et al., 2012a,b). Together, these findings strongly suggest that Tens play relevant roles in the maintenance and physiology of fully developed tissues outside of the CNS. Surprisingly, despite the depicted importance of these proteins and the wide range of biological effects that these proteins seem to mediate, little information is available about the regulatory mechanisms of their expression, including the modulatory effects of well-known cellular signaling pathways. As mentioned in the introductory section, the similarities between Ten functions and those described for Wnt signaling, mainly in the context of the CNS, prompted us to evaluate our former hypothesis and determine whether activation of the Wnt pathway could modulate Ten expression.

As a starting point, we evaluated the promoter region of TEN-3 for the presence of TCF binding motifs. Remarkably, the in silico analysis revealed 14 potential binding sites for the TCF family of transcription factors, with ten of these sites corresponding to TCF7/Lef2 (**Figure 1**). In this regard, although several TCF family members (1 to 4) have been shown to exert opposing regulatory effects when bound to β-catenin, TCF7/Lef2 has been systematically shown to increase the expression of its target genes (Nakano et al., 2010; Cadigan and Waterman, 2012; Ramakrishnan and Cadigan, 2017). Our initial screening further indicated the potential involvement of the Wnt signaling pathway, particularly the canonical branch, in the modulation of Ten-3 expression.

Based on this initial finding, we proceeded to evaluate the involvement of the ligand Wnt3a, a main representative of the canonical branch of the Wnt pathway, on the expression levels of Ten-3. Considering that most of the information regarding Tens has been reported for the nervous system and for cancer, we decided to use the SH-SY5Y cell line, a neuronal model and a representative neuroblastoma-derived cell line. Moreover, to properly evaluate the effects of exogenous Wnt3a on Ten-3 expression levels, we used C59, a specific inhibitor of porcupine; porcupine is an exclusive regulatory enzyme of Wnt ligand palmitoylation, which is mandatory for Wnt ligand secretion and bioactivity (Herr et al., 2012; Proffitt et al., 2013; Wend et al., 2013; Ho and Keller, 2015; Koo et al., 2015; Bernatik et al., 2017; Nigmatullina et al., 2017; Nusse and Clevers, 2017). Indeed, under our experimental conditions, C59 almost completely abolished Wnt ligand secretion without affecting cell survival, at least at the 1 µM and 10 µM concentrations (**Figure 2**). Notably, even though we observed that exogenous Wnt3a significantly increased the expression levels of Ten-3 at both the mRNA and protein levels, Wnt3a was unable to induce Ten-3 expression when C59 was present (**Figures 3A,B**). To corroborate the effectiveness of the exogenous Wnt3a treatment, we evaluated the mRNA levels of Cyclin D1, a conserved canonical Wnt pathway target gene, in the same samples. Interestingly, we observed that Wnt3a was able to prevent the reduction in Cyclin D1 expression caused by C59, confirming that Wnt3a induced the effects of C59. Based on these findings, we hypothesized that Wnt3a was able to induce Ten-3 expression but that this effect involved the secretion and activity of a secondary Wnt ligand. Moreover, considering that the canonical ligand was unable to counteract the effects of C59 on Ten-3 mRNA levels, as observed with Cyclin D1, we inferred that non-canonical Wnt signaling may be part

promoter A, two binding sites were located at –1000 and –400 bp upstream of the exon 1 transcription start site. On the other hand, at promoter B, the two binding sites were located in the first 150 bp upstream of the exon 2 transcription start site. Green: TCF7Lef2. (B) Wnt5a mRNA levels were significantly increased in both the Wnt3a and Wnt3a+C59 groups compared with the control groups (1.69 ± 0.077, ∗∗p < 0.01 and 1.44 ± 0.094, <sup>∗</sup>p < 0.05; respectively). (C) Cells treated with 150 ng/ml Wnt5a had 247% higher Ten-3 mRNA levels than control cells (3.47 ± 0.199, ∗∗∗p < 0.001).

of the molecular mechanism involved in the upregulation of Ten-3 expression.

Accordingly, we conducted an additional in silico analysis to screen for TCF binding motifs, this time in the promoter region of Wnt5a, a well-studied representative non-canonical Wnt ligand that has been suggested as a potential target of the canonical Wnt pathway (Hödar et al., 2010). Our analysis showed that both promoter regions of the WNT5a gene contain two TCF7/Lef2 binding motifs, suggesting that both Wnt5a isoforms are subject to canonical Wnt modulation. Indeed, when we evaluated the levels of Wnt5a mRNA in the samples previously exposed to Wnt3a and Wnt3a+C59, we observed significant increases in Wnt5a mRNA expression of up to 70 and 45%, respectively (**Figures 4A,B**). This finding suggested that Wnt3a probably induced the increased expression of Wnt5a but that because of the inhibitory effect of C59, the Wnt5a ligand was not palmitoylated, affecting its secretion and activity. Thus, we investigated whether exogenous Wnt5a could overcome

that several TCF7/Lef2 motifs are present in the promoter region of the TEN-3 gene and that Wnt5a can signal through β-catenin via ADP-ribosylation factor 6

(ARF6) activity seem to further support this suggested cooperative mechanism between the canonical and non-canonical Wnt pathways.

the inhibitory effects of C59. Remarkably, when we introduced Wnt5a into the system, we observed a dramatic increase in the

levels of Ten-3 mRNA, which reached values up to 247% of those in control cells. In addition, to establish the specific role of Wnt5a in these effects, we used Wnt7a as a secondary canonical Wnt ligand. In this case, no significant increase in Ten-3 mRNA levels was observed (**Figure 4C**).

Together, our results demonstrate that Ten-3 expression can be regulated by Wnt signaling. Moreover, they suggest that even though the non-canonical branch can induce Ten-3 expression independently through the ligand Wnt5a, the canonical branch requires a cooperative mechanism involving both the canonical and non-canonical Wnt pathways. Similar cooperation between canonical signals and non-canonical signals, specifically Wnt5a, has been reported previously (Schulte et al., 2005; Andersson et al., 2013). Moreover, as has been stated for the canonical Wnt pathway, Wnt5a has been found to be related to neuronal development, axonal guidance, neuronal branching, and organ innervation (Kumawat and Gosens, 2016). Remarkably, it has been shown that Wnt5a not only activates the non-canonical pathway but also can activate canonical signaling through activation of the GTPase ADP-ribosylation factor 6 (ARF6) after FZD4-LRP6 binding, allowing β-catenin-related gene transcription (Grossmann et al., 2013). Considering the various TCF7/Lef2 binding sites in the promoter region of Ten-3, it is possible that the protein expression of Ten-3 is mediated by crosstalk between the canonical and non-canonical Wnt pathways, specifically between the ligands Wnt3a and Wnt5a, with Wnt5a acting as the final effector of this modulatory mechanism through β-catenin/TCF7/Lef2 signaling (**Figure 5**). Furthermore, because we used SH-SY5Y cells and observed increases in Ten-3 mRNA expression, our results are in agreement with the recent report of Szemes et al. (2018), which indicates that in neuroblastoma, Wnt3a acts as a differentiation factor (making the cancer less malignant). Considering that Ten-3 expression has also been linked to reduced neuroblastoma

malignancy, we hypothesize that Wnt3a/Wnt5a-mediated Ten-3 expression might be associated with the Wnt3a/Wnt5a context-dependent protective effects against neuroblastoma (Grossmann et al., 2013).

### CONCLUSION

The Ten family has emerged as a fascinating family of proteins because of the critical roles Tens play in the development of the CNS. Importantly, Tens have also been demonstrated to be critical for the maintenance and physiological functioning of adult tissues. However, information regarding the regulatory mechanisms of Tens is completely absent. Moreover, considering that Tens have been linked to important pathological processes, the relevance of novel regulatory mechanisms and the roles of significant cellular pathways in the modulation of Tens should not be overlooked (Südhof, 2018). In this work, we report not only the very first mechanism of the regulation of Ten-3 expression but also that this mechanism involves interplay between the canonical Wnt ligand Wnt3a and the non-canonical Wnt ligand Wnt5a. We believe that these two findings are of great relevance to understanding the roles of Tens, particularly Ten-3, under physiological conditions and to understanding how Ten proteins might interact with molecular pathways that define cell fate in specific contexts, including in different pathophysiological processes such as cancer and neurodegeneration. Therefore, we believe that our work demonstrates, for the first time, the Wntmediated upregulation of Ten-3 through a novel Wnt3a-Wnt5a

#### REFERENCES


complementary signal representing a coupling of the canonical and non-canonical Wnt pathways.

We must highlight that this work constitutes an initial approach to elucidate the involvement of Wnt signaling in the regulation of Tens expression. In this sense, although our results demonstrate a Wnt-Ten interaction, new questions have emerged, including those regarding the mechanisms underlying Wnt5a-induced Ten-3 expression given the dual action of Wnt5a as a canonical and non-canonical activator of Wnt signaling. Further studies will be necessary to properly address these questions, but we believe that our work offers an interesting starting point from which to develop new research aimed at establishing the molecular mechanisms involved in Tens expression.

### AUTHOR CONTRIBUTIONS

SB-C and MM conducted the experiments and revised the final manuscript. JZ, SB-C, and NI designed the experiments and wrote the manuscript.

# FUNDING

This work was supported by grants from the Basal Center of Excellence in Aging and Regeneration AFB 170005, FONDECYT (No. 1160724) to NI and FONDECYT (No. 11170212) to SB-C.

non-canonical Wnt-5a/Ror2/Disheveled pathway activation. Front. Cell. Dev. Biol. 5:47. doi: 10.3389/fcell.2017.00047


the human genome using CART method. BMC Genomics 11:348. doi: 10.1186/ 1471-2164-11-348


system in situ and promotes homophilic cell-cell adhesion in vitro. Development 129, 4697–4705.


**Conflict of Interest Statement:** The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Copyright © 2019 Bastías-Candia, Martínez, Zolezzi and Inestrosa. This is an openaccess article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

digital media

of impactful research

article's readership