EMERGING CELLULAR STRESS SENSORS IN NEUROLOGICAL DISORDERS: CLOSING IN ON THE NUCLEOLUS AND THE PRIMARY CILIUM

EDITED BY : Rosanna Parlato and Kerry Lee Tucker PUBLISHED IN : Frontiers in Cellular Neuroscience

#### 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 978-2-88963-711-9 DOI 10.3389/978-2-88963-711-9

#### 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

# EMERGING CELLULAR STRESS SENSORS IN NEUROLOGICAL DISORDERS: CLOSING IN ON THE NUCLEOLUS AND THE PRIMARY CILIUM

Topic Editors: Rosanna Parlato, University of Ulm, Germany Kerry Lee Tucker, University of New England, United States

Citation: Parlato, R., Tucker, K. L., eds. (2020). Emerging Cellular Stress Sensors in Neurological Disorders: Closing in on the Nucleolus and the Primary Cilium. Lausanne: Frontiers Media SA. doi: 10.3389/978-2-88963-711-9

# Table of Contents


Micaela Lucarelli, Chiara Di Pietro, Gina La Sala, Maria Teresa Fiorenza, Daniela Marazziti and Sonia Canterini

*91 Targeted Depletion of Primary Cilia in Dopaminoceptive Neurons in a Preclinical Mouse Model of Huntington's Disease*

Rasem Mustafa, Grzegorz Kreiner, Katarzyna Kamińska, Amelia-Elise J. Wood, Joachim Kirsch, Kerry L. Tucker and Rosanna Parlato

# Editorial: Emerging Cellular Stress Sensors in Neurological Disorders: Closing in on the Nucleolus and the Primary Cilium

Rosanna Parlato1,2 \* and Kerry L. Tucker <sup>3</sup> \*

*1 Institute of Applied Physiology, University of Ulm, Ulm, Germany, <sup>2</sup> Institute of Anatomy and Cell Biology, Department of Medical Cell Biology, University of Heidelberg, Heidelberg, Germany, <sup>3</sup> Department of Biomedical Sciences, Center for Excellence in the Neurosciences, College of Osteopathic Medicine, University of New England, Biddeford, ME, United States*

**Emerging Cellular Stress Sensors in Neurological Disorders: Closing in on the Nucleolus and**

Keywords: primary cilia, nucleolus, autophagy, cellular stress, cell homeostasis

**Editorial on the Research Topic**

**the Primary Cilium**

#### Edited by:

*Thomas Fath, Macquarie University, Australia*

#### Reviewed by:

*Surya Nauli, University of California, Irvine, United States*

> \*Correspondence: *Rosanna Parlato*

*rosanna.parlato@uni-ulm.de Kerry L. Tucker ktucker2@une.edu*

#### Specialty section:

*This article was submitted to Cellular Neuropathology, a section of the journal Frontiers in Cellular Neuroscience*

> Received: *28 January 2020* Accepted: *04 March 2020* Published: *20 March 2020*

#### Citation:

*Parlato R and Tucker KL (2020) Editorial: Emerging Cellular Stress Sensors in Neurological Disorders: Closing in on the Nucleolus and the Primary Cilium. Front. Cell. Neurosci. 14:64. doi: 10.3389/fncel.2020.00064* The primary cilium and the nucleolus represent signaling hubs that regulate cell homeostasis and stress responses. Primary cilia have long been considered vestigial organelles but are now viewed as sensory antennae that transduce extracellular signals, including mechano- and chemotransduction. The nucleolus is traditionally viewed as the site where ribosomal RNA synthesis and ribosome biogenesis occurs, which is now considered both a sensor and a mediator of the cellular stress response. Both organelles transduce developmental and homeostatic pathways, including Wnt signaling, mechanistic target of rapamycin (mTOR) and autophagy. Both organelles adjust their structure and function in response to changes not only in the extracellular environment but also in the intracellular milieu. Interestingly, neither are typical membrane-bound organelles, and they both adapt their structure in a dynamic fashion, suggesting a crucial role in cell homeostasis. These conceptual similarities, along with the newly discovered impact of deficits in primary ciliary and nucleolar function on neuronal homeostasis, raised the idea of the first collection of reviews and original articles addressing primary cilium- and nucleolus-dependent mechanisms in normal neuronal function and neurological disorders.

The Research Article by Monaco et al. describes a novel method to differentially isolate primary cilia from motile cilia by flow cytometry. Differently from motile cilia, primary cilia express type III adenylyl cyclase (AC3), a primary cilium-localized, cAMP-generating enzyme, and prominin, a glycoprotein typical of neural stem cells. Intriguingly, different populations of primary cilia characterized by the expression of specific receptors in an age-dependent fashion are identified here, validating this method for the investigation of primary cilia-dependent signaling pathways in health and disease conditions.

A second Research Article by Zhou et al. focuses on AC3 for its fascinating genetic association in human studies with autistic spectrum and major depressive disorders. The authors use mouse knockout models of the AC3 gene to look at differences in expression of phosphorylated protein isoforms in the brains of control and AC3-knockout mice. Interestingly, the authors also distributed their data sets by sex, and came up with hundreds of sex-specific phosphoprotein expression patterns, one third of which have shown association with autism.

The mini-review by Sarkisian and Semple-Rowland looks at connections between primary cilia and glioma. Gliomas are tumors arising from glial cell populations in the brain and spinal cord, comprising the majority of malignant brain tumors. The authors focus upon the most aggressive tumor type, the glioblastoma, and examine correlations and causative experiments that have been recorded and performed, respectively, to test whether primary cilia influence tumor development and progression. The data are decidedly mixed and do not seem to deliver a consistent answer whether tumor cells are ciliated, and conversely, the tumor identity of ciliated cells identified in biopsies. This uncertainty is mirrored by functional experiments, showing that either the presence or the absence of primary cilia on tumor precursor cells can promote tumor progression. In a second mini-review, the authors Park et al. make an excellent summary of the multiple, diverse, essential functions that primary cilia have in the formation of the brain, including patterning, layering, and neuronal migration within the forebrain, expansion of precursor pools in the cerebellum, and more subtle effects observed in learning and memory. They touch upon the best-investigated pathways, including Hedgehog, Wnt, mTOR, and they even discuss autophagy in this context.

In the Brief Research Report by Mustafa et al. we present the initial characterization of a mutant mouse lacking primary cilia in dopaminoceptive neurons. These models will be useful to investigate the impact of primary cilia in diseases affecting the dopamine system, such as Huntington's disease (HD). We show that either altered structure or loss of primary cilia is associated with increased mTOR activity in a progressive mouse model of HD. In addition loss of primary cilia results in bigger mutant Huntingtin nuclear inclusions, suggesting a more advanced pathological stage. Future studies should address the disease phenotype at later stages. A second Brief Research Report by Lucarelli et al. looks at the distribution of primary cilia in a murine model of the neurodegenerative Niemann-Pick C1 (NPC1) disease. NPC1 is characterized by biochemical changes in lipid and cholesterol metabolism resulting in dopamine imbalances and corresponding locomotor defects. In their mouse model, the authors observe a decrease in dopamine transporter expression, and also an increase in the number of primary cilia, in the dorsal striatum. In line with this concept, both reports strongly suggest that primary cilia are involved in homeostatic responses to changes in striatal dopamine. Future studies should address the impact of primary cilia of mTOR dysregulation and autophagy in neurodegeneration.

On the other side, the excellent Review by Pfister provides an outstanding overview of the regulation of autophagy by the second stress sensor organelle object of this Research Topic, the nucleolus. The author discusses recent findings on the crosstalk between mTOR signaling and nucleolar activity, and how autophagy represents a response to nucleolar stress in numerous human diseases. She provides an unprecedented outline of nucleolar proteins functioning in autophagy regulation. A second exciting Review by Latonen provides a comprehensive overview of the most recent functions discovered for the nucleolus in neuronal homeostasis. The author identifies the nucleolus as a central hub of cellular proteostasis and focuses on its important new role in cell homeostasis, affected by changing its material properties. The role of nucleolar aggresomes is discussed as a potential mechanism of nuclear aggregate accumulation common to many neurodegenerative diseases, such as HD and amyotrophic lateral sclerosis. This Review also presents the emerging role of long non-coding RNA in ribosomal RNA transcription and stress response.

Although there is no current evidence that primary cilia and nucleoli are interdependent, this Research Topic poses the provocative thought that the structure and function of these two key stress sensors and mediators might be connected. This collection promises to inspire original ideas and innovative studies, to provide a deeper understanding of physiological processes and disease mechanisms, and to identify new strategies for the maintenance of neuronal health and disease treatment.

#### AUTHOR CONTRIBUTIONS

RP and KT drafted the manuscript.

**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 Parlato 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.

# A Flow Cytometry-Based Approach for the Isolation and Characterization of Neural Stem Cell Primary Cilia

Sara Monaco† , Katja Baur † , Andrea Hellwig, Gabriele Hölzl-Wenig, Claudia Mandl and Francesca Ciccolini\*

Interdisciplinary Center for Neurosciences (IZN), Department of Neurobiology, University of Heidelberg, Heidelberg, Germany

In the adult mammalian brain, the apical surface of the subependymal zone (SEZ) is covered by many motile ependymal cilia and a few primary cilia originating from rare intermingled neural stem cells (NSCs). In NSCs the primary cilia are key for the transduction of essential extracellular signals such as Sonic hedgehog (SHH) and platelet-derived growth factor (PDGF). Despite their importance, the analysis of NSC primary cilia is greatly hampered by the fact that they are overwhelmingly outnumbered by the motile cilia. We here take advantage of flow cytometry to purify the two cilia types and allow their molecular characterization. Primary cilia were identified based on immunoreactivity to the marker adenylate cyclase type III (AC3) and differential levels of prominin-1 whereas motile cilia displayed immunoreactivity only to the latter. Consistent with the morphological differences between the two classes of cilia, enrichment of motile cilia positively correlated with size. Moreover, we observed age-dependent variations in the abundance of the two groups of ciliary organelles reflecting the changes associated with their development. The two cilia groups also differed with respect to the expression of signaling molecules, since PDGF receptor (PDGFR)α, smoothened (Smo) and CXC chemokine receptor (CXCR)4 were only detected in isolated primary but not motile cilia. Thus, our novel method of cilia isolation and characterization by flow cytometry has the potential to be extended to the study of cilia from different tissues and organs, providing a powerful tool for the investigation of primary cilia in physiological and pathological conditions.

#### Edited by:

Kerry Lee Tucker, University of New England, United States

#### Reviewed by:

Xuanmao Chen, University of New Hampshire, United States Matthew Sarkisian, University of Florida, United States

\*Correspondence:

#### Francesca Ciccolini

ciccolini@nbio.uni-heidelberg.de

†These authors have contributed equally to this work

Received: 28 September 2018 Accepted: 12 December 2018 Published: 14 January 2019

#### Citation:

Monaco S, Baur K, Hellwig A, Hölzl-Wenig G, Mandl C and Ciccolini F (2019) A Flow Cytometry-Based Approach for the Isolation and Characterization of Neural Stem Cell Primary Cilia. Front. Cell. Neurosci. 12:519. doi: 10.3389/fncel.2018.00519 Keywords: primary cilium, ependymal cilium, subependymal zone, Sonic hedgehog, platelet-derived growth factor

# INTRODUCTION

Primary cilia have an emerging function in the transduction of developmental and homeostatic pathways and their dysfunction is associated with a number of human diseases, collectively referred to as ciliopathies (Berbari et al., 2009; Tobin and Beales, 2009). Primary cilia are present throughout the brain and they are involved in several functions including neurogenesis, migration, autophagy and development (Guemez-Gamboa et al., 2014). The adult subependymal zone (SEZ) is the largest germinal niche in the adult brain, where neural stem cells (NSCs) mainly generate new interneurons for the olfactory bulb throughout adulthood. The apical side of the SEZ is lined with a myriad of motile cilia stemming from ependymal cells and a few primary cilia protruding from NSCs. The two cell types form a characteristic pinwheel structure at the apical SEZ surface in which ependymal cells encircle a NSC (Mirzadeh et al., 2008).

**6**

Like adult NSCs, ependymal cells are generated from radial glia precursors perinatally and continue to develop during the first weeks after birth (Merkle et al., 2004; Spassky et al., 2005). As a consequence, primary cilia extending from radial glia represent the prevailing cilia type before birth. Both ependymal cells and NSCs undergo changes during postnatal aging. In particular, pinwheel structures become rarer in aged mice (Shook et al., 2012) likely due to NSCs losing the apical attachment (Obernier et al., 2018). Ageing also affects the ependymal layer with loss of ependymal cells into the ventricular space (Del Bigio, 2010). Primary and motile cilia are defined by striking morphological differences. The primary cilia axoneme, which is 1–9 µm long (Dummer et al., 2016) and 0.2–0.3 µm wide, is constituted at its core by a ring of nine microtubule pairs (9 × 2 + 0). In contrast, the generally longer motile cilia, whose length is extremely variable in different tissues (Lee, 2013), have a 9 × 2 + 2 core structure with the ring of external microtubule doublets connected by inner and outer dynein arms and an additional central pair of microtubule singlets. Despite being structurally and functionally different both primary and motile cilia express prominin-1, a glycoprotein commonly used to isolate stem and progenitor cells from the developing and adult nervous system, which is selectively localized in membrane protrusions (Weigmann et al., 1997), including cilia (Pfenninger et al., 2007; Coskun et al., 2008). However, whereas prominin-1 expression is a constant feature of ependymal motile cilia, a subset of primary cilia in the SEZ lacks the expression of the glycoprotein at the cell membrane (Codega et al., 2014; Khatri et al., 2014). Modification of the tubulin residues such as acetylation and glycosylation, which increase the stability of the microtubule and the length of the axoneme, are found in both types of cilia. However, the type 3 adenylate cyclase (AC3) localizes to primary cilia only and it is considered to be a marker of primary cilia in all regions of the mouse brain (Bishop et al., 2007).

The beating of ependymal motile cilia contributes to the movement of the cerebrospinal fluid (CSF) in the ventricular system of the brain representing an essential component of a protecting barrier whose integrity is important to maintain the size of the ventricles (Shook et al., 2014) as well as to create concentration gradients for the guidance of migrating neurons (Sawamoto et al., 2006). Underscoring the importance of the ependymal ciliary function, several neurological conditions (Ikeda et al., 2005; King, 2006; Suzuki et al., 2009) such as hydrocephalus (Lee, 2013; Jiménez et al., 2014) and schizophrenia (Palha et al., 2012) have been associated to impaired circulation of the CSF. The flow of the CSF is also disrupted in Huntington's disease, which leads to an increase in the length of ependymal cilia (Keryer et al., 2011). In contrast, primary cilia in radial glia have been associated to the regulation of cell cycle progression (Tong et al., 2014; Izawa et al., 2015). This function reflects the fact that the mother centriole in the basal body of the primary cilia is needed for the mitotic spindle formation. Furthermore, the organelle is essential to transduce signals involved in the regulation of progenitor proliferation such as Sonic hedgehog (SHH) and platelet-derived growth factor (PDGF) signaling (Youn and Han, 2018). Although genes coding for ciliary function are specifically enriched in NSCs and cilia depletion affects NSC quiescence in the SEZ (Beckervordersandforth et al., 2010), not all quiescent NSCs in this region display a primary cilium (Khatri et al., 2014). Moreover ablation of primary cilia affects proliferation only in NSCs of the ventral SEZ (Khatri et al., 2014; Tong et al., 2014). Thus, the function of primary cilia in NSCs is still unclear.

The direct analysis of the expression of signaling molecules in primary cilia would significantly contribute to elucidate their function. Here we describe an innovative flow cytometry-based method to isolate motile and primary cilia from the SEZ and provide evidence for its suitability to analyze the molecular composition of cilia both at the population as well as at the single cilia level.

#### MATERIALS AND METHODS

#### Analysis and Purification of Cilia From the SEZ by Flow Cytometry

#### Deciliation and Immunostaining

All animal experiments were approved by the Regierungspräsidium Karlsruhe and the local authorities of Heidelberg University. Adult mice were killed by CO<sup>2</sup> inhalation followed by cervical dislocation whereas E18 embryos were sacrificed by decapitation. The brain was removed from the skull and the SEZ was dissected in ice-cold dissection medium (150 µM sucrose, 125 µM NaCl, 3.5 mM KCl, 1.2 mM NaH2PO4, 2.4 mM CaCl2, 1.3 mM MgCl2, 2 mM HEPES, 6.65 mM D-(+) glucose; Khatri et al., 2014). The dissected tissue was put in sort medium (NS-A basal medium and L15 medium (1:1), 2% B27 supplement, 1% fetal calf serum (FCS), 0.6% D-(+) glucose, 10 ng/ml huFGF-2, 0.001% DNase) containing APC-conjugated anti-prominin-1 antibody (Becton Dickinson, BD) for 30 min at 4 ◦C and then washed at room temperature with sort medium to eliminate the excess of antibody. Deciliation was performed by combining the two most common methods for cilia detachment: mechanical shear and calcium shock (Mitchell et al., 2009). Briefly, the tissue was incubated in sort medium containing 10 mM CaCl<sup>2</sup> and subjected to mechanical agitation on a rotatory shaker (200 rpm) at 4◦C. After 20 min, the medium, containing cilia, was centrifuged for 1 min at 2,000 rpm at 4◦C to remove cellular debris. The supernatant was collected and immunostained with anti-AC3 antibodies (Santa Cruz) and Alexa Fluor 488-conjugated secondary antibodies for 30 min at 4◦C.

#### Fluorescence Activated Cell Sorting

Cilia preparations were sorted on a FACSAria II cytometer (BD) at single event precision. Sorting gates were set based on florescence levels of samples stained with secondary antibodies only or samples which were not incubated with any antibodies (autofluorescence).

To standardize Forward (FSC) and Side (SSC) scatter values we used as reference beads of known size, i.e., 3 µm (Rainbow Fluorescent Particles, BD) and 6 µm (Accudrop beads, BD). Sorted ciliary particles that based on the reference beads had a size smaller than 3 µm, between 3 µm and 6 µm or greater than 6 µm were separately collected and analyzed. Hereafter they will be referred to as 2 µm (<3 µm), 5 µm (3–5 µm) and 8 µm (≥6 µm) particles.

#### Characterization of the Cilia Types (Primary vs. Motile)

For each size group, double staining with anti-AC3-Alexa 488 and prominin-APC antibodies was analyzed and four gates were set: AC3+/Prom<sup>−</sup> (AC3+), AC3+/Prom<sup>+</sup> (double positive, DP), AC3−/Prom<sup>+</sup> (Prom+), AC3−/Prom<sup>−</sup> (double negative, DN).

#### Functional Characterization

For each size group immunostaining was used for detecting one of the following antigens: Smoothened, PDGFRα or CXC chemokine receptor (CXCR)4. For PDGFRα and CXCR4 staining, the dissected tissue was incubated in sort medium containing anti-prominin-APC (Miltenyi Biotec) and anti-PDGFRα or CXCR4 antibodies conjugated with PE (Invitrogen) for 30 min and washed with sort medium before proceeding to deciliation and staining with anti-AC3 antibody (Invitrogen) manually conjugated with Dylight 488 (Abcam, cat# ab201799). For Smoothened analysis, the tissue fragments were incubated with or without SAG (Cayman Chemical, cat# 11914, 200 nM) in NS-A medium (Euroclone/Biozol) containing 2 mM L-glutamine (Gibco), 100 U/ml penicillin/streptomycin (Gibco), 2% B27 supplement (Invitrogen), 10 ng/ml huFGF-2 (Peprotech) overnight at 37◦C. The samples were labeled with anti-Smoothened antibodies (Novus Biologicals) and antirabbit-APC secondary antibodies for 30 min. The samples were then washed, deciliated and the supernatant was subjected to AC3 staining as described above.

#### Western Blot

For western blot analysis, a specific number of particles was sorted into PBS containing proteinase inhibitor and after ultracentrifugation at 4◦C at 27,000× g for 40 min resuspended in RIPA buffer. The samples were subjected to standard immunoblot analysis with mouse anti-acetylated tubulin antibody (Sigma-Aldrich). Immunoreactivity was quantified using ImageJ and the results were normalized by the number of particles collected in each sample.

#### Whole Mount Immunostaining

Whole mount dissection was performed as previously described (Mirzadeh et al., 2008). The dissected tissue was fixed in 3% formaldehyde/4% sucrose (dissolved in PBS) for 2 h, permeabilized with 0.5% NP-40 for 10 min, incubated in 100 mM glycine to inactivate residual aldehyde groups, and blocked in 5% FCS for 1.5 h. The samples were then incubated with primary antibodies against acetylated tubulin (Sigma Aldrich) and either β-catenin (Santa Cruz), AC3 (Invitrogen) or prominin-1 (kind gift from Denis Corbeil, Technical University of Dresden) in 0.1% NP-40 over night at 4◦C. After washing, the samples were incubated with secondary antibodies and DAPI for nuclear counterstain for 2 h and analyzed using a C2 Plus confocal microscope with NIS software (Nikon) or a TCS SP8 confocal microscope with LAS X software (Leica).

### Antibodies

A list of all primary antibodies used, source, catalog/lot number and concentration is provided in **Supplementary Table S1**.

#### Electron Microscopy

Scanning electro microscopy (SEM) was performed as previously described (Khatri et al., 2014). Briefly, the samples were fixed with 2% glutaraldehyde in 0.1 M sodium phosphate buffer. After washing and postfixation with 2% osmium tetroxide/1.5% potassium ferrocyanide for 1 h, they were washed and dehydrated with an ascending series of ethanol and pure acetone before critical point drying. The samples were then sputter-coated with an 80% gold, 20% palladium alloy and examined with a ULTRA 55 field-emission scanning electron microscope (ZEISS).

#### Statistical Analysis

Statistical significance tests (ANOVA with Tukey's post hoc test or Student's t-test) of at least three independent experiments were calculated using GraphPad Prism and OriginPro 2016. Data represent means ± SEM. P-values are indicated in the figures as follows: <sup>∗</sup>p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001.

### RESULTS

#### A Flow Cytometry-Based Method for Isolation of Primary Cilia From the SEZ

As schematically illustrated in **Figure 1**, we took advantage of flow cytometry to set up a novel approach to identify and isolate primary cilia from the murine SEZ. We performed this method with samples obtained from the SEZ of adult mice, in which motile ependymal cilia are much more abundant than primary cilia, and the corresponding germinal area dissected from embryos at embryonic day (E) 18, when primary cilia represent the vast majority of the cilia. Such age-dependent differences in cilia type proportion are readily visible upon immunostaining of the apical side of the germinal area at the two ages with antibodies directed against acetylated tubulin, which is expressed in both types of cilia (**Supplementary Table S2**; **Figure 2A**). Critical steps of the procedures are the deciliation, i.e., the detachment of cilia from the cell body, the identification and the separate collection of the two cilia types. To promote deciliation we used the traditional method of mechanical shearing in the presence of a high calcium ion concentration (Hastie et al., 1986; Mitchell et al., 2009; Ishikawa and Marshall, 2013). Comparing SEM images of the apical surface of the lateral ventricle wall of E18 (**Figure 2B**) and adult mice (**Figure 2C**), after deciliation or no treatment (Control), showed that the treatment was effective in removing primary but not motile cilia. In untreated embryonic tissue, primary cilia were readily visible (**Figures 2Ba,a**<sup>0</sup> , highlighted in green) whereas, after deciliation, they were largely absent (quantification in **Supplementary Figure S1A**). Sometimes a remaining stump could be detected (**Figures 2Bb,b**<sup>0</sup> , highlighted in green). This suggests that deciliation also gives rise to fragments of different sizes, as previously reported (Mitchell et al., 2009). In contrast, many tufts of ependymal cilia were still present after deciliation, (**Figure 2C**) although their number was significantly reduced (quantification in **Supplementary Figure S1B**). In order to distinguish motile from primary cilia we have taken advantage of differential expression of prominin-1 (Prom) and AC3 in the two cilia types. As summarized in **Supplementary Table S2**, whereas prominin-1 in the adult tissue is expressed more strongly in ependymal cells, AC3 has only been detected in primary cilia (Bishop et al., 2007). Consistent with this, immunostaining of AC3 in whole mount preparations of the SEZ of mice of different ages revealed AC3 expression in primary but not motile cilia and that the immunostaining decreased with age (**Figure 3A**). On the contrary, prominin-1 expression was detected in both cilia groups (**Figure 3B**). However, whereas in motile cilia prominin-1 was expressed at consistently high levels (**Figure 3B**, white arrow), primary cilia displayed high variability in expression levels, ranging from very high (**Figure 3B**, red arrow) to undetectable (**Figure 3B**, green arrow). To address variability in cilia length, ranging between 1 µm and 9 µm (Dummer et al., 2016), we have used beads of known size to set the forward (FSC) and side light scatter (SSC) parameters (**Figures 1A,B**) and monitor changes in particle size during sorting. Although for bigger particles, like for example cells, the first parameter reflects cell size whereas the second indicates complexity (granularity), for smaller particles the measurement of both parameters provides a better detection of changes in size. Independent of size, we detected four types of particles: primary cilia particles immunopositive for either AC3 only (AC3+) or double immunopositive (DP) for both AC3 and Prom, motile cilia particles immunopositive for Prom only (Prom+) and double immunonegative fragments (DN; **Figure 1C**). To further investigate their nature, the four populations of particles were separately collected by FACS, concentrated by ultracentrifugation and then analyzed by

western blot to investigate the expression of acetylated tubulin (**Figure 4**). Particles were first sorted based on differences in size only (**Figure 4A**). After normalization according to the number of sorted particles, quantitative analysis of the western blots showed that although acetylated tubulin was present in all sorted populations, its amount increased with particle size (**Figure 4B**). This corroborates that the sorted material indeed contains cilia. We next sorted the particles based on differential

staining. We collected the same number of particles (200,000) for DN, AC3<sup>+</sup> and Prom<sup>+</sup> particles but only 50,000 of the less abundant DP particles (**Figure 4C**). Densitometric analysis after normalization for the number of particles revealed that AC3+, DP and Prom<sup>+</sup> populations similarly contain at least double the amount of acetylated tubulin than DN particles (**Figure 4D**). This indicates that the three groups of immunopositive particles indeed contain cilia material whereas the DN particles include contaminating material and a few ciliary fragments.

#### Flow Cytometric Analysis of Cilia Markers AC3 and Prominin-1 Allows Identification of Primary and Motile Cilia

To further validate our method for the identification of primary and motile cilia by flow cytometry, we took advantage of the fact that before birth motile cilia are very rare whereas primary cilia in the SEZ decrease with aging (**Figure 2A**). Therefore, we next used flow cytometry to investigate the abundance of the three ciliaenriched fractions defined by the differential antigen expression and subdivided according to size in preparations obtained either from E18 embryos or 8 week or 25 week-old mice (**Figure 5**). As illustrated by representative FACS plots (**Figure 5A**), at each age and for each size we measured the same number of total particles. At all ages examined collected cilia consisted mostly of 2 µm particles followed in decreasing order of abundance by 5 µm and 8 µm particles (**Supplementary Figure S2**). This is probably due to the fragmentation of the cilia during the process of deciliation (illustrated in **Figure 2Bb**<sup>0</sup> ). Quantitative analysis of the number of immunopositive particles in each size-defined population highlighted that the abundance of AC3<sup>+</sup> (**Figure 5B**) and DP (**Figure 5C**) particles greatly declined with age. However,

whereas the number of AC3<sup>+</sup> particles progressively declined with age, the number of DP particles did not vary significantly between samples obtained from E18 embryo and 8-week-old mice, but abruptly decreased thereafter. Moreover, whereas at all ages AC3<sup>+</sup> particles were similarly distributed across the various size groups (**Figure 5B**), DP cilia were mostly found in the 8 µm subpopulation. These observations suggest that the AC3<sup>+</sup> and DP particles represent distinct types of primary cilia, which is consistent with the variability of prominin-1 expression observed in primary cilia by immunostaining (see **Figure 3B**). In contrast to primary cilia, the fraction of Prom<sup>+</sup> particles increased with age, reaching its maximum level in preparations of 25-week-old mice (**Figure 5D**). Moreover, in samples obtained from adult but not E18 mice this group was particularly enriched in the 8 µm fraction, consistent with the fact that motile cilia reach full length after birth and are generally longer than primary cilia.

Taken together, these results show that the number of AC3<sup>+</sup> particles is highest in the embryonic preparations and progressively decreases in samples obtained from older mice in contrast to Prom<sup>+</sup> particles, which increase with age (**Figure 5E**). Since apical primary cilia in the SEZ are present on NSCs, the observed age-dependent decline in primary cilia particles is consistent with previous findings showing that the neurogenic capacity of the SEZ declines with age due to a progressive decrease in NSCs (Maslov et al., 2004; Luo et al., 2006; Bouab et al., 2011; Capilla-Gonzalez et al., 2014).

# Analysis of PDGFRα and CXCR4 Expression in Sorted Particles

To further confirm the identity of the sorted particles as primary or motile cilia, we next investigated the expression of signaling molecules which have been found in primary cilia, such as PDGFRα and CXCR4 (**Supplementary Table S2**; Schneider et al., 2005; Busillo and Benovic, 2007; Christensen et al., 2017; Schmid et al., 2018). To this end we obtained cilia preparations from the tissue of E18, 8 week and 25 week-old mice and analyzed particles of each given size either for PDGFRα (**Figures 6A,B**) or CXCR4 (**Figures 6C,D**) immunoreactivity. This analysis revealed that, independent of the size of the particles, the expression of both receptors drastically decreased after birth and continued to decrease with age in the case of PDGFRα. In contrast, CXCR4 was hardly detectable in cilia obtained from adult mice and no significant difference was observed between the two age groups (**Figures 6A–D**). However, independent of the age there was a positive correlation between the size and the relative expression of both antigens, which may indicate discontinuous expression of the two receptors along the cilia, leading to absent immunoreactivity in some of the fragments. We next investigated similar cilia preparations after triple immunostaining with antibodies specific for prominin-1 and AC3 and either PDGFRα or CXCR4 (**Figures 6E,F**). This analysis revealed that in every sub-population of size and age, almost all the particles displaying PDGFRα (**Figure 6E**) or CXCR4 (**Figure 6F**) immunoreactivity

had also a marker profile of primary cilia, i.e., AC3<sup>+</sup> and DP and not of Prom<sup>+</sup> motile cilia. Indeed, depending on size and age, this fraction represented between 0% and 0.7% of the labeled particles. Moreover, although at every age analyzed the majority of the cilia immunoreactive for either receptor were AC3+, with increasing age the abundance of PDGFRα <sup>+</sup> and CXCR4<sup>+</sup> cilia displaying a DP profile significantly increased, suggesting that not only the number but also the type of primary cilia expressing either receptor changes with age.

FIGURE 6 | Platelet-derived growth factor receptor α (PDGFRα) and CXC chemokine receptor 4 (CXCR4) expression in primary cilia. (A,C) FACS histograms depicting PDGFRα and CXCR4 total expression in particles of different size in E18 embryos (E18), 8 week-old (8w)- and 25 week-old (25w) mice. The gates include PDGFRα-PE or CXCR4-PE positive particles. (B,D) Quantification of the average percentage of PDGFRα <sup>+</sup> and CXCR4<sup>+</sup> particles in at least three independent experiments. (E,F) Quantitative analysis of the co-localization of PDGFRα and CXCR4 with AC3 (green bars), prominin-1 (Prom; red bars) and both AC3 and Prom (blue bars), as percentage of PDGFRα <sup>+</sup> or CXCR4<sup>+</sup> total particles. Values represent the average percentage of particles in each population. Error bars represent SEM. Statistically significant differences are indicated with asterisk (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001).

# Analysis of Smoothened Expression in Sorted Particles

Another protein known to be associated with primary cilia is Smoothened (Smo; **Supplementary Table S2**) which, upon binding of SHH to Patched1, translocates to the primary cilium (Corbit et al., 2005; Eggenschwiler and Anderson, 2007; Rohatgi et al., 2007; Wilson et al., 2009). To activate SHH signaling, we incubated E18 tissue fragments in medium containing

smoothened agonist SAG (Bijlsma et al., 2012; Fan et al., 2014; Lewis and Krieg, 2014; Milenkovic et al., 2015) and analyzed the effect of the treatment on Smo expression in primary cilia (**Figure 7A**). Quantitative analysis of AC3 and Smo co-expression showed that, independent of the particles size, exposure to SAG increased the expression of Smo within the population of AC3<sup>+</sup> primary cilia, especially in the subpopulation of bigger (8 µm) particles (**Figure 7B**). This observation is consistent with the notion that activation of Smo leads to its translocation to the primary cilium and it further confirms the identity of the AC3<sup>+</sup> particles as primary cilia.

#### DISCUSSION

Primary cilia have recently come to the forefront of the scientific interest after the discovery that a wide range of human diseases, collectively referred to as ciliopathies, are caused by defective functions of these organelles (Waters and Beales, 2011). For cilia purification and characterization, we took advantage of flow cytometry, which unlike the traditional approach of purification in sucrose density gradient (Raychowdhury et al., 2005; Mitchell et al., 2009) allows combining high throughput characteristics with highly sensitive analysis of multiple parameters in single particles of interest. Because of the analytical power of flow cytometry we were able to distinguish primary from motile cilia and investigate variability within these cilia types. For the differential identification of cilia we exploited the fact that in the brain AC3 is abundantly expressed in primary cilia, but not in motile cilia of the ependymal layer (Bishop et al., 2007). A possible exception may be represented by the cilia lining the 3rd ventricle, which express AC3 (Chen et al., 2016) however it was not investigated whether these organelles are motile cilia or primary cilia derived from tanycytes (Jarvis and Andrew, 1988; Mirzadeh et al., 2017). Consistent with the notion that it is present in primary but not motile cilia, we showed here that AC3 readily stains the primary cilia of embryonic radial glia but not the motile cilia in the adult SEZ. We here found that the majority of primary cilia do not express prominin-1, which is in keeping with previous observations in the adult SEZ (Codega et al., 2014; Khatri et al., 2014). Although prominin-1 was found widely expressed in embryonic radial glia, the expression was often associated with the apical membrane and not only with primary cilia (Dubreuil et al., 2007). Besides marker expression, several lines of evidence support our conclusion that the approach allows to discriminate primary vs. motile cilia. First, we found that the abundance of the two groups varies during aging reflecting the known changes in the presence of apical progenitors (Maslov et al., 2004) and ependymal cells (Capilla-Gonzalez et al., 2014). Second, the expression of the investigated signaling molecules was only observed in the group of primary cilia particles. In neural precursors several receptor-dependent pathways require a functioning primary cilium (Guemez-Gamboa et al., 2014). Despite the mechanism being still unclear, it is well established that SHH stimulation leads to increased presence of Smoothened in the primary cilium (Corbit et al., 2005; Rohatgi et al., 2007). Our data that PDGFRα localizes to primary cilia is also consistent with previous observations showing that neural precursors depend on primary cilia for transduction of PDGFRα signaling (Carter et al., 2012) and indicate that, like in fibroblasts (Schneider et al., 2005), also in neural precursors PDGFRα localizes to the primary cilium. Both CXCR4 and CXCR7 have been shown to localize to the cilia of developing interneurons (Wang et al., 2011). In the postnatal brain CXCR4 has been shown to be present also in neural progenitors of the SEZ (Tran et al., 2007). Moreover CXCR4-mediated signaling affects homing, (Kokovay et al., 2012), proliferation (Wu et al., 2009) and migration (Imitola et al., 2004; Carbajal et al., 2010) of neural precursors in culture and in vivo. Since the expression of CXCR4 has been shown throughout the ependymal layer of the SEZ (Stumm et al., 2002) our data suggest that the receptor does not localize to the motile cilia. We also found that the expression of all signaling molecules analyzed was particularly pronounced in AC3 single positive but not DP primary cilia. This may reflect the different nature of the two ciliary particles. Indeed, the nature of the population expressing PDGFRα in the adult SEZ is still a matter of debate (Jackson et al., 2006; Chojnacki et al., 2011) and it is known that NSCs more differentiated precursors, but not ependymal cells respond to SHH in the adult SEZ (Ahn and Joyner, 2005). The fact that the expression of both PDGFRα and CXCR4 declines with age especially in AC3 single positive particles further supports the hypothesis of two distinct groups of primary cilia. However, further analysis would be necessary to conclusively address this issue.

Further evidence that we could distinguish between motile and primary cilia is the fact that the deciliation method used here allowed us to enrich for the latter organelle type. An increase in the intracellular concentration of calcium was first used for the detachment of flagella from lower eukaryotes (Watson and Hopkins, 1962; Gibbons, 1965; Hansma and Kung, 1975; Adoutte et al., 1980). In these studies exposure to high calcium ion concentration, often in the presence of detergents, was used to elicit the lysis of the membranous fraction and the detachment of the flagella at the level of the transition zone. Intracellular calcium elevations were shown to trigger shedding of flagella in protists by a mechanism that involves microtubule severing activity and contraction of the fibers in the transition zone, which may also lead to membrane fission (Quarmby and Hartzell, 1994; Sanders and Salisbury, 1994; Lohret et al., 1998). Calcium influx leads to deciliation also of olfactory (Anholt et al., 1986) and primary cilia (Overgaard et al., 2009). The relative inefficacy of the approach on the detachment of the motile cilia likely reflects the different expression of molecules regulating calcium homeostasis and sensing in motile cilia compared to primary cilia as well as flagella (Satir and Christensen, 2007; Delling et al., 2013; Doerner et al., 2015; Inaba, 2015; Lishko and Kirichok, 2015). The method triggers membrane fusion after deciliation, thereby minimizing contamination from cellular material (Satir et al., 1976). Nevertheless, contaminating cellular material including mitochondria was often observed in classical preparations. In addition, osmotic lysis of the membrane leading to the separation of the skeletal axoneme from the ciliary membrane was often observed. In contrast, the particles we purify in the present study likely contain both cilium components as for their purification we used antibodies binding to integral or peripheral proteins of the ciliary membrane and by western blot analysis we observed an enrichment in acetylated tubulin, which is found in the axoneme. However, the fact that we observed acetylated tubulin not associated with ciliary membrane proteins, i.e., in the DN particles, may indicate the presence of ciliary axoneme structures dissociated from the membrane.

#### REFERENCES


Thus, our novel flow cytometry-based approach for the isolation of primary cilia is a new tool for the investigation of these organelles whose function is still poorly understood.

# AUTHOR CONTRIBUTIONS

SM and KB: data collection and analysis, manuscript writing. AH: data collection, manuscript editing. GH-W and CM: data collection. FC: conception and design and overall data interpretation, manuscript writing.

# FUNDING

SM and KB were funded by a Deutsche Forschungsgemeinschaft (DFG) grant (CI 43/11-1). We acknowledge financial support by DFG also within the funding program ''Open Access Publishing,'' by the Baden-Württemberg Ministry for Science, Research and the Arts and by the University of Heidelberg.

# ACKNOWLEDGMENTS

We would like to thank the Electron Microscopy Core Facility (EMCF) of the University of Heidelberg and especially Sebastian Wurzbacher for his support in the preparation of the samples. Confocal images in **Figure 2** were taken at the Nikon Imaging Center at the University of Heidelberg.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fncel. 2018.00519/full#supplementary-material

FIGURE S1 | Quantification of primary and motile cilia on SEM images. (A) Quantification of the number of primary cilia in four SEM fields of SEZ whole mounts of E18 mice before (control) and after deciliation. (B) Quantification of the number of tufts of motile cilia in three SEM fields of SEZ whole mounts of 8 week-old mice (8w) before and after deciliation. Bar graphs show mean ± SEM, p-values are calculated with Student's t-test. (∗p < 0.05, ∗∗∗p < 0.001).

FIGURE S2 | Sorting of particles according to size. Quantification of the relative abundance of sorted particles of different sizes in embryo (E18) 8 (8w) and 25 (25w) week-old mice from three independent experiments. Values represent the average percentage of particles in each population.

TABLE S1 | List of antibodies used in this study.

TABLE S2 | List of markers used in this study to identify and characterize primary and motile cilia.


ciliary accumulation. Chem. Biol. 21, 1680–1689. doi: 10.1016/j.chembiol.2014. 10.013


**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 Monaco, Baur, Hellwig, Hölzl-Wenig, Mandl and Ciccolini. 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.

# Comparative Phosphoproteomic Profiling of Type III Adenylyl Cyclase Knockout and Control, Male, and Female Mice

Yuxin Zhou<sup>1</sup> , Liyan Qiu<sup>1</sup> , Ashley Sterpka<sup>1</sup> , Haiying Wang<sup>2</sup> , Feixia Chu<sup>1</sup> and Xuanmao Chen<sup>1</sup> \*

<sup>1</sup> Department of Molecular, Cellular and Biomedical Sciences, University of New Hampshire, Durham, NH, United States, <sup>2</sup> Department of Statistics, University of Connecticut, Storrs, CT, United States

Type III adenylyl cyclase (AC3, ADCY3) is predominantly enriched in neuronal primary cilia throughout the central nervous system (CNS). Genome-wide association studies in humans have associated ADCY3 with major depressive disorder and autistic spectrum disorder, both of which exhibit sexual dimorphism. To date, it is unclear how AC3 affects protein phosphorylation and signal networks in central neurons, and what causes the sexual dimorphism of autism. We employed a mass spectrometry (MS)-based phosphoproteomic approach to quantitatively profile differences in phosphorylation between inducible AC3 knockout (KO) and wild type (WT), male and female mice. In total, we identified 4,655 phosphopeptides from 1,756 proteins, among which 565 phosphopeptides from 322 proteins were repetitively detected in all samples. Over 46% phosphopeptides were identified in at least three out of eight biological replicas. Comparison of AC3 KO and WT datasets revealed that phosphopeptides with motifs matching proline-directed kinases' recognition sites had a lower abundance in the KO dataset than in WTs. We detected 14 phosphopeptides restricted to WT dataset (i.e., Rabl6, Spast and Ppp1r14a) and 35 exclusively in KOs (i.e., Sptan1, Arhgap20, Arhgap44, and Pde1b). Moreover, 95 phosphopeptides (out of 90 proteins) were identified only in female dataset and 26 only in males. Label-free MS spectrum quantification using Skyline further identified phosphopeptides that had higher abundance in each sample group. In total, 204 proteins had sex-biased phosphorylation and 167 of them had increased expression in females relative to males. Interestingly, among the 204 gender-biased phosphoproteins, 31% were found to be associated with autism, including Dlg1, Dlgap2, Syn1, Syngap1, Ctnna1, Ctnnd1, Ctnnd2, Pkp4, and Arvcf. Therefore, this study also provides the first phosphoproteomics evidence suggesting that gender-biased post-translational phosphorylation may be implicated in the sexual dimorphism of autism.

Keywords: Type III adenylyl cyclase (AC3), primary cilia, major depressive disorder (MDD), autistic spectrum disorder (ASD), phosphoproteomics, delta catenin, sexual dimorphism of autism, gender-biased phosphorylation

#### Edited by:

Rosanna Parlato, University of Ulm, Germany

#### Reviewed by: Victor Faundez, Emory University, United States Francesca Ciccolini,

Universität Heidelberg, Germany \*Correspondence: Xuanmao Chen

Xuanmao.Chen@unh.edu

Received: 30 October 2018 Accepted: 23 January 2019 Published: 13 February 2019

#### Citation:

Zhou Y, Qiu L, Sterpka A, Wang H, Chu F and Chen X (2019) Comparative Phosphoproteomic Profiling of Type III Adenylyl Cyclase Knockout and Control, Male, and Female Mice. Front. Cell. Neurosci. 13:34. doi: 10.3389/fncel.2019.00034

# INTRODUCTION

Primary cilia are tiny microtubule-based, membrane-ensheathed signaling devices present in most mammalian cells (Singla and Reiter, 2006). They depend on a special intraflagellar transport system for trafficking select cargo into and out of the cilium (Rosenbaum and Witman, 2002). Primary cilia are present in virtually every neuron in the brain, although they do not harbor synaptic junctions. Thus far, no ionotropic GABA<sup>A</sup> receptor or glutamate receptors have been identified in neuronal cilia (Qiu et al., 2016; Sterpka and Chen, 2018). Although their physiological function is not well-understood (Louvi and Grove, 2011), defects in neuronal primary cilia lead to obesity, psychiatric diseases, intellectual disability, and neurodevelopmental disorders in humans (Fliegauf et al., 2007). Additionally, neuronal primary cilia have abundant expression of G-protein coupled receptors (GPCRs), such as type 3 somatostatin receptor (Wang et al., 2009; Einstein et al., 2010), type 6 serotonin receptor (Brodsky et al., 2017), and melaninconcentrating hormone receptor 1 (Green et al., 2012). This suggests that neuronal primary cilia depend on metabotropic signal pathways, rather than electrical input from synapses, to modulate neuronal activity. Most, if not all, ciliary GPCRs (Schou et al., 2015) are found to be either Gαs- or Gαi-protein coupled receptors (Qiu et al., 2016; Sterpka and Chen, 2018), which rely on heterotrimeric G-proteins, adenylyl cyclases, and cyclic adenosine monophosphate (cAMP) to send signals to the soma of neurons (Qiu et al., 2016; Sterpka and Chen, 2018).

AC3 represents a key enzyme mediating the cAMP signaling pathway in neuronal cilia (Bishop et al., 2007; Qiu et al., 2016) and is highly expressed in olfactory sensory cilia and in neuronal primary cilia throughout the central nervous system (CNS). It is known that AC3 in olfactory sensory neurons is essential for olfactory signal transduction in the main olfactory epithelium, and loss of AC3 leads to anosmia (loss of smell) (Wong et al., 2000). In the CNS, the physiological function of AC3 is yet to be established, but multiple lines of genetic evidence have associated AC3 with major depressive disorder (MDD) (Wray et al., 2012), obesity (Nordman et al., 2008; Stergiakouli et al., 2014), and autism spectrum disorders (ASD) (Skafidas et al., 2014; Yuen et al., 2017) in humans. Moreover, our previous studies have demonstrated that AC3 ablation in mice leads to pleiotropic phenotypes, including olfactory deficit (Wong et al., 2000; Chen et al., 2012), social interaction deficit (Chen et al., 2016), and depression-like behaviors (Chen et al., 2016). However, thus far it is unknown how AC3 or cAMP generated in neuronal primary cilia regulates signal transduction of central neurons.

Post-translational modifications (PTM) regulate signaling pathway and cellular processes, mediating intracellular communication and neuronal function. Protein phosphorylation is a major type of PTM, which can cause allosteric structure changes of proteins, activation, or inhibition of enzymes, alterations in protein's subcellular localization, and proteinprotein interactions (Johnson, 2009). The major downstream effector protein of cAMP in cells is protein kinase A (PKA), whose activation leads to the phosphorylation of various proteins to propagate the cAMP signaling. We hypothesized that cAMP generated by AC3 locally in neuronal primary cilia can trigger a series of phosphorylation events, thereby modulating the structure and function of many downstream proteins and consequently affecting neuronal function. Therefore, identification of protein phosphorylation affected by AC3 will help delineate AC3-signaling network in CNS neurons. To systematically identify phosphorylation that is modulated by AC3, we employed a mass spectrometry-based quantitative phosphoproteomic approach, a powerful method to elucidate many signal pathways including the cAMP signaling pathway (Gunaratne et al., 2010; Roux and Thibault, 2013; Humphrey et al., 2015), to conduct a comparative phosphoproteomic profiling analysis. In this study, using a high-performance liquid chromatography-tandem mass spectrometry technology (HPLC-MS/MS), we identified thousands of peptides from prefrontal cortical tissues, some of which are differentially phosphorylated in AC3 wild type (WT) and knockout (KO) samples.

To date, although phosphoproteomic profiling analyses have identified a high throughput of phosphorylation sites (p-sites) in a variety of tissues, mouse strains, and different brain regions (Huttlin et al., 2010), virtually no studies have specifically compared phosphoproteomic differences between male and female samples. However, many neurodevelopmental disorders or psychiatric diseases such as MDD, attention deficit and hyperactivity disorder (ADHD) and ASD show a profound sexbias (Halladay et al., 2015). For example, females have a higher risk of MDD than males (Labaka et al., 2018), whereas ASD affects more males than females with a male to female ratio of 4:1 (Kogan et al., 2009). It is unclear what causes the sexual dimorphism of these diseases. To evaluate the possibility that gender may differentially impact protein phosphorylation in the frontal cortex, we specifically compared phosphoproteomic datasets of two genders. This effort led to identification of over 200 proteins, whose phosphorylation were sex-biased. More female-biased phosphopeptides were identified than male-biased. Surprisingly, a high percentage of these targets (31%) are autismassociated proteins/genes, which include Dlg1, Dlgap2, Syn1, SynGap1, Ctnna1, and four delta catenin proteins (Ctnnd1, Ctnnd2, Pkp4, and Arvcf) (Yuan and Arikkath, 2017). Hence, this study provides the first phosphoproteomic evidence suggesting that gender-biased protein phosphorylation may contribute to the sexual dimorphism of autism.

#### MATERIALS AND METHODS

**Supplemental Information** provides additional Methods and Materials including Mice, Immunofluorescence Staining and Confocal Imaging, Western blot methods and detailed statistical methods.

### Tissue Preparation, Protein Extraction, and Phosphopeptide Enrichment

Prefrontal cortex tissues were isolated from 18 to 20 week old mice after euthanization, flash-frozen in liquid nitrogen, and stored in −80◦C until analysis. Samples were then homogenized and lysed by grinding on ice in tissue lysis buffer (50 mM Tris-HCl, pH 8.0, 150 mM KCl, 1% TritonX-100, 0.5 mM PMSF, 0.5 mM EDTA) containing proteinase inhibitor cocktail (Cat. No. 04693159001, Roche, Germany) and phosphatase inhibitor cocktail (Cat. No. 04906837001, Roche, Germany). Lysates were cleared twice by centrifugation at 14K RPM for 20 min at 4◦C. Protein centration was measured with Qubit fluorometer and ∼3 mg of brain lysate from each sample was loaded in 12– 20% gradient SDS-PAGE. In-gel digestion with reduction (final concentration 10 mM dithiothreitol, 56◦C, 1 h) and alkylation (final concentration 55 mM iodoacetamide, 45 min in dark) were carried out at 37◦C for 4 h. Phosphopeptides were enriched by MOAC (TitansphereTM Phos-TiO Kit; GL Sciences Inc., Tokyo, Japan). Briefly, 500 µl of Speed Vac enriched trypsin digested peptides (1 mg/ml) were mixed with 1,000 µl binding solution (25% lactose acid, 60% acetonitrile, 0.3% trifluoroacetic), loaded onto Phos-TiO tip with 3 mg titanium dioxide (TiO2) resin. The resin was washed with 80% acetonitrile and 0.4% trifluoroacetic and eluted with 50 µl 5% ammonium hydroxide followed by 50 µl 5% pyrrolidine. Enriched phosphopeptides were concentrated via Speed Vac for pyrrolidine removal and mass spectrometric analysis.

### Mass Spectrometry and Database Searching

HPLC-MS/MS data was acquired on a LTQ Orbitrap Elite mass spectrometer (Thermo Fisher, CA) coupled to a NanoAccuity UPLC (Waters, MA) in Whitehead MS Facility at MIT (Boston, MA). Peptides were separated by a C18 column at 250 nL/min flow rate and 90-min gradient program. LC-MS data were acquired in an information-dependent acquisition mode, cycling between a MS scan (m/z 395–1,800, resolution 240,000) acquired in the Orbitrap, followed by 10 low-energy CID analysis in the linear ion trap. The centroided peak lists of the CID spectra were generated by PAVA (Guan and Burlingame, 2010) and searched against SwissProt.2017.11.01 Mus Musculus protein database, using Batch-Tag, a program module in Protein Prospector version 5.21.2 (University of California, San Francisco). A precursor mass tolerance of 20 ppm and a fragment mass tolerance of 0.6 Da were used for protein database search with S/T/Y phosphorylation included in variable modifications. Protein hits are reported with a Protein Prospector protein score ≥22, protein discriminant score ≥0.0 and a peptide expectation value ≤0.01 (Chalkley et al., 2005). With similar parameters, false discovery rate (FDR) of all samples was <1.5% when searched against the SwissProt random concatenated database. A threshold of SILP score > 6 was imposed for false phosphorylation site assignment <5%.

#### Label-Free Quantification

Label-free quantification was performed using Skyline ver 4.1.0.18169 via MS1 full-scan filtering with the library generated by ProteinProspector (Cut-off score = 0.95; Precursor charge = 2, 3, 4, 5; Max Miss Cleavages = 1) and the SwissProt Mus Musculus protein FASTA file (Schilling et al., 2012). MS results of three fractions from each sample were combined into one project. Peak areas of identified peptides were generated from Skyline and normalized to the protein concentration of lysate samples. Phosphopeptides with different phosphorylation states, such as mono-phosphorylation and di-phosphorylation, were considered as different entries for quantitation. Identical phosphopeptides from different gel fractions of a same sample were combined for quantitation. Since methionine oxidation can be introduced during sample handling, phosphopeptides with different methionine oxidation states were combined for quantitation. Phosphopeptides with identical sequence in homologous proteins were included in the calculation of protein phosphorylation level for homologous proteins.

#### Bioinformatics Analysis

The phosphoprotein lists generated from ProteinProspector were analyzed by AmiGO 2 (Mi et al., 2017) for pathway/network enrichment. The kinase substrate motif search was performed by web-based Motif-X v1.2 10.05.06 (motifx.med.harvard.edu/motif-x.html) and analyzed basing on the Human Protein Reference Database (www.hprd.org) (Keshava Prasad et al., 2009; Chou and Schwartz, 2011). Phosphopeptides with site assignment confidence level higher than 95% were aligned in Motif-X. The motif widths were adjusted to 6 amino acids from each side of the phosphorylation site. The occurrences were set as 5 and significances were set as 0.000004, which led to a maximal number of motifs and p < 0.001. Protein-protein interaction network analysis was performed by the Cytoscapebased Search Tool for the Retrieval of Interacting Genes/Proteins (STRING, string-db.org) (Szklarczyk et al., 2015). All the proteins with phosphorylation that revealed differences between AC3 KO and WT, or between female and male were searched in PubMed and AutDB (Autism Gene Database, updated in Sept. 2018) (Basu et al., 2009), an autism candidate gene database, to explore possible association between the disease and phosphoproteome.

# Data Analysis

Data analysis and figure constructs were performed with Origin Pro and Graphpad Prism 7 software for Student'st-test, normality test. Phosphopeptides detected in no <3 KO samples but none in any WT control samples (out of n = 8 pair) were considered statistically significantly enriched in AC3 KO sample group (determined by Two Population Proportions Comparison). For phosphopeptides that were detected in both genotypes or genders, label-free quantitation of was used to identify statistically significant (p < 0.05) differences in phosphorylation between KO and WT, or female and male. Phospho-peptides with ln peak area of KO# average peak area of all ≤ 0 showed lower phosphorylation levels of the target peptide site in AC3 KO samples than in WT samples. Conversely, ln peak area of KO# average peak area of all ≥ 0 showed higher phosphorylation levels of the target phosphopeptide site in AC3 KO samples than in WT samples. For label-free quantification between two genders, phosphopeptides detected in 3 of 8 gender pairs covered both of two genders were analyzed. Phosphopeptides with ln peak area of F# average peak area of all ≤ 0 and − ln peak area of M# average peak area of all ≤ 0 had lower phosphorylation levels in female samples than in male samples. In vice versa, ln peak area of F# average peak area of all <sup>≥</sup> 0 and <sup>−</sup> ln peak area of M# average peak area of all ≥ 0 means phosphorylation levels of target phosphopeptide site in female samples were higher than that in males. All peptides spectra presented in the figure and table were reviewed and verified manually. If not otherwise indicated in the figure legends, statistical analysis was a paired student t-test with a two-tailed distribution. n.s. not significant, <sup>∗</sup>p < 0.05, ∗∗p < 0.01, ∗∗∗ p < 0.001. Data were considered as statistically significant if p < 0.05. Data in the graph were presented as mean ± standard error of the mean.

#### Two Population Proportions Comparison

We used "Two Population Proportions" for comparison to set the "3 out of the n = 8 samples" cut-off to determine "a phosphopeptide is enriched in one sample group." Mass spectrometer randomly picks a peptide separated by HPLC to be sequenced and identified. High abundance peptides are detected more frequently. We calculated the p-value by comparing two population proportions between two groups as the following. Specifically, we wanted to test the null hypothesis of p<sup>1</sup> = p<sup>2</sup> against the research hypothesis of p<sup>1</sup> 6= p<sup>2</sup> in the following.

For an individual phosphopeptide, in control group, <sup>p</sup>b<sup>1</sup> <sup>=</sup> X1 n1 = 3 8 ; in AC3 KO group, <sup>p</sup>b<sup>2</sup> <sup>=</sup> X2 n2 = 0 8 ; in the combined group pˆ = X1+X<sup>2</sup> n1+n<sup>2</sup> = 3 <sup>16</sup> ; We calculated the test statistics value according to the z-score z = <sup>r</sup> <sup>p</sup>b1−pb<sup>2</sup> pˆ(1−pˆ) 1 n1 + 1 n2 <sup>=</sup> 1.92, and

used the standard normal distribution, N (0,1) to approximate the p = P N (0, 1) > z = 0.027 < 0.05. Similarly, if a phosphopeptide was detected in 2 control samples and was not detected in any KO samples, then z-score = 1.52, p = 0.0655. If a phosphopeptide was detected in three control samples and detected in 1 KO sample, Z = 1.62, p = 0.0526. Conclusion: If a phosphopeptide was detected in three or more than three control samples (out of n = 8), but in none of 8 KO samples, then the proportion that this phosphopeptide in the Control group was significantly higher than that in the KO group. Similarly, this calculating method was applied to the other three groups (KOs, females, and males).

#### Statistics to Determine "More Female-Biased Phosphorylation Than Male-Biased"

We used two statistical methods "Two Population Proportions" comparison and "Student t-test" to determine if there were "more female-biased phosphorylation than male-biased." Using Two Population Proportions comparison, we calculated a z-score = 9.87, p < 0.0001. Conducting an unpaired Student t-test yielded a p = 0.023 < 0.05. Detailed methods were provided in the **Supplemental Information**.

#### Statistics to Determine "A High Percentage of the Sex-Biased Phosphoproteins Are From ASD-Associated Genes"

AutDB collects a total of 1,053 ASD gene entries. The human genome is estimated to have 20,000 genes. Thus, the 1,053 ASD genes are estimated to represent 5.2% of all human genes in human genome. We have identified 204 sex-biased phosphorylation, among which 32 proteins (15.6%) were listed in the AutDB as ASD genes. The ASD gene percentage is 32/204 = 15.6%. We used "Two Population Proportions" method to compare two groups and produced a Z-score = 6.57 and p < 0.0001. Detailed statistical method was provided in the **Supplemental Information**.

# Statistics of Cross Comparison of Four Groups

We used Pearson's Chi-square test to carry out a cross comparison of four groups (Female KOs, Females WTs, Male KOs, Male WTs). Detailed method as well as the results of cross comparison of 4 groups were provided in the **Supplemental Information**.

### Choosing Appropriate Statistical Method for Data Comparison

Throughout this manuscript, we mostly used two population proportions comparison (3 out of n = 8 cut off) as well as Student's t-test for data comparison. Rationales were provided in the **Supplemental Information**.

# RESULTS

#### AC3 Is Highly Enriched in Neuronal Primary Cilia, but Not in Astrocyte Cilia or Microglia in the Prefrontal Cortex

Because AC3 is associated with MDD (Wray et al., 2012; Chen et al., 2016) and ASD (Skafidas et al., 2014; Yuen et al., 2017), we chose the prefrontal cortex, a brain region important for social behaviors, personality and emotion, and mood state (Siddiqui et al., 2008), in our phosphoproteomics analysis. First, we tested if AC3 is expressed in the prefrontal cortex and if so, in which cell types it is expressed. Immunofluorescence staining images demonstrate that AC3 is highly expressed in neuronal primary cilia (**Figure 1A**), but not microglia (**Figure 1B**) or astrocytic primary cilia (**Figure 1C**) in the prefrontal cortex. These results are consistent with previous reports (Kasahara et al., 2014; Sipos et al., 2018; Sterpka and Chen, 2018), showing that AC3 is primarily confined to neuronal primary cilia in the brain. These data also suggest that AC3 mostly modulates the function of neurons, but not of astrocytes or microglia.

Given the major target protein of cAMP in cells is PKA, which can phosphorylate numerous downstream proteins, we set out to determine if AC3 affects post-translational phosphorylation in neurons in the prefrontal cortex. To circumvent developmental complications, we utilized AC3 floxed:Ubc-Cre/ERT2 KO mouse strain (Chen et al., 2016) to ablate AC3 temporally in adult mice. Ubc-Cre/ERT2 is a tamoxifen-inducible Cre strain with Cre expression driven by the ubiquitin C promoter (Ruzankina et al., 2007). Administration of tamoxifen to the AC3 floxed:Ubc-Cre/ERT2 mice removed more than 95% AC3's immunostaining signal in the prefrontal cortex (**Figure 1F**), whereas vehicle injection had no effect (**Figure 1G**), demonstrating that AC3

not microglia or astrocyte cilia. (A) Co-staining of NeuN (a neuronal marker, red) and AC3 (green) demonstrates that AC3 in primary cilia mostly localize very closely with neuronal soma (red). (B) Co-staining of IBA1 (a microglia marker, magenta) with AC3 (green) demonstrates the absence of AC3 in microglia. Note that green staining generally does not overlap with magenta. (C) Co-staining of GFAP (an astrocyte marker, red) with AC3 (green) shows that most astrocytes do not have AC3 in cilia. (D–G) AC3 was highly expressed neuronal primary cilia in AC3 WT (D,F), but minimally in AC3 inducible KO tissues (E,G). (F,G) are enlarged from (D,E), respectively.

was successfully ablated in the adult mouse brain after tamoxifen injection.

# Mass Spectrometry-Based Phosphoproteomic Analysis Using AC3 WT and KO, Male and Female Samples

To efficiently identify phosphorylation differences in the prefrontal cortex between AC3 KO and WT mice, we utilized a MS-based phosphoproteomic approach to identify phosphopeptides in large scale. Proteins of the prefrontal cortex (isolated from WT, KO, male, and female mice, respectively) were extracted and digested with trypsin. Resultant phosphopeptides were enriched using TiO<sup>2</sup> enrichment column and then subjected to HPLC-MS/MS analysis and database search for identification using UCSF Protein Prospector (**Figure 2**). Additionally, we determined the relative abundance of (p)Ser, (p)Thr, and (p)Tyr residues in the dataset. On average, ∼94% of detected unique peptides were phosphorylated peptides, indicating that the phosphopeptide TiO<sup>2</sup> enrichment was efficient and successful (**Figure 2**). Among all phosphorylated peptides (total 94%), (p)Ser accounted for 73% of total phospho-peptides, (p)Thr 19%, and (p)Tyr 2% (**Figure 2**), which is consistent with previous phosphoproteomic reporting (Olsen et al., 2006; Huttlin et al., 2010).

In total, 4,655 different phosphopeptides were detected from 1,756 proteins (**Table S1**), among which 2,390 (51.4%) were present at least in three out of the 8 biological replicas in AC3 KO group, 2,244 (48.2%) in AC3 WT group, 2,427 (52.2%) in Female group, 2,158 (46.4%) in Male group (**Table S2**). Five hundred and sixty five phosphorylation sites from 322 proteins were detected out of all 16 samples (**Table S3**). In each sample group (n = 4 for each), 62% of phosphorylated peptides were detected in more than 2 of 4 samples, and 27% of phosphorylated peptides were detected in all 4 samples (**Figure 3A**), demonstrating reproducibility of the phosphoproteomic analysis. Note that the same peptide having two p-sites was counted as two different modifications. Histograms of the average of eight pairs of KO/WT MS1 peak area and 8 pairs of female/male MS1 peak area demonstrate that our datasets largely fell in approximate

statistic normal distribution (**Figure 3B**). Our experiments using biological replicas generated a range of 43–62% for peptidelevel repeatability and a range of 53–75% for protein-level repeatability, representing a good range for phosphoproteomics analyses (**Figure 3C**). Forty-six to fifty-two percentages of phosphopeptides were present at least in three out of the 8 biological replicas in one sample group. These numbers were consistent with previous report (35–60%) and verified data solidity (Tabb et al., 2010).

For global phosphoproteomic profiling, the average MS1 peak area of KOs relative to WTs (x-axis), and the average MS1 peak area of males relative to females (y-axis) were constructed into a scatter plot (**Figure 3D**). This shows that the phosphorylation levels of most peptides had no general difference between genotypes or between genders, whereas only some modifications exhibit marked differences between AC3 KOs and WTs, or between females and males (**Figure 3D**). To validate our phosphoproteomic data, we also looked into Western blotting as an orthogonal method, but we were limited by the availability of commercial antibodies. We chose three anti-phosphopeptide antibodies (anti-pCaMK2a/b T286/287, anti-pSyn1 S605, antipERK1/2 T203/Y205 T183/Y185) for Western blot assays to verified their MS data. The anti-pCaMK2 antibody recognizes both the CaMK2a (p)Thr 286. The Western blot signal shows that CaMK2a (p)Thr 286 had no difference between WT and KO samples, and between male and female samples, which were consistent with their MS1 quantification data (**Figure S1**). Two other antibodies (anti-pSyn1 and anti-pERK1/2) yielded similar results (**Figure S1**).

To determine which classes of proteins were enriched in our MS-based phosphoproteomic analysis, the dataset of all phosphopeptides was subjected to Gene Ontology (GO) enrichment analysis (Mi et al., 2017). The GO analysis demonstrated that synaptic vehicle exocytosis proteins, gluconeogenesis, synaptic membrane protein, SNARE complex, tubulin, kinases, and SNAP receptor activity were highly enriched in our dataset, while G protein-coupled receptors, protease inhibitors, extracellular space proteins, ligand-gated ion channels, proteins mediating immune responses, and phagocytosis, cell recognition proteins, transferase, and nuclease activity were under-represented (**Figure S2**).

#### Motifs Matching Proline-Directed Kinases' Substrate Motifs Had Decreased Abundance AC3 KO Samples

cAMP regulates many kinases' activity including PKA and ERK1/2 (Waltereit and Weller, 2003; Sassone-Corsi, 2012). To classify MS-identified phosphopeptides into different motif categories and to determine if AC3 ablation affects the overall

FIGURE 3 | Data quality of phosphoproteomic analysis. (A) Venn-diagram of phosphorylated peptides detected from 4 groups (female KO, female WT, male KO, male WT). In each group, around 62% of phosphorylated peptides were detected in more than 2 of 4 samples, and around 27% of phosphorylated peptides were detected in all 4 samples. (B) Histogram of relative phosphopeptide abundance of KOs compared to WTs (left), and females compared to males (Autism Genome Project et al., 2007). The histogram indicates that the dataset fits into a normal distribution. The X-axis denotes average ratio of phosphopeptide intensity in AC3 KO relative to WT (left), or females relative to males (Autism Genome Project et al., 2007). (C) Peptide and protein repeatability. The 28 random pairs of our biological replicates (8 per groups) showed a range for peptide-level repeatability as 43–62% and a range for protein-level repeatability as 53–75%. (D) Scatter plot of phosphopeptide abundance of KOs relative to WTs, and of females relative to males. (D-a) The X-axis denotes average ratio of phosphopeptide intensity of AC3 KOs relative to WTs. The Y-axis denotes average ratio of phosphopeptide intensity of females relative to males. (b–e) Zoom-in plot of (a) for phosphopeptides that are highly enriched in female WTs (b); female KOs (c); male WTs (d); and male KOs (e). Cut-off lines ln (KO/WT) or ln (female/male) were set to 1, meaning over 2.7-fold increase in phosphopeptide MS1 spectra intensity. The majority of p-sites exhibit similar expression levels, whereas some p-sites have different phosphorylation levels among AC3 KOs and WTs, or males and females.

activity of certain group of kinases, we used Motif-X (see **Figure S3** for detailed method workflow) to determine the abundance of phospho-motifs in our samples, which infers corresponding kinases' activity (Pinna and Ruzzene, 1996). Motif-X is a software tool to extract overrepresented motifs from all phosphopeptide sequences of a dataset (Chou and Schwartz, 2011). Out of the four sample groups, we found that prolinedirected kinase's substrate motifs were the most common motifs in our samples (37% of all p-sites in AC3 KOs and 38% of all p-sites in WTs), largely in line with previous report (Huttlin et al., 2010). The motif [...(p)S-P...] was highly enriched in all four groups of samples (**Figure 4A**). Moreover, we compared the Motif-X sequence logos with Mouse International Protein Index (IPI) database and found that motifs ([Sxxx(pS)P], [Exxxx(pS)P], [(pS)xxSP], [SxPx(pS)P], [(pS)PxxE], [(pS)P]) were enriched in AC3 KO mice, while motifs ([Sxxx(pS)P], [(pS)xxSP], [(pS)P], [SPxx(pS)P], [(pS)PxxE]) were more abundantly present in AC3 WT mice (**Figure 4A**). Notably, the enriched proline-directed kinase-recognized motifs in our dataset matched to the kinase substrate motifs of GSK-3, CDK5, ERK1, ERK2, and mitogenactivated protein kinase-activated protein kinase 3 (Mapkapk3) reported on the Human Protein Reference Database (http:// hprd.org/serine\_motifs). We discovered that the motifs matching proline-detected kinases' recognition sites had significantly higher abundance in AC3 WT than in AC3 KO mice (**Figure 4B**), and the difference was gender unspecific. These data suggest that ablation of AC3 decreases the overall activity of proline-directed kinases in the prefrontal cortex.

#### Differential Expression of Phosphopeptides in AC3 KO and WT Mice

To identify phosphopeptides that were differently expressed in AC3 KO and WT mice, we compared AC3 KO and WT datasets. Phosphopeptides that were expressed differently (p < 0.05 by Two Populations Proportions comparison) in AC3 WT and KOs were pooled in **Tables 1**, **2**. **Table 1** lists 14 phosphopeptides (from 14 proteins/genes such as Map2 and Spast) that were identified only in AC3 WT dataset. **Table 2** contains 35 phosphorylation sites (out of 35 genes including Arhgap20, Arhgap44, Dlgap2, Pde1b, and Sptan1) that were exclusively detected in AC3 KO dataset. In addition to the presence-or-absence detection, phosphopeptides' mass spectra in a dataset also exhibit variant abundance in different samples, reflecting the phosphopeptides expression levels in tissues. To determine if there are quantitative spectra differences between AC3 WT and KO mice, we used Skyline software (Schilling et al., 2012) to conduct label-free quantification. The Skyline-guided quantification was based on the MS1 peptide level. After completion of the MS1 qualification, we used differentially expressed peptides in WTs and KOs to construct a heat map. Comparison of WT data with KO data led to identification of 20 phosphopeptides


Phosphopeptides listed in the table were detected in no <3 times in AC3 WT dataset (n = 8), but never detected in any AC3 KO dataset (n = 8). Proteins on top of the table were annotated in the AutDB (or per PubMed literature, labeled with \* ) to be associated with ASD or neurodevelopmental disorders. The number in brackets indicates the position of amino acid just before the peptide.

(from 19 proteins/genes), which had significantly different phosphorylation levels (p < 0.05 by unpaired Student's t-test) between AC3 WT and KO dataset (**Figure 5A**). Six of them had higher expression in AC3 KOs (**Figure 5A**, Top), whereas 14 of them (from 13 proteins/genes) were more abundantly expressed in WTs (**Figure 5A**, Bottom). The overall phosphorylation levels of all phosphopeptides and the 20 differentially expressed phosphopeptides have no significant difference between AC3 WT and KO datasets (**Figure 5B**). **Figure 5C** shows that three representative phosphopeptides (identified from Cntnap2, Atp2b2, and Ctnnd2) exhibited significant different MS1 peak area between WT and KO dataset.

In total, we identified 65 proteins either with phosphopeptides exclusively present in AC3 WTs or KOs (as listed in **Tables 1**, **2**) or with phosphopeptide abundance having significant differences in WTs or KOs (as shown in **Figure 5A**). Since ADCY3 is associated with MDD (Wray et al., 2012; Chen et al., 2016) and ASD (Skafidas et al., 2014; Yuen et al., 2017), we asked if AC3 affects phosphorylation of other autism-associated proteins. While MDD does not have a candidate gene databank thus far, ASD does have several databases, among which AutDB provides rich resources on ASD candidate genes (http://autism. mindspec.org/autdb/search) (Basu et al., 2009). We used these 65 proteins to search against AutDB as well as PubMed and found that 9 out of 65 were annotated in AutDB and other 7 were annotated in PubMed to be associated with ASD, which include phosphopeptides from Dlgap2, Map2, Ank2, Ctnnb2, Cntnap2, Sptan1, and Spast (**Tables 1**, **2** and **Figure 5A**). To explore if Adcy3 ablation affects ciliary protein phosphorylation, we compared AC3 WT and KO datasets with Ciliome Database, a comprehensive ciliary proteome database maintained by the Leroux lab (http://www.sfu.ca/\$\sim\$leroux/ciliome\_database. htm). Nine of the sixty five of proteins (14%) were listed in the Ciliome Database. Five of them (Fam126a, Ggct, Pde1b, Ctnnd2, and Gapvd1) were highly phosphorylated, whereas 4 (Rabl6, Eif3b, Klc2, and Numbl) exhibited lower phosphorylation in AC3 KOs. We did not find many hits (only 14%) that match with proteins that are listed in the Ciliome Database. We reasoned that our TiO<sup>2</sup> enrichment method harvested phosphopeptides from whole tissue lysate of prefrontal cortex (mostly cytosol), and primary cilia or the basal body only contributed to a little portion of the peptide pool. Still, there are several interesting targets such as Pde1b (Li et al., 2004), Rabl6 (Blacque et al., 2005), Klc2 (Li et al., 2004), and Numbl (Ramamurthy et al., 2014) that have been reported to be involved in ciliary biology.

To determine if some protein interaction network may be associated with the phosphoproteins differentially expressed in AC3 KOs and WTs, the 65 proteins were mapped onto the mouse Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) database (string-db.org) of known

#### TABLE 2 | Phosphopeptides exclusively detected in AC3 KO dataset.


Phosphopeptides listed in the table were detected in no <3 times in KO dataset (n = 8), but never detected in any AC3 WT dataset (n = 8). Proteins on top of the table were annotated in the AutDB (or per PubMed literature, labeled with \* ) to be associated with ASD or neurodevelopmental disorders. The number in brackets indicates the position of amino acid just before the peptide.

protein interactions (**Figure 5D**). STRING is a Cytoscape-based protein-protein interaction network analysis software, which maps and predicts protein-protein interaction (Szklarczyk et al., 2015). AC3 (Adcy3) as well as PKA (Prkaca) were added into the STRING database manually to examine whether these 65 proteins also have direct or indirect interaction with each other. We found 34 of 65 proteins had direct or indirect interaction with each other (**Figure 5D**). Thirteen out of the thirty-four had decreased expression level in AC3 KOs, whereas 19 of them had increased expression in AC3 KOs. Two of them (Map1a and Map2) had both up-regulated and down-regulated phosphopeptides (**Figure 5D**). Notably, the protein interaction map centered around Sptan1 (encoding α-II Spectrin). In the protein-protein interaction network, 45% of connectivity are established between Sptan1 with other genes, removing Sptan1 would dramatically change the network (**Figure S4**). Sptan1 has a bi-phosphorylation site (S1186|T1188|S1190) detected in 4 AC3 KO samples, but in none of 8 WTs (see **Table 2**). Sptan1 is a principal membrane skeleton component and provides a spectrin-actin cytoskeleton interface to integrate signals (Machnicka et al., 2012, 2014). Sptan1 plays a critical role in neurodevelopment and mutations of Sptan1 leads to encephalopathy, intellectual disability and ASD (Syrbe et al., 2017). Note that AC3 is associated with ASD (Skafidas et al., 2014; Yuen et al., 2017) and primary cilia regulate neurodevelopment (Valente et al., 2014).

#### Differential Expression of Phosphopeptides in Male and Female Samples

Males and females differ greatly in cognition, behaviors, and disease susceptibility (Zagni et al., 2016). Remarkably, many psychiatric diseases, such as ADHD, MDD, and ASD, exhibit a sexual dimorphism (Zagni et al., 2016). But to date it is unclear what causes the sexual dimorphism of these disorders. We hypothesized that post-translational phosphorylation may correlate with the sexual dimorphism of ASD. To determine if there are any gender-biased phosphorylations in our samples, we compared female and male datasets. All phosphopeptides exclusively enriched (p < 0.05 by Two Populations Proportions comparison) in males or females are listed in **Tables 3**, **4**. In total, 95 phosphorylated peptides (out of 90 proteins such TABLE 3 | Phosphopeptides exclusively detected in female dataset.


(Continued)

#### TABLE 3 | Continued


(Continued)

#### TABLE 3 | Continued


Phosphopeptides listed in the table were detected in no <3 times in female dataset (n = 8), but never detected in any male dataset (n = 8). Proteins on top of the table were annotated in the AutDB (or per PubMed literature, labeled with \* ) to be associated with ASD or neurodevelopmental disorders. The number in brackets indicates the position of amino acid just before the peptide.

as Atp1a3, Srgap2, and Dlgap2) were detected only in female samples, whereas 26 phosphopeptides (out of 26 proteins such as Ctnnd1, Ctnnd2, Efnb3, and Caskin1) were only found in male samples. Among these p-sites, one striking example is catenin δ-2 (p)Ser7, which was highly enriched in males. This site has been reported in male mice on UniPort database (Huttlin et al., 2010;

#### TABLE 4 | Phosphopeptides exclusively detected in male dataset.


Phosphopeptides listed in the table were detected in no <3 times in male dataset (n = 8), but never detected in any female dataset (n = 8). Proteins on top of the table were annotated in the AutDB (or per PubMed literature, labeled with \* ) to be associated with ASD or neurodevelopmental disorders. The number in brackets indicates the position of amino acid just before the peptide.

UniProt Consortium, 2018). This p-site was detected in six out of eight male samples (3 KOs and 3 controls), but in none of the eight female samples, indicating that catenin δ-2 (p)Ser7 is a male-specific phosphorylate site. Notably, mutations in catenin δ-2 cause autism in female-enriched multiplex autism families (Turner et al., 2015)

Similarly, to compare phosphorylation level differences between females and males, Skyline-based label-free quantification was also performed on female and male datasets. We further found 96 phosphopeptides (out of 88 proteins) having increased phosphorylation levels in females relative to males, whereas only 11 peptides had significantly decreased phosphorylation levels in females (**Figures 6A,B**). Strikingly, 31 proteins (with 34 peptides) out of 88 (∼35%) are associated with ASD, either annotated in the AutDB (Basu et al., 2009) or per PubMed reporting. Thirty-six of phosphopeptides out of the 33 proteins are presented in **Figure 6B**. Note that the overall phosphorylation levels had no differences between male and female samples, while the sum MS1 peak area of the 107 differentially expressed phosphopeptides were much higher in females than in males (**Figure 6C**). Phosphopeptide abundance of four representatives, Rims1, Pacs1, Syngap1, and Ctnnd2 (both associated with ASD) are shown in **Figure 6D**. These data suggest that more proteins in females are highly phosphorylated

area of all data) for male dataset. (B) Heat map of 11 peptides out of 11 proteins having increased phosphorylation levels in males relative to females. 2 phosphopeptides from 2 Autism-related proteins were on the bottom whereas other nine proteins were on the top. (C) The sum spectrum peak areas of all phosphopeptides have no difference in females (n = 8) and males (n = 8) (left). The sum spectrum peak areas of 107 phosphopeptides out of 99 proteins (as listed in A,B) were much higher in females than in males (p < 0.01, by unpaired Student's t-test). (D) Representative plots of four phosphopeptides (from Rims1, Pacs1, Syngap1, and Ctnnd2) of MS1-based quantification of phosphorylation levels in females compared to males.

in the frontal cortex than in males, with a high percentage of them being autism-associated proteins.

It is common that one protein activity can up- or downregulate the activity of its interacting proteins in shared signal transduction pathways (von Mering et al., 2003). To assess if these ASD-related proteins interact with each other, 204 proteins with different phosphorylation levels in males or females (as listed **Tables 3**, **4** and **Figures 6A,B**) were mapped onto the mouse STRING database to examine protein-protein interaction. **Figure 7A** depicts that 82 out of 204 proteins were connected in a tight network, 69 proteins of the 82 had higher phosphorylation in females, three of them (e.g., Ctnnd2) had different phosphorylated peptides with increased phosphorylated levels in both females and males. Further, 63 proteins out of the 204 (∼31%) with different phosphorylation levels in females or males were clearly associated with ASD. Thirty-two out of the sixty three of proteins were autism candidate entries in the AutDB, whereas the other 31 were reported in PubMed. It is noteworthy that all delta catenin proteins (Ctnnd1, Ctnnd2, Pkp4, and Arvcf, with names marked in red **Figure 7B**) exhibit gender-biased modifications. Furthermore, 50 out of the 63 ASD proteins had increased phosphorylation levels in female samples, whereas only eight proteins increased in male and five proteins had different phosphorylated peptides increased in both genders. STRING analysis further demonstrated that 38 of these 63 ASD proteins showed strong protein-protein interactions with each other (**Figure 7B**). These data demonstrate that 204 proteins had sex-biased phosphorylation and 31% of them were ASDrelated genes which showed strong protein-protein interactions, suggesting that autism-related proteins may be highly regulated by post-translational phosphorylation in the female brain.

# DISCUSSION

Malfunctions of primary cilia cause a broad spectrum of diseases in humans, particularly developmental disorders. AC3 is highly enriched in neuronal primary cilia, and not well-expressed in mature astrocyte cilia or microglia in the frontal cortex (**Figure 1**). AC3 is genetically associated with many human diseases including obesity (Nordman et al., 2008; Stergiakouli et al., 2014), ASD (Skafidas et al., 2014; Yuen et al., 2017), and MDD (Wray et al., 2012). It is unknown how AC3 regulates signaling network in central neurons. Besides activating the CNG channel in olfactory sensory neurons, AC3 (or cAMP) in olfactory cilia also sends signals to the cytosol of olfactory sensory neurons (DeMaria and Ngai, 2010) and regulates gene transcription (Serizawa et al., 2003). Given PKA is the major downstream protein of cAMP in most tissues, we postulated that identification of protein phosphorylation modulated by AC3 could help delineate AC3-regulated signaling network in central neurons.

We utilized a high-efficiency method to conduct comparative phosphoproteomics analyses combining TiO<sup>2</sup> phosphopeptide enrichment with HPLC-MS/MS analysis. This approach allows for large-scale identification of phosphopeptides. In our assay, more than 1,500 phosphorylated peptides and 30,000 spectra were detected from each sample (**Figure 3**). We analyzed 16 samples and in total 4655 phosphopeptides were identified from 1756 proteins. We have manually verified all mass spectra presented in **Figures 5**–**7** and **Tables 1**–**4**. This work provided a list of phosphoproteins that help elucidate the function of AC3 in the brain and unravel gender-biased protein phosphorylation. Intriguingly, we identified more gender-biased modifications than those of genotype-biased, suggesting that gender difference is much bigger than genotype differences in the frontal cortex in the phosphoproteomic assay.

To compare phosphorylation levels in proteins that are involved in cAMP signaling pathway, phosphopeptides identified from G-proteins, adenylyl cyclases, PKA, phosphodiesterases (PDE), phosphatases, as well as their regulating proteins were summarized in **Table S4**. It shows that phosphatases (such as Ppp1r1b) were highly phosphorylated, while Gprotein α-subunits were not. cAMP-dependent protein kinases' catalytic subunits were less phosphorylated than their regulatory subunits (**Table S4**). Phosphorylation of two adenylyl cyclases (Adcy5 and Adcy9) have been detected and apparently that Adcy9 is highly phosphorylated in the frontal cortex. Among differentially expressed phosphopeptides between WTs and KOs, one interesting hit was Pde1b (p)Ser18. Pde1b is a calcium/calmodulin-dependent phosphodiesterase that breaks down both cAMP and cGMP (Sharma et al., 2006). Pde1b (p)Ser18 was detected in 4 of 8 AC3 KO mice (one female and three male animals), but in none of the control mice (**Table 2**), suggesting that Pde1b may function downstream of AC3. Additionally, Ppp1r14A (p)Ser19 (protein phosphatase 1 regulatory subunit 14A), which was detected in 4 of 8 WT mice (2 female and 2 male animals), but in none of 8 KO mice (**Table 1**). Ppp1r14A is a C-kinase phosphorylation-dependent inhibitor protein of phosphatase, implicated in cerebellar longterm synaptic plasticity (Eto et al., 2002). In contrast, Ppp1r1b, which is a cAMP/PKA-dependent phosphoprotein and regulates the activity of phosphatase-1, had at least four sites identified (Ser42|45|46, Ser97, Ser130, Thr182|192) in the dataset. However, we did not detect significant differences in phosphorylation levels in these four sites between WT and KO datasets (**Table S4**). Additionally, both Rho GTPase-activating protein 20 (Arhgap20) and Rho GTPase-activating protein 44 (Arhgap44) have p-sites that were exclusively detected in KO dataset (**Table 2**), suggesting these RhoGaps may be regulated by AC3 in the frontal cortex.

Male and female brains differ in many aspects including connectivity (Ingalhalikar et al., 2014), disease susceptibility (Zagni et al., 2016), and gene expression (Trabzuni et al., 2013). Numerous studies have attempted to use different approaches including next-generation sequencing technology to decipher the differences between male and female brains (Trabzuni et al., 2013; Werling et al., 2016; Gershoni and Pietrokovski, 2017). However, current research apparently has overlooked posttranslational modification differences in male and female brains, and phosphoproteomics databanks (UnitProt and Phosphosites) thus far have not collected any data specifically from female brain samples. This study filled this gap and we have conducted systematic phosphoproteomic profiling using prefrontal cortical samples of both genders. Consequently, we have identified 95 phosphopeptides only present in female samples, and 26 phosphopeptides restricted to male samples (**Tables 3**, **4**). Label-free mass spectrometric quantification further revealed that 96 phosphopeptides have higher phosphorylation levels in females, while 11 phosphopeptides are more abundantly expressed in males. We found that phosphorylation of many autism-associated proteins, including but not limited to Dlg1, Dlgap2, Syn1, SynGap1, and Srgap2 (**Figures 6**, **7** and **Tables 3**, **4**), are gender-biased, occurring more in females than in males. As shown in **Figure 6B** and **Tables 3**, **4**, 63 out of 204 phosphopeptides (∼31%) are from autism-associated proteins. Some proteins/genes are not listed in the AutDB, but they directly interact with autism proteins or were found to be associated with autism per literature in PubMed. For example, Caskin1 Ser 728 was only identified in males (**Table 4**). Caskin-1 interacts with neurexins, which bind to neuroligins in the synapses. Both neurexins and neuroligins are strongly associated with autisms (Südhof, 2017). Caskin1 itself is implicated in autism (Daimon et al., 2015).

Among all differentially modified proteins in genders, one protein family is of particular interest. That is the delta catenin family, which contains ARVCF (encoded by Arvcf), catenin δ-1 (Ctnnd1), catenin δ-2 (Ctnnd2), and plakophilin-4 (Pkp4) (Yuan and Arikkath, 2017). All the delta catenin proteins are expressed in the central nervous system and regulate neural development, and they are strongly implicated in neurodevelopmental disorders (Turner et al., 2015; Yuan and Arikkath, 2017). Remarkably, the activity of delta catenin family is highly regulated by post-translational modifications such as ubiquitination and phosphorylation to modulate protein-protein interaction with cadherin, membrane localization, and protein stability (Yuan and Arikkath, 2017). Of note, this family has two regions flanking the ARM domain that are highly enriched with phosphorylation sites (Yuan and Arikkath, 2017). Interestingly, all four delta catenin proteins: ARVCF (Suzuki et al., 2009), catenin δ-1 (Hussman et al., 2011), catenin δ-2 (Turner et al., 2015), and plakophilin-4 (Hussman et al., 2011) are implicated in ASD or neurodevelopmental disorders. Relevant to the sexual dimorphism of autism, mutations of catenin δ-2 has been found to cause severe autism in female-enriched multiplex autism families (Turner et al., 2015). Loss-of-function of catenin δ-2 causes severe developmental phenotypes in animal model. Our MS analysis revealed that catenin δ-2 Ser 7 was restricted to males. Moreover, catenin δ-2 Thr 469 (another p-site) have increased phosphorylation in females relative to males. We have also detected p-sites of the delta catenin family which are also differentially modified in males or females: catenin δ-1 Ser 346 was only detected in males (**Table 4**); ARVCF Thr 643 has decreased phosphorylation in females; PKP4 Ser 220 had higher expression in females (**Figure 6**). Together, these data suggest that the delta catenin family may participate in the regulation of gender-biased posttranslational phosphorylations, consequently affecting neuronal development in the prefrontal cortex. This interpretation is further supported by STRING analysis, showing that these four proteins are connected in a String interaction map with gender-biased phosphorylation (**Figure 7B**).

In summary, this comparative phosphoproteomic profiling has generated several interesting findings: (1) AC3 ablation leads to decreased activity of proline-directed kinases in the frontal cortex; (2) There is a gender-biased phosphorylation in 204 proteins, 31% of which are associated with ASD; (3) Four delta catenin family members, all associated with autism, contain gender-biased phosphorylation sites. Hence, although future work is warranted, this study provides useful phosphoproteomic clues to elucidate the function of AC3 in the CNS. It also presents the first proteomic evidence suggesting that sex-biased post-translational phosphorylation is implicated in the sexual dimorphism of autism.

#### DATA DEPOSITION

Proteome raw and complete datasets (4 groups: AC3 controls, KOs, males and females, 16 mice) have been submitted to ProteomeXchange via the PRIDE-EBI database (PXD012259).

#### REFERENCES


#### ETHICS STATEMENT

All animal-related procedures were approved and conducted in accordance with the guidelines of the Institutional Animal Care and Use Committee of the University of New Hampshire.

#### AUTHOR CONTRIBUTIONS

YZ and XC designed experiments. YZ conducted phosphoproteomic assays, immunostaining, and data analysis. LQ maintained mouse colony and provided support for phosphoproteomic assay. AS contributed to immunostaining. FC provided expertise in phosphoproteomic experiments and data analysis. HW provided expertise in statistical methods. YZ and XC interpreted data and wrote the paper.

#### FUNDING

This study was supported by National Institutes of Health Grants MH105746, AG054729, and COBRE GM113131-7006 to XC; HD093783 to FC. A Cole Neuroscience and Behavioral Faculty Research Award to XC; and UNH Summer TA Research Fellowships to YZ and AS.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fncel. 2019.00034/full#supplementary-material


disorder genes using deep resequencing. PLoS Genet. 7:e1001318. doi: 10.1371/journal.pgen.1001318


**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 Zhou, Qiu, Sterpka, Wang, Chu and Chen. 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.

# Emerging Roles of Primary Cilia in Glioma

#### Matthew R. Sarkisian1,2 \* and Susan L. Semple-Rowland<sup>1</sup>

<sup>1</sup> Department of Neuroscience, McKnight Brain Institute, University of Florida College of Medicine, Gainesville, FL, United States, <sup>2</sup> Preston A. Wells, Jr. Center for Brain Tumor Therapy, McKnight Brain Institute, University of Florida College of Medicine, Gainesville, FL, United States

Primary cilia are microtubule-based organelles that are typically present on cells during the G0 or G1-S/G2 phases of the cell cycle. Recent studies of glioblastoma (GBM) biopsies, a brain tumor that is notorious for its aggressive growth and resistance to treatment, show that many cells in the tumor lack cilia. At this point, it remains unclear whether primary cilia promote or suppress glioma tumorigenesis. In this review, we will discuss the different roles that have been proposed for primary cilia in glioma and how cilia may contribute to the resistance of these tumors to current therapies.

Keywords: glioblastoma, cilium, brain cancer, temozolomide, ciliary signaling

# INTRODUCTION

# Edited by:

Rosanna Parlato, University of Ulm, Germany

#### Reviewed by:

Young-Goo Han, St. Jude Children's Research Hospital, United States Sergio Gradilone, The Hormel Institute, United States

#### \*Correspondence: Matthew R. Sarkisian msarkisian@ufl.edu

Received: 30 November 2018 Accepted: 04 February 2019 Published: 20 February 2019

#### Citation:

Sarkisian MR and Semple-Rowland SL (2019) Emerging Roles of Primary Cilia in Glioma. Front. Cell. Neurosci. 13:55. doi: 10.3389/fncel.2019.00055 Primary cilia are microtubule-based organelles that relay signals to the cell on which they reside and release signals into the cellular microenvironment (Wood and Rosenbaum, 2015; Garcia et al., 2018), but their functions in glioma remain unclear. To our knowledge Schuster (1964) were the first to report the presence of ciliated fibroblasts in a human brain tumor. Despite detecting cilia bearing long axonemes, these investigators concluded that "it is difficult to imagine these cilia fulfilling any useful function, confined and atypical as they are." A few years later, Tani and Ametani (1970) described ciliated cells in human gliomas and concluded that the cilia ". . .might be called vestigial; merely formed because of an inherited tendency of centrioles to form cilia, rather than structures necessarily performing highly specialized functions". The widely held view that these cilia were of little consequence was largely based on the observations that they lacked the central microtubule pair that is a feature of motile cilia. The early 2000s saw renewed interest in mammalian primary cilia and the emergence of a new hypothesis that primary cilia are key signaling cellular organelles that function throughout the body (for review see: Singla and Reiter, 2006; Goetz and Anderson, 2010). With the discovery of new roles for primary cilia in regulating neural stem cell proliferation and migration in embryonic and adult brain regions (e.g., Breunig et al., 2008; Han et al., 2008; Spassky et al., 2008; Han and Alvarez-Buylla, 2010; Baudoin et al., 2012; Higginbotham et al., 2012, 2013; Guo et al., 2015), this hypothesis has triggered new studies designed to understand the function of primary cilia in brain tumors.

Primary cilia are now postulated to be involved in the pathogenesis of various cancers including brain tumors such as medulloblastoma (Han et al., 2009) and choroid plexus tumors (Li et al., 2016); however, the role(s) that these important organelles have in the pathogenesis of glioma, the most common form of brain cancer in adults, are just now beginning to be examined (also see Alvarez-Satta and Matheu, 2018). In this review we focus primarily on the most aggressive form of glioma, glioblastoma (GBM) and discuss the prevalence of ciliated cells in these tumors, the roles that these

cilia may play in controlling glioma tumorigenesis, and findings that suggest the ciliated state of these cells may affect their susceptibility to standard of care GBM (glioblastoma) therapies.

## PREVALENCE OF PRIMARY CILIA IN GLIOMA BIOPSIES AND CELL LINES

A first step toward understanding the functions of primary cilia in tumors is to document the prevalence of these organelles in tumor tissues. For a summary of the status of cilia expression and function in non-glioma cancer subtypes, the reader is referred to a recent review (Eguether and Hahne, 2018). Our group has examined GBM biopsies collected from over 20 patients and have detected primary cilia in all biopsies (Sarkisian et al., 2014). The numbers of ciliated cells in these biopsies ranged from <1 to ∼25% of the population of tumor cells. Cilia were found on cells expressing Ki67, a cellular marker for proliferation, and on cells associated with the vasculature and pseudopalisading necroses, common pathological features of GBM. Closer examination of the cilia using EM analyses revealed the presence of normal appearing primary cilia and of cilia that appeared to have defects affecting the basal body or cilium. A mixture of normal and abnormal cilia in GBM biopsies was also reported by Moser et al. (2014) who examined 7 patient biopsies and reported finding normal cilia in one biopsy and cilia possessing various ultrastructural abnormalities that affected the basal body or axoneme in the remaining 6 biopsies. It is not clear whether these ultrastructural abnormalities affect cilia function. Currently, there are very few methods that allow assessment of cilia function. The best characterized is the sonic hedgehog (SHH) signaling pathway using SHH and subsequently monitor activation of downstream signaling genes (for review see: Goetz and Anderson, 2010). However, it should be noted that primary cilia also host many other signaling pathways including WNT, Notch, Hippo, platelet-derived growth factor (PDGF), insulin-like growth factor (IGF), mechanistic target of rapamycin (mTOR), and multiple G protein-coupled receptors (for review see: Liu et al., 2018; Wheway et al., 2018).

It is important to note that documentation of ciliated cells in human tumor biopsy samples could be significantly affected by the region of the tumor from which the biopsy is obtained, as well as the type of fixative used to preserve the sample and delays in fixation (e.g., Hua and Ferland, 2017). A typical research lab that receives a biopsy may know the general brain region from which the biopsy was obtained and details about the fixation of the tissue. In our experience, immunostaining of many cilia marker proteins in biopsy tissues is significantly improved if the tissues were immediately fixed in 4% paraformaldehyde. Clearly, additional analyses of GBM biopsies will be required to determine the relative frequency of ciliated cells within tumors and to improve the ultrastructural characterization of these cilia.

What cell types are ciliated in GBM tumors? The GBMmicroenvironment can contain multiple cell types that either resemble or derive from glia, microglia, oligodendrocytes, neurons, fibroblasts, and vascular cells (Charles et al., 2012; Rich, 2016). In our analyses of patient-derived xenograft (PDX) tumors, ciliated cells appear to be similarly, distributed in the large core of the tumor and in distal tumor satellite growths (Hoang-Minh L. B. et al., 2016). At this point, the distribution of primary cilia across the various cell types in human GBM in situ remains unknown. To determine which cells in GBM tumors are ciliated, it will be necessary to obtain and section larger regions of human brain that contain these tumors and then analyze the sections using combinations of immunohistochemistry, immuno-electron microscopy (EM), and reconstruction of serial sections.

In vitro analyses of tumor cell lines is a second approach used to study the role(s) that cilia might have in regulating tumor cell biology. Moser et al. (2009) performed the first immunocytochemical and quantitative EM analyses of various GBM cell lines (U-87 MG, T98G, U-251 MG, U-373 MG, and U-138 MG) and found that these cells rarely gave rise to cilia, or if the cells were ciliated, the cilia were often ultrastructurally abnormal. These particular GBM cell lines have fallen out of favor with many neurooncology researchers in part because the DNA profiles of the cell lines differ from those of the original tumor cells (Allen et al., 2016). It is unclear how these genetic changes might affect ciliogenesis. In view of this, we have studied ciliogenesis in five different recently derived human and mouse primary GBM cell lines and have found that approximately 5–30% of the cells across these cell lines were ciliated and that the cilia were ultrastructurally normal and stained positively for proteins known to localize to the ciliary axoneme and basal body (e.g., IFT88, ARL13B, SMO, GLI3, ADCY3, gamma and acetylated alpha tubulin, and PCM1; Sarkisian et al., 2014; Hoang-Minh L. et al., 2016; Hoang-Minh et al., 2018).

Can ciliogenesis be induced in GBM cells? Serum withdrawal is one way to induce differentiation and ciliogenesis (Santos and Reiter, 2008); however, we and others have been unable to stimulate ciliogenesis in cultured GBM cells using serum withdrawal (Moser et al., 2009; Sarkisian et al., 2014). These observations suggest that it may not be possible to induce ciliogenesis in glioma cells that if true may explain why many of the commonly used GBM cell lines studied in vitro typically lack cilia. Factors that may contribute to the low numbers of ciliated cells present in various cell lines, include structural cilia defects, the rapid turnover of the cultured cells, and heterogeneity of the cells with regard to their ability to generate or retain cilia. GBM growth is aggressive and so it is possible that the rapid turnover of cells within these tumors narrows the window of time during which cilia would be present. Alternatively, it may be that only a small fraction of cells in the tumor are capable of growing cilia. We examined this latter possibility by isolating cell clones from two PDX cell lines that normally display ∼10–25% ciliated cells at any given time and found that most of the clones that we isolated gave rise to ciliated progeny (Hoang-Minh L. B. et al., 2016). This finding indicates that even though ciliation was relatively low, most of the cells in these cell lines were capable of giving rise to ciliated daughter cells.

In summary, the consensus among GBM tumor biopsy and cell line studies indicates that anywhere from ∼<1 up to ∼30% of the cells in glioma biopsies and in these cell lines are ciliated at any given time. Future studies that characterize ciliated glioma

lines should make reference, if possible, to the frequency of ciliated cells in the biopsy from which they were derived. If we are able to associate patient outcomes with the numbers and functions of ciliated cells within GBM tumor biopsies, then it may be possible use this information to better inform patient prognoses and treatments.

#### CILIA AND GLIOMAGENESIS

Cilia are organelles typically associated with differentiated cells but are also assembled by dividing cells. In dividing cells, cilia are assembled by the mother centriole during G1 and can persist throughout the cell cycle but disappear during mitosis (Ford et al., 2018). Because cilia are intimately involved in cell division, it is possible that mutations that disrupt ciliogenesis could promote tumorigenesis as a result of a loss of cell cycle control (Plotnikova et al., 2008; Basten and Giles, 2013). In this section we will briefly review research data that support diametrically opposed roles for cilia in controlling tumor cell proliferation in glioma.

Recent studies of the lysophosphatidic acid receptor 1 (LPAR1) and cell cycle-related kinase (CCRK) and its substrate, intestinal cell kinase (ICK), suggest that proliferation of normal astrocytes and glioma cells is enhanced in cells that have either lost or have not synthesized primary cilia. The cilia of normal human astrocytes contain elevated levels of the LPAR1 (Loskutov et al., 2018), a receptor whose downstream signaling cascade activates the G-protein, Gα12/ Gαq (Goldsmith et al., 2011). Loskutov et al. (2018) found that proliferation can be induced in immortalized human astrocytes lacking primary cilia by activating LPAR1 signaling, signaling that is normally limited in ciliated cells because Gα12 and Gαq are excluded from the cilium. The increase in proliferation of the deciliated cells was found to be due to redistribution of LPAR1 to the plasma membrane of the cell where it was able to actively signal through association with Gα12/ Gαq. Furthermore, they found that treatments of deciliated astrocytes and of intracranial tumors in a mouse model of GBM with a small molecule inhibitor of LPA signaling significantly reduced proliferation of the astrocytes and growth of the intracranial GBM tumors, respectively. The results of this study support the idea that one function of primary cilia is to limit GBM proliferation, and that loss of tumor cell cilia may lead to the redistribution of LPAR1 to the plasma membrane and lead to cell proliferation as a result of increased LPAR1 signaling.

In addition to LPAR1 signaling, CCRK and its substrate, ICK, have been linked to the regulation of ciliogenesis and proliferation of tumor cells (Tian et al., 2012). Yang et al. (2013) found that knockdown of CCRK in cultured U251 GBM cells increased the numbers of ciliated cells in the cultures from ∼2 to 8% and slowed proliferation of the cells. They also found that depleting CCRK in NIH3T3 fibroblast cells increased levels of ICK at the tips of the cilia and prevented the cells from reentering the cell cycle. Our group has also found that blocking ciliogenesis in one PDX tumor cell line by expressing a dominant negative form of Kif3a in the cells accelerated proliferation of these cells and their tumorigenic capacity in a PDX model (Hoang-Minh L. B. et al., 2016). Collectively, the results of these studies indicate that deciliation of transformed GBM cells may lead to loss of control of the cell cycle and increased tumor growth.

Interestingly, there is evidence that tumor cells may actively repress ciliogenesis thereby promoting tumor growth. For example, an increase in the levels of the transcription factor EZH2 in melanoma cells, which normally suppresses expression of genes regulating ciliogenesis, can lead to WNT/β-cateninmediated tumorigenesis (Zingg et al., 2018). Notably, inhibition of EZH2 in GBM with elevated levels of EZH2 has been reported to suppress growth of these tumors (Jin et al., 2017). Thus, future studies should examine how EZH2 affects ciliogenesis in GBM.

There are also studies that suggest that cilia may promote tumor cell proliferation. Cells throughout the brain are ciliated (Fuchs and Schwark, 2004; Bishop et al., 2007; Guemez-Gamboa et al., 2014; Sarkisian and Guadiana, 2015) and it appears that cilia on progenitor cells may activate distinct signaling pathways that regulate cell proliferation. For example, in the hippocampal dentate granule zone of developing and adult brain, the cilia of neuronal progenitor cells mediate SHH signaling thereby promoting neurogenesis (Breunig et al., 2008; Han et al., 2008). Interestingly, a significant number of gliomas also respond to SHH (Dahmane et al., 2001; Bar et al., 2007; Clement et al., 2007; Gruber Filbin et al., 2013; Morgenroth et al., 2014). In a study of a PDX cell line reported to be responsive to SHH, we found that SHH only promoted proliferation of the tumor cells if they were ciliated or were capable of forming cilia, and that proliferation was blocked if ciliogenesis was inhibited using a dominant negative form of KIF3a or if SMO function in cilia was inhibited using cyclopamine (Hoang-Minh L. B. et al., 2016). We also found that CRISPR/Cas9-mediated suppression of PCM1, a protein that localizes to centriolar satellites and is required for ciliogenesis, inhibited the proliferation of two GBM cell lines apparently through enhancement of apoptotic cell death (Hoang-Minh L. et al., 2016).

Glioblastoma cells may also release factors from their cilia that could influence tumorigenesis. Recently, we used piggyBac transgenesis to label GBM cilia with Arl13b:GFP that allowed us to use live imaging to monitor the cilia (Hoang-Minh et al., 2018). We observed that a small fraction of GBM primary cilia released vesicles from their distal tips. We also found that culture media conditioned by ciliated GBM cells enhanced cell proliferation of GBM cells while media conditioned by cultures of GBM cells lacking cilia did not, an observation that suggests cilia themselves may release factors that enhance cell proliferation. The cells that appeared to release ciliary vesicles were Ki67<sup>−</sup> suggesting that quiescent glioma cells may release factors into the microenvironment that could influence tumor growth. Our findings are noteworthy because they are consistent with recent reports that show primary cilia tip excisions occur in multiple non-cancer cell lines and that these excision events appear to be a mechanism that can drive quiescent cells back into the cell cycle (Nager et al., 2017; Phua et al., 2017; Ford et al., 2018). These observations indicate that examination of the content and the mitogenic activities of glioma-derived primary cilia vesicles may provide insight into mechanisms underlying tumor growth or recurrence.

FIGURE 1 | Possible influences of primary cilia on glioma cell proliferation and the response of these cells to standard-of-care therapies. The upper panels illustrate three ways in which cilia could impact glioma cell proliferation. (A) Loss or malformation of the cilium could alter signaling pathways in the cell leading to dysregulation of the cell cycle. (B) Cilia signaling may actively promote cell proliferation (e.g., in response to SHH) or may, through vesicle release, enhance tumor growth, and proliferation. (C) Cilia may not have an impact on cell proliferation. The lower panels illustrate three ways cilia could influence the response of glioma to therapy with the caveat that the type of therapy, e.g., chemotherapy versus irradiation, may shape the cell's response differently. (D) Loss or malformation of the cilium could have one of two very different effects: it could increase the sensitivity of the cell to therapy under conditions in which cilia normally support cell survival, or it could increase the resistance of the cell to therapy if cilia-dependent signaling normally activates cell death pathways. (E) Cilia signaling could increase cell resistance to therapy by triggering cell survival pathways through either autocrine or paracrine signaling. (F) Glioma cells may have evolved so that their response to therapy occurs independently of cilia.

### CILIA AND RESISTANCE OF TUMORS TO THERAPY

The standard of care therapies for GBM are surgery, temozolomide (TMZ) chemotherapy, and gamma-irradiation. Very little is known about how the presence or absence of cilia on glioma cells affects the efficacies of these treatments. We found that some ciliated cells express ZEB1, a transcription factor that promotes glioma initiation, invasiveness, and resistance to TMZ chemotherapy (Siebzehnrubl et al., 2013; Sarkisian et al., 2014). Using two PDX cell lines in which we inhibited cilia formation using either CRISPR/Cas9 ablation of either PCM1 or KIF3a, or of both, we found that the loss of either KIF3A or PCM1 in these cells was associated with an increase in the cell's sensitivity to TMZ exposure (Hoang-Minh L. et al., 2016). Depletion of both KIF3A and PCM1 did not increase the sensitivity of the cells to TMZ above that observed in cells in which only one protein was ablated. These results suggest that cilia signaling may contribute to the survival of cells exposed to TMZ chemotherapy. Additional studies will be needed to determine whether this phenomenon is consistently observed in ciliated PDX cell lines and in intracranial models of GBM.

The results of a recent study of multiple types of drugresistant cancer cell lines (e.g., rhabdoid tumor, non-small cell lung carcinoma, and lung adenocarcinoma) suggest that drug resistance in these cell lines is accompanied by an increase in ciliogenesis and cilia signaling as measured by SHH pathway activation (Jenks et al., 2018). These authors found that increased cilia length was sufficient to confer drug-resistance and that blocking either ciliogenesis or cilia signaling pathways was able to reverse drug-resistance. It would be worthwhile to determine whether these mechanisms are active in GBM cells. It is noteworthy that stimulation of ciliogenesis in cultured U251 GBM cells by knocking out CCRK also promoted the localization of members of the SHH pathway to cilia, an effect that could possibly enhance the signaling capacity of these structures (Yang et al., 2013). Similarly, we reported that Arl13b overexpression promotes SMO localization to GBM cilia in the absence of SHH, and that high levels of ARL13B and SMO expression in the tumors of glioma and gastric tumor patients was positively correlated with shortened post-diagnosis survival in these patients (Shao et al., 2017; Hoang-Minh et al., 2018). Thus, it is possible treatments or conditions that stimulate glioma cell ciliogenesis could induce drug resistance in these cells and lead to tumor recurrence.

# SUMMARY AND FUTURE DIRECTIONS

Taken together, the roles or primary cilia in GBM may depend on the relative abundance of ciliated cells in these tumors and the signaling capacities of the cilia. So far, data from existing experiments indicate that glioma cilia may have dual roles,

either restricting or promoting gliomagenesis (**Figure 1**). If we can obtain high quality tumor biopsies, it may be possible to determine the numbers of ciliated cells in specific tumors and what signaling pathways may be active within these tumors. The question is whether this information could be useful in predicting how GBM tumors would respond to therapies employing SHH pathway inhibitors or chemotherapy/radiation. The hypotheses raised in this review require further testing in PDX models. Identification of signaling pathways activated by glioma cell cilia may point to new strategies to slow or inhibit the aggressive growth of these tumors.

#### REFERENCES


# AUTHOR CONTRIBUTIONS

MS and SS-R wrote the manuscript.

# FUNDING

This work was supported in part by funds from an American Brain Tumor Association Discovery Grant supported by an Family Foundation (MS), and a University of Florida Foundation Fund (SS-R).

in PTEN-deficient glioblastoma: new therapeutic opportunities. Nat. Med. 19, 1518–1523. doi: 10.1038/nm.3328



**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 Sarkisian and Semple-Rowland. 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.

# Phase-to-Phase With Nucleoli – Stress Responses, Protein Aggregation and Novel Roles of RNA

Leena Latonen\*

Institute of Biomedicine, University of Eastern Finland, Kuopio, Finland

Protein- and RNA-containing foci and aggregates are a hallmark of many age- and mutation-related neurodegenerative diseases. This article focuses on the role the nucleolus has as a hub in macromolecule regulation in the mammalian nucleus. The nucleolus has a well-established role in ribosome biogenesis and functions in several types of cellular stress responses. In addition to known reactions to DNA damaging and transcription inhibiting stresses, there is an emerging role of the nucleolus especially in responses to proteotoxic stress such as heat shock and inhibition of proteasome function. The nucleolus serves as an active regulatory site for detention of extranucleolar proteins. This takes place in nucleolar cavities and manifests in protein and RNA collections referred to as intranucleolar bodies (INBs), nucleolar aggresomes or amyloid bodies (A-bodies), depending on stress type, severity of accumulation, and material propensities of the macromolecular collections. These indicate a relevance of nucleolar function and regulation in neurodegeneration-related cellular events, but also provide surprising connections with cancer-related pathways. Yet, the molecular mechanisms governing these processes remain largely undefined. In this article, the nucleolus as the site of protein and RNA accumulation and as a possible protective organelle for nuclear proteins during stress is viewed. In addition, recent evidence of liquid-liquid phase separation (LLPS) and liquid-solid phase transition in the formation of nucleoli and its stress responses, respectively, are discussed, along with the increasingly indicated role and open questions for noncoding RNA species in these events.

#### Edited by:

Rosanna Parlato, University of Ulm, Germany

#### Reviewed by:

Stephen Lee, University of Miami, United States Miguel Lafarga, University of Cantabria-IDIVAL, Spain

> \*Correspondence: Leena Latonen leena.latonen@uef.fi

#### Specialty section:

This article was submitted to Cellular Neuropathology, a section of the journal Frontiers in Cellular Neuroscience

> Received: 09 January 2019 Accepted: 08 April 2019 Published: 26 April 2019

#### Citation:

Latonen L (2019) Phase-to-Phase With Nucleoli – Stress Responses, Protein Aggregation and Novel Roles of RNA. Front. Cell. Neurosci. 13:151. doi: 10.3389/fncel.2019.00151 Keywords: nucleoli, stress responses, protein aggregation, amyloidosis, proteasome inhibition, non-coding RNA

# INTRODUCTION

Nucleoli are the site of ribosome biogenesis. They are formed in nuclei around tandem head-to-tail gene repeats of ribosomal DNA (rDNA) in the so called nucleolar organizing regions (NORs). In human cells, NORs are located in the short arms of five acrocentric chromosomes, and their size ranges from 50 kb to >6 Mb (Mangan et al., 2017). The nucleoli are initiated upon and structurally depend on active transcription of rDNA. The nucleoli are dispersed during mitosis as the rDNA transcription is halted. During telophase, the rDNA transcription resumes, and the nucleoli begin to reform as small nucleoli around individual NORs. As the cell cycle progresses, nucleoli fuse, forming larger, mature nucleoli containing multiple NORs (Hernandez-Verdun, 2011; **Figure 1**).

In human cells, the mature nucleoli are associated with perinucleolar heterochromatin (PNH), DNA sequences located distal and proximal to NORs on the acrocentric chromosomal

**46**

arms (McStay, 2016), which is likely to contribute to positioning of the nucleoli to the 3D context in the nuclei. Currently, the sequences of the acrocentric arms are missing from human genome drafts. Yet, what is known is that the sequences on the centromeric side of rDNA are heavily segmentally duplicated and likely do not contain NOR regulatory elements (Floutsakou et al., 2013; Mangan et al., 2017). The telomeric sides of NORs contain regions called distal junctions (DJs). Their sequences are shared between the acrocentric chromosomes and dominated by around 100 kb inverted repeats and seem to have a complex chromatin structure (Floutsakou et al., 2013; Mangan et al., 2017). DJ sequences have been suggested to anchor rDNA to the PNH (Mangan et al., 2017). Other anchors for the spatial positioning of the nucleoli are intermediate filament proteins, especially lamins A/C, B1 and B2, that connect the nucleoli to nuclear matrix and contribute to maintaining nucleolar structure and functions (Martin et al., 2009; Louvet et al., 2014; Matsumoto et al., 2016; Buchwalter and Hetzer, 2017; Sen Gupta and Sengupta, 2017).

The nucleoli belong to a group of membraneless organelles (MLOs), and as such, they are dynamic structures with highly mobile constituents that can diffuse in and out to the nucleoplasm. Recently, the role of liquid-liquid phase separation (LLPS) in formation of MLOs has been increasingly recognized (Shin and Brangwynne, 2017; Sawyer et al., 2018). LLPS has a role in the formation and internal organization of the nucleoli to functional substructures (Feric et al., 2016). The current view thus holds that formation of the nucleoli is a combination of both active recruitment of factors and LLPS.

The nucleoli have a tripartite structure (**Figure 1**) with the three substructures functionally separate. The fibrillar centers (FCs) contain non-transcribed rDNA and rDNA chromatin associated factors. The rDNA transcription occurs at the interface between FCs and dense fibrillary component (DFC), in the latter of which occurs the early processing of precursor ribosomal RNA (rRNA). Late processing of rRNA and assembly of ribosome units takes place at the granular component (GC), surrounding the FCs and DFCs. Interestingly, yeast and other lower eukaryotes lack FCs, which may be connected to the closed nuclear division and intactness of the nucleoli through the cell cycle (Thiry and Lafontaine, 2005).

The tripartite nucleolar structure in human cells depends on the active transcription of rDNA, as several studies have shown inhibition of rDNA transcription by RNA polymerase I (RNApolI) disperses the nucleoli (reviewed in Grummt, 2013). The start of rRNA transcription has long been thought to be the initiating event for nucleolar reformation at the end of mitosis. This view was challenged by Dousset et al. (2000), who's work indicate that postmitotic nucleologenesis results from direct recruitment of processing factors and pre-rRNAs to UBF-associated NORs before or at the onset of rDNA transcription. This is followed by fusion of prepackaged prenucleolar bodies into the nucleolus, suggesting that pre-ribosomal ribonucleoproteins synthesized in the previous cell cycle may contribute to nucleolar formation at the end of mitosis (Dousset et al., 2000).

# NUCLEOLAR CONTENTS AND LLPS

The nucleolus is packed with protein – protein density in nucleoli is approximately double of that of the nucleoplasm (Handwerger and Gall, 2006). Although very dense, the nucleolus is also very dynamic: many nucleolar proteins are constantly moving between the nucleolus and the nucleoplasm (Leung and Lamond, 2003; Hernandez-Verdun, 2006; Sirri et al., 2008). The proteome of the nucleolus before and after stress is well described (Andersen et al., 2005; Moore et al., 2011). Most nucleolar molecules function in transcription and different maturation steps of rRNA (Andersen et al., 2005). However, there are at least dozens, if not hundreds, of nucleolar proteins with no apparent role in the formation of ribosomes (Andersen et al., 2005). Recently, it has become clear that the nucleolus contributes to biogenesis of multiple ribonucleoprotein particles, and the regulation of cellular events such as mitosis, the cell-cycle, and responses to several types of stress (Boisvert et al., 2007; Boulon et al., 2010; Lindström and Latonen, 2013).

RNA content of the nucleoli is not fully described. The well-recognized components, such as rRNA and snoRNA, are well known for their functions in ribosome production, but other non-coding components are not comprehensively described. In addition to the traditionally viewed roles in processing pre-rRNA and formation of ribosomal particles, nucleolar RNA is increasingly seen to have a role through contributing to nucleolar formation through promoting LLPS (Sawyer et al., 2018). MLOs typically harbor specific RNAs and intrinsically disordered, multivalent hub proteins, both contributing to the LLPS characteristics (Sawyer et al., 2018). It has been shown that the disordered domains in FBL and NPM (key components of DFC and GC, respectively) are required for droplet formation, and that RNA recognition motifs are required for maintaining phase separation (Brangwynne et al., 2011; Mangan et al., 2017). The sequence-encoded features of these proteins influencing their LLPS behavior also lie behind nucleolar compartmentalization, driven by different biophysical properties of the droplets, especially surface tensions (Feric et al., 2016). While specific RNAs themselves may be capable of phase separation as in the case of e.g., extended repetitive RNA motifs in clinical disorders, LLPS for MLOs is viewed to be driven more by RNA-protein interactions than RNAs as such (Sawyer et al., 2018). Long RNA molecules may potentially interact with several other proteins and RNAs simultaneously, favoring and strengthening the interactions between droplet-forming proteins. In addition, RNA-protein ratio and RNA multivalency may also be critical factors for MLO LLPS (reviewed in Sawyer et al., 2018). Interactions between NPM and rRNA promote LLPS in the nucleolar formation and supports the idea of active rDNA transcription spatially and temporally coordinating with critical, intrinsically disordered region (IDR)-containing LLPS drivers (Mitrea et al., 2018; Sawyer et al., 2018). It is likely that, in addition to rRNA, there are other contributing RNA species for LLPS in nucleolar formation yet to be identified. Likely candidates are at least the lncRNAs coded by the DJ regions (Mangan et al., 2017).

Changes in relative levels of the RNA components, likely to have profounding effects of nucleolar activity as well as organization in terms of LLPS, are not well known. Nucleolar rRNA is so abundant compared to other RNA species in the cell in general that rRNA sequences are often excluded in sequencing assays. The sequence repetitivity and lack of the reference genomes for rDNA areas makes it currently infeasible to align rRNA sequences for most quantitative expression analyses via next generation sequencing approaches. Most importantly, the lack of NORs and adjacent regions from genomic assemblies hampers the expression analyses of these areas and studies for the roles of the ncRNA expressed from these.

#### NUCLEOLAR ALTERATIONS UPON DNA DAMAGE AND TRANSCRIPTIONAL STRESS

Nucleolar structure changes significantly in response to several types of stress (**Figure 2**). If and when the rRNA production is halted resulting from e.g., double strand break-inducing DNA damage by ionizing radiation (IR) or RNA pol I inhibition by actinomycin D, nucleolar segregation occurs and so called nucleolar caps are formed. Nucleolar caps are bipartite structures containing FCs and DFCs which surround the GC components (Puvion-Dutilleul et al., 1992). A different structural reorganization, the nucleolar necklace, is formed under certain conditions where RNApolI transcription remains active, but rRNA processing is impaired (**Figure 2**). This is evident upon treatment of cells with doxorubicin (DRB), a DNA intercalating agent and inhibitor of DNA topoisomerase II (Louvet et al., 2005).

DNA damage in the form DNA bulges, by e.g., various cytotoxic drugs or UV radiation, cause inhibition of rDNA transcription by RNApolI, resulting in nucleolar disruption. The dispersal of the nucleolus releases proteins to nucleoplasm that normally do not reside in there, a mechanism by which certain stress responses are induced. E.g., p53 nucleolar and ribosomal proteins binds to the MDM2 protein following disruption of ribosome biogenesis. This leads to inhibition of MDM2 E3 ligase activity and thus to p53 activation (reviewed in Lindström and Latonen, 2013). It is interesting that the nucleolar responses to UV and IR differ (Moore et al., 2011). In addition to the different DNA damage types, these insults induce also damage to other macromolecules and partly different cellular responses (Laiho and Latonen, 2003; Goldstein and Kastan, 2015). Currently, it is unclear which other cellular events are involved in dictating the differential nucleolar stress responses upon these stresses (Moore et al., 2011).

It is well established that disruption of the nucleolus triggers a p53-dependent cellular stress response referred to as "nucleolar stress" (Zhang and Lu, 2009; Lindström and Latonen, 2013). This is frequently called also "ribosomal stress," although not all abnormalities in ribosome biogenesis lead to dispersal of the nucleolus. Nucleolar and/or ribosomal stress, mediated to a large extent by interactions of translocated ribosomal and other nucleolar proteins and rRNA, activates signaling pathways leading to cell cycle arrest, apoptosis, differentiation or

senescence, in a cell type and stress severity-dependent manner

and heat schock. This may or may not involve inhibited RNApolI activity.

(reviewed in Lindström and Latonen, 2013). Translocation to the nucleolus is also a regulatory mechanism under several cellular conditions. Initially, nucleolar sequestration as a concept was introduced by Bachant and Elledge (1999) based on work showing that exit from mitosis in budding yeast is regulated by detention of Cdc14 in the nucleolus (Shou et al., 1999; Visintin et al., 1999). The concept was further supported by the notion that, in mammalian cells, tumor suppressor Arf sequesters Mdm2 in the nucleolus to ensure activation of p53 during oncogene activation and replicative senescence (Weber et al., 1999). Detention in the nucleolus has been described for many proteins especially under different stress conditions. For example, MDM2, which is a ubiquitin ligase for tumor suppressor p53 among others, localizes to the nucleolus also upon transcriptional inhibition by Actinomycin D and possesses lower mobility there (Lohrum et al., 2003; Kurki et al., 2004; Mekhail et al., 2005). MDM2 is also transferred to nucleoli upon DNA damage by PML in an ARF-dependent manner (Bernardi et al., 2004). DNA damage induces translocation of also other proteins to nucleoli. For example, IR restores the disturbed association of telomerase protein with the nucleoli in transformed cells (Wong et al., 2002). Acidosis triggers pH-dependent interaction von Hippel-Lindau tumor suppressor protein (VHL) with rDNA, a phenomenon which is promoted by activation of hypoxia inducible factor HIF (Mekhail et al., 2004a,b, 2006). The authors suggest that this is a way for oxygen-starved cells to maintain energy equilibrium by gauging the environmental H+ concentration to statically retain VHL in nucleoli to restrict ribosomal production (Mekhail et al., 2006).

#### NUCLEOLAR DETENTION UNDER PROTEIN STRESS – AGGRESOMES, AMYLOIDOGENESIS AND RELEVANCE TO DISEASE

Different from insults directly affecting rRNA transcription and processing, protein stress causes extranucleolar proteins and RNA to be detained in the nucleolus. Initially, certain stress-responsive proteins, such as p53, Mdm2 and PML body proteins, were described to translocate to nucleoli stress-signal-dependently (Klibanov et al., 2001; Mattsson et al., 2001; Xirodimas et al., 2001; Latonen et al., 2003). Later, this phenomenon was described to apply to a number of nuclear UPS client proteins and represent formation a de novo stress response organelle (Latonen, 2011; Latonen et al., 2011).

This foci formation takes place in nucleolar cavities, and intranucleolar bodies (INBs) can already be detected in S-phase cells and even after certain types of DNA damage (Abella et al., 2010; Hutten et al., 2011). Upon severe protein stress upon e.g., heat shock, chemical inhibition of proteasome activity, and acidosis, an expanded organelle is formed (Latonen et al., 2011; Audas et al., 2012a, 2016). The intranucleolar stress-responsive macromolecular collections have also been called to occur in so called detention centers, and intranucleolar macromolecular collections showing amyloid properties have been termed amyloid bodies (A-bodies) (Jacob et al., 2013; Audas et al., 2016).

Currently it is not clear how these structures relate to each other, but they share striking similarities: (1) they all form in the nucleoli but are clearly not belonging to normal components of the nucleoli, (2) they involve accumulation of protein which are not normal components of the traditional nucleolar structures, and many of these deposits have been shown to contain at least some of the same proteins, (3) often there is also RNA, which does not belong to normal nucleolar components, accumulating, and (4) formation of many of these structures has been shown to depend on intactness of the nucleoli, with the help experiments utilizing Actinomycin D-mediated nucleolar disruption. Thus, although INBs, nucleolar aggresomes and A-bodies have not been proven to represent same structures, with such striking similarities it seems plausible that they represent a range of sizes and states resulting from same phenomena.

The initial papers describe events occurring upon different cellular stresses (proteasome inhibition, acidosis, heat stress, DNA damage), and certain differences exist in the contents of the organelles. Thus, cellular context- and stress-dependency of RNA and protein recruitment remains to be investigated in future studies. The common denominator seems to be to clear nuclear proteins from the nucleoplasm to regulate cellular activities for the duration of the stress situation, and at least upon

certain insults, the formation of the intranucleolar collections can be transient. For certain proteins, there is evidence for functional impact in the localization to nucleolar aggressomes, such as for TTRAP, which regulates rRNA processing during cellular response to proteasome inhibition (Vilotti et al., 2012). Considering, however, that nucleolar aggresomes can form even as a result of overexpression of exogenous proteins or increased protein synthesis due to a viral infection (reviewed in Latonen, 2011), the role of the nucleolar aggresomes may be, at least at times, to protect the nucleoplasmic environment from excess proteins. In fact, the formation of A-bodies has been suggested as a form of so called protective or functional amyloidosis (Lyons and Anderson, 2016; Woodruff et al., 2018). Amyloid-bodies are solid condensates (Woodruff et al., 2018), and as such, resemble Balbiani-bodies in Xenopus oocytes, forming by amyloid-like assembly of a disordered protein Xvelo (Boke et al., 2016). Proteins in nucleolar aggresomes exhibit decreased mobility (Latonen et al., 2011), while INBs are likely soluble, exhibiting liquid-like spherical appearance (Hutten et al., 2011).

Thus, a plausible, yet speculative, sequence of events (**Figure 2**) in nucleolar aggresome formation involves an initial liquid phase in the nucleolar cavity or detention center (Wang M. et al., 2018). With prolonged accumulation of macromolecules to the structure, the proteins turn immobile (Latonen et al., 2011), liquid-solid phase transition occurs, and may proceed to amyloidogenesis (Audas et al., 2016). RNA seeding is involved in the seeding, at least for the amyloidogenic phase (Audas et al., 2016; Lyons and Anderson, 2016). It is possible that amorphous gel like intermediate states also exist to maturate concentrates of initially liquid state (Woodruff et al., 2018), although this is currently purely speculative. Thus, the exact material properties in each condition, and the mechanisms leading to the possible phase transitions remain to be investigated.

It seems that the composition of the nucleolar aggresomes is somewhat dependent on the stress insult. In general, the proteins are collected to the nucleolar aggresome along with RNA, and most often the aggresomes contain conjugated ubiquitin, SUMO, and heat shock factors (reviewed in Latonen, 2011). Although nucleolar aggresomes bare similarities with cytoplasmic aggresomes especially in the presence of ubiquitin conjugates, heat shock factors and links to hampered protein degradation, they are clearly different structures from cytoplasmic aggresomes (Latonen, 2011). Furthermore, inhibition of lysosomal proteases does not affect nucleolar aggresomes (Latonen et al., 2011; Salmina et al., 2017), indicating that nucleolar aggresome formation is separate from general protein degradation defects. Nucleolar aggresomes can, however, be released to the cytoplasm during mitosis and processed through the autophagocytosis pathway (Salmina et al., 2017). Nucleolar aggresomes can occur in several cell types, their prominence being greatest in normal diploid cells (Latonen et al., 2011). It is possible that the proliferative activity and transformation status of the cells affecting nucleolar activity and organization also affects formation of nucleolar aggresomes.

The roles and identities of the RNA components in nucleolar aggresomes remain to be investigated fully. Non-coding RNA transcribed from rDNA (IGS16RNA, IGS22RNA and IGS28RNA) has been shown to recruit proteins to aggresomes upon hypoxia/acidosis and heat shock (Audas et al., 2012a; Jacob et al., 2012, 2013). Nucleolar aggresomes formed after proteasome inhibition contain polyA-tailed RNA, suggestive of either mRNA, lncRNA or both (Latonen et al., 2011). Due to lack of comprehensive extraction and sequencing studies, the full range of RNA species localized in nucleolar aggresomes is yet to be discovered.

While nucleolar aggresomes formed upon proteasome inhibition or protein overexpression have not yet been shown to be reversible, this has been reported for acidosis and heat shock-induced events (Audas et al., 2016). In addition, certain nucleolar aggresome-inducing stress events seem to inhibit rDNA transcription (Jacob et al., 2013) while others do not (Latonen et al., 2011). Thus, inhibition of rDNA transcription may not be necessary for nucleolar aggresomes to form, but a co-occuring or a following event under certain stress conditions.

#### Nucleolar Aggresomes and Neurodegeneration

The formation of nucleolar aggresomes in cultured cells resembles – and models – the situation in certain neurodegenerative disorders where proteins and RNA accumulate and aggregate to nuclei of cells, such as Huntington's disease (HD) and spinocerebellar ataxias (SCA). A hallmark of numerous neurodegenerative diseases is ubiquitin, SUMO- and RNA-containing inclusion bodies that link to expression of aggregation-prone mutant forms of disease-related proteins or RNA and to impairment of UPS (Dorval and Fraser, 2007; Lehman, 2009; Huang and Figueiredo-Pereira, 2010). The nuclear inclusions in HD in vivo resemble nucleolar aggresomes in vitro, although they localize adjacent to the nucleoli and not similarly within the nucleolus (Davies et al., 1997). Upon treatment of sensory ganglion neurons with proteasome inhibitors, and in motor neurons with severe dysfunction of proteostasis in a mouse SMA model, nuclear poly(A) RNA granules are formed frequently adjacent to the nucleolus, but not within it (Palanca et al., 2014; Narcís et al., 2018). These data may indicate differences in the in vivo vs. in vitro nucleolar state, as in post-replicative cells the nucleoli have adhered to a fully mature form. On the other hand, SUMO1-positive intranucleolar spots lacking nascent RNA and associated with a nucleolar reorganization of fibrillar centers have been found in vivo in motor neurons in the spinal muscular atrophy (SMA) (Tapia et al., 2017). Furthermore, nucleolar aggresomes have been shown to occur in ex vivo prostate tissue (Latonen et al., 2011) and in human breast and prostate cancer tissue (Audas et al., 2016).

Although the focus on the nucleoli with respect to neurodegeneration has been on effects of diseased mutants on ribosomal production and activity, several neurodegeneration relevant proteins have been shown to localize to nucleoli. A specific form of mutant Htt localizes to the nucleolus in mouse neuronal progenitor cells (Trettel et al., 2000). In addition, artificial β-sheet proteins known to form prefibrillar and fibrillar aggregates have been shown to accumulate

to nucleoli in cultured cells (Woerner et al., 2016). The most compelling in vitro evidence for nucleolar aggresome relevance for neurodegenerative disease exists for C9orf72. Expansion of the a GGGGCC hexanucleotide repeat in the first intron of this gene is the most common genetic alteration leading to hereditable amyotrophic lateral sclerosis (ALS). Glycine/arginine and proline/arginine repeats resulting from non-ATG translation of these repeats are recruited to nucleoli and hamper ribosome biogenesis, resulting in cell death (Kwon et al., 2014). Accumulating evidence shows that these dipeptide repeats locate to the GC where they phase-separate with NPM, disrupting nucleolar function (Haeusler et al., 2014; Lee et al., 2016; Lin et al., 2016). These repeats in fact interact with several IDR-containing proteins, many being RNA binding proteins (RBPs) and/or MLO proteins (Lee et al., 2016). While many repeat-expanded proteins accumulate to ribonucleoprotein (RNP) granules (Van Treeck and Parker, 2018), it is not clear what dictates accumulation of C9orf72 repeat peptides to nucleoli. Interestingly, the dipeptide repeats of C9orf72 have been shown to function as polyamines and promote intermolecular RNA-RNA interactions (Van Treeck et al., 2018). Although the functional significance of this in the toxicity of the mutant remains to be shown, it seems likely that these interactions affect the LLPS and/or nucleolar interactions of these peptides.

#### Nucleolar Aggresomes, Cancer and p53

In general, the nucleolus and interactions with nucleolar proteins and rRNA species is central in regulation of certain tumor suppressor and oncogene activities, the most recognized being p53 and c-myc, respectively (Ruggero and Pandolfi, 2003; Boisvert et al., 2007). As nucleolar aggresomes have been detected in human breast and prostate cancer tissues (Audas et al., 2016), and they can be induced by proteasome inhibition in ex vivo prostate tissue (Latonen et al., 2011), nucleolar aggresomes may also be relevant for cancer. Dozens of cancer-related nuclear factors can be targeted to nucleolar aggresomes (Latonen, 2011; Latonen et al., 2011). While some of these have implications in regulation of nucleolar activity and ribosome production, such as c-Myc (reviewed in Lindström and Latonen, 2013), most have no identified function in the nucleoli and are likely regulatory targets of the aggresomes under stress.

p53 was one of the first proteins showed to exhibit stress-responsive nucleolar localization (Klibanov et al., 2001; Xirodimas et al., 2001; Latonen et al., 2003). p53 is a tumor suppressor and the most often mutated gene in human cancers (Muller and Vousden, 2013), and it has several connections to nucleolar-related proteins such as NPM, ARF and MDM2 (Mayer and Grummt, 2005). p53 and p53-derived fragments have been shown to aggregate in vitro (Silva et al., 2014), and several p53 mutants have been found as amyloid aggregates in tumor cell lines (Xu et al., 2011) and breast cancer biopsies (Levy et al., 2011). These aggregates inactivate p53 by sequestering the protein, thus blocking its transcriptional activity and pro-apoptotic function (Xu et al., 2011). A cell-penetrating peptide, ReACp53, designed to inhibit p53 amyloid formation, rescues p53 function in cancer cell lines and in organoids derived from high-grade serous ovarian carcinomas (HGSOC) (Soragni et al., 2016).

Tumor suppressor p53 translocates to nucleolus upon treatment of cells by proteasome inhibitors in cultured cells and ex vivo tissue (reviewed in Latonen, 2011). In addition, the chemical compounds PRIMA1 and PRIMA-1MET have been reported to induce nucleolar translocation of p53 (Rökaeus et al., 2007; Russo et al., 2010), although contradicting evidence also exists (Rangel et al., 2019). PRIMA1 is a mutant p53 reactivating compound (Bykov et al., 2002) which has been shown to reactivate unfolded p53 mutants to native, functional conformation and, recently, to prevent mutant p53 aggregate accumulation in cancer cells (Rangel et al., 2019). Thus, PRIMA-1 can rescue amyloid state of mutant p53, which has implications for future cancer treatment strategies (Rangel et al., 2019).

#### Signals Behind Nucleolar Localization of Proteins

Signals in amino acid sequence that target proteins to the nucleolus are referred to as nucleolar localization signals (NoLS). They are arginine/lysine rich and range from seven to approximately 30 aa residues, but they are relatively rare and not a requirement for nucleolar localization. In fact, nucleolar localization of a protein is viewed to most often result from either direct or indirect interaction with nucleolar molecules, either rDNA, its transcripts, or protein components (reviewed in Emmott and Hiscox, 2009).

It is not clear what signals direct the localization and detention of extranucleolar proteins to nucleoli and nucleolar aggregates under stress. These likely depend on molecular interactions and involve changes in phase separation and transition balance due to presence of new molecules, but which specific molecules function in the seeding of the detention remain an open question for several conditions. Recently, using FUS family of proteins as an example it was shown that tyrosine residues in prion-like domains and arginine residues on RNA-binding domains govern the saturation concentration of phase separation (Wang J. et al., 2018). Interestingly, Mekhail et al. (2007) identified a common peptide motif in the proteins detained in the nucleoli during acidosis, including VHL, HSC70, RNF8 and cIAP2. This so called nucleolar detention signal regulated by H<sup>+</sup> (NoDSH+) is different from the canonical nucleolar localization signal (NoLS) and is composed of an arginine motif combined to several hydrophobic repeats (Mekhail et al., 2007; Jacob et al., 2012). Up to 9% of all proteins harbor a NoDS, indicating that a substantial amount of the proteome may potentially be regulated in a similar fashion (Jacob et al., 2012). Thus, arginine rich motifs seem as a recurrent event in MLO and nucleolar targeting. The species of ncRNA expressed upon stress signals also have a role in recruiting the proteins to nucleolar aggresomes (discussed below), as specific RNAs, e.g., extended repetitive RNA motifs in clinical disorders, are capable of phase separation (Jain and Vale, 2017; Sawyer et al., 2018).

Post-translational modifications, especially ubiquitin family conjugates, may have a key role in localizing extranucleolar proteins to nucleolar aggresomes. In addition to ubiquitin and

SUMO found in the physiological disease-relevant inclusion bodies (as discussed above), the nucleolar aggresomes have revealed several family members to be relevant for nucleolar aggresomes. The structures contain conjugated ubiquitin, indicating that at least some of the accumulated proteins harbor this modification (Latonen et al., 2011). Especially interesting is the role of SUMO-proteins, which are found in INBs and nucleolar aggresomes (Hutten et al., 2011; Latonen et al., 2011; Souquere et al., 2015). UBC9, the E2 SUMO-conjugating enzyme is also located in INBs, it's depletion reduces INB size, and SUMO-1 mutant unable to conjugate proteins does not localize to INBs, indicating that SUMO conjugation is relevant for INB biology (Brun et al., 2017). Yet another ubiquitin homolog, NEDD8, was recently shown to localize to nucleolar aggresomes formed upon heat shock and proteasome inhibition (Maghames et al., 2018). Similarly to SUMO, this localization was linked to NEDD8 conjugation and even NEDD8/ubiquitin hybrid chain formation (Maghames et al., 2018). Thus it seems that nucleolar localization and aggregation of extranucleolar proteins is regulated by ubiquitin family of protein conjugation, requiring further investigation to understand the exact underlying mechanisms and functional consequences.

#### EMERGING ROLES OF NON-CODING TRANSCRIPTS IN THE NUCLEOLUS

The nucleolus is packed with non-coding RNA. After the 8S, 18S and 5.8S rRNAs have been transcribed by RNApolI and cleaved from their 47S precursor, they are post-transcriptionally modified through interaction with small nucleolar ribonucleoproteins (snoRNPs) and additional processing factors. For a long time, other rRNA sequences have been neglected as garbage sequences or non-specific degradation products. Recently, it has become clear that many non-coding RNA species in addition to the classical rRNA and snoRNA contribute to nucleolar biology. For example, rDNA is transcribed in antisense orientation to produce RNA contributing to epigenetic silencing of rDNA (Bierhoff et al., 2010).

Perhaps the most interesting resource of ncRNA in the rDNA sequence to be fully explored is the intergenic spacer (IGS). This sequence differs considerably from the rRNA coding sequences and has a high variability in nucleotide composition and length. Mayer et al. (2006) showed that some of these transcripts are required for establishing and maintaining a specific heterochomatic configuration at the promoter of a subset of rDNA arrays via NoRC, a chromatin remodeling complex. The transcripts here are 150–300 nt long and are complementary to the sequences in rDNA promoter (pRNA). During mid-S phase in the cell cycle, these pRNAs increase by 2-fold to repress rRNA synthesis in late replication (Santoro et al., 2010). Interestingly, the pRNA-dependent establishment of heterochromatin condensation of rRNA genes initiates highly condensed chromatin structures outside the nucleolus (Savic et al., 2014). This promotes transcriptional activation of differentiation genes, and is a mechanism shown to be inactivated in pluripotent embryonic stem cells (Savic et al., 2014). Thus, pRNA regulates chromatin plasticity and pluripotency.

Intergenic spacer also produces several stimuli-specific ncRNAs. Stress conditions such as heat shock and acidosis induce transcription of IGS to produce several transcripts shown to be involved in nucleolar aggresome formation, including IGS16RNA, IGS22RNA and IGS28RNA (Audas et al., 2012b; Jacob et al., 2012, 2013). These transcripts are produced from stimuli-specific loci (Audas et al., 2012a). Whether there are more IGS transcripts that are relevant for protein detention in the nucleolus and nucleolar aggresome formation remains to be investigated.

Although most lncRNAs are processed like typical mRNAs to be 5<sup>0</sup> capped and 3<sup>0</sup> polyadenylated, other lncRNAs are stabilized by alternative mechanisms. One mechanism for this adaptation of snoRNA processing to produce snoRNA-ended lncRNAs (sno-lncRNAs) and 5<sup>0</sup> snoRNA-ended and 3<sup>0</sup> -polyadenylated lncRNAs (SPAs). Some sno-lncRNAs and SPAs have been shown to be involved in the regulation of pre-rRNA transcription and alternative splicing of pre-mRNAs (Xing and Chen, 2018). For example, a box H/ACA small nucleolar RNA (snoRNA)-ended long non-coding RNA (lncRNA) was described to enhance pre-rRNA transcription (SLERT) (Xing et al., 2017). SLERT requires box H/ACA snoRNAs at both ends for its biogenesis and translocation to the nucleolus. SLERT interacts with DEAD-box RNA helicase DDX21 via a 143-nt non-snoRNA sequence, following which DDX21 forms ring-shaped structures surrounding multiple RNApolI complexes and suppresses pre-rRNA transcription (Xing et al., 2017).

In C. elegans, there was recently a new class of antisense ribosomal siRNAs (risiRNAs) identified that downregulate pre-rRNA through the nuclear RNAi pathway (Zhou et al., 2017). risiRNAs exhibit sequence characteristics similar to 22G RNA while being complementary to 18S and 26S rRNA. risiRNAs induce translocation of the nuclear Argonaute protein NRDE-3 from the cytoplasm to nucleus and nucleolus, in which the risiRNA/NRDE complex binds to pre-rRNA and silences rRNA expression. Interestingly, exposing Caenorhabditis elegans to cold shock or UV radiation, risiRNAs accumulate, turning on the nuclear RNAi-mediated gene silencing pathway (Zhou et al., 2017). Whether similar mechanisms exist in mammalian cells and contribute to nucleolar stress responses remains to be explored.

# CONCLUDING REMARKS

It has become clear that, likely due to its phase separating propensities, the nucleolus can serve as a protective site for proteins following several environmental stimuli and stress signals. This detention may lead to formation of nucleolar aggresomes, and targets varying species of RNA and differential pools of proteins dependently on the cellular context and stress signal. What remains to be determined is the general mechanisms

that dictates these responses. The requirements for the phase transition steps are to be studied in detail, and surely more RNA effectors are to be found. Investigation of protein amino acid sequence signals, regulation by conjugation of ubiquitin protein family members, and interactions between RNA-protein and protein-protein domains promoting aggregation and amyloid formation in the nucleolus may enlighten the cellular and molecular routes to target in pathological nuclear aggregation.

#### REFERENCES


#### AUTHOR CONTRIBUTIONS

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

# FUNDING

The work was supported by Finnish Academy (Grant No. 317871).


Kwon, I., Xiang, S., Kato, M., Wu, L., Theodoropoulos, P., Wang, T., et al. (2014). Poly-dipeptides encoded by the C9orf72 repeats bind nucleoli, impede RNA biogenesis, and kill cells. Science 345, 1139–1145. doi: 10.1126/science.1254917


of neurodegenerative diseases and cancer. Bioessays 33, 386–395. doi: 10.1002/bies.201100008


Ruggero, D., and Pandolfi, P. P. (2003). Does the ribosome translate cancer? Nat. Rev. Cancer 3, 179–192. doi: 10.1038/nrc1015


protein phosphatase Cdc14 from nucleolar RENT complex. Cell 97, 233–244. doi: 10.1016/s0092-8674(00)80733-3


**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 Latonen. 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.

# Emerging Role of the Nucleolar Stress Response in Autophagy

#### Astrid S. Pfister\*

Institute of Biochemistry and Molecular Biology, Faculty of Medicine, Ulm University, Ulm, Germany

Autophagy represents a conserved self-digestion program, which allows regulated degradation of cellular material. Autophagy is activated by cellular stress, serum starvation and nutrient deprivation. Several autophagic pathways have been uncovered, which either non-selectively or selectively target the cellular cargo for lysosomal degradation. Autophagy engages the coordinated action of various key regulators involved in the steps of autophagosome formation, cargo targeting and lysosomal fusion. While non-selective (macro)autophagy is required for removal of bulk material or recycling of nutrients, selective autophagy mediates specific targeting of damaged organelles or protein aggregates. By proper action of the autophagic machinery, cellular homeostasis is maintained. In contrast, failure of this fundamental process is accompanied by severe pathophysiological conditions. Hallmarks of neuropathological disorders are for instance accumulated, mis-folded protein aggregates and damaged mitochondria. The nucleolus has been recognized as central hub in the cellular stress response. It represents a sub-nuclear organelle essential for ribosome biogenesis and also functions as stress sensor by mediating cell cycle arrest or apoptosis. Thus, proper nucleolar function is mandatory for cell growth and survival. Here, I highlight the emerging role of nucleolar factors in the regulation of autophagy. Moreover, I discuss the nucleolar stress response as a novel signaling pathway in the context of autophagy, health and disease.

Keywords: ribosome biogenesis, rRNA processing, nucleoli, nucleolar stress, autophagy, neuron, mitochondria, aggregates

#### INTRODUCTION

Various high quality reviews are available on principles of ribosome biogenesis, nucleolar stress, apoptosis and autophagy, respectively. Given their essential role, it is well accepted that a misregulation of each is tightly linked to pathogenic conditions (Levine and Kroemer, 2008; Boulon et al., 2010; Freed et al., 2010; Ghavami et al., 2014; Schneider and Cuervo, 2014). In this review, the emerging connection of nucleolar stress to autophagic processes serves as a basis to discuss novel concepts and cure of diseases connected to nucleolar stress.

#### Edited by:

Kerry Lee Tucker, University of New England, United States

#### Reviewed by:

Cláudia Pereira, University of Coimbra, Portugal Grzegorz Kreiner, Polish Academy of Sciences, Poland

> \*Correspondence: Astrid S. Pfister astrid.pfister@uni-ulm.de

#### Specialty section:

This article was submitted to Cellular Neurophysiology, a section of the journal Frontiers in Cellular Neuroscience

Received: 21 December 2018 Accepted: 08 April 2019 Published: 30 April 2019

#### Citation:

Pfister AS (2019) Emerging Role of the Nucleolar Stress Response in Autophagy. Front. Cell. Neurosci. 13:156. doi: 10.3389/fncel.2019.00156

# The Nucleolus as a Stress Sensor

fncel-13-00156 April 29, 2019 Time: 10:31 # 2

#### The Nucleolus: Place for Ribosome Biogenesis

Nucleoli represent membrane-free, sub-nuclear compartments, where transcription and processing of rRNA takes place. Nucleoli can be considered as an assembly platform. They host several hundreds of essential rRNA binding and processing factors, which are involved in the highly complex process of ribosome biogenesis. Nucleoli form around repetitive rDNA clusters in a dynamic and cell cycle-dependent manner during G1 phase (Potmesil and Goldfeder, 1977; Mangan et al., 2017).

The rDNA clusters are transcribed into their respective large precursor rRNA by RNA polymerase I (RNA pol I); RNA polymerases II and III are as well essential for ribosome biogenesis, by driving the expression of ribosomal proteins (RNA pol II) and 5S rRNA (RNA pol III) (Eichler and Craig, 1994; Fatica and Tollervey, 2002). The complex mechanism of prerRNA processing involves the action of a multitude of ribosome biogenesis factors. These are assembled in pre-ribosomal complexes involved in cleavage and chemical modification of the maturating transcript (Fatica and Tollervey, 2002; Granneman and Baserga, 2004; Mullineux and Lafontaine, 2012).

The nucleolar size correlates with the rRNA transcription, cell growth and the metabolic rate of a cell (Boulon et al., 2010). Importantly, nucleolar size and function is changed during aging (Tiku et al., 2016; Buchwalter and Hetzer, 2017; Zlotorynski, 2017). Thus, the nucleolus emerges as critical regulator of cellular aging (Tiku and Antebi, 2018).

Large amounts of ribosomes are especially needed in highly proliferating cells, such as during embryonic development or cancer (Montanaro et al., 2008). Therefore, a lack of functional ribosomes impairs cellular growth and survival and is incompatible with life.

#### The Nucleolar Stress Response

Nucleoli are highly dynamic structures, closely connected to growth and survival (Mangan et al., 2017). The nucleolus is being recognized as a key hub in the cellular stress response by sensing and reacting to various stimuli.

Perturbation of the nucleolar structure and/or function ultimately impairs ribosome biogenesis and triggers the so-called nucleolar stress response (Boulon et al., 2010). A key mechanism involves the release of ribosomal proteins (RPs) from the nucleoli into the nucleoplasm. As a consequence of nucleoplasmic RP accumulation, the E3-ubiquitin ligase MDM2 is inhibited (Dai et al., 2004). MDM2 keeps the levels of the tumor suppressor protein p53 low by earmarking p53 for proteasomal degradation. Upon nucleolar stress, RPs are released and inhibit MDM2, which results in p53 accumulation. Finally, stabilized p53 induces cell cycle arrest and/or apoptosis (Pestov et al., 2001; Rubbi and Milner, 2003; Yuan et al., 2005; Fumagalli et al., 2012). The nucleolar stress response is further connected to the induction of senescence and DNA damage, by commonly engaging the classical p53 pathway (Rubbi and Milner, 2003; Lindstrom et al., 2018). A simplified model of the classical p53 nucleolar stress response is given in **Figure 1**.

More recently, novel pathways have been added to the increasing list, which demonstrate that nucleolar stress can also be propagated in the absence of functional p53 (Holmberg Olausson et al., 2012; James et al., 2014; Lindstrom et al., 2018).

In summary, nucleolar integrity reflects a general prerequisite for normal cellular function. Given that many tumor types are characterized by inactivation of p53, p53-independent pathways open novel avenues toward more customized anticancer therapies (Burger and Eick, 2013).

# Autophagy Pathways

#### Macroautophagy

(Macro)autophagy is essential for cellular homeostasis by mediating destruction and recycling of bulk cytoplasmic material, defective organelles or proteins via lysosomal degradation (Mizushima, 2007; Marx, 2015). A mis-regulation of autophagy is tightly linked to the formation of diverse pathological conditions (Levine and Kroemer, 2008; Jiang and Mizushima, 2014; Schneider and Cuervo, 2014).

Autophagy can be induced by various cellular stresses, such as lack of nutrients, low energy or oxidative stress. As an upstream signaling pathway the conserved mTOR pathway plays an influential role in the regulation of autophagy (see section "mTOR Signaling Couples Autophagy and Ribosome Biogenesis") (Pattingre et al., 2008).

A central structure implicated in the process of (macro)autophagy is the double-membranous autophagosome, which mediates cellular cargo sequestration. Autophagyrelated proteins (ATGs) govern autophagosome formation at different levels. Beclin1, the mammalian homologue of yeast Atg6, is mandatory for the initial steps of autophagosome formation (Pattingre et al., 2008). Originally, Beclin1 has been identified as an interaction partner of the anti-apoptotic factor Bcl-2 (B cell lymphoma 2) (Liang et al., 1998). Beclin1 is part of phosphoinoside 3 kinase (PI3K) complexes and functions in diverse membrane trafficking processes (Levine et al., 2015). Furthermore, Beclin1 interacts with kinases, de-/ubiquitinating enzymes and multiple other factors, among them p53 (Levine et al., 2015). Certain ATGs are necessary for the engulfment of cargo destined for lysosomal degradation in the autolysosome. Members of the Atg8 protein family are directly conjugated to the autophagosomal membrane by phosphaditylinositol lipidation. Hence, Atg8/LC3 (light chain 3) represents a key marker of autophagosomes. Atg8/LC3s are essential for autophagosome maturation and cargo sequestration. They can be sub-divided into LC3 and GABARAP proteins and fulfill critical tasks (Nguyen et al., 2016).

Note that intracellular Atg8/LC3-positive autophagosome accumulation is central for autophagy flux induction as well as inhibition. Opposing mechanisms result in the accumulation of autophagosomes, thereby requiring careful interpretation of experimental results. An increased rate of autophagosome formation (increase of autophagic flux), as well as decreased autophagosome clearance in the lysosome (impaired autophagic flux), resembles autophagosome accumulation under basal conditions. Thus, care has to be taken when interpreting results on "active" or "inhibited" autophagy. Meanwhile, several excellent reviews are available, which help to unravel these

issues (Mizushima and Yoshimori, 2007; Mizushima et al., 2010; Klionsky, 2011; Gottlieb et al., 2015; Klionsky et al., 2016). So-called autophagic flux studies have become detrimental for understanding mechanisms of autophagy. Experts agree on combining various independent methods to allow solid data interpretation (Klionsky et al., 2016).

In general, autophagy is noticed as a protective mechanism by lowering the cellular stress.

Apoptosis and autophagy are both stimulated by similar stressors. However, they can be seen as opposing signaling events. Whereas autophagy acts in an anti-apoptotic manner and precedes apoptosis (Boya et al., 2005), apoptosis induction can block autophagy for instance by removing pro-autophagy proteins by caspases. However, this can even generate pro-apoptotic fragments of ATG autophagy regulators thereby triggering a fast forward response (Marino et al., 2014). Therefore, obviously a tight crosstalk exists, which goes into both directions, depending on the context. As a proof of principle, low sub-lethal levels of stress favor autophagy induction as a protective mechanism, whereas sustained stress beyond a certain threshold induces apoptosis. For instance, over-activation of autophagy can be a pro-death signal for autophagic cell death, and autophagy inducers can trigger apoptosis (Marino et al., 2014). Likewise, many stimuli that activate apoptosis can also stimulate autophagy. Extrinsic stress factors include chemotherapeutics, ionizing irradiation, lack of growth factors or nutrients. Intrinsically, p53 (see below), oncogenes (e.g., Myc), BH3-only proteins (Bcl-2 homology3) or serine/threonine kinases (such as JNK or AKT/PKB) are involved in the regulation of autophagy/apoptosis pathways

(Marino et al., 2014). For instance, the pro-apoptotic Beclin1 and anti-apoptotic Bcl-2 are commonly affected. Both interact with each other and thereby regulate the balance between autophagy or apoptosis. Also, mitochondrial integrity, caspase and ATG activation, mTOR signaling and multiple others are implicated (Marino et al., 2014).

Collectively, a mis-regulation of autophagy in either direction is connected to numerous pathophysiological conditions, and the same holds true for apoptosis (Maiuri et al., 2007; Jiang and Mizushima, 2014; Marino et al., 2014).

#### Selective Autophagy

Specific cellular cargo can be selectively targeted by autophagy (Kirkin et al., 2009). The pathways have been named according to their type of cargo, for instance mitophagy for specialized autophagy of mitochondria (Ding and Yin, 2012; Hamacher-Brady and Brady, 2016; Khaminets et al., 2016), nucleophagy for removal of the nucleus (Park et al., 2009; Mijaljica and Devenish, 2013), ribophagy for ribosomes (Beau et al., 2008; Kraft and Peter, 2008; Frankel et al., 2017), and aggrephagy for protein aggregates (Yamamoto and Simonsen, 2011).

Mitophagy is fundamental for the mitochondrial homeostasis and a mis-regulation of mitophagy is clearly implicated in the development of neurodegeneration (see section "Distinct Neurodevelopmental Pathologies, Common Concepts – A Short Overview"). The key players of mitophagy, Parkin and PINK1 [tumor suppressor phosphatase and tensin homolog (PTEN) induced putative kinase 1], are mutated in patients with Parkinson's disease (Pickrell and Youle, 2015). Parkin functions as an E3-ubiquitin ligase, which is recruited to impaired mitochondria (Narendra et al., 2008). Parkin is required for ligation of ubiquitin marks to defective mitochondrial cargo (Ding and Yin, 2012; Harper et al., 2018). Parkin depends on the proper function of PINK1 (Vives-Bauza et al., 2010), an ubiquitin kinase located at the mitochondrial outer membrane (MOM) (Chin et al., 2010).

Mitophagy and apoptosis are both characterized by similar upstream events (Mukhopadhyay et al., 2014). Induction of mitophagy, for instance, is accompanied by activation of Bcl-2-associated X protein (BAX). This induces MOM perforation (MOMP), depolarization and release of cytochrome c from the mitochondrial intermembrane space (IMS) into the cytosol. As a consequence, PINK1 becomes stabilized at depolarized mitochondria and Parkin is subsequently translocated from the cytosol into the MOM (Vives-Bauza et al., 2010). By concerted action of both, mitochondria become decorated by ubiquitin marks (Lazarou et al., 2015) and can then be recognized by autophagy/mitophagy receptors such as the ubiquitin adaptor protein p62/sequestosome (Lamark et al., 2009; Geisler et al., 2010). In fact, autophagosome formation is mediated by different key mitophagy receptors such as optineurin and NPD52, which promote the recruitment of the autophagy initiating kinase ULK1 (Wong and Holzbaur, 2014; Heo et al., 2015; Lazarou et al., 2015). Nguyen et al. (2016) found that the GABARAP subfamily is essential for mitophagy. Meanwhile, the number of novel players involved in the complex process of selective autophagy is constantly expanding. Note that accumulation of damaged mitochondria, which are eliminated by mitophagy to a certain

point, sets the threshold for apoptosis as a point of no return (Marino et al., 2014).

The autophagy receptor p62/sequestosome contains (i) interaction domains for ubiquitin, but at the same time also a LIR (LC3 interacting region) domain capable of binding to (ii) LC3, which itself is a key component of autophagosomal membranes. As a result, p62 recruits autophagosomal membranes to its selective, autophagosomal cargo (Lamark et al., 2009; Knaevelsrud and Simonsen, 2010).

The same principle of autophagy receptor (e.g., p62)/ubiquitin/LC3 cargo sequestration accounts also for removal of mis-folded protein aggregates, by a selective process termed aggrephagy (Lamark et al., 2009; Knaevelsrud and Simonsen, 2010; Yamamoto and Simonsen, 2011; Lamark and Johansen, 2012). Aggrephagy is also central to neurodegeneration (see section "Distinct Neurodevelopmental Pathologies, Common Concepts – A Short Overview"). A schematic for aggrephagy, bulk autophagy and mitophagy is depicted in **Figures 2**, **3**.

### mTOR Signaling Couples Autophagy and Ribosome Biogenesis

The mammalian target of rapamycin (mTOR) pathway couples the intake of to nutrients, growth factors, energy and stress to the regulation of cell metabolism, growth, survival and autophagy (Pattingre et al., 2008). Deregulation is linked to various diseases and cancer formation.

Mammalian target of rapamycin signaling is recognized as essential pathway for proper neuronal development, neuronal survival and morphogenesis. Consequently, changes in mTOR signaling have been correlated with a spectrum of neuropathologies, such as epilepsy, intellectual disability, autism, brain injury, brain tumor formation and neurodegeneration (Crino, 2016; Switon et al., 2017). Likewise, mTOR inhibitors such as the bacterial macrolide rapamycin and its analogs are growingly used as therapeutic drugs and tested in clinical trials for effects in diverse neuropathological conditions (Laplante and Sabatini, 2009; Crino, 2016).

Mammalian target of rapamycin is a conserved serinethreonine kinase, which belongs to the phosphoinoside 3 kinase (PI3K) family. It assembles two large protein complexes, mTORC1 and mTORC2. The mTORC1 complex is considered as rapamycin sensitive complex (Pattingre et al., 2008). Rapamycin binds to FKBP12 and mTOR, thereby inhibiting mTORC1. mTORC1 signaling affects cell growth, metabolism and autophagy: Upon favorable conditions, mTOR is activated to allow cell growth by anabolic processes, such as rRNA biogenesis and protein translation. Upon nutrient deprivation and lack of growth factors, mTOR signaling is inhibited and cell growth is suppressed, whereas catabolic processes such as autophagy are induced to allow cell survival under unfavorable conditions. mTORC1 controls autophagy by regulating ULK1, ATG13 and FIP200, as well as by a reported rapamycininsensitive mechanism (Laplante and Sabatini, 2009).

mTORC1 also regulates mitochondrial metabolism and biogenesis: mTORC1 inhibition impairs the MOM potential, reduces oxygen consumption and ATP levels. mTORC1 inhibition further decreases mitochondrial DNA levels and hampers mitochondrial biogenesis by affecting the transcriptional activity of the nuclear factor PGC1α (PPARγ co-activator 1) (Cunningham et al., 2007; Laplante and Sabatini, 2009).

Also p53 can regulate mTOR: DNA damage-induced p53 stabilization activates AMPK, which is a sensor of energy status and in turn results in mTORC1 inhibition. p53 also negatively controls mTORC1 by increasing PTEN expression, which functions as mTORC1 inhibitor. Inhibition of mTOR signaling diminishes nucleolar size and function and promotes longevity in different model organisms (Tiku and Antebi, 2018). However, the precise mechanisms regulating the crosstalk between ribosome biogenesis and autophagy remain to be determined.

A simplified model of mTORC1 signaling and the role of p53 is given in **Figure 4**.

#### NUCLEOLAR STRESS AND AUTOPHAGY: A TIGHT REGULATION BETWEEN HEALTH AND DISEASE

Defective ribosome biogenesis on the one hand and impaired autophagy on the other hand are largely contributing to several diseases. In the following, an overview is provided on common

concepts of three key classes of diseases, classically or recently connected to nucleolar stress and autophagy with specific focus on neurodegeneration, cancer and ribosomopathies. For a more detailed overview see for instance (Parlato and Kreiner, 2013; Ghavami et al., 2014; Danilova and Gazda, 2015; Woods et al., 2015; Hetman and Slomnicki, 2019).

# Distinct Neurodevelopmental Pathologies, Common Concepts – A Short Overview

The nervous system is vulnerable to intrinsic and extrinsic factors, which can give rise to distinct neurodevelopmental pathologies such as microcephaly, psychiatric disorders, autism, intellectual disability, epilepsy and neurodegeneration (please, be referred to review Hetman and Slomnicki, 2019). Causes include, for instance, gene mutations, infections or neurotoxins. As common concepts, gene expression, quality control mechanisms, cell proliferation, differentiation and apoptosis are misregulated. Apoptosis might give rise to microcephaly by eliminating, e.g., neuronal stem cells or post-mitotic neuronal cells (Hetman and Slomnicki, 2019). Likewise Zika virus infection, as an extrinsic factor for neurodevelopmental disorders, is tightly coupled to microcephaly. It has recently been demonstrated that it decreases mTOR signaling and over-activates autophagy (Liang et al., 2016). At the same time, ribosome biogenesis defects are emerging (reviewed in Hetman and Slomnicki, 2019).

Given the striking role of the nucleolus in coordinating mentioned neuropathological routes, deregulation of ribosome biogenesis rises as a potent upstream mechanism. In addition, also autophagy is activated in this context.

#### Neurodegeneration and Aging

fncel-13-00156 April 29, 2019 Time: 10:31 # 7

Aging represents a general risk factor for the formation of neurodegenerative diseases and consequently, neurodegeneration accumulates within our society. Despite the rapid advances made in medicine, not all negative aspects of aging can simultaneously be addressed. Along these lines, increasing the society's age has to go together with improving anti-aging therapies. Our scientific knowledge on distinct neurodegenerative diseases has uncovered several common mechanisms, among them loss of neurons (Parlato and Kreiner, 2013; Parlato and Liss, 2014) and a prominent contribution of aggregate accumulation, induction of apoptosis and a misregulation of autophagy (Yamamoto and Simonsen, 2011; Ghavami et al., 2014). More recently, nucleolar stress has been connected to the induction of various types of neurodegenerative diseases, like Alzheimer's, Parkinson's, and Huntington's Disease (see below) (Hetman and Pietrzak, 2012). In line, aging functions as susceptibility factor for neurodegeneration. It is characterized by a loss of rDNA and is accompanied by reduction of nucleolar size and a decline in rRNA processing (Garcia Moreno et al., 1997; Hetman and Pietrzak, 2012; Parlato and Kreiner, 2013). Therefore, the nucleolus is tightly connected to lifespan regulation (Tiku and Antebi, 2018).

As both routes of apoptosis and autophagy are interwoven and not yet fully understood, mechanistic research is essential as a basis for development of therapeutic approaches. As a mandatory goal, novel drugs have to be tested for specificity and efficacy.

#### Neurodegeneration – Underlying Concepts

In protein mis-folding diseases, also known as proteopathies, proteins loose their normal structure and/or function. As a result, mis-folded proteins accumulate and cause a toxic intracellular environment. Normally, proper cellular homeostasis is maintained by several machinery: the ubiquitin-proteasome system (UPS) degrades proteins, whereas autophagy is capable of removing proteins and whole organelles by the lysosome. Hence, both routes are essential for a healthy cell and both have been implicated in the development of neurodegenerative diseases (Levine and Kroemer, 2008; Ghavami et al., 2014). Note that increased apoptosis induction plays a crucial role in eliminating these damaged cells. In addition, chronic inflammation and oxidative stress are observed in neurodegenerative disorders.

Several neurodegenerative disorders, such as PD, AD, and HD are characterized by accumulation of mis-folded, ubiquitinated proteins, which damage the affected cell (Yamamoto and Simonsen, 2011). The accumulating proteins form inclusion bodies, which can differ between the distinct pathologies. The formation of inclusion bodies/aggresomes/amyloid structures represents a double-edged sword: aggregate formation is not only toxic, it can actually be considered as beneficial and neuroprotective mechanism by reducing the toxic nature of mis-folded ubiquitin-containing protein aggregates (Rubinsztein, 2006; Yang et al., 2007).

With increasing age the autophagic program looses efficiency, thereby increasing the likelihood of aggregate accumulation.

Neurons are highly sensitive to accumulation of protein aggregates and require proper autophagy mechanisms to keep the intracellular toxicity low. Supporting data demonstrating the significant implication of the autophagic machinery were, for instance, obtained in mice lacking ATG7 in the central nervous system. ATG-deficient mice, which fail to perform autophagy, display an accumulation of inclusion bodies followed by neuronal loss (Komatsu et al., 2006).

However, autophagy might have dual functions with respect to neurodegeneration: On the one hand functional autophagy is neuroprotective, by removing defective mitochondria via mitophagy. On the other hand pro-death autophagy is considered to induce neuronal cell death.

Massive inhibition of autophagy can trigger apoptosis, which is observed by loss of neurons in neurodegeneration. Interfering with autophagy regulators and blocking autophagy, results in accumulation of cargo-filled autophagosomes and lysosomes, again being toxic for the cell. As a consequence, lysosomemediated cell death occurs (Kroemer and Jaattela, 2005). Impairment of mitophagy causes accumulation of defective mitochondria, which in turn induces reactive oxygen species (ROS) formation and mitochondrial apoptosis (Seo et al., 2010).

Induction of apoptosis by the chemotherapeutic agent staurosporine is accompanied by mitophagy and autophagy induction in dopaminergic cells. Additional block of autophagy by Bafilomycin or inhibition of mitophagy in PINK null mice sensitizes cells to staurosporine-induced apoptosis. Autophagy and mitophagy seems to be neuroprotective upon staurosporine-mediated apoptosis induction in dopaminergic neurons (Ha et al., 2014). However, with respect to loss of dopaminergic neurons, it is not fully resolved whether autophagy is beneficial or pathogenic.

Keeping mitochondria healthy is a prerequisite for counteracting neurodegenerative diseases. Concepts include for instance maintaining mitochondrial membrane integrity and functionality. Mitochondrial membrane permeabilization is tightly coupled to apoptosis induction (Kroemer et al., 2007). Also anti-oxidants and ROS scavenging appear as beneficial strategies. Inhibitors of apoptosis are used as therapeutic drugs to inhibit neuronal loss. Anti-apoptotic drugs prevent mitochondrial apoptosis by blocking release of cytochrome c from the mitochondria or activation of pro-apoptotic BAX (Westphal et al., 2014). Alternatively, the activity or abundance of anti-apoptotic factors can be elevated.

#### Alzheimer's Disease

Alzheimer's disease is the most common neurodegenerative disorder and has both sporadic and hereditary origin. Mostly, AD is diagnosed as sporadic form by the age of 65 years and represents the primary cause of dementia within the elderly generation (Seshadri et al., 1997; Ding et al., 2005).

Classically, prominent accumulation of defective mitochondria, increase of ROS, and apoptosis induction are found in patient's neurons. A decline in autophagy during aging further promotes the mitochondrial release of cytochrome c, which serves as pro-apoptotic stimulus.

Also decreased nucleolar volume has been detected in AD patients (Iacono et al., 2008; Pietrzak et al., 2011). Mechanistically, inhibition of rDNA transcription by rDNA promoter methylation, reduced 28S/18S ratio, reduced tRNA abundance and increased rRNA oxidation has been linked to AD (Payao et al., 1998; da Silva et al., 2000; Ding et al., 2005; Pietrzak et al., 2011). In a similar manner, a decline of 28S rRNA was found in elderly healthy probes when compared to younger control groups (Payao et al., 1998; da Silva et al., 2000). The patient data suggest that impairment of ribosome biogenesis and protein synthesis is one of the earliest events observed in the pathogenesis of AD characterized by mild cognitive impairment (Ding et al., 2005).

#### Parkinson's Disease

fncel-13-00156 April 29, 2019 Time: 10:31 # 8

Parkinson's disease is the second most common neurodegenerative disease after AD (Parlato and Liss, 2014). PD is characterized by loss of dopaminergic neurons in the brain stem. Patients display tremor, dementia and depression. Also in PD, most cases are sporadic. Typical risk factors are aging and exposure to (mitochondrial) toxins. A key feature of PD is deposition of Lewy bodies, which reflects deposition of α-Synuclein oligomers (Ghavami et al., 2014). These oligomers trigger mitochondrial damage. Hereditary forms of PD involve mutation of the key mitophagy regulators PINK and Parkin. Mutations of both result in impaired mitophagy. Given that α-Synuclein serves as a substrate of the E3-ubiquitin ligase Parkin, accumulation of α-Synuclein is also detected in a Parkin mutated background. Also ER stress is implicated in PD. However, mild ER stress is attributed to function rather in a neuroprotective manner by inducing pro-survival autophagy. PD is mimicked by treating neuronal cells with chemicals, such as MPP+ and rotenone, which trigger mitochondrial dysfunction and cell death (Nicklas et al., 1987).

More recently, disruption of nucleolar integrity has been observed in human post mortem samples of patients with PD (Rieker et al., 2011). In support of a nucleolar contribution, the ribosome biogenesis factor Nucleolin interacts with α-Synuclein. Consequently, damaged mitochondria, ROS, as well as autophagosomes accumulate and cause apoptosis (Rieker et al., 2011).

Disruption of nucleoli, cell cycle arrest and p53-mediated apoptosis is observed by depletion of the transcription initiation factor IA (TIF-1A), required for the recruitment of RNA pol I, in mouse embryonic fibroblasts (Yuan et al., 2005). Ablation of TIF-1A in DA neurons of mice results in Parkinsonism and progressive loss of DA neurons (Rieker et al., 2011) and likewise, reduced expression of TIF-1A is detected in PD patient samples (Evsyukov et al., 2017). Treatment with the neurotoxin MPTP worsens the effect of nucleolar stress. In this model, also p53 is stabilized and mTOR signaling decreased. Finally, ROSmediated oxidative stress is induced and defects are detected in mitochondria, such as impaired mitochondrial transcription and COX (cytochrome c oxidase) activity (Rieker et al., 2011). Therefore nucleolar stress, by inhibiting mTOR signaling, can impair mitochondrial function, which represents a key hallmark of several neurodegenerative diseases.

To determine effects of specific PD mutations on nucleolar function irrespective of neuronal loss, pre-symptomatic, digenic PD models were analyzed. Mild overexpression of mutant human α-Synuclein in PINK1 null background (hA53T-SNCA/PINKKO) revealed differential nucleolar activity: On the one hand, reduced nucleolar activity and impaired nucleolar integrity was found in a subset of DA neurons, whereas others showed elevated nucleolar function, thereby suggesting possible compensatory mechanisms. In contrast, inactivation of PINK1 and DJ-1 showed no alterations, pointing to mutated α-Synuclein as the main contributor of nucleolar stress in the hA53T-SNCA/PINKKO model (Evsyukov et al., 2017).

Hemoglobin (Hb) is strongly expressed in dopaminergic neurons in the substantia nigra and is found in patient samples of AD and PD. Hb has recently been connected to mitochondrial function and apoptosis. Hb can form toxic aggregates in the nucleolus after stimulation with MPP+ and rotenone in a cellular model of PD. In turn, Hb overexpression impairs pre rRNA processing, induces nucleolar stress and sensitizes cells to apoptosis (Codrich et al., 2017). The authors further demonstrate decreased phosphorylation of the mTOR target 4E-BP1, decreased numbers of lysosomes in neurons and decreased levels of LC3-II following rotenone treatment, being indicative for inhibition of autophagy.

#### Huntington's Disease

Huntington's disease is caused by autosomal dominant mutation of the Huntingtin gene (Htt) and the onset of the disease is in average much earlier than AD and PD. Patients with HD display uncontrolled chorea movements and cognitive impairment (Ghavami et al., 2014). The onset of age also inversely correlates with the increasing number of repetitive Glutamine motifs present in mutated Huntingtin. Interestingly, capture of mutant Htt inside inclusion bodies was shown to be less toxic in comparison to accumulating free mutant Htt (Zuccato et al., 2010). Also in HD patients, apoptosis and mitochondrial damage is detected. Additionally, studies have demonstrated that rRNA transcription is affected in HD (Parlato and Kreiner, 2013). Triggering autophagy in mice models of HD can remove aggregates and increases their life span (Zheng et al., 2010).

Targeted disruption of nucleoli by conditional knockout of TIF-1A essential for the recruitment of RNA pol I in striatal neurons results in striatal degeneration and typical HD-like phenotypic alterations in mice (Kreiner et al., 2013). TIF-1A loss induces nucleolar disruption and nucleolar stress, which precedes neurodegeneration. Nucleophosmin (NPM) represents a multifunctional, nucleolar key factor involved in ribosome biogenesis, which fulfills a plethora of pro-survival processes (Colombo et al., 2011; Lindstrom, 2011). A down-regulation of NPM serves as readout for nucleolar stress induction and can be linked to neurodegeneration in several models of neurodegeneration (Marquez-Lona et al., 2012). In line, as a key marker for nucleolar stress (Colombo et al., 2002), NPM is reduced and p53 is stabilized in this model (Kreiner et al., 2013). As an early and p53-dependent pro-survival response, the p53 target PTEN (Stambolic et al., 2001) is induced in neurons. Given that the tumor suppressor PTEN counteracts the mTOR pathway, downstream targets of mTOR were analyzed for phosphorylation. It was found that p-S6 and p-4E-BP1 are reduced in the model. Inhibition of mTOR is connected to activation of autophagy and the same holds true in the HD model (Kreiner et al., 2013). Thus, transient over-activation of autophagy seems to be induced as initial, neuroprotective mechanism in response to impaired ribosome biogenesis. However, after sustained nucleolar stress, apoptosis of striatal neurons is inevitable (Kreiner et al., 2013; Parlato and Liss, 2014).

#### Cockayne Syndrome

fncel-13-00156 April 29, 2019 Time: 10:31 # 9

DNA damage and impaired rRNA transcription are connected to Cockayne Syndrome (CS), which is a rare, congenital, autosomalrecessive neurodegenerative disorder (Karikkineth et al., 2017; Hetman and Slomnicki, 2019). Patients are characterized by premature aging, dwarfism, microcephaly, and have an average life expectancy of 12 years. Commonly mutated genes are CSA (20% of cases) and CSB (80% of cases), which are both, besides their key role in nucleotide excision repair, also required for RNA pol I transcription (Lebedev et al., 2008; Koch et al., 2014). CSB was found to localize to mitochondria and bind to mtDNA (Aamann et al., 2010; Kamenisch et al., 2010). CSB-deficient cells show increased ROS production, increased mitochondrial content and accumulation of damaged mitochondria, in line with impaired mitophagy (Scheibye-Knudsen et al., 2012). Further, CSB promotes acetylation of α-tubulin (Majora et al., 2018), which is a modification involved in cargo transport along microtubules to facilitate autophagosome/autolysosome fusion and aggresome clearance (Xie et al., 2010; Li et al., 2011). CSB deficiency abrogates autophagy and results in increased number of dilated lysosomes with impaired function (Majora et al., 2018). In human CS cells, translational infidelity is observed, most likely due to accumulation of error-prone ribosomes as a consequence of impaired ribosome replacement. CS cells exhibit ER stress and an over-activated unfolded protein response, which can be counteracted by addition of pharmacological chaperones (Alupei et al., 2018).

#### Epilepsy – A Disease Related to Neurodegeneration

Epilepsy is characterized by recurrent seizures and represents a disease related to neurodegeneration. Abrogated morphogenesis and synaptic function is observed upon nucleolar stress and could be connected to epilepsy (reviewed in Hetman and Slomnicki, 2019). For example, pharmacologically induced shortterm seizures in mice transiently affect RNA Pol I activity in hippocampi and result in decreased de novo synthesized 18S and 28S rRNAs. In contrast, long-term seizures were associated with increased ribosome biogenesis (Vashishta et al., 2018). Epilepsy is further tightly linked to mTOR hyper-stimulation and autophagy over-activation (Cao et al., 2009; Zeng et al., 2009). Strikingly, administration of the mTOR inhibitor rapamycin counteracts seizures and thus functions in an anti-epileptogenic manner (Zeng et al., 2009).

#### Rare, Pediatric Neurodegenerative Diseases – A Short Outlook

Also rare, pediatric neurodegenerative diseases are characterized by alterations in autophagy. As an example, the multisystemic Vici syndrome is neurologically characterized by microcephaly and cognitive impairment. Accumulation of ubiquitinated autophagic cargo, p62 and damaged mitochondria is observed, reminiscent of neurodegeneration (Ebrahimi-Fakhari et al., 2014). However, whether also ribosome biogenesis is also altered here, remains to be determined.

#### Cancer and Cancer Treatment

Key hallmarks of cancerous cells involve for instance misregulation of signaling pathways, rapid cell proliferation, accelerated tumor growth and inhibition of apoptosis (Hanahan and Weinberg, 2000, 2011). Large amounts of ribosomes are essential for cancerous cell growth and large nucleoli serve as prognostic marker in many tumor types (Montanaro et al., 2008; Penzo et al., 2019).

The nucleus arises as an essential target for cancer therapy (Woods et al., 2015; Pfister and Kuhl, 2018). Anti-tumor therapies utilize the high demand of cancer cells for the production of ribosomes by inhibiting RNA pol I. Inhibiting nucleolar structure/function and RNA pol I function has been characterized as beneficial in terms of triggering apoptotic cell death of cancer cells (Burger and Eick, 2013). Classical chemotherapeutics used in the clinics are for instance actinomycin D, 5-fluorouracil and metotrexat, which interfere with the nucleolar function (Boulon et al., 2010; Burger and Eick, 2013).

Novel drugs, which specifically impair rRNA transcription, are currently tested in clinical trials. The small molecule drug CX-5461 specifically inhibits transcription of RNA pol I and stabilizes p53, whereas RNA pol II is not affected. Also, translation and DNA replication is not impaired in human cancer cell lines (Drygin et al., 2011). The drug is further reported to impair proliferation in a p53-independent manner in cancer cell lines, whereas the effect on normal cell lines is minimal (Drygin et al., 2011).

Selective inhibition of RNA pol I by CX-5461 also robustly stimulates pro-death autophagy. Nucleolar stress and autophagy seem to be tightly coupled in distinct models and setups. Recently, CX-5461 was loaded on a nanoplatform to enrich for nucleolar accumulation of the drug in order to enhance the anticancer effect, without causing significant side effects. In vivo and in vitro assays confirm induction of pro-death autophagy in HeLa cells, as well anti-proliferative and anti-tumor effects (Duo et al., 2018).

Besides autophagy, CX-5461 induces also senescence in a p53-independent manner. In U2OS osteosarcoma cells, CX-5461 induces G2 arrest, but not apoptosis (Li et al., 2016). In response to CX-5461, p53 accumulates and p21 is induced. In addition, increased levels of LC3-II are detected under basal conditions. Using TEM analysis, the authors noticed expanded vacuole-like structures filled with organelles, however, they report lack of clear identification of autophagosomal character. Knockdown of p53 by siRNA rescues p21 up-regulation and LC3-II accumulation and increases cell survival. The authors conclude that CX-5461 triggers p53-dependent autophagy. Autophagy occurs via the AMPK/mTOR pathway in U2OS cells, as evidenced by reduced p-mTOR and increased p-AMPK levels (Li et al., 2016). The p53 target p21 is shown to be up-regulated during autophagy and a p53-independent increase of p21 is reported in MNNG cells with mutant p53 (Li et al., 2016).

Taken together, autophagy induction as a response to nucleolar stress seems to be an initial surveillance mechanism in several models. However, also in terms of cancer, autophagy induction can have two modes of action: Autophagy induction is clearly beneficial for cells by preventing genotoxic stress and DNA damage. It removes cellular sources of ROS, such as defective mitochondria or proteins (Mrakovcic and Frohlich, 2018). In contrast, inhibition of autophagy represents an oncogenic event. At later stages over-activation of autophagy facilitates oncogenic drug-resistance. Autophagy inhibitors chloroquine and hydroxychloroquine have therefore been tested in clinical trials for cancer therapy (Yang et al., 2011; Marino et al., 2014).

#### Ribosomopathies

fncel-13-00156 April 29, 2019 Time: 10:31 # 10

Impairment of ribosome biogenesis is connected to a diverse class of human diseases collectively termed ribosomopathies (Freed et al., 2010; Narla and Ebert, 2010; Danilova and Gazda, 2015; Yelick and Trainor, 2015). Patients exhibit either mutations and/or haploinsufficiency of RPs or ribosome biogenesis factors. Several players associated with ribosomopathies have been described (see also below). Classically, the nucleolar stress response and the tumor suppressor p53 are activated (Freed et al., 2010; James et al., 2014). Despite a common mechanism of interfering with ribosome biogenesis, the patient's phenotypes differ among the distinct syndromes. Intriguingly, though, some phenotypes are common and include defects of the craniofacial cartilage, anemia and increased cancer susceptibility. The elevated cancer risk appears paradoxical, given the great need of tumor cells for large amounts of ribosomes (Montanaro et al., 2008). Accordingly, pathways and mechanisms might well exist, which let both co-exist (Pfister and Kuhl, 2018). For example, specialized onco-ribosomes have recently been uncovered to increase the cellular fitness by mediating preferential translation of anti-apoptotic Bcl-2, as observed for the ribosome mutant RPL10-R98S in leukemia cells (Xue and Barna, 2012; Sulima et al., 2017; Kampen et al., 2018). However, the question on cause and consequence of ribosomopathy-induced cancer formation is still under debate.

Recently, examination of murine hepatocellular carcinoma and hepatoblastoma has revealed ribosomopathy-like features of nucleolar stress, such as deregulated expression of RPs and accumulation of unprocessed rRNA precursors. Despite the fact that p53 is stabilized, no growth inhibition occurs (Kulkarni et al., 2017). Therefore, compensating mechanisms might counteract apoptosis, involving up-regulation of antiapoptotic Bcl-2, silencing of p19 ARF or cytosolic sequestration of p53. Those events would in turn inhibit the tumor suppressive mechanisms of cell cycle arrest and apoptosis in these cancers (Kulkarni et al., 2017).

Clearly, ribosome biogenesis is a highly energy-consuming process. An implication of autophagy comes in mind, which might compensate for the low levels of functional ribosomes observed in ribosomopathies. Many important questions arise. Is there a connection to bulk autophagy or selective autophagy of ribosomes (ribophagy) in ribosomopathy patients? Until now broad studies are lacking, which precisely address their implication in these issues. However, first data are collected, which indeed unravel involvement of autophagy in these processes (see section "Nucleolar Factors and Nucleolar Stress in the Regulation of Autophagy and Vice Versa").

An overview summarizing the emerging connection between nucleolar stress and autophagy in the diseases presented here is given in **Table 1**.

#### NUCLEOLAR FACTORS AND NUCLEOLAR STRESS IN THE REGULATION OF AUTOPHAGY AND VICE VERSA

Recently, inhibition of RNA pol I has been connected to autophagy, revealing that nucleolar stress functions upstream of autophagy. In the following, evidence is collected, which links the ribosome biogenesis machinery and the nucleolus to autophagy, and vice versa. As a common principle, different groups suggest implication of mTOR signaling in nucleolusmediated autophagy (see below). Also here, p53-dependent and -independent pathways are being identified.

# The p53 Family

Besides the classical role of p53 as guardian of the genome, by mediating cell cycle arrest and apoptosis, p53 has been reported

#### TABLE 1 | The role of nucleolar stress in mentioned diseases and effects on autophagy.


<sup>1</sup>Please, see text for references and details.

to exert distinct roles in autophagy (Wang et al., 2014). This depends on its subcellular localization: nuclear, cytosolic or mitochondrial, respectively.

The effect of nuclear p53 as an inducer of autophagy mostly depends on its role as transcription factor. Nuclear p53 induces expression of ATGs thereby driving autophagy. In line, many promoters of autophagy related factors, such as ATGs or Parkin, are occupied by p53 (Zhang et al., 2011; Fullgrabe et al., 2016). Induction of pro-apoptotic target genes implicated in mTOR activation, such as TCS2, AMPK/PTEN and Sestrin, result in autophagy activation. The p53 target DRAM is directly involved in autophagosome formation. P53 further induces BAX and Bcl-2 or DAPK, which in turn induce Beclin1 (Mrakovcic and Frohlich, 2018). Moreover, the p53 family members p63 and p73 induce expression of the autophagy regulators ATG5 and ATG7. Also E2F, an important co-regulator of p53, is involved in the transcriptional regulation of autophagy related genes (Polager and Ginsberg, 2009; Fullgrabe et al., 2016). Note that other p53 family members p63 and p73 can, in principal, compensate for a loss of p53 (Kenzelmann Broz et al., 2013; Fullgrabe et al., 2016). However, it remains to be determined, whether p53-independent pathways actually depend on the role of p63 and p73 or whether they are unrelated.

Cytosolic p53 counteracts autophagy by transcriptionindependent mechanisms. P53 inhibits AMPK and activates mTOR, p53 further interacts with Beclin and induces Beclin ubiquitination and degradation (Mrakovcic and Frohlich, 2018). Cytosolic p53 interacts with Parkin, which is the key regulator of mitophagy. It was reported that p53 counteracts Parkin recruitment to mitochondria, thereby impairing mitophagy (Hoshino et al., 2013).

Mitochondrial p53 has a plethora of functions: it triggers MOM permeabilization, ROS production, mitophagy and autophagy and is therefore implicated in neuropathological conditions (Marino et al., 2014; Wang et al., 2014).

#### mTOR Inhibition by Rapamycin

mTORC1 inhibition by rapamycin abrogates the nucleolar stress response induced by low, cytostatic doses of the chemotherapeutic actinomycin D (Goudarzi et al., 2014). As a result, p53 stabilization and p21 induction is impaired. The authors observe decreased interaction of RPL11 with MDM2 upon rapamycin and actinomycin D co-treatment and suggest a mechanism related to the classical RPL11-MDM2-p53-pathway. Also, they detected decreased RPL11 levels, as well as MDM2 stabilization, which might in part contribute to the rapamycinmediated effects on p53. Interestingly, inhibition of mTOR by caffeine at physiologically relevant doses is capable of abrogating the actinomycin D-induced p53 response (Goudarzi et al., 2014). Thus, a complex network between mTOR inhibitors/autophagy, nucleolar stress and cancer treatment is established.

# RNA pol I Inhibition and NPM

RNA pol I inhibition by actinomycin D or adriamycin is well known to trigger nucleolar disruption. Recently, the observation has been made that also autophagy can be induced with these drugs (Katagiri et al., 2015). The same holds true for knocking down the RNA pol I transcription factors TIF1A and POLR1A. Induction of autophagy has been monitored in flux experiments by counting LC3 punctae (being indicative for autophagosomal number) and determining levels of lipidated LC3-II (representing the autophagosome-bound form of LC3). At the same time the autophagy substrate p62/sequestosome is reduced, indicating increased turnover by autophagy. Autophagy induction can be rescued by treating cells with autophagy inhibitors or knocking down key autophagy regulators ATG5 and Beclin1. In contrast, nucleolar disruption is not rescued (Katagiri et al., 2015). Together, this finding places nucleolar disruption upstream of autophagy. As nucleolar stress is generally characterized by redistribution of nucleolar factors (such as NPM), or p53 stabilization, Katagiri et al. (2015) have determined, which accounts for the effects observed. They found that induction of autophagy by TIF1A knockdown is independent of p53, but depends on NPM. The induction of autophagy can be rescued by NPM knockdown, without reducing p53 levels. In contrast, neither the depletion of NPM affects starvation-induced autophagy; nor does nutrient deprivation have an impact on nucleolar integrity. This suggests that NPM might rather play a role in a specialized form of nucleolar stress-induced autophagy, than starvationinduced bulk autophagy.

# The Nucleolar Factor – PICT-1

The nucleolar factor PICT-1/GLTSCR2 is considered a tumor suppressor, as it binds and stabilizes PTEN. In contrast, PICT-1 deletion is linked to cancer formation and functions as oncogenic regulator of the E3-ubiquitin ligase MDM2 by preventing nucleolar release of RPL11 (Sasaki et al., 2011; Suzuki et al., 2012). Consequently, p53 stabilization, G1 cell cycle arrest and apoptosis are observed, thereby counteracting tumor growth (Sasaki et al., 2011). Homozygous PICT1 deletion in mice is lethal and impairs survival of mouse ES cells. PICT-1 binds to rDNA and the RNA pol I transcription factor upstream binding factor-1 (UBF-1). It inhibits transcription of rRNA, which depends on its localization to nucleoli (Chen et al., 2016). With respect to autophagy, it has been shown that PICT-1 overexpression induces GFP-LC3 punctae and reduces p62 levels, and that it inhibits the AKT/mTOR/p70S6K pathway (Chen et al., 2016). As the autophagy inhibitor 3-MA (3-methyladenine) enhances cell death upon PICT-1 overexpression, the authors suggest induction of pro-death autophagy. Together, the authors conclude that PICT-1 overexpression triggers prodeath autophagy, without inducing the classical nucleolar stress response, such as p53 stabilization and nucleolar disruption (Chen et al., 2016).

# The Ribosomopathy Factor – SBDS

Recent findings suggest that autophagy might be affected in patients with ribosomopathies. mTOR signaling regulates a variety of essential cellular processes, among them autophagy. In leukocytes derived from patients with the ribosomopathy syndrome Shwachman Bodian Diamond Syndrome (SBDS), a hyper-activation of mTOR phosphorylation is observed (Bezzerri et al., 2016). Also in intestinal epithelial cells autophagy

is over-activated. In this context, autophagy is independent of mTOR or p53 and is induced as a consequence of nucleolar/ribosomal stress (Bezzerri et al., 2016).

#### The Ribosomopathy Factor – RPS19

Disrupted ribosome biogenesis by knocking down RPS19 (ribosomal protein S19) is connected to the ribosomopathy syndrome Diamond Blackfan Anemia (DBA). RPS19 knockdown further affects autophagy in patient cells, and autophagy induction is also observed in red blood cells of zebrafish embryos (Heijnen et al., 2014).

RP-deficiency also recapitulates these effects in cells derived from SBDS. An increase in P-S6 as well as ROS is observed, whereas anti-oxidant treatment reverses p-S6, autophagy and p53 stabilization (Heijnen et al., 2014). Thus, the observed effects turn out to be ROS-dependent and suggest a contribution of oxidative stress in ribosomopathies.

#### The Ribosomopathy Factor – pwp2h

The zebrafish titania mutant (ttis450) harbors a recessive, lethal mutation of the pwp2h gene encoding a factor of the small ribosomal subunit (Boglev et al., 2013). In this mutants reduced 18S rRNA, impaired ribosome biogenesis and p53 stabilization is observed. pwp2h is highly expressed in intestinal, epithelial cells, but also in the brain retinal pigmented epithelium, liver and pancreas, which are rapidly dividing tissues. Also here, defects in craniofacial formation, typical hallmarks of ribosomopathies, can be detected.

Intestinal epithelial cells of the mutant larvae display accumulation of autophagosomes. In autophagic flux experiments using the autophagy inhibitor chloroquine and activator rapamycin, increased accumulation of LC3-II is observed in the mutants, indicating autophagy induction (Boglev et al., 2013). Also p-RP-S6, reflecting mTORC1 activity, is increased. Inhibiting autophagy by morpholino-mediated knockdown of ATG5 triggers apoptosis of intestinal epithelial cells specifically in the mutants, whereas the wildtype is not affected. This suggests that autophagy induction counteracts apoptosis as survival mechanism in response to nucleolar stress. Also, no signs of apoptosis are detected in the mutants, ruling out toxic levels of autophagy activation. Interestingly, autophagy induction in the zebrafish mutants occurs in mTOR and p53-independent manner (Boglev et al., 2013). However, the molecular mechanisms and pathways affected remain elusive.

# The Drosophila Nopp140

Nopp140 is in structure and function related to Treacle, representing an essential gene in the ribosomopathy syndrome Treacher Collins syndrome (TCS) (Valdez et al., 2004; Sakai and Trainor, 2009; Dai et al., 2016). Depletion of the nucleolar phosphoprotein Nopp140 in the imaginal wing disks of Drosophila results in nucleolar stress, loss of ribosomes and p53 independent apoptosis (James et al., 2013). Since there are no detectible levels of MDM2, NPM/B23 and ARF in Drosophila, the authors conclude that an alternative nucleolar stress response might exist. They consider JNK as an interesting link, which has earlier been shown to induce autophagy in response to oxidative stress and induce transcription of ATG genes (Wu et al., 2009; Zhou et al., 2015). In addition, oxidative stress induces accumulation of the autophagy marker GFP-LC3 and lysosomes in the intestinal epithelium, which is dependent on JNK signaling (Wu et al., 2009). As also in larval polyploidy midgut cells mCherry-ATG8a is accumulated after Nopp140 depletion, the authors conclude an accumulation of autophagosomes and a premature induction of autophagy regulated by loss of Nopp140 (James et al., 2013).

#### NAT10 and Glucose Deprivation Stress

NAT10 drives ribosome biogenesis by mediating acetylation of the RNA pol I transcription factor UBF-1 and facilitating processing of the 18S rRNA (Kong et al., 2011; Ito et al., 2014). Under normal conditions, NAT10 is auto-acetylates and promotes recruitment of PAF53 and RNA pol I to mediate rRNA biogenesis (Cai et al., 2017), whereas autophagy is inhibited (Liu et al., 2018).

The mechanisms, which link inhibition of rRNA biogenesis to induction of autophagy in response to energy stress were determined by Liu et al. (2018). They demonstrate that NAT10 binds and acetylates the autophagy regulator Che1 (AATF) in a p53-independent manner. As a consequence of acetylation, the transcriptional activation of target genes Redd1 and Deptor is off. Thus, Che1 enhances autophagy by activating the transcription of Redd1 and Deptor, two critical inhibitors of mTOR signaling (Desantis et al., 2015a).

Interestingly, Che-1 is also important for RNA pol II, the DNA damage response (DDR) and drives p53 expression. Upon DDR, Che1 increasingly interacts with p53 and drives the expression of genes implicated in cell cycle regulation, for instance p21 (Desantis et al., 2015b).

Upon energy stress and glucose deprivation, Sirt1 deacetylates NAT10. ChIP analysis has demonstrated that deacetylated NAT10 does not bind to rDNA upon glucose deprivation and thus NAT10-mediated ribosome biogenesis is inhibited. Under this condition, the inhibition of Che1 is released (Liu et al., 2018).

In HCT116 cells, LC3-II levels are increased both in presence or absence of chloroquine, showing that NAT10 knockdown increases basal autophagic flux. Also, p62 is reduced upon NAT10 depletion. Strikingly, the effects observed are independent of p53, as demonstrated in HCT116 p53−/<sup>−</sup> cells. In rescue experiments with HCT116-NAT10-Cas9 knockout clones, overexpression of NAT10 reverses the effects observed on p62 and LC3-II, whereas acetylation mutants fail to do so (Liu et al., 2018). Glucose withdrawal triggers release of NAT10 from nucleoli. Treatment of cells with RNAseA also leads to a loss of NAT10 from nucleoli and impairs binding to UBF-1, suggesting that rRNA binding is essential for nucleolar accumulation of NAT10. In contrast, the acetylation status does not matter as determined by use of a NAT10 mutant.

Together, Sirt1 mediated deacetylation of NAT10 impairs rRNA biogenesis and results in release of NAT10 from nucleoli. Therefore, both regulate the switch between ribosome biogenesis and autophagy as a response to energy stress to maintain cellular survival.

# The Autophagy Regulator LC3/Atg8 – Present in the Nucleolus

Besides its localization in the autophagosomal membranes and the cytosol, LC3 can rapidly shuttle into and out of the nucleus. Endogenous and overexpressed LC3 has been found to be associated with nucleoli. However, the authors point out a weak signal being indicative for a low degree of enrichment (Kraft et al., 2016). A triple arginine motif is essential for the nucleolar targeting of LC3. The arginine motif mediates protein-protein interaction and binding to RNAs, suggesting accumulation at nucleoli by interaction with rRNAs and/or nucleolar RNA binding proteins (Zhou et al., 1997; Behrends et al., 2010). Likewise, LC3 has been connected to interaction with the 60S ribosomal subunit (Kraft et al., 2016). In addition, a hydrophobic binding interface contributes to nucleolar localization. In contrast, the lipidation site of LC3 is dispensable for nucleolar targeting. Interestingly, several 40S RPs such as S27, S5, S18, and S20, have been identified as interaction partners of LC3 by MS analysis. The authors speculate that nucleolar LC3 might counteract p53 stabilization by preventing interaction of RPs with MDM2 (Kraft et al., 2016).

#### TP53INP2/DOR

TP53INP2/DOR (tumor protein p53 inducible protein nuclear protein 2) is a nuclear protein, which interacts with the LC3-related GABARAP, GABARAP-like2, as well as with LC3 via its N-terminus. TP53INP2/DOR also binds to VMP1, a transmembrane protein of autophagosomes. Upon starvation by rapamycin or EBSS incubation, TP53INP2-EGFP translocates from nuclei to the cytoplasm where it co-localizes with LC3 family proteins, indicating autophagosomal recruitment (Nowak et al., 2009). Knockdown of TP53INP2 in HeLa cells reduces rapamycin-induced accumulation of LC3-II, as well as the number of RFP-LC3 punctae per cell. Also, less Beclin1 is recruited to the autophagosomes upon knockdown of TP53INP2. The rapamycin-induced recruitment of TP53INP2 to autophagosomes is dependent of autophagy: it depends on ATG5, as demonstrated in ATG5−/<sup>−</sup> MEFs, and it is stimulated by rapamycin. Further, it depends on PI3K as revealed by inhibition by the PI3K inhibitor wortmannin (Nowak et al., 2009).

Huang and Liu (2015) determined the molecular mechanisms underlying recruitment of LC3 from the nucleus to the cytoplasm (Liu and Klionsky, 2015). By use of an NES mutant of TP53INP2/DOR, which cannot exit the nucleus, LC3 is captured in the nucleus under starvation. This data suggest that TP53INP2/DOR mediates export of nuclear LC3 during autophagy (Huang and Liu, 2015). Also, loss of Beclin1 and ATG14 inhibits the exit (Huang and Liu, 2015). Upon nutrient deprivation, SIRT1 deacetylates nuclear LC3 at K49 and K51. SIRT1 is activated by metabolic stress and functions as essential activator of autophagy by deacetylating its substrates, among them p53 and ATGs (Vaziri et al., 2001; Lee et al., 2008; Lapierre et al., 2015). This increases the interaction of TP53INP2/DOR with deacetylated LC3 and mediates cytosolic export of LC3. The authors show that LC3 derived from the nucleus is the primary source of membrane-bound LC3. As a consequence, LC3 is lipidated and autophagosome biogenesis is initiated. Therefore, TP53INP2/DOR has a key role as a scaffold for LC3, by mediating LC3 lipidation. SIRT1 is essential for deacetylation, which is required for the interaction with ATG7 (Huang and Liu, 2015).

The autophagy regulator TP53INP2/DOR has recently been found in nucleoli. TP53INP2/DOR-RFP co-localizes with Nucleophosmin-GFP and Fibrillarin in HeLa cells. Upon co-expression of a dominant negative NPM mutant, the double leucine mutant NPMdL, which cannot exit the nucleus (Maggi et al., 2008), TP53INP2/DOR is still able to perform nucleo-cytoplasmic shuttling. However, on longer term its localization to the nucleolus is impaired (Mauvezin et al., 2012). Localization of TP53INP2/DOR is mediated by a C-terminal nucleolar localization signal (NoLS) (Xu et al., 2016). ChIP analysis identified binding to rDNA, and a nucleolar exclusion of TP53INP2/DOR led to impaired rRNA synthesis. TP53INP2/DOR is capable of directly binding to the RNA pol I pre initiation complex.

In summary, TP53INP2/DOR has a dual function by mediating ribosome biogenesis in the nucleolus and regulating autophagy.

#### Drosophila – RPD3

Rpd3 represents a Drosophila homologue of histone deacetylase 1 (HDAC1). At early stages of starvation, Rpd3 accumulates in the nucleolus and co-localizes with the nucleolar factor Fibrillarin in Drosophila. Following starvation, 18S rRNA and Rpd3 mRNAs increases transiently. ChIP analysis has demonstrated that Rpd3 binds to the rDNA promoter. Rpd3 upregulates rRNA synthesis, whereas a Rpd3 knockdown reduces the number of polysomes during starvation (Nakajima et al., 2016). Also, knockdown flies die faster in response to starvation. In the knockdown flies, reduced levels of Atg9 mRNA are detected, which is induced in response to starvation in controls, revealing decreased tolerance to starvation-mediated stress.

#### Drosophila – Nucleostemin-Like 2

In Drosophila, nucleostemin-like protein NS2, which is a homologue of human NGP1/GNL2, localizes to nucleoli. Loss of NS2 results in ribosomal stress and block of nucleolar release of 60S subunits as evidenced by increased GFP-RPL11 in nucleoli (Wang and DiMario, 2017). In polyploid midgut cells, mCherry-Atg8a positive autophagosomes are detected by immunofluorescence analysis, as well as autophagosomes containing mitochondria by TEM analysis. In contrast, in larval imaginal disks induction of apoptosis is observed (Wang and DiMario, 2017).

# CLIP – Nucleophagy of Nucleolar Factors in Yeast

In budding yeast, nutrient deprivation and TORC inhibition triggers nucleophagy, the selective degradation of the nuclear compartment. In particular, nucleolar factors are degraded,

whereas rDNA is excluded (Mostofa et al., 2018). The authors establish that autophagy induction by rapamycin triggers the redistribution of nucleolar proteins and rDNA, thereby separating rDNA from nucleophagy. CLIP and cohibin, which are essential for tethering rDNA to the inner nuclear membrane, are responsible for the repositioning of rDNA and nucleolar proteins in yeast. They are also required for the nucleophagic degradation of nucleolar factors. In contrast, rDNA is not degraded by macronucleophagy (Mostofa et al., 2018).

Thus, starvation-mediated autophagy, at least in yeast, removes specifically nucleolar factors. It seems likely that autophagy allows inhibition of the energy-consuming process of ribosome biogenesis by selectively removing the processing machinery.

A simplified model summarizing the emerging connection between nucleolar stress and autophagy presented in this review is given in **Figure 5**.

# CLOSING REMARKS AND PERSPECTIVES

Despite the recent advances made in uncovering the relationship between nucleolar stress and autophagy, our understanding is far from being complete. A critical need to elucidate the underlying causes is apparent.

On a mechanistic level it seems likely that p53-dependent as well as -independent effects account for the induction of the autophagic nucleolar stress response.

It might well be that reduced ribosome biogenesis, by reducing protein synthesis, triggers autophagy as a general stress response. Nevertheless, extra ribosomal functions might as well coincide.

As precise underlying mechanisms are currently missing, many questions and possibilities rise: How are both signaling pathways integrated with each other? Which effects are cell type and context dependent? Under which conditions is autophagy either beneficial or a contributor to the pathology? Is there a p53 or compensatory p63/p73 autophagy response triggered in nucleolar stress? Does selective autophagy and other forms of autophagy such as micro-autophagy play a fundamental pathogenic role in other diseases connected to nucleolar stress?

The initial and common concepts of nucleolar stress and autophagy open novel avenues for investigating specific therapeutic approaches. Many autophagy inhibitors and activators might be contemplated as therapies for nucleolarstress mediated diseases. They might well be combined with, e.g., RNA pol I inhibitors and tested for synergy to increase selectivity but simultaneously reduce toxicity for patients.

To get a deeper understanding of the underlying events, a challenge will be to elaborate state of the art autophagy methods on monitoring the autophagy flux in cellular models of diseases, patient tissues and blood samples (Jiang and Mizushima, 2014). Research on the autophagic transcriptome and spatio-temporal expression patterns of regulators of the autophagic machinery or nucleolar response will advance our knowledge on mechanisms coupling the nucleolar stress response to autophagy.

# AUTHOR CONTRIBUTIONS

AP performed the literature research, conceptualization, and wrote the review.

# FUNDING

AP lab is supported by Ulm University (Baustein 3.2, LSBN0101) and by Deutsche Krebshilfe (70112329).

# ACKNOWLEDGMENTS

I would like to thank D. Dannheisig for initial help with the graphical illustrations. My apologies to all authors contributing, excellent work I could not cite owing space limitations.

# REFERENCES


Granneman, S., and Baserga, S. J. (2004). Ribosome biogenesis: of knobs and RNA processing. Exp. Cell Res. 296, 43–50. doi: 10.1016/j.yexcr.2004.03.016


rapamycin-related signaling pathways in osteosarcoma. OncoTargets Ther. 9, 5985–5997. doi: 10.2147/ott.s104513


the accumulation of damaged mitochondria by promoting mitochondrial autophagy. J. Exp. Med. 209, 855–869. doi: 10.1084/jem.20111721


**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 Pfister. 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.

# Roles of Primary Cilia in the Developing Brain

#### Sang Min Park<sup>1</sup> , Hee Jin Jang<sup>1</sup> and Jeong Ho Lee1,2 \*

<sup>1</sup> Biomedical Science and Engineering Interdisciplinary Program, Korea Advanced Institute of Science and Technology, Daejeon, South Korea, <sup>2</sup> Graduate School of Medical Science and Engineering, Korea Advanced Institute of Science and Technology, Daejeon, South Korea

Essential to development, primary cilia are microtubule-based cellular organelles that protrude from the surface of cells. Acting as cellular antenna, primary cilia play central roles in transducing or regulating several signaling pathways, including Sonic hedgehog (Shh) and Wnt signaling. Defects in primary cilia contribute to a group of syndromic disorders known as "ciliopathies" and can adversely affect development of the brain and other essential organs, including the kidneys, eyes, and liver. The molecular mechanisms of how defective primary cilia contribute to neurological defects, however, remain poorly understood. In this mini review, we summarize recent advances in understanding of the interactions between primary cilia and signaling pathways essential to cellular homeostasis and brain development.

#### Edited by:

Rosanna Parlato, University of Ulm, Germany

#### Reviewed by:

Elisabeth Traiffort, Institut National de la Santé et de la Recherche Médicale (INSERM), France Esther Stoeckli, University of Zurich, Switzerland

> \*Correspondence: Jeong Ho Lee jhlee4246@kaist.ac.kr

#### Specialty section:

This article was submitted to Cellular Neurophysiology, a section of the journal Frontiers in Cellular Neuroscience

> Received: 16 January 2019 Accepted: 30 April 2019 Published: 14 May 2019

#### Citation:

Park SM, Jang HJ and Lee JH (2019) Roles of Primary Cilia in the Developing Brain. Front. Cell. Neurosci. 13:218. doi: 10.3389/fncel.2019.00218 Keywords: primary cilia, Wnt, MTOR, autophagy, ciliopathy, FMCD

#### INTRODUCTION

Primary cilia are microtubule-based cellular organelles that aid in sensing and transducing environmental signals during development (Goetz and Anderson, 2010): primary cilia transduce or regulate several signaling pathways, including Sonic hedgehog (Shh) and Wnt signaling (Corbit et al., 2005; Rohatgi et al., 2007; Gerdes and Katsanis, 2008; Goetz et al., 2009). This antenna-like structure was first observed using electron microscopy at the lumen of kidney tubules in 1898 and has since been intensively studied (Zimmermann, 1898). Emanating from the apical surface of basal body, the axoneme structure of primary cilium comprises a radial array of nine microtubule doublets lacking a central pair (9+0 structure) (**Figure 1A**; Davenport and Yoder, 2005). At the base of primary cilium, protein entry and exit are regulated via a transition fiber that anchors the axoneme to the ciliary membrane, compartmentalizing this distinct organelle from the cytosol while remaining continuous with the plasma membrane (Gherman et al., 2006; Marshall, 2008; Reiter et al., 2012).

Leading to defective primary cilia, mutations in genes necessary for ciliogenesis and ciliary structure and function are known to cause a group of human genetic disorders described as "ciliopathies." To date, 187 mutated genes in 35 known ciliopathies and 241 ciliopathyassociated genes essential to ciliary structure and function that could potentially be causative for ciliopathies have been documented (Reiter and Leroux, 2017). These ciliopathies affect the body's essential organs, including the kidneys, eyes, brain, liver, and skeleton, during development and tend to share clinical phenotypes (Hildebrandt et al., 2011): for example, as primary cilia are critical to development of the central nervous system (CNS), many ciliopathies, such as Joubert syndrome (JBTS), Meckel syndrome (MKS), and orofaciodigital syndrome (OFD), commonly exhibit neurological defects of CNS malformation, intellectual disability, ataxia, and retina

FIGURE 1 | Primary cilia and signal transduction. (A) Structure of primary cilium. Microtubules extend from the centriole constituting the basal body and form the axoneme. The surrounding membrane is called the ciliary membrane, distinct from other membranes. Unlike motile cilium with a 9+2 structure, primary cilium has a 9+0 structure. Only ciliary proteins are allowed to access the cilium, and the transition zone performs the filtering function. (B) In the absence of Shh, Patched (Ptch) inhibits the translocation of Smoothened (Smo), a membrane G protein-coupled receptor (GPCR)-like protein. Suppressor of Fused (SUFU) in the tip of the cilium represses Gli, causing Gli to be present in an inactive form. In the presence of Shh, Shh binds to Ptch, making it no longer able to suppress Smo. Smo, relieved from suppression, translocates into the cilium and represses SUFU. Gli is then converted to its active form and translocates from the cilium to the nucleus and transcribes target genes. (C) Primary cilia in Wnt signaling. In the absence of Wnt ligand, the β-catenin destruction complex, comprising glycogen synthase kinase 3β (GSK3β), adenomatous polyposis coli (APC), and Axin, phosphorylates β-catenin, leading to its proteasomal degradation, which occurs near the basal body. Wnt signaling is activated by binding extracellular Wnt ligand to the membrane-bounded receptor family, frizzled and low-density lipoprotein receptor-related protein (LRP). Then, the β-catenin destruction complex is destabilized by anchoring Axin to the plasma membrane through Dishevelled (Dvl), which leads to stabilization and nuclear localization of β-catenin for transcriptional activation of target genes under T-cell factor/lymphoid enhancer-binding factor (TCF/LEF) promoters (van Amerongen and Nusse, 2009). Inversin, for which ciliary localization has been reported, mediates proteasomal degradation of Dvl to regulate Wnt signaling.

dystrophy (Valente et al., 2014). Here, we summarize our recent understanding of the interactions between primary cilia and several signaling pathways essential for cellular homeostasis and for brain development.

### SIGNALING PATHWAYS AND BIOLOGICAL PROCESSES MEDIATED BY PRIMARY CILIA

#### Shh Signaling

Sonic hedgehog signal transduction is accomplished through binding of Shh to the transmembrane receptor Patched (Ptch). In the absence of Shh, Ptch inhibits the ciliary localization of Smoothened (Smo), a membrane G protein-coupled receptor (GPCR)-like protein. When present, Shh binds to Ptch to alleviate the suppression of Smo, which, in turn, translocates to the primary cilium to repress Suppressor of Fused (SUFU) (Rohatgi et al., 2007; Tukachinsky et al., 2010; Sasai and Briscoe, 2012). This initiates the activation of Gli transcriptional factors (**Figure 1B**; Alcedo et al., 1996): Gli regulates the transcription of target genes that regulate the Shh signaling (e.g., Ptch1, Gli1), proliferation (e.g., cyclin D1, MYC), and apoptosis (e.g., Bcl-2) (Scales and de Sauvage, 2009). Shh, Ptch, Smo, and Gli are expressed in the developing and adult brain (Echelard et al., 1993; Ekker et al., 1995).

Sonic hedgehog signaling is critical to spatial patterning of the neuroepithelium, cellular identity in CNS, axonal guidance, wiring of the neural network, and neuronal activity. As Shh signaling-related proteins are located within primary cilium, Shh signaling is not properly achieved in mice with defective cilia, resulting in several defects in brain development such as defective neural patterning, cerebellar hypoplasia, and defective hippocampal neurogenesis (reviewed in Ferent and Traiffort, 2015). For example, mice with mutation in Kinesin family member 3A (Kif3a), which is important for ciliogenesis, show reduced Gli1 expression, thereby leading to defects in hippocampal neurogenesis and formation of neural stem cells (Han et al., 2008). Stumpy mutant mice exhibit defective ciliogenesis leading to prenatal hydrocephalus and severe polycystic kidney disease. The Stumpy gene, located on mouse chromosome 7, encodes B9 protein domain 2 (B9D2), which is localized with basal bodies along ciliary axonemes and appears to play a role in ciliogenesis in association with intraflagellar transport proteins (IFTs) (Town et al., 2008). Conditional Stumpy mutant mice (driven by Nestin-Cre) have been found to show abrogated Shh signaling and Gli processing in the hippocampus (Breunig et al., 2008).

#### Wnt Signaling

Wnt signaling is another developmental signaling pathway mediated by primary cilia (**Figure 1C**; May-Simera and Kelley, 2012). Wnt signaling can be divided into (1) β-catenin dependent (canonical) and (2) β-catenin independent (non-canonical) signaling (reviewed in Berbari et al., 2009; Oh and Katsanis, 2013). Several studies have shown that perturbation of ciliary genes aberrantly activates canonical Wnt signaling and disrupts non-canonical Wnt signaling (Lin et al., 2003; Cano et al., 2004; Ross et al., 2005; Simons et al., 2005; Gerdes et al., 2007; Corbit et al., 2008). In particular, Inversin, which is localized in cilium, mediates the proteasomal degradation of Dishevelled (Dvl) to regulate both canonical and non-canonical Wnt signaling (Simons et al., 2005). Perturbations in several ciliopathy-related genes have been found to elicit de-regulated canonical Wnt signaling in fibroblasts and in the forebrain (Willaredt et al., 2008; Abdelhamed et al., 2013; Wheway et al., 2013). Also, ciliary ablation in adult-born dentate granule cells in the hippocampus has been shown to aberrantly activate canonical Wnt signaling, concomitant with severe defects in dendritic refinement and synapse formation (Kumamoto et al., 2012). Additionally, researchers have recently demonstrated that defective neuronal ciliogenesis caused by hyperactivating mutation in MTOR elicits abnormal activation of canonical Wnt signaling and inactivation of non-canonical Wnt signaling, resulting in defective neuronal migration due to disrupted neuronal polarization (Park et al., 2018). Notwithstanding, several studies have reported that ciliary dysfunction does not affect Wnt signaling, as evidenced by constant activity of Wnt signaling reporter and expression of Wnt target genes (Huang and Schier, 2009; Ocbina et al., 2009). Interestingly, however, deregulated canonical Wnt signaling via deletion of adenomatous polyposis coli (APC) was found to cause a loss of primary cilia in association with radial progenitor malformation in the neocortex (Nakagawa et al., 2017). In these regards, the interaction between primary cilia and Wnt signaling may be context-dependent, and more thorough studies will be necessary to elucidate it.

## MTOR Signaling

MTOR is a serine/threonine kinase essential for protein translation, lipid synthesis, and autophagy, and plays an essential role in neural differentiation, neuronal migration, and synaptic formation, all of which are crucial to brain development. Disruption of MTOR signaling has been documented in numerous pathological conditions, including cancer, neurological disorder, and metabolic disorder (Lipton and Sahin, 2014; Saxton and Sabatini, 2017) Polycystic kidney disease (PKD) patients with inherited mutations in ciliary genes show deregulation of the MTOR signaling pathway (Shillingford et al., 2006; Wahl et al., 2006). In particular, polycystin-1, which is recurrently mutated in PKD, has been shown to inhibit MTOR through interaction with TSC2, a component of the TSC complex that inhibits MTOR (Shillingford et al., 2006). Studies have also identified that, by bending stimulus via fluid flow, primary cilia downregulate AMPK-MTOR signaling, which, in turn, induces autophagy to control cell size via LKB1 localized at primary cilia in the kidneys (Boehlke et al., 2010; Orhon et al., 2016). Recently, research has demonstrated that, during brain development, mice with defective cilia present abnormal increases in MTOR signaling, leading to enlarged apical domains of radial glial cells (RGCs) and subsequent dilatation of brain ventricles (Foerster et al., 2017). Conversely, in TSC1- or TSC2-null fibroblasts, TSC1 and TSC2 have been found to positively

regulate ciliogenesis without using the TSC-MTOR signaling axis (i.e., rapamycin-insensitive) (Hartman et al., 2009). Nevertheless, a recent independent study which also performed with TSC1- or TSC2-null fibroblasts reported different ciliary phenotypes marked by a longer ciliary length in TSC1-null cells and a shorter ciliary length in TSC2-null cells, and there was a discrepancy between the two cell types in that only the elongated ciliary phenotype in TSC1-null cells could be rescued by rapamycin treatment (Rosengren et al., 2018). In focal malformations of cortical developments (FMCDs), such as hemimegalencephaly (HME) and focal cortical dysplasia (FCD), which are highly associated with intractable epilepsy and autism-spectrum disorders, brain somatic activating mutations in MTOR eliciting blockage of autophagy have been described as disrupting neuronal ciliogenesis in brain tissues from FMCD patients (Wegiel et al., 2010; Lim and Crino, 2013; Park et al., 2018; **Figures 2A,B**). Given the complex reciprocal interaction between primary cilia and MTOR signaling, further mechanism studies are yet needed.

### Autophagy

Autophagy is a cellular degradative process by which a cell gains nutrients to maintain homeostasis (Mizushima, 2007). Two independent studies have provided evidence of links between ciliogenesis and autophagy: In serum-nutrient conditions, basal autophagy degrades the ciliary protein IFT20, while OFD1 protein at centriolar satellites inhibits ciliogenesis. Upon serumwithdrawal, the centriolar satellite pool of OFD1 is degraded by inducible autophagy, while IFT20 initiates ciliary trafficking, leading to ciliogenesis (Pampliega et al., 2013; Tang et al., 2013). Cilia-mediated Shh signaling has been found to assemble several proteins needed for autophagy at the periciliary region, thus activating Shh signaling induces autophagy flux. Interestingly, one study described defects in serum-withdrawal-induced autophagy and biogenesis of autophagosome among Ift20- or Ift88-compromised cells (Pampliega et al., 2013). In line with this, others have found that treatment of Shh to culturedhippocampal neurons upregulates several autophagy-related genes and enhances autophagy (Petralia et al., 2013).

Meanwhile, dysregulated ciliogenesis in relation to defective autophagy has been described in several diseases. Previous studies have reported suppressed autophagic flux in ciliary dysfunctional PKD mouse models and impairment of autophagosome formation in cells derived from PKD patients (Belibi et al., 2011; Zhu et al., 2017). Interestingly, Hürthle cells found in thyroid cancer were found to show defective ciliogenesis, and this defect in Hürthle cells eliciting high basal autophagic flux was restored by autophagy inhibition (Lee et al., 2016). Dysregulated ciliogenesis has also been reported in Huntington's disease (HD), for which defective autophagy has been well recognized (Keryer et al., 2011; Kaliszewski et al., 2015). Although Huntingtin, which is mutated in HD, is known to regulate autophagy, it also participates in trafficking pericentriolar material 1 (PCM1) to the centrosome. When Huntingtin is mutated, aberrant accumulation of PCM1 occurs, which, in turn, causes dysregulated ciliogenesis (Keryer et al., 2011; Rui et al., 2015). Further studies, however, are required to

elucidate the direct interaction between ciliary abnormality and defective autophagy in HD. Finally, as stated in the previous section on MTOR signaling, defective autophagy and resultant disruption of ciliogenesis have been demonstrated in FMCDs (Park et al., 2018). Brain somatic activating mutations in MTOR, which are causative for FMCDs, block autophagy, resulting in defective neuronal ciliogenesis due to aberrant accumulation of OFD1 (**Figure 2C**; Park et al., 2018). Although defective autophagy has been implicated in many neurodevelopmental and neurodegenerative disorders, much more studies are needed to elucidate the pathophysiological role of primary cilia in these disorders (Levine and Kroemer, 2008; Lee et al., 2013; Marsh and Dragich, 2018).

# DEVELOPMENTAL FUNCTIONS OF NEURONAL CILIA

# Patterning and Morphogenesis of the Forebrain

One of the well-established developmental functions of primary cilia is to control forebrain patterning, which is severely defected in human ciliopathies (Vogel et al., 2012). Patterning in the neural tube of the CNS occurs through progressive subdivisions along the dorsal to ventral and the rostral to caudal axes (Dessaud et al., 2008). Cobblestone mice (Ift88cbs/cbs), which show severe disorganization of the telencephalon, have been found to exhibit an ambiguous dorsal-ventral forebrain boundary and dorsal telencephalic-diencephalic boundary, concomitant with increased levels of full-length Gli3 (Willaredt et al., 2008). Meanwhile, mice lacking primary cilia as a result of mutation in the ciliopathy gene Rpgrip1l exhibited defective phenotypes associated with MKS and JBTS. In these mice, a mislocalized olfactory bulb-like structure and ambiguous dorsalventral patterning were observed, both of which were rescued upon introduction of constitutively active repressor form of Gli3 (Besse et al., 2011). Other mice with defective cilia through loss of Ttc21b (encoding retrograde Ift139) have also been shown to display defects in dorsal–ventral patterning and rostral– caudal patterning. Researchers have reported that Ttc21b-null mice exhibit normal anterograde transport that permits entry of Shh signaling components into cilium but partially defective retrograde transport. Overactivation of GLI2 and GLI3A in these mice could accounts for activation of Shh signaling and abnormal patterning, suggesting a periciliary role for retrograde IFT in GLI3 processing and GLI2 activity (Huangfu and Anderson, 2005; May et al., 2005; Tran et al., 2008; Stottmann et al., 2009). Crossing Ttc21b-null mice with mice heterozygous for a null allele of Shh to reduce levels of Shh ligand was found to partially rescue their defective phenotypes (Tran et al., 2008; Stottmann et al., 2009). This paradoxical effect of retrograde IFT mutants on Shh signaling (i.e., activation of Shh signaling) was well reviewed by Pal and Mukhopadhyay (2015) and Bangs and Anderson (2017). In line with the importance of primary cilia in cortical patterning, mice with mutant Slb (encoding anterograde Ift172) that possess very short axonemes with no visible microtubules

in their cilia have also been found to exhibit a global brainpatterning defect along the dorsal–ventral and rostral–caudal axes (Gorivodsky et al., 2009). Collectively, these studies suggest the importance of Gli3 processing, which could be mediated by primary cilia, in forebrain patterning.

The establishment of polarized radial glial scaffold formation during cortical development has been shown to be regulated by Arl13b, a small GTPase enriched in primary cilia (Higginbotham et al., 2013). Arl13b mutant mice (Arl13bhnn/hnn), which display defective cilia, surprisingly exhibit dramatic inversion of the apicobasal polarity of the radial glial scaffold, which, in turn, changes the location of the progenitor zone to the outer part of the cortex. This defect is also observed when Arl13b is removed just before the formation of radial glia from neuroepithelial cells using Foxg1-driven Cre. However, Arl13b removal after establishment of the radial glial scaffold using Nestin- or GFAP-driven Cre has not been found to generate this intriguing phenotype, suggesting that the role of Arl13b in polarization is temporally restricted to the period during which the radial glial scaffold develops (Higginbotham et al., 2013). This reversed apicobasal polarity seen in Arl13b mutant mice has not been reported in mice deficient of other ciliary genes. In line with this temporally restricted role of a ciliary gene during development, a recent study performed conditional knockout (KO) of Kif3a, Ift88, and Ttc21b in a series of specific spatiotemporal domains. Intriguingly, the observed neurological defects, including forebrain expansion, cortical malformations, impaired olfactory bulb development, and ventriculomegaly, were significantly different across four different Cre transgenic alleles (Foxg1-Cre, Emx1-Cre, Wnt1- Cre, and AP2-Cre), indicating that the roles of cilia in cortical development are discretely spatiotemporal. For example, the deletion of Ift88 at earlier stages of development, driven by Foxg1-Cre, was found to increase brain size; however, when deleted at later stages, driven by Emx1-Cre, there were no significant differences in brain size (Snedeker et al., 2017). Taken together, these experimental studies of perturbations of primary cilia in mice support the crucial role of primary cilia in forebrain development, which, when disrupted, lead to severe morphological malformations in the brain.

#### Neuronal Migration

fncel-13-00218 May 11, 2019 Time: 14:10 # 6

During brain cortical development, newly born neurons follow one of two main migratory routes to reach their final destination, known as radial and tangential migration (Hatten, 1999). The role of primary cilia in radially migrating pyramidal neurons remains unclear. In a previous study, cerebral cortical lamination was not apparently affected in Arl13b-conditional KO mice driven by Nex-Cre or in Stumpy-, Kif3a-, and Ift88-conditional KO mice driven by Nestin-Cre (Arellano et al., 2012; Higginbotham et al., 2012; Tong et al., 2014). However, a recent in vivo systemic study revealed that genes linked to ciliopathies affect cortical development, including neuronal radial migration, neural progenitor development, neuronal differentiation, and early neuronal connectivity (Guo et al., 2015). In this study, knockdown of 17 ciliopathy genes in the developing neocortex resulted in delayed neuronal radial migration in association with transient multipolar stage, multipolar-to-bipolar transition, and glial-guided radial migration (Guo et al., 2015). Meanwhile, others have suggested that the defective phenotypes seen in ciliary conditional KO mice driven by Nestin-Cre, the expression of which is broad and is also seen in non-RGCs at later developmental stages, may be partly attributable to non-cellautonomous effects (Foerster et al., 2017; Chen et al., 2018). Accordingly, the cell-autonomous function of primary cilia in neural stem cells at the late developmental stage was investigated via acute knockdown of ciliary genes (Kif3a or Ift88) in the developing cortex, which resulted in defective neuronal differentiation and migration; this was accompanied by delays in neural stem cell cycle progression and failures in interkinetic nuclear migration (Chen et al., 2018). In FMCDs, the molecular mechanism of how somatic mutations in MTOR lead to focal cortical dyslamination appears to be related to defective neuronal ciliogenesis by MTOR mutation, which affects the multipolar to bipolar transition essential for neuronal radial migration (**Figures 2C,D**; Park et al., 2018).

Interneurons originating from the medial ganglionic eminence migrate along a tangential path in the cortical plate, either at the marginal zone or intermediate zone. These neurons change their orientation from a tangential to radial path to colonize the cortical plate and to differentiate into GABAergic neurons (Guo and Anton, 2014). Recent studies have demonstrated that disruption of primary cilia at interneurons affects the interneuronal migratory process, suggesting the importance of primary cilia in neuronal migration (Baudoin et al., 2012; Higginbotham et al., 2012). In a previous study, mice with conditional KO of Arl13b genes driven by Dlx5/6-Cre showed disruption of interneuron placement at Arl13b-deleted cortices caused by defects in interneuron migration and branching. However, when the Arl13b gene was disrupted in post-migratory interneurons driven by PV-Cre, there was no significant difference in the morphology of dendritic arbors during neuronal differentiation. Such defective interneuron migration may be a potential pathophysiological mechanism underlying neurological defects in JBTS with Arl13b mutations (Higginbotham et al., 2012). Meanwhile, a recent study showed that disruption of cilia in interneurons using Nkx2.1-Cre to ablate Ift88 or Kif3a genes led to defective migration of interneurons, marked by an aberrant accumulation of interneurons at their first tangential path, not radial streams in the cortical plate, as well as abnormal positioning and density of interneurons in the postnatal cortex (Baudoin et al., 2012). Taken together, these studies emphasize that primary cilia in immature interneurons are crucial to interneuron migration in the developing cortex and the underlying mechanisms thereof (Baudoin et al., 2012). Nevertheless, the role of primary cilia in neuronal migration needs further study, especially in regards to which other ciliamediated signaling pathways are involved. These future studies on neuronal migration should shed light on the pathophysiology of ciliopathies presenting with cognitive deficits, which may be attributable to defective neuronal migration and disrupted neural connectivity.

#### Cerebellum Development

Ataxia due to cerebellar malformation is one major symptom of ciliopathies, particularly JBTS. During development of the cerebellum, Purkinje cells secrete Shh, which regulates the proliferation of cerebellar neuronal precursor (Wechsler-Reya and Scott, 1999; Haldipur et al., 2012): Shh signaling mediated by primary cilia occurs in late embryonic stages of cerebellum development, during which Shh is required for expansion of the granule neuron precursor population, but not for the subsequent differentiation of these cells (Lewis et al., 2004). Primary cilia control cerebellar morphogenesis by promoting the expansion of the granule progenitor pool. Loss of either Ift88 or Kif3a causes severe cerebellar hypoplasia and inhibits the expansion of the granule cell progenitor population. Although Ift88 and Kif3a are not required for the specification and differentiation of cerebellar cell types, they affect granular cell progenitor proliferation. Researches have shown that the expression of Gli1 at the external granule cell layer in Kif3a mutant mice is lower than that in controls (Chizhikov et al., 2007; Spassky et al., 2008). These studies demonstrate that ablation of primary cilia impairs Shh signaling, which regulates the development of the cerebellum, particularly expansion of cerebellar progenitors. This mechanism could explain the hypoplasia of the cerebellum and cerebellar ataxia seen in patients with ciliopathies such as JBTS.

#### Learning and Memory

Patients with ciliopathies commonly present with intellectual disability (ID) as a neurological deficit. Although there are not enough studies to explain the mechanism underlying ID as a result of defects in primary cilia, several studies have shown that depletion of primary cilia leads to hippocampal-dependent learning and memory deficits (Breunig et al., 2008; Han et al., 2008; Amador-Arjona et al., 2011). Hippocampal neurogenesis plays a role in fear conditioning, recognition, spatial memory, and pattern separation (Cameron and Glover, 2015). Also, Shh

signaling mediated by primary cilia controls the proliferation of neural progenitor cells (NPCs). As evidence thereof, Emx1-Credependent Shh-conditional KO mice were found to show smaller dorsal telencephalons at E18.5 and decreased proliferation and increased apoptosis of neural stem cells (NSCs) and NPCs in the dorsal pallium (Komada et al., 2008). Meanwhile, Nestin-Cre- and hGFAP-Cre-dependent Smo KO mice exhibited smaller postnatal dentate gyruses and fewer proliferating cells than those in controls (Machold et al., 2003; Han et al., 2008). In Kif3a mutant and Ift88 mutant mice, neuronal proliferation in the subgranular zone (SGZ) and subventricular zone (SVZ) was decreased, and Gli1 expression was abolished (Han et al., 2008; Tong et al., 2014; Gazea et al., 2016) These demonstrate that Shh signaling is involved in the proliferation and survival of NSCs and NPCs. Hypomorphic Ift88-mutant mice, in which Ift88 mRNA and protein are reduced by 70–80%, showed disorganization of the midbrain dopominergic neuron (mDA) progenitors domain, and the number of mDA neurons was severly reduced (Gazea et al., 2016). Considering that primary cilia regulate neurogenesis by mediating Shh signaling, defects in learning and memory and ID might be caused by decreased neurogenesis due to aberrant Shh signaling. Expanding on the above research, additional studies are needed to determine how defective neuronal cilia lead to learning and memory impairment and ID.

#### CONCLUDING REMARKS

In this review, we briefly summarized the interaction between primary cilia and several signaling pathways, including Shh, Wnt, MTOR, and autophagy, all of which are essential to cellular homeostasis. Given that neurological disorders are typical symptoms of ciliopathies, the importance of primary cilia during brain development cannot be emphasized enough. At present, however, we are unable to fully understand the underlying mechanisms by which defective neuronal cilia contribute the pathogenesis of neurological defects in ciliopathies. Future studies of neuronal cilia in relation to signaling pathways may provide better understanding of the functions of neuronal cilia and may help to develop novel treatments for ciliopathies by targeting neuronal cilia-mediated signaling pathways. Compared

#### REFERENCES


with considerable studies on neuronal cilia, there is a lack of research on the role of primary cilia in glial cells, another major component of the brain. Glial cells, such astrocytes and oligodendrocytes, also possess a single primary cilium (Berbari et al., 2007; Bishop et al., 2007; Kasahara et al., 2014). Recently, studies reported defective ciliogenesis in glioblastoma (GBM), a high-grade astrocytic malignancy, and demonstrated the potential role of primary cilia on GBM progression (Moser et al., 2014; Hoang-Minh et al., 2016a,b; Loskutov et al., 2018). Further studies on the role of primary cilia in glial cells, as well as neurons, will advance our understanding of the role of primary cilia in brain development.

To help achieve a better understanding of human ciliopathies resulting from mutations in ciliary genes, CRISPR/Cas9 genome editing of identified mutations in human ciliopathies or 3D culture of brain organoids derived from pluripotent stem cells from ciliopathy patients can be used to generate more pathophysiologically relevant models for ciliopathy, thereby overcoming the gap between animal models and human subjects. Also, increasing studies on interactions between primary cilia and other signaling pathways may uncover a previously unknown role for cilia in the pathogenesis of human neurological diseases that are not considered classical ciliopathies, such as FMCD, brain cancer (medulloblastoma and GBM), and neurodegenerative diseases (Alzheimer's disease and HD) (Sotthibundhu et al., 2009; Keryer et al., 2011; Eguether and Hahne, 2018). Future studies of these disorders may help us to further the molecular basis of human diseases characterized by defects in primary cilia.

#### AUTHOR CONTRIBUTIONS

SP, HJ, and JL conceived the idea and wrote the manuscript.

# FUNDING

This work was supported by grants from the Suh Kyungbae Foundation (to JL); Korean Health Technology R&D Project, Ministry of Health and Welfare, Republic of Korea (H15C3143 and H16C0415 to JL).

mouse neocortex. J. Compar. Neurol. 520, 848–873. doi: 10.1002/cne. 22793





in a novel mouse model of Meckel-Gruber syndrome. Dev. Biol. 377, 55–66. doi: 10.1016/j.ydbio.2013.02.015


**Conflict of Interest Statement:** JL is a co-founder of SoVarGen, Inc. that develops new diagnostics and therapeutics for brain disorders.

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 Park, Jang and Lee. 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.

# Anomalies in Dopamine Transporter Expression and Primary Cilium Distribution in the Dorsal Striatum of a Mouse Model of Niemann-Pick C1 Disease

Micaela Lucarelli1,2, Chiara Di Pietro<sup>3</sup> , Gina La Sala<sup>3</sup> , Maria Teresa Fiorenza<sup>1</sup> , Daniela Marazziti<sup>3</sup> and Sonia Canterini<sup>1</sup> \*

<sup>1</sup> Division of Neuroscience, Department of Psychology, Center for Research in Neurobiology 'Daniel Bovet', Sapienza University of Rome, Rome, Italy, <sup>2</sup> PhD Program in Behavioral Neuroscience, Sapienza University of Rome, Rome, Italy, 3 Institute of Cell Biology and Neurobiology, Italian National Research Council, Rome, Italy

#### Edited by:

Rosanna Parlato, Ulm University, Germany

#### Reviewed by:

Martin Witt, Rostock University Hospital, Germany Bronwyn Maree Kivell, Victoria University of Wellington, New Zealand Andreas Wree, University of Rostock, Germany

> \*Correspondence: Sonia Canterini sonia.canterini@uniroma1.it

#### Specialty section:

This article was submitted to Cellular Neurophysiology, a section of the journal Frontiers in Cellular Neuroscience

> Received: 23 January 2019 Accepted: 06 May 2019 Published: 24 May 2019

#### Citation:

Lucarelli M, Di Pietro C, La Sala G, Fiorenza MT, Marazziti D and Canterini S (2019) Anomalies in Dopamine Transporter Expression and Primary Cilium Distribution in the Dorsal Striatum of a Mouse Model of Niemann-Pick C1 Disease. Front. Cell. Neurosci. 13:226. doi: 10.3389/fncel.2019.00226 The Niemann-Pick type C1 (NPC1) is a rare genetic disease characterized by the accumulation of endocytosed cholesterol and other lipids in the endosome/lysosome compartments. In the brain, the accumulation/mislocalization of unesterified cholesterol, gangliosides and sphingolipids is responsible for the appearance of neuropathological hallmarks, and progressive neurological decline in patients. The imbalance of unesterified cholesterol and other lipids, including GM2 and GM3 gangliosides, alters a number of signaling mechanisms impacting on the overall homeostasis of neurons. In particular, lipid depletion experiments have shown that lipid rafts regulate the cell surface expression of dopamine transporter (DAT) and modulate its activity. Dysregulated dopamine transporter's function results in imbalanced dopamine levels at synapses and severely affects dopamine-induced locomotor responses and dopamine receptor-mediated synaptic signaling. Recent studies begin to correlate dopaminergic stimulation with the length and function of the primary cilium, a non-motile organelle that coordinates numerous signaling pathways. In particular, the absence of dopaminergic D2 receptor stimulation induces the elongation of dorso-striatal neuron's primary cilia. This study has used a mouse model of the NPC1 disease to correlate cholesterol dyshomeostasis with dorso-striatal anomalies in terms of DAT expression and primary cilium (PC) length and morphology. We found that juvenile Npc1nmf164 mice display a reduction of dorso-striatal DAT expression, with associated alterations of PC number, length-frequency distribution, and tortuosity.

Keywords: Niemann-Pick C1, mouse model, striatum, primary cilium, dopamine

# INTRODUCTION

Niemann-Pick type C1 (NPC1) is a rare lysosomal lipid storage disorder caused by mutations in the NPC1 gene, whose protein mediates the egress of cholesterol from lysosomes/endosomes (Peake and Vance, 2010). NPC1 patients develop severe neurological-neurovisceral disorders, including cerebellar ataxia, dysarthria, dysphagia, seizures, and progressive dementia (Vanier, 2010, 2013).

Niemann-Pick type C1 cells display defective synthesis and mobilization of endocytosed cholesterol to the plasma membrane, which affects the functions of neurotransmitters, and their receptors (Fiorenza et al., 2013, 2018).

Dopamine transporter (DAT) regulates the spatio/temporal dynamics of dopamine (DA) neurotransmission by regulating the reuptake of extracellular DA into presynaptic terminals. DAT localization and function are directly regulated by cholesterol (Jones et al., 2012). For instance, cholesterol depletion studies have shown that lipid rafts regulate DAT cell-surface expression (Foster et al., 2008; Gabriel et al., 2013), and modulate its activity (Adkins et al., 2007). In addition, cholesterol has been shown to stabilize DAT conformation and DA binding (Hong and Amara, 2010). A number of G protein-coupled receptors including dopaminergic D2-receptors (D2R) are localized to the primary cilium (PC) of mammalian neurons and the lack of D2 dopaminergic input increases striatal PC length (Marley and von Zastrow, 2010; Miyoshi et al., 2014).

We have recently reported an impairment of Sonic hedgehog (Shh) signaling and PC density and length, in hippocampal neurons of Npc1−/<sup>−</sup> mice and fibroblasts of NPC1 patients (Canterini et al., 2017). PC is a non-motile organelle that plays critical roles in coordinating numerous neuronal/developmental signaling pathways. Alterations of PC morphology and localization are responsible for ciliopathies, disorders that manifest, like the NPC1 disease, a constellation of clinical features including ataxia, retinal degeneration, behavioral disturbance, and intellectual disability (Waters and Beales, 2011; Guo et al., 2015).

In this study we investigate DA signaling and reception in the striatum of mouse of the Npc1nmf 164 strain, that bears a point mutation in the Npc1 gene (D1005G), resulting in a milder, and late-onset form comparable to the most part of human cases (Maue et al., 2012). In this study, we show that juvenile Npc1nmf 164 mice display a reduction of DAT expression and alteration of PC number and length-frequency distribution in the dorsal striatum.

Our findings identify early and subtle anomalies in striatal dopaminergic neurotransmission that might contribute to the subsequent appearance of NPC1 disease manifestations.

#### MATERIALS AND METHODS

#### Animals and Treatments

Homozygous Npc1nmf 164 mice maintained on BALB/cJ background were derived from heterozygous matings. Genotypes were identified by PCR analysis of tail DNA (Palladino et al., 2015).

Animal experimental protocols and related procedures were approved by the Italian Ministry of Health-General Directorate of Animal Health (995/2016; D.Igs. 26/2014). All efforts were made to minimize animal suffering, according to European Directive 2010/63/EU.

#### Tissue Dissection

Brains of postnatal (PN) day 30 Npc1nmf 164 mice and wild-type (wt) littermates (5 mice/genotype) (**Supplementary Table S1**) were collected on ice-cold PBS and cut along the mid-sagittal plane. For Western Blotting, punches of the dorsal striatum (DS) were obtained from coronal slices of one hemisphere using a steel needle (1.5 mm diameter) (Colelli et al., 2010; Campus et al., 2017). The other hemisphere was fixed overnight in 4% paraformaldehyde and cut on Leica-Vibratome (S1000, Leica), for immunohistochemistry (IHC).

#### Western Blot Analysis

Tissue punches of DS were processed for protein extraction and Western blot analysis as previously described (Di Pietro et al., 2017). Primary and secondary antibodies used are listed in **Supplementary Table S2**. Band intensity was normalized to α-tubulin signal. The average values were expressed in arbitrary units, as a ratio to wt mean values.

#### IHC of Primary Cilium Markers and Measurements

Double IHC staining on free-floating sections (30 µm) was performed as previously described (Canterini et al., 2012). Primary and secondary antibodies data are listed in **Supplementary Table S2**.

Only striatal PC "clearly" double-stained for γ-tubulin (basalbody) and ACIII (PC-shaft) were selected for morphological analysis using Neurolucida analysis system (MBF Bioscience, Williston, VT, United States), connected to Olympus BX53 microscope (100X/1.25 numerical aperture) with 40X/100X immersion objective lens. An average of 6 PC was measured from 10 random fields per mice (n = 318 PC/genotype).

#### Statistics

A Mann-Whitney U-test was used to determine the difference in protein levels (GraphPad Software, Inc). The D'Agostino & Pearson omnibus normality test was used to assess the distribution of values. Differences in PC distribution were determined by Chi-squared test, whereas PC lengths and related parameters were analyzed by non-parametric Mann-Withney U-test and Spearman's correlation analysis. P-values < 0.05 were considered significant.

# RESULTS

#### Reduction of Striatal DAT Expression Without Alteration of D2R and TH Levels in Npc1nmf164 Mice

It is know the functional relation between DAT and TH expression in the striatum (Salvatore et al., 2016) and the physical association between D2R and DAT proteins at striatal presynaptic terminals (Lee et al., 2007). To examine whether the dysregulation of cholesterol homeostasis in Npc1-mutant mice affects dopaminergic signaling in DS, Western blot analysis for the mature form of DAT (mDAT), D2R, TH was performed on protein extracted from tissue punches of PN30 Npc1nmf 164 mice, and wt littermates (**Figure 1A**). Only striatal mDAT levels were found significantly reduced in Npc1nmf 164 mice compared to control littermates (P < 0.01; **Figure 1B**).

# 3D Analysis of Striatal PC: Novel Measurement Method Displays Comparable Average Length of PC but a Different Length Frequency Distribution

Mann-Withney U-test. Arrowhead indicates likely immature form of DAT.

Npc1nmf 164 mice show a reduction of mDAT expression that is likely associated to increased dopaminergic stimulation, which controls PC length through the cAMP pathway (Neve et al., 2004; Ou et al., 2009). To study possible variations of PC morphology in striatal sections from Npc1nmf 164 and wt mice, we performed a double IHC labeling with antibodies against γ-tubulin (as basal body marker) and ACIII (as neuronal ciliary shaft marker) (**Figure 2A**) coupled to the Neurolucida acquisition system. The latter allows the application to PC analysis of standard tools used for neuron tracing, as simultaneous and precise 3D measurements (±0.5 µm) of length, diameter, and tortuosity.

The experimental results revealed no significant difference between the two genotypes in the mean length of neuronal PC (**Figure 2B**), as well as in other parameters, including ciliary area, volume, tortuosity, and diameter (**Figure 2C**). In Npc1nmf 164 mice, however, significant differences were observed in the number of ACIII-positive PC (P = 0.03) and distribution of PC lengths (P = 0.02), with an increment of PC with very short (2 µm), or very long length (17 µm and over), compared to control mice (**Figure 2D**).

Finally, Spearman's correlation analysis was used to investigate the relationships between three major PC morphological parameters: length, diameter, and tortuosity. The analysis between ACIII-positive PC length and tortuosity, defined as greater bends or kinks in the ciliary axoneme, surprisingly demonstrated a perfect Spearmann's correlation in Npc1nmf 164 mice, in contrast to wt mice. In mutant mice, in fact, the tortuosity progressively rises with increasing length, whereas in

wt mice a considerable tortuosity variability is observed both in long and short cilia. Concerning length-diameter and diametertortuosity relationships, similar positive Spearman coefficients were found in both wt and mutant samples (**Figure 2E**).

#### DISCUSSION

The prominent feature of the NPC1 disease is a distinctive progressive neurodegeneration, with cerebellum and Purkinje neurons being particularly vulnerable (Higashi et al., 1993; Nusca et al., 2014). Some clinical features, however, overlap between lysosomal storage disorders as NPC1 and Parkinson's disease, suggesting that the two disorders may be pathogenically linked (Storch et al., 2004; Deng et al., 2015).

To characterize the likely contribution of striatal component in the etiology of NPC1 disease, we analyzed the expression of effectors of DA signaling, including mDAT, D2R, and TH by Western Blot analysis of striatal samples from Npc1nmf 164 mice at a juvenile, asymptomatic age in comparison to wt littermates.

The reduction of mDAT protein levels in PN30 Npc1nmf 164 mice is in agreement with the marked symmetrical loss of striatal DAT, especially in the putamen, observed in NPC1 patients by DAT-scan analysis (Terbeek et al., 2017; Tomic, 2018).

The absence of D2R-mediated stimulation increases cAMP level, which in turn leads to neuronal PC elongation (Besschetnova et al., 2010; Miyoshi et al., 2014). It is also known that PC length and density exhibit brain regionspecific changes (Sipos et al., 2018) and ciliary D1-receptor translocates to and from cilia in response to environmental cues (Domire et al., 2011). In addition, DAT, TH, and D2R proteins colocalize in nigrostriatal terminals and their expression levels are often affected in neurological/ neurodegenerative disorders. The distribution of D1R and D2R varies along the rostro-caudal axis of the DS (Gangarossa et al., 2013), whereas DAT and D2R directly interact to facilitate the recruitment of DAT to the plasma membrane (Lee et al., 2007).

We have recently reported that there is a reduction of PC density and length in hippocampal neurons of Npc1−/<sup>−</sup> mice as well as in fibroblasts of NPC1 patients, with associated dysregulation of expression/subcellular localization of Shh pathway components (Canterini et al., 2017). As no previous information was available on Npc1 deficiency-dependent morphological changes of striatal PC, we performed a 3D analysis of ciliary images for understanding structural determinants of normal and pathological PC function.

The remarkable length of PC of striatal neurons of either wt or Npc1nmf 164 mice is in agreement with a previous study that reported the presence in the striatum of a large number of long ACIII-positive PC (Bishop et al., 2007). The absence of statistically significant differences between the two genotypes in the average values of ACIII-, γ tubulin-positive PC length, and related parameters indicates that the mild alteration of DAT expression that we found does not lead to a structural remodeling of dorsal striatal PC in mutant mice. However, a more detailed analysis showed that Npc1nmf 164 mice display an increased number of ACIII-, γ-tubulin-positive PC and a different distribution of their lengths, together with increased tortuosity in a length-dependent manner, suggesting anomalies of ciliary functions.

The wider range of lengths and the positive correlation between length and tortuosity observed in Npc1nmf 164 suggest that mutant cilia are "unstable." Such instability possibly reflects a mis-regulation of axonemal length. Mutant cilia could undergo excessive elongation and fragmentation that would explain the increment of either very short or very long PC, which is observed in mutant mice. Similar ciliary instability was reported in Kdm3a mutants (Yeyati et al., 2017).

The reason of regional difference of PC expression in NPC1 disease is still unclear. It could be attributable to multiple factors such as regional changes in dopaminergic signaling or projections, spatial regulation of Shh released from dendrites and axons of dopaminergic neurons or to differences in intracellular cAMP levels that positively regulate the length of PC through the modulation of protein kinase A activity.

In conclusion, our findings identify for the first time subtle changes occurring in the striatum of juvenile asymptomatic Npc1nmf 164 mice that could contribute to NPC1 disease neurological manifestations. This is in agreement with: (i) our previous studies that demonstrated early developmental defects which occur postnatally in the cerebellum of Npc1-deficient mice and largely anticipate motor deficits, typically observed during adulthood (Nusca et al., 2014; Caporali et al., 2016); (ii) reported embryonic abnormalities in the metabolism of cholesterol in striatal neurons of Npc1-deficient mice (Henderson et al., 2000); (iii) DAT KO-mice display ataxic symptoms, tremors, dystonia and saccade-failure (Cyr et al., 2003), typical of related-dopamine-transporter-deficiency syndrome (Ng et al., 2014) and late-onset NPC1 disease (Vanier, 2010).

Although the anomalies in DAT expression and PC of Pn30 Npc1 mice we report in this study appear mild, we expect later stages of the disease to be landmarked by more robust alterations, with a decrement of DAT expression and PC length, as consequence of a progressive worsening of the perturbations of plasma membrane lipid content (Peake and Vance, 2010).

# ETHICS STATEMENT

Animal experimental protocols and related procedures were approved by the Italian Ministry of Health-General Directorate of Animal Health (995/2016; D.Igs. 26/2014). All efforts were made to minimize animal suffering, according to European Directive 2010/63/EU.

# AUTHOR CONTRIBUTIONS

ML performed IHC, image acquisition, and statistical dataanalysis. CDP and GLS performed Western Blot and statistical data-analysis. MF contributed with advise and discussion. SC conceived the project. SC and DM directed the project, prepared the figures, and wrote the manuscript.

# FUNDING

This study was supported by FFABR to SC and Ateneo La Sapienza (RM11615501ED6577) to MF.

# ACKNOWLEDGMENTS

We are grateful to Raffaele Matteoni for critically reading the manuscript.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fncel. 2019.00226/full#supplementary-material

# REFERENCES

fncel-13-00226 May 23, 2019 Time: 15:21 # 6


C. Parkinsonism. Relat. Disord. 36, 107–108. doi: 10.1016/j.parkreldis.2016. 12.029


to regulate cilia stability. J. Cell Biol. 216, 999–1013. doi: 10.1083/jcb.20160 7032

**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 Lucarelli, Di Pietro, La Sala, Fiorenza, Marazziti and Canterini. 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.

# Targeted Depletion of Primary Cilia in Dopaminoceptive Neurons in a Preclinical Mouse Model of Huntington's Disease

Rasem Mustafa1,2 , Grzegorz Kreiner <sup>3</sup> , Katarzyna Kaminska ´ 4,5 , Amelia-Elise J. Wood<sup>6</sup> , Joachim Kirsch<sup>2</sup> , Kerry L. Tucker <sup>6</sup> and Rosanna Parlato1,2 \*

1 Institute of Applied Physiology, University of Ulm, Ulm, Germany, <sup>2</sup> Institute of Anatomy and Cell Biology, Medical Cell Biology, University of Heidelberg, Heidelberg, Germany, <sup>3</sup>Department of Brain Biochemistry, Maj Institute of Pharmacology, Polish Academy of Sciences, Kraków, Poland, <sup>4</sup>Department of Pharmacology, Maj Institute of Pharmacology, Polish Academy of Sciences, Kraków, Poland, <sup>5</sup>Jagiellonian Center for Experimental Therapeutics, Jagiellonian University, Kraków, Poland, <sup>6</sup>Department of Biomedical Sciences, Center for Excellence in the Neurosciences, College of Osteopathic Medicine, University of New England, Biddeford, ME, United States

Multiple pathomechanisms triggered by mutant Huntingtin (mHTT) underlie progressive degeneration of dopaminoceptive striatal neurons in Huntington's disease (HD). The primary cilium is a membrane compartment that functions as a hub for various pathways that are dysregulated in HD, for example, dopamine (DA) receptor transmission and the mechanistic target of rapamycin (mTOR) pathway. The roles of primary cilia (PC) for the maintenance of striatal neurons and in HD progression remain unknown. Here, we investigated PC defects in vulnerable striatal neurons in a progressive model of HD, the mHTT-expressing knock-in zQ175 mice. We found that PC length is affected in striatal but not in cortical neurons, in association with the accumulation of mHTT. To explore the role of PC, we generated conditional mutant mice lacking IFT88, a component of the anterograde intraflagellar transport-B complex lacking PC in dopaminoceptive neurons. This mutation preserved the expression of the dopamine 1 receptor (D1R), and the survival of striatal neurons, but resulted in a mild increase of DA metabolites in the striatum, suggesting an imbalance of ciliary DA receptor transmission. Conditional loss of PC in zQ175 mice did not trigger astrogliosis, however, mTOR signaling was more active and resulted in a more pronounced accumulation of nuclear inclusions containing mHTT. Further studies will be required of aged mice to determine the role of aberrant ciliary function in more advanced stages of HD.

#### Keywords: primary cilium, dopamine system, Huntington's disease, mTOR, p62

**Abbreviations:** ACIII, adenylate cyclase III; AD, Alzheimer's disease; DA, dopamine; DAPI, 4<sup>0</sup> ,60 -diamidino-2 pheylindol; DOPAC, 3,4-dihydroxyphenylacetic acid; D1R, dopaminoceptive D1-receptor; GFAP, glial fibrillary acidic protein; HD, Huntington's disease; HPLC-EC, High-performance liquid chromatography-electrochemical detection; 5-HT, 5-hydroxytryptamine; HVA, homovanillic acid; IF, immunofluorescence; IFT-B, intraflagellar transport B; IHC, immunohistochemistry; mHTT, mutant Huntingtin; MSNs, medium spiny neurons; mTOR, mechanistic target of rapamycin; NeuN, neuronal nuclei; NPC1, Niemann-Pick type C1; PC, primary cilia; PD, Parkinson's disease; PFA, paraformaldehyde; phospho-S6, phosphorylated ribosomal protein S6; p62/SQSTM1, p62/sequestosome 1; TH, tyrosine-hydroxylase.

#### Edited by:

Thomas Fath, Macquarie University, Australia

#### Reviewed by:

Jiro Kasahara, Tokushima University, Japan Erik B. Malarkey, Vertex Pharmaceuticals, United States

\*Correspondence: Rosanna Parlato rosanna.parlato@uni-ulm.de

Received: 10 July 2019 Accepted: 05 December 2019 Published: 20 December 2019

#### Citation:

Mustafa R, Kreiner G, Kaminska K, ´ Wood A-EJ, Kirsch J, Tucker KL and Parlato R (2019) Targeted Depletion of Primary Cilia in Dopaminoceptive Neurons in a Preclinical Mouse Model of Huntington's Disease. Front. Cell. Neurosci. 13:565. doi: 10.3389/fncel.2019.00565

# INTRODUCTION

Huntington's disease (HD) is an autosomal dominant progressive neurodegenerative disorder caused by the toxic expansion of CAG trinucleotide repeats at the N-terminus of the Huntingtin gene. The mechanisms underlying selective vulnerability of dopaminoceptive medium spiny neurons (MSNs), resulting in impaired control of voluntary movement in HD, remain elusive (Ghosh and Tabrizi, 2018). The variability of disease onset and progression depends on the CAG number, and on genetic modifiers interacting with the Huntingtin mutation [Genetic Modifiers of Huntington's Disease (GeM-HD) Consortium (2015)]. Multiple signaling pathways and cellular functions are affected by mutant Huntingtin (mHTT), including protein aggregate degradation, which results in the accumulation of toxic proteins (Saudou and Humbert, 2016).

Primary cilia (PC) are single, non-motile microtubule-based organelles resembling a cellular antenna that represents a hub for receptors and components of numerous signaling pathways (Malicki and Johnson, 2017). Lack of HTT results in reduced and aberrant PC growth, and increased mHTT results in increased ciliogenesis (Keryer et al., 2011). Notably, longer PC have been observed in immortalized cellular models of HD in culture, and ependymal cilia in the lateral ventricles are disorganized in a mouse model of HD and in HD human post-mortem brains (Keryer et al., 2011). Another study showed that photoreceptor cilia pathology accounts for their degeneration in the retina of R6/2 transgenic mice overexpressing exon 1 of the human mHTT (Karam et al., 2015). It has been previously proposed that PC altered structure might affect the function of signaling pathways whose components are localized in the PC (Maiuri et al., 2013; Kaliszewski et al., 2015).

Interestingly, increased PC length results in the induction of autophagy by inhibition of the mechanistic target of rapamycin (mTOR) kinase activity (Kaliszewski et al., 2015), and components essential for ciliogenesis are degraded by autophagy (Pampliega et al., 2013; Tang et al., 2013). Because autophagy is altered in HD (Ravikumar et al., 2004), as well as dopamine (DA)-mediated signaling (Chen et al., 2013), it is possible that HD pathophysiology depends, at least in part, on defective cilia. These previous studies investigated neither PC dysfunction in the most vulnerable striatal neurons, nor the impact of defective PC on HD pathogenesis in the striatum. A deeper understanding of the role of PC in mHTT-dependent neurotoxicity might help to identify new determinants modifying HD progression.

To this end, we monitored neuron- and stage-specific changes of PC structure in a full-length progressive mouse model of HD, called zQ175 (Menalled et al., 2012; Carty et al., 2015). This knock-in model carries a chimeric human/mouse HTT exon 1 containing expanded CAG repeats within the murine htt gene and recapitulates several hallmarks of HD pathology (Menalled et al., 2012; Farrar et al., 2014; Carty et al., 2015; Ma et al., 2015). Moreover, we generated a new genetic mouse model of defective ciliary function in striatal neurons, as a tool to investigate the specific impact of PC loss on striatal neuron maintenance and on HD neuropathological hallmarks.

#### MATERIALS AND METHODS

#### Mice

To generate a mutant mouse in which the Ift88 gene is conditionally ablated by the Cre-LoxP system in MSNs, we employed the B6.FVB/N-Tg(D1RCre)Gsc (D1R:Cre) transgene, which expresses the Cre recombinase under the control of the dopamine 1 receptor (D1R) promoter (Lemberger et al., 2007). The D1R:Cre mice were crossed to Ift88tm1.1Bky mice carrying the Ift88 floxed allele (Ift88flox/flox; Haycraft et al., 2007) to generate Ift88flox/flox; D1R:Cre mice (Ift88D1RCre; here abbreviated Ift88 cKO mice) that lack PC in MSNs. Htttmtm1Mfc/190tChdi (zQ175 knock-in) mice were received courtesy of the CHDI Foundation from the Jackson Laboratory. The analysis of the genotype was performed by PCR of tail snips as previously described (Levine et al., 1999). For the experiments reported here, male and female mice were used and wild-type and mutant littermates were analyzed. The zQ175 knock-in mice carry ca. 190 CAG repeats in a chimeric human/mouse exon 1 of the murine huntingtin gene (Menalled et al., 2012). The zQ175 mutation was kept in heterozygosity, to limit toxicity and mimic a genetic situation more relevant for the disease, as it is autosomal dominant (Menalled et al., 2012). These mutant mice were born at the expected Mendelian ratio; they showed normal lifespan and no gross abnormalities (monitored until 1-year-old).

For genotyping of D1R:Cre and floxed Ift88 alleles by PCR the following primer pairs was used: forward-primer/Cre (5<sup>0</sup> -GGA AAT GGT TTC CCG CAG AAC-3<sup>0</sup> ) and reverse-primer/Cre (50 -ACG GAA ATC CAT CGC TCG ACC-3<sup>0</sup> ), BY919 (5<sup>0</sup> - GGTCCTAACAAGTAAGCCCAGTGTT-3<sup>0</sup> ) and BY598 (5<sup>0</sup> - GCCTCCTGTTTCTTGACAACAGTG-3<sup>0</sup> ), respectively. For the zQ175 we used forward-primer/Neo: 5<sup>0</sup> -GAT CGG CCA TTG AAC AAG ATG– 3<sup>0</sup> and reverse-primer/Neo: 5<sup>0</sup> -AGA GCA GCC GAT TGT CTG TTG– 3<sup>0</sup> .

#### Brain Dissection and Tissue Preparation

For histological analysis, brains were either transcardially perfused or post-fixed in 4% paraformaldehyde (PFA; pH 7.2) overnight at 4◦C. After washing in PBS (pH 7.2) the brains were cryoprotected by incubating them in 10%, 20%, and 30% sucrose for 3 days at 4◦C. The brains were embedded in a coronal orientation (Tissue freezing medium, Leica), frozen by a mixture of liquid nitrogen and dry ice, and stored at −80◦C until sectioning (Leica CM3050S cryostat). For the analysis of adult brains, we have used coronal sections from striatum collected serially on Superfrost Ultraplus glass slides (12 µm) and free-floating striatum (30 µm). We analyzed the striatum in the region comprised between Bregma +1.18 mm and −0.34 mm based on the adult mouse brain Atlas (Franklin, 2008). For immunofluorescence on paraffin sections, one brain hemisphere was fixed in 4% PFA in PBS, pH 7.2 overnight at 4◦C and paraffin-embedded.

#### Immunofluorescence

Immunofluorescence (IF) on cryosections was performed according to established protocols (Gazea et al., 2016). Cryosections (12 µm) were pre-incubated with 5% normal pig serum in PBS for 30 min at room temperature before overnight incubation at 4◦C with the primary antibodies appropriately diluted in the blocking solution. After washing in PBS the sections were incubated with the secondary antibodies (diluted in 5% pig serum in PBS) for 30 min at room temperature. After further washes in PBS the sections were stained for 10 min with 4 0 ,60 -diamidino-2-pheylindol (DAPI, Thermoscientific) diluted in PBS.

Free-floating cryosections (30 µm) were treated in a similar way; however, PBST (0.2% Triton X-100 in PBS) was used for washing steps and blocking solution. After staining the sections were placed on glass slides. Paraffin sections (7 µm) containing the striatum comprised between Bregma +0.14 mm and −0.98 mm were incubated with primary antibodies overnight at 4◦C. Visualization of antigen-bound primary antibodies was carried out following antigen retrieval (HK086- 9K, Biogenex). The following primary antibodies were used: EM48 (1:100, MAB5374, Millipore), NeuN (1:100, MAB377, Millipore), adenylate cyclase III (ACIII; 1:500, SC588, Santa Cruz Biotechnology, Dallas, TX, USA), TH (1:500, AB1542, Millipore), D1R (1:500, D2944, Sigma), glial fibrillary acidic protein (GFAP; 1:300, G-3893, Sigma), phospho-S6 (S235/236; 91B2; 1:100, 4857S, Cell Signaling), p62/SQSTM1 (1:100, P0067, Sigma). Incubation with the appropriate secondary antibodies marked with the fluorophores Alexa 594 or Alexa 488 (1:100, Thermoscientific) was followed by DAPI staining.

# Confocal Microscopy Imaging and Image Analysis

Images were acquired by a Leica SP8 confocal and a Leica LASX software. For PC number and length quantification, we used a 63× oil-immersion objective maximal intensity projection z-stacked images (1 µm interval). To determine the number of striatal NeuN positive neurons showing PC stained by ACIII antibody eight non-consecutive free-floating cryosections per mouse were analyzed (one every fourth section, each 30 µm thick) in control and mutant mice. Either the number or the length of PC per microscopic field at 63× magnification was counted. In general, about 80 cells per mouse were measured by the ImageJ software after tracing the ACIII signal in line with established protocols (Miyoshi et al., 2014; Parker et al., 2016). For semi-quantitative analysis of D1R and TH immunoreactivity, and of GFAP and phosphoS6 positive cells, we used a 20× oil-immersion objective for single planes. For p62 positive cells, we used a 63× and for the EM48 a 100× oil-immersion objective. To determine D1R and TH immunoreactivity the optical fiber density was measured in cryosections (12 µm) at different ages by ImageJ software. The quantification was performed in eight coronal serial sections per mouse on 8-bit images (grayscale). To measure the mean optical intensity the ''Mean Gray Value'' was determined. The respective immunoreactivity was measured by subtracting the mean gray value of the respective background from the mean gray value of dorsal striatum. The measurements were limited to the drawn region of interest and the same area was selected in all sections (0.05 mm<sup>2</sup> ). The number of GFAP, pS6 and p62 and EM48 positive cells per microscopic field was counted in at least four independent paraffin sections per mouse in the dorsolateral striatal region. The EM48 signal area is calculated in µm<sup>2</sup> . We used the Quick LUT view to avoid the acquisition of images with under- and over-saturated pixels. The quantification and counting were performed blind to genotype and age.

# HPLC Analysis of Dopamine Content

The tissue levels of DA, 3,4-dihydroxyphenylacetic acid (DOPAC), homovanillic acid (HVA) and 3-methoxytyramine (3-MT), were measured using high-performance liquid chromatography with electrochemical detection (HPLC-EC). The concentrations of endogenous DA and its metabolites (DOPAC, HVA) were measured using HPLC-EC, according to a previously described method (Sikora et al., 2016). Briefly, the striatum was isolated by the adult mouse brain matrix (coronal slices, World Precision Instruments) in the 2 mm region comprised between Bregma +1.42 mm and −0.58 mm. After weighing the tissue samples were deep-frozen in dry ice and stored at −80◦C until further use. Prior to analysis, the samples were homogenized in ice-cold 0.1 M HClO<sup>4</sup> and were centrifuged at 10,000 g for 10 min at 4◦C. The supernatant (5 µl) was injected into the HPLC system. The chromatography system consisted of an LC-4C amperometric detector with a cross-flow detector cell (BAS, IN, USA), an Ultimate 3000 pump (Thermoscientific, USA) and a Hypersil Gold analytical column (3 µm, 100 × 3 mm, Thermoscientific, USA). The mobile phase consisted of 0.1 M KH2PO4, 0.5 mM Na2EDTA, 80 mg/L sodium 1-octane sulfonate, and a 4% methanol, adjusted to pH 3.8 with an 85% H3PO4. The flow rate was 1 ml/min. The potential of a 3-mm glassy carbon electrode was set at 0.7 V with a sensitivity of 5 nA/V. The temperature of the column was maintained at 30◦C. The Chromax 2007 program (Pol-Lab, Warszawa, Poland) was used for data collection and analysis. The external standard consisted of noradrenaline (NA), DA, 5-hydroxytryptamine (serotonin; 5-HT) and 5-hydroxy indole acetic acid (5-HIAA; SIGMA) at a concentration of 50 ng/ml.

### Statistical Analysis

All data are expressed as mean ± SEM. Two-tailed unpaired Student's t-test was used for single comparisons. Multiple comparisons were performed either by one-way ANOVA or by two-way ANOVA with post hoc analyses as indicated in the figure legends (GraphPad Prism Software Inc.).

# RESULTS

# Altered Primary Cilia Length in the Striatum but Not in Cortex of zQ175 Mice

By IF and confocal analysis of brain sections, we observed the presence of sporadic mHTT inclusions at 4 months in the striatum. From 8 months mHTT accumulated mostly in the striatum but also in the cortex, in agreement with previous results (Carty et al., 2015; **Figures 1A–F**). To test the hypothesis that altered PC impacts HD neuropathology, we first investigated PC alterations in the heterozygous zQ175 knock-in model of HD. We measured the stage-specific alterations of PC length in both

PC (green) on cryosections from control and zQ175 mice in the dorsolateral striatum and cortex. Nuclei are visualized by 4<sup>0</sup> ,60 -diamidino-2-pheylindol (DAPI) staining (blue). Scale bar: 25 µm. (G–N) Representative confocal images of immunofluorescent stainings on cryosections by NeuN, as a neuronal marker (red), ACIII (green), in the striatum (G,J) and cortex (K,N) to identify PC that protrudes from NeuN labeled neurons. Scale bars: 25 µm (G–J, M,N), 12 µm (K,L), 8 µm (insets, I and J). Arrows point to the area in the inset. (O,P) Diagrams showing the analysis of PC average length in striatal and cortical neurons at 4 and 8 months in control and zQ175 mice; N = 3, 5 control, and N = 4, 4 zQ175. Values represent means ± SEM. <sup>∗</sup>p < 0.05, ∗∗∗p < 0.0005, two-way ANOVA followed by Tukey's post hoc test for multiple comparisons.

striatum and cortex at 4 and 8 months in control and zQ175 mice, focusing on cells labeled by the neuronal nuclei marker NeuN (**Figures 1G–P**). PC were identified by ACIII, a major PC marker in many regions of the adult mouse brain (Bishop et al., 2007). We found that PC in dorsolateral striatum of zQ175 mice were shorter than their littermate controls at 4 months, while they were longer than respective controls at 8 months (**Figure 1O**). Notably, PC length in the striatum of controls, decreased between 4 and 8 months, but this age-dependent decrease did not occur in the zQ175 mice (**Figure 1O**). Next, we analyzed the cell-specificity of this phenotype by measuring the average PC length in cortical cells (**Figures 1K–N**). Interestingly, cortical cells of zQ175 mice did not show changes in PC length at any of the considered stages (**Figure 1P**).

Hence, mHTT accumulation either directly or indirectly affects PC length in a stage- and region-specific fashion, suggesting that altered PC function might contribute to striatal vulnerability in HD.

### Conditional Ablation of the Ift88 Gene in Dopaminoceptive Neurons Leads to a Mild Increase of Dopamine Metabolite Levels in the Striatum

To investigate the effects of PC loss on HD neuropathology, we first generated inducible mutant mice conditionally lacking Ift88, a gene encoding a microtubular component essential for the formation of the PC (Haycraft et al., 2007). These mice are characterized by the expression of Cre recombinase under the control of dopaminoceptive D1-receptor (D1R; Lemberger et al., 2007; **Figures 2A–D**). The conditional ablation of the Ift88 gene resulted in D1R:Cre;Ift88flox/flox (abbreviated as Ift88 cKO) mice lacking PC in most of the striatal neurons stained for neuronal nuclei (NeuN; **Figure 2E**). The Ift88 cKO mice did not show any gross abnormalities. We monitored changes in body weight (g) over time and we found evidence of increased weight in the Ift88 cKO at 1 year (eight male control, 34.4 ± 6.1 vs. six male mutants, 44.9 ± 4.9; p = 0.004 two-tailed Student's t-test). This conditional model enabled us to mimic a condition of loss of PC in dopaminoceptive neurons to identify their context-specific function. Next, we analyzed D1R expression as a read-out of the survival of positive neurons in control and Ift88 cKO at different ages (1, 3, 6, and 12 months; **Figures 2F–H**). To investigate the impact of PC loss on neuronal survival, we analyzed the immunoreactivity of D1R in dorsolateral striatum at different ages in control and Ift88 cKO mice by IF and semi-quantitative analysis of D1R mean signal intensity upon confocal imaging (**Figure 2H**). This analysis showed no significant differences in D1R immunoreactivity at any of the considered ages in control and Ift88 cKO mutant mice, if at all a tendency to higher D1R immunoreactivity, suggesting that PC is not required for survival of D1R MSNs under basal conditions.

To further characterize the phenotype of the Ift88 cKO, we investigated dopaminergic input to dopaminoceptive striatal neurons (**Figure 3**). To this end, we focused on young (1 month) and older (1 year) mice. We measured tyrosine-hydroxylase (TH) immunoreactivity by IF (**Figures 3A,B**). Interestingly, TH immunoreactivity increased in the mutant at 1 year (**Figure 3C**). Next, we asked whether the loss of PC in D1R neurons affects DA content in the striatum (**Figures 3D–G**). High performance liquid chromatography followed by electrochemical detection (HPLC-EC) showed that 1-year-old Ift88 cKO mutant mice are characterized both by a tendency to increased levels of total DA content in the striatum as well as by a mild but significant increase of the two most important DA metabolites DOPAC, and HVA (**Figures 3D–F**).

These observations indicate subtle PC-dependent crosstalk between PC-depleted dopaminoceptive neurons and dopaminergic neurons to cope with a potentially altered dopaminergic neurotransmission.

#### Loss of PC in the zQ175 HD Mice Results in mTOR Activation and in Larger mHTT Nuclear Inclusions

To study the impact of PC loss on HD neuropathology, we generated zQ175 heterozygous mice lacking PC in striatal but not in cortical neurons in an Ift88 cKO background (double mutants, dm; **Supplementary Figure S1**). To establish whether PC loss is toxic in combination with the zQ175 mutation and whether the combination of these mutations results in a more severe HD phenotype, we analyzed control, Ift88 cKO, zQ175, and dm at 8 months (**Supplementary Figure S2** and **Figure 4**). We found no significant difference in the number of GFAP positive cells, as a marker of astrogliosis, associated with neurodegeneration, suggesting that there is no induction of massive degeneration at least until this stage (**Supplementary Figures S2A–D,I**). To monitor the clearance of protein aggregates, we quantified changes in the number of sequestosome 1 (SQSTM1/p62) protein immunopositive cells in the dorsolateral striatum by IF in the four experimental groups above (**Supplementary Figures S2E–H,J**). p62 is a cargobinding protein associated with proteotoxic stress (Johansen and Lamark, 2011). As shown in **Supplementary Figures S2G,H**, the localization of p62 in the zQ175 and in the dm mice was intra-nuclear rather than cytoplasmic, as in controls and Ift88 cKO, however, the number of p62 positive cells was comparable between all groups. Next, we investigated mTOR kinase activity by IF based on the levels of one of its targets, phosphorylated ribosomal protein S6 (phospho-S6, pS6; **Figures 4A–E**). Interestingly, zQ175 and dm mice showed a higher number of pS6 positive cells in comparison to controls, and this was significantly higher in the dm in comparison to the zQ175 (**Figure 4E**), suggesting that PC loss in the zQ175 mice promotes mTOR activation. Moreover, we analyzed mHTT nuclear inclusions that are absent in controls and Ift88 cKO, and visible in the zQ175 and dm (**Figures 4F–I**). At the age examined (8 months), the number of mHTT positive cells was similar in the zQ175 and dm mice (**Figure 4J**). The area of the mHTT signal in the dm is ca. Thirty percent larger in comparison to the zQ175 mice, indicating that PC limit mHTT accumulation in this model (**Figure 4K**).

# DISCUSSION

In the present study, we addressed the following three questions: (1) whether PC is specifically altered in different neuronal types and HD stages; (2) what are the consequences of PC disruption in striatal neurons for their survival; and (3) to what extent these contribute to HD neuropathology in a neuronal population vulnerable in HD such as the striatum. We showed

that PC length is affected in the striatum in association with mHTT accumulation. PC disruption in MSN does not affect neuronal survival. Subtle compensatory mechanisms might be activated by dopaminergic neurons in response to PC loss on dopaminoceptive neurons; however, a detailed characterization awaits future experiments. PC disruption in striatal neurons of an HD mouse model results in increased mTOR activation and larger mHTT nuclear inclusions, suggesting that PC is required in a pathological context.

Changes in PC length have been reported in previous animal and cellular models expressing mHTT; however, this is the first study focusing on PC pathology in the striatal neurons that are pathologically affected in HD and addressing the impact of loss of PC in a progressive mouse model of HD. Here, we identified changes in PC length that are concomitant with mHTT accumulation: PC length is altered in striatal but not in cortical neurons in the zQ175 mice. Previous work has indicated that HTT is important for the transport of proteins required for ciliogenesis (Keryer et al., 2011). Ciliogenesis is a tightly regulated process, and it is crucial for its proper function that PC is produced in the right size and time (Avasthi and Marshall, 2012). Given the time- and cell-specific impact of mHTT on PC, it will be important to better understand how differences in mHTT state differentially modulate PC maintenance.

Interestingly, various HD models show altered levels of DA and its metabolites (Smith et al., 2014; Koch and Raymond, 2019), and lack of dopaminergic inputs results in longer PC in the striatum (Miyoshi et al., 2014). Hence changes of DA input in the zQ175 mice might alter PC length. Because DA receptors are located in the PC membrane

respective controls by unpaired Student's t-test.

means ± SEM. <sup>∗</sup>p < 0.05, ∗∗∗p < 0.001 by unpaired Student's t-test with respective controls.

(Marley and von Zastrow, 2010; Domire et al., 2011; Leaf and Von Zastrow, 2015), it will be important to investigate striatal synaptic transmission, and motor and psychiatric phenotypes in the HD mice lacking PC. Future studies should address the behavioral implications of the mild increase in DA metabolite levels and increased TH expression in the striatum. These studies might help to understand the impact of current DA-based symptomatic HD treatments on PC physiology and homeostasis.

The generation of tools enabling targeting of PC function independently of changes in their structural integrity might be necessary to define subtle regulatory functions. The Ift88 cKO approach, in which both PC function and structure are affected, does not allow us to dissect out the specific role of PC signaling from PC structure, however, it does allow us to identify cell-autonomous effects as well as non-cell autonomous responses to ciliary impairment. In aged homozygous zQ175 mice mTOR activity increases (Abd-Elrahman and Ferguson, 2019). We found in the zQ175 and in the dm mice increased phospho-S6 positive cells. Hence, another possible mechanism to explain changes in PC length might be linked to a block of autophagy by mTOR activation (Pampliega et al., 2013). Together, it will be important to analyze older zQ175 and dm mice. Conditional ablation of Ift88 in neural progenitors by the Nestin-Cre transgenic line revealed a significant increase in mTOR pathway activity and phospho-S6 ribosomal protein levels, resulting in hydrocephaly at late embryonic stages (Foerster et al., 2017). In the D1R-Cre driven conditional Ift88 knock-out, phospho-S6 does not increase in comparison to controls. These results indicate the cell-type, developmental stage and age-specific function of PC alterations, and support the need to further address this question in a mature and aging brain.

(N = 5) mice. (F–I) Representative confocal images of mHTT immunostaining (green) by EM48 antibody and DAPI (blue) on paraffin sections from dorsolateral striatum in control, Ift88 cKO, zQ175, and dm at 8 months. Scale bar: 10 µm. (J,K) Quantification of the EM48 positive cells expressed as a percentage of DAPI positive cells and of the mean EM48 signal area (in µm<sup>2</sup> ) per microscopic field. Values represent means ± SEM. <sup>∗</sup>p < 0.05 (p = 0.019) based on two-tailed unpaired Student's t-test; zQ175 (N = 8) and dm (N = 6) mice.

Diseases primarily caused by ciliary dysfunctions are commonly referred to as ciliopathies. As of yet, we cannot ascribe HD among the primary ciliopathies, nevertheless, PC structure and function are altered in association with a more severe phenotype. Notably, the role of PC in neuronal homeostasis, and in the control of the cellular stress response signaling cascades is emerging in other neurodegenerative disorders. PC pathology has been described for example in Alzheimer's disease (AD), in Parkinson's disease (PD) and in spinocerebellar ataxias, another polyglutamine disease, although distinct mechanisms are probably involved (Steger et al., 2017; Bowie et al., 2018; Dhekne et al., 2018; Vorobyeva and Saunders, 2018). Interestingly, PC were elongated in the hippocampus of the APP/PS1 mouse models of AD compared with wild-type mice, and serotonin 5-HT6 receptors playing a critical role in AD development regulate the morphology and function of neuronal PC (Hu et al., 2017). Moreover, in mouse models of PD expressing mutant LRRK2 R1441C, PC was affected in cholinergic neurons while the overall ciliation of neurons in the striatum was not significantly different from wild type (Steger et al., 2017; Dhekne et al., 2018). Nevertheless, defective ciliogenesis in striatal cholinergic neurons might impair a protective mechanism involving non-cell autonomous Sonic hedgehog between cholinergic and dopaminergic neurons (Gonzalez-Reyes et al., 2012). The Spinocerebellar ataxia type 11-associated mutation of the serine/threonine kinase Tau tubulin kinase 2 dominantly interferes with ciliogenesis and cilium stability (Bowie et al., 2018). Shortened primary cilium length and dysregulated Sonic hedgehog signaling were also reported in Niemann-Pick type C1 (NPC1) disease, a neurodegenerative lysosomal storage disorder caused by mutations in the NPC1 gene (Canterini et al., 2017).

In summary, although we obtained similar mean PC length as previously reported for striatum and cortex in mice (Sipos et al., 2018), a systematic quantitative comparison of PC marker expression and length in various neurodegenerative disease models and in human tissues will benefit from the use of automatized segmentation approaches (Vorobyeva and Saunders, 2018), and high content automated image acquisition. These future studies will be important to determine in other disease models which cells and tissues display the cilia defects and at what stages, and try to understand when, whether and how those changes lead to specific neuronal loss in the brain.

#### DATA AVAILABILITY STATEMENT

All datasets generated for this study are included in the article/**Supplementary Material**.

#### REFERENCES


# ETHICS STATEMENT

Procedures involving animal care were approved by the Committee on Animal Care and Use (Regierungspräsidium Karlsruhe, Germany) in accordance with the local Animal Welfare Act and the European Communities Council Directives (2010/63/EU and 2012/707/EU; Ref. number: G-252/17).

# AUTHOR CONTRIBUTIONS

RM, GK, KK, and RP: data acquisition. RM, GK, JK, A-EW, KT, and RP: data analysis and interpretation. RP: study concept and design, drafting of the manuscript. All authors revised the submitted manuscript.

#### FUNDING

This work was supported by the European Huntington's Disease Network (EHDN) seed-fund project 0907, by the ''Deutsche Forschungsgemeinschaft'' (DFG): DFG PA 1529/2-1 to RP, by start-up funds from the Center for Excellence in the Neurosciences at the University of New England (KT), and by the 2017/25/B/NZ7/02406 (Opus13) grant from the National Science Center and statutory funds of the Institute of Pharmacology (PAS) to GK and KK.

#### ACKNOWLEDGMENTS

We acknowledge the CHDI Foundation for providing the zQ175 mice used to establish the colony.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fncel.2019.005 65/full#supplementary-material.

zQ175 mouse model of Huntington's disease: an in vivo high-content imaging study. PLoS One 10:e0123527. doi: 10.1371/journal.pone.0123527


**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 © 2019 Mustafa, Kreiner, Kaminska, Wood, Kirsch, Tucker and Parlato. ´ 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.

digital media

of impactful research

article's readership