MODEL ORGANISMS: A PRECIOUS RESOURCE FOR UNDERSTANDING OF THE MOLECULAR MECHANISMS UNDERLYING HUMAN PHYSIOLOGY AND DISEASE

EDITED BY : Maria Grazia Giansanti and Roberta Fraschini PUBLISHED IN : Frontiers in Genetics

#### Frontiers Copyright Statement

© Copyright 2007-2019 Frontiers Media SA. All rights reserved. All content included on this site, such as text, graphics, logos, button icons, images, video/audio clips, downloads, data compilations and software, is the property of or is licensed to Frontiers Media SA ("Frontiers") or its licensees and/or subcontractors. The copyright in the text of individual articles is the property of their respective authors, subject to a license granted to Frontiers.

The compilation of articles constituting this e-book, wherever published, as well as the compilation of all other content on this site, is the exclusive property of Frontiers. For the conditions for downloading and copying of e-books from Frontiers' website, please see the Terms for Website Use. If purchasing Frontiers e-books from other websites or sources, the conditions of the website concerned apply.

Images and graphics not forming part of user-contributed materials may not be downloaded or copied without permission.

Individual articles may be downloaded and reproduced in accordance with the principles of the CC-BY licence subject to any copyright or other notices. They may not be re-sold as an e-book.

As author or other contributor you grant a CC-BY licence to others to reproduce your articles, including any graphics and third-party materials supplied by you, in accordance with the Conditions for Website Use and subject to any copyright notices which you include in connection with your articles and materials.

All copyright, and all rights therein, are protected by national and international copyright laws.

The above represents a summary only. For the full conditions see the Conditions for Authors and the Conditions for Website Use. ISSN 1664-8714 ISBN 978-2-88963-183-4 DOI 10.3389/978-2-88963-183-4

#### 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

# MODEL ORGANISMS: A PRECIOUS RESOURCE FOR UNDERSTANDING OF THE MOLECULAR MECHANISMS UNDERLYING HUMAN PHYSIOLOGY AND DISEASE

Topic Editors: Maria Grazia Giansanti, National Research Council (CNR), Italy Roberta Fraschini, University of Milano-Bicocca, Italy

Citation: Giansanti, M. G., Fraschini, R., eds. (2019). Model Organisms: A Precious Resource for Understanding of the Molecular Mechanisms Underlying Human Physiology and Disease. Lausanne: Frontiers Media. doi: 10.3389/978-2-88963-183-4

# Table of Contents


Ksenia Smurova and Peter De Wulf

*82 Altered Expression of Mitochondrial NAD+ Carriers Influences Yeast Chronological Lifespan by Modulating Cytosolic and Mitochondrial Metabolism*

Ivan Orlandi, Giulia Stamerra and Marina Vai

*95 Ras-Induced miR-146a and 193a Target Jmjd6 to Regulate Melanoma Progression*

Viviana Anelli, Anita Ordas, Susanne Kneitz, Leonel Munoz Sagredo, Victor Gourain, Manfred Schartl, Annemarie H. Meijer and Marina Mione

#### *108 High MYC Levels Favour Multifocal Carcinogenesis* Manuela Sollazzo, China Genchi, Simona Paglia, Simone Di Giacomo, Annalisa Pession, Dario de Biase and Daniela Grifoni

*123 p53-Sensitive Epileptic Behavior and Inflammation in* Ft1 *Hypomorphic Mice*

Romina Burla, Mattia La Torre, Giorgia Zanetti, Alex Bastianelli, Chiara Merigliano, Simona Del Giudice, Alessandro Vercelli, Ferdinando Di Cunto, Marina Boido, Fiammetta Vernì and Isabella Saggio

*135 Modeling Congenital Disorders of N-Linked Glycoprotein Glycosylation in*  Drosophila melanogaster

Anna Frappaolo, Stefano Sechi, Tadahiro Kumagai, Angela Karimpour-Ghahnavieh, Michael Tiemeyer and Maria Grazia Giansanti

*143 SUMO-Targeted Ubiquitin Ligases (STUbLs) Reduce the Toxicity and Abnormal Transcriptional Activity Associated With a Mutant, Aggregation-Prone Fragment of Huntingtin*

Kentaro Ohkuni, Nagesh Pasupala, Jennifer Peek, Grace Lauren Holloway, Gloria D. Sclar, Reuben Levy-Myers, Richard E. Baker, Munira A. Basrai and Oliver Kerscher


# Editorial: Model Organisms: A Precious Resource for the Understanding of Molecular Mechanisms Underlying Human Physiology and Disease

*Maria Grazia Giansanti1\* and Roberta Fraschini2\**

*1 Istituto di Biologia e Patologia Molecolari del CNR, Dipartimento di Biologia e Biotecnologie, Università Sapienza di Roma, Roma, Italy, 2 Dipartimento di Biotecnologie e Bioscienze, Università degli Studi di Milano-Bicocca, Milano, Italy*

Keywords: model organisms, human diseases, human physiological model, DNA damage, human disorders

**Editorial on the Research Topic** 

**Model Organisms: A Precious Resource for the Understanding of Molecular Mechanisms Underlying Human Physiology and Disease**

#### INTRODUCTION

#### *Edited and Reviewed by:*

*Inês Barroso, University of Cambridge, United Kingdom*

#### *\*Correspondence:*

*Maria Grazia Giansanti mariagrazia.giansanti@uniroma1.it Roberta Franschini roberta.fraschini@unimib.it*

#### *Specialty section:*

*This article was submitted to Genetic Disorders, a section of the journal Frontiers in Genetics*

*Received: 05 July 2019 Accepted: 08 August 2019 Published: 10 September 2019*

#### *Citation:*

*Giansanti MG and Fraschini R (2019) Editorial: Model Organisms: A Precious Resource for the Understanding of Molecular Mechanisms Underlying Human Physiology and Disease. Front. Genet. 10:822. doi: 10.3389/fgene.2019.00822*

This issue includes eight reviews and five research articles, which highlight how research in model organisms lays the foundations for the comprehension of molecular mechanisms underlying human diseases. Although budding yeast and humans are separated by a billion years of evolutionary history, more than 400 essential yeast genes can be replaced with their human orthologs. Pioneering genetic studies in yeast have contributed to understand the mechanisms involved in autophagy and vesicle trafficking, two processes involved in cancer and neurodegenerative disorders (Novick et al., 1980; Takeshige et al., 1992; Mizushima et al., 1998). More recently, production of yeast strains expressing human genes ("humanized yeast") has been essential for the detailed analysis of normal and pathogenic variants (Laurent et al., 2016). *Drosophila melanogaster* provides an extremely valid resource to investigate the mechanisms involved in organ formation and in the pathology of human diseases. Nearly 65% of human genes have orthologs in *D. melanogaster*, and nearly 75% of the genes involved in human disease have functional orthologs in flies (Reiter et al., 2001; Chien et al., 2002). The sophisticated genetic tools offered by *Drosophila* allow rapid generation of models for human disease, assaying the functional effects of human variant alleles and testing new therapeutic drugs (Moulton and Letsou, 2016; Wangler et al., 2017). *Danio rerio* shares vertebrate-conserved characteristics with human including very similar organs and is a highly suitable model system for investigating gene functions involved in hematopoiesis and screening for novel potential drugs (Wangler et al., 2017). Mouse models of human diseases are the most commonly used, reflecting the genetic and physiological similarities between humans and mice (Perlman, 2016).

#### Using Budding Yeast to Study the Molecular Pathways That Are Altered in Human Diseases

Orlandi et al. used yeast as a model system to study aging of post-mitotic mammalian cells. Their data describe a connection between nicotinamide adenine dinucleotide (NAD+) content,

**5**

mitochondrial functionality, and chronological life span. They show that, during chronological aging, an altered expression of the specific mitochondrial NAD+ carriers deeply influences the metabolic reprogramming that enables cells to acquire features required to maintain viability during aging.

Ohkuni et al. used yeast as a cellular model of neurodegenerative disorders such as Huntington's disease (HD). They show that the SUMO-targeted ubiquitin ligase (STUbL) Slx5 reduces the toxicity and abnormal transcriptional activity associated with a mutant fragment of huntingtin (Htt) that induces aggregation, the causative agent of HD. Importantly, RNF4, the human ortholog of Slx5, limits the aberrant transcriptional activity of aggregation-prone Htt in yeast and in several cultured human cell lines. Thus, this study uncovers a conserved pathway that counteracts the accumulation of aggregating, transcriptionally active Htt, on chromatin in both yeast and in mammalian cells.

Fraschini's review is focused on the molecular pathways and proteins involved in the control mitotic spindle morphogenesis and function, which are highly conserved from yeast to humans and whose impairment is connected with the development of human diseases. Fraschini illustrates the processes of spindle formation and orientation in yeast and in humans and includes many examples of misregulation that lead to the development of cancer and other human diseases.

Smurova and De Wulf report the role of centromeres and kinetochores in preserving genetic stability; in particular, they describe how centromere transcription contributes to faithful kinetochore function, how pericentromeric chromatin is silenced by RNA processing, and the transcriptional misregulation of (peri)centromeres during stress, natural aging, and disease.

Bonetti et al. illustrate how DNA ends are processed in budding yeast in order to maintain genome stability. DNA double-strand breaks (DSBs) are dangerous lesions that can be repaired by homologous recombination (HR) that occurs after DNA ends are correctly processed by several nucleases by a process called DNA end resection. The same nucleases function also during DNA replication in the processing of replication fork structures. The authors describe current knowledge of the mechanism of resection at DNA DSBs and replication forks.

Natali and Rancati describe the mutator phenotype in budding yeast. The mutator phenotype enhances genome instability and generation of phenotypic variation in a population of cells and increases the probability that some of these variations undergo selection and clonal expansion in challenging environments. This issue is particularly relevant for human health since cancer cells experience an increased mutational charge during early steps of carcinogenesis. The authors discuss the activity of the DNA replication and repair machineries and how mutations can confer increased genome instability. In addition, they describe recent clinical evidences in cancer biology indicating that these lessons can be applied to tumor development.

#### Using Model Organisms to Model Tumor Formation and Progression

Sollazzo et al. used *Drosophila* larval wing epithelium to investigate the impact of MYC upregulation on cells carrying mutations in neoplastic tumor suppressor genes (nTSGs). MYC overexpression confers to cells mutant for different nTSGs, the ability to initiate multifocal, three-dimensional growth, a hallmark of mammalian pre-cancerous fields.

By using three different models of Ras induction and tumor formation in zebrafish, Anelli et al. demonstrate that six microRNAs increase following expression of a constitutively active *HRASG12V* allele. Two Ras-induced microRNAs, namely, miR-146a and miR93a, target the Jmjd6 gene, which encodes a JumonjiC domain protein. Results in this study show that Jmjd6 plays a critical role in zebrafish melanoma development and that miR-146a and miR93a function as tumor suppressors, antagonizing Jmjd6 activation.

Mirzoyan et al. provide a comprehensive insight into the signaling pathways involved in tumorigenesis that are conserved in flies and highlight the ease to genetically manipulate these circuits to study cancer biology. In addition, they describe examples of *Drosophila*  cancer models and their use to identify new therapeutic strategies.

Merigliano et al. describe the link between vitamin B6, diabetes, and cancer. Although several data indicate that diabetes and cancer are correlated, the molecular mechanisms involved remain to be clarified. Recent results obtained in *Drosophila* indicate that vitamin B6 deficiency causes hyperglycemia and increases DNA damage. These data suggest that, in diabetic patients, high PLP levels should increase the frequency of DNA damage thus contributing to cancer formation and progression.

#### Using Model Organisms to Study the Neurological Defects Associated With Human Diseases

Burla et al. (2018) demonstrate that mutant mice with progeroid traits, caused by reduced expression of the *Ft1* gene (the ortholog of human *AKTIP*), display repeated seizures not linked to overt brain morphological alterations or severe neurodegeneration. However, Ft1 reduction is associated with the activation of the inflammatory markers IL-6 and TGF-β. Remarkably, reduction of the guardian of the genome p53 rescues the epileptic behavior and reverses back the expression of IL-6 and TGF-β in *Ft1* mutant mice, suggesting an involvement of DNA damage response in these phenotypes.

Two reviews describe the use of *D. melanogaster* for dissecting the molecular mechanisms underlying the neurological defects in inherited human diseases. Congenital disorders of glycosylation (CDGs) are multisystemic diseases caused by mutations in genes controlling the glycosylation pathways (Freeze et al., 2015). Most CDGs are associated with neurological defects, including mental retardation and seizures. As described by Frappaolo et al. *D. melanogaster* is emerging as a well-suited model organism for modeling congenital disorders of N-linked glycoprotein glycosylation due to a well-characterized glycome and a plethora of electrophysiological and behavioral assays that can be used to test neurological alterations in the whole organism.

The fragile-X (Fra-X) syndrome, caused by mutations in the fragile-X mental retardation (*Fmr1*) is associated with intellectual disability, autism, hyperactivity and language delay, long face and large ears, macroorchidism, and irregular spermatids (Santoro et al., 2012). The *Drosophila* Fra-X disease model recapitulates many phenotypic aspects of the Fra-X syndrome including defective neuronal architecture and synaptic function and altered germline development. Specchia et al. describe the involvement of dFmr1/FMRP protein in the piRNA pathway and in the DNA genome response, which may open up new perspectives in the search of potential therapeutic targets.

## AUTHOR CONTRIBUTIONS

MGG and RF wrote the editorial.

#### REFERENCES


#### FUNDING

RF researches are supported by grants from PRIN (Progetti di Ricerca di Interesse Nazionale) and from the University of Milano Bicocca (FA). Financial support from the Italian Ministry of University and Research (MIUR) through grant "Dipartimenti di Eccellenza- 2017 "to University of Milano Bicocca, Department of Biotechnology and Biosciences is also acknowledged. Research in MG lab is supported by a grant from Fondazione AIRC per la ricerca sul Cancro (AIRC), grant IG 2017, Id 20779.

*Drosophila* melanogaster. *Genome Res.* 11 (6), 1114–1125. doi: 10.1101/ gr.169101


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

# The Mutator Phenotype: Adapting Microbial Evolution to Cancer Biology

#### *Federica Natali1,2 and Giulia Rancati1\**

*1 Institute of Medical Biology (IMB), Agency for Science, Technology and Research (A\*STAR), Singapore, Singapore, 2 School of Biological Sciences, Nanyang Technological University, Singapore, Singapore*

The mutator phenotype hypothesis was postulated almost 40 years ago to reconcile the observation that while cancer cells display widespread mutational burden, acquisition of mutations in non-transformed cells is a rare event. Moreover, it also suggested that cancer evolution could be fostered by increased genome instability. Given the evolutionary conservation throughout the tree of life and the genetic tractability of model organisms, yeast and bacterial species pioneered studies to dissect the functions of genes required for genome maintenance (caretaker genes) or for cell growth control (gatekeeper genes). In this review, we first provide an overview of what we learned from model organisms about the roles of these genes and the genome instability that arises as a consequence of their dysregulation. We then discuss our current understanding of how mutator phenotypes shape the evolution of bacteria and yeast species. We end by bringing clinical evidence that lessons learned from single-cell organisms can be applied to tumor evolution.

#### *Edited by:*

*Roberta Fraschini, University of Milano-Bicocca, Italy*

#### *Reviewed by:*

*Michael McDonald, Monash University, Australia Sylvester Holt, INSERM U1081 Institut de recherche sur le cancer et le Vieillissement, France*

#### *\*Correspondence:*

*Giulia Rancati giulia.rancati@imb.a-star.edu.sg*

#### *Specialty section:*

*This article was submitted to Genetic Disorders, a section of the journal Frontiers in Genetics*

*Received: 15 October 2018 Accepted: 05 July 2019 Published: 06 August 2019*

#### *Citation:*

*Natali F and Rancati G (2019) The Mutator Phenotype: Adapting Microbial Evolution to Cancer Biology. Front. Genet. 10:713. doi: 10.3389/fgene.2019.00713*

Keywords: mutator phenotype, cell-to-cell heterogeneity, adaptation, selective pressure, asexually reproducing organisms

# INTRODUCTION

The mutator phenotype was proposed by Loeb almost 40 years ago to reconcile the observations that while cancer cells display widespread DNA and chromosomal changes (Alexandrov et al., 2013; Loeb, 2016), the rate of spontaneous mutation in somatic cells is low (Milholland et al., 2017). This hypothesis also suggested that an increased genome instability favors cancer evolution. Indeed, by reshuffling cancer cell genomes, the mutator phenotype generates cell-to-cell heterogeneity, which is the presence of cells with different genotypes and phenotypes within a population (Loeb, 2016). Since clonal competition within a tumor mass favors the expansion of fitter cells, the presence of cells with different phenotypes within a cancer sample increases the likelihood that some of them might be more aggressive and ultimately leads to poor patient survival (Greaves, 2015; McGranahan and Swanton, 2017). Thanks to the ease with which they can be grown in the laboratory and the wealth of genetic resources, in the last decades, model organisms have been extensively used to identify and dissect the function of genes required for genome stability. For instance, some of the available yeast genome-wide libraries include the systematic knockout collection (Winzeler et al., 1999; Giaever et al., 2002), the green fluorescent protein-tagged collection (Huh et al., 2003), and loss-of-function alleles for essential genes (Ben‐Aroya et al., 2008; Breslow et al., 2008; Li et al., 2011). Though progress in our ability to perform high-throughput screens and to manipulate mammalian cells has greatly improved (Behan et al., 2019), it is still difficult to envision a close future in which large-scale screens performed in model organisms could be easily and cost-effectively

**8**

Natali and Rancati Microbial Evolution Adapted for Cancer Biology

reproduced in mammalian cells. A more effective way is to directly translate findings of interesting candidates from model organisms to mammalian cells. For instance, a recent screen in budding yeast showed that, contrary to what was previously thought, heterozygous mutations in gatekeeper genes can cause genome instability (Coelho et al., 2019). Introduction of these mutations in human orthologs triggered genome instability also in human cells (Coelho et al., 2019). Interestingly, only some of the identified genes have been previously found mutated in cancer cells, supporting the idea that findings in budding yeast hold great potential for cancer cell biology.

### CELLS LOSING BALANCE

Studies on model organisms have been instrumental to understand mechanisms generating cell-to-cell heterogeneity and its consequence in evolutionary outcomes. While an important route to heterogeneity in yeast is sexual outbreeding (Long et al., 2015; Vazquez‐Garcia et al., 2017; Li et al., 2019), this point will not be discussed as evolution of cancer cells is comparable to evolution of asexually reproducing organisms. A route to heterogeneity common between single-cell model organisms and cancer cells involves the dysregulation of the genome integrity network. Initial identification of players of this network in yeast and bacteria was soon followed by the realization that cancer cells carry recurrent mutations in their respective human orthologs and provided ground for key findings. Some popular examples include the identification in cancer cells of cohesion mutations (Hill et al., 2016), of the links between microsatellite instability and the mismatch repair pathway (Kolodner and Marsischky, 1999), and of predicting novel therapeutic targets based on synthetic lethality (Ashworth et al., 2011; Beijersbergen et al., 2017). Though most of the core machinery required for genome maintenance and replication is conserved throughout evolution, there are important exceptions. The most notable differences include the lack of a welldefined nucleotide sequence of the origin of replication in human cells (Gerhardt et al., 2006) and the presence of the human DNA replication inhibitor, geminin (Symeonidou et al., 2012). Yeast cells also lack obvious orthologs of key human DNA repair enzymes, such as breast related cancer antigens 1/2 (BRCA1/2) or poly (ADP-ribose) polymerase (PARP) (Morales et al., 2014; Walsh, 2015). Given the therapeutic importance of some of these proteins, "humanized" yeast strains carrying cancer-associated mutations in BRCA genes have been generated (Guaragnella et al., 2014; Maresca et al., 2018) and used to screen for novel therapeutics.

Due to space constraints, we briefly summarize the role of key DNA repair genes and DNA replication genes. For a detailed overview of the field, please refer to the reviews by Kunkel and Erie (2015), Lujan et al. (2016), and Liu et al. (2017). For a summary of the genes discussed in this paragraph, please refer to **Table 1**.

#### Mutations Affecting DNA Polymerases

DNA replication of the lagging and the leading strands depends on the activity of two high-fidelity DNA polymerases, DNA polymerases δ (Polδ) and ε (Polε), respectively (Maslowska et al., 2018). The faithfulness of the process relies on the accuracy of nucleotide incorporation coupled with the 3′–5′ exonuclease proofreading activity (Pavlov and Shcherbakova, 2010). Indeed, biochemical assays showed that while purified human Polδ catalyzes one base substitution error every 22,000 nucleotides, the error rate decreased at least 10-fold in presence of a functional proofreading domain (Schmitt et al., 2009). Mouse models lacking a functional Polδ proofreading activity develop spontaneous cancers at high frequency (Goldsby et al., 2002), confirming *in vivo* the importance of the domain for genome stability and cancer formation. Moreover, germline mutations in the proofreading domain of Polδ and Polε have been identified in a number of families with increased susceptibility to colorectal adenomas and carcinomas (Palles et al., 2013; Heitzer and Tomlinson, 2014). Accordingly with an inability to repair mispaired bases inserted during DNA replication, tumors from affected patients increased the rate of base substitution mutations while maintaining microsatellite stability (Palles et al., 2013). Mutations in the proofreading activity of *S. cerevisiae* that mimic mutations in tumors resulted in a mutator phenotype and elevated spontaneous base substitution rates (Murphy et al., 2006; Nick McElhinny et al., 2007).

However, Polδ and Polε mutations outside of the proofreading domain were also mapped in sporadic cancers and cancer cell lines (Briggs and Tomlinson, 2013). Introduction of one of such variant, pol3-R696W (human POLD1-R689W), in heterozygosity in *S. cerevisiae* increased 30-fold the rate of forward mutations. At a biochemical level, pol3-R696W was shown to be an error-prone DNA polymerase with an increased nucleotide misinsertion rate and a specific mutational pattern (Daee et al., 2010) that is consistent with the one observed in colorectal cancer lines bearing the POLD1-R689W variant (Mertz et al., 2017). Collectively, these observations suggest that mutations affecting both the polymerase and the 3'–5' exonuclease domains confer a mutator phenotype that can be translated from bacteria and yeast to human cells.

#### Mutations Affecting Mismatch Repair Genes

The mismatch repair (MMR) pathway is a conserved surveillance system that recognizes and resolves misincorporated bases (Fishel, 2015). In prokaryotes, the MMR machinery is relatively simple and involves proteins detecting DNA mismatches (MutS), processing the damage (MutH), and bridging these two proteins together (MutL) (Fukui, 2010). While mutations in *mutS* and *mutL* human orthologs were found in the germline of patients with hereditary nonpolyposis colorectal cancer/Lynch syndrome (HNPCC/LS), and other cancer-predisposing Lynch variant syndromes (Lynch and de la Chapelle, 2003; Morales-Burgos et al., 2008; Wimmer and Kratz, 2010), they also somatically occur in up to 15% of sporadic colorectal, gastric, or endometrial carcinomas (Boland et al., 1998). Experimentally, engineered mice lacking functional MMR proteins are genomically unstable and predisposed to spontaneous cancer onset (Lee et al., 2016). Modeling cancer-related MMR mutations in yeast has been instrumental to dissect the consequences on cellular physiology TABLE 1 | The table reports a list of mutations discussed in the section "Cells losing balance" and details phenotypic consequences arising from such mutations.


cancer, and prostate cancer (Clyne

*(Continued)*

et al., 2009)

Natali and Rancati


of a non-functional mismatch repair pathway. For instance, mimicking MMR mutations found in HNPCC (Kurzawski et al., 2002) in yeast cells caused an increase in the rate of spontaneous (Drotschmann et al., 1999) and forward mutations (Clyne et al., 2009). Moreover, consistent with the observation that human cancer cell lines with dysregulation in the expression of the MMR proteins are genomically unstable (Ryan et al., 2017; Wilczak et al., 2017), tinkering with the expression levels of the yeast orthologs in *S. cerevisiae* resulted in significant increase of repeats' instability and forward mutations (Shcherbakova and Kunkel, 1999; Chakraborty et al., 2018).

# THE RISE AND THE FALL OF THE MUTATOR PHENOTYPE

While maintenance of genome stability is key for reproductive success of prokaryotes and eukaryotes, laboratory and clinical evidence suggests that tinkering with such pathways favors cellular adaptation and population expansion during exposure to challenges. Below we discuss some such evidence.

#### Lessons from Model Organisms

Several clinical isolates and natural populations of pathogenic bacteria and fungi were reported to have an enhanced mutation rate mostly mapped to defects in the methyl-directed mismatch repair system (Oliver et al., 2000; Bjorkholm et al., 2001; Chopra et al., 2003; Wang et al., 2013; Healey et al., 2016), suggesting that a mutator phenotype could be selected in fluctuating or hostile environments, such as the presence of drugs or adaptation to new ecological niches. Experimental studies have supported this idea. For instance, *MSH2*-defective *Cryptococcus neoformans*, *Candida glabrata*, and *Cryptococcus deuterogattii* strains increased mutation rates and underwent rapid adaptation to antifungal drugs (Healey et al., 2016; Billmyre et al., 2017; Boyce et al., 2017). Similarly, hypermutator *Staphylococcus aureus* bacteria strains impaired in the DNA mismatch repair pathway developed vancomycin resistance more rapidly than control strains (Schaaff et al., 2002). Moreover, mutant strains defective in DNA repair and characterized by increased mutation rates outcompeted wild-type strains and were fixed in 6 out of 12 *E. coli* populations in the Long-Term Evolution Experiment (Tenaillon et al., 2016). A link between increased mutation rates and adaptability comes also from observations that impairment in the activity of DNA repair pathways was often found to co-segregate with mutations conferring antibiotic resistance (Gould et al., 2007). At the theoretical level, a mutator phenotype potentially endows populations with a higher adaptability by generating cell-to-cell heterogeneity and a pool of allelic variants on which selection could select upon (**Figure 1**). Accordingly, mutator *msh2Δ S. cerevisiae* strains acquired resistance to the toxic arginine analog canavanine up to 20-fold faster than wild type (Bowers et al., 1999). Adaptive mutations encompassed single-nucleotide misincorporations and deletions of the canavanine influx pump gene (Sokolsky and Alani, 2000 ). Since such mutations are edited by the MMR pathway, these observations suggest that crippling with DNA replication or repair pathways could generate beneficial

TABLE 1 |

Continued

allelic variants. However, mutators with no direct effect on cellular fitness in asexually evolving unicellular organisms could sweep in a population if they are linked to beneficial mutations, a process called mutator hitchhiking. A large body of evidence coming from theoretical and experimental studies in both bacteria and yeast showed that the probability of hitchhiking has been linked to the population size and the fitness effects of beneficial mutations in complex-to-predict scenarios (Taddei et al., 1997; Notley-McRobb et al., 2002; Shaver et al., 2002; Wahl et al., 2002; Andre and Godelle, 2006; Thompson et al., 2006; Gerrish et al., 2007; Gentile et al., 2011; Raynes et al., 2014; Bui et al., 2015; Good and Desai, 2016). For instance, large population sizes are known to increase clonal interference, which has been shown to either delay or enhance fixation of mutators in different conditions (Raynes et al., 2014; Good and Desai, 2016). However, in populations at local fitness minima or experiencing fluctuating environments, mutators can hitchhike to a higher frequency when linked to strong beneficial mutations. It comes, therefore, as no surprise that mutator multidrug-resistant bacterial strains are a common feature of chronic infections, like

cystic fibrosis or urinary tract infections (Oliver et al., 2000; Labat et al., 2005; Feliziani et al., 2010; Macia et al., 2014), where bacterial strains endure pulses of antibiotic treatments. In the laboratory, coupling of cycles of antibiotic or carbon source selection with mutagenesis increased the percentage of strains carrying mutations in mismatch repair genes up to 50–100% (Mao et al., 1997), further supporting the notion that fluctuating environments positively select for mutator strains. However, since the vast majority of mutations are detrimental, genome instability in populations at their fitness peaks comes with a cost (**Figure 1**). Indeed, a general role for asexual pathogenic mutators in the emergence of drug resistance is still being debated, possibly because mutators are selected against once beneficial mutations have been acquired. For instance, *mutS* mutant *S. aureus* laboratory strains characterized by a 78-fold increased mutation frequency did not increase the rate of adaptation to vancomycin (O'Neill and Chopra, 2002). Also, *MSH2* mutations in *C. glabrata* clades were shown to be present as polymorphisms within different natural populations that were equally sensitive or resistant to antifungal drugs (Carrete et al., 2018). At the same time, *C. glabrata* clinical isolates carrying *MSH2* mutations did not show increased resistance to azole or echinocandin (Singh et al., 2018). Lastly, an *in vivo* model for chronic bone infection in the rat showed that *MSH2* mutant *S. aureus* strains carried a decreased fitness and did not acquire antibiotic resistance (Daurel et al., 2007). The reported discrepancies on the effect of mutators on evolving populations of asexual singe-cell organisms could be linked to clonal interference, mutation effects, or population fitness or could be the result of the negative selection that mutators face once beneficial mutations have been acquired. Indeed, laboratory evidence showed that while mutator strains were initially selected for, both bacteria and yeast mutants experienced reduced transmission and recolonization abilities as well as rapid fitness decline upon prolonged passaging (Giraud et al., 2001; Trindade et al., 2010). These observations suggest that mutator strains could be counterselected once adaptation to the novel environment is achieved. Accordingly, mathematical modeling of *E. coli* population dynamics showed a sharp decline in the frequency of mutator strains once adaptation was achieved (Taddei et al., 1997). At a molecular level, experimental evolution correlated fitness drop of evolving strains with acquisition of detrimental mutations in genes required for optimal fitness (Andersson and Hughes, 1996; Funchain et al., 2000), suggesting that the mutational load of mutator strains could become a selective pressure itself. Accordingly, decreased cellular fitness after prolonged passaging of *msh2Δ* mutator *S. cerevisiae* strains in non-challenging environments was followed by restoration of genome stability by increasing the buffering ability of heat shock proteins (McDonald et al., 2012). Alternatively, restoration of genome stability arose either by acquisition of antimutator suppressor alleles or by replacing the mutator alleles with functional ones through horizontal gene transfer (Denamur et al., 2000; Wielgoss et al., 2013). Taken together, all of this evidence suggests that while the mutator phenotype is initially selected for during adaptation, it could be selected against once adaptive mutations are fixed.

#### Evidence From Cancer Evolution

Intra- and inter-tumor heterogeneity has been observed in early (Jamal-Hanjani et al., 2017) as well as in advanced stages of tumor progression (Caswell and Swanton, 2017; Jamal-Hanjani et al., 2017). Its presence suggests that cancer proceeds through a branched evolutionary pathway (Nowell, 1976; McGranahan and Swanton, 2017). Specifically, single-cell–derived clones carrying different genomes, epigenomes, and karyotypes compete in a non-linear model that favors the expansion and the coexistence of clones containing distinct beneficial mutations under challenging environments (Merlo et al., 2006). The presence of cell-to-cell heterogeneity promotes cancer progression (Gerlinger et al., 2012; Loeb, 2016; McGranahan and Swanton, 2017) by potentially increasing the number of clones with penetrant driver mutations (Jamal-Hanjani et al., 2017), with resistance to drugs or poor environmental conditions (Scalerandi et al., 1999; Calcagno et al., 2010), or immune to interaction with host immune cells (Seliger, 2005).

Cancer cell populations evolve as asexually reproducing organisms and can be modeled as bacteria or mating-type locked laboratory yeast strains. As discussed above, studies on mutator populations in these model organisms indicate that while a mutator phenotype can initially promote adaptation to a variety of selective pressures, it has detrimental effects once adaptive mutations have been fixed. Does this dynamic also occur during cancer progression? We would like to propose this to be the case. In recent years, the tumultuous advances of deep-sequencing technologies increased our ability to perform and analyze large single-cell sequencing data sets (Gerlinger et al., 2012). By looking at mutations present in cancer cells in spatially distinct regions at different stages of cancer progression, a few lessons have emerged. First, consistent with a positive contribution of genome instability to cancer development and evolution in response to challenges, cancer cells display a high level of intraand inter-tumor heterogeneity (McGranahan and Swanton, 2017). Experimentally, several mouse models of genome instability display an increased spontaneous incidence of cancer onset (Liu et al., 2007; Weaver et al., 2007) and increased tumor relapse when challenged by oncogene withdrawal (Sotillo et al., 2010). This suggests that mutator phenotype could increase aggressiveness or drug resistance of cancer cells. Second, early evolution stages of different types of cancers display genome instability. Thanks to the long latency and frequent biopsies patients are subjected to, one of the cancer types that undergoes the most frequent longitudinal sampling is Barrett's esophagus. This neoplastic lesion frequently gives rise to esophageal adenocarcinoma and is associated with a high level of genomic instability (Reid et al., 2001). Consistent with the idea that mutator phenotype is an enabling characteristic of tumor development, heterogeneity of premalignant Barrett's esophagus populations is a prognostic marker that correlates with increased probability of esophageal adenocarcinoma development (Maley et al., 2006). Another line of evidence that a mutator phenotype is an early event comes from clinical evidence that mutations in mismatch repair genes and genome instability occur early in HNPCC and colon cancer evolution. For instance, microsatellite instability was found in premalignant adenomas (Shibata et al., 1996), consistent with the idea that mutations in mismatch repair genes occur prior to hallmark mutation markers for colon cancer (Huang et al., 1996). Lastly, mathematical modeling favors a positive contribution of mutator phenotype in early events of cancer progression leading to rapid tumor growth (Beckman, 2009). Therefore, similarly to yeast and bacteria adaptation to hostile environments, the mutator phenotype can facilitate early stages as well as later stages of cancer evolution. Consistently, it was recently shown that metastatic cells have higher mutations rates than non-metastatic cancer cells (Bertucci et al., 2019). However, extreme genomic instability was reported to have a negative effect on tumor growth, leading to massive cancer cell death (Kops et al., 2004; Janssen et al., 2009). Similarly to what observed in model organisms, it was proposed that excessive mutational burden decreased cellular fitness as cells cannot tolerate high levels of genome instability (Komarova and Wodarz, 2004). Accordingly, clinical evidence suggests that high levels of chromosomal instability are a marker for better prognosis than intermediate ones in nonsmall-cell lung carcinoma (Jamal-Hanjani et al., 2015). Similar observations have been made in other epithelial tumors, such as ovarian and squamous non-small-cell lung cancer and gastric adenocarcinoma (Birkbak et al., 2011). Taken together, all of this clinical evidence suggests that cancer cells, pretty much like mutator yeasts, can evolve adaptive mechanisms to decrease the rate of genome instability once fitter and more aggressive cancer clones have emerged. This view is also supported by recent studies showing that at different stages of tumor progression, cancer cells exhibit distinct types of genome instability (Nik-Zainal et al., 2012). For instance, sequencing of spatially distant clear-cell renal carcinoma masses within patients showed that, while the bulk of the primary tumor was stable and diploid, cells from metastatic regions derived from a tetraploid intermediate and were genomically unstable (Gerlinger et al., 2012). Moreover, phylogenetic reconstruction of breast cancer tissues carrying BRCA mutations showed that while early mutations during cancer development were consistent with patients' germline mutations, late-stage genome instability had a significantly different mutational pattern consistent with localized hypermutation with specific base substation (Nik-Zainal et al., 2012). Taken together, all of this clinical evidence suggests that different types of genome instability of tumor cells can be selected to better adapt to cycles of selective and non-selective environments as well as different selective pressures.

#### FUTURE PERSPECTIVE

As speculated above, to better adapt to a variety of different selective and non-selective environments, cancer cells could tinker with their genome instability to either generate cell-to-cell heterogeneity or stabilize fitter clones or change their mutational

#### REFERENCES


landscape. Since different types of mutations could allow cells to differently hike the fitness landscape (Pavelka et al., 2010), the ability of cancer cells to switch between different mutational patterns could equip them with different "gears" to successfully adapt to challenges. Therefore, to successfully eradicate cancer cells, strategies to curb their incredible genome plasticity should be found. Given the similarity in the evolution of mutator phenotypes between single-cell model organisms and cancer cells, we predict that dissecting the molecular mechanisms that allow yeast or bacteria to fine-tune their genome instability will pinpoint targets to curb cancer genome plasticity.

#### AUTHOR CONTRIBUTIONS

FN and GR conceived and wrote the manuscript.

#### FUNDING

This study is funded by the NRF Investigatorship (ref. no. NRF-NRFI05-2019-0008) awarded to GR.

#### ACKNOWLEDGMENTS

We thank Dr. Sebastien Teissier and Realite Design for assisting with Figure 1.


cancer stem cell characteristics. *J. Natl. Cancer Inst.* 102 (21), 1637–1652. doi: 10.1093/jnci/djq361


**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 Natali and Rancati. 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.*

# Drosophila melanogaster: A Model Organism to Study Cancer

Zhasmine Mirzoyan<sup>1</sup> , Manuela Sollazzo<sup>2</sup> , Mariateresa Allocca<sup>1</sup> , Alice Maria Valenza<sup>3</sup> , Daniela Grifoni <sup>2</sup> and Paola Bellosta1,2,3,4 \*

<sup>1</sup> Department of Cellular, Computational and Integrative Biology, University of Trento, Trento, Italy, <sup>2</sup> Department of Pharmacy and Biotechnology, University of Bologna, Bologna, Italy, <sup>3</sup> Department of Biosciences, University of Milan, Milan, Italy, <sup>4</sup> Department of Medicine, NYU Langone Medical Center, New York, NY, United States

Cancer is a multistep disease driven by the activation of specific oncogenic pathways concomitantly with the loss of function of tumor suppressor genes that act as sentinels to control physiological growth. The conservation of most of these signaling pathways in Drosophila, and the ability to easily manipulate them genetically, has made the fruit fly a useful model organism to study cancer biology. In this review we outline the basic mechanisms and signaling pathways conserved between humans and flies responsible of inducing uncontrolled growth and cancer development. Second, we describe classic and novel Drosophila models used to study different cancers, with the objective to discuss their strengths and limitations on their use to identify signals driving growth cell autonomously and within organs, drug discovery and for therapeutic approaches.

Keywords: Drosophila cancer modeling, cancer biology, oncogene, tumor suppressor, tissue growth, signaling, metabolism, therapeutic approaches

#### INTRODUCTION

The fruit fly, Drosophila melanogaster, is used as a model organism to study disciplines ranging from fundamental genetics to the development of tissues and organs. Drosophila genome is 60% homologous to that of humans, less redundant, and about 75% of the genes responsible for human diseases have homologs in flies (Ugur et al., 2016). These features, together with a brief generation time, low maintenance costs, and the availability of powerful genetic tools, allow the fruit fly to be eligible to study complex pathways relevant in biomedical research, including cancer. Indeed, publications that use flies to model cancer have exponentially increased in the last 10 years, as shown in the graph of **Figure 1**, suggesting the relevance of this model to cancer research.

In this review we first describe the basic biological mechanisms responsible for uncontrolled growth conserved between humans and flies. We placed a particular emphasis on the characterization of epithelial tumors from most studied models (gut and brain), to novel approaches for studying tumor-induced angiogenesis, prostate, thyroid and lung cancers, with the goal to discuss their strengths and limitations. In the second part, we analyze few physiological mechanisms that uncover potential non-autonomous mechanisms controlling growth, including the relation between the immune cells (macrophages) and the growth of epithelial cells, or the function of lipid metabolism in cancer growth. Finally, we discuss how Drosophila models are used to find novel interesting therapeutic approaches.

#### Edited by:

Maria Grazia Giansanti, Consiglio Nazionale Delle Ricerche (CNR), Italy

#### Reviewed by:

Alysia Vrailas-Mortimer, Illinois State University, United States Fiammetta Vernì, Sapienza University of Rome, Italy

> \*Correspondence: Paola Bellosta paola.bellosta@unitn.it

#### Specialty section:

This article was submitted to Genetic Disorders, a section of the journal Frontiers in Genetics

Received: 19 September 2018 Accepted: 21 January 2019 Published: 01 March 2019

#### Citation:

Mirzoyan Z, Sollazzo M, Allocca M, Valenza AM, Grifoni D and Bellosta P (2019) Drosophila melanogaster: A Model Organism to Study Cancer. Front. Genet. 10:51. doi: 10.3389/fgene.2019.00051

**18**

### PROPERTIES OF EPITHELIAL CANCER CELLS

Cancer cells are characterized by unrestrained proliferation that results from defects in signaling driving cellular growth, apoptosis and changes in metabolic pathways. At cellular level, the hyperproliferative status of cancer cells is mainly due to the activation of growth signals induced by proto-oncogenes (e.g., the RAS/RAF/MAPK axis), which function downstream of receptor signaling cascades, and are deregulated in 25% of human tumors (Samatar and Poulikakos, 2014). Tumor cells escape the anti-proliferative effect of tumor suppressor genes, such as RB (retinoblastoma-associated) and TP53 genes (Duronio and Xiong, 2013), through mutations in these genes, which result in uncontrolled growth (Hanahan and Weinberg, 2000, 2011; Hariharan and Bilder, 2006). Apoptotic cell death represents another physiological mechanism to maintain cellular homeostasis, and cancer cells have developed strategies to evade apoptosis, i.e., by increasing the activity of anti-apoptotic genes (Bcl-2, Bcl-xL, Bcl-w) and of pro-survival factors (Igf-1, Igf-2) or by downregulating the action of pro-apoptotic genes (Bax, PUMA, Bin) (Hanahan and Weinberg, 2011). Another characteristic of cancer cells is the reactivation of telomerase, present in 90% of human cancers, that allows them to replicate unlimitedly (Kumar et al., 2016).

Cancer cells also exhibit alterations in metabolic pathways that contribute to their survival. Rapidly proliferating cells have a high metabolic rate and suffer from low oxygen conditions (hypoxia). In epithelial tumors, this condition triggers the socalled angiogenic "switch" where the quiescent vascular network is induced to proliferate by the secretion of pro-angiogenic factors, such as VEGF (Vascular Endothelial Growth Factor) and FGF (Fibroblast Growth Factor) (Hida et al., 2018), allowing the formation of new vessels that penetrate into the tumor mass to supply oxygen and nutrients (Carmeliet and Jain, 2011). Cancers cells also exhibit a metabolic switch where they reprogram their metabolism to use an alternative and less abundant anabolic pathway to sustain their growth. In particular they switch from oxidative phosphorylation to anaerobic glycolysis, where glucose is used to produce lactate, through a process called the "Warburg effect" (Pavlova and Thompson, 2016; Vander Heiden and DeBerardinis, 2017). This metabolic switch is not yet completely characterized but is supported by the activation of oncogenes, including Myc that also activates glutaminolysis to fuel the TCA cycle with anaplerotic reactions to produce the intermediates necessary for cellular biosynthesis (Hsieh and Dang, 2016).

The last stage of tumorigenesis is represented by the invasive and metastatic capabilities of tumor cells to disrupt the apicalbasal cell polarity, a process that is associated with the downregulation of cell-cell contact molecules and the release of metalloproteases (MMP1), lytic enzymes that degrade the extracellular matrix (ECM) allowing tumor cells to escape and colonize an environment that suites them and to acquire new oncogenic properties (Massague and Obenauf, 2016; Lambert et al., 2017). A variety of studies are now focused on how the tumor micro environment (TME), a specific niche composed of fibroblasts, lymphocytes and immune cells, that may shape precancer cells for their progression into cancer cells and it may select the development of metastasis (Massague and Obenauf, 2016). Emergent evidence suggests also a key role for nonautonomous signals released by the cells composing the niche, particularly from cancer-associated fibroblasts (CAFs), that are essential to support the growth of cancer cells in this "new" metabolic environment (Lambert et al., 2017).

# CANCER MODELING IN DROSOPHILA

Most of the signaling pathways controlling cell growth and invasion in mammals have a conserved function in flies allowing their modulation into models that mimic tumor's biology in a simple model organism like Drosophila (Millburn et al., 2016). The combination of genetic screens with the availability of powerful recombination techniques enabled also a rapid characterization of the primary function of conserved oncogenes and of tumor suppressor genes in a whole animal (Sonoshita and Cagan, 2017). In addition, recent studies using Drosophila imaginal discs explored the mechanisms that govern growth in epithelial tumors and their interaction with the local TME and stromal cells, including some steps in the recruitment of the immune cells (macrophages) to the tumor mass (Herranz et al., 2016; Muzzopappa et al., 2017).

# EPITHELIAL TUMORS IN DROSOPHILA

About 90% of human cancers are of epithelial origin (Hanahan and Weinberg, 2000). Epithelial tissues are characterized by a specific cell architecture composed of junctions and apical and baso-lateral membrane domains that are crucial for the maintenance of cell-physiological functions. Loss of cell adhesion and cell polarity, with an increase of cell motility, are indeed characteristic early cancer traits. In this context, Drosophila larval imaginal discs are a monolayer epithelium that is limited apically by a squamous epithelium (peripodial membrane) and, basally to the notum, by a layer of myoblasts embedded in Extracellular Matrix, and constitute a perfect system in which to model the onset of epithelial cancer progression. These larval organs are indeed morphologically and biochemically comparable to mammalian epithelia (Wodarz and Nathke, 2007). Moreover, the prominent signaling pathways that regulate growth in humans are conserved in the fruit fly (**Figure 2**), allowing the use of this animal model to examine the hallmarks of cancer (St. Johnston, 2002). During the last few years, the imaginal wing and eye discs have been used successfully to study tumor growth and invasion, to investigate the function of cancer genes, and to perform chemical screenings (Tipping and Perrimon, 2014). The imaginal discs also represent an excellent model to analyze oncogenic cooperation: thanks to the use of the MARCM system (Lee and Luo, 1999), it is feasible to induce simultaneously in single cells mutations in tumor suppressor genes (e.g., mutations in cell polarity genes and Hippo pathway components and interactors) and oncogenic activating mutations, or to overexpress specific genes (e.g., EGFR, Ras, Myc, Yorki), resulting in tissue overgrowth, alteration of the normal tissue architecture, disruption of the basement membrane, and invasive/metastatic behavior (Brumby and Richardson, 2003; Pagliarini and Xu, 2003; Wu et al., 2010).

# Marks of Alteration in Epithelial Cells

#### Loss of Cell Polarity

Cellular junctions and a proper apical-basal cell polarity are fundamental for the maintenance of epithelial tissue architecture and function. During early cancer stages, tissues lose these properties and cells subvert their normal growth rate and acquire invasive and migratory behaviors (Wodarz and Nathke, 2007; Bryant and Mostov, 2008). In Drosophila, three complexes establish and maintain epithelial polarity: the Crumbs/Stardust/PATJ/Bazooka, the Par6/aPKC (atypical protein kinase-C) and the Scrib/Dlg/Lgl (Scribble/Discs large/Lethal giant larvae) complexes, which are respectively placed at the apical, subapical and baso-lateral membrane domains. Alterations in these proteins provoke continued cell proliferation, loss of differentiation and complete loss of tissue architecture, resulting in neoplastic overgrowth (Bilder, 2004; Grzeschik et al., 2010; Johnson and Halder, 2014). lgl was the first neoplastic tumor suppressor gene discovered in Drosophila and its loss leads to an abnormal growth of the imaginal structures and the larval brain. In addition, lgl mutant tissues, and tissues bearing dlg or scrib mutation, have the ability to form secondary tumors in the thorax, brain, wings, muscles, intestine and ovaries (Woodhouse et al., 1998). The loss of cell polarity impacts cell proliferation through the deregulation of the Hippo (Hpo) pathway, a signaling cascade involved in organ size maintenance (Lu et al., 2010). It is not yet fully known how lgl activity interacts with the Hpo cascade, but it was observed that its downregulation

mimic the growth of epithelial cancer cells and their ability to undergo metastasis in Drosophila have been established by inducing the cooperation between oncogenes (RED) like the active form of Ras (RasV12) together with the loss of function of cell polarity genes (GREEN) (Brumby and Richardson, 2003; Pagliarini and Xu, 2003). Alteration of cell polarity with the downregulation of the SWH (Salvador-Hippo-Warts) pathway, together with RasV12, triggers downstream events, including activation of the MAPK signaling that stabilize Myc protein (Galletti et al., 2009) resulting in robust cellular growth. Activation of the JNK signaling, with the concomitant loss of cell polarity, induces metalloproteases (MMP-1) and confers to the epithelial cells the distinct characteristics of migration and invasion, hallmarks of tumor growth (Uhlirova et al., 2005; Igaki et al., 2006; Ma et al., 2017).

up-regulates cell cycle genes (such as Cyclin E and E2F1) (Grzeschik et al., 2007) and permits the nuclear translocation of Yorkie (Yki), the downstream effector of the Hippo pathway, causing the activation of its target genes, including MYC, that was found to be important for the growth of lgl mutant clones in a competitive environment (Froldi et al., 2010). In humans, two lgl homologs have been discovered, HUGL-1 and HUGL-2, with HUGL-1 rescuing all the defects of the fly lgl mutant (Grifoni et al., 2004). HUGL-1 loss of function has been associated with a series of human malignancies (Schimanski et al., 2005; Grifoni et al., 2007; Lu et al., 2009). Finally, while the human genome encodes for only one homolog of the tumor suppressor scrib, a number of homologs are known for dlg which have been implicated in different types of cancer (Halaoui and McCaffrey, 2015).

#### Growth Signaling

The **Salvador-Warts-Hippo (SWH)** tumor suppressor pathway was discovered first in Drosophila as a regulator of organ size (Pan, 2010; Yu et al., 2015) and later in humans, where it was found to be fundamental in the regulation of cancer growth (Harvey et al., 2013). The physiological activation of the Hippo (HPO) kinase, (MST1/2 in human) (Harvey et al., 2003) consists in the phosphorylation of Warts (WTS), (LATS1/2 in human) (Genevet et al., 2010; Yu et al., 2010) and in the activation of the phosphorylated core complex, that includes Salvador (SAV in human) (Tapon et al., 2002) and Mob/MATS, that in turn, phosphorylate Yki (YAP/TAZ in humans) (Oh and Irvine, 2008). Phosphorylated Yki is sequestered and degraded in the cytoplasm, resulting in the inhibition of its nuclear transcriptional activity and oncogenic function (Harvey et al., 2013). Upstream, the Hippo cascade is regulated by components of cell junctions, including cell adhesion molecules such as Merlin, a homolog of the human Neurofibromatosis Type 2 (NF2) (Genevet et al., 2010; Yu et al., 2010), which acts as tumor suppressor; the cadherin Fat in complex with Dachsous; and by cell polarity regulators such as Crumbs (Robinson et al., 2010; Harvey et al., 2013). Alterations in the composition of the core proteins (HPO, WTS, SAV, MATS) of the pathway trigger Yki translocation into the nucleus that binds tissue-specific partners and induces the expression of its target genes, among them: CyclinE, dIAP1 and MYC (Harvey et al., 2003; Pantalacci et al., 2003; Neto-Silva et al., 2010; Ziosi et al., 2010). This articulated system is also tightly regulated by other signaling pathways: for example, in the Drosophila imaginal wing disc, Lgl or aPKC deregulation results in JNK activation to promote Yki nuclear translocation via phosphorylation of Ajuba (Jub), an upstream regulator of the cascade that binds to and inhibits Wts kinase activity (Sun and Irvine, 2013). In addition to the regulation of cellcell interaction signals, components of the Hippo pathway have been found to be sensitive to mechanical stress (Panciera et al., 2017). This mechanotransduction function is critical in the control of physiological pathways, and its deregulation may contribute to the abnormal cell behavior in diseases such as cancer, where the cells in the tumor have to sustain physical forces generated by tissue overgrowth. Interestingly, this last function has shown differences in the behavior of Yki between human and flies: indeed, in Drosophila the Yki protein does not respond to integrin stimulation, while in mammals integrin signaling promotes YAP/TAZ activity. One possible explanation for this different behavior may be that the Nterminus of Yki is missing a domain necessary to bind PDZcontaining proteins, which is found in its human counterpart YAP, and is necessary for the activation of the integrin-Src adhesion branch of the pathway (Elbediwy and Thompson, 2018). However, an interesting and potential explanation for this difference comes from a comparative analysis of the Yki protein and the evolution of the different epithelia: this analysis outlines how in Drosophila the apical membrane of the columnar epithelium is well differentiated in its function to activate the Hippo pathway, whereas in mammals the multilayer of cells lacks a functional apical domain, and the activation of YAP/TAZ relies on the activation/signal from the integrin adhesion pathways of the stem cells on the basal layer of the epithelium (Elbediwy and Thompson, 2018).

The **RAS/RAF/ERK** signaling cascade is one of the most conserved pathways in all organisms, including Drosophila. This pathway is part of the MAP kinase signaling that, in addition to ERK1/2, also includes JNK1/2/3, p38/MAPK, and ERK5, which mainly respond to stress activators (Morrison, 2012). Highly conserved in flies, ERK1/2 are activated by growth factors such as EGF or FGFs. These ligands bind to receptor tyrosine kinases (RTKs) to activate downstream signaling, in particular its core complex, which is represented by the guanidine exchange factor Son of Sevenless (SOS) that, in turn, activates the small G proteins RAS on the cell membrane. This leads to RAF activation and to the formation of the complex with the kinase D-Sor also called MAPKK or MEK that, upon phosphorylation of Rolled, the fly homolog of MAPK or ERK kinases, induces the activation of its final targets (Shilo, 2014). ERK in flies has much fewer targets than those described in vertebrates, the most common being the ETS-domain protein Pointed (Pnt). In particular PntP2, needs to be phosphorylated for its activation and is the principal activator of transcription downstream of many RTKs, and PntP1 is transcriptionally induced by MAPK (Shilo, 2014). A second transcriptional repressor is Capicua (Cic), an HMG box-containing protein highly conserved in vertebrates (Simon-Carrasco et al., 2018). Interestingly, in the last couple of years, this protein was found to possess oncogenic properties and be overexpressed in many tumors (Simon-Carrasco et al., 2018). In addition, Cic activity regulates co-target genes upon Yki activation, placing this protein at the crossroads of RTKs and SWH pathways (Simon-Carrasco et al., 2018).

Even though MAPK targets in Drosophila are less abundant than in mammals, its activation and translocation to the nucleus results in a growth phenotype mimicking a few characteristic steps of growth in tumor cells (Brumby et al., 2011). Activation of Ras is considered a cancer distinctive trait both in Drosophila and humans, and it represents one of the strategies to model human cancer in flies. In Drosophila there are three Ras genes but only Ras1 has functional homology with mammalian RAS. In the epithelial cells of the wing imaginal disc, Ras1 activation triggers hyperproliferation but also determines cell fate (Prober and Edgar, 2000). Ras activation is at the crossroads of other growth factor signaling cascades: recently, a link to Hpo function was shown in Drosophila epithelial cells, where Ras activation was able to induce the tissues to switch from a pro-differentiative to a pro-growth program by modulating SWH's transcriptional output (Pascual et al., 2017). Ras increases cell proliferation also through the transcriptional regulation of growth factors and their receptors. For example, it helps promote angiogenesis-like mechanisms in tracheal development through secretion of the FGF/EGFR molecules (Petit et al., 2002; Grifoni et al., 2015); its activation stabilizes pro-growth signals including MYC (Prober and Edgar, 2000), and inhibits pro-apoptotic molecules like Hid (Bergmann et al., 1998). Because of all these functional homologies to human RAS, its activation in Drosophila is considered a good method to establish models that mimic tumor growth.

The **JNK Signaling Pathway** is activated mainly by oxidative stress, producing reactive oxygen species (ROS), and by Eiger, the Drosophila homolog of TNF-α. Its function is variable and depends also on the cellular environment: it can indeed induce cell proliferation and migration, but its major role is to induce apoptosis (Igaki, 2009). The signaling core is characterized by Hemipterus/Hep (JNKK) (Glise et al., 1995), Basket/Bsk (JNK) (Stronach, 2005) and the AP-1 complex, that functions as negative feedback by up-regulating the expression of the Puckered phosphatase (Martin-Blanco et al., 1998). The AP-1 complex is composed of Fra (Fos-Related Antigen) and dJun (Drosophila Jun) and is the final effector of the cascade (Kockel et al., 2001). Upstream Hep is phosphorylated by many JNKK kinases (Tak1-12, Mekk1, Ask1, Slpr) and can also be activated by different indirect stimuli (e.g., RAS, JNKKKK/Msn, and Eiger). Cell death is induced by the expression of the proapoptotic genes hid, reaper and grim, whose activity inhibits the pro-survival protein dIAP1 (Weston and Davis, 2007). In Drosophila cancer cells, the JNK pathway plays a dual role, by suppressing or promoting growth depending on the context (Brumby and Richardson, 2003; Uhlirova et al., 2005; Cordero et al., 2010). lgl, scrib, and dlg mutant cells undergo JNK-mediated apoptosis resulting in a mechanism of tumor suppression (Brumby and Richardson, 2003; Uhlirova et al., 2005; Igaki et al., 2006). On the contrary, in tumor cells with active RAS, apoptosis is blocked and JNK signaling acts as a tumor promoter transcribing genes involved in growth and invasion such as MMP1 (Igaki et al., 2006; Uhlirova and Bohmann, 2006). The overexpression of activated RAS together with Hep (rasv12hepwt) gives cells invasive and metastatic abilities, highlighting how these pathways converge to induce transformation in epithelia.

The **PI3K/Target of rapamycin (TOR)** signaling pathway is a highly conserved key regulator of growth. The binding of insulin-like peptides (ILPs) (fly's insulin) to the receptor (InR) results in the phosphorylation of chico/IRS1-4, and the production of phosphatidylinositol-3, 4,5-triphosphate (PIP3) by PI3K, a reaction that is counteracted by the lipid phosphatase PTEN (Grewal, 2009). PIP3 recruits several Ser/Thr kinases to the plasma membrane, including Akt/PKB and PDK1 (3′ phosphoinosite-dependent protein kinase-1), while its activation results in the inhibition of Glycogen Synthase Kinase-beta (GSK3-β), a conserved kinase that not only controls energy metabolism by inactivation of Glycogen Synthase, but also regulates Wnt signaling by controlling β-catenin/armadillo (Xu et al., 2009) and Myc stability (Bellosta and Gallant, 2010). Activation of Akt also inhibits Tuberous Sclerosis Complex 1 and 2 (TSC1/2), a tumor suppressor binary complex that negatively regulates Rheb, a GTPase upstream of TOR kinase responsible for the activation of TORC1. TOR is found in two complexes: TORC1, which includes Raptor and LST8 adaptor molecules, is sensitive to amino acids and is inhibited by rapamycin; and TORC2, that is composed of LST8 and Rictor adaptor molecules, and does not respond to amino acids or rapamycin (Saxton and Sabatini, 2017). Activation of TORC1 results in phosphorylation of ribosomal protein kinase p-70- S6 (S6K) and of eukaryotic translation initiation factor 4Ebinding protein-1(4E-BP1), thereby triggering protein synthesis and initiation of translation. Insulin and TOR activities are also balanced by a negative feedback mechanism that is activated when S6K is hyper-activated to counteract insulin activity. Under this condition, S6K phosphorylates IRS1-4/chico triggering its internalization and subsequent proteasomal degradation. This feedback mechanism is reduced in pathological conditions, such as the Tuberous Sclerosis Complex syndrome (TSC), where cells carrying tsc1 or tsc2 mutations display an abnormal increase in size and exhibit constitutive phosphorylation of S6K (Saxton and Sabatini, 2017). As members of PI3Ks and TOR signaling are frequently activated in human tumors, they are attractive targets for cancer treatment.

#### Myc and Cell Competition

MYC is one of the most studied oncogenes, and its misexpression is associated with various tumor types including meningioma, Burkitt's lymphoma, medulloblastoma and hepatocellular carcinoma (Hsieh and Dang, 2016). Drosophila Myc is the sole fly member of the family of transcription factors that in mammals is composed of three genes (N-, L-, and c-MYC) (Gallant et al., 1996; Schreiber-Agus et al., 1997). Hypomorphic alleles of myc in flies are developmentally delayed and show a reduction in cell size resulting in smaller flies (hence the name of the mutant as diminutive = small) (Johnston et al., 1999), while null mutants die during larval stage (Pierce et al., 2004). Notably, ubiquitous expression of myc increases cell mass resulting in enrichment of genes encoding components of the nucleolus and of the ribosome; this evidence, concomitantly with Myc's ability to indirectly stimulate RNA pol I and III (Grewal et al., 2005; Hulf et al., 2005; Orian et al., 2005), contribute to revealing its role in the control of ribosomal biogenesis, thus mass and size. Myc activity is finely regulated, and while its expression is required at physiological levels during development, an excess of its activity triggers autonomous cell death and unbalanced growth (Grifoni and Bellosta, 2015). Therefore, Myc is strictly controlled both transcriptionally and post-translationally, where its protein stability is controlled by phosphorylation events downstream of RAS/ERK and GSK3β kinases with a signaling conserved in flies and mammals (Galletti et al., 2009; Parisi et al., 2011). Myc regulation of the cellular metabolic milieu is highly similar in Drosophila to the regulation found in tumor cells (DeBerardinis et al., 2008), indeed it was shown that in cells undergoing to a metabolic stress (starvation or competitive environment), expression of Myc switched their metabolism to increase glycolysis, glutaminolysis (Parisi et al., 2013; de la Cova et al., 2014; Hsieh et al., 2015), or lipid metabolism to favor survival by inducing autophagy (Parisi et al., 2013; Paiardi et al., 2017). Fascinatingly, these evolutionary functions of Myc to control mass and metabolism, resulted in the selective advantage of growth of epithelial cells described as cell competition and characterized in the monolayer epithelia composing Drosophila's imaginal discs (Johnston, 2014). Briefly, cells expressing Myc create a competitive environment and they grow at the expense of wild-type cells that are killed by non cell-autonomous apoptosis (de la Cova et al., 2004; Moreno and Basler, 2004). Myc cells thus behave as "winners" and they are able to repopulate the space of the dying "loser" cells that are killed by unidentified Myc-dependent mechanisms (Johnston, 2014). Myc-induced cell competition was also shown to be necessary in vertebrates to eliminate unfit cells (losers) during early embryogenesis (Claveria and Torres, 2016). More recently, evidence that sustains a central role for Myc-induced cell competition in the early steps of tumor formation have shown Myc present at high levels in cells surrounding the tumor near dying cells, potentially allowing the winner cells to expand and to eliminate the surrounding wild-type cells, thus establishing the first evidence of Myc involved in a tumor growth competitive environment (Johnston, 2014; Di Giacomo et al., 2017). Another form of cell competition is regulated by cell polarity genes (lgl, scrib, dlg) and by endocytic genes (such as Rab5). Cells mutant for these genes behave as losers and were eliminated by wild-type cells (Brumby and Richardson, 2003; Menendez et al., 2010); notably the expression of oncogenes in those loser clones provided them with super-competitive characteristics, i.e., lgl mutant cells over-expressing MYC send death signals to the adjacent wild-type proliferating cells (Froldi et al., 2010), suggesting the presence of another mechanism of cell competition driven by different growth forces working in combination with cell polarity genes and oncogenic signals.

#### ORGANOTYPIC DROSOPHILA CANCER MODELS

#### Gut Cancer

Similar to mammalian counterparts, the Drosophila adult gut is specialized in the digestion of food, the absorption of nutrients, and for controlling the defense response against infection (Tian et al., 2018). Based on these distinct functions, the Drosophila gut is composed of three parts: foregut, midgut, and hindgut. Among them, the midgut has a distinct architecture that resembles the digestive tract of vertebrates. The epithelium is a monolayer that is replenished by Intestinal Stem Cells (ISCs) that differentiate to either enteroblasts (EB) or pre-enteroendocrine cells (pre-EE), that then differentiate into absorptive enterocytes (EC) or secretory enteroendocrine cells (EE). Thanks to significant similarities in the physiology between the Drosophila gut and the intestine of vertebrates (Apidianakis and Rahme, 2011), Drosophila adult midgut epithelium has been used to study the contribution of signaling pathways (i.e., EGFR, Notch, Hedgehog, and Wg/Wnt) to Intestinal Stem Cells (ISCs) renewal (Jiang and Edgar, 2009; Biteau and Jasper, 2011; Jiang et al., 2011).

In vertebrates, the majority of sporadic cases of colorectal cancer and familial adenomatous polyposis (FAP) cancer syndrome are associated with activation of Wnt signaling (Bienz and Clevers, 2000). In humans, abnormal expression of Wnt in ISCs promotes adenoma formation, while deletions in mouse ISCs of the tumor suppressor adenomatous polyposis coli gene APC triggers the initial step of colon-adenoma formation (Barker et al., 2009), underlying the relevance of both mutations in this malignancy. In Drosophila, loss of the Apc gene, leads to the over proliferation of ISCs in the gut, resulting in loss of epithelial cell polarity, hyperplasia and epithelial overgrowth resembling that of intestinal adenomas induced by the loss of APC (Yu et al., 1999). Remarkably, the over-proliferation of the Apc <sup>−</sup>/<sup>−</sup> cells was rescued by lof mutation of Ras (Wang et al., 2013). On the contrary Apc−/<sup>−</sup> cells expressing an active form of Rasv<sup>12</sup> showed a malignant transformation including loss of cell polarity and invasive phenotype, highlighting the conserved functional cooperation between RAS and APC in controlling proper growth in the gut. In Drosophila, intestinal progenitors mutant for the Apc gene expand at the expense of the surrounding wild-type cells that die by apoptosis; because of this behavior these cells have been defined as "super-competitors" (Suijkerbuijk et al., 2016). Apc mutant cells exhibit higher Yki/YAP activity and increased JNK signaling, that was also detected at the border between Apc−/<sup>−</sup> and wild-type cell; moreover, inhibition of apoptosis prevented Apc mutant cells from further expansion, suggesting that a competitive behavior in these cells is controlling Apc dependent tumor growth (Suijkerbuijk et al., 2016).

The JNK-Wg signaling is important to control the physiology and regeneration of intestinal cells, as ISCs damage leads to an overactivation of the JNK pathway and an increase in Wg ligand (Biteau et al., 2008; Cordero et al., 2012b). Wg activity in the enterocytes (ECs) indirectly drives the expansion of the ISCs by upregulating the JAK-STAT ligands Upd2 and Upd3, acting nonautonomously on ISCs proliferation (Tian et al., 2018). Moreover, activation of Wnt drives Myc upregulation in ISCs leading to non-autonomous upregulation of Upd3 in the ECs (Cordero et al., 2012a). Similarly, loss of Apc1 in the midgut (ISCs) also results in JAK-STAT and EGFR pathway hyper-activation, and their removal suppresses the intestinal hyperplasia resulting from Apc1 loss, revealing an underlying conserved signaling between flies and mammals that controls ISCs proliferation and gut homeostasis (Cordero et al., 2012a).

Another aggressive oncogene that is hyper-activated upon Apc loss, in mouse and human intestinal adenomas is the nonreceptor tyrosine kinase c-Src (Yeatman, 2004). This protooncogene is amplified or activated in more than 20% of human tumors, and its activity has been demonstrated to play a central role in the formation of colorectal cancer (CRC). In mice, expression of c-Src increases in the proliferative progenitor cells of the "cripta" favoring hyperplastic adenoma formation (Cordero et al., 2014). In Drosophila the expression of c-Src orthologs (Src42A and Src62B) induces proliferation of the ISCs cells in wild-type animals, and reduction of their expression is sufficient to inhibit ISCs' hyper-proliferation of Apc mutant cells (Cordero et al., 2014). Notably, these results recapitulate an important part of the function of mammalian c-Src in the progenitor cells of the intestine during homeostasis and adenoma formation, suggesting a conserved role of this gene in flies in controlling proper ISCs proliferation.

Recently, Drosophila was also used to generate multigenic models of colon cancer using data from patients from The Cancer Genome Atlas. Interestingly, the outcomes of these models mimicked important properties of human cancers, and can be explored and used in chemical screens to find new combinations of cancer-relevant drugs (Bangi et al., 2016). Studies, using Drosophila models, to characterize intestinal human pathophysiology, revealed the high conservation between these species of the mechanisms underlaying colorectal tumorigenesis (Christofi and Apidianakis, 2013), and further revealed also the mechanisms that control the processes leading to bacterial-mediated inflammation (Lemaitre and Hoffmann, 2007).

#### Brain Cancer

Meningioma are the most common intracranial tumors (Claus et al., 2005; Rogers et al., 2015) and frequently linked with mutations in the PI3K catalytic subunit p110α isoform encoded by the gene (PI3KCA), or in the v-akt murine thymoma viral oncogene homolog 1 (AKT1) gene. Complex interactions were found between members of the PI3K/AKT/mTOR pathway and MAPK-, JAK/STAT, and Notch-1-mediated pathways that contribute to meningioma progression (El-Habr et al., 2014). Increased risk of meningiomas was associated also with neurofibromatosis type II syndrome, where mutations within the tumor suppressor gene Suppressor of fused (SUFU) was associated with hereditary meningiomas (Aavikko et al., 2012) and with medulloblastomas (Taylor et al., 2002). In Drosophila SUFU regulates Hedgehog (Hh) signaling (Ohlmeyer and Kalderon, 1998), with a similar function in humans, where loss of SUFU results in the aberrant activation of the Hedgehog (Hh) pathway (Aavikko et al., 2012).

Of all glioblastomas, the glioblastoma multiforme (GBM) is the most aggressive form of gliomas, accounting for approximately 50% of all glial tumors (Phillips et al., 2006). In GBM, Notch activity is associated with the control of Glioma Stem Cell (GSC), since its activity regulates asymmetric cell division and Notch unbalanced expression leads to uncontrolled growth and high malignancy (Mukherjee et al., 2016), Several studies demonstrate a role for Notch signal in controlling growth and stem cell maintenance of the brain also in flies (Song and Lu, 2011). Because of its conserved function, Notch pathway is now an important target for therapeutic intervention in brain cancer treatment (Yuan et al., 2015).

The current understanding of asymmetric cell division and its relation to tumorigenesis is largely derived from studies on Drosophila neuroblasts (NBs), where mutation of a single gene, brain tumor (brat), was shown to alter asymmetric stem cell division in larval development, and to generate massive neoplastic growth and enlarged adult brain formed entirely of neoplastic NBs (Caussinus and Gonzalez, 2005; Betschinger et al., 2006). Suppression of brat expression was used to establish a model of glioma stemness in Drosophila, where the upregulation of Notch, induced by reducing brat, was the critical node to maintain self-renewal and proper stemness (Mukherjee et al., 2016). This observation was also confirmed in glioblastomas where the human ortholog of brat, the tripartite motif-containing protein-3 (TRIM3), was shown to be necessary to suppress NOTCH1 signaling and to control stem cell activity during development to reduce tumor growth (Chen et al., 2014; Mukherjee et al., 2016). Glioma stem cells divide asymmetrically under the guidance of cell polarity complexes that control the proper apical and basolateral polarization and cell division, a process that was originally identified in Drosophila and later confirmed for the mechanism driving differentiation in human glia for members of the Hugl-1/Llgl-1 complexes (Prehoda, 2009). We recently developed a neurogenic brain tumor model by impairing asymmetric cell division through the loss of function of lethal giant larvae (lgl) the Drosophila ortholog of Hugl-1, in the type II NBs of the central brain (Paglia et al., 2017). In our model, PI3K activation mimics PTEN loss of function and hampers Lgl localization at the apical membrane by aPKC cortical recruitment (Paglia et al., 2017). These data connect the function of HUGL-1 in the maintenance of glioma stem cells with the loss of function of the tumor suppressor PTEN (Gont et al., 2013) and together with those in glioma (Read et al., 2009) show a conserved function for PI3K and EGFR overexpression in these tumors recapitulating many features of the neurogenic subtype of human glioblastoma. Inhibition of PI3K/Akt activity is currently used as a therapy in GBM (Zhao et al., 2017).

Other brain tumors such as oligodendrogliomas, that account for 10% of all cancers of the central nervous system, are characterized by mutations in the capicua (cic) gene (Bettegowda et al., 2011), a conserved transcriptional repressor that regulates MAPK effector genes downstream of receptor tyrosine kinase (RTK) (Simon-Carrasco et al., 2018). The development of correct animal model also for these tumors will be essential to develop specific treatments that can tackle these different brain tumors in vivo.

#### The Paradigm for Angiogenesis

In the fruit fly, the circulatory system is open, the heart pumps the hemolymph into the body cavities and the exchange of gases takes place directly within the organs (Medioni et al., 2009). Moreover, Drosophila is equipped with a complex branched system of interconnected tubules that is responsible for the oxygen transport, the tracheal system, an organ that is comparable in structure and function to the circulatory system of mammals (Affolter et al., 2009). In Drosophila's epithelia, the induction of clones bearing lgl, RasV<sup>12</sup> mutations identified how tumors are able to recruit vessels to oxygenate the growing mass (Grifoni et al., 2015; Calleja et al., 2016). These tumor cells showed ectopic expression of Bnl (branchless), the Drosophila homolog of Fibroblast Growth Factors (FGFs,), and suffered from oxygen shortage (hypoxia). In addition, it was observed a trans-differentiation of tumor cells into pseudo-tracheal cells with and the formation of new vessels, mimicking human FGF-mediated vascularization in cancer (Grifoni et al., 2015).

Cell under hypoxia condition changes their cellular metabolism to favor growth, particularly in solid tumors (Pavlova and Thompson, 2016; Vander Heiden and DeBerardinis, 2017). Interesting studies in flies showed how reduction of the SCF (Skp/Cullin/F-box)-type ubiquitin ligase Ago, homolog of human Fbw7, increased tracheogenesis through up-regulation of the hypoxia-inducible transcription factor Sima/dHIF and of its target, the FGF ligand Bnl (Mortimer and Moberg, 2013). Fbw7 is known to inhibit tumor growth by targeting proteins to the proteasome pathways, and is mutated in a wide range of primary human cancers, this data suggests that its role as a tumor suppressor may be conserved also in the modulation of HIF-regulated angiogenesis in the tracheal system of the fly (Mortimer and Moberg, 2013). This process of neo-tracheogenesis is now considered a novel cancer

are positive for the junctional marker aPKC (white) and secrete MMP1 (red). The reconstruction along the z-axis shown in the upper part of the magnified images reveals a tubule-shaped structure encircling a lumen, indicating these cells are forming tracheal-like structures.

hallmark in fly, which may help to explore the relation between angiogenesis and tumor growth in humans (Herranz et al., 2016; **Figure 3**).

#### Lung Cancer

Lung cancer is a major cause of death in the world, and the standard therapeutic strategy used is chemotherapy because target therapies only decrease tumor growth and result in high toxicity. Recently, a new Drosophila lung cancer model was developed exploiting the tubular structure of the tracheal network (Levine and Cagan, 2016), and considered functionally and anatomically comparable to the vertebrate airways (Andrew and Ewald, 2010). Both in Drosophila and mammals, airways are formed by interconnected branches that depends on the secretion of Bnl/FGFs by the neighboring cells (Ghabrial et al., 2003; Grifoni et al., 2015). Using a binary system, RasV<sup>12</sup> was ectopically expressed specifically in the tracheal cells while downregulating PTEN, a negative regulator of the PI3K/AKT signaling (Hafen, 2004; Ortega-Molina and Serrano, 2013). As a result, the cells of the tracheal branches over-proliferated to form tumors that ultimately killed the animals (Levine and Cagan, 2016). This model was successfully used in a screen for chemical compounds approved by the Food and Drug Administration (FDA), which resulted in the identification of several compounds able to reduce cell overproliferation and to improve tracheal physiological functions (Levine and Cagan, 2016), further highlighting the strong potential of the use of fruit fly models for cancer-related chemical screens.

### Prostate and Thyroid Cancer

The prostate is an exocrine gland of the male reproductive system responsible for the maturation and production of the seminal fluid, with its activity depending on androgens mostly produced by the testis. During organogenesis, the differentiation of the prostate's epithelium occurs along with that of stroma and depends on the complex coordination of many transcription factors and hormones that control the maturation of the quiescent organ (Toivanen and Shen, 2017). The adult prostate epithelium has a low turnover rate and its hyperplasia characterizes the majority of benign prostatic tumors. On the contrary, adenocarcinoma of the prostate is an aggressive tumor that rapidly progresses to a metastatic stage that can be partially blocked by androgen therapy (Shiao et al., 2016). Studies on flies' male accessory gland revealed many parallels with the physiology of human prostate epithelium (Wilson et al., 2017), i.e., a genetic screen using the Drosophila accessory gland identified genes that promote growth and migration of the secondary cells as homologs of genes expressed in human prostate cancer (Ito et al., 2014).

Like in human prostate, Drosophila's accessory gland presents a secondary layer of epithelial cells that continue to proliferate; this homology allowed the development of models that mimic tumors of endocrine origin, including human prostate and thyroid adenomas (Das and Cagan, 2013, 2018). For example, the multiple endocrine neoplasia type 2 (MEN2) syndrome, is characterized by different mutant-translocations involving the RET genes that result in multiple cancer phenotypes, including pheochromocytoma, parathyroid adenoma and the aggressive medullary thyroid carcinoma (MTC) (Das and Cagan, 2013). A recent study demonstrated that the papillary carcinoma of the thyroid (PTC), also caused by another genomic mutations of RET gene, can be profitably studied using the accessory gland of Drosophila to delineate and understand the mechanisms that characterize PTC in the context of the whole animal, including the relationship between tumor and normal cells in an environment that mimics tumor of endocrine origin in humans (Levinson and Cagan, 2016).

The prostate epithelium is characterized by the abundance of exosomes, microvesicles secreted from the endosomal multivesicular body (MVB) that fuse with sperm to modulate its activity and protect its homeostasis (Wilson et al., 2017). The exosomes are particularly relevant in cancer biology for their implication in tumor progression and survival, since they deliver survival factors, metabolites and miRNAs, that help creating a favorable microenvironment for cancer growth; in addition they also favor drug-resistance by activating mechanisms that favor the elimination of toxic chemicals such as chemotherapeutic products (Ruivo et al., 2017; Namee and O'Driscoll, 2018). Since the accessory gland has a similar structure as the prostate

epithelium, characterized by the abundance of exosomes, it could be an optimal model to better study exosome biology in tumors of endocrine origins.

# LIQUID TUMORS

The signaling pathways regulating blood cell differentiation are conserved from Drosophila to humans (Lebestky et al., 2003; Jung et al., 2005). In addition, fly macrophages originate via selfrenewal from progenitor cells localized in the lymph gland, a specialized hematopoietic organ that can be compared to the hematopoietic stem cell niche of the mammalian bone marrow (Krzemien et al., 2007; Mandal et al., 2007). These similarities with vertebrate hematopoiesis outline the utility of using fly models to elucidate the basic mechanisms of hematopoietic differentiation and homeostasis responsible for severe diseases, including leukemia. Drosophila has already been used to study Acute Myeloid Leukemia (AML), a widespread form of leukemia, and to identify the genes responsible for the disease. AML1 is a transcription factor, responsible for activating myeloid differentiation, which has a counterpart in the fly (Sinenko et al., 2010). In vertebrate tumors, the fusion of AML1 with the repressor ETO inhibits the differentiation of the multilineage progenitor cells, while their proliferation is activated, leading to AML1. Interestingly, AML1 fused with ETO in Drosophila also causes the inhibition of hematopoietic cell differentiation, confirming that the fly is a good genetic model to study the mechanisms that drive leukemia in humans (Osman et al., 2009; Sinenko et al., 2010). Myeloproliferative neoplasms (MPNs) have also been reproduced in the fly through gain-of-function mutations in the JAK pathway, finding a role for the downstream effector of the SWH pathway Yki in priming the expansion of Drosophila blood cells, which undergo malignant behavior following JAK activation (Anderson et al., 2017).

# CANCER AND IMMUNE SYSTEM

Inflammation in tumor development acts as "tug and war" since it may promote survival of tumor cells by favoring angiogenesis, by reducing the natural immune responses and by altering responses to chemotherapeutic agents (Mantovani et al., 2008; Wu and Zhou, 2009). The inflammatory response of cancer cells has been attributed to a response of the immune system to eradicate the tumor, but it can also be seen as a way to provide growth and survival, as inflammation contributes to genomic instability by releasing cytokines and through production of reactive oxygen species (ROS) that may induce genetic and genomic alterations (Negrini et al., 2010). Normal cells detect and repair DNA damage, ensuring the maintenance of the correct number of chromosomes and tissue homeostasis, instead often cancer cells have increased mutation-rates leading to high chromosomal instability (CIN) that triggers aneuploidy and advances tumorigenesis (Negrini et al., 2010). Chromosomal instability is a process conserved also in Drosophila, and it was shown to contribute to the invasive behavior of epithelial cells, with a mechanism called "compensatory proliferation" activated to counteract CIN-induced cell death (Clemente-Ruiz et al., 2016; Benhra et al., 2018).

The mechanisms controlling cancer immune response are somehow conserved also in flies as studies in Drosophila have shown that infiltration of macrophages (called hemocytes) in cancer cells requires the activation of the JAK-STAT, JNK, TNF-α, and Toll/Imd/TLR signaling pathways (Bangi, 2013). Of particular interest is TNF-α that plays an important role in controlling apoptosis and the inflammation processes (Ham et al., 2016). TNF-α in tumors has distinct and overlapping functions to promote tumor growth and proliferation and to activate cell death, functions that are mainly mediated by the activation of TNFR1 that is ubiquitously expressed while TNFR2, mainly expressed in immune cells, is less well understood. Thus these opposite signaling pathways activated by TNF signals depend on the adaptor complexes recruited by the receptors and by the cellular context, and they may create a problem for the development of therapeutic strategies that target TNF signaling in tumors (Ham et al., 2016). In Drosophila the sole TNF-α, called Eiger (Egr), binds two receptors called Wengen (Kanda et al., 2002) and Grindelwald (Andersen et al., 2015), the latter shown necessary for the growth of RasV12/scribble−/<sup>−</sup> tumors (Andersen et al., 2015). An interesting mechanism links the possibility that ROS, induced by stress or local inflammation, triggers Egr expression in the hemocytes, to control JNK signaling, in a phenomenon called Apoptosis-Induced Proliferation (AIP), a sort of compensatory proliferative response of the epithelial cells that responds to cues from local "activated" hemocytes (Fogarty et al., 2016). Other studies highlighted the role of hemocytes in the interplay between inflammation and cancer, i.e., using a classic cancer model that recapitulates the hallmarks of epithelial cancer cells (Rasv12/scribble−/−), it was shown that cancer cells induce hemocyte's recruitment and proliferation in vivo by activating JNK signaling to cause the expression of JAK/STAT cytokines (Pastor-Pareja et al., 2008). Using a similar model it was shown that Egr expression was higher in the hemocytes derived from cancer animals, and that its activity was necessary to stimulate invasive migration of tumor cells (Cordero et al., 2010). On the contrary, Egr acts as a tumor suppressor to drive apoptosis in cancer cells upon activation of Toll/NF-κB signaling by the fat body (adipocytes) in response to the secretion of Egr by the circulating "activated" hemocytes (Parisi et al., 2014). Work using allograft transplantation experiments, identify also a function for the hemocytes in tumor initiation, that is independent on Eiger, but relays rather on the activation by external stimuli (i.e., CIN, abnormal growth) of JNK pathway and on the complex of non-autonomous and autonomous signals between tumor cells and those composing the tumor microenvironment; a similar mechanism has been proposed in vertebrates suggesting a conserved response for JNK signaling in fly to control initial tumor growth (Muzzopappa et al., 2017).

In summary, all these data suggest the existence of conserved mechanisms between the immune and tumor cells in flies that may recapitulate some of the most evolutionary conserved aspects described in cancer cells.

# CANCER AND LIPID METABOLISM, OBESITY

In tumor biology, evidences highlight the relevance of lipid metabolism in influencing tumor growth (Katheder and Rusten, 2017; Weber et al., 2017). In this context, a recent role was identified for adipose triglyceride lipase (ATGL) whereby it hydrolyzes triacylglycerols into fatty acids (FAs) that may act as signaling molecules to induce growth both cell autonomously and in neighboring cells (Walther and Farese, 2012). The contribution of ATGL to cancer growth is controversial, indeed several studies showed that its depletion reduced proliferation in colorectal cancer cells and in non-small-cell lung carcinoma (Ou et al., 2014; Zagani et al., 2015), and in breast and pancreatic carcinoma its upregulation contributed to tumorigenesis (Grace et al., 2017; Wang et al., 2017). On the contrary, lack of ATGL favored pulmonary neoplasia in mice, and in few human tumors ATGL expression was found reduced highlighting the complex role of lipids in tumorigenesis (Al-Zoughbi et al., 2016). Cancer cells activate de novo lipogenesis by upregulation of key enzymes in lipid metabolism, some of which, such as AcetylCo-A Lyase (ACLY), AcetylCo-A Carboxylase (ACC) and Stearoyl-CoA desaturase-1 (SCD), are targets of pharmacological inhibitors to decrease cancer proliferation (Zaidi et al., 2012; Zu et al., 2013; Peck and Schulze, 2016; Stoiber et al., 2018). Recent work associated the mechanism of lypolysis with the induction of autophagy, a mechanism used by the cells to re-cycle part of their cytoplasm or cellular content to survive when nutrients are reduced (Dall'Armi et al., 2013). The relevance in cancer of the link between lipids and autophagy was shown when ATGLmediated lipolysis in a peritumoral area, increased autophagy and tumor survival using a non-autonomous mechanisms (Martinez-Outschoorn et al., 2011; Gnerlich et al., 2013). Interestingly, we observed that Myc in Drosophila induced autophagy in the fat body and this was enough to enhance survival of the whole animals upon starvation (Parisi et al., 2013). We linked this effect with the ability of Myc to increase desat1, a Stearoyl-CoA desaturase-1 (SCD1) key enzyme in the synthesis of lipids, that we found co-expressed with Myc in human prostatic tumors (Paiardi et al., 2017).

Metabolic disorders and obesity are associated with cardiovascular disease and type II diabetes (T2D), however numerous cohort studies reported that overweight people are more likely to develop certain types of cancer including endometrial, breast, liver, and ovarian cancer (Cancer, 2012; Chen et al., 2012; Riboli, 2014; Wang and Xu, 2014; Dougan et al., 2015; Hirabayashi, 2016). Obese people have often increased levels of circulating hormones like insulin that has been associated to higher levels of IGF-1 in colon, kidney, prostate and endometrial cancer (Roberts et al., 2010; Gallagher and LeRoith, 2015). Another hormone, leptin, a cytokine produced by the adipocytes to control satiety in a signaling circuit of the brain, has also been found up-regulated in tissues from obese people, particularly in women post-menopause, and increased levels of leptin have been associated with higher incidence of breast and other tumors (Ray, 2018). The adipose tissue produces pro-inflammatory cytokines including IL-6, IL-8, IFNγ, and TNF-α among others (Scheller et al., 2011; Arango Duque and Descoteaux, 2014), and their over-production in fats from obese, activates the infiltration of macrophages into the adipose tissue inducing a low level of chronic inflammation or adipocyte tissue macrophage infiltration called ATM (Lafontan, 2014; Kuroda and Sakaue, 2017). This low level of inflammation increases the levels of ROS and induces DNA and protein damage that may increase the risk of cancer (Lafontan, 2014; Mraz and Haluzik, 2014). The role of the inflammatory response to combat infection and tissue injury, through the activation of the immune cells, is conserved also in Drosophila's circulating hemocytes (Lemaitre and Hoffmann, 2007), where most of the signals activated in the fat body results also in ROS production (Dionne, 2014; Vlisidou and Wood, 2015). Indeed, we showed, using a genetic model that harbors an inflammation state in the fat body of larvae that mimic ATM, that reduction of ROS, using exogenous anti-oxidants components like flavonoids and anthiocianins, decreased hemocyte's migration and JNK activation in the cells of fat body (Valenza et al., 2018), suggesting that the converging signaling between the fat body and hemocytes on lipid metabolism and ROS/cytokines in response to stress is conserved also in Drosophila.

# CANCER STEM CELLS

Cancer stem cells (CSCs) have more features than tissue stem cells because they are able to initiate the tumor growth and fuel its maintenance and metastasis (Malanchi et al., 2011; Kreso and Dick, 2014). In addition, CSCs are highly resistant to conventional therapy, both radiation and chemotherapy, and they are responsible for the recurrence of disease (Mueller et al., 2009). Since the mechanisms underlying the ability of stem cells to support cancer progression are still unclear, Drosophila is convenient to use as it provides many tools for genetic and molecular investigations. Adult stem cells are required for tissue homeostasis and repair after injury and in adult flies, populations of stem cells are present in the posterior midgut, testis, and ovarian follicle rendering it again a good system to dissect these stem cell programs (Hou and Singh, 2017). Drosophila was used to better understand the functions of the centrosome and microtubule-organizing center (MTOC) in the division of stem cells (Tillery et al., 2018). Drosophila and mammalian stem cells are similar and they are regulated by homologous signals corroborating the use of the fly in this field of tumor biology. CSCs can arise from normal stem cells whose long lifespan favors the accumulation of genetic mutations responsible for the malignant phenotype. The progression from normal progenitors to stem-like cancer cells was first explored in leukemia, although nowadays we know that several solid tumors such as brain, breast, lung and colon cancer originate from cells with stem features (Krivtsov et al., 2006). Several Drosophila models of stem cell tumors are now available, and a drug screening was successfully carried out highlighting several compounds active on the signaling promoting cancer growth (Markstein et al., 2014).

#### DROSOPHILA CANCER MODELS FOR THE IDENTIFICATION OF THERAPEUTIC DRUGS

Therapeutic drug discovery requires chemical screening, a procedure allowing for the identification of potential new drugs. The spread of sequencing, automation, and miniaturization has made High Throughput Screening (HTS) the leading contributor to early-stage drug discovery. HTS consists of random screening of chemicals to find an affinity for a specific protein or biological activity characteristic of a disorder. Once identified in vitro, the compounds need to be validated in vivo to assess efficacy and toxicity during a long and expensive period of drug development. The high throughput assays depend on the existence of a specific target, assuming an in depth understanding of a disease that is not always available. Phenotype screening is an eligible option when the knowledge about the mechanisms underlying a disease process is not well defined. It is a process by which small molecules are screened for their effect on the phenotype in cells, tissue or whole animals, where a more physiological environment better describes the pharmacokinetics and toxicological effects of a drug. The great availability of genetic tools and the low cost of maintenance makes the fruit fly an ideal to model to study human diseases including cancer, in fact the fly has considerably contributed to understand tumor biology.

Chemical screens have been successfully performed in Drosophila for several disorders affecting the central nervous system, kidney and metabolism (Whitworth et al., 2006; Gasque et al., 2013; Hofherr et al., 2016), as well as for a type of thyroid cancer, the multiple endocrine neoplasia type 2A and 2B (MEN2) (Vidal et al., 2005). Regarding cancer, JAK- STAT, APC, Wnt, Notch and other signaling molecules, deeply characterized in Drosophila and shared with humans, are precious for cancer drug development. The availability of Drosophila models for multiple cancer types makes pharmacological screens possible against several drugs that aim to restrict proliferation and metastasis. The identification of anticancer compounds is possible using the adult fly, but also larvae, embryos and cells. The combined effect of anti-cancer drugs with radiation has been investigated in Drosophila larvae, producing similar findings to those observed in human cancer cells (Edwards et al., 2011). Moreover, Drosophila avatars, consisting of patient-specific tumors modeled in transgenic flies, are very promising for personalized medicine. Drosophila and other small model organisms are helpful to quickly analyze the mode of action of several active compounds in vivo, nevertheless mammalian models are indispensable in the successive phase of drug development to define important pharmacokinetic parameters such as absorption, distribution and metabolism.

# DISCUSSION

The communication between tumor cells and their microenvironment is largely implicated in neoplastic growth, hence the substantial difficulty to recapitulate the features of malignant transformation in cellular systems. Cancer research needs in vivo investigations, and the use of model organisms contributes to answer this request. In this review we described most relevant approaches in Drosophila, used to explore cancer mechanisms and therapeutics that contribute to our understanding on tumor initiation and progression. In spite of some limitations, because of the anatomical differences between flies and humans, the use of Drosophila's cancer models has been fundamental to understand some basic processes that regulate human cancers, such as the competitiveness of cancer stem cells (CSCs), the importance of tumor microenvironment, cancer cachexia, drug resistance and tumor-associated vasculogenesis, which was recently found to be functionally conserved in fly's cancer. Additional cancer hallmarks such as genomic instability, resistance to cell death, cell metabolism reprogramming, tumorpromoting inflammation and evasion from the immune system, have been studied and extensively characterized in Drosophila. Finally, although the evolutionary difference between Drosophila and humans certainly represents a restriction to the use of the fruit fly in drug discovery and development, phenotypic screenings have proven relevant to identify potential drugs that would elude the classic screens in the absence of targets. Drosophila is also offering a significant contribution to the investigation of organotypic cancers, since despite the evident differences at the macroscopic level, organ cells and functional units are usually well conserved at the biochemical and structural levels respectively. This conservation allowed to develop thyroid, lung, prostate, gut, brain and blood cancer models starting from the most characteristic genetic lesions found in the same human cancers. These models, as described in the review, are greatly helping in dissecting the contribution of specific molecular pathways to the final cancer phenotype. Given the heterogeneous nature of mammalian solid cancers, new strategies are being developed to decipher cancers at singlecell resolution. The international Drosophila community has always been engaged in the development of novel, sophisticated genetic tools, which allowed in the last 30 years to revolutionate functional gene analysis. For this reason, we anticipate that the use of the fruit fly will move fast into the field of precision medicine, contributing to seminal findings in this new era of cancer research.

# AUTHOR CONTRIBUTIONS

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

# FUNDING

Funding was provided from the Cariplo Foundation 2014–0703 and EHDN 689 for PB and AV, AIRC-IG17252 for DG and MS.

# ACKNOWLEDGMENTS

We apologize to colleagues whose work was not cited in this manuscript due to space limitations. We thank all members of Bellosta's and Grifoni's laboratories and Sheri Zola for their comments on the review.

# REFERENCES


**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 Mirzoyan, Sollazzo, Allocca, Valenza, Grifoni and Bellosta. 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.

# *Drosophila melanogaster* as a Model to Study the Multiple Phenotypes, Related to Genome Stability of the Fragile-X Syndrome

*Valeria Specchia1 , Antonietta Puricella1 , Simona D'Attis1 , Serafina Massari1 , Angela Giangrande2,3,4,5 and Maria Pia Bozzetti1 \**

*1Dipartimento di Scienze e Tecnologie Biologiche ed Ambientali, DiSTeBA, Università del Salento, Lecce, Italy, 2Institut de Génétique et de Biologie Moléculaire et Cellulaire, Illkirch, France, 3Centre National de la Recherche Scientifique, UMR7104, Illkirch, France, 4Institut National de la Santé et de la Recherche Médicale, U964, Illkirch, France, 5Université de Strasbourg, Illkirch, France*

#### *Edited by:*

*Maria Grazia Giansanti, Consiglio Nazionale Delle Ricerche (CNR), Italy*

#### *Reviewed by:*

*Ezio Rosato, University of Leicester, United Kingdom Sergio Pimpinelli, Sapienza University of Rome, Italy Fabian M. Feiguin, International Centre for Genetic Engineering and Biotechnology, Italy*

> *\*Correspondence: Maria Pia Bozzetti maria.bozzetti@unisalento.it*

#### *Specialty section:*

*This article was submitted to Genetic Disorders, a section of the journal Frontiers in Genetics*

*Received: 01 October 2018 Accepted: 11 January 2019 Published: 13 February 2019*

#### *Citation:*

*Specchia V, Puricella A, D'Attis S, Massari S, Giangrande A and Bozzetti MP (2019) Drosophila melanogaster as a Model to Study the Multiple Phenotypes, Related to Genome Stability of the Fragile-X Syndrome. Front. Genet. 10:10. doi: 10.3389/fgene.2019.00010*

Fragile-X syndrome is one of the most common forms of inherited mental retardation and autistic behaviors. The reduction/absence of the functional FMRP protein, coded by the X-linked *Fmr1* gene in humans, is responsible for the syndrome. Patients exhibit a variety of symptoms predominantly linked to the function of FMRP protein in the nervous system like autistic behavior and mild-to-severe intellectual disability. Fragile-X (FraX) individuals also display cellular and morphological traits including branched dendritic spines, large ears, and macroorchidism. The *dFmr1* gene is the Drosophila ortholog of the human *Fmr1* gene. *dFmr1* mutant flies exhibit synaptic abnormalities, behavioral defects as well as an altered germline development, resembling the phenotypes observed in FraX patients. Therefore, *Drosophila melanogaster* is considered a good model to study the physiopathological mechanisms underlying the Fragile-X syndrome. In this review, we explore how the multifaceted roles of the FMRP protein have been addressed in the *Drosophila* model and how the gained knowledge may open novel perspectives for understanding the molecular defects causing the disease and for identifying novel therapeutical targets.

Keywords: FMRP/dFmr1, Fragile-X syndrome, piRNA pathway, DNA damage response, transposon elements, neurological diseases

#### INTRODUCTION

Fragile-X syndrome (FXS, MIM300624) is the most common form of mental retardation in the human population. This affects approximately 1/7,000 males and 1/11,000 females (Hunter et al., 2014), and patients exhibit intellectual disability, autism, hyperactivity, long face, large ears, language delay, hyper arousal anxiety (Johannisson et al., 1987; O'Donnell and Warren, 2002; Santoro et al., 2012) macroorchidism, and malformed spermatids (Johannisson et al., 1987; Slegtenhorst-Eegdeman et al., 1998). The most frequent cause of the syndrome is a CGG trinucleotide repeat expansion (greater than 200 repeats) in the 5′ of the Fragile-X locus in Xq27.3, which leads to the hypermethylation of the gene promoter. The final effect is the transcriptional silencing of the Fragile-X Mental Retardation (*Fmr1*) gene, with a consequent loss of the encoded FMRP protein (Godler et al., 2010).

FMRP is a complex protein that displays distinct motifs: a nuclear localization signal (NLS) and a nuclear export signal (NES), two tandem Tudor domains that are likely involved in protein-protein interactions and/or in the DNA binding, as well as three RNA-binding domains including two KH domains and one Arg-Gly-Gly (RGG) box (**Figure 1**) (O'Donnell and Warren, 2002; Ramos et al., 2006; Santoro et al., 2012). In mammals, FMRP is nearly ubiquitous, but it is heavily expressed in neurons, particularly in the cortex, hippocampus, and Purkinje cells where it regulates specific messenger targets. FMRP is also expressed at high levels in testes. Accordingly, the main effects of the FMRP loss in humans are in the nervous system and in the gonads (Santoro et al., 2012). In neurons, the absence of FMRP may alter the processing, the localization, and/or the translational regulation of mRNAs encoding pre- and postsynaptic proteins. These defects can account for the abnormal maturation of dendritic spines in FXS patients, which are longer, thinner, and denser than the normal ones (Swanger and Bassell, 2011; Bardoni et al., 2012; Maurin et al., 2014), representing the cellular defects underpinning the neuronal dysfunctions characterizing the Fragile-X disorder.

In addition to CGG triplet expansion, different mutations in the *Fmr1* gene, leading to FXS, have been reported. They include deletions and missense and nonsense mutations, which are listed in the Human Gene Mutation Database for FXS1 . Mutations occur all along the coding sequences and affect different domains, which may explain why the FraX patients display common as well as specific defects (Reeve et al., 2008; Santoro et al., 2012; Alpatov et al., 2014; Okray et al., 2015; Suhl and Warren, 2015; Quartier et al., 2017).

Two autosomal homologs of *Fmr1* have been identified in the human genome: the Fragile-X mental retardation autosomal homolog 1 (FXR1) and 2 (FXR2), together with the *Fmr1* gene, form the Fragile-X gene family (Siomi et al., 1995; Zhang et al., 1995). Both homologs encode for RNA-binding proteins, FXR1P and FXR2P, with similar and/or complementary functions to those of FMRP, respectively (Penagarikano et al., 2007; Ascano et al., 2012).

A particular aspect linked to FXS is that individuals with a number of CGG repeats from 55 to 200 present a condition known as premutation and display an increased amount of *Fmr1* mRNA. It was proposed that the symptoms, exhibited by these subjects, are related to the *Fmr1* mRNA overproduction. Males with the premutation are at risk to developing Fragile-X-associated tremor/ataxia syndrome (FXTAS, MIM300623), whereas females with the premutation have an increased probability to develop Fragile-X-associated primary ovary insufficiency (FXPOI) (Amiri et al., 2008; Kronquist et al., 2008; Rossetti et al., 2017).

The function of FMRP has been primarily studied in the nervous system of mammals and *Drosophila*, focusing on its role as a translational regulator acting: either by repressing translational initiation (Schenck et al., 2003; Napoli et al., 2008; Aitken and Lorsch, 2012) or by interacting with the translating ribosomes (Siomi et al., 1996; Tamanini et al., 1996; Feng et al., 1997; Ishizuka et al., 2002; Darnell et al., 2005). It has also been proposed that FMRP may exert its translational control through the miRNA pathway (Siomi et al., 1996; Caudy et al., 2002; Ishizuka et al., 2002; Jin et al., 2004; Xu et al., 2008). Many screenings, aiming at identifying FMRP targets (mRNAs and proteins), contributed to the understanding of the role of FMRP, mainly in the nervous system. Many of these targets are involved in synaptic activity, which may account

<sup>1</sup> http://www.hgmd.cf.ac.uk/ac/gene.php?gene=FMR1

for the FXS phenotypes, such as defects in the development of neuronal architecture and in synaptic dysfunction (Darnell et al., 2011; Ascano et al., 2012).

FMRP regulates the local translation of a subset of mRNAs at synapses following the activation of the metabotropic glutamate receptors (mGluRs) (Huber et al., 2002; Bear et al., 2004; McBride et al., 2005). Deregulation of local protein synthesis is considered a core mechanism in FXS, underlying altered synaptic plasticity and consequent cognitive impairment. The role of FMRP in the regulation of translation was better characterized in the Drosophila quiescent oocyte in which the translation of stored mRNAs is a crucial point for the correct development of embryos (Greenblatt and Spradling, 2018).

Animal models of FXS have been developed in zebrafish, mouse, and rat (Tucker et al., 2004; McBride et al., 2005, 2012; Hamilton et al., 2014). Over the last decades, *Drosophila* has also provided key contributions to further understand the molecular pathways defective in FXS, thanks to the many advantages in the use of this versatile organism (Tessier and Broadie, 2012; Sears and Broadie, 2017; Drozd et al., 2018; Dockendorff and Labrador, 2019). The resulting imprecise excisions provided *Fmr1* alleles that lack dFmr1 expression, a situation comparable to the loss of function mutations observed in FXS patients (Wan et al., 2000). dFmr1 is equally similar to the three mammalian gene products (~35% identity, ~60% similarity) and shows particularly high sequence conservation (~70% identity) in critical domains such as the Tudor/Agenet domain that is involved in DNA binding, the RNA-binding domains, and the nuclear localization signals (Zalfa et al., 2007; Zhang et al., 2007; Xu et al., 2008).

The *Drosophila melanogaster* dFmr1 protein is expressed from embryonic stages to adult, and it is enriched in the nervous system (Morales et al., 2002). In the brain, dFmr1 is highly expressed in the mushroom bodies, the main structure of the brain involved in cognitive functions. dFmr1 highly accumulates in the dendrites and in the axons of Kenyon cells, the intrinsic neurons of the mushroom bodies (**Figure 2A**). Its expression is ubiquitous in the neurons of the adult brain, whereas very low levels have been detected in glial cells (Wan et al., 2000; Zhang et al., 2001; Morales et al., 2002; Coffee et al., 2010). Outside the nervous system, dFmr1 is presented at a high level in larval and adult testes with a strong expression in spermatocytes (Zhang et al., 2004; Bozzetti et al., 2015). dFmr1 is also a component of the polar granules of the embryo where it interacts with other specific proteins present in these structures such as Vasa, Cup, and Hsp83 (Verrotti and Wharton, 2000; Cziko et al., 2009; Pisa et al., 2009; Lasko, 2013).

The *Drosophila* animals that completely lack dFmr1 recapitulate many of the phenotypes exhibited by patients with the Fragile-X syndrome. At the cellular level, mutants present defective neuronal architecture and synaptic function. The neurons of *dFmr1* null mutant animals exhibit abnormally organized synapses in both the peripheral and central nervous systems. The neuromuscular junctions (NMJs) of the *Drosophila* larva are simple synapses that represent a good model to study synaptic plasticity. The lack of *dFmr1* causes pronounced synaptic overgrowth at the NMJs (Zhang et al., 2001; Schenck et al., 2003; Pan et al., 2004). Mutant flies display altered behaviors, such as reduced courtship activity of males and irregular circadian rhythms, like the eclosion timing, even though the mRNAs for the two clock elements Per and Tim are not affected. In addition, *dFmr1* mutants exhibit defects in locomotor activity and an acute impairment of long-term memory (Sehgal et al., 1994; Dockendorff et al., 2002; Morales et al., 2002; Bolduc et al., 2008, 2010).

In the ovary, dFmr1 plays a role in translational regulation (Costa et al., 2005), where it controls germ stem cell differentiation through the miRNA-mediated pathway (Yang et al., 2007) and cell proliferation through the proto-oncogene *cbl* (Epstein et al., 2009).

Interestingly, dFmr1 is also involved in the piRNA pathway in the *Drosophila* gonads as well as in the DNA damage response in *Drosophila* and mouse (Zhang et al., 2012; Alpatov et al., 2014; Bozzetti et al., 2015) These findings provide a direct link between dFmr1/FMRP (from here onward, we will name dFmr1 the Drosophila protein as FMRP the mammalian protein) and genome instability, which may represent the common denominator for the multiple phenotypes described in the Fragile-X syndrome and in animal models for the disease.

In this review, we will predominantly treat the roles of dFmr1 related to the genome instability in the gonads and in the nervous system.

#### THE ROLE OF dFmr1 IN THE piRNA PATHWAY

#### *dFmr1* Mutations Affect the Regulation of the *Crystal-Stellate* System and of the Transposable Elements in the Gonads

In 2015, our group demonstrated, for the first time, the role of dFmr1 in the piRNA-mediated silencing of transposable elements and repetitive sequences in the *Drosophila* gonads (Bozzetti et al., 2015). Piwi-interacting RNAs or piRNAs are small RNA molecules protecting animal germ cells and their somatic precursors from the insertion of transposons and other repetitive elements hence preserving genome stability (Malone et al., 2009; Patil and Kai, 2010; Zhang et al., 2011; Anand and Kai, 2012; Specchia et al., 2017). The genomic clusters that act as sources of piRNAs contain multiple and also defective transposon sequences. Most of the piRNA clusters produce piRNAs from both genomic strands, and the other clusters produce piRNAs only from one genomic strand.

The molecular mechanism underlying the silencing of transposable elements reached a deep level of knowledge following studies performed in the ovaries. Argonaute proteins, belonging to the Piwi subfamily groups (P-element-induced Wimpy Testes or Piwi, Aubergine or Aub, and Ago3), play a crucial role in these processes (Aravin et al., 2007). Aub and Ago3 localize to the nuage (**Figure 2B**), a perinuclear structure found in animal germ cells. Piwi localizes predominantly in the nucleus of both germ and somatic cells of the ovary.

Two pathways for piRNA biogenesis and function have been established the primary and the ping-pong pathways (**Figure 3**) (Aravin et al., 2007; Malone et al., 2009).

In the primary pathway, transcript precursors, arising from specific genomic clusters, are processed into primary piRNAs that are bound by specific Piwi proteins. *Drosophila* ovarian somatic cells use exclusively the primary pathway. In these cells, the process occurs in perinuclear Yb bodies, which are discrete cytoplasmic compartments that take their name from the principal player in the process, the protein Yb, in the somatic cells of the ovary and testis (Szakmary et al., 2009). piRNA factors, such as Armitage, Shutdown, and Vreteno, accumulate in the Yb bodies. Upon the formation of the 3′ end of the precursors by Zucchini, the mature primary piRNAs are loaded onto Piwi, which then enters the nucleus and induces transcriptional transposon silencing (Saito et al., 2010).

Germ cells use predominantly the ping-pong amplification process in which the primary piRNAs are subjected to an amplification loop that increases their amount. In this case, the Piwi subfamily proteins, Aub and Ago3, bind the piRNAs and use the sequence homology to recognize the corresponding transposon transcript. Aub and Ago3 cooperate in the pingpong pathway to amplify the piRNAs (Aravin et al., 2007).

The primary and the ping-pong pathways are also present in *Drosophila* testes (**Figure 2C**). In this tissue, the most abundant piRNAs associated with Aub and Ago3 correspond to the "*crystal*" piRNAs (Aravin et al., 2001, 2003; Vagin et al., 2006; Nishida et al., 2007; Bozzetti et al., 2012). The *crystal-Stellate* system represents the first reported natural case of piRNA-mediated regulation, where the repetitive euchromatic *Stellate* sequences are silenced by the piRNAs produced by the heterochromatic *crystal* locus. *Stellate* and *crystal* are composed of tens to hundreds of copies of repetitive sequences organized in tandem (Livak, 1984; Palumbo et al., 1994; Belloni et al., 2002; Tritto et al., 2003; Egorova et al., 2009; Bozzetti et al., 2012). At the molecular level, the loss of the *crystal* region or the "loss of function" mutations of genes involved in the *crystal-Stellate* regulation, called *crystal-Stellate* modifiers, results in the production of a testes-specific *Stellate* mRNA of 750 bases, coding for the Stellate protein. This results in the formation of needle or star-shaped crystalline aggregates that can be revealed by using a specific antibody (Bozzetti et al., 1995). The phenotype induced by *crystal-Stellate* misregulation has provided an efficient tool to identify several genes involved in the piRNA pathway. The majority of the *crystal-Stellate* modifiers has a role in the silencing of germinal and somatic transposons and participates in the primary as well as in the ping-pong pathway. Interestingly, mutants for these genes affect fertility, at various degrees, both in females and males (Pane et al., 2007; Specchia et al., 2008, 2017; Specchia and Bozzetti, 2009; Bozzetti et al., 2012; Sahin et al., 2016).

Null *dFmr1* mutations affect the piRNA pathway in the gonads and the fertility of males and females (Zhang et al., 2004; Bozzetti et al., 2015). In the mutant testes, the levels of the "*crystal*" specific piRNAs are reduced, leading to the formation of the crystalline aggregates. In addition, dFmr1 was demonstrated to have a role in the piRNA-mediated silencing of both germline and somatic transposable elements (TEs) (Bozzetti et al., 2015). For all these reasons, dFmr1 should be considered as a *bona fide* component of the piRNA pathway, at least in the gonads. More recently, the role in the silencing of TEs was confirmed by the work of Jiang et al. who demonstrated that the expression of selfish genetic elements increases in the ovaries of *dFmr1* mutant females (Jiang et al., 2016).

#### dFmr1 Genetic and Biochemical Interaction With Argonaute Proteins in the Gonads and in the Nervous System

The Argonaute proteins are key players of the small RNA-mediated silencing pathway, being the components of the RNA-induced silencing complex (RISC). By using small RNA molecules, they mediate the post-transcriptional control of repetitive sequences, transposons, and genes in different tissues (Kalmykova et al., 2005; Brennecke et al., 2007; Klattenhoff and Theurkauf, 2008; Zhou et al., 2008; Li et al., 2009; Malone et al., 2009). The *Drosophila melanogaster* genome contains five genes coding for proteins of Argonaute family: Ago1 and Ago2 belong to the Ago subfamily and work in the miRNA (micro RNA) and siRNA (small interfering RNA) pathways. As mentioned above, Ago3, Piwi, and Aub act predominantly in the gonad-specific piRNA pathway (Li et al., 2009; Thomson and Lin, 2009).

Ago1 is commonly associated to the miRNA pathway, but data from our lab assign to this protein an additional role in the piRNA pathway as well. Indeed, Ago1 affects the silencing of the transposons in the gonads of both sexes, is involved in *crystal-Stellate* regulation in the *Drosophila* testis (Bozzetti et al., 2015; Specchia et al., 2017), and localizes at the "nuage" in the subcellular compartment in which other piRNA components localize, at least in testes (Kibanov et al., 2011; Nagao et al., 2011). Accordingly, an Ago1-mediated function was demonstrated to be required for the formation of piRNAs in follicle cells, linking together the two pathways (Mugat et al., 2015). The Ago1 protein, hence, has a promiscuous role in small RNA regulation.

A strong argument supporting the role of dFmr1 in the small RNA-mediated pathways is the finding that dFmr1 interacts with the Argonaute proteins. One of the first evidence was provided by the biochemical interaction of dFmr1 with Ago2 and with the components of the RISC in S2 *Drosophila* cells (Caudy et al., 2002; Ishizuka et al., 2002).

Since this discovery, many efforts were made to clarify the molecular role of FMRP in the RNA-mediated silencing pathways based on the genetic and biochemical interactions with the Argonaute proteins. Almost all the Argonaute proteins of both subfamilies have been connected to dFmr1 in the gonads as well as in the nervous system. We here present the main findings related to the specific role of FMRP in the small RNA pathways in the two tissues, disclosing multifaceted connections.

dFmr1 interacts with Ago1 and with the *bantam* microRNA in the *Drosophila* ovary to regulate the fate of germline stem cells (Yang et al., 2007, 2009). Ago1 was also implicated in terminal dendrites elongation (Lee et al., 2015) and is required for a correct function of dFmr1 at the NMJ (Jin et al., 2004; Bozzetti et al., 2015).

dFmr1 also interacts genetically with Aub, whose overexpression in the germline, as well as in the somatic tissues of the *dFmr1* mutant animals, rescues the phenotypes related to the regulation of transposable elements and to the *crystal-Stellate* interaction mediated by piRNAs (Bozzetti et al., 2015). dFmr1 is widely distributed in the gonads, and it overlaps with Aub at the nuage and at the "piRNA nuage giant bodies" (piNG bodies) (**Figure 4**), a giant structure in the nuage of testes where the piRNA components are located and function (Bozzetti et al., 2015). The biochemical interaction between dFmr1 and Aub, in S2 cells, also supports the data obtained with the genetic experiment (Bozzetti et al., 2015). Aub and dFmr1 were demonstrated also to genetically interact in the larval neuromuscular junctions, as the neuronal overexpression of *aub* rescues the *dFmr1* defective NMJs (Bozzetti et al., 2015). Since the presence of Aub in the nervous system is still debated (see the following paragraphs), it has been proposed that the overexpressed Aub may work by taking on the function of Ago1, a protein that is definitely present and has a well-studied role in the nervous system (Lee et al., 2015).

Another crucial Argonaute interactor of dFmr1 is Piwi. A recent study from Jiang et al. in 2016 reported that dFmr1 the piNG body.

and Piwi are present in the same complex in ovarian extracts and act together in the piRNA-mediated transcriptional silencing on the transposable elements in both somatic and germline tissues of the *Drosophila* ovary (Jiang et al., 2016). *dFmr1* mutations also influence the amount of a specific piRNA regulating the *roo* transposable elements. The N-terminal region of dFmr1, where the Tudor/Agenet domain is present (Ramos et al., 2006; Adams-Cioaba et al., 2010; Bozzetti et al., 2015; Iwasaki et al., 2015), is required for the interaction with Piwi.

Finally, no interaction has been reported between dFmr1 and mammalian FMRP with Ago3, another Argonaute protein that operates in the biogenesis of piRNAs in combination with Aubergine (Li et al., 2009).

#### piRNA-Related dFmr1 Interactors Other Than Argonaute Proteins

The role of dFmr1 in the piRNA pathway is supported by its interaction with other components of the piRNA pathway, including Vasa, which is considered a molecular platform for the key components of the piRNA machinery, the so-called Amplifier complex (Xiol et al., 2014; Specchia et al., 2017). **Figure 4** shows the colocalization of dFmr1 and Vasa at the nuage in testes, in particular at the piNG bodies. Emblematic examples have been described above where the direct interaction with four Argonaute proteins has been reported.

The zinc finger protein RP-8 (Zfrp8) also stands out as a very interesting interactor of dFmr1, even though its role in the piRNA pathway or in the human syndrome is still poorly understood.

Zfrp8 was initially identified for its fundamental role in the lymph glands, the site of larval hematopoiesis in *Drosophila* (Minakhina et al., 2007). In this tissue, Zfrp8 controls cell proliferation. Zfrp8 has also an essential role in follicle cells and in germline (Minakhina and Steward, 2010; Minakhina et al., 2014). This function is conserved during evolution, and the vertebrate Zfrp8 homolog, Pdcd2, is required for stem cell maintenance (Mu et al., 2010; Granier et al., 2014). Zfrp8 genetically interacts with several components of the piRNA pathway in the ovary including *vasa, ago3, spindle-E*, and *squash* (Stapleton et al., 2001; Pane et al., 2007; Li et al., 2009; Lasko, 2013; Tan et al., 2016). In addition, the distribution of Maelstrom, one of the known components of the piRNA pathway, is strongly affected in *Zfrp8* KD (Knock Down) ovaries and in germ stem cell (GSC) clones, in which the Zfrp8 protein had been silenced. The argument that strongly supports the role of Zfrp8 in the piRNA pathway is that its reduction affects the expression of the transposable elements in the ovaries (Minakhina et al., 2014), as also seen in animals, that are mutant for the member of the piRNA pathway. Notably, dFmr1 was found as a component of the Zfrp8 protein complex together with Nufip (nuclear FMRP interacting protein) and Trailer hitch (Tral) (Minakhina et al., 2014). Both these proteins were already identified as dFmr1 interactors: Nufip is one of the known interactors of FMRP in mammals (Bardoni et al., 2003), whereas Tral is a component of the RNP granules in *Drosophila* neurons (Barbee et al., 2006). Zfrp8 may have a role in the early assembly of ribosomes with translational repressors and, as a consequence, influences different processes during oogenesis, including transposons silencing (Tan et al., 2016). Very intriguingly, Hsp83, a known component of piRNA-mediated silencing pathway in the *Drosophila* gonads (Specchia et al., 2010; Gangaraju et al., 2011; Tan et al., 2016), was found in the Zfrp8 complex as well (Tan et al., 2016).

Finally, the TDP-43 protein involved in amyotrophic lateral sclerosis (ALS) also interacts with dFmr1. The physical association of these two proteins in ribonucleoproteic complexes was observed *in vivo*, in an ALS *Drosophila* model, and *in vitro*, in neuronal derived cells. FMRP deficit causes developmental defects and autistic behavior, whereas lack of TDP-43 leads to age-dependent neurodegeneration (Fallini et al., 2012; Yu et al., 2012; Coyne et al., 2015, 2017; Majumder et al., 2016). The unexpected link between TDP-43 and FMRP opens novel perspectives to understand the physiopathological mechanisms underlying these seemingly different pathologies.

#### piRNA and TEs in the Nervous System

Although piRNAs were first identified in the gonads of mouse and *Drosophila* as regulators of transposable elements and repetitive sequences (Girard et al., 2006; Grivna et al., 2006; Vagin et al., 2006; Watanabe et al., 2006; Gunawardane et al., 2007; Nishida et al., 2007; Li et al., 2009; Malone et al., 2009), a specific set of piRNAs was found in the mouse hippocampus and in neuronal cultures (Lee et al., 2011). In addition, Ghildiyal et al. (2008) identified small RNA molecules in *Drosophila* heads displaying features resembling piRNAs (piRNA-like RNA molecules, pil-RNAs) (Ghildiyal et al., 2008). More recently, piRNAs with a role in the regulation of learning-related synaptic plasticity were also identified in the nervous system of *Aplysia* (Rajasethupathy et al., 2012). These discoveries represented the starting point for studies demonstrating the presence of piRNAs in somatic tissues and in particular in the brain of several organisms including *Drosophila* and humans (Baillie et al., 2011; Thomas et al., 2012; Perrat et al., 2013; Reilly et al., 2013; Ross et al., 2014; Weick and Miska, 2014). Furthermore, RNA-seq analyses revealed the presence of thousands of retrotransposonderived piRNA-like molecules as well as the presence of factors, involved in the piRNA biogenesis, such as Mili and Maelstrom in hippocampal mammalian neurons. Mice lacking one or the other protein exhibit defects in locomotor activity and behavior (Matsumoto et al., 2015; Nandi et al., 2016). The presence of piRNAs in the nervous system suggests a role in the transposon silencing and hence in genome stability, which may impact on brain heterogeneity, aging, and also neurological diseases. Using different organisms, it was demonstrated that the deregulated expression of the transposable elements can induce their mobilization, which causes *de novo* insertions in the genome and hence triggers genomic variability in neuronal cells (Muotri et al., 2005; Coufal et al., 2009; Baillie et al., 2011; Evrony et al., 2012; Rajasethupathy et al., 2012; Perrat et al., 2013; Ross et al., 2014; Weick and Miska, 2014; Upton et al., 2015; Jachowicz et al., 2017).

Long-interspersed line-1 element (L1) is the only active element in the human genome (Beck et al., 2011) and can transpose in the neuronal precursor stem cells of the rat hippocampus. The new insertions were found in neuronal protein coding genes (Muotri et al., 2005). Engineered human L1 *in vitro* mobilization was also reported in neuronal precursor cells isolated from human fetal brains and embryonic stem cells. These discoveries strongly suggest that L1-mediated transposition has the potential to contribute to genotypic variation in neurons.

Whole genome sequencing and the analysis of the new insertions of a *gypsy*-construct support the idea that piRNAmediated transposition also triggers cellular heterogeneity in the neurons of the *Drosophila* mushroom bodies, which are considered as the functional homolog of the mammalian hippocampus (Li et al., 2013; Perrat et al., 2013). The mobilization of the TEs occurs in a specific neuronal population, the αβ neurons, which contain a lower amount of Aub and Ago3 compared to the γδ neurons (Perrat et al., 2013), raising the concrete hypothesis that transposition may have a functional role in brain physiology. More recent data, however, do not seem to confirm the correlation between the increment in the expression of TEs and new integration sites in aging (Treiber and Waddell, 2017).

Clearly, the field is still very young and more studies will be required to firmly reach a consensus. However, even though the number of new genomic insertions does not exactly correlate with that expected from the remarkable increment of TE expression, a role of transposition in the nervous system must be considered, due to the growing amount of data on the topic.

Recent reports from many laboratories, conducted in *Drosophila*, in postmortem human tissues and in mammalian cells, support the relation between retrotransposition and neurological disorders (Muotri et al., 2010; Douville et al., 2011; Li et al., 2012; Tan et al., 2012; Rajan and Ramasamy, 2014; Krug et al., 2017; Morandi et al., 2017; Prudencio et al., 2017; Faulkner and Billon, 2018; Guo et al., 2018; Short et al., 2018). Significant examples are reported below. Parkinson's disease (PD) is a neurodegenerative disorder that strongly affects movements. Aging represents a risk factor for the occurrence of sporadic PD (Martin, 2011). piRNAs and piRNA-like molecules are differentially expressed in "induced Pluripotent Stem Cells" (iPSCs) from patients during differentiation (Schulze et al., 2018).

Alzheimer disease (AD) is the neurodegenerative disorder that represents the most common cause of dementia. As a remarkable feature, the analysis of postmortem brains from Alzheimer patients reveals the presence of misfolded proteins, namely the β-amyloid peptide and the Tau protein. In addition, transposable elements are also deregulated in these tissues compared with normal brains and in adult brains of *Drosophila* expressing human Tau protein associated with AD (Qiu et al., 2017; Roy et al., 2017; Guo et al., 2018). Intriguingly, the Tau-induced neurological phenotypes can be partially rescued by manipulating DNA damage response key factors, providing a further link between transposition, genomic instability, and DNA (Guo et al., 2018).

Amyotrophic lateral sclerosis and frontotemporal dementia (FTD) are neurological disorders exhibiting a specific phenotypic spectrum causing dementia and cognitive impairment. They have been associated to a defect in TAR-DNA-binding protein 43 (TDP-43) (Douville et al., 2011; Li et al., 2015; Prudencio et al., 2017). Retrotransposition of one of the peculiar TEs with a functional similarity to viruses possessing also a "capsid," whose name is *gypsy*, has been associated to ALS and FTD in a model expressing human TDP-43 (Krug et al., 2017). Even in the model of ALS, the modulation of DNA damage response (DDR) factors partially rescues the neurological phenotypes as occurs in Alzheimer's disease model.

Finally, Fragile-X-associated tremor/ataxia syndrome (FXTAS) is a progressive neurological disorder associated to the premutation in the *Fmr1* gene reported before (expansion up to 90 RGG repeats in the regulatory region) (Amiri et al., 2008; Kronquist et al., 2008). Transgenic *Drosophila* lines that carry the FXTASassociated expansion exhibit an increased expression of *gypsy*, hence providing the first link between the activation of transposons and neurodevelopmental disorders (Tan et al., 2012).

## DNA DAMAGE RESPONSE AND FRAGILE-X IN *DROSOPHILA* AND MAMMALS

Damage to DNA can arise for different reasons and can generate multiple lesions including single- and double-strand breaks (SSBs and DSBs). These lesions set in motion the DNA repair machine that repairs the damage and prevents massive genome instability. This involves changes in the chromatin structure and cell cycle arrest.

Different factors are sequentially involved in the repairing process like the MRN complex, which is a eukaryotic protein complex consisting of Mre11, Rad 50 and Nbs1 proteins, followed by the ATM kinase, in turn phosphorylating several targets including p53 Chk2, BRCA1, and the key histone variant H2AX in mammals (Lou et al., 2006; Matsuoka et al., 2007; Lavin, 2008; Ciccia and Elledge, 2010). Proteins and processes participating in "DNA Damage Response" (DDR) cascade are conserved during evolution. In *Drosophila*, the majority of the information comes from studies on the meiotic checkpoint in ovaries, whose defects affect the fate of the embryonic dorsal cells (Ghabrial and Schupbach, 1999; Abdu et al., 2002; Staeva-Vieira et al., 2003; Cotta-Ramusino et al., 2011). Females displaying defects in this checkpoint process produce embryos with fused dorsal appendages and the mutations affect the so-called spindle class genes (Gonzalez-Reyes et al., 1997).

Interestingly, piRNA mutants also display defects in embryonic axis specification, which are thought to be a consequence of DNA damage mediated by the activation of transposable elements (Chen et al., 2007; Klattenhoff et al., 2007; Pane et al., 2007 Klattenhoff and Theurkauf, 2008). Mutations in *aub* and in other genes of the piRNA pathway such as *armitage* (Cook et al., 2004), *spindle-E* (Stapleton et al., 2001), *zucchini*, and *squash* (Pane et al., 2007), which belong to the spindle class genes, lead to the accumulation of the H2Av histone variant (Klattenhoff et al., 2007).

#### DDR, Transposons, and Neurological Diseases

DNA lesions have been linked to neuronal decline in aging, oxidative stress conditions, and in neurological diseases (Ferrante et al., 1997; Adamec et al., 1999; Lu et al., 2004; Rass et al., 2007; Dobbin et al., 2013), even though the underlying molecular mechanisms remain poorly understood. Recently, the hyperactivation of the PARP-mediated DNA repair of single-strand breaks has been reported to be associated with neurodegeneration and ataxia in humans and mice (Nouspikel and Hanawalt, 2003; Katyal et al., 2014; Hoch et al., 2017).

As described above, transposable elements represent a considerable fraction of the eukaryotic genome and are regulated by the small RNA pathways, in particular the piRNA pathway. Defects in the small RNA-mediated regulation trigger their activation in the germline and in the somatic tissues of the *Drosophila* gonads, hence generating genome instability (Sarot et al., 2004; Kalmykova et al., 2005; Vagin et al., 2006;

Chen et al., 2007; Pane et al., 2007; Specchia et al., 2010; Piacentini et al., 2014). A strong correlation between transposon mobilization and the DNA damage response also exists in human cells where the insertion of the Line-1 non-LTR retrotransposon depends on the DNA repair machine (Belgnaoui et al., 2006; Gasior et al., 2006). In addition, enhanced L1 mobilization has been reported in ataxia telangiectasia, a neurological disorder due to mutations in the ATM gene implicated in DNA repair (Coufal et al., 2011). These observations, linking the transposable elements and the DNA damage response, have led to the hypothesis that DNA breaks accumulate in piRNA mutants, where the transposons are massively activated (Klattenhoff and Theurkauf, 2008). This opens novel perspectives in understanding the causes of devastating neurological diseases, which, in the long term, will result in better therapeutical targets.

#### DDR Has a Physiological Role in Neuronal Development

Emerging evidence support the hypothesis that activation of the DDR mediated by the double-strand breaks plays a physiological role in neuronal activity, by promoting the expression of the so-called early response genes in mice (Madabhushi et al., 2015). In neurons, the "early-response genes" code for transcription factors that are activated soon after the stimulation and regulate the cellular response by activating the expression of the "late response genes" (West and Greenberg, 2011). The "early" genes play a key role in synapse development and maturation and are hence required for learning and memory (Perez-Cadahia et al., 2011). Madabhushi et al. (2015) demonstrated that DSBs occur after neuronal activity at the transcriptional start sites of the early genes (and are related to the TopoII β activity). This facilitates the rapid response of these genes, whose promoters are bound to the "paused" RNA pol II in basal condition, that is, in the absence of stimuli (Kim et al., 2010). It is interesting to note that RNA pol II pausing is also observed at the promoters of genes that are expressed in response to environmental stimuli, and these genes are targeted by the *Drosophila* "HSP90 chaperone" (Sawarkar et al., 2010). This finding represents an intriguing link among "early" gene activation, HSP90, and DNA breaks.

The activation/movement of the transposable elements in the nervous system may induce genome instability, which in turn could connect DDR machinery and synaptic activity.

#### dFmr1/FMRP Has a Role in the DNA Damage Response

FMRP may have a crucial role in this scenario because it has been related to the DNA damage response. Liu and collaborators demonstrated that *dfmr1* mutant flies display disproportioned cell death, related to DNA breaks and to marked genome instability, upon inducing DNA lesions (Liu et al., 2012). dFmr1 and FMRP had been previously shown to regulate cell cycle progression and differentiation in the germline as well as in the brain (Epstein et al., 2009; Yang et al., 2009; Callan et al., 2010; Papoulas et al., 2010), exerting their function in the early DDR through its Agenet and KH domains (Zhang et al., 2014). Soon after this observation, a result in mouse also supported a role of FMRP in the DNA damage response, regulating H2Ax phosphorylation, BRCA complex formation, and accumulation in embryonic fibroblasts and in mouse spermatocyte (Alpatov et al., 2014). This role is thought to be independent of the canonical function in the translational control of mRNAs involved in the synaptic plasticity (Brown et al., 2001; O'Donnell and Warren, 2002; Bassell and Warren, 2008) and requires FMRP N-terminal Tudor/Agenet domain for its binding to the H3 histone (Alpatov et al., 2014). All these discoveries assign a role to FMRP/ dFmr1 in the DDR cascade, identifying this multifaceted protein as a hub for multiple cellular processes. Clearly, one of the most exciting and difficult features of FMRP is the presence of multiple domains involved in a variety of molecular processes, from the nuclear localization domain, the RNA-binding domains, and the Tudor/Agenet domain. This implies that a single protein has distinct roles depending on its localization in the different subcellular compartments. Future efforts will aim at disentangling the diverse functions of this molecular "Swiss knife" in development and physiology.

#### CONCLUSIONS AND FUTURE PERSPECTIVES

A growing number of studies report the identification of piRNAs, piRNA-related proteins, and piRNA-mediated transposition as key factors ensuring heterogeneity in mammalian neurons. Transposable elements are indeed emerging as novel players in neuronal development, and they may function through the DNA damage response pathway. In parallel, it has been shown that the *Drosophila* ortholog of the Fragile-X gene in humans, *dFmr1*, interacts with 4 of 5 Argonaute proteins in the gonads and in somatic tissues (Caudy et al., 2002; Ishizuka et al., 2002; Bozzetti et al., 2015; Jiang et al., 2016) and plays a role in the piRNA-mediated silencing of the repetitive sequences and transposon in the gonads (Bozzetti et al., 2015; Specchia et al., 2017). **Figure 5** illustrates the potential role of dFmr1 in the protein network involved in genome stability. These discoveries

proteins that are part of the polar granules.

the piRNA-mediated silencing of TEs. Hsp83 is connected to dFmr1 by a dotted line, because its interaction has not yet been demonstrated. Asterisks indicate the

open new perspectives for understanding the role and the mode of action of the dFmr1 protein in genome stability and pave the way to address its role in the piRNA pathway operating in the nervous system.

Key questions need now to be addressed: Does dFmr1 has a piRNA-mediated role in the brain and, if so, does its role in genome stability account for the multiple neurological phenotypes exhibited by *dFmr1* mutants and by the FraX patients? Typically, is the dFmr1 pathway linked to piRNAs involved in synaptic plasticity, learning and memory, and circadian behaviors? Should this role of dFmr1 be exerted in a specific temporal window during development as suggested by recent studies? (Weisz et al., 2015; Doll and Broadie, 2016; Doll et al., 2017).

*Drosophila* represents an attractive model for studying the Fragile-X syndrome and will help to address these questions because of the short generation time; the different types of genetic, cellular and molecular tools available; and the easy phenotype evaluation and rescue. *Drosophila melanogaster* offers a suitable *in vivo* model to prescreen numerous potential therapeutic molecules (McBride et al., 2005; Choi et al., 2010; Kanellopoulos et al., 2012; Hagerman et al., 2014), and clinical trials have been performed in human FraX patients, even though the results are not convincing. If the role of dFmr1 in

#### REFERENCES


the piRNA-mediated regulation of transposons is confirmed in the nervous system as well, new therapeutic possibility will open up. We are confident that dFmr1/FMRP will still surprise us and will help us in searching and finding potential therapeutical targets for the treatment of this devastating disease.

#### AUTHOR CONTRIBUTIONS

MB prepared the initial version of the manuscript. VS, SM, and AG significantly revised the manuscript. All authors provided intellectual contribution, edited, and approved the manuscript for publication in its complete version.

#### FUNDING

We acknowledge the financial support of Telethon-Italy (grant number GG14181) and MIUR (grant number RBFR10V8K6). The laboratory of AG is funded by INSERM, CNRS, UDS, Ligue Régionale contre le Cancer, Hôpital de Strasbourg, ARC, and ANR grants. The IGBMC was also supported by a French state fund through the ANR labex.


Fragile-X mental retardation protein KH2 domain and brain polyribosomes. *Genes Dev.* 19, 903–918. doi: 10.1101/gad.1276805


**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 Specchia, Puricella, D'Attis, Massari, Giangrande and Bozzetti. 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.*

# Divide Precisely and Proliferate Safely: Lessons From Budding Yeast

#### Roberta Fraschini\*

Dipartimento di Biotecnologie e Bioscienze, Università degli Studi di Milano-Bicocca, Milan, Italy

A faithful cell division is essential for proper cellular proliferation of all eukaryotic cells; indeed the correct segregation of the genetic material allows daughter cells to proceed into the cell cycle safely. Conversely, errors during chromosome partition generate aneuploid cells that have been associated to several human pathological conditions, including cancer. Given the importance of this issue, all the steps that lead to cell separation are finely regulated. The budding yeast Saccharomyces cerevisiae is a unicellular eukaryotic organism that divides asymmetrically and it is a suitable model system to study the regulation of cell division. Humans and budding yeast are distant 1 billion years of evolution, nonetheless several essential pathways, proteins, and cellular structures are conserved. Among these, the mitotic spindle is a key player in chromosome segregation and its correct morphogenesis and functioning is essential for genomic stability. In this review we will focus on molecular pathways and proteins involved in the control mitotic spindle morphogenesis and function that are conserved from yeast to humans and whose impairment is connected with the development of human diseases.

#### Edited by:

Maria Grazia Giansanti, Consiglio Nazionale delle Ricerche (CNR), Italy

#### Reviewed by:

Matteo Barberis, University of Amsterdam, Netherlands Ethel Queralt, Biomedical Research Institute of Bellvitge, Spain

#### \*Correspondence:

Roberta Fraschini roberta.fraschini@unimib.it

#### Specialty section:

This article was submitted to Genetic Disorders, a section of the journal Frontiers in Genetics

Received: 16 July 2018 Accepted: 22 December 2018 Published: 10 January 2019

#### Citation:

Fraschini R (2019) Divide Precisely and Proliferate Safely: Lessons From Budding Yeast. Front. Genet. 9:738. doi: 10.3389/fgene.2018.00738 Keywords: SPB, centrosome, mitotic spindle, genomic stability, aneuploidy

# INTRODUCTION

In the last decades budding yeast has been largely used as a model system to unravel molecular mechanisms of cell life. Saccharomyces cerevisiae is a unicellular eukaryotic organism, its cells have the same subcellular organization as those of multicellular organisms but they are easier to manipulate. In the budding yeast system several methodologies can be easily applied: genetic, biochemical, cytological, genomic approaches (Botstein and Fink, 1988), and high-throughput technologies. Several processes and proteins are conserved from yeast to human cells despite their distance from an evolutionary point of view and their misregulation is involved in disease development. The sequence of the genes and functional complementation studies have revealed that at least 20% of human genes known to have a role in disease have functional equivalents in yeast (Douzery et al., 2004). In addition, a systematic humanization analysis revealed that 47% of the yeast genes can be replaced by their human orthologs, indicative of conserved functions despite large sequence divergence (Kachroo et al., 2015).

The importance of basic research using budding yeast is highlight by three Nobel Prizes in Physiology or Medicine assigned to scientist for their discoveries of key proteins or processes in yeast. Professors L. Hartwell, J. Rothman, and Y. Ohsumi used classical forward budding yeast genetic approaches to reveal fundamental cell biological processes in eukaryotes; genetic analysis was conducted in parallel with microscopic analysis of the mutants and followed by biochemical

**49**

and cell biology studies. Importantly, the conservation of the basic cellular processes across eukaryotes explains the great impact of these studies for understanding biology and disease.

Studies in budding yeast have revealed the essential functions of the centromeres, DNA sequences that allow the assembly of specialized multiprotein complexes, called kinetochores, that connect chromosomal DNA with mitotic spindle fibers in a bipolar way and ensure proper chromosome segregation during mitosis. A surveillance mechanism called Spindle Assembly Checkpoint (SAC) has been described in yeast and is conserved throughout evolution; the SAC delays anaphase onset in case of lack of biorientation, that is the correct binding of the two sister chromatids kinetochores to opposite spindle poles. Importantly, altered SAC function allows premature mitotic exit and can cause chromosome missegregation thus leading to aneuploid daughter cells, i. e., with an abnormal chromosome number. Interestingly, both decreased and increased SAC gene expression are found in tumors in mice (Sotillo et al., 2007; Ricke et al., 2011), in addition mutations in MAD1, MAD2, BUB1, BUBR1, and BUB3 are found in human cancers and over expression of the same genes is associated with elevated proliferation index and metastatic potential in several solid tumors (Yuan et al., 2006; Wang et al., 2015). These data link cancer occurrence with SAC function, chromosome segregation defects and aneuploidy (for a complete overview on this topic see Simonetti et al., 2018).

The mitotic spindle is essential to allow proper partitioning of the genetic material between the daughter cells, it has a conserved structure and its malfunctions are at the basis of several human diseases, as described below. In budding yeast a bipolar spindle is formed during S phase, concomitantly with DNA duplication, while in animal cells the spindle apparatus is built during mitosis. However, in all eukaryotic cells the mitotic spindle is formed by microtubules (MTs), cylindric structures made by protofilaments of α- and β-tubulin heterodimers assembled in a head-to-tail fashion. Each MT has a dynamic fast growing end (plus end) and a slow growing end (minus end). MTs are associated with several proteins involved in regulation of spindle dynamics and with motor proteins that allow spindle positioning in the cell and intracellular transport. Before sister chromatids separation in anaphase, MTs plus ends bind the chromosomes via their kinetochores in a bipolar way thus ensuring their correct segregation in the daughter cells.

#### MTOC STRUCTURE AND FUNCTION

Eukaryotic cells contain microtubule organizing centers (MTOC) or centrosomes that allow MT nucleation in coordination with cell cycle progression. In S. cerevisiae the mitotic spindle is build thanks to the spindle pole body (SPB), the functional equivalent of multicellular eukaryotes centrosome. SPBs are approximately 2 megadaltons complexes embedded in the nuclear envelope and are able to nucleate both nuclear and cytoplasmic MTs, thus SPBs play a critical role in mitotic spindle formation and positioning. The SPB has a multilayered structure that consists of an outer plaque that faces the cytoplasm and emanates cytoplasmic microtubules, a central plaque, and an inner plaque that faces the nucleoplasm and emits nuclear MTs (Byers and Goetsch, 1974; **Figure 1A**). Attached to one side of the central plaque there is the half-bridge, essential for SPB duplication as it serves for the assembly of the satellite, the precursor of the daughter SPB. Given the crucial function of SPB, its components are encoded by essential genes. SPBs and human MTOCs share conserved proteins with common functions (Cavanaugh and Jaspersen, 2017), therefore the yeast SPB is good model to study centrosome function.

The SPB cycle is tightly connected with other cell cycle events (**Figure 1B**). The SPB duplicates once each cell cycle, just like chromosomes: during S phase the MTs bind the kinetochores and the SPBs separate from each other, thus allowing the formation of a short bipolar spindle. After separation, each SPB inherits a half bridge, essential for its duplication in the following cell cycle. During anaphase the SPBs further move away from each other toward the cortex of the mother and the daughter cell (Fraschini, 2017). SPB duplication is restricted once per cell cycle thanks to an oscillation between the activities of Cdk1 kinase and Cdc14 phosphatase. Sfi1 is a Centrin/Cdc31 binding protein, conserved from yeast to humans, that plays a key role in SPB cycle. During S phase and early mitosis Sfi1 is phosphorylated by Cdk1 and by Cdc5 and blocks the process of SPB duplication. After anaphase onset, the protein phosphatase Cdc14 is active and dephosphorylates Sfi1 thus allowing the maturation of the half-bridge and therefore daughter SPB formation (Elserafy et al., 2014).

Spindle pole body function is regulated by multiple proteins during the cell cycle and several SPB components have been shown to be phosphorylated in vivo (Keck et al., 2011). A genome-wide screen for the substrates of the cyclin dependent kinase Cdk1 identified some SPB components (Spc42, Spc29, Mps2, Bbp1, Sfi1) and suggested that Spc110, Cnm67, Kar1 may be Cdk1 substrates as well (Ubersax et al., 2003). Mps1 is a protein kinase involved in regulation of SPB duplication and phosphorylates three SPB components: Spc98, Spc110, Spc42. The polo-like kinase Cdc5 localizes to SPBs and over expression of a version of Cdc5 lacking the polo-box results in the formation of Spc42-containing structures in the cytoplasm. Pericentrin/Spc110 might be phosphorylated by Cdc5, since one of its phosphorylation sites matches the Cdc5 kinase consensus sequence and an affinity capture-western experiment showed that Cdc5 and Spc110 interact. The protein kinase Swe1 is an important cell cycle regulator, as it blocks entry into mitosis through inhibitory phosphorylation of the catalytic subunit of the cyclin-dependent kinase Cdk1 (S. cerevisiae Cdc28) in case of replication stress and alterations in actin cytoskeleton or cytokinetic structures (Booher et al., 1993). Interestingly Swe1, homolog of human Wee1, localizes at SPBs (Bartholomew et al., 2001), it is involved in mitotic spindle dynamics and it interacts with the outer plaque component Centrosomin/Spc72 (Raspelli et al., 2015, 2018).

Studies on SPB helped to reveal the protein composition and function of animal cells MTOC, that have a partly different structure. Each centrosome contains a pair of ninefold symmetrical centrioles embedded in a proteinaceous matrix known as the pericentriolar material (PCM), which comprises

proteins required for microtubule nucleation and cell cycle regulators (**Figure 1C**). Concomitantly with DNA replication, centrioles duplicate themselves to form a new centrosome then, during late G2 phase, centrosomes undergo maturation, which results in PCM expansion and recruitment of the conserved γ-tubulin rings. Polo-like kinase 1 (Plk-1) regulates the increase of PCM that accompanies mitotic entry by controlling the localization of γ-tubulin, Spd-2, and Cnn/CDK5RAP2. During prophase centrosomes separate, generate fibers and form a bipolar spindle (Nigg and Holland, 2018; **Figure 1D**).

#### MTOC MALFUNCTIONS AND GENETIC DISEASES

Studies in budding yeast have revealed the role of some proteins essential for SPB function and their homologs are involved in genetic diseases in human (**Table 1**). For example Spc110, the yeast homolog of pericentrin, is an essential component of the SPB inner plaque that stimulates binding with γ-tubulin and interacts with Cmd1, the homolog of human Calmodulin, a calcium sensor involved in the propagation of intracellular calcium signals. Missense mutations in one of the three genes coding for Calmodulin are associated with catecholaminergic polymorphic ventricular tachycardia (CPVT), early-onset severe long QT syndrome (esLQT), and idiopathic ventricular fibrillation (IVF) (Vassilakopoulou et al., 2015). Mutations that inactivate the PCNT gene, that encodes pericentrin, a structural component of the centrosome, can be found in all patients affected by Microcefalicosteodysplastic primordial dwarfism type II (MOPDII) (Ren et al., 2008). MOPDII is a rare and complex human autosomal recessive genetic disease, and individuals affected show primordial growth problems that are present before birth. Since pericentrin is



essential for mitotic spindle organization, mitotic progression and chromosome segregation, loss of its function causes defective recruitment of several proteins to the centrosome and inability to properly assemble microtubules, thus disrupting the mitotic cycle and cell division. These severe mitotic problems cause a dramatic reduction in the number of cells of both the growing embryo and the adult organism, resulting in small head, and body size (Delaval and Doxsey, 2010).

The homolog of Centrosomin (Cnn) is Spc72, an essential component of the SPB outer plaque, that interacts with Nud1, the yeast counterpart of human Centriolin. The SPBs components Cdc31 and Sf1 are the yeast counterparts of Centrin and hSFI1. Autosomal primary recessive microcephaly (MCPH) is a disorder in neurogenesis caused by at least nine genes, six of which encode centrosome components (CEP63, CEP135, CEP152, CDK5RAP2/Cnn, CPAP /MCPH6, and STIL/MCPH7) and two encode proteins associated with spindle poles (ASPM and WDR62) (Gilmore and Walsh, 2013). Mutations in these genes alter the precise centriole duplication process and therefore cause deregulation of centrosome number in cells. A translocation between chromosomes 8 and 9 disrupts Centriolin function and is associated with stem cell myeloproliferative disorder (MPD) (Ren et al., 2013). Idiopathic scoliosis is a complex disease with polygenic background that leads to a spinal deformity. Human POC5 dysfunction is associated with idiopathic scoliosis (Patten et al., 2015), being hPOC5 a protein that binds centrin and that is important for centriole duplication (Azimzadeh et al., 2009) together with Centrin and hSFI1 (Martinez-Sanz and Assairi, 2016).

#### CENTROSOME AMPLIFICATION AND CANCER

Centrosome amplification can cause the formation of multipolar spindle during mitosis, thus resulting in daughter cells with unbalanced genetic material. In yeast, SPB duplication is restricted by precise molecular mechanisms, as described above (Elserafy et al., 2014) and defective spindle or chromosome biorientation are sensed by the SAC, as described in the introduction (Stukenberg and Burke, 2015). The target of the checkpoint is the mitotic exit network (MEN), a pathway that

governs the transition from mitosis to the G1 phase of the cell cycle (Hotz and Barral, 2014). The MEN pathway is essential for proper coordination of nuclear division and exit from mitosis. Key MEN components are localized at the SPBs, that are therefore an important signaling platforms for mitosis progression. The MEN is conserved and in metazoan it is called the Hippo pathway; importantly, recently it has been shown that the core components of the Hippo pathway cooperate with p53 to suppress tumorigenesis (Furth et al., 2018). This cooperation occurs at multiple levels, for example in response to stress LATS2 blocks MDM2, a negative p53 regulator, thus causing p53 accumulation and activation (for a complete overview on this topic see Furth et al., 2018).

In the context of tumorigenesis, centrosome abnormalities and amplification are frequently detected in a wide range of solid cancers, myeloma, lymphomas and leukemias, and have been associated with multipolar cell divisions, chromosomal instability and aneuploidy (Chan, 2011). Cells with less or more than 2 centrosomes can form anastral, monopolar or multipolar spindles that can lead to chromosome missegregation and therefore aneuploidy. Centrosome defects can be found in early-low grade lesions, and are rarely observed in normal tissue, suggesting a possible role in tumor initiation. Recent studies show that centrosome aberrant number causes tumor formation in mice (Levine et al., 2017), centrosome amplification is correlated with high-grade tumors, disease progression and poor prognosis and also it enhances the aggressive nature of already transformed cells (Godinho and Pellman, 2014). Finally, several oncogenes and tumor suppressors have been localized to the centrosomes suggesting that they might contribute to centrosome anomalies (Tang et al., 2011). Centrosome alterations trigger a p53-response that arrest the cell cycle, indeed p53 proficient cells tolerate well variations of centrosomes copy number, while cancer cells defective in p53 frequently display centrosome anomalies (Lambrus et al., 2015).

Extracentrosomes can arise basically through two mechanisms: centriole overduplication or accumulation of mature centrosomes by aborted cell division, cell fusion, or centrosome clustering. It has been shown that the majority of centrosome aberrations in the primary tumor types are due to overduplication. However, in solid tumors other types of centrosome aberrations, originated from centrosome clustering or failed cytokinesis, are also found (Cosenza et al., 2017). Spindle multipolarity is strongly correlated with anaphase bridges, which cause DNA breaks that usually block cytokinesis, thus leading to centrosome amplification.

Centrosome amplification can be due to a deregulation of its duplication cycle that is controlled by many positive and negative regulators, such as members of the Cdk, Aurora/Ipl1, Polo-like, and NIMA families of conserved cell cycle kinases (Brownlee and Rogers, 2013). Some of these kinases are likely hyperactive in tumor tissue, since several centrosomal proteins are hyperphosphorylated in breast tumor cells compared to normal breast tissue (Lingle et al., 2005). Another key player is STIL, that interacts with the Polo-like kinase 4 (Plk4): its depletion leads to a decrease in centriole numbers while its excess activity causes extra centrioles (Habedanck et al., 2005). Cdk inhibitor p27Kip1 levels and localization are highly regulated during the cell cycle and it acts to ensure proper centrosome amplification (Sharma et al., 2012). Recently, p27 involvement in centrosome duplication and cancer has also been studied at the systems levels (Barberis and Verbruggen, 2017), this kind of approach integrates experimental and computational data and allows the prediction of how perturbation of a protein can influence a biological process under analysis.

# SPINDLE ALIGNMENT AND CELL POLARITY

Asymmetry is very important for the life of a cell: during development it drives the cellular fate, indeed in stem cells the asymmetric division discriminates the daughter cell that will differentiate and the other cell that will maintain the ability to proliferate. The asymmetry of a cell is built thanks to the polarization of several factors. In most eukaryotic cells astral MTs emanating from the centrosomes are captured by protein complexes at the cell cortex, align the mitotic spindle to the polarity axis of the cell and drive asymmetric division of the cell (Siller and Doe, 2009). Also budding yeast divides asymmetrically: the daughter cell originates from a bud that emerges from the mother cell, and the bud is the equivalent of the stem cells that retain the possibility to divide, while the mother gets old.

In budding yeast the localization of polarity factors determines the bud emergence site. At the beginning the bud grows in a polarized way, then the growth becomes isotropical and the bud becomes round shaped. The bud neck separates the mother from the daughter cell and it is the place where cells will divide. Since the division site is defined during late G1 phase, before DNA replication and mitotic spindle formation, in order to ensure proper chromosome partitioning during mitosis, the spindle must be correctly positioned and aligned with respect to the mother-bud axis (Lee et al., 2000). Yeast is the first model for which the mechanisms for spindle positioning have been described: an actin dependent and a dynein dependent pathway guide the process. In addition, the spindle orientation checkpoint (SPOC) blocks mitotic exit and cytokinesis in case of spindle mispositioning or misorientation (Caydasi and Pereira, 2012). If the checkpoint fails, cytokinesis occurs even if the nucleus divides into the mother cell, thus causing the formation of aneuploid cells.

The existing SPB is also called old SPB, while the one originated by duplication is called new SPB. Usually, the old SPB migrates into the bud thanks to cytoplasmatic MTs that contact the bud cortex. The two SPBs undergo different steps of regulation, for example differential Kar9 recruitment drives the movement of the selected SPB to the bud neck and helps spindle alignment (Liakopoulos et al., 2003). Several pathways contribute to SPB asymmetry: Kar9 is preferentially recruited to astral MTs emanated from the old SPB and this is governed by the SPB inheritance network (SPIN) and the MEN (Lengefeld et al., 2017). Recent data revealed that the asymmetry of the SPBs is due to spatial cues rather than different maturation

(Lengefeld et al., 2018). Similarly, in animal cells the old MTOC nucleates more astral MTs and is surrounded by more PCM than the new one, indicating that the new one is immature while the old one is mature and fully active (Lerit and Rusan, 2013).

Also in insect cells several data indicate that centrosome inheritance is consonant with cell fate decision. The stem cells of Drosophila male germline divide asymmetrically and produce a cell that differentiate and a cell that is totipotent. It has been observed that the old centrosome migrates in the cell that is able to renew while the new centrosome is inherited by the cell that is going to differentiate (Yamashita et al., 2007). Similar data were obtained in mouse radial glia progenitors and in Drosophila neuroblasts (Januschke et al., 2011), indicating that asymmetry of MTOCs and fate decision is a common feature of eukaryotic cells.

The centrosome plays an important role in brain development, indeed aberrant centrosome behavior is linked to inherited microcephaly. Microcephaly is the result of premature neural differentiation due to an insufficient number of symmetric division of neuroprogenitor cells before differentiation, that starts with the first asymmetric division, during cerebral cortex formation. Proper centrosome segregation ensures correct spindle orientation and the succession of several symmetric cell divisions before the beginning of differentiation (Morrison and Kimble, 2006). Several genes that encode for proteins implicated in centrosome function and spindle orientation are mutated in microcephaly in humans: MCPH5 or ASPM (abnormal spindle-like microcephaly associated), WDR62/MCPH2, and CEP63 (Thornton and Woods, 2009). However, not all forms of microcephaly are linked with spindle orientation defects, indicating that the causes of deficiencies in brain development are still partially unclear.

#### REFERENCES


#### CLOSING REMARKS

The model organism S. cerevisiae offers powerful genetic tools to dissect the molecular pathways that control centrosome structure and number. In budding yeast the genetic analyses and manipulation are simple and fast, it is possible to synchronize cells in different cell cycle phases, to perform genetic screenings, and in addition the yeast two-hybrid assay allows detecting labile protein-protein interactions in the centrosome.

Despite structural differences, the yeast SPB carries many conserved proteins of the centriole and/or centrosome machinery in metazoans. Thus, it is becoming of growing interest to compare the structure and function of SPB with centrosomes and studies in budding yeast can elucidate the role of centrosomal proteins in physiological conditions. Altogether the knowledge provided by the studies on SPB structure and function in budding yeast will also improve our understanding of the molecular basis of important human diseases thus helping in developing new biomarkers and therapies.

#### AUTHOR CONTRIBUTIONS

RF wrote and revised the manuscript.

#### FUNDING

RF researches were supported by grants from PRIN (Progetti di Ricerca di Interesse Nazionale) and from the University of Milano-Bicocca (Fondo di Ateneo).



primary cultured neonatal rat cardiomyocytes with hypoxia/reoxygenation injury. Pharmacogn. Mag. 9, 202–209. doi: 10.4103/0973-1296.113264


**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.

The handling Editor and author RF declared their involvement as co-editors in the Research Topic, and confirm the absence of any other collaboration.

Copyright © 2019 Fraschini. 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.

# Centromere and Pericentromere Transcription: Roles and Regulation . . . in Sickness and in Health

#### Ksenia Smurova and Peter De Wulf\*

Centre for Integrative Biology, University of Trento, Trento, Italy

The chromosomal loci known as centromeres (CEN) mediate the equal distribution of the duplicated genome between both daughter cells. Specifically, centromeres recruit a protein complex named the kinetochore, that bi-orients the replicated chromosome pairs to the mitotic or meiotic spindle structure. The paired chromosomes are then separated, and the individual chromosomes segregate in opposite direction along the regressing spindle into each daughter cell. Erroneous kinetochore assembly or activity produces aneuploid cells that contain an abnormal number of chromosomes. Aneuploidy may incite cell death, developmental defects (including genetic syndromes), and cancer (>90% of all cancer cells are aneuploid). While kinetochores and their activities have been preserved through evolution, the CEN DNA sequences have not. Hence, to be recognized as sites for kinetochore assembly, CEN display conserved structural themes. In addition, CEN nucleosomes enclose a CEN-exclusive variant of histone H3, named CENP-A, and carry distinct epigenetic labels on CENP-A and the other CEN histone proteins. Through the cell cycle, CEN are transcribed into non-coding RNAs. After subsequent processing, they become key components of the CEN chromatin by marking the CEN locus and by stably anchoring the CEN-binding kinetochore proteins. CEN transcription is tightly regulated, of low intensity, and essential for differentiation and development. Under- or overexpression of CEN transcripts, as documented for myriad cancers, provoke chromosome missegregation and aneuploidy. CEN are genetically stable and fully competent only when they are insulated from the surrounding, pericentromeric chromatin, which must be silenced. We will review CEN transcription and its contribution to faithful kinetochore function. We will further discuss how pericentromeric chromatin is silenced by RNA processing and transcriptionally repressive chromatin marks. We will report on the transcriptional misregulation of (peri)centromeres during stress, natural aging, and disease and reflect on whether their transcripts can serve as future diagnostic tools and anti-cancer targets in the clinic.

Keywords: centromere, pericentromere, kinetochore, heterochromatin, long non-coding RNA, transcription

# CENTROMERES, KINETOCHORES, AND ANEUPLOIDY

During cell division, the replicated chromatids that are associated by cohesin rings bind to the microtubules of the metaphase spindle, which extend from two opposite spindle poles (**Figure 1**). This binding is mediated by kinetochores, each of which assembles on the centromere (CEN) of each chromatid. CENP-A/CenH3, a variant of histone protein H3, recruits all kinetochore subunits

#### Edited by:

Roberta Fraschini, Università degli Studi di Milano-Bicocca, Italy

#### Reviewed by:

Tatsuo Fukagawa, Osaka University, Japan Michael D. Blower, Harvard Medical School, United States

> \*Correspondence: Peter De Wulf peter.dewulf@unitn.it

#### Specialty section:

This article was submitted to Genetic Disorders, a section of the journal Frontiers in Genetics

Received: 06 September 2018 Accepted: 04 December 2018 Published: 21 December 2018

#### Citation:

Smurova K and De Wulf P (2018) Centromere and Pericentromere Transcription: Roles and Regulation . . . in Sickness and in Health. Front. Genet. 9:674. doi: 10.3389/fgene.2018.00674

**56**

and spindle assembly checkpoint (SAC) proteins to the centromeric nucleosome(s). To prevent aneuploidy, the SAC monitors chromosome-spindle attachment at each kinetochore. The SAC arrests the cell division process at the metaphase–anaphase transition when a single chromosome pair is found to be unbound or misbound to the mitotic spindle. The SAC kinase Aurora B then phosphorylates the outer kinetochore Ndc80 protein of each misbound sister pair to detach it from the spindle structure. The delay of mitosis allows for a correct re-attachment. Only when the SAC is satisfied will all sister chromosomes separate by enzymatic cleavage of the cohesin rings. Each kinetochore-bound chromatid then moves into the daughter cells by depolymerization of the spindle microtubules and, in some eukaryotes, by additonal motor protein activity. In the end, each cell receives a full complement of the maternal genome (**Figure 1**). Abnormal CEN or kinetochore activity has been linked with cancer initiation/progression, developmental defects, and genetic disease (Holland and Cleveland, 2009; Santaguida and Amon, 2015). For more detailed information about kinetochores we refer to Fukagawa and Earnshaw (2014); McKinley and Cheeseman (2016); and Musacchio and Desai (2017). Of note, during revision of this manuscript, an excellent review was published (Perea-Resa and Blower, 2018) partially overlaps with ours in subject matter.

# CENTROMERES: EVOLUTIONARY DIVERGED SEQUENCES

The CEN was first identified as the central constriction of each chromosome during the light microscopic analysis of mitotic salamander cells (Flemming, 1880). Today, it is defined as the chromosomal region that underlies the stable transmission of the nuclear genomic content from one generation to the next. In the 1980s; the CEN of budding yeast Saccharomyces cerevisiae chromosome 3, and all three CEN of the fission yeast Schizosaccharomyces pombe were the first CEN loci to be characterized (Clarke and Carbon, 1980; Nakaseko et al., 1987; **Figures 2A,B**). The short budding yeast "point" CEN is ∼120 bp long and contains three DNA elements that wrap around a single CEN nucleosome. Alternate stretches of A and T residues, which cause DNA bending, comprise CDEII, which is bordered by palindromic motifs named CDEI and CDEIII (**Figure 2A**). In contrast to CDEII and CDEIII, CDEI is not essential for kinetochore activity but mutations in its sequence cause chromosome loss (Niedenthal et al., 1991). In S. cerevisiae, the CEN sequence per se defines CEN identity. In contrast and because of their 40–110 kb length, the CEN in fission yeast are designated as "regional." They comprise a 4–7 kb core sequence named cnt that encloses multiple CEN nucleosomes. The core is flanked by inverted, 6 kb-long innermost imr repeats that contain

sister chromatids) on the metaphase spindle along which they then segregate in opposite directions into the two daughter cells that receive a full complement of the maternal genome (green arrows). Errors made during the segregation process caused by CEN or kinetochore malfunction lead to aneuploid daughter cells (red arrows) carrying an abnormal number of chromosomes. Consequences are cell death, genetic disease (developmental defects), and cancer initiation/progression. The insert shows a more detailed representation of a sister chromosome pair whose chromosomes (original and copy) are linked by cohesion rings. The sister chromosomes are bound to the spindle microtubules via kinetochores that assemble on the CEN sequence of each chromosome.

chromatin, black regions: heterochromatin. The latter represent the pericentromeres, telomeres, LINEs, SINEs, micro- and macrosatellites, β, γ, I, II, III-satellites, rDNA, and DNA transposons (approximate lengths are indicated in the black box). The gray arrows represent the CEN alpha-satellite monomers, organized in a head-to-tail fashion. HOR, high-order repeat of alpha-satellite monomers (green arrow). A-boxes (dark green) and B-boxes (purple) are indicated, as well as the cruciform configuration of a dyad sequence. Right: Human centromeric chromatin with the CENP-A containing nucleosomes clustered and exposed in amphipathic configuration at its outside is bound by numerous kinetochores to a bundle of microtubule fibers (based on Fukagawa and Earnshaw, 2014). See text for details.

clusters of tRNA genes. Together, these three elements form the central domain, which is flanked left and right by outer repeats, otr, named dg and dh (**Figure 2B**).

The regional CEN of most higher eukaryotes are comprised of retrotransposon repeats and repeats of a simple 171-bp CEN sequence, named alpha-satellite DNA, where the CENP-A nucleosomes reside (**Figure 2C**). The surrounding pericentromeric domains contain repeats that are less ordered. In humans and most primates, the alpha-satellite sequence is organized in back-to-back fashion, forming a high-order

repeat (HOR) (Manuelidis and Wu, 1978; Willard, 1985). Within a HOR, alpha-satellite monomers are 50–70% identical (Willard, 1985). Each HOR is repeated hundreds-to-thousands of times, producing 2–5 Mb-long arrays (Aldrup-MacDonald and Sullivan, 2014; **Figure 2C**). Different chromosomes are distinguished by variations within the alpha-satellite sequences, by the number of alpha-satellite monomers, and the overall size of the HOR. Not all alpha-satellite monomers contribute to human kinetochore activity, these are labeled as "inactive." Human CEN contain alpha-satellite monomers of the A and B type, while lower primates only have A-type satellites (Alexandrov et al., 2001). Both monomers differ in a 17-bp sequence called A or B box (**Figure 2C**). The latter, also named CENP-B box, binds CEN protein CENP-B (Masumoto et al., 1989). It is unclear if a specific protein binds to the A box. Human chromosomes, except the Y chromosome, contain B-type alpha-satellite monomers (Tyler-Smith and Brown, 1987). A third type of alpha-satellite monomers contains neither an A nor a B box. The CEN in mice consist of homogeneous arrays of 120-bp minor satellite (MinSat) repeats, that are flanked by repeats of less-ordered 234-bp major gamma-satellite (MajSat) sequences (Joseph et al., 1989). The CEN repeat units in higher eukaryotes are typically around 150 bp in length [178 bp in plants (Kumekawa et al., 2001; Nagaki et al., 2003)], each enclosing one CENP-A nucleosome. However, they can be much shorter as in Drosophila melanogaster, whose CEN (200–500 kb) are made up of 10-bp repeats followed by 11/12-bp tandem repeats (Garavís et al., 2015b).

Most eukaryotes are monocentric since their chromosomes contain one CEN. In contrast, moths and butterflies, as well as nematodes such as Caenorhabditis elegans, and arachnids contain holocentric CEN that cover the entire chromosome, except for the telomeric regions (Heckmann et al., 2011; Steiner and Henikoff, 2014). While the C. elegans genome comprises few tandem repeats (Hillier et al., 2007), ∼50% of the genome is associated with CENP-A in 20 CEN domains of variable size (Albertson and Thomson, 1982; Gassmann et al., 2012). Its kinetochores hence may assemble randomly or at specific regions. While the evolutionary forces that drove holocentrism are unknown, one benefit may lie in DNA breaks. In contrast to broken monocentric chromosomes, fragmented holocentric chromosomes can still segregate in mitotic anaphase because of the multiple microtubule attachments they may contain. Nevertheless, the prevalence of monocentrism suggests selective advantages, possibly related to difficulties in segregating recombined holocentric chromosomes during meiosis (Maddox et al., 2004). For more detailed information about CEN we refer to Aldrup-MacDonald and Sullivan (2014); Bloom and Costanzo (2017); and Fukagawa and Earnshaw (2014).

#### TRANSCRIPTIONALLY ENHANCED CENTROMERE FEATURES

Centromeres evolved rapidly due to homologous recombinations between stretches of tandemly repeated sequences. Even within one organism CEN sequences differ significantly between its chromosomes. Despite this divergence, most CEN-binding kinetochore proteins are conserved. This "CEN paradox" is explained by the maintenance of CEN-specific structural themes during the co-evolution of CEN DNA and the CEN-binding kinetochore proteins (Henikoff et al., 2001). The adaptive evolution of CENP-A and its orthologs involves regions within this protein that are predicted to contact the centromeric DNA (Talbert et al., 2004; Schueler et al., 2010). In turn, CEN may not have been selected based on their DNA sequence but rather on non-canonical structures that act as beacons for kinetochores and sustain the pulling forces that CEN nucleosomes undergo during chromosome segregation. Studies of CEN from numerous species have indicated a functional significance of non-B-form DNA structures including single-stranded (ss) DNA, hairpins, triplexes, i-motifs, and cruciform extrusions as observed in vitro and/or in vivo (Zhu et al., 1996; Ohno et al., 2002; Jonstrup et al., 2008; Garavís et al., 2015a,b; Aze et al., 2016; Kabeche et al., 2018). All CEN, except those of S. cerevisiae, maintain a high level of inter-repeat sequence property, suggestive of a recombination-based mechanism that produces covalently closed stem–loop structures, which may define CEN recognition and activity. A conserved stem–loop model would demand repeat DNA sequences, explaining the evolution of the CEN's repeat-array configuration (illustrated for the S. pombe CEN in **Figure 2B**). Metazoans might require a threshold number of these loop structures to produce a functional CEN (McFarlane et al., 2010). Possibly, the single-stranded loops could be formed temporarily during replication and/or transcription to seed kinetochores.

A neocentromere, being a new CEN that originates at a site that is not centromeric usually due to disruption of the natural CEN, lack centromeric alpha-satellite DNA, but are fully competent to generate a primary constriction and assemble a functional kinetochore (Marshall et al., 2008) indicating that alpha-satellite DNA per se is not a trigger for attracting CEN proteins. However, neocentromeres actually form at chromosomal sites that not only contain pre-existing repeats but further develop extensive repetitive DNA sequences over time, indicating the advantage of acquiring an extensive repeat configuration (Marshall et al., 2008). Epigenetic mechanisms are additionally required for maintaining neocentromere identity and activity.

Drosophila melanogaster CEN are made up of short satellite DNA repeats (AATAACATAG)<sup>n</sup> followed by doceda tandem repeats (CCCGTACT[C]GGT) that show an asymmetric distribution of G and C residues. In vitro, the C-rich dodeca satellite single strand produces an "i-motif "; a cubic structure that is formed by the head-to-tail association of two parallel strands combined in antiparallel fashion (Garavís et al., 2015b; **Figure 2C**). Similar i-motif structures arise in vitro between human alpha-satellite monomers in which the C-rich strand of one A-box associates with that of a neighboring A-box. CEN-B boxes also form i-motifs, while those produced from an A- and B-box strand are somewhat unstable in vitro (Garavís et al., 2015b). Murine Y CEN satellite DNA that lacks an A/B-box has a sequence capable of forming an i-motif in an equivalent position (Garavís et al., 2015a). As i-motifs can

form upon transcriptionally induced supercoiling (Sun and Hurley, 2009) and since the transcription of alpha-satellite DNA is required for CEN function (Chan et al., 2012), negative superhelicity may favor i-motif formation under physiological conditions.

In vivo evidence for the phasing of CENP-A nucleosomes showed that their positioning is a physical requirement for CEN function (Hasson et al., 2013; Zhang et al., 2013). In most higher eukaryotes CEN chromatin contains blocks of CENP-A that are interspersed with blocks of histone H3-containing nucleosomes (Bodor et al., 2014; Fukagawa and Earnshaw, 2014; **Figures 2A,C**). CENP-A nucleosomes may associate laterally and exclude the H3-containing nucleosomes. The flexibility observed in the chromatin that flanks the CENP-A nucleosomes facilitates these interactions (Panchenko et al., 2011; Hasson et al., 2013). In humans, the phasing of CENP-A nucleosomes on alpha-satellite DNA places the A- and B-boxes at the beginning and at the end of the nucleosome (Hasson et al., 2013). Models of CEN chromatin folding into an amphipathic helix, loop, or boustrophedon that expose the CENP-A nucleosomes at the chromatin surface have been suggested to facilitate kinetochore formation (Blower et al., 2002; Bloom and Costanzo, 2017). A hierarchical mechanism of chromatin folding based on A- and B-box interactions and i-motif formation may determine the 3D organization of the CEN. Although CENP-B null mice are viable (Kapoor et al., 1998), CENP-B is required for de novo CEN formation on artificial chromosomes (Ohzeki et al., 2002) and enhances chromosome segregation fidelity (Fachinetti et al., 2015). Possibly, B-box i-motifs contribute to a nucleosome environment that improves kinetochore assembly and activity.

While examining the CEN from different species, Kasinathan and Henikoff (2018) identified clade-specific variations in <10-bp dyad symmetries predicted to adopt stable non-B-form cruciform extrusions (**Figure 2C**). Satellites lacking CENP-B boxes were highly enriched in these palindromes. Non-B-form DNA regions were abundant in human alpha-satellite and murine MinSat sequences from activated B cells, while reduced levels were observed in non-proliferating cells, suggesting that replication induces cruciform extrusions at CEN in dividing cells (Kasinathan and Henikoff, 2018). The authors propose that CEN are either highly enriched with dyad sequences or less-enriched in dyads that flank a nearby binding site for a DNA-bending protein whose association may stimulate dyad cruciform formation. The four-way junctions of the cruciform could be recognized by the HJURP chaperone (Scm3 in yeast) that loads CENP-A into the centromeric nucleosome (Dunleavy et al., 2009; Foltz et al., 2009; Sanchez-Pulido et al., 2009). Non-B form elements may also facilitate CEN transcription initiation and elongation by RNA polymerase II (RNAPII), enabling the loading of CENP-A during nucleosome remodeling. Also, CENP-B may be dispensable for CEN where HJURP is recruited by CENP-C and the MIS18 complex (Nardi et al., 2016) (see below). Hence, A/B boxes and dyad sequences may organize and activate CENP-A loading into CEN nucleosomes.

downstream kinetochore components as in vertebrates. See text for details.

## POST-TRANSLATIONAL MODIFICATIONS OF CENTROMERIC AND PERICENTROMERIC CHROMATIN

Within the CEN domain, CENP-A nucleosomes are interspersed with canonical nucleosomes whose histone H3 tails are methylated at Lys4 (H3K4me1, H3K4me2) and Lys36 (H3K36me2, H3K36me3) (**Figure 3A**). These modifications underlie open chromatin, promote RNAPII activity, and are essential for HJURP targeting and CENP-A assembly (Bergmann et al., 2011; Duda et al., 2017). They also differentiate the CEN chromatin from the surrounding pericentromere regions, which are marked differently (see below and **Figure 3A**) (Sullivan and Karpen, 2004; Eymery et al., 2009; Gopalakrishnan et al., 2009; Bergmann et al., 2011, 2012). Intriguingly, H3K9me3, typically associated with transcriptional repression, also labels the centromeric nucleosomes (Bergmann et al., 2012) indicating that CEN chromatin epitomizes both silent heterochromatin and transcribed euchromatin (Sullivan and Karpen, 2004).

Histone H4 mono-acetylation at Lys5 and Lys12, which correlates with transcribed chromatin, is enriched at CEN and is essential for CENP-A deposition in chicken cells (Shang et al., 2016; **Figure 3A**). H4 mono-methylation at Lys20, which marks human and chicken CENP-A nucleosomes, and is associated with transcriptional activation, is a prerequisite for kinetochore assembly (Sullivan and Karpen, 2004; Vakoc et al., 2006; Wang et al., 2008; Bergmann et al., 2011; Hori et al., 2014). Histone H2B mono-ubiquitination at Lys119, catalyzed by the E3 ubiquitin ligase RNF20/40 (Brl1 in S. pombe), is required for CEN transcription (Zhu et al., 2011; Sadeghi et al., 2014). Depleting RNF20 reduces CEN transcription and nucleosome turnover, and causes chromosome missegregation in human cells and S. pombe (Sadeghi et al., 2014; Zhang et al., 2017). The ubiquitin ligase BRCA1 preserves CEN identity by ubiquitinating histone H2A at Lys119, producing a repressive mark. BRCA1 depletion, led to CEN transcript overexpression, impaired CEN cohesion and SAC activity, and chromosome missegregation (Di Paolo et al., 2014).

## CENTROMERE TRANSCRIPTION, PROMOTERS, AND TRANSCRIPTION FACTORS

Although electron microscopy-based studies had localized RNA at kinetochores in union and salamander cells in the 1970s (Braselton, 1975; Rieder, 1979), CEN were long considered transcriptionally silent since they are confined in transcriptionally inert heterochromatin. Today we know that CEN are actively transcribed by RNAPII, which has been detected at CEN in S. pombe, flies, and human cells, at centromeric chromatin on human artificial chromosomes (HACs), and at neocentromeres (Wong et al., 2007; Li et al., 2008; Chueh et al., 2009; Ferri et al., 2009; Bergmann et al., 2011; Choi et al., 2011; Ohkuni and Kitagawa, 2011; Lyn Chan and Wong, 2012; Quénet and Dalal, 2014; Rošic et al., 2014 ´ ; Catania et al., 2015). Despite the evidence of RNAII polymerase transcribing the CEN, very little is known about the promoters and transcription factors involved.

In S. cerevisiae, RNAPII-mediated CEN transcription is driven by transcription factors Cbf1 and Ste12. Cbf1 promotes transcription from the sense strand, Ste12 from the antisense strand. Silencing protein Dig1 inhibits Ste12. Transcriptional silencers Sir1, Hst1–Sum1, and Cdc14–Net1 associate with the CEN sequence, possibly to antagonize RNAPII. While deleting CBF1 or STE12 did not prevent kinetochore assembly, each mutant experienced chromosome loss. This phenotype was rescued by driving CEN transcription from an inducible promoter introduced next to the Cbf1- or Ste12-binding site, illustrating that CEN transcription is imperative for kinetochore activity (Ohkuni and Kitagawa, 2011). CEN transcripts in S. cerevisiae remained unidentified until exosome activity (which degrades non-coding RNAs) was removed, indicating a fast turn over of these transcripts. This approach revealed a 1.2 kb CEN3 RNA species, revealing that RNAPII proceeds into the pericentromere (Houseley et al., 2007). Low-level CEN transcription is required for kinetochore activity in budding yeast. Disproportionate CEN expression driven by the galactoseinducible PGAL1 promoter placed adjacent to CEN3 on a plasmid caused plasmid loss (Hill and Bloom, 1987) since kinetochores were not able to assemble. When PGAL1 was positioned next to chromosomal CEN3 that was marked with a GFP-array, growth in galactose prevented spindle binding of labeled sister chromatids 3. Following glucose addition, the sisters bi-oriented on the metaphase spindle (Tanaka et al., 2005).

In S. pombe, the CENP-A binding region contains numerous transcription start sites and promoters on the forward and reverse strands. However, very low levels of transcripts are produced, due to transcript turnover as well as RNAPII stalling (Choi et al., 2011; Sadeghi et al., 2014), which could result from collisions with the replisome or transient H2B (de)ubiquitination activity that negatively affects chromatin accessibility (Chen et al., 2008; Sadeghi et al., 2014). In fission yeast mutants unable to restart stalled RNAPII, CENP-A became actively deposited on the CEN, suggesting that halting RNAPII, which results in a low-quality transcription environment, allows for CEN chromatin remodeling and/or CENP-A loading (Shandilya et al., 2014; Catania et al., 2015).

The PRAT CEN satellite monomer in the beetle Palorus ratzeburgii contains a putative RNAPII promoter site that overlaps with the most conserved part of the PRAT sequence. This concurrence could be the result of selection pressure to preserve the transcription activity of this satellite DNA. TATAbox-like motifs, multiple transcription initiation and termination sites were also mapped within the monomer. The presence of a 5 0 -RNA cap and 3<sup>0</sup> -poly(A) tails in a portion of the beetle CEN transcripts indicates RNAPII-dependent transcription. Indeed, treatment of larvae with alpha-amanitin at concentrations that selectively inhibit RNAPII activity reduced the amount of PRAT transcripts. These transcripts derived from one, two, or three monomers, and were produced from both strands (albeit 10 times less from the antisense strand) (Pezer and

Ugarkovic, 2008 ´ ). Within the human alpha-satellite sequence, a candidate TATA box has been identified, as well as an SV40 enhancer-core sequence with spacing and orientation characteristic of RNAPII-transcribed genes (Vissel et al., 1992). In human cells, RNAPII has been found especially enriched at prometaphase, metaphase, and anaphase CEN, as well as at kinetochore-active neocentromeres. Consistent with active transcription, FCP1, a phosphatase that is specific for the carboxy-terminal domain of RNAP II and stimulates transcript elongation by RNAP II (Mandal et al., 2002), was identified at mitotic human and murine kinetochores (Chan et al., 2012).

#### TRANSCRIPTION AND POST-TRANSLATIONAL MODIFICATIONS PROMOTE CENP-A INCLUSION INTO CENTROMERIC CHROMATIN

While CENP-A represents the epigenetic mark of CEN identity in most eukaryotes (Vafa and Sullivan, 1997; Warburton et al., 1997) its presence per se is not enough for CEN formation since trypanosomes and insects with holocentric chromosomes lack a CENP-A ortholog (Akiyoshi and Gull, 2014; Drinnenberg et al., 2014). CENP-A nucleosomes in humans are also found at non-CEN sites, including neocentromeres (Bodor et al., 2014). Both observations underscore the need for additional CEN-specifying criteria, including structural themes embedded within the CEN DNA sequence (see above). Via its N- and C-terminal tails and through its central histone-fold domain, CENP-A recruits the other kinetochore proteins, including CENP-C with which it makes direct physical contact (Chen et al., 2000; van Hooser et al., 2001; Regnier et al., 2005; Liu et al., 2006; Carroll et al., 2009, 2010; Guse et al., 2011; Fachinetti et al., 2013; Kato et al., 2013; Folco et al., 2015; Logsdon et al., 2015; Westhorpe et al., 2015; **Figure 3B**). In contrast to histone H3, CENP-A may form a more rigid interface with its partner histone H4, which is further stabilized by CENP-C. Nucleosomes containing CENP-A bind less firmly to the DNA, profoundly affecting CEN transcription and distinghuishing it from the surrounding closed-state chromatin (Hasson et al., 2013; Falk et al., 2015). During chromosome replication, CENP-A becomes diluted 1:2 with histone H3 variant H3.3, which is deposited as a temporary placeholder allowing kinetochores to assemble in early metaphase (**Figure 3B**). In mammals, CENP-A becomes incorporated in late telophase/early G1, when its chaperone HJURP localizes to CEN and H3.3 is removed (Foltz et al., 2009; Dunleavy et al., 2011). CENP-A deposition also requires the MIS18 complex (MIS18α, MIS18β, MIS18-binding protein 1/KNL2) (Hayashi et al., 2004). In D. melanogaster, HJURP and MIS18 activities appear to be combined in the Cal1 protein (Erhardt et al., 2008; Chen et al., 2014).

In S. cerevisiae, the CEN nucleosomes are evicted and kinetochores disassembled at S-phase entry, allowing for the replication of the CEN sequences, which are the first loci to be replicated in budding yeast. It is unclear whether CEN transcription is downregulated during this process. The expelled CENP-A then becomes degraded. Within 5 min after passage of the replisome, the CEN nucleosomes reassemble by the inclusion of new CENP-A by the Scm3 chaperone (ortholog of HJURP). Kinetochores then reassemble to attach the still-replicating chromatids to the interphase spindle (Kitamura et al., 2007; Wisniewski et al., 2014).

During G1 in human cells, the MIS18 complex recruits the KAT7 histone acetyltransferase complex to maintain an acetylated CEN chromatin state, which facilitates the assembly of new CENP-A nucleosomes (Ohzeki et al., 2016). CENP-C contributes to CENP-A inclusion and stability by interacting directly with CENP-A, HJURP, and MIS18-binding protein 1 (Moree et al., 2011; Dambacher et al., 2012; McKinley and Cheeseman, 2014; Tachiwana et al., 2015; **Figure 3B**). Furthermore, CENP-C, the remodeling and spacing factor complex RSF, and the MgcRacGAP Male germ cell Rac GTPase-activating protein maintain CENP-A once incorporated (Perpelescu et al., 2009; Lagana et al., 2010; Falk et al., 2015). In contrast, cyclin-dependent kinase (CDK) activity negatively regulates CENP-A incorporation. In D. melanogaster, the turnover of S/G2 phase cyclin A in mitosis is key for the deposition of CENP-A (Erhardt et al., 2008; Mellone et al., 2011). In human cells, CDKs phosphorylate the MIS18-binding protein 1 to reduce its CEN localization (Silva et al., 2012) and to avert the recruitment of the MIS18α and MIS18β beyond G1 (McKinley and Cheeseman, 2014). CDK phosphorylation of HJURP also disrupts its CEN localization (Müller et al., 2014; **Figure 3B**). In contrast, the kinase PLK1 targets the MIS18 complex to promote its CEN localization and to license the CEN for CENP-A delivery. Bypassing both CDK and PLK1 activities led to CENP-A deposition throughout the cell cycle, causing severe mitotic defects (McKinley and Cheeseman, 2014). Clearly, CENP-A must be loaded only in G1 to ensure correct CEN function.

The de novo loading of CENP-A, as detailed above, requires CEN transcription as catalyzed by RNAPII (Lyn Chan and Wong, 2012; Quénet and Dalal, 2014; Rošic et al., 2014 ´ ; Grenfell et al., 2016; **Figure 3B**). In Drosophila, Cal1 recruits RNAPII and the chromatin-remodeling complex FAcilitates Chromatin Transcription (FACT) (Foltz et al., 2006; Chen et al., 2015). Studies suggest that FACT activity weakens the histone core-DNA contact, facilitating the passage of RNAPII, and protecting the nucleosome from falling apart before it is remodeled and the new CENP-A nucleosome assembled. FACT also binds to the CEN CENP-T/W complex, possibly to promote also its deposition (Prendergast et al., 2016). Of note, FACT localizes at CEN at all stages of the cell cycle and is responsible for CENP-A loading in human cells (Okada et al., 2009). In fungi, FACT activity prevents the ectopic incorporation of CENP-A beyond CEN, rather than promoting CENP-A assembly at CEN nucleosomes (Deyter and Biggins, 2014). In Drosophila, CEN transcription and chromatin remodeling are required for

CENP-A to transition from an unstable chromatin-associated state to a stable nucleosome-incorporated state (Bobkov et al., 2018).

Alpha-satellite arrays amplified from human CEN and cloned into a BAC plasmid form a functional HAC that recruits kinetochores and stably propagates in HT1080 fibrosarcoma cells (Maloney et al., 2012). In HACs containing engineered tetO operator sequences within the alpha-satellite DNA, and cells expressing transcriptional activators or silencers fused with the tetO-binding TetR protein both destabilized kinetochore formation (Bergmann et al., 2011). Transcriptional silencing led to a gradual loss of CENP-A from the centromeric chromatin, due to reduced recruitment of HJURP. Enhancing alpha-satellite transcription ∼10-fold by tethering a minimal NF-κB p65 activation domain did not affect kinetochore formation or activity. However, tethering TetR with the activation domain of herpes virus transcription factor VP16 elevated transcription ∼150-fold, approaching the expression level of a housekeeping gene. The consequent increase in RNAPII occupancy provoked a loss of CENP-A, probably through nucleosome eviction (Bergmann et al., 2012).

Post-translational modifications of CENP-A are required for its loading (**Figure 3A**). Before becoming deposited, CENP-A is phosphorylated at Ser16 and Ser18 (Bailey et al., 2013); Ser18 is a substrate for the cyclin E1/CDK2 kinase (Takada et al., 2017). A loss or hyperphosphorylation of both sites causes chromosome missegregation (Bailey et al., 2013; Takada et al., 2017). Drosophila CENP-A is phosphorylated at Ser75 and Ser77, which could be the analogs of Ser16 and Ser18 in human CENP-A (Boltengagen et al., 2016). Biochemical evidence suggests that mono-ubiquitination of CENP-A at Lys124 by the E3 ligase activity of the CUL4A–RBX1–COPS8 complex promotes HJURP binding and CENP-A deposition (Niikura et al., 2015, 2017). However, disputing gene replacement experiments showed that non-ubiquitinatable mutant CENP-A still can replace endogenous CENP-A and support cell viability (Fachinetti et al., 2017). In humans, the starting methionine of pre-inclusion CENP-A is removed and the exposed Gly1 residue trimethylated by the enzyme NRMT1 (Bailey et al., 2013; Sathyan et al., 2017). Both this modification and phosphorylation of Ser16 and Ser18 persist after CENP-A loading (Bailey et al., 2013). Subsequent modifications of the incorporated CENP-A include Ser7 phosphorylation, which is responsible for the indirect recruitment of CENP-C, and ubiquitination of Lys124, shown to be involved in CENP-A binding to HJURP (Srivastava et al., 2018). Mutations in Ser7, Ser16, and Ser18 sites lead to chromosome missegregation, abnormal spindles, and errors in cytokinesis (Srivastava et al., 2018). Nevertheless, chromosomes carrying CENP-A mutants that cannot be phosphorylated at Ser68 or ubiquitinated at Lys124 establish functional CEN (Fachinetti et al., 2017). Since the same amount of CENP-A is renewed at each G1 stage, errors in CENP-A incorporation caused by abnormal CEN transcription, assembly factor activity, and/or post-translational modifications could permanently alter its

levels at centromeric chromatin, contributing to chromosomal instability.

## CENTROMERE TRANSCRIPTION THROUGH THE CELL CYCLE

Centromere transcription dynamics through the cell cycle have only been studied recently. The levels of alpha-satellite RNAs localizing at CEN did not change through the cell cycle, indicating a complex dynamic between CEN RNA synthesis, turnover, and stable incorporation in the CEN chromatin (McNulty et al., 2017). CEN RNA and DNA FISH experiments using identical HOR probes labeled with different fluorophores showed a co-localization of the transcripts to their originative CEN, indicating they are maintained in cis (McNulty et al., 2017). As discussed earlier, CEN transcription is required for CENP-A loading in human and Drosophila cells (Quénet and Dalal, 2014; Bobkov et al., 2018). Human CEN transcription mediated by RNAPII, in conjunction with the TATA-box binding protein, occurs through early G1 when mammalian CENP-A is deposited. When inhibiting transcription in G1, CENP-A levels dropped with ∼50% (Quénet and Dalal, 2014). Targeting the transcript with shRNA, while not impeding RNAPII activity, diminished CENP-A levels and induced mitotic defects (Quénet and Dalal, 2014). Reversely, depleting CENP-A reduced CENP-C concentrations at kinetochores, but CEN transcript levels were not affected, suggesting that CEN transcription occurs before the recruitment of CENP-A and CENP-C (McNulty et al., 2017). However, inhibiting active transcription resulted in CENP-C destabilization, suggesting that CEN transcription may also act downstream of CENP-A loading to promote CENP-C binding (Chan et al., 2012).

While most regions within condensed chromosomes are transcriptionally silent during mitosis, CEN are not (Chan et al., 2012; Lyn Chan and Wong, 2012; Liu et al., 2015), therewith differentiating them from the rest of the genome. Indeed, as indicated earlier, RNAPII localized at human and murine CEN from prometaphase through anaphase (Chan et al., 2012). Mild CEN transcription through the cell cycle ensures stable kinetochores and CEN cohesion (Liu et al., 2015).

In human cells, the cohesin-protecting protein Sgo1 (Shugoshin) is recruited to early mitotic kinetochores by the Bub1-phosphoryated centromeric histone H2A [phosphorylated at Thr120; (H2A T120P)]. Next, Sgo1 binds to RNAPII and travels along with it to the inner CEN (region between the two sister CEN) where it binds to the cohesin rings to protect them from precocious cleavage by the protease separase (Liu et al., 2015). Transcription by RNAPII and chromatin remodeling activities could open the chromatin, allowing Sgo1 access to cohesin. When transcription elongation was inhibited during mitosis with alpha-amanitin or when RNAPII subunit Rbp2 was degraded, Sgo1 still localized at kinetochores but did not relocate to the inner CEN. Besides RNAPII activity, the CEN RNAs themselves may facilitate Sgo1 relocation to the inner CEN. Indeed, since nonspecific RNA competed with H2A T120P for binding to Sgo1, CEN RNA could bind to Sgo1, releasing it

from H2A T120P and allowing Sgo1 to travel with RNAPol II toward the inner CEN.

In contrast to human alpha-satellite transcripts, murine MinSat transcripts are absent in G0/G1. They appear in S-phase, peak at G2/M, and become undetectable after mitosis, when cells re-enter the cell cycle (Ferri et al., 2009). This dynamic mirrors the accumulation of the chromosomal passenger complex (CPC) at the murine CEN, implicating a role of MinSat RNAs in CPC localization and activity. Indeed, MinSat RNAs accumulate at CENP-A chromatin and interact with CPC subunits Aurora B and Survivin at mitotic onset. We will describe the interactions between CEN RNA and the CPC components in detail further below.

Schizosaccharomyces pombe CEN are transcribed during DNA replication, which may generate transcription–replication conflicts. Encounters between RNAPII and the replisome may cause RNAPII to halt and produce immature transcripts (Lu and Gilbert, 2007; Chen et al., 2008). RNAPII stalling generates RNA–ssDNA hybrids, known as R-loops (Reddy et al., 2011), which have also been observed at human CEN chromatin (Kabeche et al., 2018). R-loops must be resolved; otherwise, they can provoke chromosome breaks and repeat-sequence recombinations. R-loops forming in centromeric chromatin (or at pericentromeres or across the genome) trigger Aurora B-mediated phosphorylation of local histone H3 at Ser10, as shown in yeast, C. elegans, and human cells. This mark stimulates confined chromatin condensation and restricts DNA replication and transcription (Castellano-Pozo et al., 2013; Oestergaard and Lisby, 2016). Since the FACT complex resolves R-loops in yeast and human cells (Herrera-Moyano et al., 2014), it could remove toxic R-loops prior to mitotic entry. FACT activities including the stimulation of CEN chromatin remodeling and transcription, the subsequent promotion of CENP-A assembly, and the resolution of R-loops may reflect the dynamic state of the CEN environment during cell cycle progression (Duda et al., 2017).

### POST-TRANSCRIPTIONAL PROCESSING OF CENTROMERE TRANSCRIPTS

In S. pombe, 5<sup>0</sup> -capped and 3<sup>0</sup> -polyadenylated non-coding CEN RNAs that are produced from the central domain are quickly degraded by the exosome (Choi et al., 2011). No evidence exists for small CEN RNA processing products as documented for the transcripts derived from the pericentromeric chromatin (see below). The RNase activity of exosome subunit Dis3 is required for correct kinetochore assembly and kinetochore–microtubule interactions (Bühler and Moazed, 2007; Mukarami et al., 2007) suggesting that degradation of CEN transcripts independent of the RNA interference (RNAi) pathway contributes to CEN activity in fission yeast.

Genome-wide screens with Drosophila and human cells identified splicing factors that are required for cell division (Goshima et al., 2007; Kittler et al., 2007; Somma et al., 2008; Neumann et al., 2010). Also, purifications of the spliceosome from HeLa cell nuclear extracts revealed the presence of microtubule- and mitotic chromatin-interacting proteins (Makarov et al., 2002). The processing of CEN RNAs may occur in mitosis since splicing factors are co-transcriptionally recruited to the elongating RNAPII transcripts (Listerman et al., 2006; David et al., 2011) (and because RNA-splicing factor Prp4 localizes to mitotic kinetochores in HeLa cells (Montembault et al., 2007; **Figure 3B**). Splicing factors also interact with MinSat transcripts in murine cells (Maison et al., 2011). The co-transcriptional recruitment of the RNA processing machinery to nascent mitotic transcripts in Xenopus is an important step in kinetochore and spindle assembly. Indeed, long non-coding CEN RNAs localize to mitotic chromosomes, chromatin, and spindles (Blower, 2016). At spindles, the transcripts regulate Aurora B and MCAK activities (Grenfell et al., 2016). Inhibiting the spliceosome, which co-IPs with CEN transcripts and CENP-C, in metaphase-arrested Xenopus egg extracts caused an accumulation of long CEN antisense transcripts representing up to six frc1 monomer repeats, which are much longer than the standard CEN RNAs containing one to two frc1 repeats. A globally reduced recruitment of CENP-A, CENP-C, and Ndc80 was observed (Grenfell et al., 2016), suggesting that fcr1 antisense RNA is processed and then freely diffuses between CEN in trans, similar to observations in Drosophila where CEN RNAs derived from the X chromosome also move to the CEN of autosomal chromosomes (Rošic´ et al., 2014). However, the RNA signals appear not to have been completely removed from the autosomes after RNase treatment (Rošic et al., 2014 ´ ) suggesting that FISH detected CEN DNA rather than the CEN RNA in trans (Bobkov et al., 2018).

In maize, CEN RNAs identified in IPs of CENP-A are produced from both strands and derived from the 156-bp CentC satellite monomer and transposable elements that are arranged in nearly continuous, intermingled arrays, and clusters. The transcripts are heterogeneous in length (40–200 nt) but predominantly contain 40 and 75-nt species (Du et al., 2010). Although these transcripts lie outside the range of microRNAs or siRNAs (20–30 nt) generated by RNAi pathways, their sizes indicate processing. The CEN RNAs are maintained in a singlestranded state within the maize kinetochore and are firmly bound to centromeric histone protein H3 (Topp et al., 2004), which may protect them from Dicer double-strand cleavage activity. Importantly, genuine siRNAs present in total RNA extracted from maize were not associated with CENP-A chromatin (Du et al., 2010).

Mouse embryonic stem (ES) cells knocked out in dicer-1 (DCR1/1) are defective in global RNAi activity but retained ES cell characteristics. Although viable, they proliferated more slowly (Kanellopoulou et al., 2005). No aberrant chromosome structures or aneuploidy was observed but the cells displayed differentiation defects. The Dicer-negative cells contained increased levels of long, polyadenylated CEN MinSat, and pericentromere MajSat transcripts (>200 nt). Heterozygous mutant cells (DCR1/+) produced 150-nt MinSat and MajSat species, as well as 21–30 nt long specimen, suggesting the contribution of Dicer (Kanellopoulou et al., 2005). Further supporting the involvement of (peri)CEN RNA processing was the identification of protein WDHD1, which may stabilize the

association of Dicer with MinSat and MajSat RNAs (Hsieh et al., 2011).

In tammar wallaby cells, 34–42 nt double-stranded (ds) RNAs with homology to the CEN retroelement kLTR (Ferreri et al., 2011) were identified in small-RNA pools (Carone et al., 2009; Lindsay et al., 2012). In rice, RNAs of ∼40 nt derive from the CentO CEN satellites (Lee et al., 2006). These rice and tammar wallaby CEN RNA species have been termed crasiRNAs (CEN repeat-associated short interacting RNAs). Targeting the small RNAs produced from the kLTR disrupted CENP-A localization in late telophase (Carone et al., 2009; Lindsay et al., 2012). Tight regulation and processing of these crasiRNAs seem integral to the epigenetic framework that is required for CEN establishment.

Hammerhead ribozyme structures associated with transcribed satellite DNA sequences have been identified in salamanders (Epstein and Gall, 1987), schistostome flatworms (Ferbeyre et al., 1998), and Dolichopoda cave crickets (Rojas et al., 2000). All hammerhead ribozymes self-cleave multimeric satellite transcripts into monomer RNAs.

### CENTROMERE PROTEINS THAT BIND TO CENTROMERE RNA

Centromere transcripts or small CEN RNA derivatives underlie the formation of ribonucleoprotein complexes that specify the CEN domains and establish correct kinetochore assembly and architecture. These complexes comprise CENP-A, HJURP, CENP-B, CENP-C, the CPC, and Sgo1. While it is not clear how each protein interacts with the CEN transcripts, CENP-B, CENP-C, Sgo1, and the CPC have in common that their RNA-binding capacity serves as a second chromatin-recruitment mechanism. Indeed, it complements their promiscuous DNA binding activity (CENP-C), their binding to a specific satellite monomer box (CENP-B), their recruitment by the CEN H2A T120P modification (Sgo1), and their recruitment by Thr3-phosphorylated CEN histone H3 and CEN H2A T120P (CPC).

#### CENP-A and HJURP

The interaction between CENP-A and CEN RNA was first observed at a human neocentromere. LINE-1 elements within the CENP-A-binding region of a neocentromere on 10q25 were transcribed into non-coding RNAs that integrate into the CENP-A chromatin (Chueh et al., 2009). Both CENP-A and HJURP interact with CEN RNA as shown in alpha-satellite transcript pull-down experiments (Quénet and Dalal, 2014). In silico predictions of potential RNA-binding sites indicated that 286 out of the 748 HJURP residues, and 79 out of the 140 CENP-A residues, have RNA-binding capacity. However, the vast majority of these CENP-A residues may well be buried inside the nucleosome and/or be bound by CENP-C and CENP-N. The predicted residues lie in the N-terminal half of CENP-A, the protein's most rapidly evolving part (Henikoff et al., 2001; Malik and Henikoff, 2001), which is required to stabilize CENP-A at centromeric nucleosomes (Logsdon et al., 2015). Possibly, the disparities in composition and length of the N-terminal halves of all CENP-A orthologs could allow for their interaction with the rapidly evolving CEN DNA and, consequently, CEN RNA sequences.

# CENP-C

CENP-C, which acts as a dimer via its C-terminal dimerization domain (Cohen et al., 2008), binds to CENP-A, CEN DNA, and RNA (**Figure 3B**). Single-stranded alpha-satellite RNA localizes CENP-C to CEN in interphase, which then together with CENP-A recruits the other kinetochore proteins. Two regions in human CENP-C (one central and one C-terminal) preferably bind to CEN RNAs as shown in competition assays with rRNA, tRNA, and murine MajSat RNA (Wong et al., 2007). However, both sequences also bind to CEN DNA (Sugimoto et al., 1997; Yang et al., 1996). Each DNA-binding element contains a 21–22 amino acid motif via which CENP-C also contacts CENP-A (Kato et al., 2013). Mutating three lysine residues adjacent to CENP-A's central DNA-binding motif also abrogated RNA binding in that region (Wong et al., 2007). Noteworthy, CENP-C's central RNA-binding domain shares homology with the RNA-binding hinge domain region of the pericentromeric heterochromatin proteins (HP) HP1α, β, and γ (Du et al., 2010; Muchardt et al., 2002).

In maize, a C-terminal 122-residue CENP-C region encoded by exons 9–12 binds RNA and DNA, and is required for its CEN localization in vivo. While maize CENP-C binding to CEN RNA occurs without any sequence specificity (in contrast to human CENP-C), CEN DNA binding is stabilized by long ssRNA in vitro. The RNAs that stabilize this contact correspond to the ssCEN RNAs present in kinetochores (Du et al., 2010). Possibly, CEN ssRNA may stabilize CENP-C by enhancing its binding to CEN DNA, adjacent to where it interacts with the CENP-A nucleosome. Indeed, disrupting CEN RNA destabilizes CENP-C at the CEN. Treating mitotic human cells with alpha-amanitin lowered CENP-C levels at kinetochores and caused an increase in lagging chromosomes. A relatively greater reduction of CENP-C occurred on the lagging chromosomes compared to the chromosomes that segregated (Lyn Chan and Wong, 2012). Impeding transcription initiation or splicing also led to decreased CENP-C levels at kinetochores in Xenopus (Grenfell et al., 2016). In Drosophila, X chromosome-specific SatIII transcripts localize to CEN and associate with CENP-C (Rošic et al., 2014 ´ ). Following CENP-C depletion, the SatIII RNA signals at CEN dropped. Reversely, when depleting SatIII RNAs, the presence of newly synthesized CENP-C and CENP-A at CEN was reduced. This negative effect cascaded up through the kinetochore (Rošic et al., ´ 2014). Taken together, results with human cells, Drosophila, maize, and Xenopus suggest that the non-coding CEN RNAs recruit and stabilize CENP-C, supporting CENP-A deposition and stability.

CENP-C bound to CEN DNA and RNA also interacts with chromatin modifying proteins to create the unique epigenetic environment of the CEN domain. CENP-C recruits DNA methyltransferase 3A-B (DNMT3A-B) to reduce local transcription by promoting the methylation of CEN DNA and histone H3. Consequently, CENP-C depletion caused increased CEN transcription (Gopalakrishnan et al., 2009).

CENP-C also binds to MIS18 complex components MIS18αand MIS18-binding protein 1 (Moree et al., 2011; Kim et al., 2012), which control CEN histone acetylation (Fujita et al., 2007). Mis18α through its interaction with DNMT3A-B can also control DNA methylation and histone modifications (Kim et al., 2012), whereas CENP-C through its interaction with M18BP1 promotes the recruitment of HJURP for CENP-A loading (Moree et al., 2011). Possibly, CEN RNA stabilizes CENP-C:DNMT3A-B:MIS18 to target HJURP:CENP-A.

#### The Chromosomal Passenger Complex

The binding of CENP-A and CENP-C to CEN DNA and alpha-satellite RNA promotes kinetochore assembly, including the recruitment of the 4-protein CPC (INCENP, Survivin, Borealin, and Aurora B), which regulates chromosome-spindle attachment and activates the SAC upon chromosome misalignment (Hindriksen et al., 2017). The CPC moves from the inner CEN to the spindle midzone in late anaphase to regulate cytokinesis (Warecki and Sullivan, 2018). Aurora B also phosphorylates CENP-A at Ser7 (Zeitlin et al., 2001). Both proteins coincide at the CEN in metaphase and move to the contractile ring in cytokinesis. Possibly, CEN RNA acts as a scaffold to promote their re-localization.

Knocking down alpha-satellite RNA in human cells (Ideue et al., 2014) or inhibiting transcription in Xenopus egg extracts (Blower, 2016) reduced the CEN levels of Aurora B, resulting in unaligned chromosomes caused by improper spindle attachment. Overexpressing MinSat RNA equally mislocalized Aurora B in murine cells, instigating chromosome misalignment and aneuploidy (Bouzinba-Segard et al., 2006). Moreover, Aurora B kinase activity was regulated by MinSat RNA levels (Ferri et al., 2009). Nonetheless, ectopic overexpression of satellite I RNA did not significantly affect chromosome segregation and CEN functions in human cells (Ideue et al., 2014).

The RNA-dependent inner kinetochore localization of the CPC is mediated by at least two RNA-binding domains: one that is present in Aurora B and one in Survivin or Borealin (Blower, 2016). Aurora B and recombinant CPC also bind to RNA in vitro. RNA stimulates Aurora B kinase activity in vitro and in vivo, and a positive feedback loop exists between its kinase activity and its metaphase localization (Wang et al., 2011; Jambhekar et al., 2014). CPC assembly and Aurora B activity were sensitive to RNase treatment. However, kinase activity was rescued with RNA, perhaps via allosteric effects on Aurora B binding (Ferri et al., 2009; Ideue et al., 2014; Jambhekar et al., 2014). Pull-downs of MinSat RNA from murine cells recovered CENP-A, Aurora B, Survivin, and INCENP (Ferri et al., 2009). Reciprocally, CEN RNAs of murine and human cells co-immunoprecipitated with CENP-A, Aurora B, Survivin, and INCENP (Ferri et al., 2009; Ideue et al., 2014).

Besides CEN RNA, Xenopus Aurora B also interacts with other RNAs (including mRNAs) to form ribonucleoprotein complexes, as observed in anti-Aurora B immunoprecipitation experiments with interphase and mitotic cells, followed by RNA-sequencing. Over 600 RNAs were identified, 465 of which were specific for mitosis, suggesting a cell cycle-regulated binding of target RNA. Identified RNAs encode proteins of the cytoskeleton, centrosome, transcription factors, and RNAs that are enriched on spindle microtubules (Jambhekar et al., 2014). While the RNA pool showed an overrepresentation of adenines, Aurora B interacted rather promiscuously with RNA, and bound in vitro only with minor preference to the Xenopus fcr1 CEN satellite transcript (Blower, 2016).

### HETEROCHROMATIC PERICENTROMERES INSULATE THE CENTROMERE

Centromeric chromatin in fission yeast and metazoans is flanked by constitutive heterochromatin. The pericentromeric domains bind specific proteins and carry epigenetic marks that keep them in a transcriptionally inert state thereby insulating themselves from the enclosed CEN. Pericentric chromatin stabilizes the CEN domain by preventing internal recombinations between intra-CEN repeat sequences (Hetrr and Allis, 2005). It also actively recruits cohesin (via the SUV4-20H2 methyltransferase enzymes that trimethylate histone H4 at Lys20) to promote the bi-orientation of and tension development between the sister chromatids (Bernard et al., 2001; Sakuno et al., 2009; Yamagishi et al., 2010; Yi et al., 2018).

Similar to the CEN sequence, pericentromeres comprise simple repeat sequences such as alpha-satellite DNA, beta-, gamma-, I, II, and III satellite sequences (5–200 bp). They further contain DNA transposons (1 kb), long terminal repeat (LTR)-endogenous retroviral elements (10 kb), non-LTR autonomous retrotransposons (transposons that are formed after reverse transcription of an intermediate RNAPIII-generated transcript) including long interspersed elements (LINEs, 6 kb) and short interspersed elements SINE (100–300 bp) (**Figure 2C**). Pericentromeres harbor promoter elements that recruit various transcription factors, including Ikaros in human cells (Gurel et al., 2008), the ubiquitous YY1 at murine gamma-satellites (Shestakova et al., 2004), Nanog and Sall1 in mouse ES cells (Lopes Novo and Rugg-Gunn, 2016) to regulate transcription by RNAPII or RNAPIII (Pezer and Ugarkovic, 2008 ´ ). The repeat sequences are not conserved between or within a species, suggesting that pericentromere transcription is epigenetically controlled. Indeed, it contains histone H3 variants H3.3 and H2A.Z (Drané et al., 2010; Santenard et al., 2010) and binds the conserved HP1, which propagates the heterochromatic state and coordinates chromatin silencing, cohesion, and replication activities (Saksouk et al., 2015). The pericentric histones are hypoacetylated, resulting in chromatin fiber compaction. Methylation marks are enriched on histone H3; H3K9me2, H3K9me3 (recognized by HP1), H3K27me2, and H3K27me3, but also on histone H4; H4K20me2, H4K20Me3, and on cytosine and adenine (Gopalakrishnan et al., 2009; Rose and Klose, 2014; **Figure 3A**). Notwithstanding this repressive environment, pericentromeres are transcribed in many organisms. A delicate balance between pericentromere and CEN transcription ensures chromosomal stability (see next).

## PERICENTROMERE TRANSCRIPTION AND TRANSCRIPT PROCESSING ENSURE ITS SILENT STATE

In S. pombe, repressive H3K9 methylation occurs at the outermost dg and dh pericentromere repeats and ends at the tRNA clusters inside the innermost repeats that surround the CEN's central domain. Their presence prevents the pericentromeric heterochromatin from expanding into the CENP-A chromatin (Cam et al., 2005; **Figure 4**). The tRNA clusters are transcribed by RNAPIII, which further delineates the CEN core domain from the flanking pericentromeres (Partridge et al., 2000; Scott et al., 2006). RNAPIII barrier transcription activity does not depend on the orientation of the tRNA genes, but on the DNA sequence that is required for formation of the RNAPIII complex (Scott et al., 2006, 2007). The retrotransposon SINE, found throughout the mammalian genome, is also transcribed by RNAPIII at pericentromeres. SINE expression has been linked to establishing boundary elements and chromatin insulators across the genome (Lunyak et al., 2007; Román et al., 2011). Similarly, SINE transcription and/or that of other pericentric DNA elements could insulate the CEN from the bulk chromatin.

Transcription of pericentromeric chromatin occurs in many species and, except for the tRNA genes in fission yeast, is largely devoid of protein-encoding sequences (Brown et al., 2012; Hall et al., 2012; Saksouk et al., 2015). In S. pombe, small-interfering RNAs (siRNAs) produced after the processing of longer transcripts are required for the propagation and maintenance of the heterochromatic identity of pericentromers (Volpe et al., 2002). The finding that transcription of pericentromeric chromatin is functionally significant led to a re-assessment of the definition of "silent" heterochromatin. Specifically, RNAPII bi-directionally produces pre-RNAs from cryptic and TATA-like promoter sequences within the dh and dg elements of the otr regions that border the central CEN (Reinhart and Bartel, 2002; Djupedal et al., 2005; Kato et al., 2005; **Figure 4**). Both otr elements are not required for CEN function during mitosis but provide a platform for the heterochromatin component of S. pombe CEN (Kagansky et al., 2009). The produced single-stranded polyadenylated transcripts are converted into dsRNA species by the RNA-directed RNA polymerase-containing RDRC complex, which Dicer (Dcr1) next processes into short siRNAs that are transferred by ARC (Argonaute siRNA chaperone complex) to the Argonaute (Ago1)-containing RNAinduced transcriptional silencing complex RITS (Volpe et al.,

RNAPII of an outer repeat otr element that flanks the central core of the CEN. Lowe panel: Regulation of the heterochromatic state of CEN sequences that flank the central core domain. The RNA-induced transcriptional silencing (RITS) complex binds to ssRNA transcripts generated from the otr sequence repeats, by siRNA–RNA base pairing interactions and via nucleosomes by localizing to histone H3 methylated at Lys9 (H3K9me). RITS then recruits RDRC/Dicer activity, promotes dsRNA synthesis, the production of siRNAs, and CLRC H3K9 methyltransferase-mediated H3K9 methylation. The Argonaute siRNA chaperone complex (ARC) catalyzes the transfer of the siRNAs from RDRC/Dicer to the RITS complex. The transcript ssRNAs present in the siRNAs become degraded by the exosome. Chromodomain HP1 proteins Swi6 and Chp2 are recruited by the H3K9me mark and silence transcription of the chromatin by localizing the chromatin remodeling Snf2/HDAC repressive complex (SHREC), which inhibits RNAPII activity. Adapted from Holoch and Moazed (2015).

2002; Martienssen et al., 2005). Through their interaction with Ago1, the siRNAs load RITS onto the cognate pericentromeric chromatin via base-pairing with the nascent transcripts. The RITS complex then recruits the CLRC complex that contains the histone methyltransferase Clr4 (SUV39H in mammals), which methylates H3K9. The latter recruits chromodomain proteins Swi6 (S. pombe HP1 ortholog) and the SHREK-associated protein Chp2, as well as histone deacetylase Clr3 (HDAC1), which removes the local permissive H3K14ac marks. The SHREK complex inhibits RNAPII activity, resulting in silent heterochromatin (**Figure 4**).

Deleting RNAi pathway genes (dcr1, ago1 or RdP1) caused chromosome missegregation due to defective silencing of the pericentromeric heterochromatin. S. pombe strains mutated in RNAPII subunits Rpb2 and Rpb7 also suffered from increased chromosomal instability, impaired transcriptional silencing, and a reduced association of H3K9me and Swi6 at dg/dh (Djupedal et al., 2005; Kato et al., 2005). Pericentromere transcription and siRNA production in S. pombe peak in S-phase. Hence, pericentromere silencing may be alleviated in S-phase as heterochromatin markers H3K9me and Swi6 become distributed on the replicated strands (Chen et al., 2008; Kloc et al., 2008). Without RNAi, homologous recombination repairs the stalled forks (Zaratiegui et al., 2011) suggesting that transcriptional silencing of pericentromeric heterochromatin prevents replication stress (Castel and Martienssen, 2013).

The importance of Dicer-dependent processing of pericentromere RNAs for heterochromatin assembly in vertebrates was demonstrated with chicken DT40 cells carrying a human chromosome (Fukagawa et al., 2004). Eliminating Dicer provoked an accumulation of long pericentric alpha-satellite and SatIII transcripts, and caused mitotic defects due to precocious sister chromatid separation; attributed to HP1 loss and a misregulation of cohesin and SAC protein BubR1. Similarly, conditionally depleting Dicer in mouse ES cells led to an accumulation of short MajSat transcripts (40 to >200 nt) and the normally repressed long interspersed repeated DNA and high-copy-number LTR retrotransposons. These findings indicated a role for Dicer in repressing pericentromere regions and other usually silent genetic elements (Kanellopoulou et al., 2005). Since the binding of HP1 to heterochromatin requires RNA (Maison et al., 2002; Muchardt et al., 2002), the Dicer-processed siRNAs were assumed to represent them. However, other than in chicken cells (Fukagawa et al., 2004), 21–25 nt siRNAs deriving from the pericentromeric domains have been difficult to identify in vertebrates. Irrespective of how or if the RNAi pathway contributes, pericentromere transcripts in mammals seem involved in the formation and maintenance of heterochromatin. For example in mice, protein WDHD1, which plays a role in RNAPII transcription and RNA processing, interacts with MajSat transcripts. Depleting WDHD1 enhanced MajSat levels and reduced pericentromeric heterochromatin condensation, resulting in proliferation defects (Hsieh et al., 2011). Additional work with mouse early embryos showed that injections of satellite dsRNAs can localize HP1β to pericentromeres revealing that HP1 is targeted in an RNA-dependent, sequence-specific manner. However, a functional association with the RNAi machinery was not assessed (Santenard et al., 2010).

Long non-coding transcripts corresponding to several MajSat satellite repeat units specifically associate with SUMOylated HP1, which is stabilized by H3K9me3, in murine cells. RNase treatment released HP1 and altered the spacing of the pericentromeric histones. HP1 preferentially binds to the forward strand of these RNAs, which remains bound to the site of transcription. Additional HP1 molecules then accumulate, connecting pericentromere transcription with heterochromatin formation (Maison et al., 2011). In primary mouse embryonic fibroblasts, pericentromeric heterochromatin transcription is proliferation- and cell cycle-dependent (Lu and Gilbert, 2007). A first pool of long, heterogeneous MajSat transcripts (1 kb to >8 kb) is produced by RNAPII through G1 and peaks in G1/S-phase, right before pericentromere replication (mid-to-late S-phase). Since the transcripts accumulate at the site of pericentromere replication, local transcription could promote heterochromatin reassembly at the replication fork. A pool of shorter transcripts (∼200 nt) is produced at mitotic onset, coinciding with transcription factors and other proteins becoming cleared from the heterochromatin. This transcript population/transcription activity could be involved in heterochromatin formation, maintenance, and reinforcement during the later stages of mitosis when cohesin at pericentromeres has been removed (Wu et al., 2006). Indeed, while HP1 is dispatched from heterochromatin during M-phase (Muchardt et al., 2002; Fischle et al., 2005), H3K9me3 and the short M-phase RNAs could contribute to the anaphase recruitment of HP1 (Saksouk et al., 2015). SUV39 (Suv39h) histone lysine methyltransferase promotes constitutive heterochromatin compaction and transcriptional repression by catalyzing the H3K9me2/3 modification in humans and mice. SUV39 is incorporated and stabilized in constitutive heterochromatin by chromatin-associated non-coding RNAs (Johnson et al., 2017; Velazquez Camacho et al., 2017).

Heterochromatin activity in D. melanogaster is also associated with histone H3K9 methylation by Su(var)3-9 and HP1 recruitment (Ebert et al., 2006). Involvement of siRNA pathways acting in heterochromatin formation in Drosophila has been hypothesized since a nuclear pool of transposable element-derived siRNAs (21 nt) was shown to promote heterochromatin formation in somatic cells of Drosophila. Components of the RNAi pathway contributed to heterochromatin maintenance (Fagegaltier et al., 2009). As in S. pombe and mammals, these siRNAs might tether complementary nascent transcripts of satellite DNAs and transposons, and guide chromatin-modifying enzymes, including Su(var)3-9. RNAi activity seems to help establish heterochromatin in the early embryo, which can then be maintained in the absence of RNAi in somatic tissues (Huisinga and Elgin, 2009). Contrary to D. melanogaster, plants often contain a significant portion of methylated repetitive DNA. In fact, siRNAs guiding the methylation of histones and DNA at the loci they were derived from (Zakrzewski et al., 2011). Processing

of satellite-derived transcripts by the RNAi pathway into siRNAs (21–24 nt) has been reported for Arabidopsis, rice, and sugar beet (May et al., 2005; Lee et al., 2006; Zakrzewski et al., 2011). Small RNAs with a predominant size of 24 nt cognate to the satellite TCAST (Ugarkovic et al., 1996; ´ Feliciello et al., 2011) have been detected in the beetle Tribolium castaneum and are more abundant in embryos than in later developmental stages (Pezer and Ugarkovic, 2008 ´ ; Pezer et al., 2012). The sequences of components of the RNAi pathway are present in the genome of T. castaneum, including Argonaute and Dicer, but not the RNA-dependent RNA polymerase gene (Tomoyasu et al., 2008), which insects and vertebrates appear to lack.

#### CENTROMERE AND PERICENTROMERE TRANSCRIPTION DURING DEVELOPMENT AND DIFFERENTIATION

Satellite DNA has been associated with differentiation and development. Repetitive DNA is not transcribed in adult tissues presumably because it is hypermethylated (Jeanpierre et al., 1993) while it is hypomethylated in fetal tissues (Miniou et al., 1997). Antisense MajSat transcripts accumulate in the central nervous system of mouse embryos 11.5 days post coitum (dpc), and become replaced by sense MajSat transcripts from 12.5 until 15.5 dpc. In adult mice, MajSat transcripts were identified only in highly proliferative tissues such as liver and testis (Rudert et al., 1995). In chicken and zebrafish, alpha-satellite expression from the sense and antisense strands occurs in a regulated pattern during embryogenesis, possibly to control gene expression following transcript processing (Li and Kirby, 2003). Before headfold formation in the chick and at 0–2 h post-fertilization (hpf) in zebrafish, blastodiscs expressed the alpha-repeat sequences. By stage 9 and at 6–8 hpf, respectively, the expression localized to the head mesoderm, myocardium, pharyngeal endoderm, and cardiac neural crest. Because the expression occurred so early in zebrafish, the authors looked for the alpha-repeat transcripts within the maternal RNAs in single-cell and four-cell stage embryos. These stages occur within minutes of fertilization and before the start of zygotic transcription at 3 hpf. High levels of the transcripts were found, supporting their maternal origin (Li and Kirby, 2003).

# ANOMALOUS CENTROMERE AND PERICENTROMERE TRANSCRIPTION DURING STRESS AND DISEASE

Since the centromeric and pericentromeric regions are epigenetically controlled, any loss/reduction in repressive marks such as DNA and histone methylation or increased removal of active acetylation marks can provoke satellite overexpression from the centromeric and pericentromeric regions as observed during stress, senescence, aging, and in cancer cells. Pathological transcription of either region dramatically affects CEN insulation and activity, resulting in disturbed kinetochore formation and genetic instability.

# (Peri)centromere Transcription During Stress

In human cells, the transcription of certain pericentromeric satellite sequences, in particular SatIII, is induced upon heat shock and exposure to heavy metals, chemicals, UV radiation, hyperosmotic, or oxidative conditions (**Figure 5A**). Importantly, while SatIII transcripts were up-regulated following heat shock, CEN transcripts were not, indicating that each domain is subject to different transcriptional control mechanisms (Jolly et al., 2004; Rizzi et al., 2004; Valgardsdottir et al., 2008; Eymery et al., 2009). SatIII expression levels also depend on the type of stress that is experienced: MMS, etoposide, aphidicolin, and oxidative stress are weak inducers; UV and hyperosmosis have a moderate effect; and heat shock and cadmium are very strong activators. In unstressed cells, SatIII sequences exist in a transcriptionally silent, closed heterochromatin conformation. Following heat shock or stress, SatIII transcription is induced (Valgardsdottir et al., 2008). Specifically, monomeric transcription factor Heat Shock transcription Factor 1 (HSF1) becomes upregulated and binds as a phosphorylated homotrimer to the SatIII sequences. HSF1 then recruits the histone acetylase CREB-binding protein CBP to trigger histone hyperacetylation while the death domainassociated protein DAXX, which acts as a chaperone for pericentromeric histone H3.3, promotes SatIII transcription by RNAPII. Upon DAXX depletion, SatIII expression levels in heat-shocked cells dropped, while less H3.3 was incorporated (Morozov et al., 2012). A set of RNA-binding and processing proteins associate with the SatIII transcripts. RNAi knock-downs of these transcripts that range between 2 and 5 kb (Jolly et al., 2004; Rizzi et al., 2004) reduced the recruitment of RNA processing factors, including the splicing factor SF2/ASF (Chiodi et al., 2004; Metz et al., 2004). The RNA-binding factors and SatIII transcripts produce ribonucleoprotein complexes that combine into many perichromatin granules. Together, they correspond to mature nuclear stress bodies that accumulate at the pericentromeres (Denegri et al., 2002; Jolly et al., 2004; **Figure 5A**). The number and size of the nuclear stress bodies correlate directly with SatIII expression (Valgardsdottir et al., 2008). During recovery from the stress, increased levels of heat shock protein HSP70 trigger the disassembly of the HSF1 trimers, which leave the nuclear stress bodies together with the histone acetyltransferase CBP and RNAPII. Next, the granule clusters dissociate, the RNA-binding proteins redistribute through the nucleoplasm but the SatIII transcripts stay bound to the granules. At the same time, granules that are H3K9 methylated appear adjacent to the disassembling nuclear stress bodies. The transcripts are then cleaved, and a complex similar to the S. pombe RITS complex may then localize the transcripts to the chromatin to silence the SatIII DNA arrays (Biamonti, 2004; Biamonti and Vourc'h, 2010; **Figure 5A**). Depending on the stress that is experienced, different transcription factors promote SatIII activation. For example, the tonicity-responsive enhancer binding protein TONEBP induces SatIII expression under hyperosmotic stress

(Valgardsdottir et al., 2008). Satellite transcript accumulation during heat stress also occurs in insects (Pezer et al., 2012) and plants (Tittel-Elmer et al., 2010). In the beetle T. castaneum pericentromere TCAST satellites are transcribed by RNAPII and processed into 21–30 nt siRNAs. The production of these siRNAs is developmentally regulated but is strongly induced upon heat shock. During recovery, siRNA expression and histone modifications are restored to normal. Transient heterochromatin remodeling seems part of a stress-activated gene-expression program in beetles (Pezer et al., 2012), and possibly other organisms as well. In Arabidopsis, a temperature upshift alleviated the silent state of CEN satellite sequences, pericentric 5S rDNA arrays, transposable elements, and 106B interspersed repeats. Surprisingly, the pattern of repressive epigenetic marks within the heterochromatin was not affected, suggesting that the temperature-stimulated transcription activity bypassed these regulatory modifications (Tittel-Elmer et al., 2010).

Centromere MinSat transcription increases when murine cells are exposed to chemical stress (DMSO, 5-aza-2<sup>0</sup> -deoxycytidine, apoptosis inducer staurosporine). Comparable ectopic overexpression of MinSat DNA led to decondensed CEN and mitotic defects such as multiple spindle attachments, loss of sister chromatid cohesion, aneuploidy, and cell death (Bouzinba-Segard et al., 2006).

Extensive evidence points to an interplay between the DNA damage response and satellite DNA expression. Ectopic expression of satellite RNA in cultured human mammary epithelial cells induced numerous foci of γ-H2A.X, the phosphorylated histone H2A.X variant that marks dsDNA breaks (Zhu et al., 2011). These cells also exhibited bridged and lagging chromosomes as well as disorganized metaphase spindles (Zhu et al., 2018). Similarly, MajSat RNA overexpression compromised DNA damage repair, resulting in high DNA mutation rates in cultured murine pancreatic cells (Kishikawa et al., 2016a, 2018). Elevated levels of γ-H2A.X were also observed after nuclear injection of satellite RNA in human cells, indicating that high transcription intensity per se does not trigger the DNA damage response (Zhu et al., 2018). CRISPR-mediated activation of MinSat and MajSat expression in murine cells incited chromosomal instability (Zhu et al., 2018). Vice versa, genotoxic etoposide treatment (causes dsDNA breaks) triggered MinSat transcription and CENP-A eviction from the mouse CEN, which relied on the p53-dependent DNA damage pathway and chromatin chaperone/remodeling factors (Hédouin et al., 2017). In the absence of functional p53, DNA demethylation as induced by 5-aza-2<sup>0</sup> -deoxycytidine resulted in massive transcription of MajSat RNA in mouse fibroblasts (Leonova et al., 2013). Moreover, ectopic overexpression of MajSat RNA stimulated by injection of sh-p53 RNA causing p53 knockdown led to tumor formation (Zhu et al., 2018).

How do (peri)centromere transcription/transcripts contribute to a stress response and recovery from it? The transcripts processed via an RNAi-dependent or -independent pathway could mediate heterochromatin reformation, as in S. pombe. Analogous to X-chromosome inactivation by the long non-coding Xist RNA, the transcripts might recruit chromatin remodelers and DNA methyltransferases to establish a silent pericentric state. Also, SatIII transcripts could protect a fragile region of the genome from stress-induced DNA damage (the SatIII-enriched 9q12 region is often rearranged in pathologies, including cancer). Possibly, the transcripts could regulate local RNA splicing during the stress response by sequestering splicing factors. Via a position-effect mechanism, they might counteract the repressive nature of heterochromatin and activate nearby genes in cis or trans (Eymery et al., 2009; Saksouk et al., 2015).

# (Peri)centromere Transcription During Senescence and Aging

Heterochromatin structure and expression change during aging. An up-regulation of MajSat expression in senescent cardiac muscle cells of aging mice but not in their brain or kidneys may be linked to mitochondria-induced oxidative stress (Gaubatz and Cutler, 1990). Transcriptional activation of pericentromeres has been observed in replicative senescence and aging. Upon extensive passaging of human fibroblasts, the cells entered replicative senescence, which correlated with an increased expression of pericentromere transcripts. The pericentromeric heterochromatin was decondensed and exhibited reduced DNA methylation. Here, the produced transcripts may not have a specific biological role but rather be the consequence of a senescent state of the heterochromatin itself (Enukashvily et al., 2007). Aberrant overexpression of SatIII from 9q12 was observed in the Hutchinson–Gilford progeria syndrome (Shumaker et al., 2006). The latter arises from mutations in the laminA gene, which encodes a component of the nuclear lamina that maintains the structural integrity of the nucleus. Lamins are crucial for pericentromeric heterochromatin organization in interphase cells (Solovei et al., 2013).

Pericentromeric heterochromatin was show to lose H3K9me3 and HP1 proteins in older flies and human cells, leading to an anomalous expression of satellite sequences (Scaffidi and Misteli, 2006; Shumaker et al., 2006; Larson et al., 2012). Loss of pericentric silencing may drive age-related genome instability and death since the cells from older individuals or progeria patients are characterized by a global loss of heterochromatin marks, and increase in DNA damage (Scaffidi and Misteli, 2006; Shumaker et al., 2006; Leung et al., 2015). Inactivation of heterochromatin silencing components in flies cuts their lifespan in half whereas a moderate overexpression of HP1α extends their lifespan with 15%, suggesting that HP1α loss in older animals promotes aging (Larson et al., 2012). Finally, the transcriptional de-repression of satellite sequences has been linked to tau-induced neurodegeneration, as in Alzheimer's disease (Frost et al., 2014).

### (Peri)centromere Transcription in Cancer and Disease

The transcriptional misregulation of the SatII and SatIII pericentromeric satellite sequences, and altered epigenetic state of pericentromeric chromatin characterizes many cancers and genetic disorders (Shumaker et al., 2006; Ehrlich, 2009; Eymery et al., 2009; Ting et al., 2011; Zhu et al., 2011; **Figure 5B**).

FIGURE 5 | (A) Regulation of pericentromere SatIII transcription in human cells following exposure to heat and other stresses. In the absence of stress, SatIII repeat sequences are epigenetically marked for silence (H3K9me, pink dot) and exist in a closed transcriptionally inert state (blue nucleosomes). Upon exposure to heat or other stresses, the monomeric HSF1 (shown in green) becomes upregulated, and forms homotrimers that after phosphorylation enter the nucleus. The HSF1 bind to the SatIII sequences and recruit the histone acetyltransferase (HAT) CREB-binding proteins to trigger histone hyperacetylation (yellow stars), which results in active SatIII transcription by RNAPII of one strand. A subset of RNA-binding/processing proteins is recruited to the SatIII transcripts, forming ribonucleoprotein complexes that associate into so-called perichromatin granules, which in turn produce clusters that correspond to a mature nuclear stress body (represented by the red oval structure). To recover from the inflicted stress, heat shock protein HSP70 induces the disassembly of the HSF1 trimers that leave the nuclear stress bodies, along with RNAPII and the HAT. The granules disassemble and the RNA-binding proteins redistribute throughout the nucleoplasm. SatIII transcripts may become processed into smaller fragments possibly by the RNAi machinery to protect and re-establish the heterochromatic state of the pericentromeric region comprising the SatIII repeats, possibly by recruiting epigenetic writing activity resulting in the establishment of the repressive H3K9me signals. Adapted from Biamonti and Vourc'h (2010) and Biamonti (2004). (B) Transcriptional regulation of (peri)centric repeat sequences as identified in various tumors. The epigenetic marks and the enzymes responsible for introducing or removing them at histones or cytosine are indicated. Blue nucleosomes: silent, purple nucleosomes: transcribed. The overproduction of (peri)centric transcripts can induce DNA damage, mitotic defects, genomic instability, and aneuploidy. See text for details.

In mouse models of pancreatic, colon, and lung cancers, satellite transcripts represent up to 50% of the total RNA, which was linked to deregulated DNA methylation. Specifically, in pancreatic ductal adenocarcinoma (PDAC) samples, 47% of all transcripts were produced from MajSat sequences. In contrast, in healthy reference tissues, only 0.02–0.4% of all transcripts originated from those repeats (Ting et al., 2011; Kishikawa et al., 2016a). The transcripts were highly heterogeneous (200–8,000 nt) and transcribed only from the forward strand. While PDAC murine cells expressed MajSats only minimally when cultured ex vivo, high expression levels similar to those observed in tumors in vivo were measured in immortalized PDAC tumor cells treated with 5-aza-2<sup>0</sup> -deoxycytidine, suggesting that transcription is regulated by DNA methylation, which might be re-established together with other epigenetic silencing mechanisms ex vivo. Furthermore, SatII expression showed a median 21-fold increase in human PDAC samples in comparison with "healthy" tissue samples (Ting et al., 2011). To determine what could be promoting SatII hyperexpression, linear regression analysis was performed to identify transcripts that were co-regulated with the mouse MajSat or human alpha-satellite sequences. Several genes involved in neuronal cell fate and stem cell pathways, that contained LINE1 transposable elements were highly expressed (Ting et al., 2011). A LINE1 insertion upstream of their transcription start sites can underlie their misregulation, contributing to cellular transformation. SatII RNA transcripts in colorectal cancer cells were reverse transcribed into DNA:RNA hybrids, and then generated dsDNAs, which were incorporated into the pericentromeric domains. Whole-genome sequencing showed that SatII copy number gain commonly characterizes human colon tumors, and is linked with low survival (Bersani et al., 2015). Healthy human testis tissue showed a high expression of pericentromeres, while in cancers their expression was silent (Eymery et al., 2009).

Methyltransferase DNMT3B, which methylates (peri)centromeric DNA at cytosines in CpG dinucleotides, is recruited by CENP-C. Impairment of this interaction causes an overproduction of CEN and pericentromere transcripts (Gopalakrishnan et al., 2009). Besides cancer, mutations in DNMT3B lead to the ICF syndrome (immunodeficiency, CEN instability, and facial anomalies) whose patients suffer from hypomethylated SatII and SatIII repeats (euchromatic gene methylation was at normal levels, Brun et al., 2011). The tumor-suppressing, heterochromatin-associating lysine demethylase 2A (KDM2A) is downregulated in prostate cancer (Frescas et al., 2008). Via its Jumonji domain, the enzyme demethylates the pericentromeric H3K36me2 modification to silence the heterochromatin. KDM2A depletion resulted in a loss of HP1 and elevated alpha-satellite and MajSat transcription in human and mouse cells, respectively. Phenotypes included genomic instability, sister chromatid misalignment, chromosome breaks, and anaphase bridges. The lower the level of KDM2A expression, the more severe the tumor grade in prostate cancer, linking hypermethylation and increased (peri)centromere transcription with cancer growth (Frescas et al., 2008; **Figure 5B**).

The histone demethylase JMJD2B acts as an oncogene in certain breast cancers (Slee et al., 2012). When overexpressed, its activity reduces H3K9me3 marks at CEN and causes chromosomal instability. While the levels of CEN and pericentromere transcripts in these tumors were not quantitated, their production was likely derepressed. A loss of SUV39H histone methyltransferase activity (mediates H3K9 methylation) facilitated the expression and/or stabilization of MinSat transcripts in mice, which accumulated as dsRNAs (Lehnertz et al., 2003; Martens et al., 2005). A forced accumulation of MinSat transcripts, in sense orientation, provoked a mislocalization of kinetochore proteins, affected chromosome segregation, sister chromatid cohesion, and induced modifications of CEN epigenetic hallmarks. Possibly, anomalous levels of CEN transcripts interfere with kinetochore and cohesin recruitment (Bouzinba-Segard et al., 2006). Of note, ectopic overexpression of alpha-satellite DNA in human cells led to chromosome loss but not to reduced methylation of the DNA. In contrast, DNA demethylation caused pathological alpha-satellite transcription and chromosome loss in human cells (Ichida et al., 2018).

SatII and SatIII transcripts were markedly overexpressed in human osteosarcoma cells depleted in tumor suppressor SIRT6, which deacetylates histone H3K16ac in pericentric heterochromatin. Its inactivation led to H3K18 hyperacetylation likely by the histone acetyltransferase GCN5, reversal of heterochromatin silencing, mitotic defects, genomic instability, and senescence. Importantly, depletion of the transcripts through RNAi rescued the phenotypes (Tasselli et al., 2016).

Mutations in the hereditary ovarian and breast cancer susceptibility gene BRCA1, which acts as a tumor suppressor, led to genomic instability. While BRCA1 acts in DNA replication and damage repair, control of the cell cycle, and many other regulatory functions, the protein was recently shown to also determine the epigenetic states of centromeric and pericentric chromatin (Zhu et al., 2011). Through its ubiquitin ligase activity, BRCA1 mono-ubiquitinates histone H2A at Lys119 (Chen et al., 2002; **Figure 5B**) to produce a repressive mark that prevents genomic instability and tumorigenesis (Zhu et al., 2011). When BRCA1 was knocked out in murine and human cells, a derepressed transcription of MinSat, MajSat, and alpha-satellite DNA was observed, respectively, concurrent with a loss of H2AK119 ubiquitination. While the latter may have produced defective heterochromatin (as indicated by reduced HP1 levels), it is unclear which factors promoted alpha-satellite transcription in the BRCA1-deficient cells. Ectopically expressing H2A fused to ubiquitin reversed the above BRCA1-loss phenotypes, whereas the ectopic expression of satellite DNA phenocopied it, resulting in DNA damage and genomic instability, cell cycle checkpoint defects, and centrosome amplification, indicating that overexpressed (peri)centromere transcripts could contribute to malignancy (Zhu et al., 2011).

The tumor-suppressing transcription factor Prep1 is associated with DNA damage control and the management of histone methylation levels (Iotti et al., 2011). Indeed, upon downregulating Prep1 in mouse or human cells, DNA damage increased. This phenotype, which is generated through an unknown mechanism, caused a widespread increase in

the repressive histone mark H3K9me3. Consequently, the transcription of MajSat in mouse, and alpha-satellite DNA in humans dropped with 62% and 45%, respectively, compared to wild-type control cells. Intriguingly, the decrease in CEN and pericentromere transcript production led to the same phenotypes as in cells overexpressing them, including aneuploidy, miniature chromosomes, Robertsonian translocations, and CEN duplications (Iotti et al., 2011).

Tumor-suppressing transcription factor p53 cooperates with DNA methylation activity to silence a large part of the mouse genome. A massive transcription of major classes of retroelements, near-CEN tandem repeat satellite DNAs, and numerous species of non-coding RNAs was observed in p53-deficient mouse fibroblasts treated with 5-aza-2<sup>0</sup> -deoxycytidine (not observed in treated p53 wildtype cells). The levels of these transcripts exceeded those of β-actin mRNA by more than 150-fold. Accumulation of these transcripts, which are capable of forming dsRNAs, was complemented by a potent apoptosis-inducing type I interferon response. The authors suggested a model in which the downregulation of these repeat sequences is controlled by p53-driven transcriptional silencing, DNA methylation-based suppression of transcription, and the suicidal type I interferon response, which eliminates the cells that escaped the first two lines of control (Leonova et al., 2013).

(Peri)centromere silencing is also regulated by the Polycomb repressive complexes PRC1 and PRC2, which are commonly misregulated in cancer (Blackledge et al., 2015). PRC2 lysine methyltransferase subunit EZH2 catalyzes the addition of one to three methyl groups to histone H3 at Lys27 (**Figure 5B**). In Rb1 mutant mice, which are defective in recruiting EZH2 to repetitive sequences, a transcriptional derepression of satellite DNA was observed, which induced susceptibility to lymphoma (Ishak et al., 2016).

# (PERI)CENTROMERE TRANSCRIPTS AS CANCER BIOMARKERS AND TARGETS IN THE CLINIC

SatII overexpression characterizes myriad cancerous and precancerous lesions, suggesting that SatII RNA levels might be a good predictor or indicator of cancer (Ting et al., 2011; Bersani et al., 2015; Tasselli et al., 2016; Hall et al., 2017). Indeed, RNA in situ hybridization analysis of SatII expression in biopsies proved a better diagnostic for pancreatic cancer than standard histopathological analysis (Ting et al., 2011). A convenient and highly sensitive method for quantitating circulating satellite repeat RNAs in blood serum (Kishikawa et al., 2016b) combines Tandem Repeat Amplification by nuclease Protector (TRAP) with droplet digital PCR. Patients with pancreatic ductal carcinoma (PDAC) were efficiently discriminated from healthy individuals, while patients with intraductal papillary mucinous neoplasm, a precancerous lesion of PDAC, could also be accurately identified. This simple and cheap test allows for early prognosis, quick screens, and regular follow-ups of PDAC progression. This method may well be adapted to quantitate additional (peri)centromere transcripts in other cancers as well.

Kinetochore subunit overexpression (Thiru et al., 2014; Zhang et al., 2016; Sun et al., 2016), CENP-A overproduction and mislocalization (Athwal et al., 2015), and de-silenced (peri)centromeric chromatin may all contribute to aneuploidy and promote cancer initiation/progression. The degree of overexpression of kinetochore protein-encoding genes, which associate with patient survival and response to therapy, could classify tumors, and serve as future prognostic cancer biomarkers (O'Brien et al., 2007; Sun et al., 2016; Zhang et al., 2016). Similarly, CEN and pericentromere transcript levels in conjunction with the (peri)centromeric methylation, acetylation, or ubiquitination state may serve as valuable readouts of cancer grade and survival. They may also represent novel therapeutic targets. In fact, various drugs inhibiting numerous epigenetic enzymes/regulators are in advanced developmental stages (Pfister and Ashworth, 2017). Nucleic acid therapeutics aimed at (peri)centromere repeats may provide alternative objectives for the future. They are transcribed in a cell- or tissue-specific manner, making them exceptional objectives. Powerful RNA structure determination assays can also map the secondary and tertiary structure of these RNAs (Wilkinson et al., 2006; Novikova et al., 2013; Lu et al., 2016). Clinical trials are already underway to similarly target highly structured bacterial or viral riboswitches using small-molecule inhibitors to treat bacterial and viral infections, respectively (Howe et al., 2015). Small-molecule ligands targeting structural elements in these CEN or pericentromere RNAs could potentially destabilize the transcript or interfere allosterically with CEN-protein binding to confer a therapeutic effect, although this remains purely hypothetical. With recent advances in genome editing methods, it is possible to achieve transcriptional silencing of (peri)centromere repeats via CRISPR interference (Gilbert et al., 2014; Koch, 2017). In a genome-wide CRISPR interference study, guide RNAs were developed to selectively and successfully inactivate >16,000 long non-coding RNA genes within the human genome (Liu et al., 2017). These experiments suggest that downregulating pathologically expressed (peri)centromeric elements could well be feasible (Zhang et al., 2016; Koch, 2017). The recently developed CRISPR/Cas13 system (Abudayyeh et al., 2017) represents another promising approach to knock down non-coding RNAs. However, only the future will tell to which extent these approaches will translate into clinical scenarios.

# PERSPECTIVES

The continuous identification and functional characterization of new epigenetic activities (enzymes, histone modifications) that impinge on centromeric and pericentromere domains via ever more sensitive mass spectrometry approaches will take our understanding of CEN, kinetochore, and pericentromere biology to the next level. In addition, transcription factors that drive (peri)centromere transcription in healthy and diseased

cells must be identified as their biology and influences on the spatiotemporal regulation of the CEN and pericentric regions remains largely unknown. The same is true for regulators that act upon the RNAPII complex to orchestrate its activity (recruitment, elongation, termination) at (peri)centromeres. Kinases and phosphatase may be prime candidates. A better understanding of RNAi pathway involvement in mammalian biology would be welcomed as well. For sure, exciting (peri)centromere biology will continue to be "written" in laboratories worldwide, and hopefully at some point in cancer clinics as well.

#### REFERENCES


#### AUTHOR CONTRIBUTIONS

KS and PDW wrote the paper and made the figures. Both authors have approved of the manuscript.

# FUNDING

We acknowledge support from the Italian Association for Cancer Research (Investigator Grant IG19250) and the University of Trento (Intramural Grants 40202054 and 40201806).




DNA sequences. Mol. Cell 64, 1074–1087. doi: 10.1016/J.MOLCEL.2016. 10.021



by polypurine/polypyrimidine DNA probes and antitriplex antibodies. Chromosoma 111, 201–213. doi: 10.1007/s00412-002-0198-0


Acta Gene Regul. Mech. 1839, 1362–1372. doi: 10.1016/J.BBAGRM.2014. 02.007


inversely regulate differentiation. Cell 152, 584–598. doi: 10.1016/j.cell.2013. 01.009


Tribolium. Mol. Biol. Evol. 13, 1059–1066. doi: 10.1093/oxfordjournals.molbev. a025668


in the human kinetochore protein CENP-C. Mol. Cell. Biol. 16, 3576–3586. doi: 10.1128/MCB.16.7.3576


replication in Saccharomyces cerevisiae. eLife 6:e28231. doi: 10.7554/eLife. 28231


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

Copyright © 2018 Smurova and De Wulf. 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.

# Altered Expression of Mitochondrial NAD<sup>+</sup> Carriers Influences Yeast Chronological Lifespan by Modulating Cytosolic and Mitochondrial Metabolism

Ivan Orlandi 1,2, Giulia Stamerra<sup>2</sup> and Marina Vai <sup>2</sup> \*

<sup>1</sup> SYSBIO Centre for Systems Biology, Milan, Italy, <sup>2</sup> Dipartimento di Biotecnologie e Bioscienze, Università di Milano-Bicocca, Milan, Italy

Nicotinamide adenine dinucleotide (NAD+) represents an essential cofactor in sustaining cellular bioenergetics and maintaining cellular fitness, and has emerged as a therapeutic target to counteract aging and age-related diseases. Besides NAD<sup>+</sup> involvement in multiple redox reactions, it is also required as co-substrate for the activity of Sirtuins, a family of evolutionary conserved NAD+-dependent deacetylases that regulate both metabolism and aging. The founding member of this family is Sir2 of Saccharomyces cerevisiae, a well-established model system for studying aging of post-mitotic mammalian cells. In this context, it refers to chronological aging, in which the chronological lifespan (CLS) is measured. In this paper, we investigated the effects of changes in the cellular content of NAD<sup>+</sup> on CLS by altering the expression of mitochondrial NAD<sup>+</sup> carriers, namely Ndt1 and Ndt2. We found that the deletion or overexpression of these carriers alters the intracellular levels of NAD<sup>+</sup> with opposite outcomes on CLS. In particular, lack of both carriers decreases NAD<sup>+</sup> content and extends CLS, whereas NDT1 overexpression increases NAD<sup>+</sup> content and reduces CLS. This correlates with opposite cytosolic and mitochondrial metabolic assets shown by the two types of mutants. In the former, an increase in the efficiency of oxidative phosphorylation is observed together with an enhancement of a pro-longevity anabolic metabolism toward gluconeogenesis and trehalose storage. On the contrary, NDT1 overexpression brings about on the one hand, a decrease in the respiratory efficiency generating harmful superoxide anions, and on the other, a decrease in gluconeogenesis and trehalose stores: all this is reflected into a time-dependent loss of mitochondrial functionality during chronological aging.

Keywords: NAD+, chronological lifespan, Ndt1, Ndt2, Saccharomyces cerevisiae

# INTRODUCTION

Significant progress has been made in elucidating fundamental processes such as human aging/ longevity as a result of studies performed in the budding yeast Saccharomyces cerevisiae. In this single-celled yeast, replicative aging and chronological aging are two complementary models that are used to simulate cellular aging of mitotically active and post-mitotic mammalian cells,

#### Edited by:

Maria Grazia Giansanti, Istituto di Biologia e Patologia Molecolari (IBPM), Italy

#### Reviewed by:

Sergio Giannattasio, Istituto di Biomembrane, Bioenergetica e Biotecnologie Molecolari (IBIOM), Italy Manuela Côrte-Real, University of Minho, Portugal

> \*Correspondence: Marina Vai marina.vai@unimib.it

#### Specialty section:

This article was submitted to Genetic Disorders, a section of the journal Frontiers in Genetics

Received: 04 September 2018 Accepted: 04 December 2018 Published: 19 December 2018

#### Citation:

Orlandi I, Stamerra G and Vai M (2018) Altered Expression of Mitochondrial NAD<sup>+</sup> Carriers Influences Yeast Chronological Lifespan by Modulating Cytosolic and Mitochondrial Metabolism. Front. Genet. 9:676. doi: 10.3389/fgene.2018.00676

**82**

respectively (MacLean et al., 2001; Longo and Kennedy, 2006; Longo et al., 2012). The former cell type is exemplified by fibroblasts and the latter by myocytes.

In the presence of nutrients, S.cerevisiae divides asymmetrically (budding) resulting in a large mother cell and a smaller daughter (bud). In this context, the replicative lifespan (RLS), namely the number of buds generated by a mother cell before senescence, indicates the reproductive potential of individual yeast cells (Steinkraus et al., 2008). The chronological lifespan (CLS), instead, refers to the rate of post-mitotic survival of a non-dividing quiescent yeast culture; viability is assessed by measuring the percentage of cells able to resume growth and form a colony after transfer from the depleted medium to the rich fresh one (Fabrizio and Longo, 2007). In a standard CLS experiment, yeast cells are grown in synthetic media with 2% glucose. When glucose becomes limiting, the diauxic shift occurs and cells shift from glucose-driven fermentation to ethanol-driven respiration. This shift determines a metabolic reprogramming, the outcomes of which influence the CLS. Afterwards cell proliferation stops and the yeast culture enters a quiescent stationary phase (Gray et al., 2004; Wanichthanarak et al., 2015). CLS is determined starting 72 h after the diauxic shift (Fabrizio and Longo, 2007).

The signaling pathways and regulators controlling RLS and CLS are evolutionary conserved (Fontana et al., 2010; Swinnen et al., 2014; Bitto et al., 2015; Baccolo et al., 2018). In particular, nicotinamide adenine dinucleotide (NAD+) homeostasis has emerged as a critical element in the regulation of aging/longevity (Imai, 2010, 2016) and accumulating evidence suggests that a reduction of NAD<sup>+</sup> levels in diverse organisms contributes to the development of age-associated metabolic decline (Imai and Guarente, 2014, 2016; Verdin, 2015). Indeed, in addition to its central role in cellular metabolism participating as essential coenzyme in many redox reactions, NAD<sup>+</sup> is absolutely required as a co-substrate by Sirtuins, a family of NAD+-dependent deacetylases, the founding member of which is Sir2 of S.cerevisiae (Houtkooper et al., 2012; Imai and Guarente, 2014). In mammals, there are seven Sirtuin isoforms (SIRT1-7) and among them SIRT1 is a key component of the systemic regulatory network called "the NAD world," a comprehensive concept that connects NAD<sup>+</sup> metabolism and aging/longevity control in mammals (Imai, 2009, 2010, 2016). The nutrient-sensing SIRT1 is the closest mammalian ortholog of Sir2 (Frye, 2000). Sir2 activity is involved in both replicative and chronological aging: in the former Sir2 extends RLS (Kaeberlein et al., 1999; Imai et al., 2000), whilst in the latter it has a pro-aging role (Fabrizio et al., 2005; Smith et al., 2007; Casatta et al., 2013; Orlandi et al., 2017a).

The other key component of the NAD world is represented by NAD<sup>+</sup> biosynthesis (Imai, 2009, 2010). From yeast to mammalian cells, NAD<sup>+</sup> synthesis occurs either de novo from L- tryptophan or through salvage pathway(s) from its precursors, namely nicotinamide riboside, nicotinic acid, and its amide form, nicotinamide (Bogan and Brenner, 2008; Canto et al., 2015). Cells mainly rely on the salvage pathway(s) for the correct maintenance of NAD<sup>+</sup> levels and it has been observed that the supplementation of NAD<sup>+</sup> precursors is sufficient to attenuate several metabolic defects common to the aging process (Johnson and Imai, 2018; Mitchell et al., 2018; Rajman et al., 2018). However, NAD<sup>+</sup> levels, as well as those of its precursors, are different depending on the type of tissue and cellular compartment (Dolle et al., 2010; Houtkooper et al., 2010; Cambronne et al., 2016) and it remains unclear in which cellular compartment(s) NAD<sup>+</sup> decrease can be relevant to aging. This has increased the interest on the role, on the one hand, of intertissue communications (Imai, 2016) and, on the other hand, of the relative subcellular localization of NAD<sup>+</sup> and its precursors during the aging process (Koch-Nolte et al., 2011; Rajman et al., 2018).

In yeast, NAD<sup>+</sup> is synthesized in the cytosol and can be imported across the inner mitochondrial membrane by two specific mitochondrial NAD<sup>+</sup> carriers, namely Ndt1 and Ndt2, which share 70% homology (Todisco et al., 2006). The physiological effects linked to an NDT1 and NDT2 double deletion and to the overexpression of NDT1, which encodes the main isoform of the NAD<sup>+</sup> transporter (Todisco et al., 2006), have been examined on cells growing with an oxidative or respiro-fermentative metabolism in batch and glucose-limited chemostat cultures (Agrimi et al., 2011).

Here, we show that during chronological aging an altered expression of the specific mitochondrial NAD<sup>+</sup> carriers deeply influences the metabolic reprogramming that enables cells to acquire features required to maintain viability during chronological aging. In particular, lack of NDT1 and NDT2 extends CLS, whereas NDT1 overexpression determines a CLS reduction. This opposite effect on CLS correlates with opposite metabolic features displayed by the two mutants.

# MATERIALS AND METHODS

# Yeast Strains, Growth Conditions and CLS Determination

The ndt11ntd21 strain and the strain overexpressing NDT1 (NDT1-over strain) were constructed in a previous work (Agrimi et al., 2011) and were derivatives of CEN.PK113-7D (MATa, MAL2-8c, SUC2). A null mutant ndt11ntd21 (ndt11::URA3 ntd21::KlLEU2) was generated by PCR-based methods in a W303-1A background (MATa ade2-1 his3-11,15 leu2-3,112 trp1- 1 ura3-1 can1-100). The accurancy of gene replacements was verified by PCR with flanking and internal primers. Cells were grown in batches at 30◦C in minimal medium (Difco Yeast Nitrogen Base without amino acids, 6.7 g/L) with 2% w/v glucose. Auxotrophies were compensated for with supplements added in excess (Orlandi et al., 2014). Cell number and cellular volumes were determined using a Coulter Counter-Particle Count and Size Analyser (Vanoni et al., 1983). Duplication time (Td) was obtained by linear regression of the cell number increase over time on a semi-logarithmic plot. For CLS experiments, cells were grown in 2% glucose and the extracellular concentration of glucose and ethanol were measured in medium samples collected at different time-points in order to define the growth profile [exponential phase, diauxic shift (Day 0), post-diauxic phase and stationary phase of the cultures] (Orlandi et al., 2013). CLS was

measured according to (Fabrizio et al., 2005) by counting colonyforming units (CFU) starting with 72 h (Day 3, first age-point) after Day 0. The number of CFU on Day 3 was considered the initial survival (100%).

# Isolation of Mitochondria

Mitochondria were prepared from chronologically aging cells essentially as described by Meisinger et al. (2006) with minor modifications. At each time-point, 10<sup>9</sup> cells were collected by centrifugation and spheroplasts were obtained by digestion with Zymolyase 20T. Then, spheroplasts were homogenized by 20 strokes using a Dounce homogenizer and mitochondria collected after differential centrifugation (Meisinger et al., 2006). Fresh crude mitochondrial pellets were used for measurements of NAD+, NADH, and protein contents.

### Metabolite Measurements and Enzymatic Assays

At designated time-points, aliquots of the yeast cultures were centrifuged, and both pellets (washed twice) and supernatants were collected and frozen at −80◦C until used. Rapid sampling for intracellular metabolite measurements was performed as previously described (Orlandi et al., 2014). The concentrations of glucose, ethanol, citrate, succinate, and malate were determined using enzymatic assays (K-HKGLU, K-ETOH, K-SUCC, K-CITR, and K-LMALR kits from Megazyme).

To measure NADH and NAD<sup>+</sup> contents, alkali, and acid extractions were performed essentially as described (Lin et al., 2001), except that before incubation of both the alkali extract and the acid one at for 30 min, an additional step was performed in order to improve cells lysis. Alkali or acid-washed glass beads were added to the two types of extracts and cells broken by vortexing (3 cycles of 1 min, interspersed with cooling on ice). NAD<sup>+</sup> and NADH concentrations were determined using the EnzyChromTM NAD+/NADH assay kit (BioAssay Systems). The rate of dye formation (formazan) at 565 nm correlates with the level of pyridine nucleotides. Duplicate reactions were performed in multi-well plates or in cuvettes. Different amounts of each sample were used in cycling reactions to obtain values within the linear portion of a standard curve that was prepared every time.

Immediately after preparation of cell-free extracts (Orlandi et al., 2014), the activities of cytosolic and mitochondrial aldehyde dehydrogenase (Ald) were assayed according to Aranda and del Olmo (2003), of phosphoenolpyruvate carboxykinase (Pck1) and isocitrate lyase (Icl1) as described in de Jong-Gubbels et al. (1995). Total protein concentration was estimated using the BCATM Protein Assay Kit (Pierce).

#### Fluorescence Microscopy

Dihydroethidium (DHE, Sigma-Aldrich) staining was performed as reported in Madeo et al. (1999) to detect superoxide anion (O<sup>−</sup> 2 ). A Nikon Eclipse E600 fluorescence microscope equipped with a Nikon Digital Sight DS Qi1 camera was used. Digital images were acquired and processed using Nikon software NIS-Elements.

# Estimation of Oxygen Consumption Rates and Index of Respiratory Competence

The basal oxygen consumption of intact cells was measured at 30◦C using a "Clark-type" oxygen electrode (Oxygraph System, Hansatech Instruments, Nortfolk, UK) as reported (Orlandi et al., 2013). The non-phosphorylating respiration and the maximal/uncoupled respiratory capacity were measured in the presence of 37.5 mM triethyltin bromide (TET, Sigma-Aldrich) and 10µM of the uncoupler carbonyl cyanide 3 chlorophenylhydrazone (CCCP, Sigma-Aldrich), respectively (Orlandi et al., 2017a). The addition of 2 M antimycin A (Sigma-Aldrich) accounted for non-mitochondrial oxygen consumption. Respiratory rates for the basal oxygen consumption (JR), the maximal/uncoupled oxygen consumption (JMAX) and the nonphosphorylating oxygen consumption (JTET) were determined from the slope of a plot of O<sup>2</sup> concentration against time, divided by the cellular concentration.

Index of respiratory competence (IRC) was measured according to Parrella and Longo (2008) by plating identical cell samples on YEP (1% w/v yeast extract, 2% w/v bacto peptone)/2% glucose (YEPD) plates and on rich medium/3% glycerol (YEPG) plates. IRC was calculated as colonies on YEPG divided by colonies on YEPD times 100%.

#### Statistical Analysis of Data

All values are presented as the mean of three independent experiments ± Standard Deviation (SD). Three technical replicates were analyzed in each independent experiment. Statistical significance was assessed by one-way ANOVA test. The level of statistical significance was set at a P value of ≤ 0.05.

# RESULTS AND DISCUSSION

## Altered Expression of the Specific Mitochondrial NAD<sup>+</sup> Carriers Affects CLS

Due to the importance of NAD<sup>+</sup> homeostasis in the aging process from yeast to humans (Baccolo et al., 2018; Rajman et al., 2018; Yaku et al., 2018), we wished to test whether changes in the cellular content of this dinucleotide would cause any effects on CLS. To this end we chose to use the ndt11ndt21 and NDT1-over mutant strains: the former lacking the two mitochondrial NAD<sup>+</sup> carriers, Ndt1 and Ndt2, identified so far and the latter overexpressing Ndt1, which is the main isoform of the carrier (Todisco et al., 2006; Agrimi et al., 2011). These strains have been previously characterized as far as NAD content is concerned (Agrimi et al., 2011). In particular, under a fully respiratory metabolism such as growth on ethanol, Ndt1 overexpression determined an increase in cellular and mitochondrial NAD<sup>+</sup> levels without affecting growth. On the contrary, on ethanol the ndt11ndt21 mutant displayed a lower cellular and mitochondrial NAD<sup>+</sup> content and a decrease in the growth rate (Agrimi et al., 2011). Here, an ndt11ndt21 double mutant generated in the W303-1A background was also included. Indeed, the W303-1A strain is commonly used in chronological aging research due to its robust respiratory capacity (Ocampo et al., 2012).

refer to mean values of three independent experiments with three technical replicates each. Standard deviation (SD) is indicated. Statistical significance as assessed by one-way ANOVA test is indicated (\*P ≤ 0.05 and \*\*P ≤ 0.01).

Initially, in the context of a standard CLS experiment (Fabrizio and Longo, 2007), we measured CLS and NAD content. As shown in **Figure 1A**, Ndt1 overexpression significantly reduced CLS, whilst the strain devoid of the two mitochondrial NAD<sup>+</sup> carriers lived longer than the prototrophic wild type (wt) CEN.PK 113-7D. The same long-lived phenotype was observed in the auxotrophic background W303-1A (**Figure 1B**) indicating that the different composition of amino acids in the medium does not influence the results. Measurements of intracellular NAD<sup>+</sup> and NADH contents indicated that in the wt, they decreased progressively after the diauxic shift (**Figures 1C,D**). We calculated values of NAD<sup>+</sup> and NADH estimating cell size with a Coulter Counter-Particle Count and Size Analyser: cell size that changes according to the yeast strain and the growth phase of the cell cycle. If we assume a yeast cell size of 70 µm<sup>3</sup> (Sherman, 2002) our measurements of 595µM NAD<sup>+</sup> (**Figure 1C**) and 215µM NADH for CEN.PK 113-7D in exponential phase (**Figure 1D**) correspond to 1.42 mM NAD<sup>+</sup>

and 0.82 mM NADH, in reasonable agreement with values of previous reports (Lin et al., 2004). In the context of the CLS experimental set-up, as the diauxic shift occurs and cells utilize the excreted fermentation by-product, ethanol, in the ndt11ndt21 mutant, and in the NDT1-over one NAD<sup>+</sup> and NADH levels decreased, but both remained constantly lower in the ndt11ndt21 mutant and higher in the NDT1-over mutant than those measured in the wt (**Figures 1C,D**). This opposite trend of the dinucleotide contents observed in the two different types of mutants is in line with that detected during exponential growth on ethanol (Agrimi et al., 2011).

It is well known that NAD<sup>+</sup> is an essential coenzyme for oxidoreductases of both cytosolic and mitochondrial redox reactions, many of which are involved in the metabolic remodeling that takes place at the diauxic shift. Indeed, at the diauxic shift carbon metabolism shifts from fermentation to mitochondrial respiration and gluconeogenesis allowing cells to be better primed for survival during chronological aging. Thus, we analyzed the metabolic features of the short-lived NDT1-over strain and those of the long-lived ndt11ndt21 one. Since the respiration-based metabolism is due to the utilization of ethanol, we initially measured the consumption of this C2


TABLE 1 | Effects of altered expression of mitochondrial NAD<sup>+</sup> carriers on the enzymatic activity of Ald isoforms.

For each time-point total Ald, Ald6, and Ald4/5 enzymatic activities were determined as in Figure 2 and the percentage of the different Ald isoforms was calculated. Day 0, diauxic shift. Data refer to mean values determined in three independent experiments with three technical replicates each. SD is indicated. Values obtained for wt strains were used as reference for comparisons with NDT1-over and ndt11 ndt21 cells. (\*\*P ≤ 0.01, one-way ANOVA test).

compound. At the diauxic shift (Day 0), the maximal amount of the extracellular ethanol was not affected either by the lack of Ndt1 and Ndt2 or by the Ndt1 overexpression (**Figures 1E,F**). Differently, during the post-diauxic phase in the ndt11ndt21 mutant ethanol decreased more slowly (**Figures 1C,D**). This is indicative of an impairment in ethanol utilization in line with the slow growth rate on medium containing ethanol as carbon source (Agrimi et al., 2011). Consequently, starting from Day 0, we determined the enzymatic activities of the acetaldehyde dehydrogenases (Alds). These enzymes are implicated in the ethanol utilization: they oxidize the acetaldehyde generated from ethanol oxidation producing acetate, which is subsequently converted to acetyl-CoA. In addition, Alds require NAD<sup>+</sup> or NADP+. No difference was detected between the wt and the NDT1-over strain in the total Ald activity levels (**Figure 2A**). On the contrary, in the ndt11ndt21 strain a significant decrease was observed (**Figure 2A**) consistent with the reduced ethanol utilization. Notably, interesting results were obtained by measuring the different isoforms of Alds, namely the mitochondrial Ald4/5, and the primary cytosolic counterpart Ald6 (Saint-Prix et al., 2004). Indeed, the activity levels of Ald4/5 were higher and those of Ald6 lower in the NDT1-over strain compared with the wt ones, whilst in the ndt11ndt21 strain the Ald6 activity prevailed (**Figures 2B,C**). Since, alterations of the mitochondrial NAD<sup>+</sup> transport are accompanied by a different prevalent subcellular localization of Ald enzymatic activities (**Table 1**), it is reasonable to speculate that in the two different mutants the metabolic pathways that are fed by mitochondrial or cytosolic acetate/acetyl-CoA could be affected. In this context, we initially measured the enzymatic activity of one of the unique enzymes of the glyoxylate shunt, such as isocitrate lyase (Icl1), and that of phosphoenolpyruvate carboxykinase (Pck1), which catalyzes the rate-limiting step in gluconeogenesis. Indeed, starting from the diauxic shift, the glyoxylate shunt becomes operative. It is an anaplerotic device of the TCA cycle, is fed by the cytosolic acetyl-CoA and is the sole possible provider for the Pck1 substrate, namely oxaloacetate (Lee et al., 2011). In the NDT1-over strain a decrease in the enzymatic activities of Icl1 and Pck1 was observed, whilst in the ndt11ndt21 mutant both activities strongly increased (**Figures 3A,B**). Since glucose-6-phosphate produced by gluconeogenesis is used for the synthesis of threalose during the post-diauxic phase, we also examined the accumulation of this disaccharide, the intracellular stores of which are advantageous for survival during chronological aging (Shi et al., 2010). In the NDT1-over strain a reduction in trehalose levels took place (**Figure 3C**), consistent with the decrease of the Pck1 activity. On the contrary, the ndt11ndt21 cells accumulated more trehalose (**Figure 3C**), consistent with the increase of the Pck1 activity. Taken together, these results indicate that the lack of the two mitochondrial NAD<sup>+</sup> carriers elicits an enhancement along the cytosolic Ald6/glyoxylate/gluconeogenesis axis, whereas Ndt1 overexpression elicits a down-regulation.

Following on, since the TCA cycle is fed by the mitochondrial acetyl-CoA, we assessed the levels of some of its intermediates, such as citrate, malate, and succinate. Starting from the diauxic shift, the levels of these C4 dicarboxylic acids in the NDT1 over strain mirrored those measured in the wt (**Figure 4**). On the contrary, in the ndt11ndt21 mutant all these metabolites significantly decreased (**Figure 4**) suggesting an impairment in the TCA cycle.

FIGURE 4 | The ndt11ndt21 mutant displays decreased levels of TCA intermediates. The indicated strains were grown as in Figure 1 and intracellular concentrations of citrate (A), succinate (B) and malate (C) were measured. Day 0, diauxic shift. Bar charts show the mean values determined in three independent experiments with three technical replicates each. SD is indicated (\*\*P ≤ 0.01).

# The ndt11ndt21 Mutant Preserves Functional Mitochondria During Chronological Aging

Considering that during respiration the TCA cycle provides the electron transport chain (ETC) with reducing equivalents through redox reactions and that respiration affects the CLS (Bonawitz et al., 2006; Ocampo et al., 2012; Baccolo et al., 2018), we measured next the respiratory activity in the NDT1 over and ndt11ndt21 strains. During the exponential phase, the respiratory parameters for the NDT1-over and ndt11ndt21 strains were very similar to those of the wt (**Table 2**) in good agreement with (Agrimi et al., 2011). Differences were observed starting from the diauxic shift (respiratory metabolism). Indeed, in the double deleted mutant, basal oxygen consumption (JR) was lower than the wt one (**Table 2**). This can be ascribed to a depletion/limitation of reducing equivalents since in the presence of the uncoupler CCCP, which dissipates the proton gradient across the mitochondrial membrane, the maximal oxygen consumption rate (JMAX) of the ndt11ndt21 cells was always lower than that of the wt (**Table 2**). Interestingly, the ndt11ndt21 cells displayed a non-phosphorylating respiration (JTET) strongly reduced compared with that of the wt, the levels of which increased as a function of time in culture (**Table 2**) as expected (Orlandi et al., 2017a,b). As a consequence, in the double deleted mutant the net respiration, which estimates the coupled respiration, was close to the wt one (**Table 2**) indicating that, despite a reduced JR, during the post-diauxic phase the ndt11ndt21 strain has a better coupling between electron transport and ATP synthesis. On the contrary, the NDT1-over strain had a J<sup>R</sup> similar to the wt one and a JMAX higher (**Table 2**). Nevertheless, in this strain the net respiration was lower due to a JTET significantly higher than that of the wt (**Table 2**) indicative of an increase of uncoupled respiration. These differences in both the level and in the state of the respiration of the two mutants were accompanied by differences in mitochondrial NAD<sup>+</sup> and NADH contents: in the ndt11ndt21 mutant and in the NDT1-over one a decrease and an increase of NAD<sup>+</sup> and NADH contents, respectively, were observed compared with those of the wt (**Figures 5A,B**). This opposite trend in the mitochondrial dinucleotide contents, as well as the different respiratory efficiency, of the two types of mutants are in line with those detected during exponential growth on ethanol (Agrimi et al., 2011).

To the best of our knowledge, Ndt1 and Ndt2 are the only mitochondrial NAD<sup>+</sup> carriers described so far in S.cerevisiae (Todisco et al., 2006) and their transport activity for NAD<sup>+</sup> is also consistent with the cellular localization of the enzymes involved in NAD<sup>+</sup> biosynthesis, which are outside the mitochondria, and with the lack of NAD+- synthesizing enzymes in the yeast mitochondria (Kato and Lin, 2014). However, since in the mitochondria of the ndt11ndt21 mutant, NAD<sup>+</sup> is present, albeit at low levels, it cannot be excluded that this dinucleotide can be imported in the mitochondria with lower efficiency by other carrier systems. Indeed, many mitochondrial transporters often exhibit some overlapping of the transported substrates


TABLE 2 |Respiratory parameters determined forNDT1-overand ndt11ndt21

JR

JMAX

strains.

Genetic Background

 Strain

Exp

> wt

NDT1-over

ndt1

wt ndt1

1ndt2

1

7.55

± 0.35

8.42\*\*

± 0.17

11.83\*\*

7.76

± 0.07

9.47

± 0.13

14.41

1ndt2

1

7.09

± 0.22

9.66\*\*

± 0.41

10.77\*\*

7.28

± 0.31

11.98

± 0.17

13.88

CEN-PK 113-7D

W303-1A Genetic Background

 Strain

wt NDT1-over

ndt1

wt ndt1

1ndt2

1

1.04

± 0.26 6 cells/s. Basal respiration rate (JR),

1.18\*\*

± 0.09

1.27\*\*

1.24

± 0.38

1.58

± 0.38

1ndt2

1

1.24

± 0.38

1.31\*\*

± 0.12

1.43\*\*

1.72

± 0.09

3.47\*\*

± 0.26

3.97\*\*

CEN-PK 113-7D

1.43

± 0.41

2.13

± 0.19

Exp

Day 0

7.12

± 0.29

11.23

± 0.29

14.24

Day 0

 Day 1

 Day 2

 Day 3

Exp

Day 0

 Day 1

 Day 2

 Day 3

W303-1A Oxygen consumption

in the measurements

each. SD is indicated. Values obtained for wt were used as reference for comparisons

 rates (J) are expressed as pmol/10

 of the respiratory parameters

 are detailed in the text. Exp, assessed by the superoxide-driven conversion of non-fluorescent dihydroethidium into fluorescent ethidium (Eth). Day 0, diauxic shift. About 1,000 cells for each sample (three technical replicates) in three independent experiments were examined. SD is indicated (\*P ≤ 0.05 and \*\*P ≤ 0.01). (D) Starting from the diauxic shift (Day 0), at indicated time-points aliquots of wt, NDT1-over, and ndt11ndt21 cultures were serially diluted and plated onto YEPD and YEPG plates in order to determine the index of respiratory competence (IRC). SD is indicated.

(Palmieri et al., 2006). In addition, other systems contribute to the homeostasis of the intramitochondrial NAD pool, as well as to balance dinucleotide pools between mitochondria and cytosol/nucleus. They include, among others, two NADH dehydrogenases (Nde1 and Nde2) distributed on the external surface of the inner mitochondrial membrane and the glycerol-3-phosphate shuttle (Bakker et al., 2001). Nde1 and Nde2 directly catalyze the transfer of electrons from cytosolic NADH to ubiquinone without the translocation of protons across the membrane. In such a way, the ETC is supplied with electrons (Baccolo et al., 2018). The expression of NDE1 and NDE2 is induced after the diauxic shift (Bakker et al., 2001). With regard to the glycerol-3-phosphate shuttle, it is a system of crucial importance under conditions where the availability of energy is limited (Rigoulet et al., 2004). In the glycerol-3-phosphate shuttle, cytosolic glycerol-3-phosphate dehydrogenase oxidizes cytosolic NADH catalyzing the reduction of dihydroxyacetone phosphate to glycerol-3-phosphate. Subsequently, into the mitochondrial matrix, glycerol-3-phosphate delivers its electrons to ubiquinone via the FAD-dependent glycerol-3-phosphate dehydrogenase, Gut2 (Bakker et al., 2001). The result is a stepwise transfer of electrons from the cytosol to the respiratory chain. Consequently, despite the low mitochondrial dinucleotide contents, the ndt11ndt21 mutant might feed the oxidative phosphorylation with the NADH produced in the cytosol.

Afterward, given the differences in the state of respiration, we decided to analyze the content of superoxide anion (O<sup>−</sup> 2 ), which is the primary mitochondrial reactive oxygen species (ROS) produced by electron leakage from the respiratory chain. It is known that O<sup>−</sup> 2 /ROS accumulation limits the long-term survival of yeast cells during CLS (Pan, 2011; Breitenbach et al., 2014; Baccolo et al., 2018). In the ndt11ndt21 chronologically cells and in the NDT1-over ones, a strong decrease and increase in O − 2 content was observed, respectively, compared to that of the wt (**Figure 5C**) consistent with non-phosphorylating respiration data (**Table 2**). Indeed, it is a state of non-phosphorylating respiration prone to generate O<sup>−</sup> 2 (Hlavata et al., 2003; Guerrero-Castillo et al., 2011). In addition, we analyzed the mitochondrial functionality by measuring the IRC, which defines the percentage of viable cells competent to respire (Parrella and Longo, 2008). Starting from the diauxic shift where all the strains were respiration-competent, a different trend of the IRC was observed for the ndt11ndt21 strain and in the NDT1-over one. In the former, a lower decrease in the mitochondrial functionality was detected and at Day 18 the IRC was still about 60% against about 40% in the wt (**Figure 5D**). In the NDT1-over chronologically aging cells, a dramatic time-dependent loss of mitochondrial functionality was observed reaching at Day 18 values close to zero (**Figure 5D**). This is in line with the increased O − 2 formation because it is known that ROS levels influence mitochondrial fitness and mitochondrial dysfunctions, in turn, lead to a higher propensity to produce ROS (Breitenbach et al., 2014).

Thus, taken together all the results clearly indicate that, in the context of a standard CLS experiment, alterations in the expression of the specific mitochondrial NAD<sup>+</sup> carriers determined by NDT1 and NDT2 double deletion and NDT1 overexpression deeply influence the metabolism with opposite outcomes on chronological longevity (**Figure 6**). We found that the former extends CLS, whereas the latter shortens it. This is a direct consequence, on the one hand, of the participation of NAD<sup>+</sup> together with its reduced counterpart, NADH, in a wide range of metabolic reactions modulating the activity of compartment-specific pathways among which the TCA cycle and the ETC in the mitochondria and the glycolysis/gluconeogenesis in the cytosol. On the other hand, the CLS is regulated by signaling pathways that coordinate the metabolic reprogramming required to ensure longevity (Breitenbach et al., 2014; Zhang and Cao, 2017). On the whole, in the ndt11ndt21 chronologically aging cells and in the NDT1-over ones an opposite metabolic remodeling is observed, involving both cytosolic (gluconeogenesis), and mitochondrial (TCA and respiration) metabolic pathways (**Figure 6**), which are operative during chronological aging. Lack of the mitochondrial NAD<sup>+</sup> carriers results in a reduced oxygen consumption that does not depend upon dysfunctional mitochondria but most likely upon a decreased amount of reducing equivalents provided by a TCA cycle, the activity of which is reduced. Nevertheless, this mutant maintains a net respiration close to that of the wt indicating that in the mutant the respiration, albeit reduced, is more efficient. This confirms previous data on ndt11ndt21 cells exponentially growing on ethanol that show a better coupling of respiration and phosphorylation (Agrimi et al., 2011). Such a state of more coupled respiration is less prone to generate hazardous O<sup>−</sup> 2 decreasing the risk of inducing oxidative stress and its detrimental effects on cell survival of non-dividing cells during chronological aging: in agreement with this, ndt11ndt21 cells are long-lived. In agreement with a short-lived phenotype accompanied by O<sup>−</sup> 2 accumulation and severe mitochondrial damage, NDT1-over chronologically aging cells display an enhanced uncoupled respiration and a lower respiratory efficiency. As in the case of the ndt11ndt21 cells, changes in the state of respiration have been already observed in NDT1-over cells exponentially growing on ethanol (Agrimi et al., 2011). In this context of fully respiratory metabolism, the NDT1 overexpression determines a decrease in the respiratory efficiency similar to that described here when cells have exhausted glucose and shift to ethanoldriven respiration. These results further underline how the mitochondrial NAD<sup>+</sup> carriers and, consequently, the availability of mitochondrial NAD+, and/or NADH is important to achieve an efficient respiration and how this aspect can influence the CLS.

Concerning gluconeogenesis, the enzymatic activity of Pck1 is generally considered the main flux-controlling step in the pathway. The gluconeogenic activity of this enzyme depends on its de/acetylation state (Lin et al., 2009; Casatta et al., 2013). Indeed, an increase in the enzymatic activity of Pck1 correlates with an increase in its acetylated active form promoting gluconeogenesis and CLS (Lin et al., 2009; Casatta et al., 2013; Orlandi et al., 2017a,b). The enzyme responsible for Pck1 deacetylation (inactive form) is the NAD+-dependent deacetylase Sir2 (Lin et al., 2009). During chronological aging, lack of Sir2 correlates with an increase of the acetylated Pck1 and with a carbohydrate metabolism shift toward glyoxylaterequiring gluconeogenesis increasing CLS (Casatta et al., 2013; Orlandi et al., 2017a,b). It is conceivable that, as the deacetylase activity of Sir2 relies on NAD+, the low level of this dinucleotide in the ndt11ndt21 mutant might decrease Sir2-mediated deacetylation of Pck1 and consequently increase gluconeogenesis and CLS. Differently, in the NDT1-over mutant, a different

availability of NAD<sup>+</sup> might favor Sir2 enzymatic activity leading to an increase of the deacetylated inactive form of Pck1 and to the observed decrease of gluconeogenesis and CLS.

Furthermore, in the ndt11ndt21 mutant and the NDT1-over one, other metabolic traits that result from an enhancement and a down-regulation, respectively, of the cytosolic Ald6/glyoxylate/gluconeogenesis axis fit-well with their CLS. Indeed, Ald6 activity requires NADP<sup>+</sup> providing NADPH, which is also provided by the pentose phosphate pathway fueled by the gluconeogenesis with glucose-6 phosphate. NADPH is a source of reducing energy and an essential cofactor for glutathione/thioredoxin-dependent enzymes that are essential for protecting cells from oxidative stress (Pollak et al., 2007). Thus, NADPH availability can contribute to influence the physiological state of the cells and consequently their survival. In this context, the ndt11ndt21 mutant might be further favored by an enhanced gluconeogenic activity that leading also to increased intracellular trehalose stores, ensures viability during chronological aging. On the contrary the down-regulation of the Ald6/glyoxylate/gluconeogenesis axis observed in the NDT1-over mutant decreasing cellular protection systems, might contribute to affect negatively the CLS.

To date, substantial number of evidence points out that lowering NAD<sup>+</sup> levels can decrease Sirtuin activities and affect the aging process both in S.cerevisiae and mammalian cells (Imai and Guarente, 2016). In particular, in yeast lack of the nicotinic acid phosphoribosyltransferase, Npt1, which in the salvage pathway generates NAD<sup>+</sup> from nicotinic acid (NA), reduces NAD<sup>+</sup> content. This is accompanied by loss of silencing and decrease in RLS (Smith et al., 2000), as NAD<sup>+</sup>

levels are not sufficient for Sir2 to function (Ondracek et al., 2017). Addition of nicotinamide riboside (an NAD<sup>+</sup> precursor) corrects the deficit in NAD<sup>+</sup> content of the npt11 mutant, promotes Sir2-dependent silencing and extends RLS (Belenky et al., 2007). Furthermore, yeast cells grown in media lacking NA has a short RLS and low NAD<sup>+</sup> levels; supplementation of isonicotinamide extends RLS in a Sir2-dependent manner by restoring NAD<sup>+</sup> content and alleviating the nicotinamide (NAM) inhibition on Sir2 (McClure et al., 2012). Indeed, NAM is an NAD<sup>+</sup> precursor that is also an endogenous noncompetitive inhibitor of Sir2 (Sauve et al., 2005). Yeast cells grown in the presence of NAM have the same phenotype of sir21 ones such as silencing defects and a short RLS (Sauve et al., 2005). In the context of chronological aging, NAM supplementation at the diauxic shift results in a phenocopy of chronologically aging sir21 cells: due to the inhibition of Sir2, Pck1 enzymatic activity, and gluconeogenesis are promoted and CLS is extended (Orlandi et al., 2017a). On the opposite, resveratrol, a Sirtuin activating compound, restricts CLS by enhancing Sir2 activity, in particular Sir2-mediated deacetylation of Pck1, and consequently gluconeogenesis is decreased (Orlandi et al., 2017b).

In conclusion, taken together all our results show that affecting the cellular distribution and the content of NAD<sup>+</sup> has a deep impact on both metabolism and chronological aging and that a critical functional role is played by the Sir2 activity. In addition, our data indicate that in order to elucidate the intimate interplay between NAD+, Sirtuins and aging, it will be important to determine how NAD<sup>+</sup> levels change in different compartments during aging and the tissue-specific regulation of NAD metabolism and Sirtuin activity.

#### AUTHOR CONTRIBUTIONS

MV conceived the project. MV and IO designed the experiments. IO and GS performed the experiments. MV wrote the manuscript. All authors have read and approved the final version of the manuscript.

#### REFERENCES


#### ACKNOWLEDGMENTS

The authors are grateful to Neil Campbell for English editing. This work was supported by CARIPLO Foundation 2015-0641 to MV and GS was supported by fellowships from SYSBIONET, Italian roadmap of ESFRI.


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

Copyright © 2018 Orlandi, Stamerra and Vai. 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.

# Ras-Induced miR-146a and 193a Target Jmjd6 to Regulate Melanoma Progression

Viviana Anelli<sup>1</sup> , Anita Ordas<sup>2</sup> , Susanne Kneitz<sup>3</sup> , Leonel Munoz Sagredo4,5 , Victor Gourain<sup>4</sup> , Manfred Schartl3,6,7, Annemarie H. Meijer<sup>2</sup> and Marina Mione<sup>1</sup> \*

<sup>1</sup> Cibio, University of Trento, Trento, Italy, <sup>2</sup> Institute of Biology, Leiden University, Leiden, Netherlands, <sup>3</sup> Physiological Chemistry, Biocenter, University of Würzburg, Würzburg, Germany, <sup>4</sup> Institute of Toxicology and Genetics, Karlsruhe Institute of Technology, Karlsruhe, Germany, <sup>5</sup> Faculty of Medicine, University of Valparaiso, Valparaíso, Chile, <sup>6</sup> Comprehensive Cancer Center, University Clinic Würzburg, Würzburg, Germany, <sup>7</sup> Hagler Institute for Advanced Study and Department of Biology, Texas A&M University, College Station, TX, United States

Ras genes are among the most commonly mutated genes in human cancer; yet our understanding of their oncogenic activity at the molecular mechanistic level is incomplete. To identify downstream events that mediate ras-induced cellular transformation in vivo, we analyzed global microRNA expression in three different models of Ras-induction and tumor formation in zebrafish. Six microRNAs were found increased in Ras-induced melanoma, glioma and in an inducible model of ubiquitous Ras expression. The upregulation of the microRNAs depended on the activation of the ERK and AKT pathways and to a lesser extent, on mTOR signaling. Two Rasinduced microRNAs (miR-146a and 193a) target Jmjd6, inducing downregulation of its mRNA and protein levels at the onset of Ras expression during melanoma development. However, at later stages of melanoma progression, jmjd6 levels were found elevated. The dynamic of Jmjd6 levels during progression of melanoma in the zebrafish model suggests that upregulation of the microRNAs targeting Jmjd6 may be part of an anti-cancer response. Indeed, triple transgenic fish engineered to express a microRNA-resistant Jmjd6 from the onset of melanoma have increased tumor burden, higher infiltration of leukocytes and shorter melanoma-free survival. Increased JMJD6 expression is found in several human cancers, including melanoma, suggesting that the up-regulation of Jmjd6 is a critical event in tumor progression.

The following link has been created to allow review of record GSE37015: http://www. ncbi.nlm.nih.gov/geo/query/acc.cgi?token=jjcrbiuicyyqgpc&acc=GSE37015.

Keywords: zebrafish, cancer models, microRNA, Jmjd6, ras, melanoma, miR-146a, miR-193a

# INTRODUCTION

Activating mutations in the RAS genes or in other members of the ras-signaling pathways are very common in cancer<sup>1</sup> and recent deep sequencing data of cancer genomes<sup>2</sup> suggest that these mutations are important primers of malignancies. Still, the initial molecular events following activation of the pathways downstream of Ras are extremely difficult to study in vivo.

<sup>1</sup>http://www.sanger.ac.uk/genetics/CGP/cosmic/ <sup>2</sup>http://cancergenome.nih.gov/

#### Edited by:

Maria Grazia Giansanti, Consiglio Nazionale delle Ricerche (CNR), Italy

#### Reviewed by:

Victoriano Mulero, University of Murcia, Spain Theodora Katsila, University of Patras, Greece

\*Correspondence: Marina Mione mariacaterina.mione@unitn.it

#### Specialty section:

This article was submitted to Genetic Disorders, a section of the journal Frontiers in Genetics

Received: 28 September 2018 Accepted: 04 December 2018 Published: 18 December 2018

#### Citation:

Anelli V, Ordas A, Kneitz S, Sagredo LM, Gourain V, Schartl M, Meijer AH and Mione M (2018) Ras-Induced miR-146a and 193a Target Jmjd6 to Regulate Melanoma Progression. Front. Genet. 9:675. doi: 10.3389/fgene.2018.00675

**95**

Transgenic models, where the expression of the oncogene leads to cancer development in a reproducible manner provide a suitable experimental system for addressing the complexity of cellular transformation in live animals. The oncogenic versions of the human RAS genes (KRAS, HRAS, and NRAS) have been the first and most successful drivers of cancer in transgenic mice (Chin et al., 1999a,b; Johnson et al., 2001; Malumbres and Barbacid, 2003). This ability of ras oncogenes to initiate and maintain cancer has been related to global molecular and epigenetic changes at early stages of transformation. Among the targets of oncogenes, microRNAs are well-suited to sustain global changes of cellular functions. Changes in several protein levels may be regulated by a single or just a few microRNAs and a number of microRNAs have been found deregulated in cancer (Harrandah et al., 2018). Yet, very few studies have investigated their roles at the onset of transformation as possible "global effectors" of oncogenesis. In this study we have investigated the link between Ras-induced transformation and microRNA expression, using genetically tractable zebrafish models where the expression of a constitutively active HRASG12V allele leads to the development of different cancer types. We found that activated Ras signaling promotes the rapid increase of six microRNAS. Interestingly, two of these microRNAs target the same gene, Jmjd6, a jumonjiC domain protein with at least two reported functions: histone arginine demethylation (Chang et al., 2007) and mRNA splicing regulation (Webby et al., 2009). Results reported here indicate that Jmjd6 is a critical player in zebrafish melanoma development and that at least two Ras-induced microRNAs antagonize Jmjd6 activation.

# RESULTS

#### Dynamic Regulation of MicroRNAs by RAS Activation

To identify miRNAs that are regulated by oncogenic Ras from the earliest stages of transformation, we used a custom Agilent microarray (see M&M). We profiled miRNA expression in transgenic zebrafish overexpressing HRASG12V in melanocytes (Santoriello et al., 2010), in brain cells (Mayrhofer et al., 2017) and ubiquitously (Santoriello et al., 2009) (**Figure 1A**). Six miRNAs (miR-21.1, miR- 21.2, miR-146a, miR-146b-1, miR-193a, miR-193a-1) were up-regulated (log2FC > 1.2 and p-value < 0.01, see **Supplementary Table S1**) in all three transgenic models at 3 day post-fertilization (dpf) and in 7 dpf melanoma, whereas no commonly down-regulated miRNAs were found (**Figure 1B** and **Supplementary Table S1**, highlighted rows).

This study, we focused on the melanoma model, and to clarify the potential roles of the upregulated microRNAs in melanoma progression, we analyzed microRNA expression levels at four time points (3, 7, and 14 dpf and in adult tumors, **Supplementary Table S1**) during melanoma progression. Next, we designed Taqman probes for the active strand (−5p) of miR21, miR-146a and miR-146b and for the active strand (−3p) of miR-193a and validated the expression levels of the six upregulated miRNAs by qPCR in the melanoma model (**Figure 1C**). The expression of the six microRNA genes showed dynamic patterns: one was upregulated up to 7 dpf (miR-146b-1-5p), while the others were upregulated to different extents up to adult melanoma (**Figure 1D** and **Supplementary Table S1**). To further validate the dependence of these microRNAs on ras we used drugs that target different pathways downstream of ras. Next, expression levels of miR-21-5p (including both mir21-1 and miR21-2), miR-146a-5p, miR-146b-1-5p, and miR-193a-3p (including also mir193a-1-3p) were quantified through qPCR, after induction of ras expression and in the presence of drugs. For simplicity, the microRNAs under study are named miR-21, miR-146a, miR-146b, miR-193a from now on.

## The Increase of miR-21, miR-146a, and miR-193a Is Ras-Dependent

To investigate whether miR-21, miR-146a, miR-146b, and miR-193a are direct targets of ras, we analyzed their expression profiles using an inducible model, Tg(hsp70l:EGFP-HRASV12)io3 , called HS-RAS, that expresses oncogenic ras upon heat shock. RNA was extracted from 3 dpf HS-RAS embryos, 6 h after a 30 min heat shock treatment at 37◦C, when robust ras activation occurs (Santoriello et al., 2009). QPCR data show that miR-21 and miR-146a are significantly upregulated in Ras overexpressing larvae (**Figure 2A**), thus supporting the hypothesis that miR-21 and miR-146a represents early-response targets, likely to be directly induced by oncogenic Ras. MiR-193a and miR-146b did not show significant upregulation in response to ras; however, we performed the inhibitor experiments also on these two microRNAs as they had shown to be upregulated in the melanoma model (**Figure 1C**).

xs To identifying which signaling pathway(s) downstream of Ras induces overexpression of these microRNAs, we blocked specific pathways in HS-RAS larvae using small chemical inhibitors of ERK, AKT and mTor phosphorylation (**Figure 2B**). To check the efficacy of the inhibitors, we collected treated larvae at 3 dpf and performed western blot analysis for known targets of the three inhibitors. As shown in **Figure 2C** PD98059, rapamycin and LY29004 were able to decrease ERK-P, AKT-P and S6-P levels, respectively. Next we checked the levels of microRNAs in inhibitor treated larvae using qPCR. As shown in **Figures 2D,E**, induction of miR-21 and -146a expression was greatly attenuated by all three inhibitors and most robustly by the ERK inhibitor. MiR-146b levels were not affected by the drug treatments (**Figure 2F**) and miR-193a levels were reduced to statistically significant levels only when larvae were treated with rapamycin (**Figure 2G**). These data suggest that miR-21, miR-146a and to a less extent, miR-193a, are rasresponsive genes and their activation is regulated mostly by the MAPK/ERK and mTOR (for miR-193a) branches of ras signaling.

### Predicted Target of MicroRNAs 146a and 193a

As the function of miR-21 is widely studied in cancer (Frezzetti et al., 2011), and miR-146b was found not to respond to Ras signaling, in this study we focused on miR-146a and - 193a, to clarify whether they have a role in melanomagenesis.

A web-based target prediction algorithm [MicroCosmTargets, now incorporated in www.tools4mirs.org (Lukasik et al., 2016)], was used to identify potential targets of zebrafish miR-146a and miR-193a. The tool is based on the genome assembly ZV9<sup>3</sup> and returned a few targets that were selected to contain seed sequences for miR- 146a, miR-193a or both, this last category included only jmjd6. We then used a web-based interactionprediction algorithm for RNA molecules, IntaRNA<sup>4</sup> (Mann et al., 2017) for the fast and accurate prediction of interactions between miR-146a and/or 193a with jmjd6 mRNAs using the newest genome assembly, GRCz11<sup>5</sup> . This search confirmed the presence of seed sequences for miR-146a and -193a (see **Supplementary Figures S1A,B**).

#### Jmjd6 Is a Target of miR-193a and miR-146a

Danio rerio jmjd6 has 1 transcript (Ensemble, ENSDAR G00000102896). The 3<sup>0</sup> UTR region of Jmjd6 contains miRNA recognition elements (MREs) for miR-146a (**Supplementary Figure S1A**), and MREs for miR-193a (**Supplementary Figure S1B**). To determine whether Jmjd6 is a bona fide target of miR-193a and miR-146a, we tested whether miR-193a and miR-146a expression affects jmjd6 levels using an in vivo GFP reporter assay. The entire jmjd6- 3<sup>0</sup> UTR was cloned downstream of green fluorescent protein (GFP) open reading frame (**Figure 3A**). In vitro synthesized mRNA from this construct was then injected into single-cell zebrafish embryos with or without miR-193a or mir-146a duplexes (**Figure 3A**), which mimic microRNA overexpression. The injection of the microRNA duplexes resulted in increased levels of microRNA (**Supplementary Figure S2A**) and caused mild

<sup>3</sup>http://mar2015.archive.ensembl.org/Danio\_rerio/Info/Index

<sup>4</sup>http://rna.informatik.uni-freiburg.de/IntaRNA/Input.jsp

<sup>5</sup>https://www.ensembl.org/Danio\_rerio/Info/Index

or no phenotypes (**Supplementary Figure S2B**). The following day, GFP expression levels were monitored by fluorescence microscopy (**Figure 3B**) and by western blot analysis with an antibody against GFP (**Figures 3C,D**). In both assays, GFP levels were reduced in embryos injected with miR-193a or miR-146a duplexes. Duplex injection of either miR-146a or miR-193a also resulted in reduction of Jmjd6 protein levels (**Figures 3E,F**).

To validate the direct interaction between miR-146a and miR-193a with jmjd6 mRNA, we performed RNA immunoprecipitation (RIP, **Figure 3G**). After immunoprecipitation (IP) with Ago2 antibody, which selectively enriched for RISC complex components (Ikeda et al., 2006), jmjd6 transcripts were readily found in embryos injected with miR-146a and miR-193a duplexes (**Figure 3H**) suggesting that a physical interaction between jmjd6 transcripts and specific microRNAs occurs in the RISC complex. These data confirmed the interaction between miR-146a and miR-193a with jmjd6 mRNA.

As we found that jmjd6 is a target of miR-146a and - 193a, we investigated whether the level of jmjd6 was lower in Ras expressing larvae compared to controls. As shown in **Supplementary Figures S3A,B** heat-shock induced Ras overexpression results in down-regulation of jmjd6 RNA level

(**Supplementary Figure S3A**, column 2) and jmjd6 protein levels (**Supplementary Figure S3B**, lane 2). To test if jmjd6 down-regulation was due to increased expression of miR-146a and/or miR-193a, we reduced microRNA expression by injecting morpholinos (MO) specific to miR-146a or miR-193a and measured jmjd6 mRNA and protein levels in morpholino injected HS-Ras embryos. All morpholinos were able to reduce expression levels of their targets (data not shown) and to increase the levels of jmjd6 transcripts, repressed by ras, to levels similar to controls (**Supplementary Figure S3A**, columns 3–4) and protein levels (**Supplementary Figure S3B**, lanes 3– 4), suggesting that ras-induced miRs are responsible for the downregulation of jmjd6 levels observed in response to ras expression.

#### Jmjd6 in Melanoma

In our model of melanoma progression, where ras is overexpressed in melanocytes, jmjd6 levels are reduced compared to control larvae at 3 and 7 dpf (**Figure 4A**). However, at 14 dpf, there is no significant difference in jmjd6 levels between kita:Ras

and controls. We then analyzed the levels of jmjd6 expression in full-blown melanoma in two different genotypes: wild type (p53+/+) and p53−/−. In the latter background, melanomas developed earlier and were highly invasive from the earliest stages (Santoriello et al., 2010). QPCR analysis of jmjd6 levels in six wild type and in three p53−/− cases showed significantly higher levels of expression of jmjd6 in melanoma, especially in tumors developing in a p53−/− genetic background (**Figure 4B**).

Given the unexpected increase of jmjd6 expression in melanoma, and its correlation with a more aggressive phenotype (Liu et al., 2017), to understand the clinical relevance of our findings, we queried online databases to investigate the expression levels of miR-146a, miR-193a and JMJD6 in melanoma samples from human patients. Data reported here on human melanoma are derived from publicly available resources, as stated in the following text and in the figure's legend. While no changes in miR-193a expression were reported in melanoma, a couple of studies (GSE18509 and GSE31568, see dbDEMC at www.picb.ac.cn/dbDEMC/index.html) reported an increase of miR-146a in melanoma (not shown). Next, we analyzed the expression levels of JMJD6 in different types of cancers, including melanoma, using the website cBioPortal for Cancer Genomics<sup>6</sup> (Cerami et al., 2012). As shown in **Figure 4C**, we found that JMJD6 is upregulated in different cancers, with melanoma being one of the cancer having higher levels of JMJD6 expression and high genomic alteration frequency (with amplification in almost 3% of cases, **Supplementary Figure S4A**). Looking carefully into the skin cutaneous melanoma TCGA dataset we found that the levels of JMJD6 are upregulated in 16% of human cutaneous melanoma (72 out of 469 patients, **Supplementary Figure S4B**).

<sup>6</sup>http://www.cbioportal.org/

Moreover, JMJD6 expression levels are higher in the samples with HRAS and BRAF mutations compared to samples with wild type HRAS and BRAF (**Supplementary Figures S4C,D**), suggesting that JMJD6 is upregulated by RAS signaling in these melanomas. Patients with alteration of JMJD6 (mutations, amplifications, deep deletions, or multiple alterations) have a worse overall survival and disease/progression-free survival compared with patients with wild type JMJD6 (**Supplementary Figures S4E,F**), suggesting a key role of JMJD6 in disease progression.

Next, we investigated the levels of JMJD6 protein in human melanoma using the website Protein Atlas<sup>7</sup> (Uhlen et al., 2017). The website reported medium (in three patients) or high (in nine patients) nuclear JMJD6 immunostaining in malignant melanoma. An example of a high JMJD6 immunostaining in melanoma is shown in comparison to control skin (where JMJD6 is not expressed) (**Figure 4D**) (Uhlen et al., 2017). These data suggest that the microRNA-mediated downregulation of jmjd6 in the zebrafish progressive melanoma model is a transient event and at later stages of melanoma development or in more aggressive melanomas in fish and in human, Jmjd6 is overexpressed.

#### Expression of miR-Resistant Jmjd6 Promotes Ras-Induced Melanoma

Given the dynamic changes in Jmjd6 expression in the Rasinduced melanoma model in zebrafish, and the overall increase in expression in human melanoma, we wanted to clarify the role of Jmjd6 in melanoma with a gain of function approach. Therefore, we produced a transgenic line that expresses miR (146a and 193a) -resistant jmjd6 in melanocytes. This was achieved by replacing the 3<sup>0</sup> UTR of jmjd6 in the transgenic construct with an artificial sequence (SV40 polyA, **Figure 5A**). With this construct, we generated a tg(UAS:eGFP-jmjd6) transgenic line using standard Tol2 transposase methods (Kawakami, 2004), and crossed it to kita:Gal4 fish. The double transgenic fish, designated as kita:Jmjd6, have no phenotype; however eGFP-Jmjd6 expression was confirmed by kita driven GFP nuclear localization in melanocytes and notochord cells (**Figures 5B–D**). To study the effects of combined expression of oncogenic ras and Jmjd6 in melanoma we generated triple transgenic fish, Et(kita:Gal4TA, UAS:mCherry)hmz1;Tg(UAS:eGFP-HRASV12)io006; Tg(UAS:eGFP-JMJD6)ka202 (designated as kita/Ras/Jmjd6, **Figure 5E**) and observe the fish for the development of melanoma at regular intervals, from 7 days to 1 month. As shown in the disease-free survival curve, 60% of double transgenic kita-ras fish show melanoma at 1 month (**Figure 5F**). However, triple transgenic fish kita/Ras/Jmjd6 developed melanoma at an earlier stage (**Figures 5E,F**) and by 1 month of age, penetrance of melanoma was 95% (**Figure 5F**). The tumors appeared in multiple locations in the same fish (**Figure 5E**) and were more invasive, as judged by the body surface showing melanoma lesions and by histological analysis (**Figures 5G,H**).

We also noticed that Kita/Ras/Jmjd6 melanomas have massive infiltration of leukocytes (L-plastin+ cells), resembling macrophages and neutrophils for their shape, much higher than in to Kita/Ras melanomas (**Figures 5I–L**).

#### DISCUSSION

In this study we used a progressive model of melanoma in zebrafish to study the changes of microRNA expression at the onset of RAS-induced transformation and throughout progression of the disease. We found six microRNAs, which are upregulated as an early response to oncogenic RAS expression in three different models. As two of these microRNAs target Jmjd6 we investigated the significance of these interactions for melanoma progression. To our surprise we discovered that the target of these increased microRNAs, Jmjd6, is overexpressed in aggressive zebrafish melanoma. This suggested that overexpression of Jmjd6 promotes melanoma progression and that the increase of the microRNAs that downregulate Jmjd6 at the onset of Ras expression is part of a defensive response against the pro-oncogenic activity of Jmjd6. We validated this hypothesis by generating triple transgenic zebrafish that express micro-RNA resistant Jmjd6. Melanoma develop faster in the kita/ras/jmjd6 fish, supporting a pro-oncogenic role of Jmjd6. Further work is needed to clarify how Jmjd6 favors melanoma development.

Jmjd6 is a JumanjiC-domain containing protein, which has been endorsed for different functions. It was initially identified as a phosphatidylserine receptor (PSR) on the surface of phagocytes (Fadok et al., 2000). Ablation of Jmjd6 in mice, and morpholino downregulation of Jmjd6/PSR in zebrafish (Hong et al., 2004) causes abnormal development and leads to neonatal lethality in mouse (Böse et al., 2004; Kunisaki et al., 2004). However, many lines of evidence argue against a function of Jmjd6 as a PSR, primarily its nuclear localization (Cui et al., 2004; Tibrewal et al., 2007), and a number of newly reported nuclear activities. Subsequently, Jmjd6 was shown to function as an arginine demethylase, which removes repressive symmetric methylation marks from histone 2 (H2R3me2s) and histone 4 (H4R3me2s) arginines (Chang et al., 2007), but other groups have been unable to confirm the arginine demethylation activity of JMJD6 (Hahn et al., 2010). Webby et al. (2009) reported that JMJD6 acts as a lysyl-5-hydroxylase of the splice factor subunit U2A65. Silencing of JMJD6 expression in endothelial cells resulted in abnormal splicing of the VEGF receptor 1 (Boeckel et al., 2011). Meanwhile, a structural study suggested that methyl groups on singlestranded RNAs (ssRNAs) might be substrates of JMJD6 (Hong et al., 2010). More recently, a proteomic approach identified JMJD6 as one of a few binding partners of the bromodomain and extraterminal (BET) domain protein Brd4 (Rahman et al., 2011), which regulates gene expression through interaction with the cdk9 subunit of the positive transcription and elongation factor, pTEFb complex (Jang et al., 2005). Further studies lead to the identification of JMJD6 as the partner of BRD4 in binding distal enhancers known as anti-pause enhancers, which regulate release from transcriptional pausing in a large subset of transcriptional units which depend on long-range interactions (Liu et al., 2013). Here, the demethylase activity of JMJD6 on

<sup>7</sup>https://www.proteinatlas.org/

fluorescence) in representative melanoma sections from transgenic fish as indicated. Calibration bar: 100 µm for (B,D), 50 µm for (I–L).

the snRNA, 7SK, results in the disassembly of the transcriptional repressive complex HEXIM1-7SK, releasing pTEFb from pausing (Liu et al., 2013). It would be interesting to study whether antipause enhancers, which are the substrates of the JMJD6-Brd4 interaction, are associated with pro-oncogenic gene expression in melanoma. Given the robust correlation between Jmjd6 overexpression and aggressive melanoma in different models and in human samples, it would be important to clarify if the prooncogenic function of JMJD6 depends on this transcriptional activity. More recently, Liu et al. (2017) have shown that the JMJD6 promotes melanomagenesis through the regulation of the alternative splicing of PAK1, a key MAPK signaling component. In recent years, also RNA binding abilities of Jmjd6 are emerging, perhaps its many functions, linked to arginine methylation of RNA binding proteins in transcription initiation complexes, spliceosome and ribosomes may be mediated by its RNA-binding abilities. Indeed Heim et al. (2014), showed that treatment with RNAse disrupt Jmjd6 nuclear localization.

In our study we found that two microRNAs induced by Ras (miR-146a and miR-193a) behave as tumor suppressors. MiR-146a and b are well-studied for their roles in immune responses. Induced by inflammation and infection through NF-kB signaling (Bhaumik et al., 2008), miR-146a, and to a lesser extent miR-146b, function in a negative feedback loop to downregulate several proinflammatory effectors. Indeed miR-146a has been implicated in attenuating pro-inflammatory responses (Saba et al., 2014), which could contribute in cancer to the elimination of early transformed cells by the immune system, and in reducing the

metastatic potential of breast cancer cells (Bhaumik et al., 2008). Increase in expression of miR-146a was reported in unclassified melanoma samples (Philippidou et al., 2010), and found to suppress brain metastasis from melanoma in experimental models (Hwang et al., 2013). These findings support a tumor suppressor role for miR-146a in melanoma. Our study provides evidence that miR-146a is induced by the same MAPK/AKT pathway that sustains melanoma growth at the earliest stages of transformation, when it could exert an anti-melanoma function by repressing Jmjd6 and perhaps also pro-tumoral inflammation. The tumor suppressive activity of miR-193a has been related to its ability to inhibit cell proliferation and to promote apoptosis of cancer cells (Nakano et al., 2013; Mamoori et al., 2018). MiR-193a undergoes epigenetic silencing in acute myeloid leukemia by the AML1/ETO fusion protein (Li et al., 2013), where miR-193 was found to repress several pro-leukemogenic factors, including KIT. Moreover miR193a has been found to participate in a regulatory loop that controls p53 family member levels (Ory et al., 2011) thus further reinforcing the hypothesis that miR193a is a tumor suppressor. The finding that both microRNAs target Jmjd6 and that this regulation is conserved in zebrafish and human melanoma suggests that lowering JMJD6 levels in cancer is another function of the tumor suppressors miR-146a and miR-193a.

The mechanisms through which Jmjd6 promote melanoma progression is still unknown. Different approaches, including transcriptome, epigenome and spliceosome analysis, can help to gain insights into the pro-oncogenic activity of JMJD6. In these studies, the transgenic lines kita:Jmjd6 and kita/Ras/Jmjd6 will provide a source of GFP tagged Jmjd6 proteins expressed specifically in melanocytes. We were puzzled by the massive presence of L-plastin positive cells (leukocytes, possibly including neutrophils and macrophages) within the kita/Ras/Jmjd6 melanomas, compared to kita/Ras melanomas. Neutrophils and eosinophil infiltrations in melanoma predict unfavorable disease outcome (Ding et al., 2018) whereas accumulation of dendritic cells and T-lymphocytes is positively correlated with survival. Therefore, the increase of L-plastin positive neutrophils and macrophages is an index of the aggressiveness of these tumors. It is intriguing that one of the microRNAs targeting Jmjd6 in this progressive model of melanoma is miR-146a, a well-known regulator of inflammatory responses. It would be interesting to investigate whether miR-146a levels differ in melanoma developing in kita/Ras (where they are increased according to RNA-Seq data) versus those developing in kita/Ras/Jmjd6.

In summary, this study has shown that several microRNAs are induced by RAS signaling in melanoma initiating cells at the onset of transformation. Induced miR-146a and - 193a target Jmjd6, which is temporarily downregulated. At later stages of melanoma development, and in human malignant melanoma with unfavorable prognosis, Jmjd6 is overexpressed, and in a zebrafish model where melanocytes express both Ras and Jmjd6, melanomas are more aggressive. Therefore, Jmjd6 has pro-oncogenic activities and the initial downregulation mediated by microRNAs may be part of an anti-oncogenic response, which is induced by the same RAS oncogene.

# MATERIALS AND METHODS

# Zebrafish Lines

Zebrafish were maintained and staged as described (Kimmel et al., 1995).


#### miRNA Array

Custom-designed 8 × 15k microarray slides were ordered from Agilent Technologies. The 15k custom design was obtained from Edwin Cuppen and Eugene Berezikov (Hubrecht Institute, Utrecht, Netherlands) and has been submitted into the Gene Expression Omnibus (GEO) database (GPL 15403). The 15k design contained a duplicate of 7604 probes of 60-oligonucleotide length. The probes consisted of 2 × 22 nucleotide sequences antisense to mature miRNAs separated by a spacer of 8 nucleotides (CGATCTTT) and with a second spacer with the same sequence at the end. From 7604 probes 546 were designed for left (5<sup>0</sup> ) and right (3<sup>0</sup> ) arms of the hairpins of zebrafish miRNAs that are known in miRBase, while the remainder 7058 probes corresponded to predicted hairpin structures in the zebrafish genome that might include additional miRNAs but were not considered in this study. Total RNA, including microRNA, were extracted from 3-7-14 dpf wt or UAS-RAS larvae (driven by kita:Gal4, zic:Gal4, or hsp:Gal4), and from adult melanoma or wt fin tissue using miRNeasy Mini Kit <sup>R</sup> (Qiagen). Three biological replicates were obtained for each condition. For dual color hybridization of the Agilent chips miRNA samples from RAS transgenics were labeled with Hy3 and samples from control fish were labeled with Hy5 with miRCURYTM LNA microRNA, Hy3TM/Hy5TM Power Labeling kit (Exiqon) using 1 microgram of total RNA according to the manufacturer's instructions. The dual color hybridization of the microarray chips was performed according to Agilent protocol GE2\_105\_Jan09<sup>8</sup> for two-color microarray-based gene expression analysis except that hybridization and washing was performed at 37◦C. The arrays were scanned with DNA Microarray Scanner G2565CA from Agilent Technologies. The arrays were scanned twice with 10% PMT and 100% PMT laser power. Microarray data was processed from raw data image files with Feature Extraction Software 9.5.3.1 (Agilent Technologies). The XDR function was used to extend the dynamic range. Processed data were subsequently imported into Rosetta Resolver 7.1 (Rosetta Biosoftware, Seattle, WA, United States) and subjected to default ratio error modeling. The raw data have been submitted to GEO under accession number GSE 37015. Values above 1.2 log2 Fold Changes and p-values < 0.01 in all three models were used as selection criteria.

#### QPCR for MicroRNAs

fgene-09-00675 December 15, 2018 Time: 15:13 # 10

To confirm microarray data, total RNA, including miRNA, were purified from 3 dpf embryos using miRNeasy Mini Kit <sup>R</sup> (Qiagen). Mature miRNA were reverse transcribed using specific primers mix for each miR to produce different cDNA for TaqMan <sup>R</sup> MicroRNA assay (30 ng of total mRNA for each reaction) (Applied Biosystems). Taqman probes were designed for the active strand (−5p) of miR-21, miR-146a, and miR-146b and for the active (−3p) strand of miR-193a. Real-time PCR reactions based on TaqMan reagent chemistry were performed in duplicate on ABI PRISM <sup>R</sup> 7900HT Fast Real-Time PCR System (Applied Biosystems). The level of miRNA expression was measured using C<sup>T</sup> (cycle threshold). For normalization, miR-133, which was unaffected by RAS overexpression, was taken as reference. Fold change was generated using the equation 2 <sup>−</sup>CT. The list of oligos used and their sequences is provided in **Supplementary Table S2**.

# Inhibitor Treatment

Three dpf HS-RAS and AB embryos were incubated in 2ml E3 medium in a 12 well plate in the presence of the following inhibitors: 1 µg/ml PD98059 (Calbiochem), 15 µM LY294002 (Cell Signaling) and 1 µM Rapamycin (Tocris). After 2 h, embryos were heat-shocked at 39◦C for 30 min. After 6 h 50 RAS-GFP+ or GFP- embryos were collected in Trizol or sample buffer (2% SDS, 10% glycerol, 60 mM Tris pH 6.8) for miRNA or protein extraction respectively.

#### Western Blot and Antisera

Ten 3 dpf embryos were homogenized in 200 µl sample buffer (2% SDS, 10% glycerol, 60 mM Tris pH 6.8). 30 µg of total extract were resolved by SDS-PAGE, transferred to nitrocellulose and tested with the following antibodies: phospho-p44/42 (1:1000, Cell Signaling), p44/42 (1:1000, Cell Signaling), Phospho-AKT (1:1000, Cell Signaling), AKT (1:1000, Cell Signaling), Phospho-S6 (1:1000, Cell Signaling), S6 (Cell Signaling 1:1000), jmjd6 (Abcam, 1:1000), GFP (1:1000, Torrey Pines, United States), actin (1:5000, MP Biomedical).

# Manual Inspection of 3<sup>0</sup> UTR of Candidate MicroRNA Targets

An initial analysis performed with MicroCosm (now incorporated in tools4mirs at https://tools4mirs.org/) which used a previous version of the zebrafish genome identified a few genes as potential targets of both miR146a and 193a. Manual inspection of the 3<sup>0</sup> UTR regions of these genes using the GRCz11 release at Ensembl<sup>5</sup> confirmed that only jmjd6 3 0 UTR region contained sequences that match the seed regions of both microRNAs. We then used a web-based interactionprediction algorithm for RNA molecules, IntaRNA<sup>4</sup> (Mann et al., 2017) for the fast and accurate prediction of the interactions between miR-146a and/or 193a with jmjd6 mRNAs (Ensemble, ENSDARG00000102896<sup>5</sup> ). We followed the web-site instructions and used jmjd6 and miR-146a-5p and -193a-3p sequences as input. The following parameters were used: minimum number of base pairs in seed: 7; temperature for energy computation: 37◦C; energy parameter set (Vienna package), Turner Model 2004 (Lorenz et al., 2011); energy interaction levels < −8.

# Morpholino, Duplexes, and Plasmids

Morpholinos, including a standard control morpholino, were obtained from Gene Tools (United States), titrated to non-toxic concentrations and injected in a final volume of 2 nl per 1 cell embryo. A list of all morpholinos, their sequences and the concentration used is provided in **Supplementary Table S3**. Synthetic miR- duplexes controls and 146a were designed and ordered from SIGMA (United States). Synthetic miR- duplexes for miR-193a were purchased from Ambion. Duplexes were dissolved in RNAse free water and diluted using annealing buffer (30 mM HEPES-KOH pH 7.4, 10 mM KCl, 2 mM MgCl2, 50 mM NH4Ac) to a final concentration of 10 µM for miR-146a 5 µM for miR-193a. The solution was incubated for 1 min at 90◦C, cooled down slowly to room temperature and injected in a final volume of 2 nl per 1-cell embryo. A list of all duplexes, their sequences and the concentration used is provided in **Supplementary Table S4**.

For Jmjd6 reporter construct, the whole Jmjd6 (Ensemble, ENSDARG00000102896) 3<sup>0</sup> UTRs was PCR amplified from cDNA using specific primers (**Supplementary Table S2**). The PCR product was subcloned into the pCS2:eGFP vector downstream of the GFP open reading frame and confirmed by sequencing.

# Generation of a Jmjd6 Transgenic Line

To express a microRNA resistant Jmjd6 in melanocytes, the full coding region excluding the stop codon of Jmjd6 (Ensemble, ENSDARG00000102896) was PCR amplified and cloned into pEntry5-no stop (Invitrogen) and then recombined using gateway technology with a pEnt5-4nrUAS (generated by cloning four non-repetitive UAS elements into pENT5<sup>0</sup> from Invitrogen), a p3E-EGFPpA (Tol2Kit clone n.366) + pDestTol2CG2 (Tol2Kit clone n. 395) using gateway (Kwan and North, 2017). Clone numbers and sequences can be found at: http://chien.neuro.utah.edu/tol2kitwiki/index.php/Main\_Page.

Final recombined clones were checked by sequencing. 4nrUAS:eGFP-Jmjd6 plasmid was injected in 1-cell embryos

<sup>8</sup>www.Agilent.com

of the Et(kita:Gal4TA, UAS:mCherry)hmz1 line to generate a transgenic line using standard Tol2 mediated transgenesis (Kawakami, 2004; Pase and Lieschke, 2009).

### Zebrafish Embryo Injections

fgene-09-00675 December 15, 2018 Time: 15:13 # 11

Zebrafish embryos at the stage of 1–2 cells were injected with morpholinos against miRs or Jmjd6 diluted in double distilled, sterile H2O. The morpholino oligonucleotides were injected at a concentration of 5 ng/nl, in a volume of 2 nl/embryo. miR duplexes mimicking mature microRNA were injected following a described protocol (Pase and Lieschke, 2009), at a concentration of 10 µM for miR-146a and miR-146b and 5 µM for miR193a in a volume of 2 nl/embryo.

Sensor injections: mRNA encoding for eGFP carrying Jmjd6 3 <sup>0</sup> UTR was in vitro synthesized and injected in 1–2 cell zebrafish embryos at 100 pg/embryo, alone or in combination with duplexes or morpholinos.

### Imaging

Photographs of whole larvae were acquired with a Nikon S100 stereomicroscope equipped with epifluorescence and multiple filters. We used a Leica SP5 confocal for analysis of melanocytes expressing Ras-eGFP, Jmjd6-GFP, and L-plastin immunoreactivity.

### AGO2 – RIP

One hundred 3 dpf embryos injected or not with miR duplex were homogenized in ice-cold buffer A containing 50 mM Tris-HCl pH 7.5, 5 mM EDTA, 5 mM EGTA, complete proteinase inhibitor (Roche) and 1 U/µl RNAse inhibitor SUPERnase-IN (Ambion). Unbroken cells were removed at 100 g for 5 min at 4 ◦C. NP40 was added at 0.5% (wt/vol) and samples incubated at 4 ◦C for 15 min with rotation and centrifuged at 10000 × g for 15 min. SN was the transferred to a new tube and the protein were measured by BCA. 2 mg of protein were diluted with buffer A and incubated overnight at 4◦C with 50 µl Dynabeads (Invitrogen) bound with 2 µg Ago2 antibody (Abcam) or IgG (negative control). The next morning beads were washed four times with 1 ml of washing buffer containing 50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM MgCl<sup>2</sup> and 0.05% NP40. Beads were then resuspended in 1ml of washing buffer and 200 µl of beads were pelleted and resuspended in 50 ul sample buffer to test Ago2 by western blot (Abcam 1:1000). The remaining 800 ul were resuspend with 500 µl of trizol and RNA was extracted by RNeasy micro Kit (Qiagen). All the RNA was retro-transcribed using Vilo (Invitrogen) and 1 µl of cDNA was used for PCR using specific primers for jmjd6 or the housekeeping gene, gapdh.

### Analysis of Data From cBioportal and Protein Atlas

Analysis of JMJD6 expression, correlation with HRAS or BRAF mutations, survival and disease/progression-free survival in patients was done using the dataset Skin Cutaneus Melanoma TCGA dataset in cBioportal<sup>6</sup> . The survival and disease/progression-free survival were done taking into account the following JMJD6 alterations: mutations, amplifications, deep deletions or multiple alterations.

We report here examples of JMJD6 protein localization in control skin (ID: 2396<sup>9</sup> ) and in a malignant melanoma (ID: 2136<sup>10</sup>) from the Protein Atlas website.

# AUTHOR CONTRIBUTIONS

VA and MM design the study, conducted experiments, and wrote the manuscript. VA, AO and AM performed the microarray analysis. SK, LS, and VG performed bioinformatics analysis. MS and AM provided conceptual insights.

### FUNDING

Work in the lab of MCM is supported the European Commission under the Horizon 2020 program (U. M. Cure/Project No. 667787), and by a 5 × 1000 LILT-2016 contribution. AM acknowledges the contribution of the European Commission 6th Framework Project ZF-TOOLS (LSHG-CT-2006-037220).

### ACKNOWLEDGMENTS

We would like to thank Cristina Santoriello for help and discussion on the experimental design, Nadine Borel and EZRC personnel for fish care, Reinhard Koester and Martin Distel for sharing the Gal4 lines used in this study, Kyuson Yun for reading and commenting on an earlier version of the manuscript, Erik Dassi for help with the heatmap in **Figure 1**, and Sabrina Burkart for technical support.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fgene.2018. 00675/full#supplementary-material

FIGURE S1 | Output of IntaRNA website, showing interaction of microRNAs 146a (A) and 193a (B) with the 3<sup>0</sup> UTR of jmjd6. An energy of −8 Kcal/mol or less is indicative of a stable interaction.

FIGURE S2 | Injections of duplexes microRNAs cause increase of the corresponding microRNAs and mild developmental defects. (A) QPCR analysis of the expression levels of the microRNAs indicated. (B) All the other injected embryos developed without visible abnormalities.

FIGURE S3 | Jmjd6 is down-regulated upon Ras expression. Jmjd6 mRNA (A) and protein levels (B) are downregulated by Ras overexpression (2nd lane) and rescued to normal levels by downregulating the increased microRNAs (lanes 3–4).

<sup>9</sup>https://www.proteinatlas.org/ENSG00000070495-JMJD6/tissue/skin <sup>10</sup>https://www.proteinatlas.org/ENSG00000070495-JMJD6/pathology/tissue/ melanoma#img

FIGURE S4 | cBioPortal data of JMJD6. (A) Alteration frequency of JMJD6 in different type of cancers. (B) JMJD6 expression in melanoma patients; each bar represents a patient. Red bars: patients with upregulated JMJD6. Blue bars: patients with downregulated JMJD6. JMJD6 expression in HRAS (C) and BRAF (D) mutated melanoma. Overall survival (E) and disease/progression-free survival (F) in patients with (red curve) or without (blue curve) JMJD6 alterations.

TABLE S1 | miRNA array results table. Column A: probe ID; column B: sequence of the probe; column C: Gene ID; column D: Gene symbol; Columns E–P: fold change and relative p-value of the samples as indicated. Fold change and p-value are calculated by comparing the miRNAs expression in kita-Ras larvae at 3, 7, and 14 dpf, HS-Ras, and zic at 3 dpf and adult melanoma with the respective control

#### REFERENCES


(AB for all the larvae and skin from adult AB wild type fish for the adult melanoma). MicroRNAs selected for further studies (see text for criteria used) and their expression values (log2FoldChanges and pValues) are highlighted.

TABLE S2 | Oligo sequences. Sequences of the oligos used for qPCR, RIP assay, jmjd6 3 0 -UTR cloning and Gateway cloning of Jmjd6.

TABLE S3 | Morpholino sequences. Sequences of the morpholinos used in the study.

TABLE S4 | microRNA duplexes. Sequences of the miR control (miR-C) and miR-146a\_l injected in the embryos. The sequence of miR-193a was not provided by the supplier (Ambion).

specifically interacts with arginine-serine-rich (RS) domains of SR- and SR-like proteins. Nucleic Acids Res. 42, 7833–7850. doi: 10.1093/nar/gku488



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

Copyright © 2018 Anelli, Ordas, Kneitz, Sagredo, Gourain, Schartl, Meijer and Mione. 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.

# High MYC Levels Favour Multifocal Carcinogenesis

Manuela Sollazzo\* † , China Genchi† , Simona Paglia, Simone Di Giacomo, Annalisa Pession, Dario de Biase and Daniela Grifoni\*

Cancer Evolution Laboratory, Department of Pharmacy and Biotechnology, University of Bologna, Bologna, Italy

#### Edited by:

Maria Grazia Giansanti, Consiglio Nazionale delle Ricerche (CNR), Italy

#### Reviewed by:

Louise Cheng, Peter MacCallum Cancer Centre, Australia Romain Levayer, Institut Pasteur, France

#### \*Correspondence:

Manuela Sollazzo manuela.sollazzo2@unibo.it Daniela Grifoni daniela.grifoni@unibo.it

#### †Present address:

Manuela Sollazzo, Cellular Biochemistry Laboratory, Department of Pharmacy and Biotechnology, University of Bologna, Bologna, Italy China Genchi, Laboratory of Membrane Trafficking, Signaling and Development, Department of Biosciences, University of Milan, Milan, Italy

#### Specialty section:

This article was submitted to Genetic Disorders, a section of the journal Frontiers in Genetics

Received: 15 September 2018 Accepted: 20 November 2018 Published: 11 December 2018

#### Citation:

Sollazzo M, Genchi C, Paglia S, Di Giacomo S, Pession A, de Biase D and Grifoni D (2018) High MYC Levels Favour Multifocal Carcinogenesis. Front. Genet. 9:612. doi: 10.3389/fgene.2018.00612 The term "field cancerisation" describes the formation of tissue sub-areas highly susceptible to multifocal tumourigenesis. In the earlier stages of cancer, cells may indeed display a series of molecular alterations that allow them to proliferate faster, eventually occupying discrete tissue regions with irrelevant morphological anomalies. This behaviour recalls cell competition, a process based on a reciprocal fitness comparison: when cells with a growth advantage arise in a tissue, they are able to commit wild-type neighbours to death and to proliferate at their expense. It is known that cells expressing high MYC levels behave as super-competitors, able to kill and replace less performant adjacent cells; given MYC upregulation in most human cancers, MYC-mediated cell competition is likely to pioneer field cancerisation. Here we show that MYC overexpression in a sub-territory of the larval wing epithelium of Drosophila is sufficient to trigger a number of cellular responses specific to mammalian premalignant tissues. Moreover, following induction of different second mutations, high MYC-expressing epithelia were found to be susceptible to multifocal growth, a hallmark of mammalian pre-cancerous fields. In summary, our study identified an early molecular alteration implicated in field cancerisation and established a genetically amenable model which may help study the molecular basis of early carcinogenesis.

#### Keywords: MYC, field cancerisation, multifocality, Drosophila, TSGs, cell competition

# INTRODUCTION

The molecular events underlying cancer initiation are largely unknown. It is commonly accepted that most cancers are monoclonal in origin, evolving from a single cell whose lineage accumulates in time multiple molecular insults (Michor et al., 2004; Vogelstein et al., 2013; Feinberg et al., 2016). In particular, driver mutations, which provide cells with a growth advantage and are positively selected during lineage evolution, are generally associated with clonal expansion and are frequently found in pre-malignant lesions (Maley et al., 2004; Lawrence et al., 2014; Curtius et al., 2017). In the 1950s, Slaughter introduced the concept of "field cancerisation": while studying oral cancers, he observed that they recurred more frequently adjacent to a resected tumour (Slaughter et al., 1953). Therefore, field cancerisation was defined as the process leading to the formation of a tissue sub-territory which, despite a normal appearance, bears a series of alterations that make cells more susceptible to malignant transformation than wild-type neighbours, giving rise to multifocal cancers (Wodarz et al., 2004). Successive studies, also fostered by the development of post-genomic technologies (Metzker, 2010), have demonstrated that this phenomenon is not specific to the oral mucosa, being rather a common feature of epithelial organs (Braakhuis et al., 2003; Dakubo et al., 2007;

Nonn et al., 2009; Zeki et al., 2011; Jakubek et al., 2016; Lu et al., 2016; Park et al., 2016; Abdalla et al., 2017; Castven et al., 2017).

Although the interest in deciphering cancer's molecular signature is obvious, it may be quite difficult to understand what mutations favour and maintain the malignant phenotype: a surprising number of driver mutations is indeed present in pre-cancerous tissues, also in those that are not likely to evolve into a frank malignancy (Hofstad et al., 1996; Martincorena et al., 2015; Kato et al., 2016). This suggests that several alterations are evolutionarily neutral and do not impact cell's phenotype, maybe depending on their temporal occurrence (de Bruin et al., 2014), the tissue context (Galandiuk et al., 2012; Gagneur et al., 2013; Vermeulen et al., 2013) and the genetic background (Chandler et al., 2013). In human tissues, a number of genetic alterations have been associated with field cancerisation (Papadimitrakopoulou et al., 1996; Braakhuis et al., 2002; Santos-Garcia et al., 2005; Haaland et al., 2009; Trujillo et al., 2011; Mohan and Jagannathan, 2014), and genetic/genomic instability (Ellsworth et al., 2004; Zaky et al., 2008; Giaretti et al., 2012), mitochondrial defects (McDonald et al., 2008; Maggrah et al., 2013; Parr et al., 2013), production of reactive oxygen species (ROS) (Bongers et al., 1995; Jaloszynski et al., 2003; Chan et al., 2017), increased expression of proliferation and apoptosis markers (Birchall et al., 1997; Bascones-Martinez et al., 2013) and epigenetic modifications (Grady, 2005; Lee et al., 2011; Kamiyama et al., 2012; Luo et al., 2014) are also repeatedly found in regions adjacent to malignant tumours from a variety of organs. Whatever the cause of these modifications, from DNA replication errors to mutagenic injuries, the ongoing precancerous field will most likely be composed of a number of genetically different clones, with the fittest one expected to colonise the entire territory over time (Driessens et al., 2012). This process of selection based on fitness comparison is a distinctive trait of cell competition (CC), a phenomenon first observed and characterised in Drosophila (Morata and Ripoll, 1975), and then demonstrated to be conserved in mammals (Penzo-Mendez and Stanger, 2014; Di Gregorio et al., 2016).

Competitive interactions are typically triggered when cells with different proliferation rates are found in close proximity: the fittest cells (winners) commit less fit neighbours (losers) to death and overgrow to replace them in the tissue (Levayer and Moreno, 2013, 2016; Tamori and Deng, 2013; Tsuboi et al., 2018). A number of molecules and signalling pathways have to date been found to play a role in CC (Moreno et al., 2002; Tyler et al., 2007; Vincent et al., 2011; Rodrigues et al., 2012; Akai et al., 2018): among these, the MYC protein was shown to be the most powerful inducer of CC (named in this case MYC-Mediated Cell Competition, MMCC) from Drosophila to mammals (Johnston, 2014), paving the way to studies that found this process implicated in a number of seemingly distant contexts, from organ development (de la Cova et al., 2004; Moreno and Basler, 2004; Claveria et al., 2013; Sancho et al., 2013; Villa del Campo et al., 2014; Villa Del Campo et al., 2016) to tissue regeneration (Oertel et al., 2006; Gogna et al., 2015; Rosen et al., 2015; Villa Del Campo et al., 2016; Shakiba and Zandstra, 2017), cell stemness (Rhiner et al., 2009; Diaz-Diaz et al., 2017) and cancer (Froldi et al., 2010; Ziosi et al., 2010; Eichenlaub et al., 2016; Suijkerbuijk et al., 2016). Of note, we and others recently demonstrated that MMCC is also active in human cancer cells (Patel et al., 2016; Di Giacomo et al., 2017). MYC upregulation is sufficient as to transform cells into super-competitors (Moreno and Basler, 2004), able to kill and replace suboptimal neighbours, and this capability has opened to speculations about a possible role for MMCC in field cancerisation (Rhiner and Moreno, 2009; Johnston, 2014). MYC family proteins are long investigated for their essential functions in cell physiology and in cancer (Stine et al., 2015); the Drosophila genome bears a single locus (diminutive, dm) encoding the MYC protein, which exerts the same functions as the mammalian orthologues (Gallant, 2013). MYC overexpression in wild-type cells may provoke a series of contradictory responses: on the one hand, it supports cell growth by accelerating biosynthesis, cell metabolism and cell cycle (Evan and Littlewood, 1993; Grewal et al., 2005; Meyer and Penn, 2008); on the other hand, it promotes potentially harmful reactions such as ROS production and genetic instability (Vafa et al., 2002; Greer et al., 2013; Kuzyk and Mai, 2014), and increases propensity to apoptotic cell death (Montero et al., 2008; McMahon, 2014). Cancer cells upregulating MYC are contrariwise protected from untimely death, primarily due to relevant changes in metabolic pathways leading to MYC addiction (Gabay et al., 2014). MYC seems thus to elicit in normal cells a number of biological responses similar to those found in mammalian pre-cancerous fields (Mohan and Jagannathan, 2014). Moreover, MYC upregulation is an early event in human prostate cancer (Gurel et al., 2008), and MYC overexpression is sufficient to transform luminal epithelial cells into pre-malignant derivatives in the mouse prostatic gland (Kim et al., 2009; Iwata et al., 2010). MYC upregulation has also been observed in cytologically normal bronchial epithelial cells of mice with pre-neoplastic lung squamous cell carcinoma lesions (Xiong et al., 2017), and it was reported to initiate gastric tumourigenesis following Hippo pathway deregulation in the pyloric stem cell (Choi et al., 2018). These observations led us to speculate that high MYC levels may be sufficient for an epithelial tissue to become responsive to the effect of second mutations that would otherwise be irrelevant when occurring in a wild-type epithelium.

In Drosophila, the tumour suppressor genes (TSGs) are historically subdivided into two classes, called "hyperplastic" and "neoplastic" according to the mutant phenotype (Hariharan and Bilder, 2006), most of which have in time been found to encode different components of the Hippo pathway (Grusche et al., 2010), a highly conserved signalling cascade central in cell growth and organ size modulation (Halder and Johnson, 2011). Broadly speaking, loss-of-function (LOF) mutants of these hyperplastic TSGs (fat, ft; dachsous, ds; expanded, ex; warts, wts; and hippo, hpo) show a substantial overgrowth of the larval epithelial organs, called imaginal discs (Aldaz and Escudero, 2010), and premature death at the pupal stage (Hariharan and Bilder, 2006), whereas LOF mutants of neoplastic TSGs do not survive beyond embryogenesis (Menut et al., 2007). An exception is made for scribble (scrib), discs large (dlg), and lethal giant larvae (lgl) neoplastic mutants which, given the abundant maternal transcript released into the zygote, survive up to the end of the larval life, showing abnormal growth of the imaginal discs with

a complete loss of the epithelial structure (Bilder et al., 2000; Bilder, 2004). In case single mutant cells are created in a wildtype background through clonal analysis techniques (del Valle et al., 2012), those bearing hyperplastic LOF mutations survive and overgrow in the target tissue (Xu et al., 1995; Buratovich and Bryant, 1997; Udan et al., 2003; Maitra et al., 2006), whereas those bearing neoplastic LOF mutations are usually eliminated during development (Agrawal et al., 1995; Enomoto and Igaki, 2011). We and others demonstrated these opposite behaviours are dictated by MMCC: while hyperplastic mutant cells upregulate MYC and behave like winners in the wild-type tissue, killing the surrounding neighbours and growing at their expense (Neto-Silva et al., 2010; Ziosi et al., 2010), neoplastic mutant cells do not upregulate MYC and behave like losers in the context, being themselves out-competed by adjacent wild-type cells (Froldi et al., 2010; Menendez et al., 2010). We previously showed that a MYCoverexpressing background strengthens the super-competitive behaviour of ft, ds and ex mutant clones, which were found to kill the surrounding cells with increased efficiency and to grow more rapidly, although it did not provide mutant cells with the capability to evolve into a malignant mass (Ziosi et al., 2010).

Here we expanded on previous work by first identifying in Drosophila MYC-overexpressing epithelial organs a series of morphological and molecular markers typically found in human pre-cancerous fields. Moreover, we investigated the impact of a MYC-overexpressing background on the cellular phenotypes consequent to mutations in neoplastic TSGs, showing it is in this case sufficient to make mutant cells able to initiate multifocal malignant transformation, a peculiar trait of human pre-neoplastic fields.

# MATERIALS AND METHODS

#### Fly Stocks and Manipulation

The following fly lines were used in the study, built using stocks obtained from the Bloomington Drosophila Stock Center, Indiana: w; UAS-GFP(Bl−6874) ; hh-Gal4(Bl−67046) – yw, PI3K92ECAAX(Bl−25908) – w; Ubi-GFPnls, FRT40A(Bl−5629) /CyO; hh-Gal4(Bl−67046) /TM6b – w; l(2)gl<sup>4</sup> P(neo-FRT)40A(Bl−36289) / In(2-3)Gla,Bc; UAS-HAdm/TM6b – w; Rab52P(neo-FRT) 40A(Bl−42702) /In(2-3)Gla,Bc; UAS-HAdm/TM6b. UAS-HAdm on III is a gift of P. Bellosta. Plain genotypes are given for each experiment in the figure legends. For all experiments, flies were kept at 25◦C. Larvae were heat-shocked once at 48 ± 4 h AEL in a water bath at 37◦C for 10 min and dissected after additional 72 h development.

#### Immunofluorescence

Frozen or fresh larvae were prepared for immunofluorescence by standard methods. The following antibodies and dilutions were used: mouse α-MYC (1:5, P. Bellosta); rabbit α-Lgl (1:400, D. Strand); rabbit α-active Caspase 3 (1:100, Cell Signalling Technologies); rabbit α-aPKCζ (1:200, Santa Cruz Biotechnology); rabbit α-pAKT (1:100, Cell Signaling Technologies); rabbit α-PH3 (1:100, Upstate Technology); mouse α-γH2Av (1:30, DSHB); mouse α-dIAP1 (1:100, B. A. Hay); rabbit α-Pc (1:400, Santa Cruz Biotechnology); mouse α-En (1:50, DSHB). Alexa Fluor 555 goat α-mouse and α-rabbit (1:500, Invitrogen) and DyLight 649-conjugated goat α-mouse and α-rabbit (1:750, Jackson ImmunoResearch Laboratories) were used as secondary antibodies. Samples were analysed with a Leica TSC SP2 laser confocal microscope and entire images were processed with Adobe Photoshop software or ImageJ free software from NIH. All the images represent a single confocal stack unless otherwise specified. Image magnification is 400× unless otherwise specified.

# ROS Detection

Larvae were dissected in PBS1X and carcasses were incubated for 30 min at room temperature in PBS1X – DHE (Dihydroethidium, Invitrogen Molecular Probes) at a final concentration of 30 µM in gentle shaking before fixation. Wing discs were immediately imaged under a Nikon 90i wide-field fluorescence microscope.

## Statistical Analysis

For the experiments shown in **Figures 3–7**, the number of wing discs analysed was 15÷25 from different larvae for each sample. For each experiment, the data presented are the average of three biological replicates. Multifocality was assessed on a total of 346 wing discs for l(2)gl<sup>4</sup> clones (see **Figures 9**, **10**), and on a total of 146 wing discs for Rab5<sup>2</sup> (see **Figure 11**). For the experiments shown in **Figures 2**, **8**, **13**, the number of discs analysed is indicated. Mean Fluorescence Intensity (MFI) (**Figures 2**, **5**– **7**), clone area (**Figures 8**, **13**) and positive signals (**Figures 3**, **4**) were calculated by ImageJ free software (NIH) on images captured with a Nikon 90i wide-field fluorescence microscope at a magnification of 200×. All measurements have been taken inside the yellowish area highlighted in **Figure 1A**. P-values were as follows: ∗∗p ≤ 0.01 and ∗∗∗p ≤ 0.001. Mean, SEM and the t-Student test p-value were calculated by using GraphPad Prism software, San Diego, CA, United States.

# RESULTS

### MYC-Overexpressing Tissues Show Several Markers Repeatedly Found in Human Pre-cancerous Fields

Pre-cancerous fields are defined as tissue areas composed of histologically normal but genetically altered cells, shown to be more susceptible than wild-type counterparts to the onset of new mutations, promoting in time the development of multifocal tumours (Slaughter et al., 1953; Dotto, 2014). Since these areas are found to surround primary masses in several epithelial malignancies (Nonn et al., 2009; Zeki et al., 2011; Park et al., 2016), a pre-neoplastic field can be considered, borrowing Paget's hypothesis, a soil providing "bad seeds" with the capacity to initiate malignant growth, including those that would normally fail. The wide series of aberrations underlying the process of field cancerisation can hardly be attributed to a single cellular event, but deregulation of a gene piloting a number of cell behaviours may greatly favour its formation.

MYC represents an excellent candidate, because its misexpression does not account on gene mutation but is rather caused by alterations in many, if not all, signalling pathways (Nussinov et al., 2016). As an example, activated forms of RAS are frequently found in human pre-neoplastic tissues (Braakhuis et al., 2003), and it is known that activated RAS stabilises MYC protein in Drosophila (Prober and Edgar, 2002) and mammals (Sears et al., 2000). Stabilised MYC is in turn able to remodulate cell growth and proliferation, metabolism and stress response (Meyer and Penn, 2008). Moreover, a founder cell upregulating MYC could easily expand into a MYCupregulating field through MMCC (Johnston, 2014). Therefore, MYC could play a causative role both in driving the expansion and in determining the intrinsic characteristics of a precancerous field. To investigate this issue we bypassed field formation, since it is well established that Drosophila epithelial cells upregulating MYC eliminate the wild-type neighbours during development and colonise a large fraction of the tissue through MMCC (de la Cova et al., 2004; Moreno and Basler, 2004).

We then took advantage of the UAS-Gal4 binary system (Brand and Perrimon, 1993) to drive MYC overexpression (hereafter referred to as MYCOVER) under the control of

FIGURE 2 | MYC and PI3KCAAX overexpression do not induce reciprocal activation. (A–A2) Immunostaining for pAKT (A,A1, cyan) and MYC (A2, red) on wing discs from late yw/yw, UAS-PI3KCAAX; hh-Gal4, UAS-GFP/+ larvae. (B) Graph comparing the Mean Fluorescence Intensity Arbitrary Units (MFI-AU) of MYC staining measured in the P (green bar) and A (grey bar) compartments of 15 wing discs from different larvae. (C,C1) Immunostaining for pAKT (cyan) on wing discs from late yw; hh-Gal4, UAS-GFP/UAS-dm larvae. (D) Graph comparing the Mean Fluorescence Intensity Arbitrary Units (MFI-AU) of pAKT staining measured in the P (green bar) and A (grey bar) compartments of 15 wing discs from different larvae. Basic genotypes are indicated on the left of the figure panel. P compartments are outlined in A2,C1, and disc axes are indicated in A,C. Magnification is 400×.

the hedgehog (hh) promoter in the posterior compartment of the wing disc, a Drosophila larval epithelial organ (Bryant, 1975). **Figure 1A** shows the Posterior/Anterior (P/A) and the Dorsal/Ventral (D/V) axes of the larval wing disc, while the yellowish region represents the area subjected to measurements and P vs. A comparisons, being the notum mostly composed of anterior cells (see P/A boundary in the notum region). As can be appreciated in **Figure 1B**, MYCOVER is confined to the P compartment (representing the pre-cancerous field), where it does not seem to cause evident alterations in tissue morphology with respect to the A compartment (representing the wild-type field), as noted in **Figures 1B1,B2**, where a Ubi-GFPnls transgene and the Lgl protein mark cell nuclei and cell membranes, respectively. To demonstrate MYC's specificity in providing cells with a complex pre-cancerisation signature, we

FIGURE 3 | MYC and PI3KCAAX overexpression increases mitotic activity. (A,A1) Immunostaining for PH3 (red) on wing discs from late yw; hh-Gal4, UAS-GFP/UAS-dm larvae. (B) Graph comparing the PH3-positive nuclei counted in the P (green bar) and A (grey bar) compartments, ∗∗∗p ≤ 0.001. (C,C1) Immunostaining for PH3 (red) on wing discs from late yw/yw, UAS-PI3KCAAX; hh-Gal4, UAS-GFP/+ larvae. (D) Graph comparing the PH3-positive nuclei counted in the P (green bar) and A (grey bar) compartments, ∗∗p ≤ 0.01. (E) Graph comparing the PH3-positive nuclei counted in the P compartments of yw; hh-Gal4, UAS-GFP/UAS-dm (striped green bar) and yw/yw, UAS-PI3KCAAX; hh-Gal4, UAS-GFP/+ (dotted green bar) larvae, ∗∗∗p ≤ 0.001. Basic genotypes are indicated on the left of the figure panels and under the graphs. P compartments are outlined in A1,C1, and disc axes are indicated in A,C. Magnification is 400×.

compared the results of each experiment with those obtained following overexpression of a membrane-tethered form of PI3K (PI3KCAAX), another potent growth inducer (de la Cova et al., 2004). We first verified if overexpression of the PI3KCAAX transgene (PI3KCAAX−OVER) caused consistent activation of the PI3K/AKT signalling pathway. As noted in **Figures 2A,A1**, the phosphorylated form of AKT was detected in the P compartment of the wing disc following PI3KCAAX−OVER (GFP<sup>+</sup> region in **Figure 2A**). Moreover, it did not impact MYC endogenous levels (**Figure 2A2**, the P/A border is outlined), being the MFI of MYC staining statistically comparable in P and A compartments (**Figure 2B**). Following MYCOVER, the levels of phosphorylated AKT in the P compartment (GFP<sup>+</sup> region, **Figures 2C,C1**, the P/A border is outlined) were also comparable to those observed

in the A compartment (**Figure 2D**), confirming that, differently from what has been observed in a previous study (Levayer et al., 2015), in our genetic system and under our working conditions, the two growth inducers do not significantly cross-regulate each other, making it suitable for the successive analyses.

We started by investigating in the MYCOVER tissue a number of markers characteristic of human pre-neoplastic fields. Since it is known that pre-malignant areas may display a higher proliferative index than normal tissues (Mohan and Jagannathan, 2014), we first checked the mitotic activity of MYCOVER cells by immunostaining for the phosphorylated histone H3 (PH3), which is known to play a key role during mitosis both in Drosophila and mammals (Kamakaka and Biggins, 2005). A mitotic index analysis highlighted a 32% increase of PH3 positive nuclei in MYCOVER P compartments with respect to their A counterparts (**Figures 3A,A1,B**), and a 20% increase in the PI3KCAAX−OVER P vs. A compartments (**Figures 3C,C1,D**). This result was not unexpected, as PI3K activation plays important

positivity measured in the P (green bar) and A (grey bar) compartments, ∗∗∗p ≤ 0.001. (C,C1) ROS production (red) in wing discs from late yw/yw, UAS-PI3KCAAX; hh-Gal4, UAS-GFP/+ larvae. (D) Graph comparing the Mean Fluorescence Intensity Arbitrary Units (MFI-AU) of ROS positivity measured in the P (green bar) and A (grey bar) compartments. (E) Graph comparing the Mean Fluorescence Intensity Arbitrary Units (MFI-AU) of ROS positivity measured in the A compartments of yw; hh-Gal4, UAS-GFP/UAS-dm (striped grey bar) and yw/yw, UAS-PI3KCAAX; hh-Gal4, UAS-GFP/+ (dotted grey bar) larvae, ∗∗∗p ≤ 0.001. Basic genotypes are indicated on the left of the figure panels and under the graphs. P compartments are outlined in A1,C1, and disc axes are indicated in A,C. Magnification is 400×.

roles in cell growth and proliferation (Leevers et al., 1996). The mitotic index of the MYCOVER tissue was, however, significantly higher than that observed in the PI3KCAAX−OVER samples, as in the graph reported in **Figure 3E**. To assess genetic instability, another feature of pre-cancerous fields with obvious mutagenic effects (Bhattacharjee and Nandi, 2016), we used an antibody against the γ variant of the phosphorylated histone H2, which is recognised as the first modification occurring following DNA double strand breaks, resulting in the assembling of multi-protein complexes which attempt to repair DNA damage (Dronamraju and Mason, 2011). As can be seen in **Figures 4A,A1**, the γH2Av foci (red) in the MYCOVER P compartment (GFP+,

outlined in **Figure 4A1**) were about twice compared to the A compartment (**Figure 4B**), while they resulted comparable in the P and A compartments of the PI3KCAAX−OVER samples (**Figures 4C,C1,D**). Our study continued by evaluating the presence and abundance of ROS in the presumptive precancerous field. As noted in **Figures 5A,A1**, a strong increase in ROS generation (red) was found in the MYCOVER P compartment of the wing disc (GFP+, outlined in **Figure 5A1**), quantified as about 20 arbitrary units (AU) MFI vs. the 6.5 AU found in the A compartment (compare green and grey bars in the graph **Figure 5B**). Contrariwise, no significant differences were found between the P (GFP+, outlined in **Figure 5C1**) and A compartments following PI3KCAAX−OVER, as is appreciable in **Figures 5C,C1,D**. As MYCOVER and PI3KCAAX−OVER samples underwent parallel enzymatic reactions, we could also compare

bars) and A (grey bars) compartments. Basic genotypes are indicated under the figure panels. P compartments are outlined in A1,A2,C1,C2, and disc axes are indicated in A,C. Magnification is 400×.

ROS levels in the respective wild-type A compartments, and found that MYCOVER A compartment showed a twofold ROS increase with respect to the PI3KCAAX−OVER A compartment (**Figure 5E**). This was an interesting finding, as ROS are diffusible ions and molecules and they may freely move away from the producing cells, thus expanding MYC's pre-cancerisation effect to adjacent tissues by a non-autonomous mechanism. In this sense, a recent study demonstrated that, in Drosophila epithelial tumours, apoptotic caspases enhance tumour malignancy by generating ROS, which in turn recruit immune cells that signal back to the epithelium to activate cancer pathways (Perez et al., 2017). Although MYCOVER tissues cannot be compared to overt cancers, similar cell-cell interactions may be at work that cooperate with MMCC to expand the pre-cancerous field. An analysis of apoptotic cell death carried out by immunostaining for the activated form of the effector Caspase 3 (Cas3) revealed that MYCOVER epithelial cells were highly prone to apoptotic death (see **Figures 6A,A1**), with about 9 MFI AU in the P compartment vs. 1.5 in the A counterpart, as can be noted in **Figure 6B**. By contrast, no significant differences were noticed between P and A compartments overexpressing PI3KCAAX (**Figures 6C,C1,D**). Consistently with Caspase 3 activation, MYCOVER cells downregulated the anti-apoptotic protein dIAP1 (Wang et al., 1999), as shown in **Figure 7A1** (**Figure 7A** shows the GFP<sup>+</sup> P compartment), with 7.4 MFI AU in the P vs. 13.5 in the A compartment (**Figure 7B**); dIAP1 indeed functions by inhibiting the initiator caspase DRONC (Meier et al., 2000) that, in turn, activates the effector caspases. Also in this case, PI3KCAAX−OVER tissues did not show significant differences in dIAP1 staining between the P and A compartments (**Figures 7C,C1,D**). Finally, with regard to changes in the epigenetic signature of human preneoplastic tissues (Grady, 2005; Lee et al., 2011), we analysed the effect of MYCOVER on the chromatin modifier Polycomb (Pc), known to shape cellular plasticity through large-scale epigenetic regulation (Klebes et al., 2005). We previously showed that Pc expression is nearly absent in Drosophila epithelial cancers (Grifoni et al., 2015). As it is known that Pc and other proteins of the Pc group (PcG) are necessary to MYC autorepression in Drosophila (Goodliffe et al., 2005; Khan et al., 2009), Pc downregulation in overt cancers may help sustain high MYC cellular levels, so allowing it to impact many different phenotypic traits. As can be observed in **Figure 7A2**, Pc resulted downregulated also in our pre-cancerisation model, with 9.8 vs. 19.4 MFI AU in the P (GFP<sup>+</sup> in **Figure 7A**, outlined in 7A2), and A compartments, respectively (**Figure 7B**). This is consistent with MYC and PcG proteins trans-regulation (Benetatos et al., 2014), and since low Pc levels result in a higher chromatin accessibility, this condition would favour additional mutational insults through inappropriate entrance of DNA cleaving enzymes (Zhang et al., 2008). Also in this case, PI3KCAAX−OVER tissues did not display significant differences compared to the wild-type counterparts (**Figures 7C2,D**).

Altogether, these results support our hypothesis that high levels of MYC are sufficient as to induce a series of molecular changes, which are likely to turn the affected tissue into a premalignant field. Moreover, this ability seems to be specific to MYC, as an active form of the growth inducer PI3K failed to promote significant alterations of the markers analysed.

#### Single-Cell Mutations of Neoplastic TSGs Initiate Multifocal Growth in a MYCOVER Tissue

With the aim to translate the evidence described above into a functional demonstration of MYCOVER's capacity to establish

a pre-malignant condition, we investigated the phenotypic consequences of the induction of second mutations in a MYCOVER background. We used a genetic model which, through a combination of the UAS-Gal4 (Brand and Perrimon, 1993) and Flp-FRP (Xu and Rubin, 1993) binary systems, allowed us to express MYC in the P compartment and to induce second mutations of interest later in time, so reproducing the temporal sequence that is likely to occur during cancer initiation. The A compartment has been used as a control, to assess the clonal phenotype promoted by the same second mutations in a region carrying endogenous MYC expression.

As described in the Introduction, we previously showed that hyperplastic TSGs (hTSGs) exploit excess MYC to grow more rapidly, but are not able to initiate malignant transformation (Ziosi et al., 2010); we thus aimed at exploring MYCOVER's effect on the clonal behaviour of neoplastic TSGs (nTSGs). We first analysed the lethal giant larvae (lgl) mutation. Lgl protein regulates the apical-basal cell polarity in the epithelia (Grifoni et al., 2013); we previously demonstrated its functional conservation from Drosophila to humans (Grifoni et al., 2004), and we and others found the human orthologue HUGL-1 involved in cancers from different organs (Grifoni et al., 2004, 2007; Schimanski et al., 2005; Lu et al., 2009). In the Drosophila wing disc, lgl mutant cells are unable to grow in a wildtype background, especially in the regions where MYC levels are high, and are eliminated by MMCC (Froldi et al., 2010). In the same wild-type background, MYCOVER in lgl mutant

clones rescues them from death and transforms lgl−/<sup>−</sup> cells from losers into super-competitors (Froldi et al., 2010). But what happens to newly formed lgl, MYCOVER cells when they are surrounded by MYCOVER neighbours? As can be seen in **Figure 8**, while lgl−/<sup>−</sup> clones were smaller than wild-type twins in the A control compartment of the disc (**Figures 8A,A1,C**), in the 28% of the wing discs analysed the lgl−/<sup>−</sup> clones growing in the MYCOVER P compartment appeared significantly larger than the wild-type twins (**Figures 8A,A1,B**). As an example, the arrow in **Figure 8A1** points to a wild-type clone (double red) which appears much smaller than the lgl mutant twin (black). In addition, the arrowhead indicates an lgl mutant clone in the hinge region of the P compartment with no apparent wild-type twin clone. This suggests that the lgl mutant cells have a greater ability to exploit the excess MYC protein than the surrounding neighbours, hence the gain of a competitive advantage over the wild-type tissue. However, the average clone area occupied by the lgl−/<sup>−</sup> cells in this system was about 5000 px<sup>2</sup> , whereas it was found to be around 24000 px<sup>2</sup> in a previous study where lgl−/−, MYCOVER cells were induced in a wild-type background (Froldi

et al., 2010), demonstrating the MYCOVER neighbours exert a competitive pressure against the growth of the lgl mutant clones, which translates into a limited capability of lgl−/<sup>−</sup> cells to form large masses in a uniform MYCOVER field.

The remaining 72% of the wing discs analysed displayed a novel phenotype: the lgl mutant tissue grew as a multitude of spots scattered all across the MYCOVER P compartment. **Figure 9** shows two typical samples that we classified as "mild multifocal" (**Figure 9A**), which represented the 38% of the total samples (**Figure 9C**), and "severe multifocal" (**Figure 9B**), which represented the 34% of the total samples (**Figure 9C**). We classified multifocality as "mild" when the lgl mutant clones (black), though colonising a large fraction of the P compartment, did not alter its width (**Figure 9A1**), and "severe" when the lgl−/<sup>−</sup> cells filled the entire P compartment, which appeared dramatically enlarged (see how the P/A border moved from P to A comparing **Figures 9A1,B1**). This deep organ alteration suggests a locally invasive, malignant behaviour of these mutant cells that may be favoured by clone confluence during growth, as it is with other tumour models in Drosophila (Menendez et al., 2010; Ballesteros-Arias et al., 2014), with MYC protein levels that appeared to increase along with phenotype severity (compare **Figures 9A,B**). lgl mutant cells displayed preferential MYC accumulation, as can be appreciated in **Figure 10A2**, where arrowheads indicate some of the mutant cells (black, see **Figure 10A1**) accumulating MYC. Again, the organs displaying larger mutant spots (**Figure 10B**, squared area) showed an obvious increase in MYC protein levels (**Figure 10B2**). The most interesting aspect of this model is that it faithfully reproduced a distinctive feature of human pre-cancerous fields, i.e., multifocality (Dotto, 2014). The multifocal phenotype has never been associated with lgl mutations in Drosophila; therefore, it represents a novel trait acquired by a cell subject to a mutation in the lgl nTSG while being part of a MYCOVER field.

Magnification is 400× in A,B, 1000× in A1,A2, and 800× in B1,B2.

To verify that a MYCOVER field represented a bona fide precancerous area, and that multifocality did not result from a

MYC-overexpressing background. Immunostaining for aPKC (cyan) on wing discs from late w; Rab5<sup>2</sup> , FRT40A/Ubi-GFPnls, FRT40A; hh-Gal4/UAS-dm larvae. In A1,B1, arrows indicate mutant cells (black) displaying membrane redistribution of the apical marker aPKC, while the arrowheads point to a normal region of the disc border where the pseudostratified epithelium shows apical aPKC staining (asterisks). The dashed line in A1 marks the disc outer border. The basic genotype is indicated on the left of the figure panel. Disc axes are indicated in A,C. A1,B1 are magnified views of the squares drawn in A,B. Magnification is 400× in A,B,C,C1 and 1000× in A1,B1.

specific interaction between lgl and MYC, we induced a LOF mutation of a different nTSG in the MYCOVER field. Rab5 is an evolutionarily conserved core component of the vesicle trafficking machinery (Lu and Bilder, 2005), implicated in various aspects of human tumourigenesis (Torres and Stupack, 2011; Mendoza et al., 2014). Like lgl, entire fly organs mutated for Rab5 show neoplastic growth (Lu and Bilder, 2005; Vaccari and Bilder, 2009), and Rab5 mutant cells induced in a wild-type wing disc suffer from cell competition and are eliminated from the organ (Ballesteros-Arias et al., 2014). Using the same clonal system as above, we induced Rab5 LOF clones in animals whose P compartments overexpressed MYC. As can be seen in **Figure 11**, the multifocal phenotype was evident also for the Rab5−/<sup>−</sup>

Immunostaining for En (red) on wing discs from late w/yw, UAS-PI3KCAAX; Rab5<sup>2</sup> , FRT40A/Ubi-GFPnls; hh-Gal4/+ larvae. The arrows in C1 indicate the mutant clones (black). (D) Graph comparing the average clone area of Rab5−/<sup>−</sup> clones in the P (black triangles) and A (grey triangles) compartments. Basic genotypes are indicated on the left of the figure panel. The P compartment is outlined in A1,C1, and disc axes are indicated in A,C. Magnification is 400×.

cells (**Figures 11A,B** and respective magnifications A1 and B1). Also in this case, mutant cells showed MYC accumulation (**Figures 11A2,B2**, arrowheads in A2 indicate some mutant nests accumulating MYC). The 100% of the organs analysed showed a multifocal phenotype, subdivided in 71% mild and 29% severe. Moreover, Rab5 mutant cells showed loss of apical-basal cell polarity, a central feature of epithelial cancers (Wodarz and Nathke, 2007): in **Figure 12**, the magnifications in A1 and B1 show a region of the disc outer border where one can appreciate that the normal epithelium (arrowheads in A1 and B1) displays a wild-type localisation of the apical marker atypical PKC (aPKC, cyan, asterisks in B1). On the contrary, the mutant cells in the region indicated by the arrows in A1 and B1 (black in A1) show a redistribution of the polarity marker from the apical side to the entire cell cortex, together with aberrant, three-dimensional

growth. In **Figures 12C,C1**, the impairment of aPKC expression (cyan) is evident across the entire MYCOVER P compartment. This characteristic is consistent with Rab5 function: the endocytic trafficking is indeed essential in the maintenance of cell polarity, and mutations in genes involved in endocytosis provoke the expansion of cell's apical domain (Shivas et al., 2010).

Altogether, this evidence indicated that MYC overexpression in an epithelial tissue is sufficient to promote multifocal malignant lesions following single-cell mutations of different nTSGs.

To assess if multifocality may be considered a trait arising from specific properties conferred by the MYC field to the mutant cells, we repeated the same experiments as above in a PI3KCAAX−OVER territory. Using the same system as above, we first analysed lgl mutant behaviour. In **Figures 13A,A1**, we can observe lgl mutant clones (GFP−, indicated by the arrows) in the PI3KCAAX P compartment (marked in red by En staining in A). They are located outside the central region of the disc where, instead, we observed the presence of wild-type clones (GFP2+), indicating that mutant twins were eliminated by MMCC. Therefore, despite the over-expression of PI3KCAAX, lgl−/<sup>−</sup> clones continue to die in this area of the wing discs where MYC is normally highly expressed (see **Figure 2A2**). A statistical analysis of the clone area in the P and A compartments showed that lgl−/<sup>−</sup> mutant clones were significantly larger in P, with an average size of about 5000 px<sup>2</sup> , compared to A, where they displayed an average size of about 2000 px<sup>2</sup> (**Figure 13B**). The most important observation was, however, the total absence of multifocal growth. We then analysed the behaviour of the Rab5<sup>2</sup> mutation in a PI3KCAAX background. In **Figures 13C,C1**, a wing disc is shown where small mutant clones of comparable size are present in both compartments (black, arrows in **Figure 13C1**). In **Figure 13D**, the graph indicates that the mutant clones do not show significant differences in size between the P and A compartments. Finally, as it was for lgl, no multifocal growth was observed in all the Rab5 samples analysed.

These latter findings indicate that MYC confers on cells mutant for different nTSGs the ability to grow in multiple foci dispersed all across the modified territory. This seems to be a specific characteristic of MYC, as the growth inducer PI3K did not promote this peculiar phenotype. MYC upregulation emerges from our study as an excellent candidate to foster field cancerisation, by inducing a complex precancerisation molecular signature able to provide cells hit by non-competitive mutations with the ability to initiate carcinogenesis.

#### DISCUSSION

Field cancerisation is studied in the effort to understand if essential events recur in tumour initiation that may help develop early therapeutic interventions. It is now recognised that many types of cancers start from cells owing some, but not all, phenotypic traits necessary for malignancy, and those traits may result from various mutagenic insults, on the basis of which the most performant cells are selected for clonal expansion (Curtius et al., 2018). This process may be driven by cell competition, which is intensively studied both in Drosophila (Merino et al., 2016) and mammals (Di Gregorio et al., 2016). In this context, we focused our attention on MMCC, a process based on steep differences in MYC levels in confronting cells, which ultimately favour the expansion of high MYC-expressing cells at the expense of the less fit neighbours (Grifoni and Bellosta, 2015). Given the broad implication of MYC protein in human cancers (Gabay et al., 2014), its myriad functions inside the cell (Dang et al., 2006) and its regulation at both the transcriptional and post-transcriptional levels by a number of signalling pathways (Nussinov et al., 2016), it seems an excellent candidate to pioneer field cancerisation (Moreno, 2008).

To address this question, we first investigated the cellular responses to MYC overexpression (MYCOVER) in the imaginal wing disc, a Drosophila epithelial tissue widely used to model development, cell competition and cancer (Herranz et al., 2016). We found that MYCOVER was per se sufficient to activate a series of cellular behaviours consistent with the formation of a pre-neoplastic field, such as ROS production, genetic instability, changes in apoptotic and proliferation activity and alteration of epigenetic markers. Moreover, we showed that these cellular responses were not elicited by a MYC's generic pro-growth function, as an active form of the powerful growth inducer PI3K was not able to induce similar phenotypes, except a mild proproliferative effect. High MYC levels seem rather to prime field cancerisation by triggering a cascade of molecular changes that cooperate in taking cells a step closer to malignancy.

This bona fide pre-cancerous tissue was then tested for the ability to initiate tumourigenesis following mutations in neoplastic TSGs (nTSGs). We previously studied the effects of MYCOVER on three hyperplastic TSGs (hTSGs) owing to the Hippo signalling pathway: ds, ft and ex, and found that mutant clones grew more rapidly while killing the MYCOVER wildtype neighbours with higher efficiency, but they did not show any signs of malignancy (Ziosi et al., 2010). We and others demonstrated that most hTSGs upregulate MYC (Neto-Silva et al., 2010; Ziosi et al., 2010), hence their competitive capability, while some nTSGs downregulate MYC, hence their elimination from the tissue (Froldi et al., 2010). It is also recognised that the behaviour of both hTSGs and nTSGs depends on tissue's MYC levels (Froldi et al., 2010; Neto-Silva et al., 2010; Ziosi et al., 2010): in a uniform background, as with our model, mutant behaviour should rather be dictated by the intrinsic features of the given mutation.

In the Drosophila wing disc, wild-type cells hit by nTSGs mutations are usually irrelevant: they are indeed eliminated rapidly or contribute to the tissue without overgrowing (Froldi et al., 2010; Ballesteros-Arias et al., 2014). The same mutations induced in a MYCOVER field were rather capable to initiate multifocal, three-dimensional growth accompanied by loss of apical-basal cell polarity and aberrant tissue architecture. This was convincing evidence that MYC upregulation was sufficient as to establish a specific, complex pre-cancerisation signature, which predisposes the tissue to undergo malignant multifocal growth following certain second mutations.

Our findings lay the basis for future studies focused on early tumourigenesis. These studies are as essential as difficult: while understanding the very first phases of cancer is mandatory to conceive novel preventive and therapeutic interventions, investigations carried out in complex systems may lead to discouraging results. In this sense, the use of a genetically amenable animal model may greatly help dissect and dismantle the intricate networks implicated in cancer initiation.

#### AUTHOR CONTRIBUTIONS

fgene-09-00612 December 8, 2018 Time: 15:8 # 12

MS and DG conceived and designed the study. MS, CG, and SP performed the experiments. DG and DdB analysed the experimental data. SDG performed the statistical analysis. AP and DG wrote the manuscript. All authors contributed to manuscript revision and read and approved the submitted version.

#### REFERENCES


#### FUNDING

This study was supported by AIRC IG 17252 to AP and Research Fellowships from the University of Bologna to MS, SDG, SP, and DG.

#### ACKNOWLEDGMENTS

We acknowledge the Bloomington Drosophila Stock Center (NIH P40OD018537) for the fly stocks used in this study, and the Developmental Studies Hybridoma Bank, created by the NICHD of the NIH and maintained at The University of Iowa, for monoclonal antibodies. We thank Paola Bellosta, Bruce Hay, and Dennis Strand for reagents and Silvia Strocchi for manuscript revision.





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

Copyright © 2018 Sollazzo, Genchi, Paglia, Di Giacomo, Pession, de Biase and Grifoni. 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.

# p53-Sensitive Epileptic Behavior and Inflammation in Ft1 Hypomorphic Mice

Romina Burla1†, Mattia La Torre1†, Giorgia Zanetti <sup>1</sup> , Alex Bastianelli <sup>1</sup> , Chiara Merigliano1,2 , Simona Del Giudice<sup>1</sup> , Alessandro Vercelli 3,4, Ferdinando Di Cunto3,4, Marina Boido3,4 , Fiammetta Vernì <sup>1</sup> and Isabella Saggio1,2 \*

<sup>1</sup> Department of Biology and Biotechnology, Sapienza University of Rome, Rome, Italy, <sup>2</sup> Nanyang Technological University, Singapore, Singapore, <sup>3</sup> Neuroscience Institute Cavalieri Ottolenghi, Torino, Italy, <sup>4</sup> Department of Neuroscience, University of Torino, Piedmont, Italy

#### Edited by:

Maria Grazia Giansanti, Consiglio Nazionale Delle Ricerche (CNR), Italy

#### Reviewed by:

Maria Rosaria D'Apice, Consiglio Nazionale Delle Ricerche (CNR), Italy Cristina Capanni, Istituto di Genetica Molecolare (IGM), Italy

> \*Correspondence: Isabella Saggio isabella.saggio@uniroma1.it

†These authors share first authorship

#### Specialty section:

This article was submitted to Genetic Disorders, a section of the journal Frontiers in Genetics

Received: 29 August 2018 Accepted: 08 November 2018 Published: 27 November 2018

#### Citation:

Burla R, La Torre M, Zanetti G, Bastianelli A, Merigliano C, Del Giudice S, Vercelli A, Di Cunto F, Boido M, Vernì F and Saggio I (2018) p53-Sensitive Epileptic Behavior and Inflammation in Ft1 Hypomorphic Mice. Front. Genet. 9:581. doi: 10.3389/fgene.2018.00581 Epilepsy is a complex clinical condition characterized by repeated spontaneous seizures. Seizures have been linked to multiple drivers including DNA damage accumulation. Investigation of epilepsy physiopathology in humans imposes ethical and practical limitations, for this reason model systems are mostly preferred. Among animal models, mouse mutants are particularly valuable since they allow conjoint behavioral, organismal, and genetic analyses. Along with this, since aging has been associated with higher frequency of seizures, prematurely aging mice, simulating human progeroid diseases, offer a further useful modeling element as they recapitulate aging over a short timewindow. Here we report on a mouse mutant with progeroid traits that displays repeated spontaneous seizures. Mutant mice were produced by reducing the expression of the gene Ft1 (AKTIP in humans). In vitro, AKTIP/Ft1 depletion causes telomere aberrations, DNA damage, and cell senescence. AKTIP/Ft1 interacts with lamins, which control nuclear architecture and DNA function. Premature aging defects of Ft1 mutant mice include skeletal alterations and lipodystrophy. The epileptic behavior of Ft1 mutant animals was age and sex linked. Seizures were observed in 18 mutant mice (23.6% of aged ≥ 21 weeks), at an average frequency of 2.33 events/mouse. Time distribution of seizures indicated non-random enrichment of seizures over the follow-up period, with 75% of seizures happening in consecutive weeks. The analysis of epileptic brains did not reveal overt brain morphological alterations or severe neurodegeneration, however, Ft1 reduction induced expression of the inflammatory markers IL-6 and TGF-β. Importantly, Ft1 mutant mice with concomitant genetic reduction of the guardian of the genome, p53, showed no seizures or inflammatory marker activation, implicating the DNA damage response into these phenotypes. This work adds insights into the connection among DNA damage, brain function, and aging. In addition, it further underscores the importance of model organisms for studying specific phenotypes, along with permitting the analysis of genetic interactions at the organismal level.

Keywords: aging, epilepsy, DNA damage, p53, DNA repair

# INTRODUCTION

Epilepsy comprises a family of disorders characterized by enduring predisposition to generate spontaneous seizures (Scharfman, 2007; Fisher et al., 2014). Seizures are underpinned by multiple mechanisms and their clinical outcome varies widely (Scharfman, 2007). Regardless of their outcome, seizures arise from disruption of mechanisms that create a balance between neuronal excitation and inhibition. Factors corrupting this balance result from alterations at different levels of brain function, from ion channels to receptors, and neuronal circuits (Stafstrom and Carmant, 2015). Epilepsies are often associated with morphological brain abnormalities (Bertram, 2013), but at least one-third have non-structural etiologies (Guerrini et al., 2014). In the last years, concomitantly with human population demographic changes, high incidence of epileptic disorders has been associated with aging, whose specific pathophysiology is under investigation (Leppik and Birnbaum, 2010).

Understanding the mechanistic path to disease is complex in humans due to ethical issues, unavailability of controls and high costs of human research. As a result, studies mostly rely on the use of models, including human 3D cultures and stem cell based systems (Riminucci et al., 2006; Simão et al., 2015), or, for organismal analyses, genetically engineered mice (Baraban, 2007; Saggio et al., 2014; Remoli et al., 2015; La Torre et al., 2018). One of the way in which epilepsy has been modeled in mice is via the inactivation of genes implicated in ion channels (Yu et al., 2006; Baraban, 2007; Glasscock et al., 2012), or of genes encoding for neurotransmitter receptors (Fonck et al., 2005). In addition, epileptic phenomena have been observed in mouse models of Alzheimer disease (Vogt et al., 2011; Ziyatdinova et al., 2011). However, no other genetic models of age related epilepsy have been yet described. Prematurely aging mouse mutants, which recapitulate aging traits over a short time-window (Blasco, 2005; Stewart et al., 2007), offer a specific advantage to model diseases caused or exacerbated by aging, including brain pathologies.

An emerging cause for brain disease and for the aging brain is DNA damage (Langie et al., 2017). DNA integrity poses a challenge for the nervous system as its development depends on a complex series of dynamic and adaptive processes associated to DNA damage (Mckinnon, 2013). Unrepaired DNA lesions have detrimental effects on the developing of a functional nervous system and neural progenitor cells rely on DNA repair systems during the developmental program. After completion of neurogenesis, DNA repair is still of paramount importance to safeguard the genome (Madabhushi et al., 2014), especially to protect the neurons against reactive oxygen species (Langie et al., 2017). DNA damage is also direct cause for cell senescence and for a related inflammatory response (Campisi, 2013; López-Otín et al., 2013).

Mouse mutants of DNA damage functions have opened the path to establish a link between DNA damage and the seizure phenotype (Shen et al., 2010; Bianchi et al., 2017). For example, the inactivation of XRCC1, a central factor in the DNA single strand break repair pathway, leads to profound neuropathology involving behavioral phenotypes consistent with epilepsy (Lee et al., 2009). Data based on studies in Drosophila suggest that nuclear architecture and lamins could play a role into epilepsy (Frost, 2016). However, the hierarchy and range of events connecting nuclear architecture, molecular DNA function to epileptic behavior is still to be dissected, along with the elements exacerbating this pathology in aging.

DNA damage affects genome near to randomly, but some chromosomal regions, such as telomeres, are more prone to DNA instability. In mammals, telomeric DNA is composed of doublestranded short tandem repeats of TTAGGG sequence forming higher-order DNA structures binding a specialized protein complex, known as shelterin (de Lange, 2005). We identified a telomere-associated protein named AKTIP in humans (and Ft1 in mouse), which interacts with the shelterin members TRF1 and TRF2 (Burla et al., 2015). AKTIP/Ft1 reduction causes telomere aberrations, DNA instability and cell senescence (Burla et al., 2015). In vivo, the genetic reduction of Ft1 causes premature aging defects including skeletal alterations and lipodystrophy (La Torre et al., 2018). AKTIP/Ft1 interacts with lamins (Burla et al., 2016a,b), pivotal elements for the control of nuclear architecture and DNA function, including DNA repair, replication and transcription (Dittmer and Misteli, 2011). Importantly, Ft1 mutant mice share similarities with lamin mutant animals, which are models of choice for human progeroid syndromes, linking the Ft1 model to premature aging and progeroid diseases (Burla et al., 2016a,b, 2018).

Here we report that Ft1 mutant mice are subjected to repeated seizures. We show that this trait is not linked to overt brain morphological alterations, but is age and inflammation linked. We also demonstrate that this phenotype is sensitive to the expression of the guardian of the genome p53, pointing to a role of DNA function in the seizure phenotype.

# RESULTS

# Seizures in Ft1 Mutant Mice

Mice with reduced levels of Ft1 were generated using the knock out first (kof) strategy based on the insertion into the target gene (referred as kof allele) of the βgeok cassette (Testa et al., 2004), which traps and truncates Ft1 nascent transcript reducing the expression of the targeted gene (La Torre et al., 2018). Previous observations of Ft1kof /kof mice revealed that mutant animals display significant reduction in body weight and lifespan, compared to controls. Twenty-one percent of the animals show a severe body size reduction and early post-natal death (we refer to these mice as severely affected Ft1kof /kof mice, abbreviated with SA Ft1kof /kof or SA mutant mice). The leftovers, non SA Ft1kof /kof mice have a mild phenotype, with a median survival of 113 weeks, allowing adult phenotype observation (La Torre et al., 2018).

To investigate mutant mouse behavior, we worked on a cohort of animals aged ≥ 21 weeks non SA Ft1kof /kof . We recorded spontaneous behavioral abnormalities in non SA Ft1kof /kof mice including episodes of motor tremors and convulsions, fast runs, jumps, and excessive salivation (**Figure 1** and **Supplementary Movies 1**–**3**). Dissection of video recordings indicated that non SA Ft1kof /kof displayed sudden movements followed by facial twitching, violent hind limbs shaking and falling, back arching, short jerks in muscles of the hind limbs and forelimbs extension, Straub tail and incontinence, followed by short jerks and fast breathing (**Figure 1A** and **Supplementary Movies 1**–**3**). Eventually, animals stood up and returned to normal activity; after few minutes switched to post-ictal phase characterized by short periods of complete immobility, interrupted by short intervals of movement (**Supplementary Movies 4**–**6**). Recording showed moving rhythmical up-down or left-right, stroking mouth with forepaws, in a repetitive motion, and appearing to be chewing and grooming (**Figure 1A** and **Supplementary Movies 4**–**6**). Temporal evaluation of the data indicated that non SA Ft1kof /kof mice display a behavioral repertoire corresponding in quality and duration to epileptic phenotypes (**Figure 1B**), as described for other mouse models (Chabrol et al., 2010; Robie et al., 2017).

Over a total of 260 animals (including Ft1+/+, Ft1+/kof , and Ft1kof /kof), 18 Ft1kof /kof exhibited at least one seizure manifestation (**Figures 2A,B**). No seizure episodes were observed in heterozygote Ft1+/kof or wt mice (**Figure 2A**). The frequency of seizures reached a maximum of one seizure episode per week. Of 18 seizure positive mutants, 7 exhibited one seizure-like manifestation during their follow-up period, while the leftover displayed more than one seizure. On average, we observed 2.33 ± 0.14 seizures during the entire follow-up period in the seizure positive mutants (**Figure 2B**). Time distribution of seizures for each mutant animal experiencing more than one seizure indicated non-random enrichment of seizures over the follow-up period, with 75% of seizures happening in consecutive weeks (**Figure 2C**).

## Seizures of Ft1 Mutant Mice Are Age and Gender Linked

To analyze the distribution of seizures during aging, we subdivided mouse lifespan in three major intervals: young (3 ≤ weeks ≤ 20), juvenile (21 ≤ weeks ≤ 60), and adult (61 ≤ weeks ≤ 100). We monitored, for each age interval, the fraction of mice exhibiting seizures for the first time (**Figure 3A**). None of young non SA Ft1kof /kof mice (n = 189) displayed seizures, while seizures were observed in juvenile and adult mice, in 16 out of 68 and 2 out of 8, respectively.

We previously reported that non SA Ft1kof /kof mice display growth defects, which start within the juvenile interval and become prominent through aging (La Torre et al., 2018). We then asked whether the seizure phenotype was paralleled by age-associated growth impairment. We considered the difference between the average body weight of wt (n = 6) and sex-matched seizure positive mutant mice (n = 8) (1body weight), and plotted it against the number of seizures observed in each mutant animal

picture). (B) Progression of behavior of two Ft1kof/kof mice, during and post seizure events, each represented by a horizontal bar. The length of each bar indicates the duration of the relative motor type. In turquoise seizure analysis of mouse ID #13489, and in purple the same analysis of mouse ID #16247.

to generate a regression curve (**Figure 3B**). The two variables were linked by positive correlation (r <sup>2</sup> = 0.60; ∗∗p < 0.01 in Pearson's R test).

Non SA Ft1kof /kof male mice were previously shown to display a stronger phenotype as compared to Ft1kof /kof females (La Torre et al., 2018). We thus decided to explore gender differences in the seizure trait. We analyzed a cohort of non SA Ft1kof /kof animals aged ≥ 21 weeks including 37 males and 39 females. Seizure positive mutants were 37.8% among males and 10.3% among non SA Ft1kof /kof female mice (**Figure 3C**).

In order to verify whether aging would induce variations in Ft1 expression we monitored two cohorts of Ft1+/<sup>+</sup> mice by QPCR. Results indicate that there are no significant variation in Ft1 expression with aging (**Supplementary Figure 1A**). In addition, the human counterpart of Ft1, AKTIP, is also expressed in brain cells (**Supplementary Figure 1B**).

Altogether these results indicate that Ft1 reduction causes seizures, and that this phenotype parallels overall organismal alterations characteristic of non SA Ft1kof /kof mice including age dependent growth impairment and sex-linked defects (La Torre et al., 2018).

#### Structure of Ft1 Mutant Brain

To define whether Ft1 function impacted on brain structure we firstly measured skull and cranial length of non SA Ft1kof /kof mice by X-ray analysis (**Figure 4**). Mutant and age- and sex-matched wt animals were monitored at 2, 4, 8, and 13 months. Length measures revealed a continuous increase during postnatal development in mutant mice as in controls, with a mild alteration induced by Ft1 reduction at the age of 8 weeks (∗p < 0.05 in Student's t-test) (**Figures 4A–D**). We then investigated brain morphology by histological analysis. We firstly evaluated the hippocampus, as hippocampal defects are often cause of epilepsy. Nissl-staining of coronal sections of Ft1kof /kof brains did not highlight overt alterations of hippocampal organization or gross lesions, such as cell loss, structural deformation or scars (**Figures 5A–B**). By immunofluorescence, we further analyzed hippocampus cytoarchitecture. Semi-quantitative analysis

of Parvalbumin-positive GABAergic interneurons showed a mild reduction in Ft1kof /kof mice as compared to controls (**Figure 5C** and **Supplementary Figure 2**). Then we evaluated the somatosensory cortex cytoarchitecture: by Nissl-staining and MAP2 immunofluorescence, we did not detect evident alterations, including cortical thinning and layering defects in Ft1kof /kof brain compared to wt (**Figures 6A–C**). The density of Parvalbumin-positive GABAergic interneurons in the cortex was non-significantly affected by Ft1 reduction (**Figure 6D** and **Supplementary Figure 3**). Moreover, we did not observe any evident macroscopic alterations in the whole brain structure, as for example shrinkage of specific cerebral regions, ventricle enlargement, corpus callosum thinning (data not shown), or in brain volume (wt 177.95 mm<sup>3</sup> , Ft1kof /kof 189.46 mm<sup>3</sup> ; **Supplementary Figure 4**). Overall, these analyses did not reveal robust differences between non SA Ft1kof /kof and control mice, thus in principle excluding the presence of severe brain degenerative processes.

### p53-Sensitive Inflammatory Response in Ft1 Mutant Brain

To investigate on molecular drivers for the seizure phenotype observed in Ft1kof /kof mice we reasoned on studies in aging and progeroid models which link DNA damage to systemic inflammation and neurodegeneration (Campisi, 2013; López-Otín et al., 2013). Given the progeroid traits observed in non SA Ft1kof /kof mice and the implication of Ft1 in telomere metabolism and DNA function we decided to monitor the level of two canonical, interrelated, inflammatory cytokines: IL-6 and TGF-β. To get a full picture also on the connection with DNA damage, we explored this aspect in non SA Ft1kof /kof mice, expressing normal levels of the guardian of the genome p53 and in mice defective for p53 expression due to heterozygous p53 inactivation (Ft1kof /kof ; p53+/ko). QPCR analysis showed significant higher levels of both IL-6 and TGF-β in Ft1kof /kof brains as compared to age matched control mice (**Figures 7A,B**; ∗∗p < 0.01). Indeed, in Ft1kof /kof ; p53+/ko mutant mice IL-6 and TGF-β activation was reversed. These results suggest that the activation of the DNA damage response pivotal player, p53, is a crucial event in the organismal response to Ft1 reduction (**Figures 7A,B**). As expected, p53 reduction did not impact on Ft1 expression, which remained significantly reduced in Ft1kof /kof ; p53+/ko as in Ft1kof /kof animals (**Figure 7C**).

Given the activation of the inflammatory cytokines and its rescue by p53 reduction, we decided to explore the link between p53, as an element directly associated with DNA function, to the seizure phenotype of Ft1kof /kof mice. To this purpose, we analyzed three cohorts of ≥ 21 weeks mice including Ft1+/+;p53+/ko , Ft1+/kof ;p53+/ko and Ft1kof /kof ;p53+/ko mice, as

compared to genetically and age matched controls. None of the mice with the p53+/ko mutation displayed seizures, suggesting on one side that p53 reduction alone does not cause seizures, and, on the other side, that it rescues the seizure phenotype of our Ft1 mutant mice (**Figure 7D**).

Altogether these results establish a link among Ft1, p53, inflammatory parameters and seizure behavior.

# DISCUSSION

Epilepsy is a complex disorder codified by the concept of predisposition to generate spontaneous seizures. Studies to interpret seizure etiology are complex in humans, indeed many open questions remain about its causes and mechanisms. An important tool to understand and interpret epileptic phenomena comes from model systems. Mutant mice in particular have been instrumental for the identification of genes and gene-determined molecular processes generating brain dysfunction and epileptic phenotypes (Fonck et al., 2005; Yu et al., 2006; Baraban, 2007; Löscher, 2011; Glasscock et al., 2012). In this work we report on the epileptic behavior of a mouse mutant obtained by genetically reducing the expression of the telomeric gene Ft1 (AKTIP in humans) and present data supportive of a link between DNA damage and epileptic behavior.

As a consequence to its biological function, AKTIP/Ft1 mutant cells display DNA replication defects, DNA and telomere damage, along with cell senescence (Burla et al., 2015, 2016a). AKTIP/Ft1 is linked to nuclear architecture through the biochemical interaction occurring with A- and B-type lamins, the main components of the nucleoskeleton, which, in turn, is a pivotal element for the control of DNA function (Dittmer and Misteli, 2011; Burla et al., 2018). Mice with reduced Ft1 expression display a segmental phenotype including reduced subcutaneous fat, growth and skeletal defects. These traits partly recapitulate the premature aging phenotypes observed in progeroid mice generated by mutations of lamins or DNA maintenance genes (Blasco, 2005; Stewart et al., 2007; Burla et al., 2018), and are exacerbated in adult individuals as compared to juvenile. Given the connection of AKTIP/Ft1 with both lamins and DNA function, Ft1 mutant mice represent an attractive model for investigating how these connections may impact on the organismal phenotype and on different tissues and organs.

We report here that Ft1 mutant mice exhibit spontaneous seizures. Digital movies document the seizure manifestation highlighting typical traits, as loss of postural equilibrium, limb jerks, and convulsions, as in other epileptic mouse models (Yang et al., 2007; Minkeviciene et al., 2009; Chabrol et al., 2010; Glasscock et al., 2012; Simeone et al., 2018). Seizures were observed after the 20th week of age suggesting age-dependent degenerative changes. Seizures also correlated with growth impairment, which, in Ft1kof /kof mice, is exacerbated with aging (La Torre et al., 2018). The frequency of seizures was higher

(upper panels), and higher magnification of the anterior hippocampus (bottom panels), at about Bregma −1.22 mm. (B) Nissl-stained hippocampus of Ft1+/<sup>+</sup> and Ft1kof/kof mice, at about Bregma <sup>−</sup>2.18 mm. (C) Confocal images showing parvalbumin-positive interneurons (PARV, in red) at the hippocampal level; cell nuclei are labeled by DAPI staining (in blue). Scale bars: (A) 500µm (upper panel), (A) 200µm (bottom panel), (B) 400µm, (C) (merge) 160µm, (C) (DAPI/PARV) 200µm.

in male mice, in line with the overall impact of Ft1 reduction observed in our model (La Torre et al., 2018).

Along with behavioral descriptions, the importance of mouse models resides in the fact that they allow investigating upstream events to behavioral phenotypes, which, in humans, is complex to explore. We thus exploited these animals to analyze the pathophysiological path to seizures in an attempt to contribute to establish experimentally proven links between DNA function and brain alterations.

The analysis of brain morphology and cytoarchitecture in non SA epileptic Ft1kof /kof mice did not highlight overt defects nor macroscopic neurodegeneration. These results were obtained by analyzing restricted brain areas and cell subtypes, and we do not exclude the possibility of other or more subtle brain alterations escaping our analysis. However, given the data, we hypothesized that molecular, rather than overt structural aberrations would underpin the epileptic behavior. This interpretation would be in line with the fact that many types of epilepsy have been associated with molecular alterations, rather than with macroscopic brain structure defects (Guerrini et al., 2014).

On the basis of this hypothesis, we investigated on putative molecular culprits for the seizure phenotype. Telomeres and cell senescence have been implicated in disease and aging. In particular, the secretion of a panel of inflammatory signals (Senescence Associated Secretory Phenotype, SASP) is considered a pivotal element in the alteration of tissue homeostasis contributing to pathological status (Campisi et al., 2001; Campisi, 2013). Given the function of AKTIP/Ft1 in telomere protection along with its importance in preventing cell senescence and DNA damage (Burla et al., 2015; La Torre et al., 2018), we hypothesized that in Ft1 mutant mice accumulation of damage and consequent senescence would provoke inflammatory cytokine production, which, in turn, could be implicated into the epileptic phenotype.

We thus measured inflammatory factors in the brain along with exploring the implication of DNA function in inflammation. In the brain of Ft1 mutant mice we detected the activation of IL-6 and TGF-β. Both these factors are related with inflammation (Coppe et al., 2010). IL-6 activation is assigned to the SASP group of factors linked to cell senescence and has been defined as a senescence biomarker in mice (Coppe et al., 2010; Hudgins et al., 2018). The cytokine TGF-β, in association with IL-6, is a further inflammatory stimulus (Sanjabi et al., 2009). Altogether these data suggested a connection between seizures and inflammation in the brain of Ft1kof /kof mice.

To further investigate on the causative cascade of pathologic events occurring in mice from Ft1 mutation to seizures, we generated Ft1 mutant mice with a heterozygous ko mutation in p53. p53 is a pivotal element in the DNA damage response and we have demonstrated that p53 is activated in AKTIP reduced primary human cells, which contributes to blocking cell proliferation and inducing senescence (Burla et al., 2015; La Torre et al., 2018). Interestingly p53 loss was shown to rescue aging traits in vivo, by releasing DNA damageinduced checkpoints and cell senescence (Varela et al., 2005). In addition, p53 regulated senescence has been defined as a pivotal element in generating the SASP phenotype. Consistently, reduction of p53 expression rescues SASP related aging traits (Baar et al., 2017). In Ft1 mutant mice, p53 reduction not only reversed back the expression of IL-6 and TGFβ, but also rescued the seizure phenotype. These results, taken together, induce to speculate that the path to seizures generated by Ft1 reduction could start with DNA damage, including telomere dysfunction, followed by p53 activation, cell senescence, and SASP. The latter could eventually induce brain seizures, through a mechanism that remains to be elucidated.

This interpretation of the data is interesting for investigating on seizure causative events in basal physiological conditions, but also in aging. In fact, aging is characterized by the exacerbation of the alteration of the biological pathways which we have taken into consideration, including senescence, genomic instability and telomere fragility (López-Otín et al., 2013). These factors act on aging in a cell intrinsic way and through extrinsic mechanisms

(Lopez-Otin et al., 2016), as secretion of SASP factors (Coppe et al., 2010) and chronic inflammation, profoundly altering tissue microenvironment (Fulop et al., 2017), which, we would suggest, could impact also on brain function.

In conclusion, this work adds new insights into the connection among DNA damage, brain function, and inflammation. In addition, it further underscores the importance of model organisms for studying molecular path to specific phenotypes, along with permitting the study of genetic interactions at the organismal level. These aspects are even more relevant in aging and brain research studies for which work in humans is inaccessible due to time, ethical and sample accessibility issues.

# MATERIALS AND METHODS

#### Mice and Ethical Statements

Mice were maintained and bred in a 12 h light/12 h dark cycle, in a pathogen free unit of the animal house at Biology and Biotechnology Department, Sapienza University. Animals were housed and treated in accordance with protocol 355/2017-PR approved by the Italian Ministry of Health. Animals carrying the knockout first mutations in the Ft1 gene (Ft1 kof) were obtained as previously described (La Torre et al., 2018). In order to obtain Ft1 and p53 mutant animals Ft1+/kof were crossed with p53+/ko animals [kindly provided by G. Piaggio and S. Soddu IFO, Italy;

(Jacks et al., 1994)]; subsequently doubly heterozygous mice were intercrossed to obtain the desired genotypes. Offspring were weaned at 3 weeks and tail biopsies were used for genotyping. When needed, mice were anesthetized by intramuscular Zoletil 20 (Virbac S.A., France), or euthanized by asphyxiation with carbon dioxide or cervical dislocation.

# Genotyping

Mice were genotyped as previously described (La Torre et al., 2018). Briefly genomic DNA was extracted from tail biopsies digested 50◦C with a proteinase K/SDS solution using the blackPREP Rodent Tail DNA Kit (Analytik Jena, Jena, Germany) following manufacturer's instructions.

Mice were PCR genotyped using the following primers:

```
Ft1 E4 F: 5′
        -GTGAAGCAGAAGCTGCCAGGAGT−3
                                           ′
                                           ;
Ft1 E6 R: 5′
         -AGCTCACCCGAGGTGGGATCAA−3
                                         ′
                                          ;
p53-X6 F: 5′
         -AGCGTGGTGGTACCTTATGAGC−3
                                         ′
                                          ;
p53-Neo19 F: 5′
            -GCTATCAGGACATAGCGTTGGC−3
                                            ′
                                             ;
p53-X7 R: 5′
         -GGATGGTGGTATACTCAGAGCC−3
                                         ′
```
#### Seizure Observation

Seizures were observed during routine mouse handling. Where indicated seizure and post-seizure events were video recorded. The movies were then analyzed for specific behavioral signs as previously described (Chabrol et al., 2010).

# X-ray and Cephalometric Measurements

X-ray images were taken using Faxitron MX-20 (Faxitron X-ray Corp.) at 24 kV for 6 s; images captured with Medical Imaging Film HM Plus (Ferrania). Skull length and cranial length were measured by Image J software as previously described (Rueden et al., 2017).

# Histological Analysis

Eight month-old mice were euthanized and brains were removed and postfixed in 4% PFA for 2 h at 4◦C. Samples were then immersed in a solution containing 30% sucrose in phosphate buffer 0.1 M at 4◦C for cryoprotection, embedded in cryostat medium (Killik; Bio-Optica, Milan, Italy), and cut on the cryostat (HM 550; Microm) in serial transverse 50µm thick sections. Some brain sections (one section every 600µm) were Nissl-stained to evaluate the gross cerebral morphology: briefly sections were mounted on 2% gelatin-coated Superfrost slides and air-dried overnight; slides were hydrated in distilled water for 1 min before staining in 0.1% Cresyl violet acetate (Sigma Aldrich, St. Louis, MO) for 10 min, dehydrated in an ascending series of ethanol, cleared in xylene and cover-slipped with Eukitt (Bioptica, Milan, Italy). Sections were examined at a Nikon Eclipse 80i microscope Equipped with a Nikon DS-Fi1 camera. Brain volume of wt and Ft1kof /kof mice was calculated considering Bregma 2.10 mm to Bregma −2.54 mm segments

in Nissl-stained serial sections reconstructed by Neurolucida software (MicroBrightField, Williston, VT, USA) and the volume (expressed in mm<sup>3</sup> ) was obtained by NeuroExplorer software (MicroBrightField). For immunofluorescence, brain sections were permeabilized with in PBS 0.3% Triton X-100 at RT on a tilting plate for 20 min; then, to block unspecific binding of the antibody, sections were incubated for 30 min at RT with 0.3% Triton X-100 and 10% normal donkey serum or normal goat serum (Sigma-Aldrich) in PBS (pH 7.4). Sections were then incubated at 4◦C overnight with the following antibodies: 1:7,500 anti-Parvalbumin (rabbit; Swant); 1:200 anti-MAP2 (mouse; Chemicon). Then, sections were incubated with appropriate fluorochrome-conjugated secondary antibodies, for 1h at RT: 1:400 Alexa Fluor 647, goat anti mouse (Jackson ImmunoResearch Laboratories); 1:200 cyanine 3-conjugated secondary antibody, donkey anti-rabbit (Jackson ImmunoResearch Laboratories); 1:200 cyanine 2 conjugated secondary antibody, donkey anti-mouse (Jackson ImmunoResearch Laboratories). Finally sections were incubated with 4,6 Diamino-2 phenyindole Dilactate (DAPI; Sigma Aldrich) in PBS 1:50 for 3 min. Samples were washed and coverslips were mounted with the anti-fade mounting medium Mowiol. For double staining and 3D reconstructions, slides were examined with a Leica TCS SP5 confocal laser scanning microscope.

#### Cell Culture

Human foreskin primary fibroblasts (HPFs), HeLa (ATCC CCL-2) and 293T (ATCC CRL-11268) cells were cultured in DMEM with 10% FBS. SH-SY5Y (ATCC CRL-2266) cells were cultured in EMEM supplemented with 10% FBS.

#### RNA Extraction and QPCR Analysis

Cells were lysed using the TRIzol reagent (Invitrogen). Brains were removed from euthanized mice and frozen in liquid nitrogen. RNA was extracted using the TRIzol reagent (Invitrogen) according to manufacturer. After DNaseI treatment (Invitrogen) RNA from cells and brains was reverse transcribed into cDNA with oligo d(T) primer and OMNISCRIPT RT KIT (Qiagen). QPCRs were performed as described (Piersanti et al., 2015) using the following primers:

Ft1 E3 F: AACCAGTCCTCCACGAAGTGCA; Ft1 E3 R: TAGGGCTTCGCTATGGGTAGAGCA; Ft1 E6 F: CCGTCTTTCACCCACTAGTTGAT; Ft1 E6 R: TTGCGAACGCTCTTTTCACA; mGAPDH F: GTGGCAAAGTGGAGATTGTTGCC; mGAPDH R: TGTGCCGTTGAATTTGCCGT; IL-6 F: CTCTGGGAAATCGTGGAAAT; IL-6 R: CCAGTTTGGTAGCATCCATC

# REFERENCES

Baar, M. P., Brandt, R. M. C., Putavet, D. A., Klein, J. D. D., Derks, K. W. J., Bourgeois, B. R. M., et al. (2017). Targeted apoptosis of senescent cells restores tissue homeostasis in response to chemotoxicity and aging. Cell 169, 132–147 e116. doi: 10.1016/j.cell.2017.02.031

```
TGF-β F: CCCTATATTTGGAGCCTGGA;
TGF-β R: CTTGCGACCCACGTAGTAGA;
AKTIP F: TCCACGCTTGGTGTTCGAT;
AKTIP R: TCACCTGAGGTGGGATCAACT
GAPDH F: TGGGCTACACTGAGCACCAG
GAPDH R: GGGTGTCGCTGTTGAAGTCA
```
#### Statistics

χ 2 test was applied for comparisons of the mouse cohorts. Independent data sets were analyzed with the Student's t-test (unpaired, two-tailed). Correlation analyses were performed via Pearson's R test.

### ETHICS STATEMENT

This study was carried out in accordance with the recommendations of the European Directive 2010/63/EU on the protection of animals used for scientific purposes. The protocol was approved by the Italian Ministry of Health (Ministero della Salute) with protocol number 355/2017-PR.

### AUTHOR CONTRIBUTIONS

RB, MLT, FDC, AV, and FV contributed to the design of the experiments and to the writing of the manuscript. MLT, GZ, AB, SDG, MB, and CM performed the experiments and analyzed the data. IS designed the experiments, analyzed the data and wrote the manuscript.

#### FUNDING

This work has been supported by Sapienza Ateneo grants to FV, IS, MLT, and RB; by grants EU FP7 Brainvectors (n. 286071), Telethon GEP15033 and PRF 2016-67 to IS. This work is in the memory of P. Bianco.

# ACKNOWLEDGMENTS

This work has been supported by Sapienza Ateneo grants to FV, IS, ML, and RB; by grants EU FP7 Brainvectors (n. 286071), Telethon GEP15033 and PRF 2016-67 to IS. This work is in the memory of P. Bianco.

#### SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fgene. 2018.00581/full#supplementary-material


a mouse model of severe myoclonic epilepsy in infancy. Nat. Neurosci. 9, 1142–1149. doi: 10.1038/nn1754

Ziyatdinova, S., Gurevicius, K., Kutchiashvili, N., Bolkvadze, T., Nissinen, J., Tanila, H., et al. (2011). Spontaneous epileptiform discharges in a mouse model of Alzheimer's disease are suppressed by antiepileptic drugs that block sodium channels. Epilepsy Res. 94, 75–85. doi: 10.1016/j.eplepsyres.2011.01.003

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

Copyright © 2018 Burla, La Torre, Zanetti, Bastianelli, Merigliano, Del Giudice, Vercelli, Di Cunto, Boido, Vernì and Saggio. 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.

# Modeling Congenital Disorders of N-Linked Glycoprotein Glycosylation in Drosophila melanogaster

Anna Frappaolo<sup>1</sup>† , Stefano Sechi<sup>1</sup>† , Tadahiro Kumagai<sup>2</sup>† , Angela Karimpour-Ghahnavieh<sup>1</sup> , Michael Tiemeyer2,3 \* and Maria Grazia Giansanti<sup>1</sup> \*

1 Istituto di Biologia e Patologia Molecolari del CNR, Dipartimento di Biologia e Biotecnologie, Sapienza - Università di Roma, Rome, Italy, <sup>2</sup> Complex Carbohydrate Research Center, University of Georgia, Athens, GA, United States, <sup>3</sup> Department of Biochemistry and Molecular Biology, University of Georgia, Athens, GA, United States

#### Edited by:

Roberta Fraschini, Università degli Studi Milano-Bicocca, Italy

#### Reviewed by:

Vladimir Lupashin, University of Arkansas for Medical Sciences, United States Giuliano Callaini, Università degli Studi di Siena, Italy

#### \*Correspondence:

Michael Tiemeyer mtiemeyer@ccrc.uga.edu Maria Grazia Giansanti mariagrazia.giansanti@uniroma1.it †These authors have contributed

equally to this work

#### Specialty section:

This article was submitted to Genetic Disorders, a section of the journal Frontiers in Genetics

Received: 10 July 2018 Accepted: 14 September 2018 Published: 02 October 2018

#### Citation:

Frappaolo A, Sechi S, Kumagai T, Karimpour-Ghahnavieh A, Tiemeyer M and Giansanti MG (2018) Modeling Congenital Disorders of N-Linked Glycoprotein Glycosylation in Drosophila melanogaster. Front. Genet. 9:436. doi: 10.3389/fgene.2018.00436 Protein glycosylation, the enzymatic addition of N-linked or O-linked glycans to proteins, serves crucial functions in animal cells and requires the action of glycosyltransferases, glycosidases and nucleotide-sugar transporters, localized in the endoplasmic reticulum and Golgi apparatus. Congenital Disorders of Glycosylation (CDGs) comprise a family of multisystemic diseases caused by mutations in genes encoding proteins involved in glycosylation pathways. CDGs are classified into two large groups. Type I CDGs affect the synthesis of the dolichol-linked Glc3Man9GlcNac<sup>2</sup> precursor of N-linked glycosylation or its transfer to acceptor proteins. Type II CDG (CDG-II) diseases impair either the trimming of the N-linked oligosaccharide, the addition of terminal glycans or the biosynthesis of O-linked oligosaccharides, which occur in the Golgi apparatus. So far, over 100 distinct forms of CDGs are known, with the majority of them characterized by neurological defects including mental retardation, seizures and hypotonia. Yet, it is unclear how defective glycosylation causes the pathology of CDGs. This issue can be only addressed by developing animal models of specific CDGs. Drosophila melanogaster is emerging as a highly suitable organism for analyzing glycan-dependent functions in the central nervous system (CNS) and the involvement of N-glycosylation in neuropathologies. In this review we illustrate recent work that highlights the genetic and neurobiologic advantages offered by D. melanogaster for dissecting glycosylation pathways and modeling CDG pathophysiology.

Keywords: Drosophila, glycosylation, congenital disorders, Golgi, model organism

# INTRODUCTION

Protein glycosylation is one of the most frequent post-translational modifications in eukaryotes; approximately one fifth of all proteins in protein structural databases are glycosylated (Ohtsubo and Marth, 2006; Freeze and Ng, 2011; Khoury et al., 2011). The oligosaccharide moieties added to glycoproteins impact their structure and biological function by contributing to protein folding, stability, and transport to appropriate sub-cellular locations. Glycans also mediate cell–cell interactions, modulate signal transduction, and regulate molecular trafficking and endocytosis. The two main types of protein glycosylation are N-linked and O-linked glycosylation. The biosynthesis and elaboration of glycoprotein N-linked or O-linked glycans, require the coordinated action

of hundreds of glycogenes, primarily glycosyltransferases and glycosidases, which are trafficked to specific locations within the endoplasmic reticulum (ER) and Golgi apparatus (Ohtsubo and Marth, 2006). Cytoplasmic and nuclear proteins are frequently modified with O-linked N-acetylglucosamine (GlcNAc) which regulates many biological processes but is beyond the scope of this review. N- and O-linked glycosylation of secreted and membrane protein starts in the ER or early cis-Golgi and is completed in later Golgi compartments. Major animal glycans contain ten monosacchararides: glucose (Glc), Galactose (Gal), N-acetylglucosamine (GlcNAc), N-acetylgalactosamine (GalNAc), fucose (Fuc), mannose (Man), xylose (Xyl), glucuronic acid (GlcA), iduronic acid (IdoA), and sialic acid (SA, either as 5-N-acetylneuraminic acid, Neu5Ac, or as 5-Nacetylglycolylneuraminic acid, Neu5Gc).

The N-glycosylation pathway starts at the ER membrane where a precursor glycan is built upon a dolichol isoprenoid lipid (**Figure 1**). This precursor glycan, with the composition Glc3Man9GlcNAc2, is transferred en bloc onto asparagine residues located within glycosylation sequons of nascent polypeptide chains either co-translationally or shortly after translation by multi-subunit oligosaccharyltransferase (OST) complexes within the lumen of the ER (Li et al., 2008; Zielinska et al., 2010; Shwartz and Aebi, 2011). The transferred Glc3Man9GlcNAc<sup>2</sup> glycan is subsequently trimmed by sequential action of ER glucosidases. Glc-trimming is an essential component of the folding process for most secretory pathway glycoproteins (Helenius and Aebi, 2004). Thus, human diseases that arise from altered biosynthesis or trimming of the N-linked precursor glycan, or ineffective transfer of the precursor to protein will impact the folding and stability of many glycoproteins and, consequently, manifest with multi-systemic and broadly severe clinical phenotypes.

Glycoproteins arrive at the cis-Golgi carrying high-Man glycans (**Figure 1**). Mannose trimming in the cis-Golgi by Golgi α-mannosidases removes Man residues to generate the Man5GlcNAc<sup>2</sup> intermediate. In medial Golgi compartments, Man5GlcNAc<sup>2</sup> is the substrate for GlcNAcT-1, a glycosyltransferase that transfers a GlcNAc residue to a terminal Man residue on the α3-arm of the Man5GlcNAc<sup>2</sup> structure, thereby initiating the synthesis of hybrid and complex N-linked glycans (Stanley, 2011; Moremen et al., 2012). The product of GlcNAcT-1 is also a substrate for core fucosylation, the addition of one or more Fuc residues (depending on the species) to the most proximal core GlcNAc residues attached to Asn. The α3 arm initiated by GlcNAT-1 can be extended with Gal, Sia and/or other residues, resulting in the production of hybrid structures. Removal of the remaining Man residues from the α6-arm allows branching with additional GlcNAc residues catalyzed by specific GlcNAcT enzymes and subsequent extension to generate fully elaborated multiantennary complex glycans in late medial and trans Golgi compartments (Stanley, 2011). The vectorial nature of N-glycan processing is facilitated by enzyme specificity and by the spatial distribution of processing steps across the Golgi apparatus. Therefore, human diseases that impact N-glycan fine structure may arise from genes that encode for processing enzymes or for proteins that regulate Golgi architecture and trafficking. Such diseases may be characterized by relatively restricted phenotypes associated with altered function, half-life, or targeting of specific glycoproteins.

In contrast to N-linked glycosylation, O-linked glycosylation does not rely on a precursor core that is transferred en bloc to the nascent polypeptide. Instead, O-glycosylation is initiated on folding or folded proteins and involves the formation of a glycosidic linkage between serine or threonine and GalNAc, GlcNAc, Man, Glc, Xyl, or Fuc residues (Stanley, 2011). Some O-glycans are specifically elaborated on well-defined protein domains and contribute to protein folding, stability, protease sensitivity, and protein function. The biosynthesis of O-glycans in the secretory pathway is initiated in the early cis-Golgi or in a transitional compartment that retains characteristics of the ER and is completed in subsequent processing steps distributed across the Golgi apparatus. Therefore, human diseases arising from genes that regulate Golgi dynamics may impact both N-linked and O-linked glycoprotein glycosylation.

Congenital disorders of glycosylation (CDGs) are inborn errors in protein and lipid glycosylation that arise from mutations in genes controlling steps in glycan addition (Freeze and Ng, 2011; Jaeken, 2013). More than 100 distinct forms of CDGs have been discovered, many of which display multisystemic defects including severe neurological impairment, highlighting the important role of regulated glycosylation in central nervous system (CNS) functions (Barone et al., 2014; Freeze et al., 2015). Because 1–2% of the genome encodes for glyco-enzymes and glycan transporters, it is likely that many other CDGs remain to be discovered (Freeze et al., 2015). CDGs have been traditionally divided into two large groups (Goreta et al., 2012; Freeze et al., 2014). Type I CDGs impair the synthesis of the dolichol pyrophosphate oligosaccharide precursor of N-linked glycoproteins or its transfer to acceptor proteins (Freeze and Ng, 2011) resulting in decreased efficiency of protein N-glycosylation. Type II CDGs (CDG-II) are characterized by defects in the processing of N-linked glycans or the biosynthesis of O-linked oligosaccharides (Freeze and Ng, 2011; Goreta et al., 2012; Freeze et al., 2014). Although most CDGs exhibit neurological impairment (Freeze et al., 2015), there are no comprehensive studies, aimed at elucidating the molecular mechanisms that link defective glycosylation to the neuropathological aspects of the disease. Animal models that faithfully recapitulate the pathological aspects of the disease, including the neurological defects, provide a valuable resource to study the molecular mechanisms underlying pathology in CDGs. In this review we focus our attention on the advantages offered by the use of Drosophila melanogaster for understanding and modeling the glycobiology of CDGs.

# Drosophila melanogaster AS A MODEL SYSTEM FOR STUDYING CDGs

Drosophila models offer many advantages for studying CDGs as well as other human diseases (Moulton and Letsou, 2016). Fundamental biological processes are highly conserved between Drosophila and humans; approximately 75% of human-disease

FIGURE 1 | N-linked glycosylation pathway. Biosynthesis of the N-linked precursor glycan begins on the cytoplasmic face of the ER where a GlcNAc residue is added in a pyrophosphate linkage to dolichol, an isoprenoid lipid. The GlcNAc-P-P-Dol is extended to form Man5GlcNAc2-P-P-Dol which is then flipped so that the glycan moiety is within the lumen of the ER. Further extension produces a Glc3Man9GlcNAc2-P-P-Dol that is a substrate for the oligosaccharyltransferase (OST) complex, which transfers the precursor glycan en bloc to a nascent polypeptide. This figure depicts the glycosylation of a glycoprotein (brown) with 3 N-linked glycosylation sites (labeled 1, 2, and 3). Once transfered to protein, the glycan precursor is trimmed of its Glc residues during folding as part of the calnexin/calreticulin quality control cycle. CDG Type I mutations affect the biosynthesis of the precursor glycan, its transfer to protein, and early trimming steps. Once successfully folded, glycoproteins bearing high-Man glycans are transported to the Golgi apparatus where Man trimming occurs. In the early cis Golgi, high-Man glycans can be trimmed to Man5GlcNAc<sup>2</sup> by complete removal of Man residues on the α3 arm and partial removal of Man residues on the α6 arm. In the medial Golgi, the first committed step toward production of a complex glycan is taken; GlcNAcT1 adds a GlcNAc to the α3 Man residue to form a hybrid type glycan (site 1 retains this structure). The GlcNAc-extended Man5GlcNAc<sup>2</sup> glycan can be core fucosylated by the addition of a Fuc residue to the internal GlcNAc (site 2). In Drosophila and other arthropods, a second Fuc residue can be added (site 3). Additional Man trimming by Golgi mannosidases provide substrates for branching in the medial and trans Golgi (site 2). Subsequent extention with Gal and capping with sialic acid (shown here, as N-acetylneuraminic acid, NeuAc) completes the maturation of complex N-linked glycans. Hybrid glycans can also be extended on the α3 arm (site 1). The abundance of hybrid and complex glycans is reduced in Drosophila compared to vertebrate species due to the presence of an hexosaminidase that removes the GlcNAc added by GlcNAcT1, thereby blocking additional branching/extension and producing a paucimanose glycan (site 3). CDG Type II mutations impact the availability of substrates and the activity of enzymes that process N-glycans in the Golgi apparatus. Graphical representation of monosaccharide residues and glycan structures is consistent with the Symbol Nomenclature For Glycans (SNFG), which has been broadly adopted by the glycobiology community (Varki et al., 2015).

related genes have a homolog in Drosophila (Reiter et al., 2001; Chien et al., 2002). Moreover the genome of D. melanogaster is far less complex than the human genome and exhibits fewer gene duplications (Hartl, 2000). All of these characteristics, and its extraordinary repertoire of readily available genetic tools, have combined to make Drosophila a valuable, emerging model system for investigating glycan-dependent functions in vivo and for understanding the link between CDG neuropathology and glycan changes (Katoh and Tiemeyer, 2013; Scott and Panin, 2014). Such studies are extremely challenging in vertebrates due to the complexity of the nervous system and the redundancy of glycosylation pathways and enzymes (Freeze et al., 2014; Scott and Panin, 2014). Drosophila combines the advantages of a well-characterized glycome and the availability of electrophysiological and behavioral assays to test neurological impairment in the whole organism (Gatto and Broadie, 2011; Dani et al., 2014; Scott and Panin, 2014). Moreover, larval neuromuscular junction (NMJ) synapses use ionotropic glutamate receptors (GluRs), providing an excellent model system for excitatory synapses in the mammalian CNS.

Various analytic techniques have revealed that the most abundant N-linked glycans on Drosophila glycoproteins are of the high-Man or paucimannosidic type. However, hybrid and complex glycans are also present, although they represent a lower fraction of the total glycan profile when compared with vertebrates (Katoh and Tiemeyer, 2013). The relative paucity of complex glycans in Drosophila is a result of an arthropod-specific glycan processing enzyme encoded by a gene named Fused Lobes (Fdl) that removes the GlcNAc residue added by GlcNAcT-1, thereby blocking further glycan elaboration (**Figures 1**, **2**). The presence of Fdl in the secretory pathway means that Drosophila glycan profiles are skewed away from the highly abundant complex profiles found in most vertebrates. Nonetheless, the glycans that escape the activity of Fdl in Drosophila are readily processed to complexity, indicating common logic underlies glycan maturation in vertebrates and Drosophila. The resulting low content of complex glycans also provides a benefit for this system because it generates a simpler profile to analyze and a larger dynamic range for detecting shifts induced by mutations. Additionally, unlike mammalian organisms, the Drosophila genome contains only a single sialyltransferase (DSiaT), which greatly simplifies in vivo analysis of glycoprotein sialylation (Aoki et al., 2007; Koles et al., 2007; Repnikova et al., 2010).

In this section we describe Drosophila mutants that offer functional models for characterized human CDGs (**Figure 2**). We discuss the phenotypic characteristics that recapitulate the pathological aspects of the human disease and the translational impact of modeling the CDG in this organism. Many other Drosophila mutants have been shown to impact glycoprotein or glycolipid glycosylation and it is likely that the impact of many others is underappreciated (Seppo et al., 2003; Baas et al., 2011; Daenzer et al., 2016; Jumbo-Lucioni et al., 2016). But here, we focus on those mutations that have immediate parallels with human type I and II CDGs.

#### PMM2-CDG (CDG-Ia)

The most prevalent CDG is known as CDG-1a or PMM2- CDG, accounting for around 80% of all diagnosed cases. PMM2-CDG is inherited as an autosomal-recessive trait resulting from mutations in human PMM2, which encodes phosphomannomutase-2, that converts mannose-6-phosphate to mannose-1-phosphate, the precursor of GDP-mannose (Freeze et al., 2014, 2015). Because GDP-mannose is the donor for the addition of the first 5 Man residues to the dolichol-linked precursor, synthesis of N-linked glycans is impacted, as is O-mannosylation. The defect results in hypoglycosylation of many types of glycoproteins including serum glycoproteins, plasma membrane glycoproteins, and lysosomal enzymes. Pediatric patients suffering with CDG-Ia present with variable clinical features that affect nearly all systems and include failure to thrive, hypotonia, psychomotor retardation, ataxia, dysmorphia and coagulopathy (Freeze, 2009; Grünewald, 2009; Jaeken, 2013). Most adult CDG-Ia patients are wheelchair bound and display peripheral neuropathy and mental retardation (Grünewald, 2009).

Overall, the Drosophila PMM2 protein displays 56% amino acid identity with human PMM2 (Parkinson et al., 2016). Parkinson et al. (2016) generated Drosophila pmm2 mutants. Similar to CDG-1a patients, pmm2 mutants displayed uncoordinated movement and reduced lifespan. Analysis of the N-linked glycome of the pmm2-null mutant larval body demonstrated a global suppression of N-linked glycosylation. Furthermore, the N-linked glycome of adult heads with neurally targeted pmm2 RNAi revealed increased abundance of pauci-mannose glycans. Analysis of the larval NMJs revealed altered glycan composition within the heavily glycosylated synaptomatrix which correlated with striking NMJ structural overgrowth and increased neurotransmission strength. Since NMJ synaptogenesis requires trans-synaptic Wnt/Wingless signaling, which in turn depends on expression of Dally-like protein (Dlp, a heparan sulfate proteoglycan) and matrix metalloproteinases (MMPs). Because knockdown of PMM2 resulted in loss of MMP2, reduced synaptic levels of Wingless, Dlp co-receptor and downstream trans-synaptic signaling, the authors propose that the matrix metalloproteome and Wnt signaling pathway might provide potential new targets for developing CDG-1a treatments.

#### SLC35-CDG (CDG-IIc)

Also known as CDG-IIc and leukocyte adhesion deficiency type II syndrome (LAD II), this disorder is caused by mutations in a GDP-Fuc transporter (GFR). SLC35-CDG patients show craniofacial dysmorphism, severe retardation and chronic infections with unusually high leukocytosis. Importantly neutrophils of these patients lack the ability to synthesize the fucosylated glycan sialyl-Lewis X, a ligand of the selectin family of cell adhesion molecules, that is necessary for their recruitment to infection sites (Yakubenia et al., 2008). Several mutations in SLC35 nucleotide sugar transporters have been identified in Drosophila, including the ER Fuc transporter (Efr), fringe connection (frc), slalom (sll) and neurally altered carbohydrate

Drosophila N-linked glycan profile is the existence of an arthropod-specific, N-acetylhexosaminidase known as Fused lobes (Fdl), which converts the precursor for complex glycans (GlcNAc1Man3−5GlcNAc2-Protein) to a paucimannose structure (Man3−5GlcNAc2-Protein) that cannot be extended further. Glycoprotein glycans that escape Fdl are fully capable of being processed into complex structures. (B) Human glycosylation disorders and phenotypic characteristics of the Drosophila model.

(nac). In nac<sup>1</sup> mutant, a conserved serine at position 29 of the Golgi GFR is replaced by a leucine, which abolishes GDP-Fuc transport in vivo and in vitro. Mass spectrometry and HPLC analysis demonstrated reduced core α1,3-and α1,6-fucosylation in nac<sup>1</sup> (Geisler et al., 2012). While Lewis-type glycans have not yet been identified in Drosophila, the commonality of altered fucosylation in SLC35 CDG-IIc and in select Drosophila mutants provides opportunities to investigate the regulation of protein fucosylation in a whole organism.

# COG-CDG (CDG-IIe and CDG-IIi)

The eight-subunit Conserved Oligomeric Golgi (COG) complex is a Golgi tether required for intra Golgi trafficking of vesicles that recycle Golgi resident proteins and is essential for proper localization of Golgi-localized glycosylation enzymes including glycosyltransferases (Ungar et al., 2002; Kranz et al., 2007; Miller and Ungar, 2012; Willett et al., 2013; Climer et al., 2015). Mutations in the genes encoding human COG1, COG2, and COG4–COG8 are associated with monogenic forms of

inherited, autosomal recessive, CDGs-II (Freeze and Ng, 2011; Climer et al., 2015). Common features of patients carrying mutations in COG proteins (COG-CDG) are feeding problems and developmental defects, including microcephaly and growth retardation associated with dysmorphic features, hypotonia and cerebral atrophy (Wu et al., 2004; Spaapen et al., 2005; Foulquier et al., 2006; Kranz et al., 2007; Morava et al., 2007; Ng et al., 2007; Paesold-Burda et al., 2009; Reynders et al., 2009; Zeevaert et al., 2009; Lübbehusen et al., 2010; Fung et al., 2012; Kodera et al., 2015). COG7–CDG patients had the highest mortality within the first year of life and presented with dysmorphic facial features, generalized hypotonia, skeletal anomalies, hepatomegaly, progressive jaundice, cardiac insufficiency, microcephaly, and severe epilepsy (Wu et al., 2004; Spaapen et al., 2005; Morava et al., 2007; Ng et al., 2007; Zeevaert et al., 2009). Defects in COG proteins have been linked to glycosylation alterations in mammalian cultured cells and in COG–CDG patients, including hyposialylation of serum proteins, abnormal synthesis of N- and O-linked glycans and altered glycolipid glycosylation (Kingsley et al., 1986; Suvorova et al., 2002; Wu et al., 2004; Spaapen et al., 2005; Morava et al., 2007; Ng et al., 2007; Zeevaert et al., 2009; Struwe and Reinhold, 2012).

Analysis of phenotypes associated with mutations in Drosophila homologs of human COG complex members highlights both the value and also the limitations of modeling COG-complex disorders in this organism. The Drosophila homolog of human COG5, Four way stop (Fws), is not essential for adult survival but is required for male fertility. Mutations in fws impair spermatocyte cytokinesis, acroblast structure and elongation and individualization of differentiating spermatids (Farkas et al., 2003; Fári et al., 2016). Thus, the Drosophila COG5 mutant presents less severe involvement than the presently known human COG-complex CDGs. On the other hand, loss of COG7 in COG7–CDG patients and in Drosophila mutants results in reduced life span and severe psychomotor defects (Frappaolo et al., 2017). Analysis of N-glycans from heads of Drosophila Cog7 mutants, revealed increased abundance of high-Man type glycans compared to wild type, accompanied by a disproportionate increase of the Man5GlcNAc<sup>2</sup> glycan, which is the precursor for all complex glycans. Additionally, a substantial increase in the abundance of a family of neuralspecific, difucosylated N-glycans known as HRP-epitopes, was detected. However, not all N-glycans were increased in Cog7 mutants. A single sialylated N-glycan was detected among the glycans harvested from adult heads and quantification relative to standard indicated that it was decreased in two mutant allelic combinations compared to wild type (Frappaolo et al., 2017). Moreover, like DSiaT mutants, Cog7 mutant flies exhibit temperature sensitive (TS) paralysis, coordination defects, and altered architecture of larval NMJ. Thus the phenotypic characteristics of our Drosophila COG7–CDG model closely parallel the pathological characteristics of COG7–CDG patients including N-linked glycome defects with hyposialylation. Ongoing analysis of the COG protein interactome is beginning to highlight molecular hierarchies and trafficking paradigms that may underlie altered protein glycosylation (Belloni et al., 2012; Miller et al., 2012; Willett et al., 2013; Hong and Lev, 2014; Climer et al., 2015; Bailey Blackburn et al., 2016; Frappaolo et al., 2017; Sechi et al., 2017).

### ATP6AP2-CDG

The multi-subunit vacuolar-type proton ATPase (V-ATPase) is a highly conserved proton pump, which acidifies intracellular compartments and is essential for endocytosis and vesicular trafficking. Rujano et al. (2018) identified missense mutations in the extracellular domain of the accessory V-ATPase subunit ATP6AP2 that cause a novel glycosylation disorder associated with hepatopathy, immunodeficiency, cutis laxa, muscular hypotonia, dysmorphic features, and psychomotor impairment. Analysis of ATP6AP2-CDG patients' serum proteins revealed hypoglycosylation, a defect that could be recapitulated by ATP6AP2 deficiency in the mouse. Null alleles of the Drosophila ortholog of ATP6AP2 cause an early lethal phenotype in Drosophila. The introduction of an ATP6AP2 transgene carrying the p.L98S mutation in the background of Drosophila ATP6AP2 null mutants, reduced viability and affected the developing optic lobes in larval brains by expanding the pool of optic lobe neuroblasts, a phenotype associated with altered Notch signaling (Vaccari et al., 2010). In agreement with the role of V-ATPase-mediated acidification in autophagic degradation (Mauvezin et al., 2015), p.L98S mutation leads to lipid accumulation and autophagic dysregulation in the liver-like fat body, associated with defects of lysosomal acidification and mTOR signaling. Thus the Drosophila ATP6AP2-CDG model has allowed the elucidation of molecular mechanisms underlying pathological aspects of the human disease.

# CONCLUSION AND PERSPECTIVES

Many impactful studies utilizing model systems (Zebrafish, flies, C. elegans, mice, etc.,) have enhanced our understanding of the underlying biochemical and phenotypic consequences of altered glycan biosynthesis associated with human CDG subtypes. Clinical phenotypes of human CDGs have parallels in these model systems. For the growing subset of CDGs modeled in Drosophila, specific phenotypes related to neural function, lifespan, viability, and glycomic diversity are replicated across these highly divergent species. This phenotypic reproducibility across species should not be surprising since the core biosynthetic pathways for protein glycosylation as well as the basic mechanisms that regulate Golgi trafficking are shared across broad swaths of evolutionary space. This conservation will allow mechanistic questions to be effectively answered in CDG models. One of these key questions is whether phenotypes arise from altered glycosylation of broad sets of glycoproteins or whether aberrant glycosylation of small subsets of glycoproteins can be linked to underlying pathologies. Once candidate proteins, whose glycosylation is altered in a given CDG, are identified by cutting edge glycoproteomics, the genetic tools offered by model systems such as Drosophila will allow unprecedented targeted investigations of the cell- and tissue-specific impacts of glycosylation deficiencies.

#### AUTHOR CONTRIBUTIONS

fgene-09-00436 September 28, 2018 Time: 16:9 # 7

All authors edited and critically revised the manuscript.

#### REFERENCES


#### FUNDING

The Research in MG lab was supported by a grant from Associazione Italiana per la Ricerca sul Cancro (AIRC), grant IG 2017, Id 20779 and a grant from Fondazione Telethon Italy (grant number GEP 14076). Work in the MT laboratory was supported by a grant from the NIGMS/NIH, P41GM103490.



**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.

The handling Editor and author MG declared their involvement as co-editors in the Research Topic, and confirm the absence of any other collaboration.

Copyright © 2018 Frappaolo, Sechi, Kumagai, Karimpour-Ghahnavieh, Tiemeyer and Giansanti. 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.

# SUMO-Targeted Ubiquitin Ligases (STUbLs) Reduce the Toxicity and Abnormal Transcriptional Activity Associated With a Mutant, Aggregation-Prone Fragment of Huntingtin

Kentaro Ohkuni<sup>1</sup>† , Nagesh Pasupala<sup>2</sup>†‡, Jennifer Peek<sup>2</sup> , Grace Lauren Holloway<sup>2</sup> , Gloria D. Sclar<sup>2</sup> , Reuben Levy-Myers<sup>2</sup> , Richard E. Baker<sup>3</sup> , Munira A. Basrai<sup>1</sup> and Oliver Kerscher<sup>2</sup> \*

<sup>1</sup> Genetics Branch, Center for Cancer Research, National Cancer Institute, National Institutes of Health, Bethesda, MD, United States, <sup>2</sup> Biology Department, College of William & Mary, Williamsburg, VA, United States, <sup>3</sup> Department of Microbiology and Physiological Systems, University of Massachusetts Medical School, Worcester, MA, United States

Cell viability and gene expression profiles are altered in cellular models of neurodegenerative disorders such as Huntington's Disease (HD). Using the yeast model system, we show that the SUMO-targeted ubiquitin ligase (STUbL) Slx5 reduces the toxicity and abnormal transcriptional activity associated with a mutant, aggregationprone fragment of huntingtin (Htt), the causative agent of HD. We demonstrate that expression of an aggregation-prone Htt construct with 103 glutamine residues (103Q), but not the non-expanded form (25Q), results in severe growth defects in slx51 and slx81 cells. Since Slx5 is a nuclear protein and because Htt expression affects gene transcription, we assessed the effect of STUbLs on the transcriptional properties of aggregation-prone Htt. Expression of Htt 25Q and 55Q fused to the Gal4 activation domain (AD) resulted in reporter gene auto-activation. Remarkably, the auto-activation of Htt constructs was abolished by expression of Slx5 fused to the Gal4 DNA-binding domain (BD-Slx5). In support of these observations, RNF4, the human ortholog of Slx5, curbs the aberrant transcriptional activity of aggregation-prone Htt in yeast and a variety of cultured human cell lines. Functionally, we find that an extra copy of SLX5 specifically reduces Htt aggregates in the cytosol as well as chromatin-associated Htt aggregates in the nucleus. Finally, using RNA sequencing, we identified and confirmed specific targets of Htt's transcriptional activity that are modulated by Slx5. In summary, this study of STUbLs uncovers a conserved pathway that counteracts the accumulation of aggregating, transcriptionally active Htt (and possibly other poly-glutamine expanded proteins) on chromatin in both yeast and in mammalian cells.

Keywords: Slx5, STUbL, SUMO, ubiquitin, Htt

#### Edited by:

Roberta Fraschini, Università degli Studi Milano-Bicocca, Italy

#### Reviewed by:

Eric M. Rubenstein, Ball State University, United States Hansen Wang, University of Toronto, Canada

> \*Correspondence: Oliver Kerscher opkers@wm.edu

†Joint first authors

‡Present address: Nagesh Pasupala, Biophysics and Biophysical Chemistry, Johns Hopkins School of Medicine, Baltimore, MD, United States

#### Specialty section:

This article was submitted to Genetic Disorders, a section of the journal Frontiers in Genetics

Received: 21 June 2018 Accepted: 27 August 2018 Published: 18 September 2018

#### Citation:

Ohkuni K, Pasupala N, Peek J, Holloway GL, Sclar GD, Levy-Myers R, Baker RE, Basrai MA and Kerscher O (2018) SUMO-Targeted Ubiquitin Ligases (STUbLs) Reduce the Toxicity and Abnormal Transcriptional Activity Associated With a Mutant, Aggregation-Prone Fragment of Huntingtin. Front. Genet. 9:379. doi: 10.3389/fgene.2018.00379

**143**

# INTRODUCTION

fgene-09-00379 September 18, 2018 Time: 12:9 # 2

Ubiquitin and SUMO are members of a conserved family of small ubiquitin-like modifier proteins (UBLs) that can be conjugated to lysine residues of target proteins to modulate their activity, function, localization, and half-life. The conjugation of both SUMO and ubiquitin to numerous target proteins is a multistep process and involves a cascade of similar, yet distinct E1 activating enzymes, E2 conjugating enzymes, and E3 ligases. Additionally, dedicated SUMO or ubiquitin-specific proteases render these protein modifiers conjugation competent and also aid in their deconjugation from modified proteins. As such, the dynamic conjugation and deconjugation of UBLs has key roles in cell growth and the maintenance of genome integrity and has been implicated in disease-related processes including cancer, inflammation, and neurodegeneration (Hoeller et al., 2006; Kerscher et al., 2006; Dorval and Fraser, 2007; Dasso, 2008; Liu and Shuai, 2008; Sarge and Park-Sarge, 2009; van Wijk et al., 2011; Cubeñas-Potts and Matunis, 2013). Mammalian cells express one form of ubiquitin and three forms of conjugatable SUMO (SUMO-1, SUMO-2, and SUMO-3), while budding yeast only expresses one form each of ubiquitin and SUMO (Smt3). Chains of ubiquitin can be formed through conjugation of internal lysines. Analogously, Smt3, SUMO-2, and SUMO-3 can form SUMO chains on the proteins they modify, a property not shared by SUMO-1, which lacks the internal lysines required for polymerization (Ulrich, 2008; Vertegaal, 2010). The majority of proteins that are modified with ubiquitin chains are targeted to the proteasome. In contrast, SUMO chains and hybrid SUMOubiquitin chains do not play a direct role in proteolytic targeting but play an important but poorly understood role in SUMOdependent signaling and the regulation of chromatin (Guzzo et al., 2012).

STUbLs, including the heterodimeric Slx5/Slx8 in budding yeast and the RNF4 homodimer in mammalian cells, are ubiquitin E3 ligases that can specifically target and bind sumoylated proteins and facilitate their ubiquitylation. Members of this unusual subfamily of ubiquitin ligases are well conserved, contain a RING domain required for their ubiquitylation activity, and use multiple SIMs (SUMO-interacting motifs) to target sumoylated substrates (Wang et al., 2006; Kerscher, 2007; Sun et al., 2007; Uzunova et al., 2007; Xie et al., 2007; Mullen and Brill, 2008; Tatham et al., 2008; Prudden et al., 2011; Alonso et al., 2012). Not surprisingly, STUbLs play an important role in the cross-regulation of proteins that can be modified with both SUMO and ubiquitin (Perry et al., 2008; Geoffroy and Hay, 2009). Deletion of SLX5 and SLX8 results in the accumulation of high-molecular weight SUMO adducts and renders cells hypersensitive to DNA damage and perturbed DNA replication (Zhang et al., 2006; Prudden et al., 2007). Similarly, depletion of RNF4 sensitizes cells to DNA damage (Tatham et al., 2008; Geoffroy and Hay, 2009; Yin et al., 2012). However, several lines of evidence suggest that STUbLs also play a critical role in protein quality control. For example, Slx5/Slx8 plays a role in degrading a mutant yeast transcriptional regulator, mot1-103, the nuclear degradation of the SUMO E3 ligase Siz1 in mitosis, and proteolysis of centromeric histone H3 variant Cse4, and also a transcription factor, Mat α2, that is not modified with SUMO (Wang and Prelich, 2009; Westerbeck et al., 2014; Hickey and Hochstrasser, 2015; Ohkuni et al., 2016). Similarly, RNF4 has been shown to regulate the SUMO- and ubiquitin-mediated proteasomal degradation of a mutant cystic fibrosis transmembrane conductance regulator (F508del CFTR), mutant attaxin (Atxn1 82Q), and possibly the reduction of SDSresistant aggregates of mutant huntingtin (Htt, 97QP) in the cytosol of mammalian cells (Ahner et al., 2013; Guo et al., 2014).

Mutant Htt is the causative agent of Huntington's disease (HD), a hereditary neurodegenerative illness that affects 2.71 per 100,000 people worldwide (Pringsheim et al., 2012). The IT15 gene, first discovered in 1993, encodes the huntingtin protein (Htt) which is essential for normal development of mammals and interacts with a variety of proteins implicated in transcription, intracellular transport, and cell signaling. However, the complete extent of Htt function remains unknown (Cattaneo et al., 2005). The amino-terminus of Htt normally contains a stretch of 17–28 glutamine (Q) residues, that is expanded to more than 36Q (and sometimes over 100) in patients with HD. These amino-terminal poly-glutamine expansions form aggregates of mutant Htt that visibly accumulate in neurons and in cell culture models, including budding yeast.

It has been suggested that Htt aggregates may be neuroprotective in that they incorporate cytotoxic Htt monomers into inert cellular inclusions (Arrasate et al., 2004). However, there is also ample evidence that cellular aggregates of Htt sequester a variety of other proteins required for vesicle trafficking, cell cycle regulation, transcriptional regulation, cytoskeletal functions, cell signaling, and protein turnover, thus contributing to the demise of cells expressing these aggregation-prone proteins (Suhr et al., 2001; Bauer et al., 2012).

The majority of Htt aggregates occur in the cytosol, but there is considerable evidence that the accumulation of Htt in the nucleus enhances its toxicity [reviewed in Davies et al. (1997), Lunkes and Mandel (1997), and Benn et al. (2008)]. For example, it has recently been found that a poly-glutamine expanded Htt protein fused with a nuclear localization signal (Htt-103Q-NLS), unlike Htt-103Q, is highly toxic to Wild-type (WT) budding yeast (Wolfe et al., 2014). Interestingly, the authors of this study found that toxicity of Htt-103Q-NLS can be suppressed by overexpression of other poly-Q rich proteins including Nab3, an RNA binding protein (Wolfe et al., 2014). The poly-Q tract on Htt is known to sequester other naturally occurring proteins with poly-Q tracts. For example, there are 14 proteins with poly-Q tracts of at least 20 glutamines encoded in the yeast genome, and at least 66 genes in the human genome that have been classified as encoding poly-Q proteins (Butland et al., 2007). Proteins with poly-Q tracts are involved in a variety of functions, but the majority are classified as transcription cofactors, coactivators, and DNA-binding proteins, or regulators of metabolic processes (Butland et al., 2007). In mammalian cells, poly-Q expanded Htt globally disrupts transcriptional regulation (Steffan et al., 2000; Dunah et al., 2002; Schaffar et al., 2015). In a yeast two-hybrid assay, transcriptional activity of aggregation-prone

Htt was dependent on the length of the poly-Q tract (Gerber et al., 1994; Benn et al., 2008; Atanesyan et al., 2012). These observations underscore the importance of understanding the functional role that Htt has in the nucleus.

We reasoned that Slx5, owing to it's protein quality-control functions, may alter the aggregation or distribution of Htt aggregates in WT cells. Therefore, we investigated whether STUbLs, Slx5, and Slx8 play a role in preventing the toxicity of poly-Q expanded Htt in budding yeast cells. We found that expression of Htt-103Q elicited a severe growth defect and was toxic in slx51 and slx81 mutants. The genetic interaction of Htt with STUbLs led us to examine the functional role of STUbLs in counteracting the toxic effects of Htt-103Q. For this we assessed the interaction of STUbLs with various Htt constructs using a reporter gene assay. Using this assay, we established that both Slx5 and RNF4, the human STUbL ortholog, reduced the transcriptional activity of Htt in yeast and human cells. Functionally, we determined that a plasmid-borne copy of SLX5 reduced the levels of both cytosolic and nuclear Htt aggregates but did not affect the levels of monomeric Htt protein in the nucleus. Finally, we completed a global RNA sequencing study to identify transcripts that are affected by Htt-103Q and modulated by an extra copy of SLX5. Therefore, our data implicates STUbLs in a conserved mechanism that prevents the accumulation of aggregating proteins such as Htt on chromatin and curbs their promiscuous transcriptional activity both in yeast and in mammalian cells.

# MATERIALS AND METHODS

#### Yeast Strains, Plasmids, Mammalian Tissue Culture, and Media

All strains and plasmids used in this study are listed in **Supplementary Table S1**. Unless noted otherwise, preparation of yeast media and manipulation of yeast strains were performed as previously reported (Guthrie and Fink, 2002). Unless otherwise noted, all yeast strains were grown at 30◦C. Yeast plasmids expressing 25Q and 103Q Htt were purchased from Addgene.org (Addgene plasmid # 1177 (GPD-25Q-GFP Htt in p416), # 1180 (GPD-103Q-GFP Htt in p416)). These plasmids were used for growth assays and microscopy. For Htt localization and auto-activation assays, Htt with 25Q, 55Q, and 97Q were PCR-amplified using NEB Q5 hot start high-fidelity polymerase 2× master mix (Cat # M0494S) and cloned into the pCR8/GW/TOPO entry vector (Life Technologies) and then recombined into either pAG414GAL-ccdB-DsRed (Addgene #14359) forming GAL-97QHtt-DsRed/TRP1/CEN (BOK 1213) or pACT2.2gtwy (Addgene # 11346) forming ADH1-GAL4AD-25QHtt/LEU2/2µ (BOK 1207), ADH1-GAL4AD-55QHtt/LEU2/2µ (BOK1209) and ADH1-GAL4AD-97QHtt/LEU2/2µ (BOK 1215). All plasmids expressing Htt encode exon I (17 amino acids) followed by poly-Q and proline-rich regions. NEBase Changer v1.2.1 software at NEB website was used for designing mutagenesis 5 0 -phospho primers and NEB Q5 hot start high-fidelity polymerase 2x master mix was used for PCR amplification. After PCR, template plasmid DNA in the reaction mixture was digested by treatment with DpnI enzyme (cat # R0176S) and PCR amplicon was ligated using T4 DNA Ligase enzyme (Cat # M0202S). All primer sequences used for cloning and mutagenesis are available upon request. Yeast cells were transformed as previously described (Amberg et al., 2005) or using the frozen-EZ yeast transformation II kit (Zymo research corporation, Irvine CA). For mammalian 2-hybrid assays, the Matchmaker Mammalian Assay Kit 2 (Clontech.com Cat. No. 630305) was used as per suppliers instructions. The pVP16 Activation Domain (AD) Htt constructs were designed in the Kerscher lab and synthesized by Genewiz (South Plainville, NJ, United States) to produce pVP16-Htt25Q-AD and pVP16-Htt55Q-AD. RNF4 was PCR amplified and cloned into EcoR1 and HindIII sites in the pM-BD plasmid to produce pM-BD-RNF4.

PC3 (Prostate Adenocarcinoma), PNT2 (Prostate Epithelium), and LNCaP (Prostate Carcinoma) cells were grown in RPMI media with 10% heat inactivated FBS (Thermo Fisher Scientific #10438018) and 1% antifungal/antibiotic (anti/anti) (Thermo Fisher Scientific #15240062). PC12 (Rat pheochromocytoma) cells were grown as above but also contained 10% horse serum. HEK 293 (Embryonic Kidney) cells were grown in DMEM media with 10% heat inactivated FBS and 1% anti-anti. Cells were transfected using Lipofectamine 2000 or 3000 reagents using supplier instruction (Thermo Fisher Scientific, Cat. No. 11668-019 or L3000-015).

#### Growth Curves

Yeast strains YOK2206-2207, YOK2209-2210, YOK2824-2828 were grown overnight in 5 ml selective media with 2% dextrose. OD readings were recorded every hour from OD<sup>600</sup> ∼ 0.15 to OD<sup>600</sup> ∼2.0 for up to 10 h (Thermo Fisher Scientific Spectronic 200). Readings were averaged and graphed in Microsoft Excel. Error bars represent the standard error of four independent cultures for each strain listed. Doubling times were calculated as previously published (Murakami and Kaeberlein, 2009).

# Spotting Assays

Yeast strains were grown overnight in 5 ml selective media with 2% dextrose. When cultures reached mid-log phase (OD<sup>600</sup> 0.8– 1.0), 1 OD of cells was harvested. Cultures were 10-fold serially diluted and 5 µl was spotted onto selective medium containing 2% dextrose. Plates were dried at ambient temperature and incubated at 30◦C for up to 3 days.

# Ortho-Nitrophenyl-β-Galactoside (ONPG) and SEAP Assays

Yeast cultures of pJ694alpha containing the appropriate AD and BD constructs were grown until cells reached mid-log phase (OD<sup>600</sup> of 1 ml = 0.5–0.8) and lac-Z reporter gene expression was determined as outlined in the Clontech Yeast Methods protocols handbook (PT3024-1). Briefly, the exact OD<sup>600</sup> was recorded when the cultures were harvested. Cells

were then washed in Z-buffer (16.1 g/L of Na2HPO4·7H2O, 5.50 g/L of NaH2PO4·H2O, 0.75 g/L of KCl, and 0.246 g/L of MgSO4·7H2O. pH 7.0). Cell pellets were resuspended in 100 µl of Z-buffer and three cycles of freeze/thaw each for 30 s was done to break open the cells. Cells were then incubated in the presence of ONPG (4 mg/ml) in Z-buffer at 30◦C until yellow color developed. Reactions were stopped using 1 M Na2CO<sup>3</sup> and cell debris was removed by centrifugation. The OD<sup>420</sup> was determined using a spectrophotometer and β-galactosidase units were calculated using the formula [β-gal units = 1000 × OD420/(t × V ×OD600)] where t is elapsed time (in minutes) of incubation, V is 0.1 ml times 5 (concentration factor) (Miller, 1972). The β-galactosidase units reported were average values of at least three independent experiments and values were graphed including +/−SD. The Great EscAPe Chemiluminescence kit (Clontech #631737) was used to detect SEAP levels in the mammalian 2-hybrid assay. Twenty-five microliters of culture media were obtained and spun for 1 min at 12,000 rpm to remove cells. The supernatant was transferred to black 96-well plates with clear, flat well bottoms (Corning #353219) and after addition of SEAP substrate solution, Chemiluminescent signals were visualized and analyzed using a Li-COR C-Digit Blot Scanner and also autoradiography film. Student's t-tests were used to analyze statistical significance of SEAP transcriptional levels.

#### Fluorescence Microscopy

Images of live cells were collected using a Zeiss Axioscope two plus microscope (Carl Zeiss Microscopy, LLC, Thornwood, NY, United States) fitted with a Qimaging RetigaTM SRV chargecoupled device digital camera (Qimaging, Surrey, BC, Canada), i-Vision software for macintosh (Bio Vision Technologies, Exton, PA, United States) and a Uniblitz shutter assembly (Vincent Associates/ UNIBLITZ, Rochester, NY, United States). Pertinent filter sets for the above applications include CZ909 (GFP), XF114-2 (CFP), Filter set 15 (DsRed1), and 49 (DAPI and Hoechst 33258) (Chroma Technology Group, Bellows Falls, VT, United States). Where applicable, images were normalized using i-vision software and pseudo-colored and adjusted using Adobe Photoshop software (vs13.0 × 64, Adobe Systems, San Jose, CA, United States).

#### Subcellular Fractionation Assay

Cells were grown in a 2% raffinose synthetic complete medium at 25◦C until reaching mid-log phase. Then, galactose was added to the media to a final concentration of 2% to induce Htt-25Q-NLS-GFP or Htt-103Q-NLS-GFP expression from the GAL promoter for 4 h at 25◦C. Whole cell extract (WCE) was purified from 50 OD<sup>600</sup> equivalent cells. Subcellular fractionation was performed as described previously (Au et al., 2008). Western blot analysis of WCE, soluble, and chromatin fraction was carried out to monitor the Htt25Q-NLS or Htt103Q-NLS levels. Tub2 and histone H3 were used as markers for soluble and chromatin fractions, respectively. Protein levels were quantified using Gene Tools software (version 3.8.8.0) from SynGene (Frederick, MD, United States). Primary antibodies were anti-GFP mouse (1:3000, 11814460001, Roche), anti-Tub2 rabbit (1:3000, Basrai laboratory), and anti-H3 rabbit (1:7500, ab1791, Abcam).

# Total RNA Isolation

Cells were grown in a 2% raffinose synthetic complete medium at 25◦C until reaching mid-log phase. Then, galactose was added to the media to a final concentration of 2% to induce Htt-25Q-NLS or Htt103Q-NLS expression from the GAL promoter for 4 h at 25◦C. Total RNAs were isolated from 3 OD<sup>600</sup> equivalent cells using MasterPureTM Yeast RNA purification kit with DNase I treatment as indicated by the manufacturer (Epicentre). All RNA samples had an RNA integrity number (RIN) of 8 and above, indicative of high sample quality. Half of the sample is used for RNA sequencing, and another half is for RT-PCR for a validation of the RNA sequencing.

#### Reverse Transcription-PCR (RT-PCR)

Total RNAs (100 ng for HBT1, 10 ng for UIP4 and UBC11, and 1 ng for ACT1) were analyzed by AccessQuickTM RT-PCR system (Promega). Primer sets and PCR conditions are available upon request. PCR products were loaded onto Ethidium Bromide-stained 1.5% agarose gels in TBE (KD Medical) and band intensities were quantified with Gene Tools software (version 3.8.8.0) from SynGene (Frederick, MD, United States). Expression levels were calculated based on the standard curve on the same gel and relative values were determined when level of the NLS-Htt25Q-GFP [Vector] was defined as 100.

#### mRNA-Seq and Analysis

Three independent RNA-seq libraries for each of 4 samples were prepared from total RNA using the Illumina TruSeq Stranded Total RNA Kit RS-122-2201. They were pooled and sequenced in a single 150 cycle paired end HiSeq run at the Frederick National Laboratory for Cancer Research (FNLCR) at the CCR Sequencing Facility, NCI, NIH, Frederick, MD 21701. Fifty-six to 81 million pass-filter reads were obtained with > 95% base calling quality of Q30. Reads were adapter-trimmed with lowquality calls removed using Trimmomatic v0.36 and aligned using STAR 2.5.1. The transcriptome reference was annotated transcripts from Saccharomyces cerevisiae S288C, assembly EF4 (Ensembl). One library (YMB10544\_c) contained 45% rRNA sequences and was removed from further analysis (all other libraries contained < 2% rRNA reads). Genewise read counts were quantitated using RSEM 1.2.22, and differential expression analysis was performed using edgeR version 3.20.9 utilizing the tool's GLM functionality. An F-like test was performed first to identify genes showing a statistically significant difference in at least one condition (3961 of 7126 total), and only these genes were included in subsequent pairwise comparisons. Analysis of identified transcripts was completed using the online Panther classification system<sup>1</sup> (Mi et al., 2013) and the Saccharomyces Genome Database<sup>2</sup> .

<sup>1</sup>www.pantherdb.org

<sup>2</sup>www.yeastgenome.org

TRP1 marked plasmid.

duplicate on selective medium (SC-TRP URA) and rich medium (YPD). Patches were then replica plated on SC-URA medium with 5FAA to counter-select against the

# RESULTS

fgene-09-00379 September 18, 2018 Time: 12:9 # 6

# Expression of Poly-Q Expanded Huntingtin Causes a Growth Defect in STUbL Mutants

STUbLs play an important role in the quality control of both SUMO-modified and non-sumoylated proteins (Wang et al., 2006; Xie et al., 2010; Westerbeck et al., 2014). Therefore, we tested our hypothesis that Slx5 and Slx8 are required for growth in the presence of a toxic, aggregation-prone model protein: exon 1 of poly-Q expanded Htt. Budding yeast is established as an exquisite model system for the study of poly-Q expanded proteins (Krobitsch and Lindquist, 2000) and hence we compared the effect of expression of Htt with either a 25-glutamine residue tract (Htt-25Q or 25Q) or an abnormal, aggregationprone 103-glutamine tract (Htt-103Q or 103Q) on the growth properties of WT, slx51, and slx81 cells. Isogenic WT, slx51, and slx81 cells were transformed with low-copy (CEN) plasmids expressing GFP-tagged Htt-25Q or Htt-103Q under control of the constitutive GPD promoter (Krobitsch and Lindquist, 2000). The resulting transformants, or an empty vector control, were grown to mid-log phase and equal numbers of cells were serially diluted and spotted on selective media (**Figure 1A**). Though slx51 and slx81 cells initially formed smaller colonies than WT cells, no severe growth defect or lethality was apparent after 2–3 days of growth at 30◦C for both the vector and Htt-25Q transformants (**Figure 1A** top and middle panel). In contrast, slx51 and slx81 cells transformed with the Htt-103Q construct showed a severe growth defect (**Figure 1A** bottom panel), supporting our hypothesis that STUbLs are required to relieve the growth-inhibiting properties of aggregation-prone poly-Q expanded Htt in budding yeast.

Next, we used liquid cultures to investigate the effect of Htt-25Q and Htt-103Q constructs on the growth of WT, slx51, and slx81. For this analysis, we compared the slopes of growth curves for WT, slx51, and slx81 cells that were transformed with either Htt-25Q or Htt-103Q constructs. First, we found that WT cells transformed with Htt-25Q or Htt-103Q grew equally well as slx51 cells once established in logarithmic phase of growth (**Figure 1B** top left). Similarly, slx51 and slx81 cells transformed with either an empty vector or the Htt-25Q construct displayed similar growth characteristics (**Figure 1B**, top right and bottom left). In sharp contrast, the growth curves for slx51 and slx81 cells transformed with 103Q constructs revealed a significant growth delay with two- to four-fold increases in doubling times from 2 to 8 h (**Figure 1B** bottom right).

Further support that STUbL subunits, Slx5, and Slx8, have a role in preventing the Htt-103Q induced growth delay or toxicity was derived from a shuffle assay used to examine the ability of a slx51 STUbL mutant to grow in the presence of Htt-25Q or Htt-103Q. For this assay, a slx51 shuffle strain (slx51; SLX5/TRP1/CEN) was transformed with Htt-25Q or Htt-103Q. All transformants showed similar growth characteristics and were patched in duplicate onto selective (-TRP -URA) media. Once patches grew in, cells were replica-plated on rich media (YPD) and then onto 5FAA media to counter-select against the TRP-marked WT SLX5 plasmid (Toyn et al., 2000). After two successive replicas onto fresh 5FAA media, the majority of cells with the Htt-103Q construct failed to grow into colonies because they had lost SLX5. In stark contrast, slx51 cells harboring Htt-25Q grew unimpeded because growth of these cells did not depend on SLX5 (**Figure 1C**). In summary, our results show that STUbLs provide an essential function for yeast cells growing in the presence of aggregation-prone, poly-Q expanded proteins.

# Slx5 Reduces the Number of Poly-Q Expanded Huntingtin Aggregates

Intrigued by the poly-Q-induced growth defect in both STUbL mutants, we decided to compare the phenotypic manifestations of aggregation-prone Htt in WT, slx51, and slx81 cells. We used a fluorescence microscope to collect images of WT, slx51, and slx81 cells transformed with either the GFP-tagged 25Q construct or a GFP-tagged 103Q construct. We predicted, based on the results of our growth assays (**Figure 1**), that STUbL mutants would affect the localization of 103Q construct but not the 25Q construct. Consistent with previous results (Krobitsch and Lindquist, 2000), the Htt-25Q-GFP construct was evenly distributed across the nucleus and cytosol of WT, slx51 and slx81 cells (data not shown). In contrast, WT cells expressing Htt-103Q-GFP revealed a mixture of speckles, aggregates, and diffuse-staining cells (see **Supplementary Figure S1**). However, the majority of slx51 and slx81 cells expressing 103Q did not reveal a GFP signal as these cells were dead, as determined by a vital stain that differentiates live and dead cells (**Supplementary Figure S1**). By comparison, WT cells expressing 103Q contained less than 5% of dead or dying cells in the culture. These results also show that constitutive expression of 103Q, unlike 25Q, results in lethality of STUbL mutants. These data are consistent with our growth assays (**Figures 1A–C**) and support our conclusion that STUbLs fulfill an essential role in preventing cytotoxicity due to poly-Q expanded proteins such as Htt-103Q.

The Htt-103Q-induced lethality in slx51 strain impeded our microscopic analysis of Htt toxicity in STUbL mutants and hence we assayed the effect of plasmid-borne SLX5 on the phenotype of 103Q aggregates in WT cells. We reasoned that Slx5, owing to it's quality-control functions, may alter the aggregation or distribution of Htt aggregates in WT cells. WT cells were transformed with GFP-tagged 103Q and either a SLX5 CEN plasmid (under control of its own promoter on a low-copy CEN vector) or an empty control vector. The transformants were grown to mid-logarithmic phase in selective media and 103Q aggregates were analyzed using fluorescence microscopy (**Figure 2A**). We determined that the incidence of 103Q aggregates was reduced by almost 14-fold while the number of diffuse-staining cells increased by at least twofold with plasmid borne SLX5 (**Figure 2B**).

Htt has been reported to reside both in the cytosol as well as the nucleus, but the majority of Htt aggregates are observed to form in the cytosol (Davies et al., 1997; Krobitsch and Lindquist, 2000). The nuclear localization of Slx5 and Slx8 (Cook et al., 2009) and the lethality of Htt103Q in slx51 and slx81 strains prompted us to further investigate the role

#### FIGURE 2 | Continued

deviation. Y-axis: percent of cells (n = 100/experiment). Y-axis: phenotypes scored (C) Aggregates of poly-Q expanded Htt are localized in the cytoplasm. WT and slx51 strains transformed with GAL-Htt-97Q-DsRed (YOK 3112 and YOK 3114) were grown overnight in SC-TRP medium with 2% raffinose. Cultures were diluted to ∼0.2 OD in a fresh medium with 2% galactose and incubated for an additional 16 h for expression of Htt-97Q-DsRed prior to imaging Htt aggregates using a fluorescence microscope. Nuclei were stained with Hoechst dye. Merged images indicate the absence of Htt-97Q aggregates in nuclei (yellow arrow-heads).

of STUbLs in the localization of Htt. For this analysis, we transiently expressed 97Q-DsRed under control of the strong inducible GAL promoter in WT and the slx51 mutant. This transient expression prevented the cytotoxicity associated with constitutive expression of aggregation-prone Htt in slx51 strains. After galactose induction, we imaged the nuclei of live WT and slx51 cells were stained with Hoechst dye (33342). In both WT and slx51 cells (n > 200), Htt aggregates or speckles were solely observed in the cytosol. A low level of diffusely staining 97Q-DsRed was evenly distributed between the cytosol and the nucleus of WT and did not appear to be enriched in either compartment. A similar localization pattern for 97Q-DsRed to that in WT cells was observed in the slx51 cells (**Figure 2C**). Based on these results we propose that 97Q, under the conditions employed, does not readily form large aggregates in the nuclei of yeast cells.

#### Regulation of Transcriptional Activity of Htt by Slx5 and RNF4

It was previously reported that poly-Q expanded Htt, in the absence of a Gal4-DNA-binding fusion (BD), induces the expression of reporter genes in a two-hybrid reporter assay (Atanesyan et al., 2012). This transcriptional auto-activation was directly related to the length of the poly-glutamine tract in Htt (Atanesyan et al., 2012). Therefore, we determined the effect of Slx5 on this poly-Q dependent transcriptional activity. Htt-25Q and Htt-55Q were fused to the Gal4 activation domain (AD) and assayed for the auto-activation of each construct in the presence or absence of BD-Slx5 or just BD. Consistent with published data (Atanesyan et al., 2012), all AD fusions of Htt, by themselves, induced expression of both a HIS3 and a lacZ reporter gene, indicating that both 25Q and 55Q associate with the Gal4-UAS independent of a BD (**Figure 3A**). We used AD-Htt-25Q and AD-Htt-55Q to avoid the potential toxicity associated with AD-Htt-97Q. Auto-activation of the HIS3 reporter gene was scored using a growth assay, transformed cells where diluted and spotted on media with (SD-Trp-Leu) or without histidine (SD-Trp-Leu-His). Concomitantly, auto-activation of the lacZ reporter was quantitated using ONPG assays that were performed in triplicate. Intriguingly, when the AD-Htt constructs were paired with BD-Slx5, the Htt-induced auto-activation was reduced to background levels (**Figure 3A**). To confirm that the reduction of the Htt-induced transcriptional auto-activation is not due to the decreased expression of Htt 25Q and Htt 55Q in the presence of

2842) or Htt-103Q-GFP with SLX5 (YOK 2843) were grown to mid-logarithmic phase in selective medium. Images of 103Q diffuse staining, aggregates, and speckles in the indicated strains were recorded and then quantitated. Average counts for three independent experiments were graphed +/– standard (Continued)

(Continued)

#### FIGURE 3 | Continued

fgene-09-00379 September 18, 2018 Time: 12:9 # 9

activity using ONPG assays (error bar – SD). (B) Effect of Slx5sim expression on the repression of poly-Q expanded Htt auto-activation as indicated by growth on SC-TRP LEU HIS medium and β-galactosidase assays. (C) Expression of the human SLX5 ortholog RNF4 represses the transcriptional activity of poly-Q expanded Htt. To assess activation of the lacZ reporter gene in the indicated strains β-galactosidase units of strains were determined and graphed +/– standard deviation. (D) RNF4 significantly reduces the transcriptional activation of Htt in mammalian cells. Mammalian two hybrid analysis of PC3, LNCaP, HEK 293, PNT2, and PC12 cells [n = 3]. Reporter gene auto-activation of 25Q and 55Q Htt was assessed by transfection of pM-EV (empty vector) and pVP16-Htt. Where indicated, pM-RNF4 was cotransfected with 25Q or 55Q Htt to measure RNF4's inhibitory function. Values of relative SEAP luminescence for 55Q are graphed as shown and values for both 25Q and 55Q are shown in the table below the graph. Assay results were normalized to the positive control. Error bars represent standard deviation. ∗∗p < 0.01, ∗∗∗p < 0.001.

BD-SLX5, we tested the expression of AD-Htt 25Q and AD-Htt-55Q by Western blotting (**Supplementary Figure S3**). The result shows that the steady state levels of both Htt-25Q and Htt-55Q were not affected by BD-SLX5.

Next we asked if the suppression of AD-Htt reporter gene activation was dependent on the SIMs in Slx5. We combined a SIM mutant of Slx5 that fails to interact with SUMO (BD-Slx5sim) with AD-55Q and assessed the auto-activation properties of our poly-Q Htt constructs on the two reporters using growth and quantitative ONPG assays. We found that Slx5sim reduced the auto-activation of the AD-55Q construct significantly less than WT Slx5 (**Figure 3B**, ONPG assay). This data suggests that SUMO-binding may support the ability of Slx5 to suppress the transcriptional activity of AD-Htt. However, SUMO-binding of Slx5 may not be a steadfast requirement to reduce Htt toxicity because the Slx5sim mutant can still suppress the Htt-103Q growth phenotype (**Supplementary Figure S2**).

To determine whether ability of Slx5 to repress auto-activation of poly-Q expanded Htt is evolutionarily conserved, we tested mammalian BD-RNF4 in combination with AD-25Q and AD-55Q. Consistent with results for Slx5, RNF4 also repressed the auto-activation activity of AD-25Q and AD-55Q constructs in our reporter assay (**Figure 3C**). In the presence of RNF4, the auto-activation activity of AD-25Q and AD-55Q was reduced threefold, when compared to AD-55Q alone. These results strongly support a role for RNF4, and other STUbLs, in counteracting the aggregation of transcriptionally active Htt and possibly other poly-Q expanded proteins associated with neurodegenerative diseases.

Finally, we tested whether RNF4 curbs the transcriptional activity of aggregation-prone proteins in a mammalian tissue culture model of Huntington's disease, employing a mammalian Matchmaker (2-hybrid) assay. For this approach, both Htt-25Q and Htt-55Q were cloned into the pVP16AD Gal4-activation domain vector and co-transformed with the reporter plasmid pG5SEAP into 5 separate cell lines (PC3, LNCaP, HEK293, PNT2, and PC12). Consistent with our finding in yeast, all mammalian cell lines recapitulated the Htt-dependent transcription of the pG5SEAP reporter (**Figure 3D** – blue bars). Next, reporter gene activation was assayed in the presence of BD-RNF4. Mammalian two-hybrid analysis in all cell lines displayed a significant decrease in the transcriptional activation of 25Q and 55Q Htt upon addition of RNF4 BD (**Figure 3D** green bars and table). The strength of RNF4's inhibitory effect ranged from 30–60% reduction on 25Q but was statistically significant for all 5 cell lines (Student's T-test), indicating that RNF4's inhibitory effect is consistent and reproducible. In all cell lines, 55Q mHtt displayed greater transcriptional reduction (40–70%), suggesting that this poly-Q expanded 55Q Htt is more amenable to RNF4's activity. Due to their neuronal origin, results from the PC12 cell two-hybrid are the most physiologically relevant model of Huntington's Disease. As an important indicator of specificity, reporter gene activation by a positive control construct (pMpVP16), was not affected by transfection of RNF4 (data not shown). In summary, we have now shown that both in yeast and mammalian cells auto-activation of Htt can be significantly modulated due to the activity of STUbLs.

#### Slx5 Reduces Chromatin-Associated Htt Aggregates in Budding Yeast

To study the physiological relevance of our assays for transcriptional activity, we examined whether Htt-103Q associates with chromatin using subcellular fractionation of whole-cell lysates after overexpression of nuclear targeted Htt-25Q-NLS-GFP or Htt-103Q-NLS-GFP. We assayed levels of Htt in whole-cell extracts, soluble fractions, and chromatin (**Figure 4A**). As expected, both aggregated (high-molecular weight) and non-aggregated (53 kD, monomer) forms of Htt-103Q, but not Htt-25Q, were clearly detectable in the chromatin fraction, indicating that both aggregated and non-aggregated Htt-103Q associate with chromatin.

To examine the role of SLX5 in modulating chromatin bound Htt-103Q, we assayed levels of Htt-103Q-NLS-GFP in the presence or absence of plasmid-borne SLX5 (**Figures 4B–E**). Consistent with the microscopy of Htt-103Q-GFP expressing cells (**Figure 2B**), the soluble, monomeric form of Htt-103Q (53 kDa) was increased with plasmid-borne SLX5 (**Figures 4B,C**). Importantly, aggregated Htt in the chromatin fraction was reduced ∼3-fold (p = 0.0262) in the strain with plasmid borne SLX5 (**Figures 4B,D**). In contrast, monomeric, chromatinbound Htt-103Q (53 kDa) remained similar in both strains (**Figures 4B,E**). These data show that increased expression of SLX5 specifically reduces Htt-103Q aggregates in chromatin. We propose that STUbLs contribute to reducing chromatinassociated Htt aggregates.

### Identification of Htt-Altered Transcripts Modulated by a STUbL in Yeast

The reduced association of aggregated Htt-103Q-NLS with chromatin in the presence of plasmid-borne SLX5 (**Figure 4B**) led us to postulate that Slx5 curbs the abnormal transcriptional activities induced by Htt-103Q-NLS. Hence, we performed genome-wide RNA-seq analysis to examine the transcriptome

#### FIGURE 4 | Continued

fgene-09-00379 September 18, 2018 Time: 12:9 # 11

histone H3 were used as markers for soluble and chromatin fractions, respectively. (B) WCEs prepared from equal numbers of cells expressing Htt103Q-NLS-GFP with (SLX5-CEN) or without (Vector) were fractionated into soluble and chromatin fractions as described in A. Htt103Q-NLS-GFP levels were monitored by western blot analysis with anti-GFP antibody. Three independent transformants were assayed and shown are the results from two of these. (C) Quantification of the 53 kD GFP signals in soluble fraction from 4B. The 53 kD GFP was normalized using Tub2 levels in soluble fraction. The graph represents the mean of three independent clones with SEM. P-value is 0.0317. (D) Quantification of the aggregate GFP signals in chromatin fraction from 4B. The aggregate GFP signal was normalized using H3 levels in chromatin fraction. The graph represents the mean of three independent transformants with SEM. P-value is 0.0262. (E) Quantification of the 53 kD GFP signals in chromatin fraction from 4B. The 53 kD GFP (shorter exposure) was normalized using H3 levels in chromatin fraction. The graph represents the mean of three independent transformants with SEM. P-value is 0.3523.

of four strains expressing either Htt-25Q-NLS or Htt-103Q-NLS with or without plasmid-borne SLX5. Consistent with the effect of Htt-103Q on transcription, our results showed that the expression of > 50% of all yeast genes (3438 genes) was altered in the presence of Htt-103Q-NLS when compared to Htt-25Q-NLS with empty vector (**Figure 5A**, 25Q [V] vs. 103Q [V]). Of the 3438 genes affected by Htt-103Q-NLS, 48.6% of the genes were up-regulated and 51.4% were down-regulated. These results show that chromatin associated Htt-103Q-NLS affects global transcription in budding yeast.

We next analyzed the effect of plasmid-borne SLX5 on the transcriptome of cells expressing Htt-25Q-NLS or Htt-103Q-NLS. Our RNA-seq data showed that SLX5 had a minimal effect on the transcriptome of cells expressing Htt-25Q-NLS as only 33 genes were differentially expressed (25Q [V] vs. 25Q [SLX5]). In contrast to this, plasmid-borne SLX5 affected the transcription of 398 genes in cells expressing Htt-103Q-NLS when compared to Htt-103Q-NLS without SLX5 (**Figure 5A**, 103Q [V] vs. 103Q [SLX5]). The majority of the 398 genes encode for proteins that reside in the cytoplasm (37%), nucleus (19%), or mitochondria (14%) (**Figure 5B** and **Supplementary Table S4**). Of the 398 genes that were significantly altered by plasmid-borne SLX5, 66% (261) genes were upregulated and 34% (137) genes were down-regulated (**Supplementary Table S2**). We observed two distinctive characteristics on the transcriptome of Htt-103Q-NLS with and without plasmidborne SLX5 (**Figure 5C**). First, for 99.4% (361 out of 363) of SLX5-affected genes, the effect of added Slx5 was inversely correlated with that of 103Q (**Supplementary Table S2** and **Figure 5A**). Plasmid-borne SLX5 upregulated the expression of 68.0% (247) of the genes that were downregulated by Htt-103Q-NLS, and downregulated the expression of 31.4% (114) of genes that were upregulated by Htt-103Q-NLS (**Figure 5A**). For example, expression of YAL008W was down-regulated by Htt-103Q-NLS, and up-regulated by plasmid-borne SLX5. In contrast, expression of YAL014C was up-regulated by Htt-103Q-NLS, and down-regulated by plasmid-borne SLX5. Only two genes (YDL182W and YGR092W) were an exception to this pattern (**Supplementary Table S2**). A second distinctive characteristic of the transcription profiles show that about 25% of the genes that are affected by plasmid-borne SLX5 are neighbors or adjacent to each other on the chromosome (e.g., YBR052C, YBR053C, and YBR054W) (**Supplementary Table S3**). The RNA sequencing data generated in this study have been deposited in NCBI's Gene Expression Omnibus (Barrett et al., 2013) and are accessible through GEO Series accession number GSE115990.

To confirm the transcriptome data from RNA-seq, we performed RT-PCR to assay the transcription of a subset of Htt-103Q-NLS/Slx5 modulated genes such as YDL223C/HBT1, YPL186C/UIP4, and YOR339C/UBC11 (**Figure 5D**). Consistent with our RNA-seq data, we found that expression of YDL223C/HBT1 and YPL186C/UIP4 is down-regulated by Htt-103Q-NLS, and up-regulated by plasmid-borne SLX5 (Down-Up). In contrast, YOR339C/UBC11 is up-regulated by Htt-103Q-NLS, and down-regulated by plasmid-borne SLX5 (Up-Down). In agreement with the RNA-seq data the expression of ACT1 was not significantly affected when assayed by RT-PCR. In summary, our data shows that chromatin-associated Htt-103Q-NLS affects global transcriptional in budding yeast. Most importantly, we define a role for Slx5 in modulating the aberrant transcriptional activity, induced by chromatin-associated Htt-103Q-NLS.

# DISCUSSION

In this study we show, for the first time, that STUbLs are required to prevent the toxicity associated with an aggregation-prone protein namely poly-Q expanded Htt and define a functional role for STUbLs in counteracting the toxic effects of Htt-103Q expression. Using reporter gene assays we determined that Slx5 and RNF4 reduce the transcriptional activity of Htt in yeast and human cells, respectively. For example, Htt fused to the Gal4 activation domain (AD) auto-activates Gal4-regulated reporter genes. However, reporter gene activation by Htt-AD is reduced to background levels in the presence of BD-Slx5 or BD-RNF4. Most importantly, our results show that Slx5 reduces cytosolic and chromatin-associated Htt-103Q aggregates and modulates the transcriptome of cells expressing Htt-103Q. Taken together we provide evidence for a conserved role of STUbLs in preventing the accumulation of aggregating proteins such as Htt on chromatin and propose that STUbLs counteract the transcriptional effect of these aggregates in yeast and mammalian cells.

In the initial stages of our analysis of Htt in STUbL mutants, we focused on the cellular distribution and aggregates formed by aggregation-prone Htt in yeast. We detected aggregates, speckles, and diffuse-staining Htt in both WT and a yeast STUbL mutant, slx51 (**Supplementary Figure S1**). Due to its toxicity in STUbL mutants, we ultimately studied the effect of an extra plasmid-borne copy of SLX5 in WT cells expressing Htt-103Q. Presence of the SLX5 plasmids increased diffusely staining Htt in WT cells while reducing the incidence of Htt aggregates

#### FIGURE 5 | Continued

fgene-09-00379 September 18, 2018 Time: 12:9 # 13

between Htt25Q-NLS and Htt103Q-NLS (25Q [V] vs. 103Q [V]) is 363 genes. The expanded view of A shows that expression of most of the 363 genes (99.4%) inversely correlates between 25Q [V] vs. 103Q [V] and 103Q [V] vs. 103Q [SLX5]. Expression of 247 of the 363 genes (68.0%) is down-regulated by Htt103Q, and this effect is reversed by plasmid-borne SLX5 (Down – Up). Expression of 114 of the 363 genes (31.4%) is up-regulated by the Htt103Q, and this effect is reversed by plasmid-borne SLX5 (Up – Down). (B) Subcellular localization of differentially expressed genes. The localization of proteins encoded by the 398 genes indicated in Supplementary Table S2 was analyzed using cellular components assignment from the PANTHER Classification System and the Saccharomyces Genome Database. Pie chart shows a ratio of the genes placed into cellular component categories. Individual genes are listed in Supplementary Table S4. (C) Schematic of gene expression in Htt25Q-NLS and Htt103Q-NLS cells and the effect of plasmid-borne SLX5 on the transcriptome of Htt103Q-NLS cells. Expression of genes A and B are downregulated and upregulated in Htt103Q-NLS cells relative to Htt25Q-NLS cells, respectively. Slx5 reduces the association of Htt with chromatin and this contributes to the reversal in gene expression such that gene A is upregulated and gene B is downregulated. (D) RT-PCR validation of gene expression analysis. Total RNAs were purified from strains expressing either Htt25Q-NLS or Htt103Q-NLS from a GAL promoter for 4 h with or without plasmid-borne SLX5. RT-PCR analyses was performed using the same samples used for the RNA-seq. Relative intensities are reported as the mean ± SD of three biological repeats. Reactions for YDL223C/HBT1 and YPL186C/UIP4 were performed in duplicate. N = 3 for YOR339C/UBC11 and YFL039C/ACT1, N = 6 for YDL223C/HBT1 and YPL186C/UIP4.

(**Figures 2A,B**). However, we failed to detect a reproducible, STUbL-dependent, reduction of Htt by western blot analysis (for example **Supplementary Figure S3**). Previously it has been reported that the STUbL RNF4 is involved in the degradation of another poly-Q expanded protein, Atxn1 82Q (Guo et al., 2014). While our data are consistent with a re-distribution of Htt aggregate, we did not observe that Slx5 altered the steadystate levels of this aggregation-prone protein. One explanation for this may be that budding yeast cells do not disassemble the nuclear envelope, making it difficult to observe the effect of nuclear localized STUbLs on aggregates of Htt in the cytosol. We overcame this limitation in our yeast model by using nucleartargeted Htt-103Q to assess both the transcriptional activity and the chromatin association of Htt in the presence or absence of SLX5 (**Figure 3** and **Figure 4**).

First, using Gal4-based two-hybrid reporter assays, we were able to show that yeast Slx5 and human RNF4, both nuclear localized proteins, curb the transcriptional activity of Htt. Therefore, we predict that the role of RNF4 in mammalian cells is to dispel transcriptionally active Htt complexes rather than to degrade cytosolic Htt aggregates. However, at this point we cannot entirely exclude the possibility that Slx5 and RNF4 form repressive promotor-associated complexes see (Cubeñas-Potts and Matunis, 2013). Regarding the Gal4-AD-fusion of Htt, similar constructs have proven invaluable in defining the aberrant transcriptional activity of poly-Q expanded proteins. The most important observation in this regard is that the poly-Q domain is necessary and sufficient for both the targeting to the Gal4 UAS and reporter gene activation (Atanesyan et al., 2012). Furthermore, introduction of a poly-Q stretch into transcription factors increases their transcriptional activity. Even though we have not yet observed a direct, physical interaction between Htt and Slx5 (we predict a transient interaction involving Httassociated proteins), our assays are consistent with an important role of Slx5 in counteracting nuclear activities of aggregation prone, chromatin associated proteins. The finding that a STUbL plays an important role in transcriptional regulation is not entirely surprising. Before it became known as a STUbL, RNF4 had already been identified as a co-regulator of androgen receptor-dependent transcription (Yan et al., 2002). Furthermore, RNF4 can act both as a transcriptional activator or a repressor depending on the proteins it interacts with (Fedele et al., 2000).

Second, using Gal-driven, nuclear-targeted 103Q constructs, we assessed the chromatin association of an aggregation-prone Htt construct. This time we were able to clearly document that an extra plasmid-borne copy of SLX5 reduced the levels of chromatin-associated Htt (**Figure 4**). The association of Htt with DNA, transcription factor recognition elements, and transcription factors has previously been reported (Benn et al., 2008). STUbLs may provide a mechanism to counteract these inappropriate associations of Htt. For example, it is tempting to speculate that Slx5 recruits Cdc48/Ufd1/Npl4 (Cdc48-UN), a SUMO-targeted STUbL effector, to dislodge Htt from chromatin (Nie et al., 2012; Bergink et al., 2013). Cdc48 has also been identified in association with Htt aggregates and we are now studying the effect that Cdc48-UN plays in Htt-mediated transcriptional activation (Wang et al., 2008).

Finally, we have completed a global RNA sequencing study to identify those transcripts that are affected by nuclear-targeted Htt-103Q and modulated by an extra plasmid-borne copy of SLX5. Our transcriptome analysis revealed that SLX5 counteracts transcriptional abnormalities of 398 genes induced by expression of 103Q-NLS. Dysregulated transcripts encode proteins localized throughout the cells, with the majority enriched in the cytoplasm (263) nucleus (33) and mitochondria (19). Functional categorization of the differentially transcribed genes showed that at least 22 are involved in transcription, transcriptional regulation, and RNA/DNA binding (RTC3, RPA43, RPB7, BUD27, TFA2, SRB7, YAP7, MCM1, PHO4, MAP1, HSP31, SNF5, RPP1, TMA22, RPS27A, MRPL49, NHP2, RPL7A, RPL7B, MAP1, HST2, and CBC2) (**Supplementary Table S4**). We posit that some of the transcriptionally active and chromatin-associated proteins identified in our study represent genuine STUbL targets. Additionally, several SLX5-modulated genes identified here have previously been described in other Htt studies [e.g., Glo2 (human HAGH1), ZTA1 (human zeta crystalline), Msb1, COA2, BUD22, ERG5, and TIR1], supporting a genuine role of STUbLs in counteracting huntingtin-mediated dysregulation (Willingham et al., 2003; Wolfe et al., 2014). We prefer a model in which Htt aggregates may contain both sumoylated and non-sumoylated proteins, including those listed above. STUbLmediated ubiquitination could then result in the recruitment of the Cdc48-UN desegregase and the subsequent proteasomal degradation of ubiquitylated proteins in the aggregates (reviewed in Kerscher, 2016).

In summary, the STUbL/Htt assay is one of the first of its kind to assess the ability of RNF4 and other STUbLs to modulate the activity of transcriptionally active, aggregation-prone proteins. This reporter assay should complement other sophisticated genetic tools used to study protein aggregation processes (Newby et al., 2017). Results from our reporter assays are consistent with biochemical and genome-wide transcriptome data and provide evidence for a role of STUbLs in preventing toxicity due to aggregation-prone Htt in the nucleus. Overall, our findings indicate that STUbLs can reduce the chromatin association and abnormal transcriptionally activity of Htt (or other aggregating proteins) and suggest that mammalian STUbLs may play neuroprotective functions in Huntington's Disease.

# AUTHOR CONTRIBUTIONS

fgene-09-00379 September 18, 2018 Time: 12:9 # 14

OK designed the study, drafted, wrote, and revised the manuscript, designed and completed the experiments, collected and interpreted the data, supervised the students and postdoc coauthors, approved the content for publication, and is accountable for all aspects of the work. KO, NP, and JP acquired, analyzed, and interpreted the data for the work, and revised the manuscript. GH analyzed and interpreted the data for the work. GS acquired, analyzed, and interpreted the data for the work. RL-M made substantial contributions to the conception and design of the work, and revised the manuscript. RB made substantial contributions to the analysis and interpretation of data for the work, and revised the manuscript. MB was involved in the study design, helped in writing the manuscript, revised the manuscript, supervised the data collection and interpretation, supervised the postdoc co-authors, approved the content for publication, and is accountable for all aspects of the work.

#### FUNDING

This research was supported by NSF grant MBC#1051970, Award No. 17-1 from the Commonwealth of Virginia's Alzheimer's and Related Diseases Research Award Fund (ARDRAF), and a Broderick Family/Goldman Sachs TDAP award to OK. KO and MB were supported by the Intramural Research Program of the National Cancer Institute, National Institutes of Health. Research support for W&M students was provided by the Bailey-Huston Research fund, Summer research fellowships and freshman research funds from the HHMI Science Education and Research program at William & Mary to GS, RL-M, GH, and a Charles Center Honors Fellowship to JP.

#### ACKNOWLEDGMENTS

We would like to thank all current and former members of the OK's and MB's laboratories, Tamara Golden for critical reading

#### REFERENCES

Ahner, A., Gong, X., Schmidt, B. Z., Peters, K. W., Rabeh, W. M., Thibodeau, P. H., et al. (2013). Small heat shock proteins target mutant cystic fibrosis transmembrane conductance regulator for degradation via a small ubiquitin-like modifier-dependent of the manuscript, Lidia Epp for assistance with sequencing, the Cyr lab for providing Htt25Q-NLS and Htt103Q-NLS plasmids, Alexis Sedgewick for strain construction, and Andrew Halleran. RNA sequencing was conducted at the Frederick National Laboratory for Cancer Research (FNLCR) at the CCR Sequencing Facility, NCI, NIH, Frederick, MD, United States.

# SUPPLEMENTARY MATERIAL

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fgene. 2018.00379/full#supplementary-material

FIGURE S1 | WT strain expressing Htt-103Q-GFP alone (YOK 2842) or Htt-103Q-GFP together with SLX5 (YOK 2843) were grown to mid-logarithmic phase in selective medium. Images of yeast cells with diffuse staining 103Q-GFP, aggregates, and speckles were recorded, counted, and graphed. Additionally, we stained cells with the LIVE/DEAD Yeast Viability Kit (Thermo Fisher) to quantitate dead or dying cells in the culture (dead). Average counts for three independent experiments were graphed +/− standard deviation. Y-axis: percent of cells.

FIGURE S2 | The slx5SIM mutant suppresses lethality of Htt-103Q in slx51. slx51 strain YOK821 expressing 103Q-Htt/URA3 was transformed with SLX5 plasmid (BOK376), slx5SIM mutant (BOK463), and an empty vector (pRS425). Resulting transformants were struck to appropriate selective media and incubated for 3 days at 30◦C.

FIGURE S3 | Steady-state protein levels of AD-25Q and AD-55Q are not grossly affected by expression of BD-Slx5. AD-25Q and AD-55Q were expressed in the presence or absence of BD-Slx5 in the yeast two-hybrid reporter strain. Proteins were extracted, separated by SDS–PAGE and western blotted with an antibody to the Gal4-AD or PGK, a loading control (Szymanski and Kerscher, 2013). EV – empty Gal4-AD vector expressing only Gal4-AD.

TABLE S1 | Strains used in this study.

TABLE S2 | Expression profiles of 398 genes regulated by an extra copy of SLX5. An extra copy of SLX5 alters the expression profiles of 398 genes in the Htt103Q background (Htt103Q-NLS [Vector] vs. Htt103Q-NLS [SLX5]). 363 out of these 398 genes are also found to be differently expressed between Htt25Q-NLS and Htt103Q-NLS (Htt25Q[Vector] vs. Htt103Q[Vector]). Expression of most of the 363 genes (99.4%) is inversely correlation between Htt25Q [Vector] vs. Htt103Q [Vector] and Htt103Q [Vector] vs. Htt103Q [SLX5]. Only two genes do not exhibit this inverse relationship (Exceptions: YDL182W and YGR092W). Up- and down-regulated transcripts are shown in red and blue, respectively. Green shows 35 genes that are not detected the comparison between Htt-25Q and Htt-103Q (Htt25Q [Vector] vs. Htt103Q [Vector]).

TABLE S3 | List of clustering genes. The 398 genes, differently expressed between Htt103Q [Vector] and Htt103Q [SLX5], are analyzed. We used systematic name to search the neighboring genes. Up- and down-regulated profiles are shown in red and blue, respectively.

TABLE S4 | Cellular Components Categories of genes modulated by Slx5. The number of individual genes placed into cellular component categories is listed in the table. Individual genes are colored red (increased expression in cells with Htt103Q-NLS and Slx5) or blue (decreased expression in cells with Htt103Q-NLS and Slx5), black (only one data-point – compare Supplementary Table S2). Note that some genes analyzed are part of multiple cellular component categories.

pathway. Mol. Biol. Cell 24, 74–84. doi: 10.1091/mbc.E12-09- 0678

Alonso, A., D'Silva, S., Rahman, M., Meluh, P. B., Keeling, J., Meednu, N., et al. (2012). The yeast homologue of the microtubule-associated protein Lis1 interacts with the sumoylation machinery and a SUMO-targeted ubiquitin ligase. Mol. Biol. Cell 23, 4552–4566. doi: 10.1091/mbc.E12-03-0195


and thereby mediate the recruitment of BRCA1 to sites of DNA Damage. Sci. Signal. 5:ra88. doi: 10.1126/scisignal.2003485



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

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

# The Relationship Between Vitamin B6, Diabetes and Cancer

Chiara Merigliano<sup>1</sup> , Elisa Mascolo<sup>1</sup> , Romina Burla<sup>1</sup> , Isabella Saggio1,2 \* and Fiammetta Vernì<sup>1</sup> \*

<sup>1</sup> Dipartimento di Biologia e Biotecnologie "C. Darwin," Sapienza Università di Roma, Rome, Italy, <sup>2</sup> Institute of Structural Biology, School of Biological Sciences, Nanyang Technological University, Singapore, Singapore

Pyridoxal 5<sup>0</sup> -phosphate (PLP), the active form of vitamin B6, works as cofactor in numerous enzymatic reactions and it behaves as antioxidant molecule. PLP deficiency has been associated to many human pathologies including cancer and diabetes and the mechanism behind this connection is now becoming clearer. Inadequate intake of this vitamin increases the risk of many cancers; furthermore, PLP deprivation impairs insulin secretion in rats, whereas PLP supplementation prevents diabetic complications and improves gestational diabetes. Growing evidence shows that diabetes and cancer are correlated not only because they share same risk factors but also because diabetic patients have a higher risk of developing tumors, although the underlying mechanisms remain elusive. In this review, we will explore data obtained in Drosophila revealing the existence of a connection between vitamin B6, DNA damage and diabetes, as flies in the past decade turned out to be a promising model also for metabolic diseases including diabetes. We will focus on recent studies that revealed a specific role for PLP in maintaining chromosome integrity and glucose homeostasis, and we will show that these aspects are correlated. In addition, we will discuss recent data identifying PLP as a putative linking factor between diabetes and cancer.

#### Edited by:

Maria Grazia Giansanti, Istituto di Biologia e Patologia Molecolari (IBPM), Consiglio Nazionale Delle Ricerche (CNR), Italy

#### Reviewed by:

Roberto Contestabile, Università degli Studi di Roma La Sapienza, Italy Fabrizio Barbetti, Università degli Studi di Roma Tor Vergata, Italy

#### \*Correspondence:

Isabella Saggio Isabella.saggio@Uniroma1.it Fiammetta Vernì Fiammetta.verni@Uniroma1.it

#### Specialty section:

This article was submitted to Genetic Disorders, a section of the journal Frontiers in Genetics

Received: 12 July 2018 Accepted: 29 August 2018 Published: 13 September 2018

#### Citation:

Merigliano C, Mascolo E, Burla R, Saggio I and Vernì F (2018) The Relationship Between Vitamin B6, Diabetes and Cancer. Front. Genet. 9:388. doi: 10.3389/fgene.2018.00388 Keywords: vitamin B6, pyridoxal 5<sup>0</sup> - phosphate, diabetes, chromosome aberrations, Drosophila, AGEs

### VITAMIN B6

The biologically active form of the vitamin B6, the pyridoxal 5<sup>0</sup> -phosphate (PLP), acts as coenzyme in about 160 distinct enzymatic activities mainly involved in amino acid, carbohydrate and lipid metabolism, and plays key roles in the synthesis and/or catabolism of certain neurotransmitters (Percudani and Peracchi, 2003; di Salvo et al., 2011). In addition, PLP works as antioxidant molecule by quenching oxygen reactive species (ROS) (Ehrenshaft et al., 1999) and counteracting the formation of Advanced Glycation End products (AGEs), genotoxic compounds associated with senescence and diabetes (Booth et al., 1997). Mammals, differently from microorganisms, are not able to synthesize PLP but they recycle it through a salvage pathway from B6 vitamers as pyridoxal (PL), pyridoxamine (PM), and pyridoxine (PN) contained in food (McCormick and Chen, 1999). In the cytoplasm PL, PM, and PN are converted into the 5<sup>0</sup> -phosphorylated vitamers by pyridoxal kinase (PDXK), while the FMN-dependent pyridoxine 5<sup>0</sup> -phosphate oxidase (PNPO) converts PNP and PMP into PLP.

Deficiency of vitamin B6 has been implicated in several clinically relevant diseases including autism, schizophrenia, Alzheimer, Parkinson, epilepsy, Down's syndrome, diabetes, and cancer.

In this review we focus on the role of PLP in diabetes and cancer suggesting more specialist readings for the other pathologies (Hellmann and Mooney, 2010; di Salvo et al., 2012).

#### VITAMIN B6 IN CANCER AND DIABETES

Epidemiological studies and meta-analysis indicate an inverse correlation between vitamin B6 and cancer development. For example, high expression levels of PDXK have been positively correlated with survival of non-small cell lung cancer (NSCLC) patients (Galluzzi et al., 2012). Furthermore, vitamin B6 intake and blood PLP levels were inversely correlated with the colorectal cancer risk (Gylling et al., 2017). PLP has been proposed to influence carcinogenesis through different pathways including those involved in DNA metabolism, suggesting that antitumor properties of vitamin B6 may be in part due to its protective role against DNA damage (Ames and Wakimoto, 2002). Vitamin B6 has also been associated to diabetes. However, it is not clear whether low PLP levels represent a cause or an effect of diabetes or both. Some studies report that low PLP levels can contribute to cause diabetes (Toyota et al., 1981; Rubi, 2012), whereas others show that diabetes decreases PLP levels (Bennink and Schreurs, 1975; Spellacy et al., 1977; Okada et al., 1999). Several groups reported that B6 administration produces beneficial effects on diabetic pathology and its complications (Cohen et al., 1984; Solomon and Cohen, 1989; Ellis et al., 1991; Hayakawa and Shibata, 1991; Jain, 2007), although underlying cellular and molecular mechanisms are not completely understood.

Pyridoxal 5<sup>0</sup> -phosphate deficiency might impact on diabetes in different ways. For example, it could act on the pathway that converts tryptophan into nicotinic acid as PLP is a cofactor of some enzymes that work in this pathway (Bennink and Schreurs, 1975; Spellacy et al., 1977; Oxenkrug, 2013). It has been shown that metabolites produced when this pathway does not work properly can interfere with biological insulin activity (Kotake et al., 1975) causing insulin resistance, a hallmark of type 2 diabetes. Moreover, it has also been proposed that PLP may impact on insulin resistance by controlling the expression of genes involved in adipogenesis (Moreno-Navarrete et al., 2016). Another hypothesis is that PLP deficiency might cause insulin resistance through an increase of homocysteine due to impairment of enzymes such as cystathionine-β-synthase (CBS) and cystathionine-γ-lyase (CGL), which require PLP as a coenzyme (Liu et al., 2016).

Cancer and diabetes are correlated as they share some risk factors. Growing evidence shows that diabetic patients have an increased risk to develop some malignance throughout multiple, not fully elucidate mechanisms, including DNA damage (Noto et al., 2011; Dankner et al., 2016). Interpret the cause-effect relationships in humans is difficult for unavailability of controls and high costs of human research. Researches in the field thus rely on model organisms as well as human 3D cultures and stem cell based systems (Riminucci et al., 2006; Simao et al., 2016). In this review we show how Drosophila has turned out to be a useful model not only to investigate the role of vitamin B6 in cancer and in diabetes but also to connect these two pathologies. Furthermore, we present evidence from flies suggesting that incorrect PLP intake could represent a cancer risk factor for diabetic patients, as it enhances DNA damage.

## PLP SAFEGUARDS GENOME INTEGRITY IN Drosophila

Using Drosophila as a model system we demonstrated that PLP plays a crucial role in genome integrity maintenance (Marzio et al., 2014). Drosophila dPdxk gene encodes the ortholog of the PDXK enzyme, which is required for vitamin B6 biosynthesis. Mutations in dPdxk gene produce, in larval neuroblasts, chromosome aberrations (CABs) (∼6 vs. 0.5% in controls), which are fully rescued by PLP. Dividing larval neuroblasts represent a suitable system to study CABs in Drosophila as they exhibit morphologically well defined chromosomes that can be stained by a variety of procedures (Gatti and Goldberg, 1991). CABs have been previously observed after X-ray treatment (Gatti et al., 1974) and as a consequence of mutations in genes which control chromatin structure (Mengoli et al., 2014) or different steps of DNA repair (Bianchi et al., 2017; Merigliano et al., 2017). Vitamin B6 antagonists, namely 4-deoxypyridoxine hydrochloride (4- DP), penicillamine, cycloserine, or isoniazid, produce high CAB frequencies (ranging from 3 to 19%) in wild type cells, further confirming that PLP plays an essential role in genome integrity maintenance. The aforementioned function is evolutionarily conserved in humans as the depletion of the human PDXK counterpart induces CABs and a copy of the human PDXK gene inserted in a Drosophila dPdxk<sup>1</sup> background is capable to rescue CABs (Marzio et al., 2014). Also in Saccharomyces cerevisiae mutations in BUD16 gene, encoding PDXK, result in gross chromosomal rearrangements. Altogether these data support the hypothesis that low PLP levels may promote cancer initiation and progression throughout the formation of CABs, which represent a cancer prerequisite (Mitelman et al., 2007; Aguilera and Gomez-Gonzalez, 2008; Bunting and Nussenzweig, 2013).

### DNA DAMAGE IS CAUSED BY HIGH GLUCOSE LEVELS IN PLP DEFICIENT CELLS

We obtained evidence that in Drosophila PLP is involved in glucose metabolism as dPdxk<sup>1</sup> mutants display, in their hemolymph, higher glucose concentrations compared to wild type individuals. dPdxk<sup>1</sup> mutants have normal insulin levels, but a weakened ability to respond to insulin signaling (Marzio et al., 2014). Remarkably, in dPdxk<sup>1</sup> mutants, high glucose levels and CABs are correlated. In dPdxk<sup>1</sup> mutant brains, indeed, 1% glucose in vitro treatment strongly increases CAB frequency (from 6 to 20%); in contrast sugar treatment of wild type larvae and brains did not result in detectable effects on chromosome integrity. The relationship between glucose and CABs, in PLP depleted cells, is evolutionarily conserved as glucose supplementation enhances chromosome damage also in

PDXK depleted HeLa cells (Marzio et al., 2014). In addition, the wild type PDXK human gene, inserted in dPdxk<sup>1</sup> flies, is able to reduce hyperglycemia. Hyperglycemia triggers the formation of AGEs that in turn produces ROS, which are harmful for DNA. It has been shown that ROS, even at low levels, can cause DNA damage that further leads to DNA double strands breaks (DSBs) (Sharma et al., 2016) and that CABs are mainly generated by unrepaired or improperly repaired DSBs. Repairing complex DSBs may result in genomic instability that can be involved in the etiology of a wide variety of human diseases including cancer (Khanna and Jackson, 2001; Kasparek and Humphrey, 2011).

dPdxk<sup>1</sup> mutant cells accumulate AGEs and treatment of dPdxk<sup>1</sup> mutants with alpha lipoic acid (ALA), a known AGE inhibitor, rescues not only AGEs but also CABs, suggesting that PLP protects from DNA damage Drosophila cells by counteracting AGE formation (Marzio et al., 2014). To the best of our knowledge only our work (Marzio et al., 2014) showed the cause effect relationship between AGEs and CABs in flies. However, Drosophila represents a good model to study AGEs as flies accumulate significant AGEs over their lifespan (Oudes et al., 1998) and, in addition, an AGE-rich diet results in ROS accumulation (Tsakiri et al., 2013). AGE formation is at the basis of many diabetic complications (Thorpe and Baynes, 1996; Brownlee, 2001; Vlassara and Palace, 2002) and it also can contribute to diabetes onset (Vlassara and Uribarri, 2014). Our data are consistent with studies indicating that vitamin B6 is beneficial for diabetes complication as, for example, nephropathy (Hayakawa and Shibata, 1991) and retinopathy (Ellis et al., 1991) and with in vivo studies showing that PLP is able to reduce AGE accumulation and protein glycation (Cohen et al., 1984; Solomon and Cohen, 1989). How PLP counteracts AGEs is not completely understood but it has been proposed that it may trap 3 deoxyglucosone (3-DG), an AGE's metabolism intermediate (Nakamura et al., 2007), although other mechanisms are possible. Besides to its antioxidant role, PLP also works as cofactor for serine hydroxymethyltransferase enzyme, which takes part to the thymidylate synthase cycle by converting dUMP in dTMP (Florio et al., 2011). However, whereas PLP depletion in yeast compromises DNA synthesis (Kanellis et al., 2007), in Drosophila it does not seems to have the same effect. Although in dPdxk<sup>1</sup> mutants there is an altered dTMP/dUTP ratio DNA syntesis is not the main cause of CABs as dPdxk<sup>1</sup> mutant are only slighty sensitive to Hydroxyurea, a drug that interferes with replication (Marzio et al., 2014). However, considering the wide range of enzymatic reactions regulated by vitamin B6, we cannot exclude that in addition to block AGE formation PLP may prevent CABs also through other mechanisms.

# Drosophila AS TYPE 2 DIABETES MODEL

Drosophila represents a good model to study diabetes as flies and humans largely share mechanisms involved in glucose homeostasis maintenance (Graham and Pick, 2017). In addition, fly genome possess well characterized orthologs of most genes working in the insulin signaling pathway that controls the glucose uptake and storage (Garofalo, 2002).

In humans and mice mutations in insulin pathway genes cause severe insulin resistance syndromes and type 2 diabetes (reviewed in Boucher et al., 2014). In Drosophila type 2 diabetes models can be generated by two different strategies: by downregulating conserved genes working on insulin pathways as for example the insulin receptor InR, the insulin substrate receptor chico/IRS1, Akt1, PI3K, and by feeding flies with a high sugar rich diet (Alfa and Kim, 2016). In both cases resulting flies exhibit diabetic hallmarks as hyperglycemia and insulin resistance allowing the study in flies of various aspects of diabetes and related human disorders. In addition, a diabetic fly model also enhances the ability to identify genes and discover functional interactions that can be exploited for disease treatment.

# PLP DEPLETION AS NEW CANCER RISK FACTOR IN DIABETIC CELLS

Meta-analysis and epidemiological studies indicate that diabetic patients have an increased risk to develop several solid and hematologic malignancies (including liver, pancreas, colorectal, kidney, bladder, endometrial, and breast cancers, and non-Hodgkin's lymphoma) although the molecular mechanisms are not completely clarified (Vigneri, 2009; Noto et al., 2011; Dankner et al., 2016). However, some risk factors have been identified including hyperinsulinemia and hyperglycemia that might rise cancer risk in diabetic patients by promoting cell growth (Shikata et al., 2013). Besides triggering cell division hyperglycemia also causes oxidative stress as glucose in excess promotes, through different pathways, ROS formation which in turn induces DNA and cellular damage (Rains and Jain, 2011). In addition, in cells from diabetic patients an impaired DNA repair, combined to a weakened antioxidant defense, contributes to enhance DNA damage (Blasiak et al., 2004). Consistently, oxidative damage and DNA strand breaks have been found in both type 1 and type 2 diabetic patients (Goodarzi et al., 2010; Tatsch et al., 2012; Anand et al., 2014). We have recently shown in Drosophila that PLP deficiency can further increase DNA damage in cells from diabetic individuals (Merigliano et al., 2018). Using two different type 2 diabetes models, the first obtained by downregulating genes involved in insulin signaling such as InR, chico (IRS1), and Akt1, and the second by feeding wild type flies with a high sugar diet (Musselman et al., 2011), we showed that the treatment of larval neuroblasts with the strong PLP inhibitor 4- DP produced a very high CAB frequency ranging from 60 to 80% (vs. 25% in wild type cells). Accordingly, genetic analysis revealed a synergistic interaction between Akt1 and dPdkx<sup>1</sup> mutations in CAB formation (Merigliano et al., 2018). AGEs are in part responsible for CABs in Drosophila diabetic PLP depleted cells as they accumulate in these cells and, more strikingly, ALA rescues either AGEs and CABs (Merigliano et al., 2018). These findings indicate that, in diabetic cells, low PLP levels heavily impact on genome integrity. Thus, if translated to humans, these data suggest that low PLP levels may contribute to increase cancer risk in diabetic patients. Although PLP deficiency is a rare condition caused by excessive alcohol consumption, unwanted effects of some drugs (i.e., isonyazide, cycloserine penicillamine), or

celiac disease and renal dialysis (Clayton, 2006), it has been demonstrated, in murine models and by epidemiological studies, that it can also be associated to diabetes (Leklem and Hollenbeck, 1990; Okada et al., 1999; Ahn et al., 2011; Nix et al., 2015). All evidence suggests the importance to maintain under strict control PLP levels in diabetic patients to avoid the chance to increase DNA damage, which could in turn contribute to cancer initiation and progression.

### CONCLUSION

Several studies have shown that insufficient intake of vitamin B6 is associated with increased cancer risk and growing evidence indicates that diabetes patients have a higher risk of developing various types of cancer. The findings reviewed here, obtained in Drosophila, provide a mechanistic link between aforementioned studies by suggesting that PLP deficiency accompanied by hyperglycemia can lead to DNA damage and may contribute to

### REFERENCES


cancerogenesis. Thus, Drosophila has proved to be a useful model system to shed light on a novel and important role of vitamin B6 deficiency in the pathogenesis of cancer and diabetes. In addition, this model organism allowed identifying PLP deficiency as one of the risk factors that contribute to correlating diabetes to cancer.

## AUTHOR CONTRIBUTIONS

CM contributed to planning and writing the manuscript. EM and RB contributed to planning the study. IS contributed to writing the manuscript. FV planned the study and wrote the manuscript. All authors performed a critical revision of the manuscript.

# FUNDING

The work was supported by MIUR Ateneo to FV and MIUR Avvio alla ricerca to CM.

of 2.3 Million adults. Am. J. Epidemiol. 183, 1098–1106. doi: 10.1093/aje/k wv290



**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.

The reviewer RC declared a shared affiliation, with no collaboration, with the authors to the handling Editor at time of review.

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

# Processing of DNA Ends in the Maintenance of Genome Stability

Diego Bonetti, Chiara Vittoria Colombo, Michela Clerici and Maria Pia Longhese\*

Dipartimento di Biotecnologie e Bioscienze, Università di Milano-Bicocca, Milan, Italy

DNA double-strand breaks (DSBs) are particularly hazardous lesions as their inappropriate repair can result in chromosome rearrangements, an important driving force of tumorigenesis. DSBs can be repaired by end joining mechanisms or by homologous recombination (HR). HR requires the action of several nucleases that preferentially remove the 5<sup>0</sup> -terminated strands at both DSB ends in a process called DNA end resection. The same nucleases are also involved in the processing of replication fork structures. Much of our understanding of these pathways has come from studies in the model organism Saccharomyces cerevisiae. Here, we review the current knowledge of the mechanism of resection at DNA DSBs and replication forks.

#### Edited by:

Maria Grazia Giansanti, Istituto di Biologia e Patologia Molecolari (IBPM), Consiglio Nazionale Delle Ricerche (CNR), Italy

#### Reviewed by:

Simonetta Piatti, Délégation Languedoc Roussillon (CNRS), France Marcus Smolka, Cornell University, United States

> \*Correspondence: Maria Pia Longhese mariapia.longhese@unimib.it

#### Specialty section:

This article was submitted to Genetic Disorders, a section of the journal Frontiers in Genetics

Received: 29 June 2018 Accepted: 29 August 2018 Published: 12 September 2018

#### Citation:

Bonetti D, Colombo CV, Clerici M and Longhese MP (2018) Processing of DNA Ends in the Maintenance of Genome Stability. Front. Genet. 9:390. doi: 10.3389/fgene.2018.00390 Keywords: checkpoint, DNA replication, double-strand break, MRX, nucleases, resection

# INTRODUCTION

DNA double-strand breaks (DSBs) are highly cytotoxic forms of DNA damage because their incorrect repair or failure to repair causes chromosome loss and rearrangements that can lead to cell death or transformation (Liu et al., 2012). They can form accidentally during normal cell metabolism or after exposure of cells to ionizing radiations or chemotherapeutic drugs. In addition, DSBs are intermediates in programmed recombination events in eukaryotic cells. Indeed, defects in DSB signaling or repair are associated with developmental, immunological and neurological disorders, and tumorigenesis (O'Driscoll, 2012).

Conserved pathways extensively studied in recent years are devoted to repair DSBs in eukaryotes. The two predominant repair mechanisms are non-homologous end joining (NHEJ) and homologous recombination (HR) and the choice between them is regulated during the cell cycle. NHEJ allows a direct ligation of the DNA ends with very little or no complementary base pairing and it operates predominantly in the G1 phase of the cell cycle (Chiruvella et al., 2013). The initial step involves the binding to DNA ends of the Ku heterodimer, which protects the DNA ends from degradation, followed by ligation of the broken DNA ends by the DNA ligase IV (Dnl4/Lig4 in yeast) complex. By contrast, HR is the predominant repair pathway in the S and G2 phases of the cell cycle and it requires a homologous duplex DNA to direct the repair (Mehta and Haber, 2014). For HR to occur, the 5<sup>0</sup> -terminated DNA strands on either side of the DSB must first be degraded by a concerted action of nucleases to generate 3<sup>0</sup> -ended single-stranded DNA (ssDNA) tails in a process referred to as resection (Cejka, 2015; Symington, 2016). These tails are first bound by the ssDNA binding complex Replication Protein A (RPA). RPA is then replaced by the recombination protein Rad51 to form a right-handed helical filament that is used to search and invade the homologous duplex DNA (Mehta and Haber, 2014).

Double-strand break occurrence also triggers the activation of a sophisticated highly conserved pathway, called DNA damage checkpoint, which couples DSB repair with cell cycle progression

**164**

(Gobbini et al., 2013; Villa et al., 2016). Apical checkpoint proteins include phosphatidylinositol 3-kinase related protein kinases, such as mammalian ATM (Ataxia-Telangiectasia-Mutated) and ATR (ATM- and Rad3-related), orthologs of Saccharomyces cerevisiae Tel1 and Mec1, respectively (Ciccia and Elledge, 2010). Once Mec1/ATR and/or Tel1/ATM are activated, their checkpoint signals are propagated through the S. cerevisiae protein kinases Rad53 and Chk1 (CHK2 and CHK1 in mammals, respectively), whose activation requires the conserved protein Rad9 (53BP1 in mammals) (Sweeney et al., 2005). While Tel1/ATM recognizes unprocessed or minimally processed DSBs, Mec1/ATR is recruited to and activated by RPAcoated ssDNA, which arises upon resection of the DSB ends (Zou and Elledge, 2003).

Most of our knowledge of the nucleolytic activities responsible for DSB resection has come from studies in the budding yeast S. cerevisiae, where DNA end resection can be monitored physically at sites of endonuclease-induced DSBs. Interestingly, the same nucleases involved in DSB resection are also responsible for the processing of stalled replication forks both in yeast and in mammals. Here we will focus on the work done in S. cerevisiae to understand the resection mechanism at DNA DSBs and replication forks and its regulation by Tel1/ATM and Mec1/ATR checkpoint kinases.

#### NUCLEASE ACTION AT DNA DOUBLE-STRAND BREAKS

Genetic studies in S. cerevisiae identified at least three distinct nucleases involved in end-resection: the MRX (Mre11-Rad50- Xrs2 in yeast; MRE11-RAD50-NBS1 in mammals) complex, Dna2 and Exo1 (DNA2 and EXO1 in mammals, respectively). In particular, the Mre11 subunit of MRX has five conserved phosphoesterase motifs in the amino-terminal half of the protein that are required for 30–5<sup>0</sup> double-strand DNA (dsDNA) exonuclease and ssDNA endonuclease activities of the protein in vitro (Bressan et al., 1998; Paull and Gellert, 1998; Trujillo et al., 1998; Usui et al., 1998). Rad50 is characterized by Walker A and B ATP binding cassettes located at the amino- and carboxyterminal regions of the protein, with the intervening sequence forming a long antiparallel coiled-coil. The apex of the coiledcoil domain can interact with other MRX complexes by Zn+ mediated dimerization to tether the bound DNA ends together (de Jager et al., 2001; Hopfner et al., 2002; Wiltzius et al., 2005; Williams et al., 2008). The ATP-bound state of Rad50 inhibits the Mre11 nuclease activity by masking the active site of Mre11 from contacting DNA (Lim et al., 2011). ATP hydrolysis induces conformational changes of both Rad50 and Mre11 that allow the Mre11 nuclease domain to access the DSB ends and to be engaged in DSB resection (Lammens et al., 2011; Lim et al., 2011; Williams et al., 2011; Möckel et al., 2012; Deshpande et al., 2014).

In the current model for resection, the Sae2 protein (CtIP in mammals) activates a latent dsDNA-specific endonuclease activity of Mre11 within the context of the MRX complex to incise the 5<sup>0</sup> -terminated dsDNA strands at both DNA ends (Cannavo and Cejka, 2014). The resulting nick generates an entry site for the Mre11 exonuclease to degrade back to the DSB end in the 30–5<sup>0</sup> direction, and for Exo1 and Dna2 nucleases to degrade DNA in the 50–3<sup>0</sup> direction away from the DSB end (Mimitou and Symington, 2008; Zhu et al., 2008; Cejka et al., 2010; Niu et al., 2010; Garcia et al., 2011; Nimonkar et al., 2011; Shibata et al., 2014; Reginato et al., 2017; Wang et al., 2017; **Figure 1**). In yeast, inactivation of either Sgs1-Dna2 or Exo1 results in only minor resection defects, whereas resection is severely compromised when the two pathways are simultaneously inactivated, indicating that they play partially overlapping functions (Mimitou and Symington, 2008; Zhu et al., 2008).

The efficiency of 5<sup>0</sup> DNA end cleavage in vitro by MRX-Sae2 was shown to be strongly enhanced by the presence of protein blocks at DNA ends (Cannavo and Cejka, 2014; Anand et al., 2016; Deshpande et al., 2016). It has been proposed that the endonucleolytic cleavage catalyzed by MRX-Sae2 allows the resection machinery to bypass end-binding factors that can be present at the break end and restrict the accessibility of DNA ends to Exo1 and Sgs1-Dna2. These end-binding factors includes Spo11, which cleaves DNA by a topoisomerase-like transesterase mechanism and remains covalently attached to the 5<sup>0</sup> end of meiotic DSBs, trapped topoisomerases, or the Ku complex (see the next paragraph) (Neale et al., 2005; Bonetti et al., 2010; Mimitou and Symington, 2010; Langerak et al., 2011; Chanut et al., 2016).

While Exo1 shows 50–3<sup>0</sup> exonuclease activity capable to release mononucleotide products from a dsDNA end (Tran et al., 2002), Dna2 has an endonuclease activity that can cleave either 3<sup>0</sup> or 5 <sup>0</sup> overhangs adjoining a duplex DNA (Kao et al., 2004). The resection activity of Dna2 relies on the RecQ helicase Sgs1 (BLM in humans) that provides the substrates for Dna2 by unwinding the dsDNA (Zhu et al., 2008; Cejka et al., 2010; Niu et al., 2010; Nimonkar et al., 2011). Furthermore, RPA directs the resection activity of Dna2 to the 5<sup>0</sup> strand by binding and protecting the 3<sup>0</sup> strand to Dna2 access (Cejka et al., 2010; Niu et al., 2010). In both yeast and humans, Dna2 contains also a helicase domain that can function as a ssDNA translocase to facilitate the degradation of 5<sup>0</sup> terminated DNA by the nuclease activity of the enzyme (Levikova et al., 2017; Miller et al., 2017).

In addition to the end-clipping function, the MRX complex also stimulates resection by Exo1 and Sgs1-Dna2 both in vitro and in vivo (Cejka et al., 2010; Nicolette et al., 2010; Niu et al., 2010; Shim et al., 2010; Nimonkar et al., 2011). Biochemical experiments have shown that MRX enhances the ability of Sgs1 to unwind dsDNA, possibly by increasing Sgs1 association to DNA ends. Furthermore, MRX enhances both the affinity to DNA ends and the processivity of Exo1 (Cejka et al., 2010; Nicolette et al., 2010; Niu et al., 2010; Nimonkar et al., 2011; Cannavo et al., 2013). The MRX function in promoting Sgs1-Dna2 and Exo1 resection activities does not require Mre11 nuclease, suggesting that it does involve the Mre11 end-clipping activity (Shim et al., 2010).

Interestingly, MRX possesses an ATP-dependent unwinding activity capable of releasing a short oligonucleotide from dsDNA (Paull and Gellert, 1999; Cannon et al., 2013) and the recent identification of the hypermorphic mre11-R10T mutation has

allowed us to demonstrate that this strand-separation function of MRX is important to stimulate Exo1 resection activity (Gobbini et al., 2018). In particular, Mre11-R10T mutant variant, whose single aminoacid substitution is located in the first Mre11 phosphodiesterase domain, accelerates DSB resection compared to wild type Mre11 by potentiating the processing activity of Exo1, whose association to DSBs is increased in mre11- R10T cells. Molecular dynamic simulations have shown that the two capping domains of wild type Mre11 dimer rapidly interact with the DNA ends and cause a partial unwinding of the dsDNA molecule. The Mre11-R10T dimer undergoes an abnormal rotation that leads one of the capping domain to wedge in between the two DNA strands and to persistently melt the dsDNA ends (Gobbini et al., 2018). These findings support a model in which MRX can directly stimulate Exo1 activity by promoting local unwinding of the DSB DNA end that facilitates Exo1 persistence on DNA. Although Exo1 is a processive nuclease in vitro, single-molecule fluorescence imaging has shown that it is rapidly stripped from DNA by RPA (Myler et al., 2016), suggesting that multiple cycles of Exo1 rebinding at the same DNA end would be required for extensive resection. Therefore, this MRX function in the stimulation of Exo1 activity at DNA ends can be of benefit to increase the processivity of Exo1 in the presence of RPA.

#### POSITIVE AND NEGATIVE REGULATION OF NUCLEASE ACTION AT DNA DOUBLE-STRAND BREAKS

Homologous recombination is generally restricted to the S and G2 phases of the cell cycle, when a sister chromatid is present as repair template (Aylon et al., 2004; Ira et al., 2004). This restriction is mainly caused by reduced end resection in G1 compared to G2. Reduced resection in G1 is due to both Ku binding to DNA ends and low cyclin-dependent kinase (Cdk1 in S. cerevisiae) activity (Aylon et al., 2004; Ira et al., 2004; Clerici et al., 2008; Zierhut and Diffley, 2008). Elimination of Ku in G1 (where Cdk1 activity is low) allows Cdk1-independent DSB resection that is limited to the break-proximal sequence, whereas the absence of Ku does not enhance DSB processing in G2 (where Cdk1 activity is high) (Clerici et al., 2008). Furthermore, inhibition of Cdk1 activity in G2 prevents DSB resection in wild type but not in ku1 cells (Clerici et al., 2008). These

findings suggest that Cdk1 activity is required for resection initiation when Ku is present. However, the finding that Cdk1 inhibition in G2-arrested ku1 cells allows only short but not long-range resection (Clerici et al., 2008) suggests the existence of other Cdk1 targets to allow extensive resection. Consistent with this hypothesis, Cdk1 was shown to promote short- and long-range resection by phosphorylating and activating Sae2 and Dna2, respectively. In fact, substitution of Cdk1-dependent phosphorylation residues in Sae2 causes a delay of DSB resection initiation, while mutations of Cdk1-target sites in Dna2 cause a defect in long-range resection (Huertas et al., 2008; Huertas and Jackson, 2009; Manfrini et al., 2010; Chen et al., 2011).

Subsequent experiments have shown that the Ku complex is rapidly recruited to DSBs and protects the DNA ends from degradation by Exo1 (**Figure 1**). The absence of Ku partially suppresses both the hypersensitivity to DSB-inducing agents and the resection defect of mre111 and sae21 cells in an Exo1 dependent fashion (Mimitou and Symington, 2010; Foster et al., 2011; Langerak et al., 2011). This finding suggests that Sae2, once phosphorylated by Cdk1, promotes resection initiation by supporting MRX function in removing Ku from the DSB ends. As Ku preferentially binds dsDNA ends over ssDNA (Griffith et al., 1992), the MRX-Sae2 endonucleolytic activity could limit DSB association of Ku by creating a DNA substrate less suitable for Ku engagement (Mimitou and Symington, 2010; Langerak et al., 2011; Chanut et al., 2016). On the other hand, as the absence of MRX, but not of Sae2 or Mre11 nuclease activity, increases Ku association at DNA ends (Zhang et al., 2007; Wu et al., 2008; Shim et al., 2010), MRX could compete with Ku for end binding. However, the finding that hyperactivation of Exo1 resection activity by the Mre11-R10T mutant variant leads to Ku dissociation from DSB ends and Cdk1-independent DSB resection close to the DSB end suggests that MRX can limit Ku association to DNA ends also indirectly by promoting Exo1 resection activity (Gobbini et al., 2018).

In addition to Ku, the Rad9 protein, originally identified as adaptor for activation of Rad53 checkpoint kinase (Sweeney et al., 2005), inhibits DSB resection (Bonetti et al., 2015; Ferrari et al., 2015; **Figure 1**). The lack of Rad9 suppresses the resection defect of Sae2-deficient cells and increases the resection efficiency also in a wild type context (Bonetti et al., 2015; Ferrari et al., 2015). Both these effects occur in a Sgs1-Dna2-dependent fashion, indicating that Rad9 inhibits mainly the resection activity of Sgs1-Dna2 by limiting Sgs1 association to DSBs. Further support for a role of Rad9 in Sgs1-Dna2 inhibition comes from the identification of the hypermorphic Sgs1-G1298R mutant variant, which potentiates the Dna2 resection activity by escaping the inhibition that Rad9 exerts on Sgs1 (Bonetti et al., 2015).

Recruitment of Rad9 to chromatin involves multiple pathways. The TUDOR domain of Rad9 interacts with histone H3 methylated at K79 (H3-K79me) (Giannattasio et al., 2005; Wysocki et al., 2005; Grenon et al., 2007). Rad9 binding to the sites of damage is strengthened through an interaction of its tandem-BRCT domain with histone H2A phosphorylated at S129 (γH2A) by Mec1 and Tel1 checkpoint kinases (Downs et al., 2000; Shroff et al., 2004; Toh et al., 2006; Hammet et al., 2007). Finally, phosphorylation of Rad9 by Cdk1 leads to Rad9 interaction with the multi-BRCT domain protein Dpb11 (TopBP1 in mammals), which mediates histone−independent Rad9 association to the sites of damage (Granata et al., 2010; Pfander and Diffley, 2011).

Rad9 association to DSB ends is counteracted by the Swr1 like family remodeler Fun30 (SMARCAD1 in mammals) (Chen et al., 2012; Costelloe et al., 2012; Eapen et al., 2012; Bantele et al., 2017) and the scaffold protein complex Slx4-Rtt107 (Dibitetto et al., 2016; Liu et al., 2017), both of which promote DSB resection (Chen et al., 2012; Dibitetto et al., 2016). The Slx4- Rtt107 complex limits Rad9 binding near a DSB possibly by competing with Rad9 for interaction with Dpb11 and γH2A (Ohouo et al., 2013; Dibitetto et al., 2016).

## DNA DAMAGE CHECKPOINT REGULATION OF NUCLEASE ACTION AT DOUBLE-STRAND BREAKS

Generation of DNA DSBs triggers the activation of the DNA damage checkpoint, whose key players include the S. cerevisiae protein kinases Mec1 (ATR in mammals) and Tel1 (ATM in mammals) (Gobbini et al., 2013). In both yeast and mammals, Tel1/ATM is activated by the MRX/MRN complex, which is required for Tel1/ATM recruitment to the site of damage through direct interaction between Tel1/ATM with the Xrs2 subunit (Nakada et al., 2003; Falck et al., 2005; Lee and Paull, 2005; You et al., 2005). By contrast, Mec1/ATR activation depends on its interactor Ddc2 (ATRIP in mammals) (Paciotti et al., 2000). While blunt or minimally processed DSB ends are preferential substrates for Tel1/ATM (Shiotani and Zou, 2009), RPA−coated ssDNA is the structure that enables Mec1/ATR to recognize DNA (Zou and Elledge, 2003). In both yeast and mammals, as the single−stranded 3<sup>0</sup> overhangs increase in length, Mec1/ATR activation is coupled with loss of ATM/Tel1 activation, suggesting that DSB resection promotes a switch from a Tel1/ATM− to a Mec1/ATR−dependent checkpoint (Mantiero et al., 2007; Shiotani and Zou, 2009; **Figure 2**). The substrates for Mec1 and Tel1 are largely overlapping and include H2A, Rad53/CHK2, Chk1, Rad9/53BP1, Sae2/CtIP, Dna2, and RPA (Ciccia and Elledge, 2010).

The DNA damage checkpoint regulates the generation of 3 <sup>0</sup>−ended ssDNA at DNA ends in both positive and negative fashions. Cells lacking Tel1 slightly reduce the efficiency of DSB resection (Mantiero et al., 2007). Tel1, which is loaded at DSBs by MRX, supports MRX persistence at DSBs in a positive feedback loop (Cassani et al., 2016), suggesting that it can facilitates DSB resection by promoting MRX function. Interestingly, Tel1 exerts this role independently of its kinase activity (Cassani et al., 2016), suggesting that it plays a structural role in stabilizing MRX retention to DSBs.

In contrast to tel11 cells, cells lacking Mec1 accelerate the generation of ssDNA at the DSBs, whereas the same process is impaired by the mec1-ad allele (Clerici et al., 2014), indicating that Mec1 inhibits DSB resection. Mec1 exerts this function at least in two ways: (i) it induces Rad53-dependent phosphorylation of Exo1 that leads to the inhibition of Exo1

phosphorylates Rad9, which in turn allows Rad53 in-trans autophosphorylation and activation (double black arrows). Activated Rad53 limits DSB resection by phosphorylating and inhibiting Exo1. Phosphorylation events by Mec1 and Tel1 are indicated by red dots, whereas green dots indicate phosphorylation events by Cdk1.

activity (Jia et al., 2004; Morin et al., 2008), (ii) it promotes retention of the resection inhibitor Rad9 at DNA DSBs through phosphorylation of H2A on serine 129 (Eapen et al., 2012; Clerici et al., 2014; Gobbini et al., 2015). The association of Rad9 at DSBs and therefore the inhibition of DSB resection is promoted also by the checkpoint sliding clamp Ddc1-Mec3-Rad17 (9-1-1 in mammals) complex (Ngo and Lydall, 2015), which is required for full Mec1 activation and binds to the ssDNA-dsDNA junction at DNA ends (Gobbini et al., 2013).

Both Mec1 and 9-1-1 have also a positive role in DSB resection. In fact, Mec1 is known to phosphorylate Sae2 and this phosphorylation is important for Sae2 function in resection of both mitotic and meiotic DSBs (Baroni et al., 2004; Cartagena-Lirola et al., 2006). Furthermore, Mec1 also phosphorylates Slx4 and this phosphorylation favors DSB resection by promoting Dpb11-Slx4-Rtt107 complex formation that leads to a destabilization of Rad9 association at DSBs (Smolka et al., 2007; Ohouo et al., 2013; Dibitetto et al., 2016).

Finally, in the absence of Rad9, the 9-1-1 complex facilitates DSB resection by stimulating both Dna2-Sgs1 and Exo1 through an unknown mechanism (Ngo et al., 2014). This effect of 9-1- 1 is conserved, as also the human 9-1-1 complex stimulates the activities of DNA2 and EXO1 in vitro (Ngo et al., 2014).

In yeast, the checkpoint response to DNA DSBs depends primarily on Mec1. However, if resection initiation is delayed, for example, in the sae21 mutant, MRX persistence at DSBs is increased, Tel1 is hyperactivated and the mec11 checkpoint defect is partially bypassed (Usui et al., 2001; Clerici et al., 2006). This persistent checkpoint activation caused by enhanced MRX and Tel1 signaling activity at DSBs contributes to the DNA

damage hypersensitivity and the resection defect of Sae2-deficient cells by increasing Rad9 persistence at DSBs. In fact, mre11 mutant alleles that reduce MRX binding to DSBs restore DNA damage resistance and resection in sae21 cells (Chen et al., 2015; Puddu et al., 2015; Cassani et al., 2018). Furthermore, reduction in Tel1 binding to DNA ends or abrogation of its kinase activity restores DNA damage resistance in sae21 cells (Gobbini et al., 2015). Similarly, impairment of Rad53 activity either by affecting its interaction with Rad9 or by abolishing its kinase activity suppresses the sensitivity to DNA damaging agents and the resection defect of sae21 cells (Gobbini et al., 2015). The bypass of Sae2 function by Rad53 and Tel1 impairment is due to decreased amount of Rad9 bound at the DSBs (Gobbini et al., 2015). As Rad9 inhibits Sgs1-Dna2 (Bonetti et al., 2015; Ferrari et al., 2015), reduced Rad9 association at DSBs increases the resection efficiency by relieving Sgs1-Dna2 inhibition.

Altogether, these findings support a model whereby the binding of MRX to DNA ends drives the recruitment of Tel1, which facilitates initiation of end resection by phosphorylating Sae2 and promoting MRX association to DNA ends (**Figure 2**). Generation of RPA-coated ssDNA leads to the recruitment of Mec1-Ddc2, which in turn phosphorylates Rad9, Rad53, and H2A. γH2A generation promotes the enrichment of Rad9 to the DSB ends, which limits the resection activity of Dna2-Sgs1. Rad9 association at DSBs also leads to the inhibition of Exo1 activity indirectly by allowing activation of Rad53, which in turn phosphorylates and inhibits Exo1 (**Figure 2**).

This Mec1-mediated inhibition of nuclease action at DSBs avoids excessive generation of ssDNA, which can form secondary structures that can be attacked by structureselective endonucleases, leading to chromosome fragmentation. Furthermore, since Mec1 is activated by RPA-coated ssDNA, inhibition of end resection by Mec1 keeps under control Mec1 itself. This negative feedback loop may avoid excessive checkpoint activation to ensure a rapid checkpoint turning off to either resume cell cycle progression when the DSB is repaired or adapt to DSBs as a final attempt at survival after cells have exhausted repair options.

#### NUCLEASE ACTION AT THE REPLICATION FORKS

Accurate and complete DNA replication is essential for the maintenance of genome stability. However, progression of replication forks is constantly challenged by various types of replication stress that generally causes a slowing or stalling of replication forks (Giannattasio and Branzei, 2017; Pasero and Vindigni, 2017). Replication forks can slow or stall at sites containing DNA lesions, chromatin compaction, DNA secondary structures (G-quadruplex, small inverted repeats, trinucleotides repeats), DNA/RNA hybrids and covalent protein-DNA adducts. Furthermore, clashes between transcription and replication machineries can impact genome stability even in unchallenged conditions (Giannattasio and Branzei, 2017; Pasero and Vindigni, 2017). Fork obstacles may result in dysfunctional replication forks, which lack their replication-competent state and necessitate additional mechanisms to resume DNA synthesis. Failure to resume DNA synthesis results in the generation of DNA DSBs, a major source of the genome rearrangements (Liu et al., 2012).

A general feature of stalled replication forks is the accumulation of ssDNA that can originate from physical uncoupling between the polymerase and the replicative helicase or between the leading and the lagging strand polymerases (Pagès and Fuchs, 2003; Byun et al., 2005; Lopes et al., 2006). The accumulation of torsional stress ahead of replication forks (Katou et al., 2003; Bermejo et al., 2011; Gan et al., 2017) can also lead to the annealing of the two newly synthesized strands and the formation of a four-way structure resembling a Holliday junction (i.e., fork reversal), which might expose DNA ends to exonucleolytic processing (Sogo et al., 2002). These tracts of ssDNA coated by the RPA complex recruit the checkpoint kinase Mec1/ATR (Zou and Elledge, 2003), whose activation prevents entry into mitosis, increases the intracellular dNTP pools, represses late origin firing, maintains replisome stability and orchestrates different pathways of replication fork restart/stabilization (Giannattasio and Branzei, 2017; Pasero and Vindigni, 2017).

In both yeast and mammals, the same nucleases involved in DSB resection are emerging as key factors for the processing of replication intermediates to allow repair/restart of stalled replication forks and/or to prevent accumulation of DSBs (Cotta-Ramusino et al., 2005; Segurado and Diffley, 2008; Tsang et al., 2014; Thangavel et al., 2015; Colosio et al., 2016). Indeed, the ability of Mre11, Sae2, Dna2, and Exo1 to resect dsDNA ends is relevant to prevent the accumulation of replicationassociated DSBs by promoting DSB repair by HR (Costanzo et al., 2001; Cotta-Ramusino et al., 2005; Segurado and Diffley, 2008; Hashimoto et al., 2012; Tsang et al., 2014; Yeo et al., 2014; Thangavel et al., 2015; Colosio et al., 2016; Ait Saada et al., 2017). In addition, controlled Dna2-mediated degradation of replication forks is a relevant mechanism to mediate reversed fork restart (Thangavel et al., 2015).

Although the nucleolytic processing of nascent strands at stalled replication forks is important to resume DNA synthesis, unrestricted nuclease access can also promote extensive and uncontrolled degradation of stalled replication intermediates and genome instability (Pasero and Vindigni, 2017). In budding yeast, the checkpoint activated by the ssDNA that arise at stalled replication forks plays a role in protecting replication intermediates from aberrant nuclease activity (Tercero and Diffley, 2001; Alabert et al., 2009; Barlow and Rothstein, 2009). In fact, in the absence of the checkpoint, relieve of Exo1 inhibition by Rad53 leads to the formation of long ssDNA gaps and fork collapse (Sogo et al., 2002; Cotta-Ramusino et al., 2005; Segurado and Diffley, 2008). Furthermore, replication stress in ATR-defective Schizosaccharomyces pombe and mammalian cells results in MRE11- and EXO1-dependent ssDNA accumulation (Hu et al., 2012; Koundrioukoff et al., 2013; Tsang et al., 2014). Interestingly, in S. cerevisiae, Tel1/ATM was recently found to counteract nucleolytic degradation by Mre11 of replication forks that reverse upon treatment with camptothecin (CPT) (Menin et al., 2018), which leads to accumulation of torsional stress by

blocking Top1 on DNA (Postow et al., 2001; Koster et al., 2007; Ray Chaudhuri et al., 2012). Fork reversal in CPT is promoted by the replisome component Mrc1, whose inactivation prevents fork reversal in both wild type and TEL1 deleted cells (Menin et al., 2018).

Interestingly, the same negative regulators of DSB resection limit nuclease action also at the replication forks. In yeast, Rad9, which is known to counteract the resection activity of Sgs1-Dna2, is important to protect stalled replication forks from detrimental Dna2-mediated degradation when Mec1/ATR is not fully functional (Villa et al., 2018). This Rad9 protective function relies mainly on the interaction of Rad9 with Dpb11, which is recruited to stalled replication forks at origin-proximal regions (Balint et al., 2015). Similarly, human cells lacking 53BP1, the mammalian Rad9 ortholog, are hypersensitive to DNA replication stress and show degradation of nascent replicated DNA (Her et al., 2018). Furthermore, the Ku heterodimer was shown to be recruited to terminally arrested replication forks and to regulate their resection in S. pombe (Teixeira-Silva et al., 2017). The lack of Ku leads to extensive Exo1-mediated fork resection, a reduced recruitment of RPA and Rad51 and a delay of forkrestart, suggesting that arrested replication forks undergo fork reversal that provides a substrate for Ku binding.

In addition to the checkpoint, other proteins are devoted to protect replication forks from degradation in mammalian cells. The absence of proteins involved in HR or in the Fanconi anemia network, including FAN1, FANCD2, RAD51, BRCA1, and BRCA2, leads to uncontrolled DNA degradation by MRE11 and EXO1 (Howlett et al., 2005; Hashimoto et al., 2010; Schlacher et al., 2011, 2012; Ying et al., 2012; Chaudhury et al., 2014; Karanja et al., 2014; Chen et al., 2016; Kais et al., 2016; Kolinjivadi et al., 2017; Lemaçon et al., 2017; Mijic et al., 2017; Taglialatela et al., 2017). Furthermore, loss of the WRN exonuclease activity enhances degradation at nascent DNA strands by EXO1 and MRE11 (Su et al., 2014; Iannascoli et al., 2015), whereas cells depleted of the biorientation defect 1-like (BOD1L) protein exhibit a DNA2-dependent degradation of stalled/damaged replication forks (Higgs et al., 2015).

#### REFERENCES


#### CONCLUSION

Defects in HR and DNA replication underlie a significant proportion of the genomic instability observed in cancer cells. Furthermore, ssDNA formed at DSBs and at replication forks can be source of clustered mutations, frequently occurring during carcinogenesis, and of error-prone repair events that can cause DNA deletions or translocations (Nik-Zainal et al., 2012; Roberts et al., 2012; Alexandrov et al., 2013; Sakofsky et al., 2014). Therefore, there is a growing interest in understanding how ssDNA is generated at both DSBs and replication forks and how its generation is regulated. Mounting evidence indicates that processing of both DSB ends and replication forks is regulated both positively and negatively by several proteins involved also in the DNA damage checkpoint, thus coupling resection with checkpoint activation. Given the importance to maintain genome stability, advancements in delineating the mechanisms that control nuclease action at both DSBs and replication forks will have far-reaching implications for human health.

#### AUTHOR CONTRIBUTIONS

MPL conceived the idea. DB and MPL wrote the manuscript. CVC and MC revised and edited the manuscript.

# FUNDING

This work was supported by the Associazione Italiana per la Ricerca sul Cancro (AIRC) (IG Grant 19783) and Progetti di Ricerca di Interesse Nazionale (PRIN) 2015 to MPL.

### ACKNOWLEDGMENTS

We thank all members of the Longhese lab for helpful discussions.



and checkpoint activation during the cell cycle. EMBO Rep. 9, 810–818. doi: 10.1038/embor.2008.121


Rad6-Bre1 and H3 methylation by Dot1. J. Biol. Chem. 280, 9879–9886. doi: 10.1074/jbc.M414453200


and promotes alternative end-joining DNA Repair. Cell Rep. 15, 2488–2499. doi: 10.1016/j.celrep.2016.05.031




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

Copyright © 2018 Bonetti, Colombo, Clerici and Longhese. 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