The Evolution of Cell Cycle Regulation, Cellular Differentiation, and Sexual Traits during the Evolution of Multicellularity

Item Type text; Electronic Dissertation

Authors Hanschen, Erik Richard

Publisher The University of Arizona.

Rights Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.

Download date 07/10/2021 00:03:04

Link to Item http://hdl.handle.net/10150/626165

THE EVOLUTION OF CELL CYCLE REGULATION, CELLULAR DIFFERENTIATION, AND SEXUAL TRAITS DURING THE EVOLUTION OF MULTICELLULARITY

by

Erik Richard Hanschen

______Copyright © Erik Richard Hanschen 2017

A Dissertation Submitted to the Faculty of the

DEPARTMENT OF ECOLOGY AND EVOLUTIONARY BIOLOGY

In Partial Fulfillment of the Requirements

For the Degree of

DOCTOR OF PHILOSOPHY

In the Graduate College

THE UNIVERSITY OF ARIZONA

2017

2

THE UNIVERSITY OF ARIZONA GRADUATE COLLEGE

As members of the Dissertation Committee, we certify that we have read the dissertation prepared by Erik R. Hanschen, titled “The evolution of cell cycle regulation, cellular differentiation, and sexual traits during the evolution of multicellularity” and recommend that it be accepted as fulfilling the dissertation requirement for the Degree of Doctor of Philosophy.

______Date: September 25, 2017 Richard E. Michod

______Date: September 25, 2017 Régis Ferrière

______Date: September 25, 2017 Michael Barker

______Date: September 25, 2017 Bradley J. S. C. Olson

Final approval and acceptance of this dissertation is contingent upon the candidate’s submission of the final copies of the dissertation to the Graduate College.

I hereby certify that I have read this dissertation prepared under my direction and recommend that it be accepted as fulfilling the dissertation requirement.

______Date: September 25, 2017 Dissertation Director: Richard E. Michod 3

STATEMENT BY AUTHOR

This dissertation has been submitted in partial fulfillment of the requirements for an advanced degree at the University of Arizona and is deposited in the University Library to be made available to borrowers under rules of the Library.

Brief quotations from this dissertation are allowable without special permission, provided that an accurate acknowledgement of the source is made. Requests for permission for extended quotation from or reproduction of this manuscript in whole or in part may be granted by the head of the major department or the Dean of the Graduate College when in his or her judgment the proposed use of the material is in the interests of scholarship. In all other instances, however, permission must be obtained from the author.

SIGNED: Erik Richard Hanschen

4

ACKNOWLEDGEMENTS

I would like to first thank my doctoral committee for their invaluable contributions to my research over the course of my time at the University of Arizona. Working for Rick Michod has taught me how to be an independent and creative scientist and how to take research projects from an initial idea through development and analysis to publication. Brad Olson has been a close and valuable collaborator who has contributed much to my development as a scientist as we analyzed volvocine genomes. Mike Barker has been an invaluable resource for genomic approaches, concepts, and ideas. More importantly, Mike Barker is a steadfast source of encouragement and cheerleading throughout the progress of research. Régis Ferrière has been invaluable to the direction of my dissertation research, expertly and brilliantly suggesting sweeping changes in research direction with a single question.

I greatly appreciate the community of scientists that have lead to this body of work, the Volvox community has been a wealth of ideas and suggestions as my research progresses. I would especially like to thank Aurora Nedelcu and Matthew Herron who have both been dear friends and sources of advice. I would like many others, including Dinah R. Davison, Deborah S. Shelton, Zach I. Grochau-Wright, Patrick J. Ferris, Tara N. Marriage, Hisayoshi Nozaki, John J. Wiens, Baltazar Chavez-Diaz Jr., David Shahnooshi, and Jeremiah Hackett.

Of course an entire community at UA EEB is due acknowledgement, including Sky Dominguez, LuAnn Casem, Lauren Harrison, Pennie Rabago, Barbara Johnson, Vincente Leon, Barry McCabe, Lilian Schwartz, Pat Waters, and Carol Freeman. Other valuable colleagues, among others, include Jen Wisecaver, William Driscoll, and Doug Cromey.

Throughout my PhD, a large number of friends and family have been critical to staying the course, whether through beer, advice, or getting outside together. This includes Tom Hanschen, Linnea Hanschen, Brendan Hanschen, Ryan Hanschen, Shannon Hanschen, Holden Hanschen, Hugo Hanschen, Sky Dominguez, Moira Hough, Tyeen Taylor, Marielle Smith, Brad Boyle, Rebekah Doyle, Alex Brummer, Jason Veatch, Justin Schaffer, Jonathan Mosher, Ryan Dunn, Katie Fisher, Connie Lam, Amanda Goldstein, Blake Heym, Pacifica Sommers, Neill Prohaska, and Ariana La Porte.

Sources of funding for this research include the NSF IGERT training grant “Program in Comparative Genomics”, the NIH training grant “Computational and Mathematical Modeling of Biological Sciences”, the National Institute of Genetics (Japan), and the UA Department of Ecology and Evolutionary Biology. I also received support from a NSF grant awarded to Brad Olson and Rick Michod and a NASA grant awarded to Rick Michod and Patrick Ferris. Other valuable sources of funding in the American Genetics Association, Graduate and Professional Student Council travel and Research and Project grants, the UA Galileo Circle Scholarship program, and Robert W. Hoshaw Memorial Scholarship, the Kavli Institute for Theoretical Physics, and the National Evolutionary Synthesis Center.

5

TABLE OF CONTENTS

ABSTRACT...... 6

INTRODUCTION...... 8

PRESENT STUDY...... 15

REFERENCES...... 24

APPENDIX A. EVOLUTIONARY TRANSITIONS IN INDIVIDUALITY AND RECENT MODELS OF MULTICELLULARITY...... 32

APPENDIX B. THE GONIUM PECTORALE GENOME DEMONSTRATES CO-OPTION OF CELL CYCLE REGULATION DURING THE EVOLUTION OF MULTICELLULARITY... 57

APPENDIX C. EARLY EVOLUTION OF THE GENETIC BASIS FOR SOMA IN THE VOLVOCACEAE...... 68

APPENDIX D. MULTICELLULARITY DRIVES THE EVOLUTION OF SEXUAL TRAITS...... 81

6

ABSTRACT

During the evolution of multicellularity from unicellular ancestors, cells transition from being evolutionary individuals to components of more complex, multicellular evolutionary individuals.

The volvocine green algae provide a powerful model system for understanding the genetic and morphological changes that underlie and are caused by the evolution of multicellularity. This dissertation concerns the role of cell cycle regulation, cellular differentiation, and sexual traits during the evolution of multicellularity. While some of these are shown to be causally important in the origins of multicellularity (Appendix B), others are driven by the evolution of multicellularity (Appendix D). We provide a review of recent mathematical models on the evolution of multicellularity, which are found to focus heavily on the later, subsequent stages of the evolution of multicellular complexity. We found that many of these models assume multicellular ancestors and instead evolve cellular differentiation, bringing attention to a gap in our understanding of the events in the initial stages of the evolution of multicellularity. We show that a focus on the early stages of the evolution of multicellularity reveals a powerful and critical role for regulation of the cell cycle at the origins of multicellularity (Appendix B). We further find that the genetic basis for cellular differentiation evolved sometime after the evolution of cell cycle regulation. We find that while the genetic basis for cellular differentiation evolved after cell cycle regulation, it also evolved earlier than previously predicted in the volvocine green algae, suggesting an important role in undifferentiated species (Appendix C). Lastly, having elucidated the origins and evolution of multicellularity, we find that multicellularity causes the evolution of sexual traits including anisogamy, internal fertilization, and subsequently sexual dimorphism (Appendix D). This work emphasizes the important role that multicellularity plays 7

in driving the evolution of sexual diversity seen across the eukaryotic tree and well as informs critical hypotheses on the evolution of anisogamous sex, among the most challenging problems in evolutionary theory.

8

INTRODUCTION

“The history of life is a history of transitions between different units of selection.”

–Leo W. Buss (1987 p. 171)

Context of the problem

Biological individuals are hierarchically nested such that evolutionary individuals are composed of groups of pre-existing individuals. This diversity of individuals includes genes, chromosomes, cells, cells made of cells (eukaryotes), multicellular organisms, and eusocial societies. Given these hierarchical levels and the knowledge that these hierarchical levels did not exist at the origin of life but rather have subsequently emerged throughout evolutionary time, the unit of selection, and individuality itself, must transition from the lower level to the higher level

(Maynard Smith and Szathmáry 1995; Szathmáry and Maynard Smith 1995; Michod 1999).

While multiple transitions between levels of individuality have occurred throughout the history of life on Earth, the evolution of multicellularity is a particularly well-studied transition in which the processes and critical steps may be tractably investigated (Bonner 1998; Grosberg and

Strathmann 2007). The evolution of multicellularity is also particularly interesting because, through multiple events, resulted in clades such as plants, animals, fungi, and brown algae, which currently dominate global ecosystems. The emergence of these clades has indisputably alter the course of evolutionary history, creating the novel niches and species interactions, which form the basis for much of life on Earth.

Given that biological entities which were previously evolutionary individuals (unicellular organisms in the example of the evolution of multicellularity) are no longer evolutionary individuals, but only a component of an evolutionary individual (multicellular organisms), it is 9

abundantly evident why the evolution of multicellularity is an interesting and concerning evolutionary question. How have biological entities lost evolutionary individuality and how has evolutionary individuality arisen at hierarchical levels in which it did not previously exist? The multiple independent evolutions of multicellularity across the tree of life provide a unique opportunity to answer not only this question, but other important questions regarding the nature of individuality. Are there generalities about why evolutionary multicellular individuals evolve and what biological forms of multicellularity are realized? Are there general biological properties that underlie the basis of multicellularity through eukaryotes? Are there common evolutionary consequences that multicellularity sets forth?

This dissertation is part of a beginning to answer the latter two questions, investigating the genetic bases and mechanisms underlying the evolution of multicellularity and investigating how multicellularity results in selection pressures which result in similarly subsequent evolution of multicellular phenotypes across the tree of life. In doing so, another other fundamental evolutionary question is informed, why have males evolved?

The evolution of males has been described as the hardest problem in evolutionary biology

(John Maynard Smith, 1999), as it attempts to address why anisogamy, the evolution of males which produce numerous, small sperm and females which produce quite few, large eggs.

Anisogamous sex is such an evolutionary conundrum because males do not contribute to reproduction; anisogamous sex results in half the number of offspring than either isogamous sex or asexual reproduction (Maynard Smith 1978). Given this twofold cost of anisogamous sex, why did males evolve? 10

While mathematical and empirical work has attempted to explain the evolution of anisogamous sex by investigating the benefits of sexual reproduction, largely through genetic recombination which confers evolutionary benefits when selection varies across time and space

(see Otto (2009) for a review), much of this work address the costs and benefits of sex, not the costs and benefits of anisogamous sex specifically. Previous theory does directly address the evolution of anisogamy (Parker et al. 1972; Bulmer and Parker 2002), predicting that the evolution of multicellularity actually causes the evolution of anisogamy. During sexual reproduction, a diploid zygote must provision the subsequent developing embryo. During the evolution of multicellularity, the cytoplasmic resources needed to provision the embryo increases as the embryo increases in size (the embryo increases in size to develop into an increasingly large multicellular organism), and this selection disrupts the isogamous evolutionary stable state

(Bulmer and Parker 2002). Then, disruptive selection, the result of a tradeoff between gamete quantity and quality (Parker et al. 1972), results in the evolution of males producing sperm and females producing eggs. Unfortunately, little research as directly tested the hypothesis that multicellularity should cause the evolution of anisogamy.

Introduction to the volvocine model system

In understanding the genetic basis of multicellularity and the evolutionary consequences of multicellularity, I utilize the volvocine green algal model system in this dissertation. The volvocine green algae were first described by Anthony van Leeuwenhoek (1700) and systematically described by Carl Linneaus (1758). Importantly, van Leeuwenhoek noted the ability of volvocine green algae to reproduce asexually, a critical component of Appendix D. 11

“Among the rest, I did observe one of the greatest round parts somewhat larger, in

a small quantity of water, before my eyes, and did perceive that the outward part

began to open, and that one of the round particles was within it, and was of a

delicate green colour, did slip out of it, and began to move in the water, as that

part had done whereout it did come.” –Anthony van Leeuwenhoek (1700 p. 511)

Other members of the volvocine green algae include unicellular species such as Chlamydomonas reinhardtii and Vitreochlamys aulata, undifferentiated colonial species of various sizes including

Gonium pectorale, Pandorina morum, and Eudorina elegans. Species with cellular differentiation are additionally extant, including species with somatic cellular differentiation only such as Pleodorina starrii and Pleodorina californica, and species with cellular specialization resulting in both somatic and germ cells such as Volvox carteri and Volvox ferrisii. Furthermore, the volvocine green algae display a large diversity of forms of sexual reproduction, from species with isogamous sexual reproduction (sexual meiotic gametes are equal sized, denoted (+) or (-)) where gametic fusion occurs in the environment (such as C. reinhardtii), to species with anisogamous sexual reproduction (sexual meiotic gametes are not equal sized, males produce smaller sperm and females produce larger eggs) where gametic fusion occurs within the body of a multicellular organism (such as V. carteri). Thus is it important to note that all known species are facultatively sexual.

Ecologically, volvocine green algae are motile and photosynthetic autotrophs, motility is important for maintaining position in the water column for photosynthesis (Reynolds 1983;

Solari et al. 2006a,b). Some volvocine algae are known to migrate huge distances to obtain nutrients, likely nitrogen and phosphorus, on a daily cycle, photosynthesizing during the day at 12

the top of the water column and obtaining nutrients near the sedimentary layer during night

(Sommer and Maciej Gliwicz 1986).

Most volvocine algae have a unique life history with regards to growth and cellular division, which proves to be important for the evolution of multicellularity. Rather than each cell alternating between growth and a single round of division, most volvocine algae undergo palintomy, where a long period of growth is followed by several rounds of division. While there are select volvocine algae which do not undergo palintomy (Volvox aureus, species in Volvox section Volvox), palintomy is present in most volvocine algae and appears ancestral (Desnitski

1995; Herron et al. 2010). Reproductive cells grow several times larger than their starting cell size, and undergo rapid, repeated rounds of cell division to produce many cells of the starting cell size (Kirk 1998). Thus volvocine species have cell numbers that are powers of two based on the number of cell divisions each reproductive cell undergoes to produce a new colony (i.e., 8, 16,

32, 64, etc.).

Lastly, the volvocine green algae are an excellent model system for the evolution of multicellularity because they have no multicellular ancestor (Leliaert et al. 2012), are recently, compared to animals or plants, diverged from unicellular relatives (Herron et al. 2009), and several genomes have been previously sequenced and multiple genetic tools, including transformation, are available (Kirk 1998; Merchant et al. 2007; Prochnik et al. 2010). For further reviews of the volvocine green algae see Coleman (2012), Kirk (1998), or Nozaki (2003).

Explanation of the dissertation format and nature of collaborative work 13

This dissertation includes four appendices, each formatted as an independent scientific

manuscript. Appendix A is a published book chapter (Hanschen et al. 2015) that reviews recent

mathematical models of the evolution of multicellularity and cellular differentiation. This

manuscript reviews the strengths and value of several mathematical models and identifies a lack

of understanding on the early stages of the evolution of multicellularity. Appendix B is a

published paper (Hanschen et al. 2016) which then investigates the early steps in the evolution of

multicellularity. This work uses a comparative evolutionary genomics approach to sequence and

analyze the genome of the undifferentiated colonial Gonium pectorale. This genome was

compared to the genome of unicellular Chlamydomonas reinhardtii and differentiated

multicellular Volvox carteri. This work illuminates the critical role of regulation of the cell cycle,

through modification of the retinoblastoma master tumor suppressor gene, in the evolution of

multicellularity. It is further demonstrated that the genetic basis for cellular differentiation had

not yet evolved in the common ancestor of G. pectorale and V. carteri. Appendix C is a

published paper (Hanschen et al. 2014) which investigates the origin of the genetic basis for

cellular differentiation in V. carteri, the regA gene cluster, finding this gene cluster in species

which independently evolved somatic cells and the typical Volvox morphology. This work

demonstrates the genetic basis for cellular differentiation evolved much earlier and required more transitional steps than previously predicted. Having shown how, genetically, multicellularity evolves, Appendix D is a manuscript currently in preparation which investigates the previous theoretical work predicting that multicellularity should cause the evolution of sexual traits such as anisogamy, the evolution of male and female sexes which produces sperm and eggs, respectively. These predictions are generalized to sexual traits, not just anisogamy, and a 14

comparative phylogenetic analysis demonstrates results strongly consistent with previous mathematical theory. The appended book chapter, papers, and manuscript are collaborative works, and my role was primary and corresponding author/researcher for all of them. In addition to writing these works, I led the genomic, genetic, and phylogenetic analysis portions of all the appendices.

15

PRESENT STUDY

The methods, results, and conclusions of this research are presented in the manuscript appended to this dissertation. As the supplemental material accompanying the accumulated manuscripts is both freely available on multiple independent repositories and of asinine length, it is not included here. This chapter is a brief summary of the research presented in the appendices and provides the conclusions of the dissertation research.

In Appendix A, we review a set of recent mathematical models on the evolution of multicellularity and cellular differentiation. Previously existing work is often framed in the context of the evolutionary transitions in individuality framework (ETI) which hypotheses that groups of individuals evolve into a new, higher-level individual through processes involving five factors: cooperation and conflict, life history trade-offs, multi-level selection, division of labor, and the decoupling of fitness at the level of the group from the level of the cell (Michod 1999,

2006, 2007; Hanschen et al. 2015). However, many of the recent mathematical models were relatively disconnected from the prevailing literature, including the ETI framework. After discussing how the components of recent mathematical work are consistent, or could be easily viewed as consistent, with core components of ETI theory, we identify three important future directions for research.

First, in order for mathematical models to provide relevant predictions, they must incorporate basic life history issues. Obviously, some generalization is necessary and desirable for a mathematical model; however, when a model forgoes basic aspects of the life history, it is difficult to evaluate a model’s predictions in specific cases. We found several cases of this in the 16

current literature. For example, Ispolatov & Doebeli (2011) model the evolution of multicellularity with cyanobacteria in mind. They model an aggregative process in which all cells reproduce. Given that multicellular cyanobacteria do not develop via aggregation, nor do all cells reproduce (Kumar et al. 2010; Muro-Pastor and Hess 2012), it is difficult to apply the predictions of Ispolatov & Doebeli (2011) even to the case they are considering. While the amount of life history details included in the model vary with the model’s intentions, some appreciation of life history is critical in order to build and apply the model.

Second, fitness components must be further connected to empirical work. Previous work

(Michod 2006, 2007; Michod et al. 2006) modeled cellular differentiation as the specialization on fitness components, viability and reproduction. While the relationship of germ cells to reproduction is easily tractable, this work was agnostic to how somatic cells influenced viability.

Recent models are also agnostic to how lower level units (e.g., cells) relate to fitness components. For example, this is seen in Van Dyken & Wade (Van Dyken and Wade 2012a,b) which assumed units of energy somehow relate to viability and fecundity as well as Rueffler et al. (Rueffler et al. 2012), which modeled units specializing on some tasks which traded off which each other.

Lastly, as much recent work as focused on understanding the later stages of the evolution of multicellularity, the evolution of cellular differentiation among them, we identified the early stages of multicellularity as lacking in both mathematical and empirical work, emphasizing the importance of continued work on the early stages of the evolution of multicellularity.

In Appendix B, we address the early stages of the evolution of multicellularity by sequencing the genome of the early branching undifferentiated colonial green alga, Gonium 17

pectorale. This species was chosen because it’s placement in the phylogenetic tree promised to illuminate the genetic changes, including gene family expansions and creation, which underlie the evolution of multicellularity. While the unicellular and multicellular genomes of the volvocine algae have already been published (Merchant et al. 2007; Prochnik et al. 2010), they were not framed in the context of evolutionary genomics, only comparative genomics, a concern addressed in Appendix B. The analysis of the genome of Volvox carteri identified important transcription factors (i.e., cyclins and regA (Duncan et al. 2007)) and cell wall genes

(pherophorins and metalloproteases) associated with the evolution of multicellularity.

Sequencing the genome of Gonium pectorale aims to further identify which genes are associated with the evolution of multicellularity, as well as the order in which they evolved. At the genomic level, there is very little genomic innovation, whether measured by protein content, gene family expansion and creation, functional protein domain innovation, de novo gene creation, signaling domain expansion and creation, genetic architectures, or small RNA molecule expansion, that correlates with the evolution of multicellularity, suggesting the early stages of multicellularity are underlied by relatively few changes in critical gene pathways.

We then utilized an a priori expectation of genes which may underlie multicellularity, focusing on cell wall genes, cell cycle regulation, and cellular differentiation. The modification to the cell wall genes and cellular differentiation is immediately apparent in comparing morphology of C. reinhardtii to V. carteri. Cell cycle regulation was investigated because while cell division is reproduction in unicellular species, cell division has been decoupled from reproduction and co-opted for development in multicellular species. Furthermore, previous work suggests a critical role of cell cycle regulation in the evolution of multicellularity (Shelton and 18

Michod 2014). Using detailed gene annotations, we have shown that cyclins and pherophorins are already expanded in the genome of G. pectorale. Of note, regA and the regA tandem duplication, which regulates somatic cells in V. carteri, is absent. Lastly, structural modification the retinoblastoma tumor suppressor and expansion of cell cycle cyclinD genes suggests an important role of cell cycle in the evolution of multicellularity, a prediction born out by genetic transformation. We transformed the retinoblastoma gene from G. pectorale into retinoblastoma deficient C. reinhardtii, demonstrating the formation of clumps in a unicellular background, thus causally linking a critical tumor suppressor gene to the origin of multicellularity. Furthermore, this result is similar to other genomic investigations of independent evolutions of multicellularity. In both green plants and metazoans, it appears most of the genes responsible for multicellularity are present surprisingly early in the phylogenetic tree (Abedin and King 2008;

Fairclough et al. 2010; Nichols et al. 2012; Ryan et al. 2013; Sebé-Pedrós et al. 2013; Suga et al.

2013; Hori et al. 2014).

Appendix C investigates the absence of the genetic basis for cellular differentiation, the regA gene cluster, in Gonium pectorale. Given the importance of the evolution of division of labor (i.e., cellular differentiation in the evolution of multicellularity, Michod 2006), understanding the evolutionary history of genes regulating somatic cells is critical to understanding the evolution of multicellularity. In order to understand the evolutionary history of the regA gene cluster in the volvocine algae, we have investigated the presence of regA in divergent species of Volvox (V. ferrisii and V. gigas). Using the presence of regA in other species of Volvox, particularly V. ferrisii, we argue that regA evolved early in the volvocine algae though after the Gonium lineage split from the Eudorina-Volvox clade. The ancestor in which regA 19

likely evolved is predicted to not have somatic cells (Herron and Michod 2008; Grochau-Wright et al. 2017), suggesting an alternative original selection pressure for regA, perhaps related to temporal regulation of cell division in a multicellular context (Grochau-Wright et al. 2017). This result is similar to the lessons learned from Appendix B, where the genetic basis for multicellularity largely evolved before the phenotypes thought to arise from these genes.

Because the common ancestor of V. ferrisii and V. carteri is predicted to have given rise to multiple lineages with and without somatic cells, this suggests a more complex relationship between somatic cells and fitness than previously appreciated. This research raises a number of questions, including under what evolutionary and ecological conditions would species with the genetic potential for soma not evolve somatic cells, and how many pathways are available for the volvocine green algae to regulate somatic cells.

Having discussed how multicellularity evolves in Appendices A, B, and C, Appendix D investigates how multicellularity affects the evolution of other traits, focusing on sexual traits.

Previous theory (Parker et al. 1972; Bulmer and Parker 2002) predicts, not only that multicellularity should correlate with the evolution of anisogamy, but that multicellularity causes the evolution of anisogamy, one of the great problems in evolutionary theory. This theory has several predictions, that (1) anisogamous species should be larger than isogamous species, (2) anisogamous species should produce larger zygotes than isogamous species, and (3) anisogamy should evolve in a multicellular ancestor. We have extended this prediction to other sexual traits

(including internal fertilization, polar bodies, and sexual dimorphism), generalizing theory to suggest multicellularity should drive the evolution of sexual traits. 20

After identifying relevant sexual traits and compiling trait information for up to 86 species, we performed statistically rigorous ancestral state reconstructions, demonstrating that derived sexual traits arose in a multicellular ancestor. Using phylogenetic t-tests, we demonstrate that nearly all sexual traits correlate with increased multicellular complexity, including anisogamy. We show that anisogamous species are larger than isogamous species, consistent with previously predictions (Parker et al. 1972; Bulmer and Parker 2002), which remained inappropriately statistically tested. Finally, having identified which states are derived for all sexual traits, we demonstrate that the number of sexual traits present in a species well correlates with three metrics of multicellularity, suggesting that the evolution of traits related to multicellularity and sexuality in the volvocine algae may offer a window into their origins and evolution across the Tree of Life.

Several conclusions emerge from the work in this dissertation. Our first motivating question was understanding the identity and evolutionary timing of the genetic basis for the evolution of multicellularity. A major conclusion of Appendix B is that the genetic basis of multicellularity is relatively few genes which can be associated with relatively few specific a priori morphological expectations. Furthermore, this genetic basis did not arise de novo, but rather through the co-option of existing genes and gene pathways, suggesting co-option as an important evolutionary mechanism during evolutionary transitions in individuality. Appendix C builds upon by further investigating the evolutionary history of the genetic basis for cellular differentiation, finding the genes which underlie somatic cells in V. carteri are predicted to have arisen in an undifferentiated species. This work predicts the regA gene cluster is present in both 21

undifferentiated and differentiated daughter species, a prediction validated in subsequent work

(Grochau-Wright et al. 2017).

This conclusion raises more questions than it answers. Given that the genes underlying somatic cells were likely present in an undifferentiated ancestor, what were these genes doing?

Regulating the division rate of all undifferentiated cells in a temporal context rather than a spatial context as observed in Volvox carteri (Nedelcu and Michod 2006; Hanschen et al. 2014;

Grochau-Wright et al. 2017)? Furthermore, given the observation of the regA gene cluster in undifferentiated species, what is its functional purpose? Lastly, the Volvox species in which the regA gene cluster has been sequenced (Appendix C) have likely evolved somatic cellular differentiation independently (Grochau-Wright et al. 2017). However, evidence that the regA gene cluster plays a role in somatic differentiation is only present for Volvox carteri. Was the regA gene cluster independently co-opted to regulate somatic cells in three (at least) lineages independently? If so, such a result would greatly inform the evolutionary repeatability (Gould

1989; Conway Morris 2003) of the evolution of cellular differentiation and multicellularity.

However, Astrephomene, the sister genus of Gonium, also has independently evolved somatic cells (Nozaki 1983; Nozaki et al. 2002, 2006; Herron and Michod 2008; Grochau-Wright et al.

2017), yet would be predicted to lack the regA gene cluster (Appendix B, Hanschen et al. 2016).

How are somatic cells regulated in Astrephomene? If they are regulated by the regA gene cluster, then did this cluster independently arise in Astrephomene than other volvocine algae? If so, this would suggest that the regA gene pathway is easily duplicated during the evolution of multicellularity, which would increase the likelihood and repeatability of the evolution of multicellularity. Alternatively, if the regA gene cluster is not involved in somatic cell regulation 22

in Astrephomene, then there exist multiple genetic pathways to evolve somatic cells, which

would also increase the likelihood and repeatability of the evolution of cellular differentiation

and multicellularity. Taken together, the absence of regA in Gonium pectorale, presence in

undifferentiated species such as Pandorina morum, and presence in species which independently

evolved somatic cells, such as Volvox ferrisii, raises more evolutionary questions, guiding the

direction of necessary and valuable future research.

The other guiding question of this dissertation was how multicellularity drives the

evolution of sexual traits (Appendix D), where we found phylogenetically robust, strong

statistical support consistent with previous theoretical predictions. This work also addresses the

repeatable nature of evolutionary processes, in many of the independent evolutions of

multicellularity across the eukaryotic tree of life, sexual traits such as polar bodies, internal

fertilization, anisogamy, and sexual dimorphism have evolved (Bell 1978, 1982, 1985). In

Appendix D, we showed that multicellularity strongly correlates with the evolution of sexual traits, strongly consistent with previous theoretical predictions (Parker et al. 1972; Bulmer and

Parker 2002). Taken together, this suggests high repeatability in the evolution of sexual traits, even including the highly costly evolution of males (Appendix D, Maynard Smith 1978).

While the two guiding questions of this dissertation seem highly divergent, the genetic basis and subsequent evolutionary pressures of multicellularity, the conclusions of each section point to the repeatability of multicellularity and highly complex sexual traits across the tree of life. Using evolutionary repeatability or convergence as an indicator of adaptive value (Simpson

1953; Mayr 1966; Bonner 2000; Schluter 2000), the repeatability observed across the multicellular innovations suggest that they have an inherent adaptive value, further suggesting 23

the repeatability of the evolution of multicellularity and the evolution of sexual traits such as anisogamy.

24

REFERENCES

Abedin, M., and N. King. 2008. The premetazoan ancestry of cadherins. Science 319:946–948.

Bell, G. 1978. The evolution of anisogamy. J. Theor. Biol. 73:247–270.

Bell, G. 1982. The materpiece of nature: the evolution and genetics of sexuality. Croom Helm,

London.

Bell, G. 1985. The origin and evolution of germ cells as illustrated by the Volvocales. Pp. 221–

256 in H. O. Halvorson and A. Monroy, eds. The origin and evolution of sex. Alan R.

Liss, New York.

Bonner, J. T. 1998. The origins of multicellularity. Integr. Biol. Issues, News, Rev. 1:27–36.

Bulmer, M. G., and G. A. Parker. 2002. The evolution of anisogamy: a game-theoretic approach.

Proc. R. Soc. B Biol. Sci. 269:2381–2388.

Buss, L. W. 1987. The Evolution of Individuality. Princeton University Press, Princeton, New

Jersey.

Coleman, A. W. 2012. A comparative analysis of the Volvocaceae (Chlorophyta). J. Phycol.

48:491–513.

Conway Morris, S. 2003. Life’s Solution: Inevitable Humans in a Lonely Universe. Cambridge

University Press, Cambridge.

Desnitski, A. 1995. A Review on the Evolution of Development in Volvox- Morphological and

Physiological Aspects. Eur. J. Prostiology 31:241–247.

Duncan, L., I. Nishii, A. Harryman, S. Buckley, A. Howard, N. R. Friedman, and S. M. Miller.

2007. The VARL gene family and the evolutionary origins of the master cell-type

regulatory gene, regA, in Volvox carteri. J. Mol. Evol. 65:1–11. 25

Fairclough, S. R., M. J. Dayel, and N. King. 2010. Multicellular development in a

choanoflagellate. Curr. Biol. 20:875–876.

Gould, S. J. 1989. It’s a Wonderful Life: The Burgess Shale and the Nature of History. W. W.

Norton & Company Inc., New York.

Grochau-Wright, Z. I., E. R. Hanschen, P. J. Ferris, B. J. S. C. Olson, and R. E. Michod. 2017.

Genetic Basis for Soma Predates Evolution of Soma in the Volvocine Green Algae. J.

Evol. Biol.

Grosberg, R. K., and R. R. Strathmann. 2007. The Evolution of Multicellularity: A Minor Major

Transition? Annu. Rev. Ecol. Evol. Syst. 38:621–654.

Hanschen, E. R., P. J. Ferris, and R. E. Michod. 2014. Early evolution of the genetic basis for

soma in the Volvocaceae. Evolution 68:2014–2025.

Hanschen, E. R., T. N. Marriage, P. J. Ferris, T. Hamaji, A. Toyoda, A. Fujiyama, R. Neme, H.

Noguchi, Y. Minakuchi, M. Suzuki, H. Kawai-Toyooka, D. R. Smith, V. Luria, A.

Karger, M. W. Kirschner, H. Sparks, J. Anderson, R. Bakaric, P. M. Durand, R. E.

Michod, H. Nozaki, and B. J. S. C. Olson. 2016. The Gonium pectorale genome

demostrates co-option of cell cycle regulation during the evolution of multicellularity.

Nat. Commun. 7:11370.

Hanschen, E. R., D. E. Shelton, and R. E. Michod. 2015. Evolutionary transitions in individuality

and recent models in multicellularity. Pp. 165–188 in A. M. Nedelcu and I. Ruiz-Trillo,

eds. Evolutionary transitions to multicellular life: principles and mechanisms. Springer,

Dordrecht, Netherlands.

Herron, M. D., A. G. Desnitskiy, and R. E. Michod. 2010. Evolution of developmental programs 26

in Volvox (Chlorophyta). J. Phycol. 46:316–324.

Herron, M. D., J. D. Hackett, F. O. Aylward, and R. E. Michod. 2009. Triassic origin and early

radiation of multicellular volvocine algae. Proc. Natl. Acad. Sci. USA 106:3254–3258.

Herron, M. D., and R. E. Michod. 2008. Evolution of complexity in the volvocine algae:

transitions in individuality through Darwin’s eye. Evolution 62:436–451.

Hori, K., F. Maruyama, T. Fujisawa, T. Togashi, N. Yamamoto, S. Mitsunori, S. Sato, T.

Yamada, H. Mori, N. Tajima, T. Moriyama, M. Ikeuchi, M. Watanabe, H. Wada, K.

Kobayashi, M. Saito, T. Masuda, Y. Sasaki-Sekimoto, K. Mashiguchi, K. Awai, M.

Shimojima, S. Masuda, M. Iwai, T. Nobusawa, T. Narise, S. Kondo, H. Saito, R. Sato, M.

Murakawa, Y. Ihara, Y. Oshima-Yamada, K. Ohtaka, M. Satoh, K. Sonobe, M. Ishii, R.

Ohtani, M. Kanamori-Sato, R. Honoki, D. Miyazaki, H. Mochizuki, J. Umetsu, K.

Higashi, D. Shibata, Y. Kamiya, N. Sato, Y. Nakamura, S. Tabata, S. Ida, K. Kurokawa,

and H. Ohta. 2014. Klebsormidium flaccidum genome reveals primary factors for plant

terrestrial adaptation. Nat. Commun. 5:1–9. Nature Publishing Group.

Ispolatov, I., M. Ackermann, and M. Doebeli. 2011. Division of labour and the evolution of

multicellularity. Proc. R. Soc. B Biol. Sci., doi: 10.1098/rspb.2011.1999.

Kirk, D. L. 1998. Volvox: Molecular-genetic origins of multicellularity and cellular

differentiation. Cambridge University Press, Cambridge.

Kumar, K., R. A. Mella-Herrera, and J. W. Golden. 2010. Cyanobacterial heterocysts. Cold

Spring Harb. Perspect. Biol. 2:a000315.

Leliaert, F., D. R. Smith, H. Moreau, M. D. Herron, H. Verbruggen, C. F. Delwiche, and O. De

Clerck. 2012. Phylogeny and Molecular Evolution of the Green Algae. CRC. Crit. Rev. 27

Plant Sci. 31:1–46.

Linnaeus, C. 1758. Systema naturae per regna tria naturae, secundum classes, ordines, genera,

species, cum characteribus, differentiis, synonymis, locis. Tomus I. Editio decima,

reformata. Editio decima revisa. Vol. 1. Holmiae, Stockholm.

Maynard Smith, J. 1978. The evolution of sex. Cambridge University Press, Cambridge.

Maynard Smith, J., and E. Szathmáry. 1995. The Major Transitions in Evolution. Oxford

University Press, Oxford.

Merchant, S. S., S. E. Prochnik, O. Vallon, E. H. Harris, S. J. Karpowicz, G. B. Witman, A.

Terry, A. Salamov, L. K. Fritz-Laylin, L. Maréchal-Drouard, W. F. Marshall, L.-H. Qu,

D. R. Nelson, A. a Sanderfoot, M. H. Spalding, V. V Kapitonov, Q. Ren, P. Ferris, E.

Lindquist, H. Shapiro, S. M. Lucas, J. Grimwood, J. Schmutz, P. Cardol, H. Cerutti, G.

Chanfreau, C.-L. Chen, V. Cognat, M. T. Croft, R. Dent, S. Dutcher, E. Fernández, H.

Fukuzawa, D. González-Ballester, D. González-Halphen, A. Hallmann, M. Hanikenne,

M. Hippler, W. Inwood, K. Jabbari, M. Kalanon, R. Kuras, P. A. Lefebvre, S. D.

Lemaire, A. V Lobanov, M. Lohr, A. Manuell, I. Meier, L. Mets, M. Mittag, T.

Mittelmeier, J. V Moroney, J. Moseley, C. Napoli, A. M. Nedelcu, K. Niyogi, S. V

Novoselov, I. T. Paulsen, G. Pazour, S. Purton, J.-P. Ral, D. M. Riaño-Pachón, W.

Riekhof, L. Rymarquis, M. Schroda, D. Stern, J. Umen, R. Willows, N. Wilson, S. L.

Zimmer, J. Allmer, J. Balk, K. Bisova, C.-J. Chen, M. Elias, K. Gendler, C. Hauser, M.

R. Lamb, H. Ledford, J. C. Long, J. Minagawa, M. D. Page, J. Pan, W. Pootakham, S.

Roje, A. Rose, E. Stahlberg, A. M. Terauchi, P. Yang, S. Ball, C. Bowler, C. L.

Dieckmann, V. N. Gladyshev, P. Green, R. Jorgensen, S. Mayfield, B. Mueller-Roeber, 28

S. Rajamani, R. T. Sayre, P. Brokstein, I. Dubchak, D. Goodstein, L. Hornick, Y. W.

Huang, J. Jhaveri, Y. Luo, D. Martínez, W. C. A. Ngau, B. Otillar, A. Poliakov, A.

Porter, L. Szajkowski, G. Werner, K. Zhou, I. V Grigoriev, D. S. Rokhsar, and A. R.

Grossman. 2007. The Chlamydomonas genome reveals the evolution of key animal and

plant functions. Science 318:245–50.

Michod, R. E. 1999. Darwinian Dynamics, evolutionary transitions in fitness and individuality.

Princeton University Press, Princeton, New Jersey.

Michod, R. E. 2007. Evolution of individuality during the transition from unicellular to

multicellular life. Proc. Natl. Acad. Sci. U. S. A. 104:8613–8618.

Michod, R. E. 2006. The group covariance effect and fitness trade-offs during evolutionary

transitions in individuality. Proc. Natl. Acad. Sci. U. S. A. 103:9113–9117.

Michod, R. E., Y. Viossat, C. A. Solari, M. Hurand, and A. M. Nedelcu. 2006. Life-history

evolution and the origin of multicellularity. J. Theor. Biol. 239:257–272.

Muro-Pastor, A. M., and W. R. Hess. 2012. Heterocyst differentiation: from single mutants to

global approaches. Trends Microbiol. 20:548–557. Elsevier Ltd.

Nedelcu, A. M., and R. E. Michod. 2006. The evolutionary origin of an altruistic gene. Mol.

Biol. Evol. 23:1460–4.

Nichols, S. A., B. W. Roberts, D. J. Richter, S. R. Fairclough, and N. King. 2012. Origin of

metazoan cadherin diversity and the antiquity of the classical cadherin/β-catenin

complex. Proc. Natl. Acad. Sci. U. S. A. 109:13046–13051.

Nozaki, H. 2003. Flagellated green algae. Pp. 225–252 in J. D. Wehr and R. G. Sheath, eds.

Freshwater Algae of North America. Academic Press, San Diego, California. 29

Nozaki, H. 1983. Morphology and taxonomy of two species of Astrephomene in Japan. Japanese

J. Phycol. 58:345–352.

Nozaki, H., F. D. Ott, and A. W. Coleman. 2006. Morphology, molecular phylogeny and

taxonomy of two new species of Pleodorina (Volvoceae, Chlorophyceae). J. Phycol.

42:1072–1080.

Nozaki, H., M. Takahara, A. Nakazawa, Y. Kita, T. Yamada, H. Takano, S. Kawano, and M.

Kato. 2002. Evolution of rbcL group IA introns and intron open reading frames within

the colonial Volvocales (Chlorophyceae). Mol. Phylogenet. Evol. 23:326–338.

Otto, S. P. 2009. The evolutionary enigma of sex. Am. Nat. 174:S1–S14.

Parker, G. A., R. R. Baker, and V. G. F. Smith. 1972. The origin and evolution of gamete

dimorphism and the male-female phenomenon. J. Theor. Biol. 36:529–553.

Prochnik, S. E., J. Umen, A. M. Nedelcu, A. Hallmann, S. M. Miller, I. Nishii, P. J. Ferris, A.

Kuo, T. Mitros, L. K. Fritz-Laylin, U. Hellsten, J. Chapman, O. Simakov, S. A. Rensing,

A. Terry, J. Pangilinan, V. Kapitonov, J. Jurka, A. Salamov, H. Shapiro, J. Schmutz, J.

Grimwood, E. Lindquist, S. Lucas, I. V. Grigoriev, R. Schmitt, D. L. Kirk, and D. S.

Rokhsar. 2010. Genomic analysis of organismal complexity in the multicellular green

alga Volvox carteri. Science 329:223–6.

Reynolds, C. S. 1983. Growth-rate responses of Volvox aureus Ehrenb. (Chlorophyta,

Volvocales) to variability in the physical environment. Br. Phycol. J. 18:433–442.

Rueffler, C., J. Hermisson, and G. P. Wagner. 2012. Evolution of functional specialization and

division of labor. Proc. Natl. Acad. Sci. 109:E326–E335.

Ryan, J. F., K. Pang, C. E. Schnitzler, A.-D. Nguyen, R. T. Moreland, D. K. Simmons, B. J. 30

Koch, W. R. Francis, P. Havlak, S. A. Smith, N. H. Putnam, S. H. D. Haddock, C. W.

Dunn, T. G. Wolfsberg, J. C. Mullikin, M. Q. Martindale, and A. D. Baxevanis. 2013.

The Genome of the Ctenophore Mnemiopsis leidyi and Its Implications for Cell Type

Evolution. Science 342:1242592–1242592.

Sebé-Pedrós, A., A. Ariza-Cosano, M. T. Weirauch, S. Leininger, A. Yang, G. Torruella, M.

Adamski, M. Adamska, T. R. Hughes, J. L. Gómez-Skarmeta, and I. Ruiz-Trillo. 2013.

Early evolution of the T-box transcription factor family. Proc. Natl. Acad. Sci. U. S. A.

2–7.

Shelton, D. E., and R. E. Michod. 2014. Group Selection and Group Adaptation During a Major

Evolutionary Transition: Insights from the Evolution of Multicellularity in the Volvocine

Algae. Biol. Theory 9:452–469.

Solari, C. A., S. Ganguly, J. O. Kessler, R. E. Michod, and R. E. Goldstein. 2006a.

Multicellularity and the functional interdependence of motility and molecular transport.

Proc. Natl. Acad. Sci. U. S. A. 103:1353–1358.

Solari, C. A., J. O. Kessler, and R. E. Michod. 2006b. A hydrodynamics approach to the

evolution of multicellularity: flagellar motility and germ-soma differentiation in

volvocalean green algae. Am. Nat. 167:537–554.

Sommer, U., and Z. Maciej Gliwicz. 1986. Long range vertical migration of Volvox in tropical

Lake Cahora Bassa (Mozambique). Limnol. Oceanogr. 31:650–653.

Suga, H., Z. Chen, A. de Mendoza, A. Sebé-Pedrós, M. W. Brown, E. Kramer, M. Carr, P.

Kerner, M. Vervoort, N. Sánchez-Pons, G. Torruella, R. Derelle, G. Manning, B. F.

Lang, C. Russ, B. J. Haas, A. J. Roger, C. Nusbaum, and I. Ruiz-Trillo. 2013. The 31

Capsaspora genome reveals a complex unicellular prehistory of animals. Nat. Commun.

4:2325–2333.

Szathmáry, E., and J. Maynard Smith. 1995. The major evolutionary transitions. Nature

374:227–232.

Van Dyken, J. D., and M. J. Wade. 2012a. Origins of altruism diversity I: The diverse ecological

roles of altruistic strategies and their evolutionary responses to local competition.

Evolution 66:2484–97.

Van Dyken, J. D., and M. J. Wade. 2012b. Origins of altruism diversity II: Runaway coevolution

of altruistic strategies via “reciprocal niche construction”. Evolution 66:2498–513. van Leeuwenhoek, A. 1700. Part of a Letter from Mr Antony van Leeuwenhoek, concerning the

Worms in Sheeps Livers, Gnats, and Animalcula in the Excrements of Frogs. Philos.

Trans. R. Soc. London 22:509–518.

32

APPENDIX A. EVOLUTIONARY TRANSITIONS IN INDIVIDUALITY AND

RECENT MODELS OF MULTICELLULARITY

Hanschen E. R., D. E. Shelton, R. E. Michod. 2015. Evolutionary transitions in individuality and

recent models of multicellularity. Pp. 165-188 in A. M. Nedelcu and I. Ruiz-Trillo, eds.

Evolutionary Transitions to Multicellular Life: Principles and mechanisms. Springer,

Dordrecht, Netherlands. With permission of Springer.

33

Evolutionary Transitions in Individuality and Recent Models of Multicellularity

Erik R. Hanschen, Deborah E. Shelton and Richard E. Michod

Abstract An evolutionary transition in individuality (ETI) is a fundamental shift in the unit of adaptation. ETIs occur through the evolution of groups of individuals into a new higher-level individual. The evolution of groups with cells specialized in somatic (viability) or reproductive functions has been proposed as a landmark of the unicellular to multicellular ETI. Several recent models of the evolution of multicellularity and cellular specialization have contributed insights on different aspects of this topic; however, these works are disconnected from each other and from the general framework of ETIs. While each of these works is valuable on its own, our interest in ETIs motivates an attempt to connect these models. We review the theory of ETIs along with these recent models with an eye towards better integrating insights from these models into the ETI framework. We consider how each model addresses key recurring topics, such as the importance of cooperation and conflict, life history trade-offs, multi-level selection, division of labor and the decoupling of fitness at the level of the group from the level of the cell. Finally, we identify a few areas in which conflicting views or unanswered questions remain, and we discuss modeling strategies that would be most suited for making further progress in understanding ETIs.

Keywords Cellular specialization Division of labor Evolution Evolutionary transition in individuality (ETI) Major· transition Modeling· Multicellularity· · · ·

E. R. Hanschen (!) D. E. Shelton R. E. Michod Department of Ecology· and Evolutionary· Biology, University of Arizona, 1041 E. Lowell St., BioSciences West, Rm. 310, Tucson, AZ 85721, USA e-mail: [email protected]

D. E. Shelton e-mail: [email protected]

R. E. Michod e-mail: [email protected]

© Springer Science+Business Media Dordrecht 2015 165 I. Ruiz-Trillo, A. M. Nedelcu (eds.), Evolutionary Transitions to Multicellular Life, Advances in Marine Genomics 2, DOI 10.1007/978-94-017-9642-2_9 34

166 E. R. Hanschen et al.

Introduction

Why do groups of individuals evolve into new kinds of individuals? Answering this question is basic to understanding the origin of the hierarchy of life: genes, chro- mosomes, cells, cells within cells (eukaryotic cell), multicellular organisms, and societies. We refer to transitions among levels in the hierarchy of life as evolu- tionary transitions in individuality or ETIs. Building on previous work (Buss 1987; Maynard Smith 1988, 1991; Maynard Smith and Szathmáry 1995; Otto and Orive 1995), Michod and colleagues developed a framework for understanding ETIs, with special focus on the transition from unicellular to multicellular life. This framework involves five interrelated components: the evolution of cooperation (and conflict) in groups, multi-level selection, life history trade-offs, division of labor among the basic components of fitness (reproduction and viability), and, finally, decoupling of fitness of the two levels in the selection hierarchy (after which the fitness of the group is no longer a simple additive function of the fitnesses of the lower level units; Michod 1996, 1997, 1999, 2003, 2006, 2007; Michod and Roze 1999, 2001; Michod and Nedelcu 2003; Michod et al. 2006). Recently, a number of interesting models have addressed the evolution of multicel- lularity and division of labor from several different perspectives. While each of these works is valuable on its own, our interest in ETIs motivates an attempt to integrate them, so we consider these models from the perspective of the ETI framework. Of course other useful and valid frameworks for the conceptual issues discussed exist, but for the purposes of this paper, we have restricted ourselves to the ETI framework as discussed below. As each of these models was developed for their own quite valid and compelling reasons, discussing how these models relate to the ETI framework is not meant as criticism. Rather, we wish to analyze these models to help us better understand the main issues relevant to ETIs. In this section we briefly review the ETI framework and the models upon which it is based. Then, in the next sections we discuss some of the more recent models and how they relate to the ETI framework as well as provide general suggestions for future work.

Overview of the ETI Framework

Because of cooperation, groups may function in ways that their members cannot. For example, a cell may not be able to swim and divide at the same time, but a group containing cells specialized at either swimming or cell division may undertake both functions simultaneously. In this way, groups of cooperating cells may break through the life history constraints governing life as a single cell. When cells start forming groups, selection operates at both the level of the cell and the level of the group and may lead to the evolution of cooperation among cells in the group. Cooperation in turn provides the opportunity for cheating and conflict among lower-level units, which must be mediated if cooperation is to be stable. Under certain conditions, conflict mediators evolve that lead to specialized reproductive and non-reproductive 35

Evolutionary Transitions in Individuality and Recent Models of Multicellularity 167 somatic cells. Division of labor among lower level units specialized in the basic fitness components of the group enhances the individuality of the group. Once cells are specialized in one of the necessary components of fitness, say reproduction or viability, they can no longer exist outside of the group and the fitness of the group is no longer the average fitness of the cells belonging to the group. In summary, a working hypothesis for the basic steps in an ETI is as follows: (i) formation of groups, (ii) increase of cooperation within groups, (iii) cheating and conflict, (iv) conflict mediation leading to enhanced cooperation, (v) division of labor in the basic components of fitness leading to (vi) fitness decoupling and individuality of the group. The extent to which these steps apply in diverse empirical cases and the implications of empirical results are active areas of research. The evolution of cooperation, the central problem of social biology, gains special significance during ETIs because altruism and other forms of cooperation lead to the transfer of fitness from the lower level (the costs of altruism) to the group level (the benefits of altruism). By “transfer of fitness” we do not mean to suggest that fitness is a conserved quantity; rather, we mean that a single change (e.g., the increase in cooperation) both decreases the fitness of the lower level and increases the fitness of the higher level. For example, in the additive model of altruism with costs to self -c and benefit to group b, c and b are not equal nor are they typically related in magnitude (although in specific cases there may be some connection between the two). Related to cooperation and altruism is specialization of group members in the two fitness components, reproduction and viability, of the group (Michod 2005, 2006). As already mentioned, when cells completely specialize at one of the two basic fitness components, they lose their overall fitness and capacity to function as evolutionary individuals if they existed outside the context of the group. By virtue of their specialization, cells have low cell-level fitness if they existed as single- celled individuals, while they contribute to increasing the fitness of the group. As a consequence, the fitness of the group is no longer the average of the individual fitnesses of the component cells. This is an example of a general principle promoting group cohesion: cell traits optimal in a group context may no longer be optimal outside of the group context; indeed group-selected traits may be deleterious at the cell level (Shelton and Michod 2014). Energy, resources, and time expended on one fitness component often detract from another component, resulting in trade-offs among fitness components. Fitness trade-offs drive the diversification of life history traits in extant species (Stearns 1992; Roff 2002)butgainspecialsignificanceduringETIsforthreereasons(Michod2006, 2007; Michod et al. 2006). First, specialization by cells in the fitness components of the group can be driven by cell-level fitness trade-offs. The curvature of the trade- off between survival and reproduction is known to be a central issue in life history evolution (Stearns 1992; Roff 2002). In the case of the origin of multicellularity, if the lower-level fitness trade-off is of convex curvature (positive second derivative) and assumptions are made about how lower-level fitness components relate to higher- level fitness components (see below), specialization by cells in the reproductive and 36

168 E. R. Hanschen et al. viability fitness components (termed “germ-soma specialization” below) of the group will be an optimal group-level strategy (Michod et al. 2006). Second, tradeoffs between fitness components at the cell level provide the basis for cooperation and division of labor in groups. Life history “trade-off genes” are genes that down regulate reproduction so as to enhance survival of the organism in stressful environments. Such genes can be co-opted to produce reproductive altruism in cell groups (somatic cells) through the shifting of their expression from a temporal (environmentally-induced) context into a spatial (developmental) context (Nedelcu and Michod 2006). For example, an important component of viability in the volvocine green algae is flagellar motility, but cell division and reproduction interfere with flagellar motility. In the unicellular members of this lineage, selection will optimize the allocation of time and energy to these two processes, but, in a group, cells that spend more time flagellated divide less frequently. As flagellar action of cells benefits group motility, cells with a greater propensity to remain actively flagellated are altruistic relative to cells that spend less time flagellated (because less flagellated cells reproduce more). Third, as discussed below in Eq. 1, fitness tradeoffs can enhance the fitness of the group through a covariance effect by which group fitness is augmented beyond the average fitness of cells according to the covariance of cellular contributions to viability and fecundity (Michod 2006).

Models Related to the ETI Framework

The ETI framework described above is based on results from two different kinds of models: two-locus population genetic modifier models (Michod 1996, 1997, 1999, 2003, 2006; Michod and Roze 1997, 1999, 2001; Michod and Nedelcu 2003; Michod et al. 2003, 2006) and optimality models of division of labor (Michod 2006; Michod et al. 2006).

Two Locus Modifier Models

The two locus modifier models seek to explain how modifiers of development evolve in response to mutation and selection at a cooperation locus. Development involves the conversion of a propagule into an adult cell group via cell division. Propagules contain a cell or cells sampled from an adult group in the previous generation (or from several adult groups in the case of aggregation). Sex may occur in the case of single celled propagules that fuse with propagules from other groups to start a new group. The cooperation locus has two alleles, C and D, which express cooperation and defection, respectively, among cells. Cooperation benefits other cells in the group at a cost to the cooperating cell. Defecting cells do not pay any cost, but receive the benefits of cooperating cells in their group. Benefit and cost are in terms of the 37

Evolutionary Transitions in Individuality and Recent Models of Multicellularity 169 cell death rate or cell division rate. During development of the adult group, there is recurrent mutation from C to D at each cell division (back mutation is ignored on the view that there are many more ways to lose a functional trait like cooperation than to gain it again). These defector mutations disrupt the functioning of the adult cell group by reducing the level of cooperation. Mutation increases the variance and opportunity for selection at the within-group or cell level. After the adult group is formed, a propagule is made. Depending on the parameters of development (mutation rate, cell replication and death rates, total number of cell divisions, the costs and benefits of cooperation, and the mode of propagule formation), a polymorphism may be maintained at the C/D locus by mutation selection balance. This polymorphism sets the stage for the evolution of modifiers of development assumed to be encoded by a second locus. The second modifier locus affects the parameters or mode of development of the adult group and so affects the degree to which the propagule produced by an adult resembles the propagule that founded the group (a measure of group heritability). By molding development, modifier alleles affect group heritability and the capacity of the groups to reproduce themselves (Griesemer 2000). Why do these modifiers evolve and how do they lead to the capacity of a group to reproduce itself? The mutation-selection balance equilibrium at the C/D locus implies that C alleles are fitter than D alleles, to compensate for mutation from C to D. Under certain conditions, alleles at the modifier locus evolve due to interaction with the fitter C allele. This has the effect of increasing the between-group variance and decreasing the within-group variance, thereby increasing the level of cooperation and the fitness of the group. Examples of conflict modifiers studied by this approach include germ- soma specialization, reduced mutation rate, policing, programmed cell death, passing the life cycle through a single-cell zygote stage, and fixed group size (reviewed in Michod 2003). By increasing the variance at the group level and decreasing the variance at the cell level, the modifiers lead to fitness decoupling between levels and enhance the capacity of group to reproduce itself. The capacity of a group to reproduce itself may be measured by the degree to which a group created by a propagule resembles the group the propagule came from. Alternatively, since the group is made from a propagule, and the two locus recurrence equations are in terms of the gene and genotype frequencies at the propagule stage, we may measure the capacity for reproduction and heritability as the degree to which the propagules produced by a group are similar to the propagule(s) that created the group.

Optimality Division of Labor Model

The evolution of specialization at the two basic components of fitness, reproduction and viability, was also studied using an optimality approach (Michod 2006; Michod et al. 2006). In these models, the phenotype of a cell is described by its effort at repro- duction (fecundity) with the remainder of effort put into the viability. In the simplest case, the viability, V, and fecundity, B, of the group are assumed to be arithmetic av- erages of the cell efforts at the two fitness components, viability and fecundity, v and 38

170 E. R. Hanschen et al. b, respectively, or V i vi /N and B i bi /N. This formulation assumes an initial isomorphism between= fitness components= at the two levels, because it assumes that the activities of the! cell at cellular viability! and cellular fecundity contribute, respectively, to group viability and group fecundity. Clearly, as cells become more specialized and integrated into the group, this isomorphism breaks down; but this assumption likely applies as cells first start joining (and leaving) groups. This for- mulation of group fitness also assumes that the two fitness components of the group, viability and fecundity, are first composed separately from cell properties, and then combined (multiplicatively as is appropriate for discrete generations) to generate the fitness of the group, W. Without this (or a related assumption), evolution of special- ization at activities which trade-off with one another at the lower level would not be possible. Group fitness would be zero, if group fitness was composed directly out of cell fitness, and cells completely specialized in one fitness component or the other. The cell fitness (vb) of specialized cells is zero since one component (v or b) is zero. In this way, the group may break through the trade-off constraints imposed at the cell level. Finally, it is assumed that the total fitness at either level is the product of viability and fecundity, as is appropriate for cells or multicellular organisms with discrete generations. One result of the model is the group covariance effect given in Eq. 1, which shows that the fitness of the cell group, W (taken as the product of V and B), is greater than the average fitness of member cells, w i vi bi /N, by an amount equal to the negative covariance of the fitness components¯ = at the cell level (viability, v, and fecundity, b). !

W VB w Cov[v, b] (1) = =¯− If the covariance between fitness components is itself negative, as it is when fitness components trade-off with one another, there is an enhanced fitness at the group level from what would be expected from the average fitness of cells. The covariance effect given in Eq. 1 translates the negative covariance of fitness components of group members into a benefit at the group level. Alternatively, if fitness components were to positively co-vary, then there would be a decrease in fitness of the group from that expected by the average cell fitness. High fitness for any unit at any level of organization requires a balance of fitness components at that level; in the explicit formulation here of discrete generations the components are multiplied together to give total fitness. The covariance effect translates a lack of balance at the lower cell level into an advantage at the group level, especially under conditions of convexity of the trade-offs. Convexity of the trade-off curvature allows for enhanced effectiveness at each fitness component for cells that specialize (Michod 2006; Michod et al. 2006). The particular mathematical representation of the covariance effect given in Eq. 1 depends upon additivity of effects on the viability and reproduction com- ponents of fitness as described above. Additivity of fitness effects is the simplest assumption possible, nevertheless, the assumption of additivity of the contributions of cells to the viability of the group may be relaxed and the general points still hold (Michod et al. 2006). 39

Evolutionary Transitions in Individuality and Recent Models of Multicellularity 171

Fig. 1 Two cells specializing in different fitness components, reproduction (white) and viability (grey). Cell i specializes in reproduction, with reproductive effort bi , with less effort put into viability functions, vi . Cell j does the reverse. Alone they would each have low fitness, because they are unbalanced and high fitness requires a balance at the two components. However, together, they may constitute a good team and bring high fitness to the group. (From Michod (2007))

As illustrated in Fig. 1, what is required for the covariance effect is that, if one cell has a high reproductive effort (and hence a low viability, and a low cell fitness), this may be compensated for by another cell with high viability (and hence a low fecundity, and also a low cell fitness) (Michod et al. 2006). Consequently, even though each of these cells by themselves would have a low fitness, together they can bring a high fitness to the group, especially under conditions of convexity of the trade-off. In effect, such oppositely specialized cells complement one another and constitute a good and integrated “team” under conditions of convexity of the trade-off curve. This kind of joint effect contributes to fitness decoupling and integration of the group, and would not be possible if group fitness were the average of cell fitness. During an ETI, fitness must be reorganized so that the individuals (e.g., cells) that were the focus of evolution and adaptation become components of a new higher- level individual (e.g., multicellular group) that is the new focus of evolution and adaptation. As already discussed, this involves two major processes. First, fitness and its variance is increased at the new group level and reduced at the lower level. Second, the previously-existing individuals, which are now members of the group, specialize in the fitness components (reproduction and viability) of the higher level group. This leads to the decoupling of fitness between the levels so that the fitness of the group may be quite high, while the fitness of the group members may be low or even non-existent were they to leave the group. The reorganization of fitness during ETIs may involve a number of cycles of cooperation, conflict and conflict mediation. Conflict through its mediation allows for more cooperation. This cycle of cooperation, conflict and conflict mediation drives ETIs. 40

172 E. R. Hanschen et al.

Table 1 Summary of how each model incorporates five interrelated components of ETIs Cooperation Multilevel Life history Division of la- Fitness in groups selection trade-offs bor decoupling Willensdorfer Implicit (see None Colony level Cells None 2009 discussion (benefit and specialize on in text) cost of soma) viability or fertility Gavrilets Implicit Viability Cell level Cells special- Present once 2010 Conflict depends on (fertility and ize on viabil- division of between colony trait, viability) ity or fertility labor evolves levels arises fertility develop- depends on mentally cell trait from mutation Ispolatov Synergistic Cell death Cell level Cells None et al 2012 cooperation rate depends (metabolic specialize on on cell trait, process A and a single cell birth rate B) metabolic depends on process colony trait Rueffler None None Colony level Cells None et al 2012 (between two specialize on tasks) a single task Van dyken Altruism Implicit Organism Organisms Implicit (see and Wade level specialize on discussion in 2012a, b (between viability or text) tasks) fertility

Recent Models

In addition to the models described above, which were built around the concepts associated with the ETI framework, several recent models have addressed various aspects associated with the evolution of multicellularity and cellular specialization. In Table 1 below we briefly relate the recent models considered here to the ETI framework presented in the previous section.

Willensdorfer 2009

Willensdorfer (2009) takes a colony-level optimality approach, defining colony fit- ness as the product of three functions. The first function represents the cost of soma, the second the cost of colony size, and the third the benefit of group living. Although not explicitly stated in the paper, by assuming costly somatic cells and germ cells that contribute less to group benefit function than do somatic cells, the model is based on 41

Evolutionary Transitions in Individuality and Recent Models of Multicellularity 173 cell-level trade-offs among fitness components. Willensdorfer is interested in when the benefit of soma to biomass production outweighs the cost of soma. Based on this condition, he predicts when somatic cells will evolve and what the proportion of somatic cells will be. Similar to Solari et al. (2006) and the optimality framework above, this model investigates the optimal proportion of soma given certain benefits and costs. Because this is a single-level model, Willensdorfer (2009) models the evolution of division of labor after an evolutionary transition to group living has occurred. We see similarities and differences between Willensdorfer’s model and the ETI framework (Table 1). We have already mentioned that fitness trade-offs are implicitly assumed and drive the division of labor in his model. In addition, cooperation between soma and germ cells is present. Multi-level selection is absent from this model because cell-level variation and cell-level selection have been explicitly removed. It is unclear how Willensdorfer’s biomass fitness definition relates to more traditional fitness definitions and life history issues (r, R0, invasion fitness, etc.). To try to understand fitness in this model from a life history perspective (say with fitness equal to the product of reproduction and viability), one could view the numerator of the first term (biomass of reproductive cells) of Willensdorfer’s fitness definition as reproduction and the rest of the equation (allometric cost of size benefit of multicellularity/total biomass) as viability (Eq. 2, Willensdorfer 2009×). However this results in a viability term that is very difficult to interpret biologically. Novel definitions of fitness are interesting, but make it difficult to connect to other models or to the biological cases we wish to understand. In terms of the steps in an ETI, Willensdorfer is investigating how multicellular organisms evolve in the later stages after obligate coloniality has evolved.

Gavrilets 2010

Gavrilets (2010) decomposes fitness into reproductive and survival fitness compo- nents and defines a trade-off between them. He assumes four genes, two functional genes for each of the two fitness components and two regulatory loci, one for each fitness component. Investment in a functional locus increases cell fertility or cell viability, but negatively influences the other fitness component. The two regulatory loci determine to what extent cells will contribute to viability and reproduction. Thus, in order to evolve cellular differentiation in this model, a colony must increase in- vestment in viability and fertility (via the functional loci) as well as regulation (via the regulatory loci) so every cell will specialize in either viability or fertility. Such cell specialization increases colony-level fitness by avoiding the cell level trade-off. Gavrilets studies how the shape of the trade-off affects the evolution of division of labor at reproductive and viability functions. The life cycle consists of development, survival selection, and reproduction selec- tion. During development, a single cell divides to found a colony and subsequent cell divisions result in a colony of a specified size. Mutation at all four loci occurs during 42

174 E. R. Hanschen et al. cell division, which produces intracolony and intercolony variation. A set proportion of cells always develop as potential soma, regardless of whether regulation actually functionally distinguishes them from germ cells. During survival selection, a colony survives based on the average of its cell investment in survival. Lastly, during repro- duction selection, cells reproduce with probability proportional to their investment in fertility. In this way, conflict between cell level fitness (which favors cell fecun- dity) and the colony level fitness (which favors a balance between cell fecundity and cell investment in colony viability) arises. Individual-based simulations are used to investigate how the shape of the tradeoff between viability and fecundity alters the evolution of division of labor and Gavrilets concludes that convex trade-offs are necessary for division of labor to evolve. The model shares many of the properties of the ETI framework above: life history trade-offs, multi-level selection, cooperation, and reproductive division of labor (see Table 1). Life history trade-offs are manifested as functional trade-offs at the cell level (survival and reproduction). When this trade-off is convex, cooperation is observed as germ and somatic cells evolve. Multi-level selection is present as selection operates at the colony level through survival selection and at the cell level through fecundity selection on variation introduced through developmental mutation. Fitness decoupling is also present as the fitness of cells decreases (if they were outside the colony) as colony-level fitness increases. Lastly, division of labor in reproduction and survival components is clearly observed and basic to the transition to a multicellular individual. Interestingly, some portions of Gavrilets’ discussion suggest that we would be mistaken to see multilevel selection and cooperation/conflict as major elements of his model. For example, Gavrilets says (2010, p. 6), “In the model, cell differentiation and the division of labor are driven by individual selection maximizing the number of colony-producing offspring of a colony-producing cell. That is, the transition to individuality can be explained in terms of immediate selective advantage to individual replicators (Maynard Smith and Szathmáry 1998). Note that mutant cells that ‘cheat’ by having increased fertility within colonies will tend to lose in competition at the colony level after they develop their own colonies. Therefore, the conflict between individual and colony level selection is largely removed.” We understand the word “individual” in the above quote to be synonymous with “cell” or “cell-level” (except when he says “individuality,” here he is talking about a multicellular individual). Given that interpretation, Gavrilets is making the claim that cell-level selection “drives” division of labor in this model. Apparently in support of this claim, Gavrilets points out that the success of “cheater” (high fertility) cells depends on how they are assorted into groups. Recall that this model assumes an extreme bottleneck: every colony originates from a single cell and eventually disin- tegrates into single cells. Recall also in this model that (i) a colony survives based on the average of its cells’ investments in survival, (ii) cells reproduce proportionally to their cell-level investment in fertility, and (iii) there is a cell level trade-off between the two investments. Following Gavrilets’ quote, consider a non-cheating cell that gives rise to a colony and a cheater that arises by mutation during development of that colony. Call the cells that descend from the original cell the F1 cells. We would 43

Evolutionary Transitions in Individuality and Recent Models of Multicellularity 175 expect the cheater genotype to be over-represented among F1 cells because a cheater does well at reproduction at the expense of its non-cheating colony-mates. Call the cells that descend from an F1 cheater cell and actually survive to begin the process of creating colonies of their own the F2c cells and cells which descend from a F1 non-cheater cell the F2nc cells. Our expectation of which genotype is doing better in the overall population (F2c and F2nc) is not obvious—it depends on the fertility advantage at the cell level and the survival disadvantage at the group level. Because the F1 cheater genotypes produce poor-surviving colonies, their ability to make it into the F2 generation is compromised as are the colonies that result from them if they make it into the F2 generation. We agree with Gavrilets that the sorting effect of the single-cell bottleneck is an important aspect of his model. However, this point does not support the idea that cell-level selection “drives” the evolution of division of labor in his model. In order to see what traits cell-level selection favors, the relevant question is not the relative representation of cheater cells in the overall cell population in F2. That would be informative about the net effect of cell-level and group-level selection combined. Rather, the effect of individual selection should be assessed by asking what the effect of a trait change is within each group (or what its effect would be if the groups were dissolved). We think that Gavrilets’ logic here—that cheater cells are not doing better overall in the F2 generation, therefore cell-level selection does not favor cheating—is an example of what others have called the “averaging fallacy” (Sober and Wilson 1998; Okasha 2006). Gavrilets’ claim that individual- level selection is driving the results of his model notwithstanding, we see this model as clearly overlapping with the ETI framework with respect to having multilevel selection. From these considerations, it is clear to us that levels of selection, cooperation and conflict are major elements of Gavrilets’ model. In addition, there exist several group-level properties built into colonies from the start of the simulations, including the unspecified traits that allow the assumption of undifferentiated colonies as a starting point for the analysis, the developmental plasticity of cells to terminally differentiate, and the set proportion of germ and somatic precursor cells in a colony. In addition, groups of cells reproduce groups of cells through a group-level life cycle in the sense that a colony produces single cell propagules that develop into colonies. Gavrilets sets the expression of the regulatory loci to be zero in the founding pop- ulation thereby starting simulations with undifferentiated colonies. However, this initial condition is an unstable starting point. In the first few generations, the regula- tory loci mutate away from zero, and germ and soma provide different contributions to colony-level fitness. Because undifferentiated colonies are unstable, Gavrilets has implicitly assumed the initial existence of division of labor and explained the subsequent evolution of more pronounced division of labor. A similar comment on Gavrilets’ model was made by Ispolatov et al. (2012). Because initial colonies have division of labor and division of labor is the hallmark of evolutionary individuals, we view the initial colonies as already possessing properties of higher level individuals. For this reason, we see the model as addressing the latter stages of an ETI. 44

176 E. R. Hanschen et al.

Ispolatov, Ackermann, and Doebeli 2012

Ispolatov et al. (2012) model the evolution of multicellularity in cyanobacteria where individual cells need to perform two incompatible metabolic processes such as oxy- genic photosynthesis and anoxic nitrogen fixation. They assume a fitness cost to producing both metabolic products simultaneously. They utilize chemical mass ac- tion equations and assume an ephemeral group state in which cells aggregate into pairs that dissociate into cells. Cells are assumed to quickly alter their metabolic investment based on the paired group state and fitness landscape to increase cell fitness. Lastly, they model the evolution of a heritable trait, stickiness, which is rep- resentative of the evolution of multicellularity, as the amount of time a cell spends in a paired group state. One innovation of this model is the changing dimensionality of the fitness land- scape based on the solitary or paired state. When a cell is solitary, its fitness is a function of its investment in each metabolic process (two dimensions). When a cell is in a pair, its fitness is a function of its and its partner’s investment in each metabolic process (four dimensions). This change in dimensionality changes a cell’s fitness landscape from a single optimal peak when solitary to a saddle point when paired. When a saddle point is present in the fitness landscape, cells in a pair will differentiate to specialize on one metabolic product. The introduction of changes in dimensionality in fitness is important because it mathematically represents the increased complexity of multilevel selection in the group state. They find that multi- cellularity in the sense of the paired state, as a function of cellular stickiness, evolves when the cost of producing both metabolic products is high enough to make the mixed partial derivative of the fitness landscape negative (representing a saddle point) and the cost of being sticky is relatively low. During their discussion, they emphasize that this model has several advantages absent in other models. Specifically, this model does not explicitly assume pre- existing groups or differentiation (e.g., Willensdorfer 2009; Gavrilets 2010) and this allows the simultaneous evolution of groups (of size two) and division of labor. In this way, the pre-existence of developmental pathways or undifferentiated groups of cells is avoided. As much research investigates how differentiated colonies evolved from undifferentiated colonies (Willensdorfer 2008, 2009; Gavrilets 2010; Rueffler et al. 2012), a model which simultaneously evolves small colonies and division of labor is an explicit demonstration these processes are not necessarily separate. This model has similarities and differences with the ETI framework (Table 1). When in a pair, cells independently alter their metabolic investment to maximize fitness. This leads to each cell specializing on a different metabolic process with- out apparent communication and cells share the metabolic products equally. In this way, metabolic adjustment appears to be synergistically cooperative, though non- altruistic. This synergism causes cell fitness to increase when in a multicellular pair, but there is no discussion of group-level fitness so we found it difficult to determine whether fitness decoupling is occurring in their model. However Ispolatov et al. do not speak of cooperation, even though it is clear that synergistic cooperation is present 45

Evolutionary Transitions in Individuality and Recent Models of Multicellularity 177 in the paired state. Modeling cell interaction in this way represents a rudimentary form of the cellular integration that is critical to the evolution of multicellularity (Folse and Roughgarden 2010; Clarke 2011). Multi-level selection is present in the Ispolatov et al. model, although they do not present it as such. Every cell experiences the same death rate and cells in groups have nearly identical birth rates (due to a small error term in metabolic investment). In comparison to solitary cells, cells in a pair have much higher fitness due to sharing of metabolic products and avoidance of the fitness cost associated with producing both products. Groups serve to produce population structure that affects the fitness of the cells. This context-dependent fitness means that group-level selection operates on pairs of cells to increase cell birth rate. However, the chemical mass action approach results in the lack of a canonical group level life cycle, in the sense that groups do not directly beget other groups. Because of the lack of a canonical group life cycle, one interpretation is that cells, not multicellular groups, are the only evolutionary individuals in this model. However, this ignores the facts that group selection is occurring and group properties are evolving in the model. Although it is difficult to see this model as explaining the evolution of multicellularity as an ETI, it is novel in explaining the origin of group structure and properties (Table 1). The model has the novel feature of applying to the early stages of an ETI as well as showing the division of labor can evolve early in an ETI. While the evolution of division of labor is observed, reproductive division of labor is not addressed by the model. With respect to division of labor, Ispolatov et al. (2012) model incompatible metabolic processes such as photosynthesis and nitrogen fixation in cyanobacteria, showing how non-reproductive division of labor may evolve. However, nitrogen fix- ing heterocysts in cyanobacteria do not reproduce (Kumar et al. 2010; Muro-Paster and Hess 2012). So while reproductive division of labor is present in cyanobacteria, it is not modeled by Ispolatov et al. (2012), which may call into question the ap- plicability of their results both to cyanobacteria and to ETIs with division of labor in reproductive functions. The observation that reproductive division of labor tends to precede other kinds of functional specialization (Simpson 2012) indicates that there is something special about reproductive specialization; whatever this is, it is not being addressed by this model.

Rueffler, Hermisson, and Wagner 2012

Rueffler et al. (2012) develop a mathematical model with general assumptions in order to identify general principles about the conditions that favor the evolution of division of labor. Their starting point is a group of two modules. Each module is characterized by a trait value, which is initially constrained to be the same (i.e., the organism is initially made up of undifferentiated parts). The trait value determines how well the module performs on two tasks. Task performance is constrained by a trade-off, such that performance at both tasks cannot be optimized simultaneously 46

178 E. R. Hanschen et al. by one module (or two undifferentiated modules). Fitness is taken to be an increasing function of performance of both tasks. Note that this is a single-level fitness concept like in Willensdorfer’s model, so Rueffler et al. (2012) is modeling the evolution of division of labor after an evolutionary transition to group living. The group of modules is the only unit that has fitness. Fitness increases as the ability of groups of modules to perform tasks increases. Rueffler et al. assume no differentiation in the ancestral state and that trait values had evolved to a fitness maximum based on performance constraints. They then ask, given an ancestor with undifferentiated trait values, what conditions favor differentiation of the modules? The main connection with the ETI theory, as Rueffler et al. (2012, p. E333) point out, is that the covariance effect described by Michod (2006) and Michod et al. (2006) can be seen as a special case of their observation that accelerating performance functions favor the evolution of differentiation. Michod (2006) and Michod et al. (2006) consider the trade-off between viability and fecundity as two contrasting basic categories of performance that combine multiplicatively to give fitness. Rueffler et al. consider a trade-off between performance of two tasks, meaning that increases in performance at one task come at the cost of decreases in performance of the other task. Performance at each task makes some positive contribution to fitness. They show that when the performance landscape is convex along the trait axes, specialization is favored. Specialization can lead to higher fitness in this case because the loss in performance at one task is due to one module’s deviation from the (constrained) optimal trait value can be more than compensated for if the other module deviates in the other direction. Aside from the shape of the performance function, Rueffler et al. also find that positional effects (in which one module is less good at a particular task due to its position) and synergistic interactions (when the joint contribution of two differentiated modules exceeds the sum of their separate contributions) can favor the evolution of specialization. Because larger organisms a priori have more distinct areas, there are more positional effects to select for many cell types. Also, note that the greater the number of cell types, the higher the heterogeneity of the organism, hence also the highest positional effects will be present. Rueffler et al.’s model highlights some commonalities underpinning the evolu- tion of functional specialization in a variety of different kinds of units (e.g., genes, appendages, cells, etc.). Their framework is so general that it does not touch on issues that may be more specific to division of labor during an ETI. For example, the Rueffler et al. model does not address what (if anything) is special about reproductive division of labor. By leaving aside the issue of how the traits affect fitness, Rueffler et al. also leave aside the insight that basic life history trade-offs present at lower levels can disappear at higher levels while new trade-offs emerge at the higher level.

Van Dyken and Wade 2012a, b

In a series of papers, Van Dyken and Wade (2012a, b) propose a resource-based model for the evolution of different kinds of altruism. The evolution of altruism is 47

Evolutionary Transitions in Individuality and Recent Models of Multicellularity 179 fundamental to ETIs, so we include these papers here, even though they focus on the evolution of sociality and do not specifically address the evolution of multicellu- larity. Van Dyken and Wade assume discrete and non-overlapping generations for a population that is sub-divided into groups. They present fitness of a focal individual as the product of two factors: one factor represents the fitness that the individual would have given infinite resources and the second factor accounts for the effects of local resource availability and efficiency of use (2012a, their Eq. A6). Group-mates affect the local resource availability and resource use of the focal individual. With respect to the evolution of division of labor, a major conclusion of Van Dyken and Wade’s work is that local resource pressure, which depends both on the external/physical environment and the behavior of group-mates, determines which specialized tasks evolve. For example, abundant resources favor specializations on tasks such as nest defense whereas scarce resources favor the evolution of specialized foragers. We think Van Dyken and Wade’s focus on the diverse ecological effects of altruistic behaviors begins to address an important shortcoming of some previous work on altruism. However, Van Dyken and Wade do not model the fundamental life history trade-off between viability and fecundity of a focal individual. They write: “vijfij is individual ij’s asymptotic fitness; that is, its resource-independent proba- bility of survival to maturity (viability, vij) times its maximum fecundity (fij)” (Van Dyken and Wade (2012a, p. 2487). However, viability and fecundity are unconnected parameters in this model, not functions of more basic variables like reproductive ef- fort or time spent at the two activities as is more commonly assumed in life history modeling. This is an interesting choice, and is probably appropriate for some cases. For example, an altruistic behavior such as alarm calling simply lowers the personal viability of the caller (and increases group fitness) without an increase in personal fe- cundity of the caller. However, in many other cases, a single behavior has contrasting effects on both personal viability and fecundity through a constrained resource such as time. For example, take a volvocine somatic cell that spends more time or energy beating its flagella than another cell. The fitness effect of this behavior would likely be a decrease in fecundity and an increase viability were it to live on its own. The behavior is costly (at the cell level) if the benefit of the behavior (increased cell-level viability) is insufficient to outweigh its costs (decreased cell-level fecundity). The cost of the behavior translates into a lack of balance in the two fitness components at the cell level. Survival-fecundity tradeoffs at the level of single individuals are common (see Roff and Fairbairn 2007), so considering this kind of situation is im- portant. The viability and fecundity terms in Van Dyken and Wade are independent parameters so they do not capture the idea that behaviors can be costly at the lower level by creating an imbalance of investment in lower level fitness components. Van Dyken and Wade’s classification of four types of altruism depends on their fitness function and the types of exchanges between individuals that are allowed, in particular, the lack of any interdependence of viability and fecundity. They write: “Altruism is typically modeled as increasing a recipients’survival or fecundity. How- ever, our model (Eq. 1) provides four different parameters that control fitness, each of which can be modified by altruism...” (Van Dyken and Wade 2012a, p. 2488). 48

180 E. R. Hanschen et al.

Although they see this approach as advantageous, we question whether the hierarchi- cal structure of the concepts involved is being ignored. That is, when one considers how increased resources increase a recipient’s fitness, it becomes clear that this in- crease must be channeled through benefits to the recipient’s fecundity, viability or both (assuming constant generation times). So it seems artificial to have four distinct ways in which altruistic behavior can benefit a partner (increased survival, increased fecundity, increased resources, increase resource use efficiency). Rather, why are there not two ways of affecting the partner’s resource (total amount or efficiency of use), which the partner then channels into personal viability and/or fecundity? The typical meaning of viability and fecundity (probability of survival to reproduction and number of offspring conditioned on survival) is closely tied to the idea that both of these features are connected through their resource-dependency. By contrast, Van Dyken and Wade use viability and fecundity to refer to resource-independent char- acteristics of organisms. It is not clear how these resource-independent notions of viability and fecundity should be applied to classifying social interactions. The very idea of one individual helping another individual to reproduce seems to require an exchange of resources; offspring are made of resources. Yet Van Dyken and Wade’s framework suggests that “fecundity altruism” occurs when one individual increases another’s maximum, resource-independent fecundity. The usefulness of this way of parsing interactions is not yet clear. In the Van Dyken and Wade approach, the “individual” fitness reflects both indi- vidual and group properties. (“Individual fitness was modeled as the physiological consequence of resource consumption in an environment composed of other con- sumers.” (Van Dyken and Wade 2012b, p. 2498)). Specifically, the amount of crowding is a property of the group that affects the “individual” fitness (Van Dyken and Wade 2012a, p. 2487, their Eq. 2). The focus on this kind of fitness function (i.e., “individual” fitness that includes group effects) makes the multi-level nature of the processes that are affecting trait change implicit rather than explicit. So while it seems likely that the Van Dyken and Wade models are examples of the evolution of traits that decouple lower- and higher-level fitness, this result is not explicit and is not explicitly discussed in the papers. For example, a somatic cell in a colony decouples fitness at the two levels because this trait would be detrimental in a global population of cells but can be favored by selection among groups of cells. Thus, when somatic cells evolve, it is clear that the groups of cells with high fitness are not simply groups that contain the cells that would do best in a global population. Note that the realized, direct fitness of cells within a group (i.e., the kind of fit- ness in the Van Dyken and Wade papers) is not helpful here. Fitness is decoupled when the strictly single-level fitness that a cell would have declines and the fitness of the group increases. The same logic applies to all of the types of altruism that Van Dyken and Wade propose, suggesting that these models investigate the evolution of altruism after an evolutionary transition in individuality. One consequence of their work is the possibility that contrasting kinds of altruism can increase together via co-evolutionary niche construction (Van Dyken and Wade 2012b). With respect to division of labor, the ETI framework focuses on the lower-level trade-off between vi- ability and fertility. However, Van Dyken and Wade explicitly assume that these two 49

Evolutionary Transitions in Individuality and Recent Models of Multicellularity 181 components of fitness are independent. Van Dyken and Wade (2012b, p. 2509) sug- gest that natural selection would act so as to “partition tasks into survival/fecundity and resource task specialists in proportions that most efficiently cope with negative environmental feedback at the colony level.” This is an interesting and novel idea that certainly could be explored more with respect to the multicellularity ETI.

Discussion

We have discussed how each model relates to five components of ETIs: the evolution of cooperation in groups, multi-level selection, life history trade-offs, division of labor, and decoupling of fitness (see Table 1). While not all the models reviewed here cleanly relate to this ETI framework, none conflict with it, and they can all be understood as contributing new information to the different components. Of course, each of these works has a scope defined by their own objectives, so the observation that not all components are included in each model is not meant as a criticism. Two major themes have emerged in our analysis of these models, (i) the role of life history traits and life cycles in evolutionary transitions and (ii) what the models assume and purport to explain, in particular, whether groups are assumed to already have the properties we wish to explain.

Life History Traits and Life Cycles

The evolution of multicellularity involves the evolution of a group life cycle from the life cycles of single cells. Thus it is natural to consider life history traits and how they change and are reorganized at the two levels during this ETI. Ignoring the critical role life history traits play in defining fitness creates ambiguities in comparing and understanding definitions of fitness. Rueffler et al. (2012), in the search for generality, did not assume explicit life history traits like survival and reproduction in their model; the detachment of this model from life history traits and reproductive specialization limits its application to ETIs. Van Dyken and Wade (2012a, b) do not explicitly model how resources affect viability and/or fecundity. They also ignore the most fundamental aspect of life history traits by assuming that viability and fecundity are independent and do not trade off in a focal individual. Because of these assumptions, it was difficult to relate their definition of fitness and types of altruists to the ETI framework, although their most general point, that resource use needs to be explicitly modeled, should be pursued in the context of ETIs. We recognize the difficulty of defining fitness in a situation where the very unit of evolution and hence fitness is changing, but this is the challenge that must be met in understanding an ETI. When fitness is based on life history traits, a tractable conception of fitness results as the level of fitness changes. Novel definitions of 50

182 E. R. Hanschen et al.

fitness (such as Willensdorfer (2009) who defined fitness as biomass production) may be interesting, but make it difficult to connect these definitions to other models or to the biological cases we wish to understand. Care must be taken to ensure fitness definitions are biologically tractable and applicable. Most of the recent models (Willensdorfer 2009; Gavrilets 2010; Rueffler et al. 2012; Van Dyken and Wade 2012a, b) along with the earlier modifier and optimality models, assume the initial existence of group-level life cycles. This leaves a critical gap in our understanding of an ETI. How might a group life cycle emerge from unicellular cycles? Life cycles can be thought of in terms of a few fundamental elements (Fig. 2). We consider the asexual life cycles of three different kinds of species of volvocine green algae—a diverse group of photosynthetic eukaryotes ranging from unicellular to complex multicellular forms (see Chapter “Volvocine Algae: From Simple to Complex Multicellularity”): Chlamydomonas reinhardtii (unicellular), Basichlamys sacculifera or Tetrabaena socialis (colonial with group size 2–4) and Gonium pectorale (colonial with group size 4–16). In C. reinhardtii,a parent cell (1c) grows and then (2c) divides through several rounds of synchronous cell divisions. Finally, offspring cells (3c) separate from the parent and from each other. In the undifferentiated multicellular volvocine algae such as G. pectorale, we see the same three basic life cycle elements applying to groups of cells on the right- hand-side of Fig. 2. A group of cells, i.e. colony, (1g) grows; next, divisions occur (2g). Although divisions are at the cell level, each cell undergoes divisions while still being a group member. Whereas cell divisions in the cell cycle (2c) are immediately responsible for the number of offspring cells, cell divisions in the group cycle (2g) correspond to the number of cells in the adult group, that is, the adult group’s body size. The major elements of the asexual life cycle are similar in the unicellular Chlamydomonas and the multicellular Gonium (i.e. 1c-2c-3c is similar to 1g-2g-3g). However, note that these two cycles (Fig. 2, left side and right side, respectively) are distinct and share no states in common. How might evolution get from one to the other? The earliest-branching colonial volvocine species (B. sacculifera and T. socialis) have a hybrid pattern that may represent a transitional stage between a cell-level and group-level life cycle (Fig. 2, grey lines). In these species, among the simplest of known multicellular forms, a colony undergoes growth (1g), then the cells of the colony separate from each other (3c), and then the cells undergo cell division (2c). From a unicellular ancestor similar to Chlamydomonas,aBa- sichlamys-like life cycle could arise from a change in timing of the cell separation state. Step 3c (separation) fails to occur at the normal time (see connection between C and D’). Instead, cells separate later (see connection between E and B), after a period of growth within the colony. A mutation that causes temporary adherence of offspring cells to the mother cell wall material could create a Basichlamys-like cycle from a unicellular cycle. If so, then growth as a colony was the first aspect of a fully colonial life cycle to evolve in volvocine algae. This would suggest that the effects of colony-level traits (e.g. colony cell number, overall colony size) on growth could form the basis for specifically group-level selection very early in the 51

Evolutionary Transitions in Individuality and Recent Models of Multicellularity 183

Fig. 2 Left side: The three major elements of a single-celled asexual cycle. (1c) growth, (2c) division, (3c) separation. Right side: The same three elements are applicable in the same order in a simple colonial (group) life cycle (e.g. as seen in G. pectorale). Note the cell and group life cycles are distinct and do not have states in common. Gray lines:Thesimplestcolonialvolvocinealgae(B. sacculifera and T. socialis) show a hybrid asexual cycle in which growth (1g), but not division (2c) or separation (3c), occurs while cells are in groups. Throughout this figure, the number of rounds of division is shown as two. In reality, this value can vary based on external conditions 52

184 E. R. Hanschen et al. transition to multicellularity. Modeling these issues in such simple life cycles can help us understand how the level of adaptation changes from the cell to the group during a transition in individuality (Shelton 2013; Shelton and Michod 2014).

Modeling Approaches and Challenges

The second theme to emerge from our analysis regards the relationship between the initial assumptions and what the model seeks to explain. In any mathematical model, the conclusions are logical outcomes of the assumptions, some of which may be implicit. Evolutionary transitions in individuality pose unique problems to the modeler, because if natural selection is to play a role in the transition, then the units of evolution are both the explanandum (phenomenon to be explained) and the explanans (the explanation of the phenomenon). We seek to explain a new higher- level unit of evolution by making assumptions about the lower level units and how they interact. The multilevel selection approach determines when the magnitude of selection increases at the higher level and decreases at the lower level. We have argued that several of the models studied here either explicitly or implicitly use this approach (Gavrilets 2010; Ispolatov et al. 2012; Van Dyken and Wade 2012a, b). The issue of assuming pre-existing group properties in a model meant to explain group features is not unique to the recent models, for example, the modifier mod- els discussed above assume the prior existence of group-level traits such as group structure and cooperation. With these assumptions, the modifier models studied the evolution of traits promoting individuality at the group level, such as germ line segre- gation and policing, which were assumed to be properties of an introduced modifier allele. The model predicted the conditions under which the modifier allele increased in frequency and the effect of the modifier allele on levels of cooperation, heritability of fitness, and individuality at the group level. Similar challenges exist in modeling division of labor. As Ispolatov et al. (2012) point out and we discussed above, Gavrilets’ (2010) model of division of labor ini- tially assumes many of the antecedents of division of labor. To help us understand this, consider again the modifier models in which there were two loci, the cooper- ate/defect locus and the modifier locus. The modifier locus changed an aspect of the life history or development, such as mutation rate, germ soma division of labor, policing, single cell bottlenecks, or genetic control of group size, which in turn in- creased the fitness of the already more fit C allele. Consequently, the evolution of division or labor at the modifier locus and its effect on individuality emerged out of an interaction between the two loci. The modifier approach allows us to track the genetic and selective factors involved in the evolution of division of labor. Gavrilets (2010) sets the proportion of soma precursor cells arbitrarily to be 25 % and the conditions for which division of labor evolves via the regulatory loci are de- termined. However we do not know whether or in what way these conditions depend upon the arbitrary figure of 25 % somatic cell precursors (although simulations were also done with 75 % somatic precursors). A modifier approach would assume that 53

Evolutionary Transitions in Individuality and Recent Models of Multicellularity 185 the proportion of soma precursors was encoded by a third locus that was initially set at 0 %. In this case, the evolution of division of labor would emerge out of an interaction between the regulatory loci and the precursor locus and not as a result of a specific preordained frequency of soma precursors. These challenges highlight the need for careful interpretation of the model. Here, we have discussed how the interpretation of a model can lead to substantial differences in understanding the causal factors underlying the evolution of multicellularity. The clearest examples of this were Gavrilets’suggestion that selection solely at the level of cells can drive the evolution of division of labor in multicellular groups and Ispolatov et al.’s suggestion that multicellularity can evolve without cooperation or increased cellular integration.

Future Directions

The papers we have reviewed here are largely focused on the later stages of ETIs, specifically the evolution of division of labor (Step v, Table 1). One exception to this was Ispolatov et al. (2012), which focused on initial group formation and increasing synergistic cooperation (Steps i and ii). Previous ETI models largely focused on con- flict mediation, division of labor among fitness components and fitness decoupling (Steps iii–vi). In comparison, there has been little recent research on group formation and the evolution of group life cycles as discussed in the previous section. In modifier models, the life cycle is summarized in terms of heritability of group traits but a more explicit treatment of the components of the life cycle and how they are reassembled at the level of the group has not yet been achieved. This represents acriticalgapinthetheorygiventheemphasisonlifehistorytraitsintheETIframe- work. Recent empirical (e.g. Rainey and Rainey 2003; Ratcliff et al. 2012, 2013), conceptual (Libby and Rainey 2013) and theoretical work (Shelton 2013; Shelton and Michod 2014) studying the evolution of group life cycles will provide ideas for this future work to address. Furthermore, we suspect a better theoretical understanding of group-level life cycles will integrate the ETI framework with the multi-level selec- tion (MLS) framework (Damuth and Heisler 1988; Okasha 2006), in which Okasha has defined the evolution of individuality as the transition among different kinds of MLS (MLS1 and MLS2) (Okasha 2006). This is an important step to integrating the large body of research on MLS1 and MLS2 with the ETI framework. We also suggest that modeling non-canonical systems would be valuable for our understanding of ETIs. Many recent models are implicitly or explicitly modeling a canonical model system (multicellular development from a single celled propagule). While it is important to understand this form of multicellularity, there are other, albeit rare, pathways to multicellularity. These pathways deserve mathematical investiga- tion to more fully appreciate what factors predispose a system to evolve towards one kind of multicellularity or another and why certain kinds of multicellularity are rare. We envision models investigating multicellularity via aggregation (such as the slim mold, Dictyostelium; Chapter “The Evolution of Developmental Signalling in 54

186 E. R. Hanschen et al.

Dictyostelia from an Amoebozoan Stress Response”) and species with unique life histories (such as the red alga, Porphyra). Such models may investigate the evolution- ary consequences of alternative pathways. Major differences, such as the difference between groups formed by aggregation (“coming together”, Tarnita et al. 2013) and groups formed by reproductive products staying together (“staying together”, Tar- nita et al. 2013), may relate to the generalized step (i) group formation differently. This distinction may also affect the timing of step (iii) cheating and conflict, as relatives that stay together may be less prone to conflict than genetically distinct individuals that aggregate. Ispolatov et al. (2012) is the only model reviewed here which models multicellularity via aggregation rather than via development. As such, this model serves as an important first step to understanding alternative pathways to multicellularity. Finally, we find it interesting that each paper has taken a different mathematical approach, made different biological assumptions, and has made new contributions to our understanding of the evolution of multicellularity. This suggests that our theoretical understanding of the evolution of multicellularity is far from complete. It seems clear that continued theoretical exploration of this subject would be highly valuable to our understanding of multicellularity and ETIs more generally.

Summary

1. An evolutionary transition in individuality (ETI) is a fundamental shift in the unit of selection, when a group of individuals become a new, higher-level individual. ETIs are thought to be responsible for the origin of the hierarchy of life: genes, chromosomes, cells, cells within cells (eukaryotic cell), multicellular organisms, and societies. 2. We discuss five interrelated components of ETIs: the evolution of cooperation in groups, multi-level selection, life history trade-offs, division of labor among the basic components of fitness, and decoupling of fitness of the new, higher-level unit from the fitnesses of the lower-level units. 3. We review the two-locus modifier models and optimality models that provide the basis for the ETI framework emphasizing the role of lower-level fitness trade-offs. 4. We discuss how six recently published papers (Willensdorfer 2009; Gavrilets 2010; Ispolatov et al. 2012; Rueffler et al. 2012; Van Dyken and Wade 2012a, b) relate to the five components of the ETI framework. 5. We identify life history traits and life cycles as critical considerations in under- standing ETIs that have been missing in some recent mathematical models. We show how group life cycles may emerge from cell life cycles in the volvocine green algae model system. 6. Incorporating recent papers into the ETI framework reveals several less un- derstood aspects of ETIs, which we suggest as directions for future research, including the evolution of group life cycles, how alternative frameworks relate to ETIs, and the need to model non-canonical systems. 55

Evolutionary Transitions in Individuality and Recent Models of Multicellularity 187

Acknowledgements This research was supported by the National Science Foundation under Grant No. DEB-0742383 to R. E. Michod, the National Science Foundation under Grant No. DGE- 0654435 to the University ofArizona, and the NationalAeronautics and SpaceAdministration under Grant No. NNX13AH41G to R. E. Michod. We would like to thank J. D. Van Dyken, M. Doebeli, A. M. Nedelcu, and two anonymous reviewers for their helpful comments on the manuscript.

References

Buss LW (1987) The evolution of individuality. Princeton University, Princeton Clarke E (2011) The problem of biological individuality. Biol Theory 5:312–325 Damuth J, Heisler IL (1988) Alternative formulations of multilevel selection. Biol and Philos 3:407–430 Folse HJ III, Roughgarden J (2010) What is an individual? A multilevel selection perspective. Q Rev Biol 85:447–472 Gavrilets S (2010) rapid transition towards the division of labor via evolution of developmental plasticity. PLoS Comput Biol 6:e1000805 Griesemer J (2000) The units of evolutionary transition. Selection 1:67–80 Ispolatov I,Ackermann M, Doebeli M (2012) Division of labour and the evolution of multicellularity. Proc Biol Sci 279:1768–1776 Kumar K, Mella-Herrera RA, Golden JW (2010) Cyanobacterial heterocysts. Cold Spring Harb Perspect Biol 2:a000315 Libby E, Rainey PB (2013) A conceptual framework for the evolutionary origins of multicellularity. Phys Biol 10:035001 Maynard Smith J (1988) Evolutionary progress and levels of selection. In: Nitecki MH (ed) Evolutionary progress. University of Chicago Press, Chicago, pp 219–230 Maynard Smith J (1991) A Darwinian view of symbiosis. In: Margulis L, Fester R (eds) Symbiosis as a source of evolutionary innovation. MIT Press, Cambridge, pp 26–39 Maynard Smith J, Szathmáry E (1995) The major transitions in evolution. W.H. Freeman, San Francisco Michod RE (1996) Cooperation and conflict in the evolution of individuality. II. Conflict mediation. Proc: Biol Sci 263:813–822 Michod RE (1997) Cooperation and conflict in the evolution of individuality. I. Multi-level selection of the organism. Am Nat 149:607–645 Michod RE (1999) Darwinian dynamics, evolutionary transitions in fitness and individuality. Princeton University Press, Princeton Michod RE (2003) Cooperation and conflict mediation during the origin of multicellularity. In: Hammerstein P (ed) Genetic and cultural evolution of cooperation. MIT Press, Cambridge, pp 261–307 Michod RE (2005) On the transfer of fitness from the cell to the multicellular organism. Biol Philos 20:967–987 Michod RE (2006) The group covariance effect and fitness trade-offs during evolutionary transitions in individuality. Proc Natl Acad Sci U S A 103:9113–9117 Michod RE (2007) Evolution of individuality during the transition from unicellular to multicellular life. Proc Natl Acad Sci U S A 104(1):8613–8618 Michod RE, Nedelcu AM (2003) Cooperation and conflict during the unicellular-multicellular and prokaryotic-eukaryotic transitions. In: Moya A, Font E (eds) Evolution: from molecules to ecosystems. Oxford University Press, Oxford, pp 195–208 Michod RE, Roze D (1997) Transitions in individuality. Proc: Biol Sci 264:853–857 Michod RE, Roze D (1999) Cooperation and conflict in the evolution of individuality. III. Transitions in the unit of fitness. In: Nehaniv CL (ed) Mathematical and computational biology: computa- tional morphogenesis, hierarchical complexity, and digital evolution. American Mathematical Society, Providence, pp 47–92 56

188 E. R. Hanschen et al.

Michod RE, Roze D (2001) Cooperation and conflict in the evolution of multicellularity. Heredity 81:1–7 Michod RE, Nedelcu AM, Roze D (2003) Cooperation and conflict in the evolution of individuality IV. Conflict mediation and evolvability in Volvox carteri. BioSystems 69:95–114 Michod RE, Viossat Y, Solari CA, Hurrand M, Nedelcu AM (2006) Life history evolution and the origin of multicellularity. J Theor Biol 239:257–272 Muro-Pastor AM, Hess WR (2012) Heterocyst differentiation: from single mutants to global approaches. Trends Microbiol 20:548–557 Nedelcu AM, Michod RE (2006) The evolutionary origin of an altruistic gene. Mol Biol Evol 23:1460–1464 Okasha S (2006) Evolution and the levels of selection. Oxford University Press, Oxford Otto SP,Orive ME (1995) Evolutionary consequences of mutation and selection within an individual. Genetics 3:1173–1187 Rainey PB, Rainey K (2003) Evolution of cooperation and conflict in experimental bacterial populations. Nature 425:72–74 Ratcliff WC, Denison RF, Borrello M, Travisano M (2012) Experimental evolution of multicellu- larity. Proc Natl Acad Sci U S A 109:1595–1600 Ratcliff WC, Herron MD, Howell K, Pentz JT, Rosenzweig F, Travisano M (2013) Experimental evolution of an alternating uni- and multicellular life cycle in Chlamydomonas reinhardtii. Nat commun 4:2742 Roff DA (2002) Life history evolution. Sinauer Associates, Sunderland Roff DA, Fairbairn DJ (2007) The evolution of trade-offs: where are we? J Evol Biol 20:433–447 Rueffler C, Hermisson J, Wagner GP (2012) Evolution of functional specialization and division of labor. Proc Natl Acad Sci U S A 109:E326–E335 Shelton DE (2013) The shifting role of cell division during an evolutionary transition to multicellular-level individuality. Doctoral Dissertation, University of Arizona Shelton DE, Michod RE (2014) Group selection and group adaptation during a major evolutionary transition: insights from the evolution of multicellularity in the volvocine algae. Biological Theory 9:452–469 Simpson C (2012) The evolutionary history of division of labor. Proc: Bio Sci 279:116–121 Sober E,Wilson DS (1998) Unto others: the evolution and psychology of unselfish behavior. Harvard University Press, Cambridge Solari CA, Kessler JO, Michod RE (2006) A hydrodynamics approach to the evolution of multicel- lularity: flagellar motility and germ-soma differentiation in volvocalean green algae. Am Nat 167:537–554 Stearns SC (1992) The evolution of life histories. Oxford University Press, Oxford Tarnita CE, Taubes CH, Nowak MA (2013) Evolutionary construction by staying together and coming together. J Theor Biol 320:10–22 Van Dyken JD, Wade MJ (2012a) Origins of altruism diversity I: the diverse ecological roles of altruistic strategies and their evolutionary responses to local competition. Evolution 66:2484– 2513 Van Dyken JD, Wade MJ (2012b) Origins of altruism diversity II: runaway coevolution of altruistic strategies via “reciprocal niche construction”. Evolution 66:2498–2513 Willensdorfer M (2008) Organism size promotes the evolution of specialized cells in multicellular digital organisms. J Evol Biol 21:104–110 Willensdorfer M (2009) On the evolution of differentiated multicellularity. Evolution 63:306–323 57

APPENDIX B. THE GONIUM PECTORALE GENOME DEMONSTRATES CO-OPTION OF

CELL CYCLE REGULATION DURING THE EVOLUTION OF

MULTICELLULARITY

Hanschen E. R., T. N. Marriage, P. J. Ferris, T. Hamaji, A. Toyoda, A. Fujiyama, R. Neme, H.

Noguchi, Y. Minakuchi, M. Suzuki, H. Kawai-Toyooka, D. R. Smith, H. Sparks, J.

Anderson, R. Bakarić, V. Luria, A. Karger, M. W. Kirschner, P. M. Durand, R. E.

Michod, H. Nozaki, B. J. S. C. Olson. 2016. The Gonium pectorale genome demonstrates

co-option of cell cycle regulation during the evolution of multicellularity. Nature

Communications 7:11370.

58

ARTICLE Received 26 Jan 2016 | Accepted 18 Mar 2016 | Published 22 Apr 2016 DOI: 10.1038/ncomms11370 OPEN The Gonium pectorale genome demonstrates co-option of cell cycle regulation during the evolution of multicellularity

Erik R. Hanschen1, Tara N. Marriage2, Patrick J. Ferris1, Takashi Hamaji3, Atsushi Toyoda4,5, Asao Fujiyama4,5, Rafik Neme6, Hideki Noguchi4, Yohei Minakuchi5, Masahiro Suzuki7, Hiroko Kawai-Toyooka7, David R. Smith8, Halle Sparks2, Jaden Anderson2, Robert Bakaric´9, Victor Luria10,11, Amir Karger12, Marc W. Kirschner10, Pierre M. Durand1,13,14, Richard E. Michod1,11, Hisayoshi Nozaki7 & Bradley J.S.C. Olson2,11

The transition to multicellularity has occurred numerous times in all domains of life, yet its initial steps are poorly understood. The volvocine green algae are a tractable system for understanding the genetic basis of multicellularity including the initial formation of coop- erative cell groups. Here we report the genome sequence of the undifferentiated colonial alga, Gonium pectorale, where group formation evolved by co-option of the retinoblastoma cell cycle regulatory pathway. Significantly, expression of the Gonium retinoblastoma cell cycle regulator in unicellular Chlamydomonas causes it to become colonial. The presence of these changes in undifferentiated Gonium indicates extensive group-level adaptation during the initial step in the evolution of multicellularity. These results emphasize an early and formative step in the evolution of multicellularity, the evolution of cell cycle regulation, one that may shed light on the evolutionary history of other multicellular innovations and evolutionary transitions.

1 Department of Ecology and Evolutionary Biology, University of Arizona, Tucson, Arizona 85712, USA. 2 Division of Biology, Kansas State University, Manhattan, Kansas 66506, USA. 3 Donald Danforth Plant Science Center, St Louis, Missouri 63132, USA. 4 Center for Advanced Genomics, National Institute of Genetics, Mishima, Shizuoka 411-8540, Japan. 5 Center for Information Biology, National Institute of Genetics, Mishima, Shizuoka 411-8540, Japan. 6 Department of Evolutionary Genetics, Max Planck Institute for Evolutionary Biology, 24306 Plo¨n, Germany. 7 Department of Biological Sciences, Graduate School of Science, University of Tokyo, Bunkyo-ku, Tokyo 113-003, Japan. 8 Department of Biology, Western University, London, Ontario, Canada N6A 5B7. 9 Division of Molecular Biology, RuXer Bosˇkovic´ Institute, Zagreb 10000, Croatia. 10 Department of Systems Biology, Harvard Medical School, Boston, Massachusetts 02115, USA. 11 Kavli Institute for Theoretical Physics, University of California Santa Barbara, Santa Barbara, California 93106, USA. 12 Research Computing Division, Harvard Medical School, Boston, Massachusetts 02115, USA. 13 Department of Molecular Medicine and Haematology, Faculty of Health Sciences and Evolutionary Studies Institute, University of the Witwatersrand, Johannesburg 2000, South Africa. 14 Department of Biodiversity and Conservation Biology, University of the Western Cape, Cape Town 7535, South Africa. Correspondence and requests for materials should be addressed to E.R.H. (email: [email protected]) or to B.J.S.C.O. (email: [email protected]).

NATURE COMMUNICATIONS | 7:11370 | DOI: 10.1038/ncomms11370 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11370 59

ulticellular organisms have independently evolved complexity, Chlamydomonas and Volvox, suggests few numerous times throughout the tree of life including genetic changes are required17, but it is unclear how and when Mplants, animals, fungi, cyanobacteria, amoeba, brown the genes important for multicellularity evolved during these 12 algae, red algae and green algae1,2. In animals, multicellularity steps14. emerged 600–950 million years ago (Myr ago) correlating with a By sequencing the genome of the undifferentiated, chlorophy- large expansion of genes encoding transcription factors, signalling cean Gonium pectorale, a species without a differentiated pathways, and cell adhesion genes that were co-opted from their ancestor16, we find co-option of cell cycle regulation, which unicellular ancestors3,4 Similarly, multicellular, terrestrial plants occurred during the initial transition to cell groups, as the genetic emerged B750 Myr ago correlating with an expansion of many basis for the evolution of multicellularity. The cell cycle regulation signalling pathways present in their unicellular relatives5,6. found in undifferentiated Gonium, co-opted in a multicellular However, because most multicellular lineages have long context and shared with germ–soma-differentiated Volvox, diverged from their unicellular relatives, the genomic signature indicates group-level adaptations in undifferentiated colonies. of the transition to multicellularity has been obscured, and The early co-option of cell cycle regulation for group-level life consequently this evolutionary process remains enigmatic. cycle and reproduction is a critical and formative step in the The volvocine green algae are a unique model system for the evolution of multicellularity. evolution of multicellularity because the unicellular ancestry is clear, the emergence of multicellularity occurred B230 Myr ago, Results and species exhibit a stepwise increase in morphological Genomic comparisons of volvocine algae. At the genomic level, complexity ranging from undifferentiated colonies to differen- 7,8 the genomes of Chlamydomonas, Gonium and Volvox are similar, tiated multicellular species (Fig. 1, Supplementary Figs 1,2). though various measures of genome compactness correlate with Unicellular Chlamydomonas reinhardtii is thought to resemble cell number, consistent with a long-term increase in organismal the unicellular ancestor of multicellular volvocines, including size18,19. Chlamydomonas and Gonium have similar GC content undifferentiated Gonium pectorale and differentiated Volvox near 64%, while Volvox has 56% GC (Table 1). Otherwise, carteri (Fig. 1). Chlamydomonas, Gonium and Volvox have decreasing gene Chlamydomonas undergoes a variant cell cycle (Supplementary densities of 159.6, 120.9 and 113.7 genes per megabase, with an Fig. 2), regulated by homologues of the retinoblastoma cell increasing average intron length of 279, 349 and 500 base pairs, cycle pathway, termed multiple-fission where it divides by a series 9–11 respectively. Although intron length increases with organismal of rapid cell divisions producing individual daughter cells . complexity, the number of introns per gene (Chlamydomonas, Gonium typically forms 8- or 16-celled undifferentiated colonies, 7.46; Gonium, 6.5; Volvox 6.8, Table 1) does not. GC content, where each constituent cell resembles a Chlamydomonas cell intron length and gene density correlate with morphological (Fig. 1). Gonium also undergoes multiple fission forming daughter complexity. colonies by keeping cells attached after multiple-fission, suggesting 12 We next examined genome-wide evolution in all three species either cell cycle regulation , or cell–cell adhesion has been to better understand the genetic basis for the evolution of modified to promote multicellularity. In Gonium, like 20 13,14 multicellularity. A prediction of lineage-specific genes shows Chlamydomonas, growth and cell division are uncoupled . few genes correlate with the evolution of multicellularity in the Asexual juvenile Gonium colonies grow (without cell division) into Volvocales (phylostratum 7; PS7) with a maximum of 180–357 adults. After cell division through multiple-fission, juvenile genes (Fig. 2a). This suggests that the evolution of multicellularity colonies hatch forming 8 or 16 daughter colonies of 8 or 16 cells does not rely upon the evolution of de novo genes. Though gene (Supplementary Fig. 2). Volvox contains approximately 2,000 regulation may be important during multicellular innovation, the small, terminally differentiated, somatic cells on the surface of the diversity and abundance of transcription factors is similar in spheroid and approximately 16 large reproductive cells embedded Chlamydomonas, Gonium and Volvox (Fig. 2b, Supplementary in extracellular matrix (ECM) inside the spheroid (Fig. 1). Volvox Table 1, Supplementary Fig. 3). Although enrichment of also has modified multiple fission where germ–soma separation is 8,14 transcription factors can correlate with the evolution of established after an asymmetric cell division . multicellularity3, this is not always the case21; we found that The transition to multicellularity in the Volvocales was thought 15,16 Gonium and Volvox have fewer transcription factors than in to involve at least 12 steps (Supplementary Fig. 1) though Chlamydomonas (Fig. 2b). These include PHD domains the genetic basis of these steps remains enigmatic. Genomic comparison of the extremes of morphological

Table 1 | Summary statistics for genome level analyses for abChlamydomonas, Gonium and Volvox. Chlamydomonas reinhardtii Tetrabaena socialis Characteristic Chlamydomonas Gonium Volvox Volvox Gonium pectorale v5.3 v1 v2 C E S Astrephomene gubernaculifera Genome size (Mb) 111.1 148.8 137.8 131.1 Pandorina morum Scaffold N50 (Mb) 7.78 1.27 1.49 2.6 S Volvox ferrisii Number of contigs/ 54 2,373 1,265 434 scaffolds E Yamagishiella unicocca % G and C 64.1 64.5 56.0 56.1 Eudorina elegans S Evolution of somatic cells Protein coding loci 17,737 17,984 15,669 14,971 Pleodorina starrii C Evolution of cell cycle control Gene density 159.6 120.9 113.7 114.1 S Volvox carteri E Evolution of expanded ECM (genes/Mb) Introns/gene 7.46 6.50 6.78 6.29 Figure 1 | Volvocine phylogenetic tree. (a) Evolution of cell cycle control Average intron 279.17 349.83 496.67 399.50 (C), expanded ECM (E) and somatic cells (S) are denoted. (b) Micrographs length (bp) of Chlamydomonas (green; scale bar, 10 mm), Gonium (blue; scale bar, 10 mm) % genes w/introns 92.4 92.6 82.8 84.0 and Volvox (black; scale bar, 25 mm) show morphological differences.

2 NATURE COMMUNICATIONS | 7:11370 | DOI: 10.1038/ncomms11370 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11370 ARTICLE 60

ab

C. reinhardtii: PS9 2,748 Chlamydomonas: PS8 0 0 0.5 1 Volvocine Chlamydomonadaceae: PS7 0 Normalized algae Chlamydomonadales: PS6 4,468 Chlorophyceae: PS5 3 gene count Chlorophyta: PS4 676 Viridiplantae: PS3 526 PHD– phylostratum Eukaryota: PS2 3,866 Chlamydomonas – C3H Cellular organisms: PS1 7,103 – SAP G. pectorale: PS9 1,334 – FHA 0 Gonium: PS8 + TRAF Goniaceae: PS7 357 Chlamydomonadales: PS6 4,386 + SBP Chlorophyceae: PS5 3 – ARID Chlorophyta: PS4 702 Gonium – HB-other Viridiplantae: PS3 506 phylostratum Eukaryota: PS2 3,883 – YABBY Cellular organisms: PS1 6,709 – NF-YA V. carteri: PS9 2,887 – WOX Volvox: PS8 0 + SAND 180 Volvocaceae: PS7 – TIG Chlamydomonadales: PS6 2,417 Chlorophyceae: PS5 3 Correlates with multicellularity – C2H2 458 Volvox Chlorophyta: PS4 – GNAT Viridiplantae: PS3 371 phylostratum – SNF2 Eukaryota: PS2 2,828 Cellular organisms: PS1 5,065 – MYB 0 7,000 Genes Volvox Chlorella Gonium Micromonas Coccomyxa c Ostreococcus Bathycoccus Families Families Chlamydomonas lost gained Multicell. LCA 206 396 d Chlamydomonas

Volvox 1,499 4,804 25

Gonium 664 5,473 27 15 58 Chlamydomonas 171 5,357 32 44 Volvocine LCA 660 3,504 Gonium 9 Volvox

Figure 2 | Genome similarity in the Volvocales. (a) Predicted number of genes in each phylostratum (PS1–PS9) for Chlamydomonas, Gonium and Volvox. (b) Heatmap of transcription factor abundance for all green algae. Significant over- ( ) and under-representation ( ) in colonial/multicellular lineages þ À (Gonium and Volvox) is denoted (G test of independence, a 0.05). Rows are clustered (left), an accepted phylogeny is depicted (top). (c) Phylogenetic ¼ analysis of gene family evolution. Bars to the left and right of the vertical axis denote the lost and gained gene families respectively, relative to its parental node. (d) Venn diagram of the species distribution of Pfam A domains unique to the volvocine algae.

(transcription and chromatin binding), DNA binding trans- adaptations are more numerous than changes correlating with cription factors and histones (Fig. 2b, Supplementary Tables 2 the evolution of multicellularity. We observe more evidence of and 3, Supplementary Fig. 3). species-specific, rather than multicellular-specific, protein inno- Using all published chlorophyte green algae genomes, we vations, suggesting species-specific adaptation (Fig. 2a,c,d, constructed Markov-based gene families (Chlamydomonas, 73%; Supplementary Tables 5–8) rather than genome-wide differences Gonium, 73%; Volvox, 70% of genes in gene families of size correlating with the evolution of multicellularity. The evolution of greater than one). Compared with 2,844 net gene families gained, multicellularity in the volvocine algae does not require large-scale which correlate with the origin of the Chlorophyceae, a genomic innovation. phylogenetic analysis of these gene families suggests little protein innovation (110 net gene families) during the evolution of multicellularity (Fig. 2c). These same green algae genomes Co-option of cell cycle regulation for multicellularity. Notably, allowed analysis into Pfam domain innovation, which may we observe that the genetic innovation correlating with multi- correlate with the evolution of multicellularity. We found cellularity, shared between Gonium and Volvox, evolved through innovation of only nine Pfam domains correlating with the co-option of existing developmental programs of cell cycle evolution of multicellularity (Fig. 2d, Supplementary Table 4). control. Volvocine algae have a common multiple-fission life Moreover, multicellular algae (Gonium and Volvox) have reduced cycle, with variation in timing and number of divisions Pfam domain diversity and abundance (compared with nine (Supplementary Fig. 1)14. Like most eukaryotes, including plants unicellular green algae, Supplementary Data 1); 394 Pfam and animals, their cell cycles are regulated by homologues A domains are significantly under-represented versus 129 over- of the retinoblastoma cell cycle regulatory pathway (Fig. 3, represented Pfam domains (Supplementary Fig. 4). Interestingly, Supplementary Figs 5–8, Supplementary Table 9)9,10, in which there is an excess of species-specific genes (Fig. 2a,c) and Pfam cyclin-dependent kinases (CDKs) bind cyclin proteins to domains (Fig. 2d) compared with multicellularity-correlated phosphorylate and regulate retinoblastoma (RB or MAT3 in the genes and Pfam domains, suggesting that species-specific Volvocales), which in turn de-represses the cell cycle (Fig. 3a).

NATURE COMMUNICATIONS | 7:11370 | DOI: 10.1038/ncomms11370 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11370 61

a Gene Cell Cell size CDKs+Cyclins MAT3/RB E2F1/DP1 transcription division

e E2F1 Chlamydomonas b CDK E1 Gonium cd96 CYCAB1 Chlamydomonas MAT3/RB Volvox female CDK E1 Chlamydomonas CycAB1 Volvox 100 E2F1 Gonium 100 100 100 MAT3/RB Volvox male CDK E1 Volvox CYCAB3 Gonium 94 E2F1 Volvox CDK I1 Gonium 95 CYCAB2 Gonium MAT3/RB Gonium DP1 Chlamydomonas CDK I1 Chlamydomonas 100 CYCAB1 Gonium 100 100 MAT3/RB Chlamydomonas 96 100 CDK I1 Volvox 100 CycB1 Volvox DP1 Gonium CDK D1 Volvox 100 CYCB1 Gonium DP1 Volvox 99 100 CDK D1 Chlamydomonas CYCB1 Chlamydomonas 100 100 87 CDK D1 Gonium CYCA1 Chlamydomonas E2FR1 Chlamydomonas CycA1 Volvox CDK C1 Gonium 100 E2FR1 Gonium 100 CDK C1 Chlamydomonas CYCA2 Gonium 100 81 CDK C1 Volvox CYCA1 Gonium E2FR1 Volvox 83 CDK H1 Chlamydomonas CycD4 Volvox 100 CDK H1 Volvox CYCD4 Chlamydomonas 87 CYCD3 Chlamydomonas CDK H1 Gonium Chlamydomonas N RB-A L1 RB-B C CDK B1 Volvox 100 100 CycD3 Volvox 100 100 aa 100 CDK B1 Gonium CYCD3 Gonium Gonium N RB-A L1 RB-B C 83 CDK B1 Chlamydomonas CycD2 Volvox CDK A1 Gonium 100 CYCD2 Chlamydomonas Volvox male N RB-A L1 RB-B C 85 88 100 CDK A1 Volvox CYCD2 Gonium 85 CDK A1 Chlamydomonas CYCD1.2 Gonium Volvox female N RB-AL1 RB-B C CYCD1.1 Gonium CDK G1 Gonium 91 100 CDK G1 Volvox CYCD1.4 Gonium 2 kb CDK G1 Chlamydomonas 95 CYCD1.3 Gonium Chlamydomonas D1 CYCD5 Gonium 5.7 kb Repeats CYCD1 Chlamydomonas Gonium D1.1 D1.2 D1.3 D1.4 CycD1.1 Volvox 14.7 kb 11.7 kb Repeats CycD1.2 Volvox Volvox D1.4 D1.2 D1.3 D1.1 CycD1.3 Volvox 100 CycD1.4 Volvox Figure 3 | Cell cycle pathway evolution. (a) The retinoblastoma cell cycle regulatory pathway. (b) Phylogeny of CDK genes. (c) Phylogeny of cyclin genes and syntenic relationships of cyclin D1 genes. (d) Phylogeny of MAT3/RB genes and comparison of MAT3/RB proteins. Domains RB-A and RB-B, the linker region in the binding pocket (L1), N-terminal conservation (N) and C-terminal conservation (C) are shown. Difference in shade of grey for Volvox male and Volvox female indicates sex-specific divergence24. Conserved putative CDK phosphorylation sites are indicated with solid arrows, species-specific sites are indicated with open arrows. (e) E2F/DP1 genes. All trees have a midpoint root and bootstrap values above 80% are indicated.

Although most of these regulators are nearly identical in RB proteins (Fig. 3d). Interestingly, in animals the C terminus Chlamydomonas, Gonium and Volvox (Fig. 3b,e), there are two of RB is intertwined with E2F/DP and changes in the notable differences. First, Volvox has a four gene expansion of phosphorylation by cyclin–CDK complexes could also alter cyclin D1 genes (Fig. 3c)17. As Volvox has tissue differentiation, E2F/DP binding28. As these phosphorylation sites are absent in these cyclin D1 genes may have been important for tissue Gonium and Volvox RB proteins (Fig. 3d), this suggests that RB development as is the case in metazoans and land plants22,23, co-option for multicellularity may result in differences in locus- supported by the fact that RB has moved into the mating locus of specific temporal expression of genes important for Gonium and Volvox and is differentially expressed between multicellularity during G1, such as cell–cell adhesion genes. mating loci (Fig. 3d)24,25. However, the tandem array expansion Given the role the RB pathway plays in regulating the cell cycle in of the cyclin D1 genes is also found in Gonium (Fig. 3c), where Chlamydomonas, its early modification found in Gonium, and its cyclin D genes display elevated dN/dS ratios compared with other co-option for complex morphology in Volvox, RB pathway cell cycle regulators (Supplementary Fig. 8), suggesting the regulation might be a key step towards multicellularity in the function of these cyclin Ds may be important for the transition volvocine algae. to undifferentiated colonies, rather than tissue differentiation. To test whether RB modifications present in Gonium and Second, there is modification of the RB gene in Gonium and Volvox (compared with Chlamydomonas) are unrelated to, cause Volvox (Fig. 3d, Supplementary Figs 5–7). Protein dimers of or are a consequence of multicellularity, we expressed the cyclins and CDKs primarily regulate RB by phosphorylating Chlamydomonas11 and Gonium RB genes in a Chlamydomonas serine or threonine residues, which is thought to regulate RB strain lacking its RB gene (rb, mat3–4 strain, Fig. 4a)9,11 using the binding of chromatin via E2F/DP transcription factors11,26,27. promoter and terminator from the Chlamydomonas RB gene to Recently it has been shown in human cells that cyclin D and ensure expression near wild-type levels (Fig. 4b,c)11. The CDK4/6 regulate monophosphorylation of RB proteins for G1 Chlamydomonas RB gene rescues the small cell size defect in phase-specific RB functions27. If similar in Gonium, this would the rb mutant (HA-CrRB::rb, Fig. 4a), while the Gonium RB gene suggest a role of the expanded cyclin D1 genes for regulating RB rescues the cell size defect and causes the Chlamydomonas rb to express multicellularity-related genes during G1 phase. If the mutant to become non-palmelloid colonial, ranging from expanded cyclin D1 proteins found in Gonium regulate 2 to 16 normal-sized cells (HA-GpRB::rb, Fig. 4a). Crossing multicellular cell cycle changes, modification of cyclin D-CDK RB gain-of-function transformed Chlamydomonas strain to a phosphorylation sites in RB is predicted. Indeed, the linker of the Chlamydomonas strain lacking DP1, a gene that dimerizes with E2F/DP binding pocket of RB is shorter in Gonium and Volvox E2F to anchor RB to chromatin11, results in suppression of the compared with Chlamydomonas (Fig. 3d, Supplementary Figs 5 colonial phenotype and large-sized cells (HA-GpRB::rb::dp1, and 6), potentially altering how RB binds to chromatin via E2F/ Fig. 4a) consistent with the phenotype of the Chlamydomonas DP. In addition, phosphorylation sites between the E2F/DP dp1 mutant itself10. This demonstrates that the Gonium RB gene pocket region (RB-A and RB-B domains) and the conserved causes colonial multicellularity (Fig. 4a) through the RB pathway carboxy (C)-terminal domain are absent in Gonium and Volvox (Fig. 3a) and suggests differences in how RB binds chromatin and

4 NATURE COMMUNICATIONS | 7:11370 | DOI: 10.1038/ncomms11370 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11370 ARTICLE 62

ac W.T. Chlamydomonas RB knockout DP1 knockout HA-CrRB::rbHA-GpRB::rbkDa from Umen & Goodenough (2001) from Fang et al. (2006) 150 rb dp1 Transform Transform 100 with C.r. RB with G.p. RB 80 HA-CrRB::rb Number of Sexual 60 HA-GpRB::rb cells/colonies rb anti-HA cross 50 observed 1,500 40 kDa 1,000 150 HA-GpRB::rb::dp1 100 500 80 60 0 50 0 10 100 1,000 anti-Tub 3 40 Cell/colony size (µm ) 30 b 25 pCrRB Gonium RB tCrRB HA

AphVIII

Figure 4 | Gonium RB causes a colonial phenotype when expressed in Chlamydomonas. (a) Transformation schematic showing resulting morphology overlaid onto cell and colony size measurements (logarithmic scale) of control Chlamydomonas RB mutant (rb or mat3–4, transformed with empty vector), complementing HA-CrRB::rb (two of five independent transformations are shown) and colonial HA-GpRB::rb (four independent transformations are shown). Crossing colonial HA-GpRB::rb to a Chlamydomonas DP1 mutant (dp1) restores unicellularity in Chlamydomonas (one of two independent matings are shown). (b) Schematic Gonium RB tagged with 3XHA with its expression driven by the Chlamydomonas RB promoter and terminator11.(c) Anti-HA immunoblotting of HA-CrRB::rb and HA-GpRB::rb with anti-tubulin loading controls. Arrows indicate proteins at their expected molecular mass. regulates the expression of cell cycle related genes in Gonium spheroid is largely composed of ECM (Fig. 1). Pherophorin and Volvox are important for co-option of these RB targeted and MMP gene families are expanded in Volvox relative genes for multicellularity (Figs 3 and 4). This gain-of-function to Chlamydomonas (Fig. 5c)17. We found the expansion of demonstrates a causal link between cell cycle regulation and the pherophorins and MMP genes in Volvox (Fig. 5c, Supplementary group level during the evolution of multicellularity, emphasizing Data 2 and 3) is not present in Gonium, though some species- that multicellularity can evolve by co-option and modification of specific expansion of MMP genes has occurred (Supplementary regulatory genes rather than extensive genomic differences or Data 2 and 3). While some expansion of ECM gene families in innovation. Gonium was expected14 to direct the cell wall layer synthesis of a Gonium colony, this layer may instead be directed through differential gene expression. Pherophorin, MMP expansion and Volvox innovations for morphological complexity. In Volvox, cellular differentiation correlate with expanded organismal size somatic differentiation is causally regulated by the regA gene rather than the origin of multicellularity, suggesting a subsequent cluster, a set of putative DNA-binding transcription factors step in the evolution of multicellularity. thought to regulate chloroplast biogenesis29–31. The regA gene cluster is absent in Chlamydomonas and Gonium (Fig. 5a,b), but is present in diverse Volvox ferrisii and Volvox gigas32, suggesting Discussion early evolution and co-option of this cluster shortly after the split We have investigated the evolution of multicellularity in the of Gonium and Volvox lineages (Fig. 1)32. Interestingly, if the volvocine algae by sequencing the genome of the undifferentiated absence of regA in Gonium is indicative of the absence of regA in Gonium. Despite morphological differences, it was known that Astrephomene, with an independent evolution of somatic cells the Chlamydomonas and Volvox genomes are strikingly similar, (Fig. 1)16, Astrephomene may determine somatic cell fate through suggesting that multicellularity required few genetic innova- a different pathway than Volvox suggesting multiple evolutionary tions17,35. However, these two genomes, positioned at the pathways and subsequent evolutionary consequences during the extremes of volvocine morphology, were unable to resolve evolution of multicellularity. Indeed, undifferentiated multi- the tempo and mode36 of the evolutionary transition to cellularity evolved once in the Volvocales16, while additional multicellularity. morphological complexity (for example, cellular differentiation The evolution of multicellularity in the volvocine algae is and large Volvox body size) has repeatedly evolved, suggesting a thought to involve 12 morphological innovations (Supplementary relative ease to gain and lose additional complexity. Fig. 1)15. Five of these steps correlate with the evolution of cell We investigated proteins related to morphological complexity, groups7, a period of rapid evolutionary change (tempo). This view pherophorins and matrix metalloprotease (MMP) proteins, in the emphasizes the importance, and subsequent modification, of volvocine algae17. These proteins are hypothesized to produce innovations correlating with undifferentiated colonies (mode). ECM and break up cell wall components during reproduction in Finding support for this view, we have generalized these 12 steps Gonium and other Volvocales15,33,34. While Chlamydomonas into three major phases (Fig. 6): the evolution of cell cycle contains no ECM and Gonium contains little ECM, a Volvox regulation to form cooperative groups via cell–cell adhesion, the

NATURE COMMUNICATIONS | 7:11370 | DOI: 10.1038/ncomms11370 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11370 63

a 710300 31 b 710100 RLS4 Chlamydomonas 710050 RPT4 IDO 710000 Volvox RLS7 Chlamydomonas

ackB RLS12 Chlamydomonas 709950 rlsD 91964 93 rlsK Volvox 709900 rlsJ Volvox

105099 sc11.g147 Gonium ChlamydomonasACK2 sc5.g127 Gonium 17 91971 23 sc11.g146b Gonium RLS1 91972 sc11.g146 Gonium 709750 RLS3 Chlamydomonas 91973 sc788.g9 RLS10 Gonium 709700 97 709650 rlsL Volvox 709600 ido RLS10 Chlamydomonas rlsC 96 rlsH Volvox 709550 RLS9 Chlamydomonas RLS6 Chlamydomonas 709500 RLS5 Chlamydomonas rlsB RLS11 Chlamydomonas regA rlsG Volvox cluster rlsE Volvox 709400 RLS8 Chlamydomonas 709350 sc275.g720 Gonium 709300 regA RLS2 Chlamydomonas 709250 rlsM Volvox rlsI Volvox IDO rlsA 97 rlsF Volvox RPT4 sc11.g233 Gonium 233 RLS1 Gonium RLS1 Chlamydomonas 232 rlsD Volvox rlsA Volvox 91985 80 regA Volvox regA 105104 rlsC Volvox cluster 78 rlsB Volvox rpt4 PR46b230 msd4 61368 c 14 61248 91993 Species Pherophorin Metalloprotease 91994

Gonium genes genes 227

ACK2 RLS1 Chlamydomonas 31 44 Gonium 35 36 Volvox 78 98 Figure 5 | Genetic changes present in the Volvox genome. (a) Gene synteny near the regA gene cluster and closely related regA-like genes (bold). Chromosome or scaffold number is indicated. Conserved genes are linked by line segments for regA-like (thick, black) and neighbouring genes (thin, grey). (b) Phylogenetic relationships of regA-like genes. The tree is a midpoint root and bootstrap values above 70% are indicated. (c) Comparison of number of pherophorin and metalloprotease genes in Chlamydomonas, Gonium and Volvox.

previously obscured in other taxa such as plants, fungi and

Evolution of animals due to genomic divergence (Fig. 6). cell cycle Evolution of Interestingly, an emerging theme throughout the evolution of regulation cellular differentiation multicellularity is that the genetic basis for the evolutionary Evolution of transition emerges much earlier than anticipated3,6,32. In plants increased body size and animals, RB proteins are important for regulating both cell Figure 6 | Conceptual model for the evolution of multicellularity. proliferation and differentiation by highly complex locus Multicellularity hinges on the evolution of cell cycle regulation in a interactions with chromatin and chromatin remodelling 37,38 multicellular context with subsequent evolution of cellular differentiation factors . Our finding that the RB pathway was co-opted (here, cell size-based) and increased body size. early for multicellularity in undifferentiated colonies suggests that the template for subsequent evolutionary innovations in developmental programs was laid out during the transition to evolution of increased organismal size, and the evolution of undifferentiated multicellularity via RB and cell cycle differentiated germ and soma cells. Having sequenced the modifications, rather than with emergence of germ and somatic genome of Gonium, along with the published genomes of cellular differentiation. Interestingly, RB has been further co- Chlamydomonas and Volvox, we can now identify the genetic opted for a role in sexual differentiation in Volvox, where there pathways associated with each of these steps. The evolution of are male- and female-specific isoforms of RB24. This suggests that undifferentiated colonies correlates with RB cell cycle regulatory the evolution of multicellular cell cycle regulation was a critical pathway evolution (Figs 3 and 4), which is further modified as step for the evolution of multicellularity. By comparing the complexity increases in the Volvocales24. Increased organismal genomes of these three volvocine green algae, we have determined size toward Volvox correlates with an expansion of pherophorins that the mechanism of multicellular evolution is primarily co- and MMPs (Fig. 5c). Finally the evolution of the regA gene cluster option and regulatory modification of existing genetic underlies somatic differentiation (Fig. 5a,b). Future sequencing of pathways39. Gene duplication forms the basis of subsequent additional Volvocales genomes should clarify the evolutionary multicellular innovations. steps required for the evolution of germ and soma. Our three- The genomic age is illuminating the genetic pathways that are phase model for the emergence of multicellularity, supported by important for the evolution of multicellularity in other organisms the genetic pathways important for their evolution, changes our where genes such as cadherins and integrins in animals3,4 and cell understanding of the tempo and mode of multicellular evolution wall biogenesis genes in plants6. These are roughly analogous to

6 NATURE COMMUNICATIONS | 7:11370 | DOI: 10.1038/ncomms11370 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11370 ARTICLE 64 metalloproteases and pherophorins in the volvocine algae counted and produced a matrix of Pfam domain diversity and abundance across highlighting convergences on similar genetic innovations for green algae. multicellularity. The substantial innovation and expansion of transcription factors and signalling networks found in animals Analysis of transcription-associated proteins. Transcription-associated proteins and plants3,6 is not present in the volvocine algae. However, the (TAPs) include transcription factors (enhance or repress transcription) and volvocine algae demonstrate the critical role of transcriptional transcription regulators (proteins which indirectly regulate transcription such as scaffold proteins, histone modification or DNA methylation). We combined three regulation of the cell cycle by RB for the formation of TAP classification rules for plants; PlantTFDB54, PlnTFDB55 and PlanTAPDB56 to undifferentiated colonies. RB proteins regulate the cell cycle of make a set of classification rules for 96 TAP families. Conflicts between the three most eukaryotes11,26, and are tumour suppressors in humans26, sets of rules were manually resolved using the rule that included more genes as suggesting a broader role for RB and cell cycle regulation during transcription-associated proteins. Each transcription family includes at least one, up to three, mandatory domains. the evolution of multicellularity. Families may include up to six forbidden domains (that is, a gene G cannot be in The implications of these findings are greater than simply family F if domain D is present); not all families have defined forbidden domains. identifying when genes evolved during the evolution of multi- All mandatory and forbidden domains were represented by a full-length, global, cellularity. Theoretical work has emphasized the need for greater Hidden Markov Model (HMM). Available HMMs were retrieved from Pfam_ls database57,58. When HMMs were not available from the Pfam_ls database, custom understanding of the origin of an integrated group life cycle 55 12,40–42 HMMs were made using multiple sequence alignments from PlnTFDB and the during the evolution of multicellularity . The field has been HMM was calculated using HMMER version 3.0 (ref. 59) using ‘hmmbuild’ with concerned with the evolution of germ–soma division of labour as default parameters and ‘hmmcalibrate—seed 00. the defining step in the evolution of multicellularity40,43–45; Gathering cutoff thresholds (GA) for the custom HMMs were set as the lowest indeed, a recent review of animal multicellularity45 does not score of a true positive hit using a ‘hmmscan’ search against several complete Chlorophyte genomes. Chlorophyte genomes including Bathycoccus prasinos48, mention the importance of cell cycle regulation and group Chlamydomonas reinhardtii35, Chlorella variabilis49, Coccomyxa subellipsoidea formation. The Gonium genome reflects the early evolution of cell C-169 (ref. 50), Micromonas pusilla CCMP1545 (ref. 51), Micromonas pusilla cycle regulation (Figs 3 and 4) in undifferentiated groups, RCC299 (ref. 51), Ostreococcus tauri52, Ostreococcus lucimarinus53, Ostreococcus conserved and modified in differentiated Volvox, that is indicative sp. RCC809 (available on the DOE Phytozome website, version 10.1) and Volvox carteri17 were searched using ‘hmmscan’ to search the library of 103 domains of the emergence of colony level adaptations. We highlight an against the predicted protein sequences. Analyses were replicated with both Volvox early and formative step, the co-option and expansion of cell cycle version 1 and version 2; however, as results were not qualitatively different, results regulation, as important for the evolution of cooperative groups from version 1 are provided (Supplementary Fig. 3). Subsequent hits were classified and impacting the evolution of more complicated body plans; one into a TAP family. Conflicts between multiple TAP families were resolved by assigning the gene to the TAP family with the highest score (Supplementary that may shed light on the evolutionary history of other Table 1). multicellular innovations and evolutionary transitions.

Construction of protein families. Protein families were created using Methods OrthoMCL60 with a variety of inflation values ranging from 1.2 to 4.0 in steps of Strain and genome sequencing. The Gonium pectorale strain K3-F3-4 0.1 (Supplementary Figs 16–17). This analysis was performed using Chlorophyte (mating type minus, NIES-2863 from the Microbial Culture Collection at National genomes available on the DOE JGI Phytozome website, version 10.1 including Institute for Environmental Studies, Tsukuba, Japan, http://mcc.nies.go.jp/) Bathycoccus prasinos48, Chlamydomonas reinhardtii35, Chlorella variabilis49, was used for genome sequencing. Gonium was grown in 200–300 ml VTAC media Coccomyxa subellipsoidea C-169 (ref. 50), Micromonas pusilla CCMP1545 (ref. 51), at 20 °C with a 14:10 h light–dark cycle using cool-white fluorescent lights Micromonas pusilla RCC299 (ref. 51), Ostreococcus tauri52, Ostreococcus 2 1 53 (165–175 mmol m À s À ). lucimarinus , Ostreococcus sp. RCC809 (available on the DOE Joint Genome For next-generation sequencing and construction of a fosmid library, total DNA Institute website) and Volvox carteri17. This analysis was repeated for both Volvox was extracted. Sequencing libraries were prepared using the GS FLX Titanium version 1 and Volvox version 2. The inflation value of 1.9 was used for both Rapid Library Preparation Kit (F. Hoffmann-La Roche, Basel, Switzerland) and the analyses for consistency and was chosen to have relatively large, coarser grained TruSeq DNA Sample Prep Kit (Illumina Inc., San Diego, CA, USA) and were run clusters that were robust to higher inflation values (Supplementary Figs 16–19). To on both GS FLX (F. Hoffmann-La Roche) and MiSeq (Illumina Inc.) machines. avoid bias introduced by not including all genes for each species, genes not Newbler v2.6 was used to assemble the GS FLX reads. A fosmid library was assigned to a gene family (singletons) were assigned to single gene families and constructed in-house using vector pKS300. The fosmid library (23,424 clones) included in all subsequent phylogenetic gene family analyses. and BAC library (18,048 clones, Genome Institute (CUGI), Clemson University, A species tree was calculated by extracting OrthoMCL gene families containing Clemson, SC, USA) were end-sequenced using a BigDye terminator kit v3 only one copy in each species, for a total of 1,457 genes. The OrthoMCL run with (Life Technologies, Carlsbad, CA, USA) analysed on automated ABI3730 capillary an inflation value of 1.5 was chosen to use larger, coarser grained clusters, thus sequencers (Life Technologies). increasing the likelihood of capturing true 1:1:1 orthologues. This species tree included Volvox carteri version 2. These genes were independently aligned using Muscle version 3.8.31 (ref. 61) and concatenated. A phylogenetic tree was produced Evidence-based gene prediction. Introns hint file generation was done through a using RAxML version 8.0.20 (ref. 62) using the Protein Gamma model with two-step, iterative mapping approach using Bowtie/Tophat command lines automatic model selection on a per gene basis via partitions for each protein. and custom Perl scripts written by Mario Stanke as part of AUGUSTUS46, A rapid bootstrapping analysis to search for the best-scoring ML tree was (available at: http://bioinf.uni-greifswald.de/bioinf/wiki/pmwiki.php?n= run with 100 bootstraps. The resulting species tree is consistent with previous IncorporatingRNAseq.Tophat). AUGUSTUS version 2.6.1 was selected because its results16,51,63–65 and had 100 bootstrap support at every node (Supplementary algorithm has been successfully tuned to predict genes in Chlamydomonas and Fig. 20). This result is also consistent with numerous morphological characteristics Volvox genomes, which contain high GC content46. Reads were first mapped to the supporting a closer relationship of Gonium and Volvox66. genome assembly with Tophat version 2.0.2 (ref. 47) and the raw alignments were Gene family evolution within the volvocine algae was analysed using Count filtered to create an initial (intron) hints file, which was subsequently provided to version 10.04 (ref. 67) to perform several parsimony analyses including symmetric AUGUSTUS during gene prediction. An exon–exon junction database was Wagner parsimony (each gene family may be gained or expanded multiple times generated from the initial AUGUSTUS prediction via a Perl script. The twice- and the gain penalty is equal to the loss penalty) and asymmetric Wagner mapped reads (once to the genome and once to the exon–exon sequences) were parsimony (each gene family may be gained or expanded multiple times and the then merged, filtered and a final intron hints file was created. From this, the final gain penalty is two times higher than the loss penalty). This analysis was repeated gene prediction with AUGUSTUS was performed. for both Volvox version 1 and version 2 genomes (Supplementary Tables 5–8).

Pfam domain analysis. Diversity and abundance of Pfam domains was deter- dN/dS analysis. During our OrthoMCL construction of protein gene families, we mined for all published green algae genomes. Chlorophyte genomes including identified 6,154 clusters with exactly one copy in Chlamydomonas (version 5.3), Bathycoccus prasinos48, Chlamydomonas reinhardtii35, Chlorella variabilis49, Gonium and Volvox (version 2). The number of genes from other unicellular (non- Coccomyxa subellipsoidea C-169 (ref. 50), Micromonas pusilla CCMP1545 (ref. 51), Chlamydomonas) Chlorophyte species was ignored. This criteria is relatively strict Micromonas pusilla RCC299 (ref. 51), Ostreococcus tauri52, Ostreococcus as it does not include any genes with a duplicate in any species (copy number lucimarinus53, Ostreococcus sp. RCC809 (US Department of Energy, Phytozome) greater than one in any species) or any genes which are not essential (no copy and Volvox carteri (both versions 1 and 2; ref. 17) were searched using direct present in any species) resulting in 1:1:1 orthologues. Given the relatively high gene submission of Pfam A and Pfam B domains using Bioperl. Subsequent hits were duplication rates in volvocine algae (data not shown), these strict criteria support

NATURE COMMUNICATIONS | 7:11370 | DOI: 10.1038/ncomms11370 | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11370 65 an interpretation of 1:1:1 orthology. Genome-wide pairwise comparisons of dN, dS mutations are linked to AphVIII, single tetrads were dissected. HA-GpRB and dN/dS were calculated (Supplementary Fig. 21; Supplementary Table 11) using was genotyped with primers in the 3XHA tag 50-AGTGCTAACAGCATGTCT 68 PAML and codeml (ML analysis ) based on nucleotide translation based AGTTAC-30, and in the 50 portion of GpRB 50-TGCGAACAACCGCTGCAGA 61 10 alignments (proteins were aligned using MUSCLE ). CCTTC-30.Thedp1 mutation was genotyped as previously described .

20 Prediction of lineage-specific genes. The phylostratigraphy method assumes Immunoblotting HA-GpRB and HA-CrRB strains complementing rb. Whole-cell Dollo’s parsimony (that is, it is more likely that a gene observed in two distant lysates from strains were prepared, separated and immunoblotted11. Briefly, the clades was present in the common ancestor and multiple independent gains are not anti-HA antibody used for detection of HA-GpRB and HA-CrRB was an anti-HA possible). This provides an entry point for testing evolutionary hypotheses related high affinity monoclonal antibody (clone 3F10, Roche) and anti-alpha-tubulin to the age of genes and to quantify how much gene-level innovation has occurred monoclonal antibody (Sigma), as previously described11. The expression levels of along each phylogenetic branch. Old genes are classified in low phylostrata (present RB in HA-CrRB strains have been previously shown to be similar to wild-type in distant species, PS1–PS7) and young genes are classified in higher phylostrata Chlamydomonas expression levels11. The expression levels of RB in HA-GpRB are (for example, genus- or species-specific genes, PS8–PS9). The resolution of each similar, if not slightly below, the expression levels of HA-CrRB, suggesting that phylostratum strictly depends on the availability of reliable outgroups (the overexpression of RB is not causing the observed colonial phenotype, but rather availability of reliable genomic outgroups is relatively low in Chlorophyte algae). modification to the Gonium RB gene. The phylogenetic classes were defined from those in each NCBI Taxonomy entry for Chlamydomonas, Gonium and Volvox, resulting in nine expected phylostrata for each species. All proteins were subjected to a BLASTP search with an E-value Measurement of cell or colony size distribution. The size of cells and groups of threshold of 0.001 against the NCBI nr database. Placement in phylostrata was cells was measured with a Moxi Z automated cell sizer/counter using type ‘S’ derived from the taxonomic information of these hits for each protein, using the cassettes (ORFLO Technologies). Sizing is based on the Coulter principle used most distant hit, and following Dollo’s parsimony. previously with Chlamydomonas reinhardtii10,11.

Phylogenetic analyses. Unless otherwise stated, all phylogenetic analyses were performed using a custom pipeline of SATe version 2.2.7 (ref. 69) coupled with References RAxML version 8 (ref. 62). Full gene protein sequences were passed to SATe 1. Bonner, J. T. The origins of multicellularity. Integr. Biol. 1, 27–36 (1998). using a FASTTREE tree estimation with a RAxML search after tree formation 2. Grosberg, R. K. & Strathmann, R. R. The evolution of multicellularity: a minor with a maximum limit of 10 iterations and the ‘longest’ decomposition strategy. major transition? Annu. Rev. Ecol. Evol. Syst. 38, 621–654 (2007). Bootstraps were made on the SATe output alignment and tree using RAxML with 3. Suga, H. et al. The Capsaspora genome reveals a complex unicellular prehistory automatic model selection, a rapid hill climbing algorithm ( f d) and 100 of animals. Nat. Commun. 4, 2325–2333 (2013). À bootstrap partitions. Bipartition information ( f a) was obtained using the SATe 4. King, N. et al. The genome of the choanoflagellate Monosiga brevicollis and the À output tree and RAxML bootstraps. origin of metazoans. Nature 451, 783–788 (2008). 5. Parfrey, L., Lahr, D. J. G., Knoll, A. H. & Katz, L. A. Estimating the timing of early eukaryotic diversification with multigene molecular clocks. Proc. Natl Chlamydomonas strains culture conditions. Wild-type Chlamydomonas Acad. Sci. USA 108, 13624–13629 (2011). reinhardtii 6145 and 21gr, and HA-CrRB (HA-MAT3::mat3–4, here referred to as HA-CrRB::rb), mat3–4 (here referred to as rb), and dp1 have been previously 6. Hori, K. et al. Klebsormidium flaccidum genome reveals primary factors for described9–11. Briefly, wild-type strains 6145 (MT ) and 21gr (MT ) are mating plant terrestrial adaptation. Nat. Commun. 5, 1–9 (2014). pairs that have been back crossed to eliminate theÀy1 mutation in 6145þ (ref. 10). 7. Herron, M. D., Hackett, J. D., Aylward, F. O. & Michod, R. E. Triassic origin The RB knockout strain has been previously characterized as a null allele, and the and early radiation of multicellular volvocine algae. Proc. Natl Acad. Sci. USA knockout mutation is the rb allele9,11. The rb mutation can be complemented by a 106, 3254–3258 (2009). amino (N)-terminally tagged version of the gene that behaves identical to wild type. 8. Kirk, D. L. Volvox: Molecular-Genetic Origins of Multicellularity and Cellular Previously, a knockout mutation in the Chlamydomonas DP1 gene, dp1,was Differentiation (Cambridge Univ. Press, 1998). identified and characterized10,11. All the strains were maintained on TAP plates. 9. Umen, J. G. & Goodenough, U. W. Control of cell division by a retinoblastoma For phenotype analysis, the strains were grown in high salt media (HSM) protein homolog in Chlamydomonas. Genes Dev. 15, 1652–1661 (2001). synchronously under 14 h of 150 mE of light, samples were fixed hourly and 10. Fang, S.-C., de los Reyes, C. & Umen, J. G. Cell size checkpoint control by the examined by light microscopy10,11. retinoblastoma tumor suppressor pathway. PLoS Genet. 2, e167 (2006). 11. Olson, B. J. S. C. et al. Regulation of the Chlamydomonas cell cycle by a stable, chromatin-associated retinoblastoma tumor suppressor complex. Plant Cell 22, Cloning of Gonium pectorale RB and transformation into rb. A 3X haemagluttin 3331–3347 (2010). (HA) tagged copy of the Gonium pectorale RB gene was cloned using InFusion 12. Shelton, D. E. & Michod, R. E. Group selection and group adaptation during a Cloning (Clontech) to be driven by the Chlamydomonas RB promoter and major evolutionary transition: insights from the evolution of multicellularity in terminator that includes a AphVIII selectable marker for Chlamydomonas trans- the volvocine algae. Biol. Theory 9, 452–469 (2014). formation (Fig. 4, (ref. 11)). Gonium pectorale genomic DNA from K4F3 was used 13. Harper, J. D. I. & John, P. C. L. Coordination of division events in the as a template and the genomic region of RB was amplified without its ATG start Chlamydomonas cell cycle. Protoplasma 131, 118–130 (1986). codon using the primers 5 -CAGATTACGCTACTAGATCTGCCGAAGCTG 0 14. Umen, J. G. & Olson, B. J. S. C. in Genomic Insights Into the Biology of Algae AACGTTTTACTGCG-3 , and 5 -CTCCGGCCGCGGTGCCTAATTTGCG 0 0 Vol. 64 (ed. Piganeau, G.) 185–243 (Elsevier, 2012). CCGTACCGCCGGA-30. These primers overlap with the 3X HA tag and 30 terminator from the previously created HA-CrRB transformation clone that 15. Kirk, D. L. A twelve-step program for evolving multicellularity and a division of complements the rb mutation11. The HA-CrRB plasmid was amplified by inverse labor. Bioessays 27, 299–310 (2005). 16. Herron, M. D. & Michod, R. E. Evolution of complexity in the volvocine algae: PCR with 50-TCTAGTAGCGTAATCTGGAACGTCATATGGATAGG-30 and transitions in individuality through Darwin’s eye. Evolution 62, 436–451 50-GCACCGCGGCCGGAGGT-30 primers. PCR products were gel purified with a QiaQuick gel extraction kit (Qiagen). Purified PCR fragments were fused by (2008). InFusion (Clontech) cloning based on overlaps in the amplified sequences and 17. Prochnik, S. E. et al. Genomic analysis of organismal complexity in the transformed into chemically competent DH5-apha cells, after which the clone was multicellular green alga Volvox carteri. Science 329, 223–226 (2010). confirmed by sequencing. 18. Lynch, M. & Conery, J. S. The origins of genome complexity. Science 302, 1401–1404 (2003). 19. Smith, D. R. et al. Organelle genome complexity scales positively with organism Transformation of Chlamydomonas reinhardtii. The rb strain was transformed size in volvocine green algae. Mol. Biol. Evol. 30, 793–797 (2013). 11 with glass beads , with the HA-GpRB clone (above) and as a control with 20. Domazet-Loso, T., Brajkovic, J. & Tautz, D. A phylostratigraphy approach to HA-CrRB and pSI103 (AphVIII selectable marker only) and selected on TAP plates 1 11 uncover the genomic history of major adaptations in metazoan lineages. Trends supplemented with 20 mgmlÀ paromycin . Candidate strains were screened by 10,11 Genet. 23, 533–539 (2007). growth morphology , and then screened for expression by immunoblotting with 21. Lang, D. et al. Genome-wide phylogenetic comparative analysis of plant an anti-HA antibody (Roche 3F10, high affinity11). Four independent strains transcriptional regulation: a timeline of loss, gain, expansion, and correlation expressing the HA-GpRB, and five independent strains expressing HA-CrRB were with complexity. Genome Biol. Evol. 2, 488–503 (2010). created. Control complementation of the rb mutation with HA-CrRB occurred at 22. Beumer, T. L., Roepers-Gajadien, H. L., Gademan, I. S., Kal, H. B. & rates similar to previous results11. The presence of the rb mutation was confirmed de Rooij, D. G. Involvement of the D-type cyclins in germ cell proliferation by replica plating on TAP plates supplemented with 10 mgml 1 emetine9,11. À and differentiation in the mouse. Biol. Reprod. 63, 1893–1898 (2000). 23. Dewitte, W. et al. Altered cell cycle distribution, hyperplasia, and inhibited Genetic analysis of HA-GpRB-expressing strains. Two lines expressing HA- differentiation in Arabidopsis caused by the D-type cyclin CYCD3. Plant Cell GpRB were crossed to a dp1 null mutation10. Because both the HA-GpRB and dp1 15, 79–92 (2003).

8 NATURE COMMUNICATIONS | 7:11370 | DOI: 10.1038/ncomms11370 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11370 ARTICLE 66

24. Ferris, P. J. et al. Evolution of an expanded sex-determining locus in Volvox. 54. Guo, A.-Y. et al. PlantTFDB: a comprehensive plant transcription factor Science 328, 351–354 (2010). database. Nucleic Acids Res. 36, D966–D969 (2008). 25. Hiraide, R. et al. The evolution of male-female sexual dimorphism predates the 55. Rian˜o-Pacho´n, D. M., Ruzicic, S., Dreyer, I. & Mueller-Roeber, B. PlnTFDB: an gender-based divergence of the mating locus gene MAT3/RB. Mol. Biol. Evol. integrative plant transcription factor database. BMC Bioinformatics 8, 42 30, 1038–1040 (2013). (2007). 26. Khidr, L. & Chen, P.-L. RB, the conductor that orchestrates life, death and 56. Richardt, S., Lang, D., Reski, R., Frank, W. & Rensing, S. A. PlanTAPDB, a differentiation. Oncogene 25, 5210–5219 (2006). phylogeny-based resource of plant transcription-associated proteins. Plant 27. Narasimha, A. M. et al. Cyclin D activates the Rb tumor suppressor by mono- Physiol. 143, 1452–1466 (2007). phosphorylation. Elife 3, 1–21 (2014). 57. Finn, R. D. et al. The Pfam protein families database. Nucleic Acids Res. 36, 28. Rubin, S. M., Gall, A. L., Zheng, N. & Pavletich, N. P. Structure of the Rb D281–D288 (2008). C-terminal domain bound to E2F1-DP1: a mechanism for phosphorylation- 58. Finn, R. D. et al. The Pfam protein families database. Nucleic Acids Res. 38, induced E2F release. Cell 123, 1093–1106 (2005). D211–D222 (2010). 29. Meissner, M., Stark, K., Cresnar, B., Kirk, D. L. & Schmitt, R. Volvox germline- 59. Eddy, S. R. Profile hidden Markov models. Bioinformatics 14, 755–763 (1998). specific genes that are putative targets of RegA repression encode chloroplast 60. Li, L., Stoeckert, C. J. & Roos, D. S. OrthoMCL: identification of ortholog proteins. Curr. Genet. 36, 363–370 (1999). groups for eukaryotic genomes. Genome Res. 13, 2178–2189 (2003). 30. Kirk, M. M. et al. regA,aVolvox gene that plays a central role in germ-soma 61. Edgar, R. C. MUSCLE: multiple sequence alignment with high accuracy and differentiation, encodes a novel regulatory protein. Development 126, 639–647 high throughput. Nucleic Acids Res. 32, 1792–1797 (2004). (1999). 62. Stamatakis, A. RAxML version 8: a tool for phylogenetic analysis and post- 31. Duncan, L. et al. The VARL gene family and the evolutionary origins of the analysis of large phylogenies. Bioinformatics 30, 1312–1313 (2014). master cell-type regulatory gene, regA, in Volvox carteri. J. Mol. Evol. 65, 1–11 63. Worden, A. Z. & Not, F. in Microbial Ecology Of the Oceans (ed. Kirchman, D.) (2007). 159–205 (Wiley, 2008). 32. Hanschen, E. R., Ferris, P. J. & Michod, R. E. Early evolution of the genetic basis 64. Nozaki, H. Origin and evolution of the genera Pleodorina and Volvox for soma in the Volvocaceae. Evolution 68, 2014–2025 (2014). (Volvocales). Biologia 58, 425–431 (2003). 33. Hallmann, A. Extracellular matrix and sex-inducing pheromone in Volvox. Int. 65. Leliaert, F. et al. Phylogeny and molecular evolution of the green algae. CRC Rev. Cytol. 227, 131–182 (2003). Crit. Rev. Plant Sci. 31, 1–46 (2012). 34. Hallmann, A. The pherophorins: common, versatile building blocks in the 66. Nozaki, H. & Itoh, M. Phylogenetic relationships within the colonial Volvocales evolution of extracellular matrix architecture in Volvocales. Plant J. 45, (Chlorophyta) inferred from cladistic analysis based on morphological data. 292–307 (2006). J. Phycol. 30, 353–365 (1994). 35. Merchant, S. S. et al. The Chlamydomonas genome reveals the evolution of key 67. Csuro¨s, M. Count: evolutionary analysis of phylogenetic profiles with animal and plant functions. Science 318, 245–250 (2007). parsimony and likelihood. Bioinformatics 26, 1910–1912 (2010). 36. Simpson, G. G. Tempo and Mode in Evolution (Columbia Univ. Press, 1944). 68. Yang, Z. PAML 4: phylogenetic analysis by maximum likelihood. Mol. Biol. 37. Sage, J. The retinoblastoma tumor suppressor and stem cell biology. Genes Dev. Evol. 24, 1586–1591 (2007). 26, 1409–1420 (2012). 69. Liu, K., Raghavan, S., Nelesen, S., Linder, C. R. & Warnow, T. Rapid and 38. Kuwabara, A. & Gruissem, W. Arabidopsis Retinoblastoma-related and accurate large-scale coestimation of sequence alignments and phylogenetic Polycomb group proteins: Cooperation during plant cell differentiation and trees. Science 324, 1561–1564 (2009). development. J. Exp. Bot. 65, 2667–2676 (2014). 39. Kapheim, K. M. et al. Genomic signature of evolutionary transitions from solitary to group living. Science 348, 1139–1144 (2015). Acknowledgements 40. Hanschen, E. R., Shelton, D. E. & Michod, R. E. in Evolutionary transitions to We acknowledge the support of the National Aeronautics and Space Administration multicellular life: principles and mechanisms. (eds Nedelcu, A. M. & Ruiz-Trillo, (grant number NNX13AH41G to R.E.M., P.J.F. and P.M.D.), the National Institute of I.) 165–188 (Springer, 2015). Health (grant number GM084905 to the University of Arizona, P20GM103638 to 41. Libby, E. & Rainey, P. B. A conceptual framework for the evolutionary origins B.J.S.C.O., P20GM103418 to B.J.S.C.O. and T.N.M., GM103785-02 to M.W.K.), the of multicellularity. Phys. Biol. 10, 035001 (2013). National Science Foundation (grant number MCB-1412395 to B.J.S.C.O. and R.E.M., 42. Rainey, P. B. & Kerr, B. Cheats as first propagules: a new hypothesis for the grant number DGE-0654435 to the University of Arizona and PHY11-25915 to U.C.S.B., evolution of individuality during the transition from single cells to K.I.T.P.), the KSU Johnson Cancer Center (Innovative Research Award to B.J.S.C.O.), multicellularity. Bioessays 32, 872–880 (2010). MEXT/JSPS KAKENHI Scientific Research on Innovative Areas ‘Genome Science’ (grant 43. Michod, R. E. The group covariance effect and fitness trade-offs during number 221S0002 to A.T.), MEXT/JSPS KAKENHI Scientific Research (A) (grant evolutionary transitions in individuality. Proc. Natl Acad. Sci. USA 103, number 24247042 to H.N.), JSPS Fellows (19–7661 and 23–5499 to T.H.) and the JSPS 9113–9117 (2006). Postdoctoral Fellowship for Research Abroad (26–495 to T.H.). Part of this work was 44. Michod, R. E. Evolution of individuality during the transition from undertaken during the U.C.S.B. K.I.T.P. ‘Cooperation and the Evolution of Multi- unicellular to multicellular life. Proc. Natl Acad. Sci. USA 104, 8613–8618 cellularity’ program. Part of the computing for this project was performed on the Beocat (2007). Research Cluster at KSU, which is funded in part by NSF grants CNS-1006860, EPS- 45. Rokas, A. The origins of multicellularity and the early history of the 1006860 and EPS-0919443. We thank M. Barker and M. Sanderson for technical assis- genetic toolkit for animal development. Annu. Rev. Genet. 42, 235–251 tance; A. M. Nedelcu for comments; three anonymous reviewers for comments and all 2008 : the technical staff of the Comparative Genomics lab at the National Institute of Genetics 46.ð Stanke,Þ M. & Morgenstern, B. AUGUSTUS: a web server for gene prediction in for their assistance. Publication of this article was funded in part by the Kansas State eukaryotes that allows user-defined constraints. Nucleic Acids Res. 33, University Open Access Publishing Fund. W465–W467 (2005). 47. Trapnell, C., Pachter, L. & Salzberg, S. L. TopHat: discovering splice junctions with RNA-Seq. Bioinformatics 25, 1105–1111 (2009). Author contributions 48. Moreau, H. et al. Gene functionalities and genome structure in Bathycoccus E.R.H. conceived the project, performed the analyses and wrote the manuscript; T.N.M. prasinos reflect cellular specializations at the base of the green lineage. Genome did gene model prediction and annotation and prepared the RNA samples; T.H. prepared Biol. 13, R74 (2012). gDNA; A.T., A.F., H.N., Y.M., M.S. and H.K.-T. did assemblies; P.J.F., D.R.S., H.S. and 49. Blanc, G. et al. The Chlorella variabilis NC64A genome reveals adaptation to J.A. did gene model annotations; R.N., R.B., V.L., A.K. and M.W.K. performed lineage- photosymbiosis, coevolution with viruses, and cryptic sex. Plant Cell 22, specific gene analyses; P.M.D., R.E.M. and H.N. conceived the project; B.J.S.C.O. 2943–2955 (2010). conceived the project, performed the analyses, performed transformations and wrote the 50. Blanc, G. et al. The genome of the polar eukaryotic microalga manuscript; all the authors contributed to preparation of the manuscript and the text. Coccomyxa subellipsoidea reveals traits of cold adaptation. Genome Biol. 13, R39 (2012). 51. Worden, A. Z. et al. Green evolution and dynamic adaptations revealed by genomes of the marine Picoeukaryotes Micromonas. Science 324, 268–272 Additional information (2009). Accession codes: This Whole Genome Shotgun project has been deposited at DDBJ/ 52. Derelle, E. et al. Genome analysis of the smallest free-living eukaryote ENA/GenBank under the accession LSYV00000000. The version described in this paper Ostreococcus tauri unveils many unique features. Proc. Natl Acad. Sci. USA 103, is version LSYV01000000. Data are also available via JGI Phytozome. 11647–11652 (2006). Supplementary Information accompanies this paper at http://www.nature.com/ 53. Palenik, B. et al. The tiny eukaryote Ostreococcus provides genomic insights naturecommunications into the paradox of plankton speciation. Proc. Natl Acad. Sci. USA 104, 7705–7710 (2007). Competing financial interests: The authors declare no competing financial interests.

NATURE COMMUNICATIONS | 7:11370 | DOI: 10.1038/ncomms11370 | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11370 67

Reprints and permission information is available online at http://npg.nature.com/ This work is licensed under a Creative Commons Attribution 4.0 reprintsandpermissions/ International License. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise How to cite this article: Hanschen, E. R. et al. The Gonium pectorale genome demon- in the credit line; if the material is not included under the Creative Commons license, strates co-option of cell cycle regulation during the evolution of multicellularity. users will need to obtain permission from the license holder to reproduce the material. Nat. Commun. 7:11370 doi: 10.1038/ncomms11370 (2016). To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

10 NATURE COMMUNICATIONS | 7:11370 | DOI: 10.1038/ncomms11370 | www.nature.com/naturecommunications 68

APPENDIX C. EARLY EVOLUTION OF THE GENETIC BASIS FOR SOMA IN THE

VOLVOCACEAE

Hanschen E. R., P. J. Ferris, R. E. Michod. 2014. Early evolution of the genetic basis for soma in

the Volvocaceae. Evolution 68(7): 2014-2025.

ORIGINAL ARTICLE 69

doi:10.1111/evo.12416

EARLY EVOLUTION OF THE GENETIC BASIS FOR SOMA IN THE VOLVOCACEAE

Erik R. Hanschen,1,2 Patrick J. Ferris,1 and Richard E. Michod1 1Department of Ecology and Evolutionary Biology, University of Arizona, 1041 E. Lowell St., Biosciences West, Rm. 310, Tucson, Arizona 2E-mail: [email protected]

Received September 30, 2013 Accepted March 16, 2014

To understand the hierarchy of life in evolutionary terms, we must explain why groups of one kind of individual, say cells, evolve into a new higher level individual, a multicellular organism. A fundamental step in this process is the division of labor into nonreproductive altruistic soma. The regA gene is critical for somatic differentiation in Volvox carteri,amulticellularspeciesof volvocine algae. We report the sequence of regA-like genes and several syntenic markers from divergent species of Volvox.We show that regA evolved early in the volvocines and predict that lineages with and without soma descended from a regA-containing ancestor. We hypothesize an alternate evolutionary history of regA than the prevailing “proto-regA”hypothesis.Thevariation in presence of soma may be explained by multiple lineages independently evolving soma utilizing regA or alternate genetic pathways. Our prediction that the genetic basis for soma exists in species without somatic cells raises a number of questions, most fundamentally, under what conditions would species with the genetic potential for soma, and hence greater individuality, not evolve these traits. We conclude that the evolution of individuality in the volvocine algae is more complicated and labile than previously appreciated on theoretical grounds.

KEY WORDS: Gene duplication, germ-soma differentiation, major transitions, multicellularity, regA, Volvox.

The familiar hierarchy of life consists of different kinds of evo- group (Michod 2007). In numerous independent evolutions of lutionary individuals, for example, prokaryotic cells, eukaryotic multicellularity across the tree of life, fitness specialization and cells, multicellular organisms, and eusocial societies. During evo- integration often takes the form of cellular differentiation into lutionary transitions in individuality, groups of individuals be- germ and soma cells, where germ cells specialize on reproduction come a new kind of individual and the level of selection is trans- and somatic cells specialize on survival (Bonner 1998; Michod ferred from the individual to the group—afterwards a new kind 2007; Simpson 2012). Because these cells specialize on only one of individual. A fundamental characteristic of these transitions fitness component, they can no longer both survive and reproduce is the functional integration of the former evolutionary individ- on their own, were they to leave the group. As a result, the group uals into a new indivisible group. During this process, fitness is becomes indivisible and hence an individual. reorganized so that the former lower level individuals specialize Understanding the evolution of reproductive differentiation in the fitness components of the group and the unit of fitness and specialization during the evolution of multicellularity requires changes from the lower level to the new higher level individual an understanding of the genetic basis underlying these traits as (Michod 1999). well as their ecological and phylogenetic context. For example, Fitness involves two basic components, reproduction and in the volvocine algae studied here, a gene critical for somatic viability. During the evolution of multicellularity, functional in- differentiation in Volvox carteri, regA,ishomologoustoagene tegration occurs when cells relinquish their total fitness by spe- in a unicellular relative, Chlamydomonas reinhardtii.InC. rein- cializing in one of the two basic fitness components of the cell hardtii, this homologous gene is expressed in stressful environ- ments in which growth and reproduction should be down regulated Data Archives: GenBank

C C ⃝ 2014 The Author(s). Evolution ⃝ 2014 The Society for the Study of Evolution. 2014 Evolution 68-7: 2014–2025 EVOLUTION OF THE GENETIC BASIS FOR70 SOMA

(Nedelcu and Michod 2006; Nedelcu 2009). Thus, it appears that VARL gene, rlsD.AlthoughrlsD in V. carteri and RLS1 in C. a life-history gene for down regulating reproduction in stress- reinhardtii are orthologs, there are no orthologs of the V. carteri ful environments in a unicellular ancestor could have been co- tandem duplication genes in C. reinhardtii.Aphylogeneticanal- opted to down regulate reproduction in somatic cells, by changing ysis found that the regA cluster genes are more closely related to its expression from a temporal to spatial context (Michod 1999; each other than to rlsD (Duncan et al. 2007). This led Duncan Michod and Roze 1999; Schlichting 2003; Nedelcu and Michod et al. (2007) to hypothesize that regA evolved from an ancient 2006; Nedelcu 2009). “proto-regA” gene which was subsequently lost in C. reinhardtii The volvocine algae are a clade of green algae including uni- (Fig. S1). cellular (e.g., C. reinhardtii), colonial (e.g., Gonium pectorale), Here, we investigate the presence of VARLgenes in a diverse and multicellular species (e.g., V. carteri). The volvocine algae do group of Volvox species: Volvoxafricanus, Volvox obversus, V. car- not have a multicellular ancestor and unlike other model systems teri f. weismannia, Volvox gigas,andVolvox ferrisii.AsVolvox is for the evolution of multicellularity and cellular differentiation, polyphyletic, these species are both closely related and divergent have a relatively recent origin of multicellularity which allows to each other and to V. carteri. Volvox gigas is thought to share an identification of the genetic changes associated with the steps to origin of somatic cells with V.carteri but V. ferrisii diverged earlier multicellularity (Kirk 2005; Herron et al. 2009; Prochnik et al. (Fig. 1) and is thought to have an independent evolution of so- 2010; Leliaert et al. 2012). In V. carteri, germ and soma differ- matic cells (Nozaki et al. 2002; Herron et al. 2009). The common entiation mutants produce a variety of developmental phenotypes ancestor of V. carteri, V. gigas,andV. ferrisii has also given rise such as the reg or “regenerator” mutant phenotype (Starr 1970; to species of the genus Yamagishiella, Eudorina,andPleodorina Husky and Griffin 1979; Harper et al. 1987; Kirk et al. 1987, (Nozaki et al. 2002, 2006). Yamagishiella species have 32 cells 1999; Adams et al. 1990). In this mutant phenotype, somatic without somatic differentiation (Rayburn and Starr 1974) whereas cells initially develop apparently normally, but then redifferenti- Eudorina comprises species with 32 cells without somatic cells ate into asexual reproductive germ cells which divide to produce (Eudorina elegans)andwithfacultativesomaticcells(Eudorina offspring. All regenerator mutants map to the regA locus and trans- illinoisensis, Goldstein 1964). Pleodorina species vary between formational rescue of this mutant with wild-type regA recovers the 32 and 128 cells, and have between 12.5 and 50% somatic cells wild-type phenotype (Kirk et al. 1999; Nishii and Miller 2010). (Iyengar and Desikachary 1981). The clade composed of V. car- The evolutionary history of regA is of interest to the evolution of teri and V. ferrisii includes species without somatic cells, with reproductive division of labor and the evolution of individuality facultative somatic cells, and with obligate somatic cells. in this lineage. Relatively, little is known about when and why The genus Volvox is diverse, representing four distinct devel- regA evolved except that the V. carteri lineage is thought to have opmental programs designated D1–D4 (Fig. 1, also see Desnitski evolved sterile somatic cells between 76.6 and 116 million years 1995). In D1, (e.g., V. gigas)germcellsareinitiallysmall,grow ago (Herron et al. 2009). substantially to their mature size, and then undergo rapid em- The regA gene is a member of the VARL gene family bryonic divisions without cell growth during embryogenesis. D2 (Volvocine Algae RegA Like) which encodes a single DNA- (e.g., V. carteri, V. africanus, V. obversus) is characterized by ini- binding SAND domain, approximately 80 amino acids long. The tially large germ cells followed by rapid divisions and unequal VARL domain is characterized by a short N-terminal extension divisions in the germ line. D3 (e.g., Volvox tertius) is character- separated from the longer core VARLdomain by a variable length ized by slow embryonic divisions and initially large germ cells. D4 linker (Duncan et al. 2007). regA is expressed in smaller somatic (e.g., V. ferrisii)ischaracterizedbyinitiallysmallgermcellsthat cells after embryonic development and is thought to be a master undergo slow embryonic divisions with cells growing between regulator transcription factor of nuclear-encoded chloroplast bio- divisions (Desnitski 1995). To date, the presence and function of genesis genes (Kirk 1998; Kirk et al. 1999; Meissner et al. 1999; VARL genes has only been demonstrated in C. reinhardtii and Duncan et al. 2006; Nedelcu et al. 2006). regA is hypothesized to V. carteri, so the possible existence and role of regA in other prevent cell division by preventing cells from growing by down developmental programs was unknown. regulating chloroplast biosynthesis, thereby creating somatic cells We report here that both V. gigas and V. ferrisii contain the (Meissner et al. 1999). tandem duplication of VARL genes including regA and that V. fer- In V. carteri,theregA gene is part of the regA cluster, a series risii has an additional gene in this tandem array. On the basis of of tandem duplications involving regA, rlsA, rlsB,andrlsC.The the results reported here, these tandem duplications are thought to RNA expression of regA, rlsB,andrlsC peaks during cellular dif- be ancestral to species with and without soma. The phylogenetic ferentiation, soon after embryonic development, suggesting the distribution of somatic cells in the volvocine algae (Fig. 1) may entirety of the regA cluster plays a role in somatic differentiation be explained by independent tandem duplications, multiple lin- (Harryman 2012). These four genes are closely related to another eages independently evolving somatic cells via regA,ormultiple

EVOLUTION JULY 2014 2015 E. R. HANSCHEN ET AL. 71

Volvox ferrisii D4 regA cluster genotyped by sequencing the nuclear rDNA ITS after polymerase G Volvox rousseletii D4 chain reaction (PCR) amplification (Coleman et al. 1994). Volvox globator D4 R Yamagishiella unicocca D1 DNA PREPARATION Eudorina elegans UTEX 1205 D1 Genomic DNA was prepared using the protocol described for V. L Eudorina elegans UTEX 1212 D1 carteri (Miller and Kirk 1999). Plasmid and cosmid DNA was prepared using the Qiaprep Spin Miniprep Kit or the Qiagen Volvox gigas D1 regA cluster Plasmid Midi Kit according to the manufacturer’s instructions. Eudorina illinoisensis D1 G Pleodorina starrii D1 L PCR CLONING OF GENES Eudorina unicocca D1 All PCR reactions (with one exception noted below) were per- Pleodorina californica D1 formed using 2X Phusion HF Master Mix (Thermo Scientific, Volvox obversus D2 regA Waltham, MA) with 3% DMSO and 25 ng of template when Volvox carteri f. weismannia D2 regA using genomic DNA. Cloning of regA cluster VARL genes was Volvox carteri f. kawakiensis D2 regA initiated by performing genomic PCR with degenerate primers. Volvox carteri f. nagariensis D2 regA cluster Three forward primers (regF1, regF2, regF3) and one reverse Volvox africanus D2 regA primer (regR) were designed using the protein sequences of Volvox tertius D3 the regA cluster genes from V. carteri f. nagariensis (Table S1). Volvox dissipatrix D4 The regF1 primer was only used successfully with V. obversus.

Figure 1. Species phylogeny of the Volvocine algae adapted The regF2 and regF3 primers are based on the same amino acid from Yamada et al. (2008) and Isaka et al. (2012). Tree is rooted sequence, but regF3 is more degenerate. Cycling conditions for (not shown) using G. pectorale (volvocine alga) as the outgroup. these experiments were 98°C for 3 min, followed by 35 cycles Species with sequences reported here are in bold. Evolution of of 98°C/10 sec, 68°C/20 sec, 72°C/60 sec, then a finishing step soma (G) and loss of soma (L) as inferred by Herron et al. (2009) of 5 min at 72°C. The PCR product produced typically included are denoted. Likely evolution of regA (R) is denoted. Herron et al. 57 coding nucleotides (not contributed by the primers) and a con- (2009) inferred an approximate date of 100 million years ago for served intron. PCR amplifications intended for sequencing were the last common ancestor of Eudorina elegans UTEX 1205 and the Pleodorina/Volvox group and an approximate date of 200 million performed in at least two (usually more) separate reactions and years ago for the last common ancestor of V. ferrisii and V. carteri then combined before sequencing to reduce the possibility of PCR (R). Filled circles represent obligate somatic cell differentiation, errors. open circles represent facultative somatic cell differentiation, and Inverse PCR was performed following Sambrook and Russell dots represent absence of somatic cell differentiation. In the sec- (2001). Approximately 500 ng of genomic DNA were digested ond column, division programs as defined in Desnitski (1995) are with SacI(forV. africanus)orPstI(forV. obversus and V. carteri shown. In the third column, species with regA and the complete f. weismannia) and the restriction enzyme was then heat inac- regA cluster are shown. For species with only evidence of regA, tivated. Following ethanol precipitation and resuspension of the the other regA cluster genes may also be present (i.e., we have no evidence of presence or absence of other regA cluster genes for DNA at a concentration of 2 ng/µl, the restriction fragments were these species). circularized by ligation with T4 DNA ligase (New England Bi- olabs, Ipswich, MA). The DNA ligase was heat inactivated, the DNA concentrated by ethanol precipitation, then used as template lineages independently evolving somatic cells via different ge- for PCR with inverse PCR primers. The products of inverse PCR netic pathways. were gel-purified (QIAquick Gel Extraction Kit) and their DNA sequence determined.

Material and Methods regA CLUSTER ISOLATION SCHEME BY SPECIES ALGAE CULTURES and CULTURE CONDITIONS Volvox carteri f. weismannia: genomic PCR with regF2 and Strains were grown in standard Volvox medium at 25°Cona16:8h regR primers yielded products of approximately 600 and 1100 light:dark cycle, at !35 µmol photons/m2/s on a rotary shaker. nucleotides; the sequence of the latter was homologous to the The following strains were used: V. carteri f. weismannia Ne- conserved intron in V. carteri f. nagariensis and V. carteri f. braska female (UTEX 1874), V. obversus male (UTEX 1865), V. kawasakiensis regA genes (Miller et al. 1999; Duncan et al. 2006). africanus (UTEX 1890), V. gigas (UTEX 1895), V. ferrisii (NIES PCR primers were used for inverse PCR of genomic DNA to ex- 2739), and V. rousseletii male (UTEX 1861). All strains were tend the sequence some 750 base pairs further upstream. We then

2016 EVOLUTION JULY 2014 EVOLUTION OF THE GENETIC BASIS FOR72 SOMA

used a direct genomic PCR approach. The V. carteri f. weis- pooled, subcloned, and sequenced. The sequence from two of mannia sequence was used to design forward PCR primers that eight subclones from the regF2/regR reactions encoded part of were paired with reverse primers (some slightly degenerate) de- the regA VARL domain. The sequence from one of 13 subclones signed from coding sequences in exon 6 (exons are numbered as from the regF3/regR reactions encoded part of the rlsB VARL do- for V. carteri f. kawasakiensis regA, Duncan et al. 2006) of the main. An oligomer hybridization primer (Table S1) was designed two known V. carteri regA genes (Cterm1, Cterm3, and Cterm4, from each of these for screening the V. ferrisii cosmid library see Table S1). Volvox carteri f. weismannia-specific reverse PCR (see below). The regA probe yielded three overlapping cosmids, primers were paired with forward primers from the 5′ end of exon one of which also hybridized to the rlsB probe. PCR using regF2

5(regNterm)andthe3′ end of exon 1 (int1). The PCR products or regF3 and regR using cosmid DNA templates identified two from the reactions using these five primers were sequenced (by products encoding additional VARLdomains (rlsA and rlsC). The primer walking when necessary). Finally, the promoter region was rlsN and ackB genes were identified during cosmid sequencing. amplified using a V. carteri f. weismannia-specific reverse PCR An oligomer hybridization primer (Table S1) was designed from primer from regA exon2 (rlsA2) with a forward primer rlsA1 (de- the DNA sequence around the starting methionine of ackB and signed from rlsA)andsequenced.ThefinalV. carteri f. weisman- used to screen the V. ferrisii cosmid library to identify an overlap- nia assembly (Genbank KF555603) extends from the promoter ping cosmid that proved to have the rlsD gene as determined by region of regA to within a few nucleotides of the predicted stop partial sequencing of the cosmid. The 80,171 nucleotide V. ferrisii codon. sequence is Genbank KF607039. Volvox obversus: genomic PCR with regF1 and regR yielded Volvox rousseletii: two sets of PCR primers (fer6F/fer6R and aproductof!2200 nucleotides. PCR primers designed from fer6F2/fer6R2, Table S1) were designed from the V. ferrisii rlsN this sequence were used for inverse PCR of genomic DNA to sequence and used in PCR to yield overlapping products that extend this sequence to include the complete nucleotide sequence span the sequence encoding the two rlsN VA R L d o m a i n s i n V. encoding the VARL domain (Genbank KF545953). rousseletii (Genbank KF615907). Volvox africanus: genomic PCR with regF2 and regR yielded two products of approximately 800 nucleotides. Both PCR prod- LIBRARY CONSTRUCTION ucts were sequenced; PCR primers were designed from the one Cosmid libraries from V. gigas and V. ferrisii were constructed most similar to regA (V. carteri f. nagariensis)andusedforin- from size-fractionated genomic DNA using the pWEB-TNC Cos- verse PCR of genomic DNA to extend this sequence to include mid Cloning Kit (Epicentre, Madison, WI), following the manu- the complete nucleotide sequence encoding the VARL domain facturer’s instructions. To create amplified pools of the libraries, (Genbank KF555602). the E. coli host strain EPI100-T1 was transduced with an aliquot Volvox gigas: genomic PCR with regF2 and regR (66°C of the packaged cosmid library and plated at a density of several annealing temperature) yielded multiple products. These were thousand colonies per plate. The colonies on individual plates cloned using the pGEM-T Easy Vector System (Promega, Madi- were resuspended separately in LB, DMSO added to 8% (v/v), son, WI) after A-tailing and transformation into NEB 5α com- then stored at 80°C. An aliquot from each pool was used to − petent Escherichia coli (New England Biolabs). Plasmid DNA inoculate a 5 ml overnight culture in LB/ampicillin (100 µg/ml) was prepared from ampicillin-resistant white colonies selected on from which cosmid DNA was prepared for use as PCR template Luria Broth (LB)/ampicillin/IPTG/X-GAL plates, then sequenced for ascertaining whether genes of interest were present within a using primers based on the T7 and SP6 promoter sequences within given pool. Pools of interest were plated on LB/ampicillin agar the pGEM-T Easy Vector. Sequences encoding part of three dif- plates at a density of 500–1000 colonies; after overnight incuba- ferent VARL domains were identified among the subclones: rlsA, tion, E. coli colonies were transferred to nitrocellulose (NitroPure rlsB,andrlsC.Threeoligomerhybridizationprimers(TableS1) supported nitrocellulose, 0.45 µm, 82 mm circle; Maine Manufac- were designed for screening the V. gigas cosmid library (see be- turing, LLC, Sanford, ME), lysed in situ, and the DNA bound to low). The regA gene was discovered during cosmid sequencing. the filter by UV crosslinking (Sambrook and Russell 2001). The rlsB appeared to be incomplete at the C-terminus, so the sequence nitrocellulose lifts were incubated overnight with radiolabeled was extended to a stop codon by TAIL-PCR, following Dent et al. probe in a rotary hybridization oven at a temperature 15°Cbe- (2005) using Taq 2X Master Mix (New England Biolabs). The V. low the Tm of the hybridization primer. The hybridization buffer gigas sequences are Genbank KF582790 and KF582789. was 5X SCP (5X SCP is 0.5 M NaCl, 0.05 M EDTA, 0.15 M Volvox ferrisii: genomic PCR with regF2/regF3 and regR was sodium phosphate, pH 6.6), 1% sodium N-lauroylsarcosine, 100 performed using an eight-step gradient in annealing temperature µg/ml fish sperm DNA, MB-grade (Roche, Basel, Switzerland). (60–72°C) yielding multiple products. The PCR products were Lifts were rinsed at room temperature in 2X SCP, 1% SDS (three

EVOLUTION JULY 2014 2017 E. R. HANSCHEN ET AL. 73

changes, 20 min each), then in 0.5X SCP, 1% SDS (two changes, the N-terminal extension and VARL domains. A codon-based 10 min each), then autoradiographed. Colonies showing positive Z-test for selection using the Nei–Gojobori method with 500 hybridization were subcloned and cosmid DNA prepared. Only bootstrap replicates was performed on codon-based nucleotide one cosmid was used from any given pool. alignments. Hybridization primers were end-labeled at 37°C for 60 min in a 20 µl reaction containing 25 pmole of primer, 50 pmole of γ-32P ATP (6000 Ci/mmol; PerkinElmer NEG035C005MC), Results and 10U polynucleotide kinase (Fermentas, Waltham, MA) in 1X PRESENCE OF VARL GENES PNK buffer A. Unincorporated ATP was removed using a Micro We sequenced almost the entire regA gene from V. carteri f. weis- G-25 Spin Column (Santa Cruz Biotechnology, Dallas, TX). mannia, as well as a portion of the regA gene of V. obversus and V. africanus that contains the VARL domain. In V. gigas,wese- DNA SEQUENCING quenced the regA cluster. In V. ferrisii,wesequencedatotalof DNA sequencing of PCR products, plasmids, and cosmids was six genes encoding VARL domains. Five of these genes encode a performed by the University of Arizona Genetics Core service single predicted VARLdomain and the last gene encodes two pre- using Applied Biosystems 3730 DNA Analyzers (Waltham, MA). dicted VARL domains. We used Pfam and SMART to predict the DNA sequences were assembled into contigs using the CLC Main presence of a SAND domain (Table S2). We have also sequenced Workbench software. five adjacent genes in V. ferrisii that are syntenic with the regA From this assembly, BLAST (version 2.2.26) was used to cluster in both V. ferrisii and V. carteri. identify the sequence encoding putative VARL domains which were confirmed with Pfam (version 26.0) and SMART (version NAMING CONVENTIONS FOR VARL GENES 7.0; Altschul et al. 1990; Finn et al. 2010; Letunic et al. 2012). A standardized nomenclature was used for sequenced genes, fol- Gene models were obtained with AUGUSTUS (version 2.6.1, lowing V. carteri f. nagariensis.Whenthephylogeneticsupport trained on C. reinhardtii and using known VARLdomains from C. of a predicted VARL domain clade was not above 0.70 (ML reinhardtii and V.carteri f. nagariensis as hints, Stanke et al. 2006) estimation), we used the syntenic relationships compared to V. and manual model building. N-terminal extensions were identified carteri f. nagariensis to provide names (Table S3). Otherwise, using custom Hidden Markov Models (HMM, HMMER version phylogenetic relationships were used to provide gene names. The 3.0, Eddy 1998). Syntenic non-VARL genes were identified by fourth gene in the regA cluster in V. ferrisii with two predicted BLAST searches and AUGUSTUS gene modeling. SAND domains was named rlsN (with domains rlsN1 and rlsN2). Lastly, the syntenic genes found in V. ferrisii were named using PHYLOGENETIC ANALYSES conventions from V. carteri f. nagariensis. All protein sequences (approximately 80 amino acids) were aligned using MAFFT (version 6.859b) with the accurate L-INS-i STRUCTURE OF THE VARL GENES option (Katoh et al. 2005). The corrected Akaike information cri- We used AUGUSTUS and manual model building to predict exon terion in ProtTest (version 3.2, Posada et al. 2011) determined and intron locations in the sequenced VARL genes in all studied LG Gastheoptimalsubstitutionmodel.Phylogenieswereper- species. In all sequences encoding VARL domains, a single con- + formed using maximum likelihood (ML), maximum parsimony, served intron position was found. This is position 4 from Duncan neighbor joining, and Bayesian methods. ML estimation was im- et al.’s (2007) analysis. The conserved intron position is further plemented using RAxML (version 7.0.4) with a rapid bootstrap evidence that we have sequenced orthologous genes in V. gigas analysis, generating 1000 ML replicate trees to estimate bootstrap and V. ferrisii. support (Stamatakis 2006; Stamatakis et al. 2008). Maximum par- The three V. carteri formae (nagariensis, kawasakiensis, and simony and neighbor joining were implemented using PHYLIP weismannia) are roughly equally diverged and show conservation (version 3.695) with 100 replicate datasets (Felsenstein 1989). For outside the VARL domain sequence but this conservation is not Bayesian analysis, MrBayes (version 3.2.1) was used to run four present in V. africanus and V. obversus.Wesearchedforconser- MCMC replicates for 1 million generations each (Ronquist et al. vation outside the sequence encoding VARL domains in genes 2012). The WAG I G model was used for direct comparison to isolated from V. gigas and V. ferrisii but were not able to identify + + the Bayesian tree in Duncan et al. (2007). After a 10% burn-in, functionally defined, conserved motifs. There are several homol- trees were sampled every 100 generations to build a 50% majority ogous sequences outside the region encoding the predicted VARL rule consensus tree. We used codeml in PAML (version 4.3, Yang domain in regA, rlsA, rlsB,andrlsC in V. carteri f. nagariensis 2007) to test for stabilizing selection on the sequence encoding and V. gigas (Supporting Information Alignments S1–S4). These

2018 EVOLUTION JULY 2014 EVOLUTION OF THE GENETIC BASIS FOR74 SOMA

short regions provided evidence of orthology and may play a role The syntenic relationships in V. ferrisii are more informa- in facilitating protein–protein interactions but are not currently tive (Fig. 3). Four syntenic gene markers (ido, rpt4, 105104,and functionally annotated. This lack of substantial homology outside 91985) lie upstream of the regA cluster in V. ferrisii and one gene the VARL domain hinders phylogenetic analysis and is consistent marker, ackB,liesdownstream.InV. carteri f. nagariensis, ackB with Duncan et al.’s (2007) attempt to identify conserved domains is immediately adjacent to rlsD but these two genes are located outside of VARL domains, suggesting that VARL genes indeed at least 1 Mb away from the regA cluster. In C. reinhardtii, ACK2 encode single-domain proteins. (the ortholog of ackB) is immediately adjacent to RLS1.InV. ferrisii, ackB and rlsD are closely linked to the regA cluster. PHYLOGENETIC RELATIONSHIPS OF THE VARL GENES We made phylogenetic trees using the boundaries of the VARL STABILIZING SELECTION ON THE VARL GENES domain and the N-terminal extension of the VARL domain as We used PAML to test for stabilizing selection on the sequence defined in Duncan et al. (2007) both for proper comparison to of predicted VARL domains. The gene rpt4 in C. reinhardtii, V. their work and because sequenced genes are consistent with these carteri f. nagariensis,andV. ferrisii (1197 nucleotides) has dN/dS domain boundaries. The VARL genes we have sequenced are values ranging from 0.030 to 0.050 (Tables S4 and S5). This in- closely related to the functional regA gene in V. carteri;however, dicates strong stabilizing selection, which is expected given that due to the short sequence of the VARL domain (87 amino acids) rpt4 is a subunit of the 26S proteasome regulatory polyprotein there is little phylogenetic support (Fig. 2). To compare our results associated with ubiquitin degradation (Finley 2009). The dN/dS to those of Duncan et al. (2007), we constructed gene phylogenies values for the regA genes we have sequenced in related and diver- using four methods (Bayesian, ML, maximum parsimony, and gent Volvox species range from 0.000 to 0.378 (261 nucleotides, neighbor joining). Tables S6 and S7). dN/dS values for RLS1 and rlsD for C. rein- We found strong support for the genes we report in V. obver- hardtii, V. carteri f. nagariensis,andV. ferrisii range from 0.087 sus, V. africanus,andV. carteri f. weismannia as regA in all four to 0.250 (261 nucleotides, Tables S8 and S9). These results indi- phylogenetic methods, which is consistent with the conserved in- cate that the predicted VARL domains of regA and RLS1 and rlsD tron locations as well as short conserved motifs outside the VARL have been subjected to stabilizing selection, which is consistent domain (Figs. 2, S2–S4). Our gene phylogenies also found strong with findings in Duncan et al. (2007). This stabilizing selection support for rlsA and rlsC in V. gigas (Figs. 2, S2–S4). suggests both regA and rlsD in Volvox and RLS1 in C. reinhardtii There are a few notable differences between our analysis and have been functionally conserved. This is consistent with the tem- the previous study (Duncan et al. 2007). First, although previous poral expression of RLS1 in stressful environments such as light work identified RLS2 in C. reinhardtii and rlsI in V. carteri f. na- and sulfur deprivation (Nedelcu 2009). gariensis as likely orthologs, there was no strong support for this relationship. In our phylogenetic analyses, RLS2 and rlsI form a ANALYSIS OF rlsN well-supported clade, which is consistent with their syntenic rela- Our analysis reconstructed rlsN2 in V. ferrisii and rlsM in V. car- tionship. Second, previous work identified RLS8 in C. reinhardtii teri f. nagariensis as a significant clade. However, because rlsN and rlsE in V.carteri f. nagariensis as potential orthologs, which is is within the tandem duplication in V. ferrisii, rlsN is likely the now supported in our analysis. Lastly, Duncan et al. (2007) found result of an expanded duplication not present in V. carteri.Alter- strong support for a clade composed of RLS1 in C. reinhardtii and natively, five VARL genes may be ancestral and V. carteri may rlsD in V. carteri f. nagariensis, which was the sister group to the have lost rlsN. rlsN also has a novel gene architecture, containing regA cluster in V. carteri f. nagariensis. In our analysis, this clade the sequence of two VARL domains that is not present in rlsM. of RLS1 and rlsD is supported, but its position as the sister group Therefore, we suspect this node does not represent evolutionary to the regA cluster is not well supported using any phylogenetic relationships, but rather phenomena, such as long branch attrac- method (Figs. 2, S2–S4). tion, that are beyond the scope of this article as this issue does not affect the remainder of our analysis. rlsN2 is the only domain that SYNTENIC RELATIONSHIPS OF THE VARL GENES SMART did not significantly annotate, so we performed the phy- We identified the syntenic structure of the genes in V. gigas and logenetic analyses with and without rlsN2 and found no difference V. ferrisii (Fig. 3). In V. carteri f. nagariensis, regA is part of in phylogenetic support. We did not find a significant N-terminal a tandem duplication including three other closely related VARL extension for rlsN2 with our HMM search. We performed the genes, rlsA, rlsB,andrlsC.InV. gigas,wefindthatrlsA, regA,and phylogenetic analyses with and without the N-terminal extension rlsB are also closely syntenic (Fig. 3). rlsC in V. gigas is located and there was no significant difference. on a separate, nonoverlapping cosmid, so the linkage relationship To investigate potential function of rlsN in V. ferrisii,wese- of rlsC to the rest of the tandem duplication is undetermined. quenced the region between predicted VARLdomains in a closely

EVOLUTION JULY 2014 2019 E. R. HANSCHEN ET AL. 75

RLS10 C. reinhardtii 99 rlsL V. carteri f. nagariensis rlsH V. carteri f. nagariensis 89 RLS9 C. reinhardtii 67 RLS5 C. reinhardtii RLS6 C. reinhardtii RLS3 C. reinhardtii rlsJ V. carteri f. nagariensis 51 88 rlsK V. carteri f. nagariensis RLS4 C. reinhardtii RLS7 C. reinhardtii RLS12 C. reinhardtii RLS2 C. reinhardtii 74 rlsI V. carteri f. nagariensis rlsG V. carteri f. nagariensis 56 RLS11 C. reinhardtii rlsE V. carteri f. nagariensis 57 RLS8 C. reinhardtii rlsF V. carteri f. nagariensis rlsD V. carteri f. nagariensis 58 rlsD V. ferrisii RLS1 C. reinhardtii rlsM V. carteri f. nagariensis 76 rlsN2 V. ferrisii rlsN1 V. ferrisii rlsA V. carteri f. nagariensis 97 rlsA V. gigas

regA V. obversus regA 70 regA V. africanus regA V. carteri f. nagariensis regA V. carteri f. weismannia

regA V. carteri f. kawasakiensis cluster regA V. ferrisii 89 51 regA V. gigas rlsB V. carteri f. nagariensis rlsC V. carteri f. nagariensis 74 rlsC V. gigas rlsB V. ferrisii rlsB V. gigas rlsA V. ferrisii 0.1 rlsC V. ferrisii

Figure 2. Phylogenetic relationships of VARL domains using maximum likelihood methods. Protein sequences from the N-terminal extension and core VARL domain were used. The tree is a mid-point root and numerical values represent bootstrap values when above 50%. Sequences in bold are reported here. The regA cluster, which includes all members of the tandem duplication, is denoted by vertical bar. related species, V. rousseletii (Isaka et al. 2012). This region has a lost in C. reinhardtii. Because we are unable to replicate strong dN/dS value of 0.037 between V. ferrisii and V. rousseletii (1176 support for their tree topology, in the next section we consider an nucleotides, 96.3% identity). This suggests that the interdomain alternate evolutionary history of regA. region is under stabilizing selection and is functionally expressed, We have investigated the evolutionary history of regA by although the function of rlsN is unknown. rlsN has a unique gene sequencing VARLgenes from four species of Volvox,bothclosely structure of two predicted VARLdomains not previously observed related and highly diverged. The sequences from V. africanus, in the volvocine algae. This novel structure suggests unique func- V. obversus,andV. carteri f. weismannia are closely related to tion which would be worthwhile investigating. previously reported regA sequences in V. carteri f. nagariensis and V. carteri f. kawasakiensis. As all these species have a D2 developmental program, we cannot infer the role of regA in other Discussion developmental programs from these species (Fig. 1). In V. gigas VARL GENE ORTHOLOGY (D1) and V. ferrisii (D4), we have demonstrated the presence of Previous analyses of the evolutionary history of regA have been the regA cluster. This is the first demonstration of regA genes in a restricted to C. reinhardtii and V. carteri,bothofwhichhave Volvox species with a non-D2 developmental program and equal published genomes (Merchant et al. 2007; Prochnik et al. 2010). embryonic cleavage. Under the assumption that these genes are Duncan et al. (2007) used these genomes to study the evolutionary functional, as our dN/dS results suggest, this result demonstrates history of regA and the VARL gene family and based on their tree that regA is not restricted to species with unequal embryonic topology hypothesized the ancient existence of a “proto-regA” cleavage, and likely plays a role in somatic differentiation in gene which ultimately gave rise to regA in V. carteri and was species with and without unequal embryonic cleavage.

2020 EVOLUTION JULY 2014 EVOLUTION OF THE GENETIC BASIS FOR76 SOMA

Figure 3. Gene synteny of VARL genes in C. reinhardtii (blue), V. carteri f. nagariensis (red), V. gigas (yellow), and V. ferrisii (green). Chromosome or scaffold number is indicated. Positions of VARL genes (gray) and neighboring genes are shown (black). Highly conserved genes are linked by line segments for both VARL genes (gray) and neighboring genes (black).

We observe phylogenetic and syntenic relationships between event giving rise to regA, or the entire regA cluster, that occurred C. reinhardtii, V. carteri,andV. ferrisii (Figs. 2 and 3). The rlsD before the speciation of the Chlamydomonas and Volvox lineages gene and the regA cluster in V. ferrisii are separated by approxi- (Fig. S1). Our data do not allow us to conclusively choose between mately 7.6 kb including one syntenic marker, ackB. This result is these alternatives. in contrast to V. carteri f. nagariensis synteny where rlsD and the Duncan et al. (2007) used an inferred close phylogenetic regA cluster are separated by at least 1 Mb of sequence (i.e., the relationship between RLS1 in C. reinhardtii and rlsD in V. carteri minimum distance if scaffolds 23 and 31 in V. carteri version 1 to hypothesize an ancient ancestor with two VARL genes, the are adjacent). ancestor to RLS1 and rlsD,whichweterm“proto-RLS1”and proto-regA (Fig. S1). Under this hypothesis, the ancestor proto- EVOLUTIONARY ORIGIN OF REGA RLS1 became the orthologous genes RLS1 and rlsD after the We have discovered that the regA cluster (rlsA, regA, rlsB, rlsN, Chlamydomonas and Volvox lineages speciated. Duncan et al. rlsC)andrlsD are closely linked in V. ferrisii and have sequenced (2007) hypothesized that the proto-regA gene was lost in the C. five genes that are syntenic with both the V. carteri f. nagariensis reinhardtii lineage but not the Volvox lineage thereby producing regA clade and C. reinhardtii RLS1 (Fig. 3). The regA cluster their observed gene phylogeny. is likely the result of tandem duplications, perhaps by repeated However, with increased sampling of the regA cluster in V. unequal crossing over or backward strand slippage (Chen et al. gigas and V. ferrisii, we have been unable to replicate strong 2005; Kondrashov and Kondrashov 2006). We hypothesize that support for the close phylogenetic relationship of RLS1 and rlsD, the regA cluster duplicated from rlsD more recently than the or for the clade containing RLS1 and rlsD as the sister group to the speciation of the Chlamydomonas and Volvox lineages (Fig. 4). In regA cluster (Figs. 2, S2–S4). The existence of a clade containing contrast, Duncan et al. (2007) hypothesized an early duplication RLS1 and rlsD sister to the regA cluster observed by Duncan et al.

EVOLUTION JULY 2014 2021 E. R. HANSCHEN ET AL. 77

Figure 4. Model schematic for the revised evolutionary origin of regA. VARL genes are indicated by gray boxes, marker genes are indicated by black boxes, and larger synteny flanking regions as black rectangles. Not to scale. Evolutionary events such as speciation, gene duplication, and recombination are denoted in hypothesized order, the order of events 4a and 4b is unknown. ackB and ACK2, acetate kinase gene.

(2007), which we have been unable to confirm with increased Michod 2006; Nedelcu 2009). Although Duncan et al. (2007) hy- sampling of Volvox species, was critical to the development of pothesize regA, with unknown function, evolved in a unicellular their “proto-regA” hypothesis. species, under our hypothesis, regA evolved in a colonial alga Due to the low support of this aspect of our phylogenetic where spatial differentiation may have been beneficial (Herron trees, our hypothesis of more recent duplication events is not et al. 2009). explicitly based on gene phylogenies as was the proto-regA The proto-regA hypothesis posits an early evolution of regA hypothesis. We also note that our observed phylogenetic trees and predicts that regA should be found in a variety of algae be- (Figs. 2, S2–S4) do not reconstruct the topology of our revised tween Volvox and Chlamydomonas.Ontheotherhand,wehy- evolutionary history either, which predicts RLS1 as the outgroup pothesize a later evolution of regA from rlsD and predict that to the regA cluster and rlsD.Wedonotthinkthisisaprob- rlsD,butnotregA, will be found in other unicellular and colonial lem because stabilizing selection on RLS1 and rlsD (Table S8) volvocine algae. The hypotheses would be conclusively tested and the limited conserved nucleotide sequence (261 nucleotides) by sequencing the genomes of colonial volvocine algae such as available for phylogenetic analysis may cause the aforementioned Tetrabaena socialis or G. pectorale as these are early-diverging uncertainty of this aspect of the tree topology (Figs. 2, S2–S4). colonial species. Understanding when the genetic basis for somatic cells evolved will shed light into how and why co-option of a life- IMPLICATIONS FOR MULTICELLULARITY history gene for suppression of reproduction may have occurred There are many species with and without somatic cells in the during the evolution of multicellularity. Our revised evolutionary clade containing V. ferrisii and V. carteri (Nozaki et al. 2002, history is consistent with the hypothesis that regA evolved by co- 2006). These include species of Yamagishiella, Eudorina,and option of a life-history gene in a unicellular ancestor by changing Pleodorina, which contain species with no somatic cells, facul- its expression from a temporal to spatial context (Nedelcu and tative somatic cells, and obligate somatic cells (Fig. 1). Previous

2022 EVOLUTION JULY 2014 EVOLUTION OF THE GENETIC BASIS FOR78 SOMA

phylogenetic analyses have predicted that somatic cells are ances- the volvocine algae. Such a result would suggest that evolving tral to Eudorina and Pleodorina but not Yamagishiella (Herron somatic cells is easier than previously understood, as multiple et al. 2009). However, based on our results, a genetic component genetic pathways are available for complex traits such as cellular thought to be essential to somatic differentiation, the regA cluster, differentiation and individuality (Ardent and Reznick 2008). is ancestral to all of these species (Fig. 1). For alternative (3), it is known that independent tandem du- Although the genetic basis for soma is ancestral to lineages plications have important effects on evolutionary processes such with and without somatic cells, it does not mean the common as host–parasite interactions and life cycle evolution (Hughes and ancestor had somatic cells. Herron and Michod (2008) used an- Friedman 2003; Danchin et al. 2010; Cooley et al. 2011), so the cestral state reconstructions to infer that the last common ancestor possibility of an independent tandem duplication of rlsD produc- of V. ferrisii and V.carteri did not have somatic cells and predicted ing the regA cluster in V. ferrisii cannot be ignored. However, that two lineages, the Euvolvox (V. ferrisii)andthelastcommon given the syntenic position of the regA cluster and similar gene ancestor of V. gigas and V. carteri,independentlyevolvedso- orientation in V. carteri and V. ferrisii (Fig. 3), we do not think matic cells. This prediction, supported by hypothesis testing, was this likely. If independent duplications did occur, it would have reached when cellular differentiation was treated as an unordered important implications on the likelihood of evolving somatic dif- three-state character (undifferentiated, differentiated soma cells, ferentiation due to the constrained number of genetic pathways differentiated soma and germ cells), but not when somatic cells and specific mutations required for this pathway. were modeled as a separate character from differentiated germ Although we are not currently able to differentiate among cells. Consequently, the possibility that somatic cells are ancestral these explanations, under all of these possibilities, there have to all species included in this analysis must be acknowledged. If been multiple events regarding the history of somatic cells and somatic cells are ancestral to all these species, it remains to be seen individuality in the volvocine algae. Our analysis demonstrates if, and how, the regA cluster is functional in species which have lost the availability of regA as the genetic basis for soma in both soma. The regA cluster may be facilitating a spatial cue to undif- V. carteri and V. ferrisii lineages as well as lineages without ferentiated cells or a temporal cue based on environmental change somatic cells. Functional analysis of regA in V. ferrisii could (Nedelcu and Michod 2006; Nedelcu 2009; Arakaki et al. 2013). differentiate between explanations (1) and (2). Detailed analy- Either way, our results imply a phylogenetic discordance sis of VARL gene presence or absence, and subsequent func- between somatic phenotype and genotype. If somatic cells are tion, in colonial species such as Yamagishiella unicocca and E. not ancestral to the species studied here, the observation that the elegans is important to testing possibility (3). If these species somatic genotype, regA, is ancestral to species without somatic all have the regA cluster (even if psuedogenized), hypothesis cells raises three interesting possibilities regarding the regulation (3) would be disproven. We are currently investigating these of soma: (1) both the V. carteri and V. ferrisii lineages regulate possibilities. somatic cells via regA,(2)theV. ferrisii lineage does not regulate The first two possibilities are particularly interesting given somatic cells via regA,or(3)asin(1)theregA cluster in V. the number of lineages that did not evolve somatic cells despite carteri and V. ferrisii is used for soma but these clusters represent their ancestor containing regA (R, Fig. 1) and thereby have not independent, recent tandem duplications of rlsD. further increased in complexity and individuality. Why have these Under (1), the V. carteri and V. ferrisii lineages have evolved species not evolved soma given the genetic basis for soma? Were the use of regA for somatic regulation in parallel from an an- the costs of soma simply too great, the required mutations not cestor containing regA but not soma. This alternative would present, or the immediate advantages nonexistent in their envi- highlight the role of repeatability and convergence in the evo- ronments? Identifying the environment in which somatic cells lution of soma. Reproductive altruism is thought to be a fun- are not advantageous would be valuable to understanding the damental property of all multicellular individuals and eusocial evolution of multicellularity and individuality more generally. groups. Although recent experimental work has investigated how If somatic cells are ancestral to the species studied here, what repeatability impacts evolution (Blount et al. 2008; Meyer et al. function, if any, has the genetic potential for soma taken on 2012), this possibility would suggest a more important role of in these species? Answers to these questions will be needed to convergence in the evolution of individuality than previously fully understand the evolution of multicellularity and individual- appreciated. ity in this group. Nevertheless, our results imply that individual- Under (2), as in (1) there is parallel phenotypic evolution ity, the trait which underlies the levels in the hierarchy of life, is of soma, but V. ferrisii uses some other genetic means for so- more open to evolutionary change than previous theoretical work matic regulation. This would also be an important result as it envisioned (Buss 1987; Maynard Smith and Szathmary´ 1995; suggests multiple genetic pathways for evolving somatic cells in Michod 1999).

EVOLUTION JULY 2014 2023 E. R. HANSCHEN ET AL. 79

ACKNOWLEDGMENTS The Pfam protein families database. Nucleic Acids Res. 38:D211– We would like to thank Q. Li, C. Dieckmann, T. Mittelmeier, M. Barker, 222. and B. J. S. C. Olson for technical help, D. Shelton, Z. Grochau-Wright, Finley, D. 2009. Recognition and processing of ubitiquin-protein conjugates and S. Miller for discussions and comments, J. Umen for genomic DNA, by the proteosome. Annu. Rev. Biochem. 78:477–513. H. Nozaki for providing algae strains, and three reviewers for helpful Goldstein, M. 1964. Speciation and mating behavior in Eudorina.J.Protozool. comments. This work was supported by the National Science Founda- 11:317–344. tion (grant numbers DEB-0742383, DGE-0654435); and the National Harper, J. F., K. S. Huson, and D. L. Kirk. 1987. Use of repetitive sequences Aeronautics and Space Administration (grant number NNX13AH41G). to identify DNA polymorphisms linked to regA, a developmentally im- portant locus in Volvox.GenesDev.1:573–584. Harryman, A. Y. 2012. Investigating the roles of regA and related genes in LITERATURE CITED the evolution of multicellularity in the volvocine green algae. Retrieved Arendt, J., and D. Reznick. 2008. Convergence and parallelism reconsidered: from ProQuest Dissertations and Theses (Thesis No. 3539470). what have we learned about the genetics of adaptation? Trends Ecol. Herron, M. D., and R. E. Michod. 2008. Evolution of complexity in the Evol. 23:26–32. volvocine algae: transitions in individuality through Darwin’s eye. Evo- Adams, C. R., K. A. Stamer, J. K. Miller, J. G. McNally, M. M. Kirk, and lution 62:436–451. D. L. Kirk. 1990. Patterns of organellar and nuclear inheritance among Herron, M. D., J. D. Hackett, F. O. Aylward, R. E. Michod. 2009. Triassic progeny of two geographically isolated strains of Volvox carteri.Curr. origin and early radiation of multicellular volvocine algae. Proc. Natl. Genet. 18:141–153. Acad. Sci. USA 106:3254–3258. Altschul, S., W. Gish, W. Miller, E. Myers, and D. Lipman. 1990. Basic local Hughes, A. L., and R. Friedman. 2003. Parallel evolution by gene duplication alignment search tool. J. Mol. Biol. 215:403–410. in the genomes of two unicellular fungi. Genome Res. 604:794–799. Arakaki, Y., H. Kawai-Toyooka, Y. Hamamura, T. Higashiyama, A. Noga, M. Husky, R. J., and B. E. Griffin. 1979. Genetic control of somatic cell differ- Hirono, B. J. S. C. Olson, and H. Nozaki. 2013. The simplest integrated entiation in Volvox. Analysis of somatic regenerator mutants. Dev. Biol. multicellular organism unveiled. PLoS ONE 8:e81641. 72:226–235. Blount, Z. D., C. Z. Borland, and R. E. Lenski. 2008. Historical contingency Isaka, N., H. Kawai-Toyooka, R. Matsuzaki, T. Nakada, and H. Nozaki. and the evolution of a key innovation in an experimental population of 2012. Description of two new monoecious species of Volvox sect. Escherichia coli. Proc. Natl. Acad. Sci. USA 105:7899–7906. Volvox (Volvocaceae, Chlorophyceae), based on comparative morphol- Bonner, J. T. 1998. The origins of multicellularity. Integr. Biol. 1:27–36. ogy and molecular phylogeny of cultured material. J. Phycol. 48:759– Buss, L. W.1987. The evolution of individuality. Princeton Univ. Press, Prince- 767. ton, NJ. Iyengar, M. O. P., and T. V. Desikachary. 1981. Volvocales. Indian Council of Chen, J., N. Chuzhanova, P. D. Stenson, C. Ferec, and D. N. Cooper. 2005. Agricultural Research, New Delhi. Meta-analysis of gross insertions causing human genetic disease: novel Katoh, K., K. Kuma, H. Toh, and T. Miyata. 2005. MAFFT version 5: im- mutational mechanisms and the role of replication slippage. Hum. Mutat. provement in accuracy of multiple sequence alignment. Nucleic Acids 25:207–221. Res. 33:511–518. Coleman, A. W., A. Suarez, and L. J. Goff. 1994. Molecular delineation Kirk, D. L. 1998. Volvox: molecular-genetic origins of multicellularity and of species and syngens in Volvocacean green algae (Chlorophyta). J. cellular differentiation. Cambridge Univ. Press, Cambridge, U.K. Phycol. 30:80–90. ———. 2005. A twelve-step program for evolving multicellularity and a Cooley, A. M., J. L. Modliszewski, M. L. Rommel, and J. H. Willis. 2011. division of labor. Bioessays 27:299–310. Gene duplication in Mimulus underlies parallel floral evolution via in- Kirk, D. L., G. J. Baran, J. F. Harper, R. J. Huskey, K. S. Juson, and N. Zagris. dependent trans-regulatory changes. Curr. Biol. 21:700–704. 1987. Stage-specific hypermutability of the regA locus of Volvox,agene Danchin, E. G. J., M. Rosso, P. Vieira, J. de Almeida-Engler, P. M. Coutinho, regulating the germ-soma dichotomy. Cell 48:11–24. B. Henrissat, and P. Abad. 2010. Multiple lateral gene transfers and Kirk, M., K. Stark, S. Miller, B. E. Taillon, H. Gruber, R. Schmitt, and D. duplications have promoted plant parasitism ability in nematodes. Proc. L. Kirk. 1999. regA,aVolvox gene that plays a central role in germ- Natl. Acad. Sci. USA 107:17651–17656. soma differentiation, encodes a novel regulatory protein. Development Dent, R. M., C. M. Haglund, B. L. Chin, M. C. Kobayashi, and K. K. 126:639–647. Niyogi. 2005. Functional genomics of eukaryotic photosynthesis using Kondrashov, K. A., and A. S. Kondrashov. 2006. Role of selection in fixation insertional mutagenesis of Chlamydomonas reinhardtii.PlantPhysiol. of gene duplications. J. Theor. Biol. 239:141–151. 137:545–556. Leliaert, F., D. R. Smith, H. Moreau, M. D. Herron, H. Verbruggen, C. F. Desnitski, A. 1995. A review on the evolution of development in Volvox- Delwiche, and O. De Clerck. 2012. Phylogeny and molecular evolution morphological and physiological aspects. Eur. J. Protistol. 31:241–247. of the green algae. Crit. Rev. Plant Sci. 31:1–46. Duncan, L., I. Nishii, A. Howard, D. Kirk, and S. M. Miller. 2006. Orthologs Letunic, I., T. Doerks, and P. Bork. 2012. SMART 7: recent updates to and paralogs of regA, a master cell-type regulatory gene in Volvox carteri. the protein domain annotation resource. Nucleic Acids Res. 40:D302– Curr. Genet. 50:61–72. 305. Duncan, L., I. Nishii, A. Harryman, S. Buckley, A. Howard, N. R. Friedman, Maynard Smith, J., and E. Szathmary.´ 1995. The major transitions in evolution. and S. M. Miller. 2007. The VARL gene family and the evolutionary Oxford Univ. Press, Oxford, U.K. origins of the master cell-type regulatory gene, regA, in Volvox carteri. Merchant, S. S., S. E. Prochnik, O. Vallon, E. H. Harris, S. J. Karpowicz, J. Mol. Evol. 65:1–11. G. B. Witman, A. Terry, A. Salamov, L. K. Fritz-Laylin, L. Marechal-´ Eddy, S. R. 1998. Profile hidden Markov models. Bioinformatics 14:755–763. Drouard, et al. 2007. The Chlamydomonas genome reveals the evolution Felsenstein, J. 1989. PHYLIP—Phylogeny Inference Package (version 3.2). of key animal and plant functions. Science 318:245–250. Cladistics 5:164–166. Meissner, M., K. Stark, B. Cresnar, D. L. Kirk, and R. Schmitt. 1999. Volvox Finn, R. D., J. Mistry, J. Tate, P. Coggill, A. Heger, J. E. Pollington, germline-specific genes that are putative targets of RegA repression en- O. L. Gavin, P. Gunasekaran, G. Ceric, K. Forslund, et al. 2010. code chloroplast proteins. Curr. Genet. 36:363–370.

2024 EVOLUTION JULY 2014 EVOLUTION OF THE GENETIC BASIS FOR80 SOMA

Meyer, J. R., D. T. Dobias, J. S. Weitz, J. E. Barrick, R. T. Quick, and R. E. analysis of organismal complexity in the multicellular green alga Volvox Lenski. 2012. Repeatability and contingency in the evolution of a key carteri.Science329:223–226. innovation in phage lambda. Science 335:428–432. Rayburn, W. R., and R. C. Starr. 1974. Morphology and nutrition of Pandorina Michod, R. E. 1999. Darwinian dynamics, evolutionary transitions in fitness unicocca sp. nov. J. Phycol. 10:42–49. and individuality. Princeton Univ. Press, Princeton, NJ. Ronquist, F., M. Teslenko, P. van der Mark, D. L. Ayres, A. Darling, S. Hohna,¨ ———. 2007. Evolution of individuality during the transition from unicellular B. Larget, L. Liu, M. A. Suchard, and J. P. Huelsenbeck. 2012. MrBayes to multicellular life. Proc. Natl. Acad. Sci. USA 104:8613–8618. 3.2: efficient Bayesian phylogenetic inference and model choice across Michod, R. E., and D. Roze. 1999. Cooperation and conflict in the evolution alargemodelspace.Syst.Biol.61:539–542. of individuality. III. Transitions in the unit of fitness. Pp. 47–91 in C. L. Sambrook, J., and D. W. Russell. 2001. Molecular cloning: a laboratory man- Nehaniv, ed. Mathematical and computational biology: computational ual. 3rd ed. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, morphogenesis, heirarchical complexity, and digital evolution. Lectures NY. on Mathematics in the Life Sciences, Vol. 26. American Mathematical Schlichting, C. D. 2003. Origins of differentiation via phenotypic plasticity. Society, Providence, RI. Evol. Dev. 5:98–105. Miller, S. M., and D. L. Kirk. 1999. glsA,aVolvox gene required for asym- Simpson, C. 2012. The evolutionary history of division of labour. Proc. R. metric division and germ cell specification, encodes a chaperone-like Soc. Lond. B Biol. Sci. 279:116–121. protein. Development 126:649–658. Stamatakis, A. 2006. RAxML-VI-HPC: maximum likelihood-based phyloge- Nedelcu, A. M. 2009. Environmentally induced responses co-opted for repro- netic analyses with thousands of taxa and mixed models. Bioinformatics ductive altruism. Biol. Lett. 5:805–808. 22:2688–2690. Nedelcu, A. M., and R. E. Michod. 2006. The evolutionary origin of an Stamatakis, A., P. Hoover, and J. Rougemont. 2008. A rapid bootstrap algo- altruistic gene. Mol. Biol. Evol. 23:1460–1464. rithm for the RAxML Web servers. Syst. Biol. 57:758–771. Nishii, I., and S. Miller. 2010. Volvox: simple steps to developmental com- Stanke, M., O. Schoffmann,¨ B. Morgenstern, and S. Waack. 2006. Gene plexity? Curr. Opin. Plant Biol. 13:646–653. prediction in eukaryotes with a generalized hidden Markov model that Nozaki, H., M. Takahara, A. Nakazawa, Y. Kita, T. Yamada, H. Takano, uses hints from external sources. BMC Bioinformatics 7:62–72. S. Kawano, and M. Kato. 2002. Evolution of rbcL group IA introns Starr, R. C. 1970. Control of differentiation in Volvox.Dev.Biol.4:59–100. and intron open reading frames within the colonial Volvocales (Chloro- Yamada, T. K., Y. Tsuchikane, J.-T. Wu, H. Sekimoto, K. Miyaji, and phyceae). Mol. Phylogenet. Evol. 23:326–338. H. Nozaki. 2008. Morphology and molecular phylogeny of Eudo- Nozaki, H., F. D. Ott, and A. W. Coleman. 2006. Morphology, molecular phy- rina sp. (Volvocaceae, Chlorophyceae) from Taiwan. Hikobia 15:135– logeny and taxonomy of two new species of Pleodorina (Volvocaceae, 143. Chlorophyceae). J. Phycol. 42:1072–1080. Yang, Z. 2007. PAML 4: a program package for phylogenetic analysis by Posada, D., D. Darriba, and G. L. Taboada. 2011. ProtTest 3: fast selection of maximum likelihood. Mol. Biol. Evol. 24:1586–1591. best-fit models of protein evolution. Bioinformatics 27:1–2. Prochnik, S. E., J. Umen, A. M. Nedelcu, A. Hallmann, S. M. Miller, I. Nishii, P. Ferris, A. Kuo, T. Mitros, L. K. Fritz-Laylin, et al. 2010. Genomic Associate Editor: L. Fishman

Supporting Information Additional Supporting Information may be found in the online version of this article at the publisher’s website:

Table S1. Sequences of relevant oligonucleotides. Table S2. E-values of sequenced VARL domains for both Pfam and SMART gene annotation methods. Table S3. Naming conventions used for specific gene names. Table S4. dN/dS values for rpt4. Table S5. dN (lower triangle) and dS (upper triangle) values for rpt4. Table S6. dN/dS values for regA. Table S7. dN (lower triangle) and dS (upper triangle) values for regA. Table S8. dN/dS values for RLS1 and rlsD. Table S9. dN (lower triangle) and dS (upper triangle) values for RLS1 and rlsD. Figure S1. Model schematic for the previous hypothesis for the evolutionary origin of regA (Duncan et al. 2007). Figure S2. Phylogenetic relationships of VARL domains using Bayesian methods, using identical protein sequences as Figure 1. Figure S3. Phylogenetic relationships of VARL domains using maximum parsimony methods, using identical protein sequences as Figure 1. Figure S4. Phylogenetic relationships of VARL domains using neighbor joining methods, using identical protein sequences as Figure 1.

EVOLUTION JULY 2014 2025 81

APPENDIX D. MULTICELLULARITY DRIVES THE EVOLUTION OF SEXUAL TRAITS

Hanschen, E. R., M. D. Herron, J. J. Wiens, H. Nozaki, R. E. Michod. Multicellularity drives the

evolution of sexual traits. Prepared for submission to The American Naturalist.

82

1 Title 2 Multicellularity drives the evolution of sexual traits 3 4 Running title 5 Multicellularity drives the evolution of sex 6 7 Authors 8 Erik R. Hanschen1* [email protected] 9 Matthew D. Herron2 [email protected] 10 John J. Wiens1 [email protected] 11 Patrick J. Ferris1 [email protected] 12 Hisayoshi Nozaki3 [email protected] 13 Richard E. Michod1 [email protected] 14 15 Affiliations 16 1Department of Ecology and Evolutionary Biology, University of Arizona 17 2School of Biological Sciences, Georgia Institute of Technology 18 3Department of Biological Sciences, University of Tokyo 19 20 *Corresponding author 21 Erik R. Hanschen, [email protected] 22 Department of Ecology and Evolutionary Biology 23 University of Arizona 24 1041 E. Lowell Street, Tucson AZ, 85712 25 520-621-1844 26 27 Author contributions 28 E.R.H. conceived the project, collected species character data, described strains, 29 sequenced chloroplast genes, constructed the species tree, performed statistical analyses, 30 and wrote the manuscript; M.D.H. conceived the project and collected species character 31 data; J.J.W. conceived experimental design; P.J.F. collected and described strains; H.N. 32 collected species character data; R.E.M. conceived the project; all authors contributed to 33 preparation of the manuscript. 34 35 Acknowledgements 36 We thank Dinah R. Davison and Daniel S. Moen for technical assistance, Deborah E. 37 Shelton for collecting strains, and Dinah R. Davison and Sarah P. Otto for their helpful 38 comments and discussion. We acknowledge the support of the National Aeronautics and 39 Space Administration (NNX13AH41G, NNX15AR33G, Cooperative Agreement Notice 40 7), the National Institutes of Health (GM084905), the National Science Foundation 41 (MCB-1412395, DEB-1457701), and MEXT/JSPS KAKENHI Grants-in-Aid for 42 Scientific Research (15K14590 and 16H02518). 43 44 Abstract 45 From the male peacock’s tail plumage to the floral displays of flowering plants, traits 46 related to sexual reproduction are often complex and exaggerated. But why has sexual

1 83

1 reproduction become so complicated? Why have such exaggerated sexual traits evolved? 2 Early work posited a connection of multicellularity with sexual traits such as anisogamy 3 (i.e., the evolution of sperm and eggs), which in turn drives the evolution of other forms 4 of sexual dimorphism. Yet, the relationship between multicellularity and the evolution of 5 sexual traits has not been empirically tested. Due to high morphological variation 6 between species, the volvocine green algae offer a tractable system for understanding the 7 interrelationship of multicellular complexity and sex. Here we show that species with 8 higher metrics of multicellular complexity have significantly more derived sexual traits, 9 including anisogamy, internal fertilization, and exaggerated sexual dimorphism. Our 10 results demonstrate that anisogamy repeatedly evolved from isogamous multicellular 11 ancestors and that anisogamous species are larger and produce larger zygotes than 12 isogamous species. In the volvocine algae, the evolution of multicellularity likely drives 13 the evolution of anisogamy, and anisogamy subsequently drives exaggerated sexual 14 dimorphism, suggesting that multicellularity sets the stage for the overall diversity of 15 sexual complexity throughout the Tree of Life. 16 17 Keywords sex, volvocine green algae, multicellularity, phylogenetics, ancestral state 18 reconstruction, anisogamy, sexual dimorphism 19 20 The diversity of sexual reproduction throughout the Tree of Life is astonishing - from the 21 fusion of monomorphic gametes in single-celled organisms to the bizarre mating rituals 22 and strikingly dimorphic plumage of the birds of paradise. What explains this diversity of 23 sexual biology and the evolution of sexual traits? More elaborate forms of sexual 24 reproduction depend upon female and male sexes, which arise after the evolution of 25 anisogamy (Lehtonen et al. 2016b), the existence of dimorphic gametes. Anisogamy has 26 almost certainly evolved repeatedly from isogamous ancestors (equal sized gametes) 27 (Maynard Smith 1982; Togashi and Cox 2011; Beukeboom and Perrin 2014), resulting in 28 the evolution of male and female. However, because males do not produce offspring, sex 29 in anisogamous species has a two-fold cost, also known as the cost of males, compared to 30 isogamous relatives (Maynard Smith 1978). Despite this cost, most multicellular taxa are 31 anisogamous (Bell 1982), thus raising the question why are multicellular taxa mostly 32 anisogamous? The evolution of anisogamy may be selected for by divergent selection for 33 increased number of gametes and increased survival of zygotes (Parker et al. 1972; Bell 34 1978, 1982), where zygote survival is an increasing function of zygote size. Multicellular 35 species are hypothesized to require increased zygotic provisioning in order to provision a 36 larger embryo (Parker et al. 1972; Bulmer and Parker 2002), selecting for increased 37 zygote size. According to this hypothesis, as zygote size increases, the isogamous 38 evolutionary stable state is disrupted, and anisogamy evolves in order to provision the 39 increasing zygote size. Thus, multicellularity is predicted to drive the evolution of 40 anisogamy by the need for increasing zygote size (Bulmer and Parker 2002). Divergent 41 selection for increased number of gametes and increased zygote survival results in 42 numerous sperm (increased number of gametes) and few, larger eggs (increased zygote 43 survival), resulting in pronounced gamete dimorphism (Parker et al. 1972). Anisogamy 44 subsequently sets the stage for the secondary evolution of sex roles and exaggerated 45 sexual dimorphism (Bateman 1948; Parker et al. 1972; Lehtonen et al. 2016b).

2 84

1 The volvocine algae are a uniquely valuable model system for studying the 2 evolution of sex and of multicellularity. Previous studies showed that the ancestor of the 3 group was unicellular (Herron and Michod 2008) and multicellularity evolved relatively 4 recently at ~230 Myr ago (Herron et al. 2009). Furthermore, several genomes have been 5 sequenced (Merchant et al. 2007; Prochnik et al. 2010; Hanschen et al. 2016). Finally, the 6 extant species exhibit remarkable morphological diversity. All species are haploid and 7 facultatively sexual; morphology includes unicellular, undifferentiated multicellular, and 8 differentiated multicellular species (Fig. 1A-F). Based on this diversity of morphological 9 complexity, we do not analyze multicellularity as a binary trait (Santelices 1999; Clarke 10 2013; Hanschen et al. 2017), but rather use three metrics of multicellular complexity: (1) 11 rounds of cell division (i.e., log2[cell number]), (2) percent of somatic cells and (3) 12 natural logarithm of body length. These metrics capture the variation in morphological 13 complexity through increased cell number, division of labor, and secretion of 14 extracellular matrix (a major component of increased body size), that is important for 15 multicellularity in the volvocine algae. 16 Sexual morphology in the volvocine algae includes isogamy (equal sized gametes 17 produced by self-incompatible (–) and (+) mating types), anisogamy (smaller sperm, 18 bundled into “sperm packets”, produced by males and larger, flagellated eggs produced 19 by females), and oogamy (a form of anisogamy where eggs are unflagellated and much 20 larger than sperm, Fig. 1H). Fusion and fertilization of these gametes can occur both 21 external and internal to a female colony (Fig. 1I). Given that fertilization internal to a 22 multicellular body is only possible after the evolution of multicellularity, internal 23 fertilization is hypothesized to correlate with the evolution of multicellular complexity. 24 After fertilization, a desiccation-resistant, diploid zygospore is formed. Upon 25 germination, in some species, four meiotic progeny enter the asexual life cycle. In other 26 species, a reduced number of meiotic progeny (one or rarely two) enter the asexual life 27 cycle while the remaining meiotic products are discarded as polar bodies (Fig. 1G). A 28 reduced number of meiotic products may be related to the evolution of multicellularity 29 because it increases zygotic provisioning per sexual offspring, similar to the effect of 30 anisogamy. When a reduced number of surviving meiotic products evolves, zygotic 31 provisioning per sexual offspring may increase substantially, assuming zygote size 32 remains constant or increases. 33 Secondary sexual dimorphism is exhibited through extra-fertile females (the 34 number of eggs in a female colony is more than twice the number of reproductive gonidia 35 in an asexual colony, sometimes called “special females” (Smith 1944; Nozaki and Itoh 36 1994; Nozaki 2003, Fig. 1J), multicellular sperm packets, and dwarf males (male colonies 37 are 1/10 to 1/2 the size of female colonies (Coleman 2012, Fig 1K). As multicellularity is 38 hypothesized to drive the evolution of anisogamy (Parker et al. 1972; Bulmer and Parker 39 2002) and anisogamy drives female and male sex roles (Lehtonen et al. 2016b), we 40 expect sexual dimorphism and multicellular complexity to correlate in the volvocine 41 algae. 42 The volvocine algae have been previously used to test theories on the evolution of 43 anisogamy. Knowlton (1974) and Bell (1978) observed that larger species tended to be 44 anisogamous. Bell (1985) provided non-phylogenetically-controlled evidence that egg 45 size and gamete dimorphism are correlated with increasing body size. Later, Randerson 46 and Hurst (2001) analyzed the correlations between egg size and gamete dimorphism

3 85

1 with body size in a phylogenetic context. However, their results were not statistically 2 supported. Therefore, the volvocine green algae still provide a valuable system to test 3 previous predictions, including that (1) sexual traits (including a reduced number of 4 meiotic products hatching, internal fertilization, anisogamy, and sexual dimorphism) 5 correlate with multicellular complexity (Parker et al. 1972), (2) anisogamous species 6 produce larger zygotes than isogamous species (Bulmer and Parker 2002), and (3) 7 anisogamy and other traits evolved in a multicellular ancestor (Bulmer and Parker 2002). 8 In order to investigate how the evolution of multicellular complexity affects the 9 evolution of sexual traits, we used ancestral state reconstruction to infer the evolutionary 10 histories of six sexual traits in the volvocine algae: (1) a reduced number of meiotic 11 products, (2) internal fertilization, (3) anisogamy, (4) oogamy, (5) extra-fertile females, 12 and (6) dwarf males. We find multiple gains and/or losses in every trait studied, resulting 13 in both evolutionary convergence and substantial sexual diversity. We find that all sexual 14 traits evolved in a multicellular ancestor, and anisogamy evolved from an isogamous 15 ancestor. Anisogamy, internal fertilization, oogamy, and sexual dimorphism were each 16 significantly correlated with metrics of multicellular complexity. As predicted, 17 anisogamous species produce larger zygotes than isogamous species. Having identified 18 which traits are derived, we find that species with higher metrics of multicellular 19 complexity have significantly more derived sexual traits, suggesting that multicellular 20 complexity drives the evolution of sexual traits. 21 22 Material and Methods 23 Our phylogenetic analyses are based upon 97 volvocine terminal units (strains and 24 species), most of these species being well characterized phenotypically and genetically. 25 However, two taxa, G. pectorale Russia and Volvox perglobator Tucson, had not yet 26 been well characterized, so we first describe phenotypic and genotypic data collection 27 from these taxa. 28 29 CULTURE CONDITIONS 30 Strains (Gonium pectorale Russia and Volvox perglobator Tucson) were grown in 31 standard Volvox medium (SVM) at 25°C on a 16:8 hour light:dark cycle at approximately 32 35 µmol photons/m2/s. 33 34 DNA PREPARATION, PCR CLONING, AND SEQUENCING 35 Genomic DNA for chloroplast gene sequencing (G. pectorale Russia and V. perglobator 36 Tucson) was prepared according to Miller and Kirk (1999). PCR reactions were 37 performed using 2X Phusion HF Master Mix (Thermo Scientific, Waltham, MA). 38 Cycling conditions were 98°C for 2 minutes, followed by 37 cycles of 98°C/10 seconds, 39 variable annealing temperature (47.4–60.1°C) for 20 seconds, 72°C for variable extension 40 time (15-150 seconds), then a finishing step of 72°C for 5 minutes (Table S1). PCR 41 amplifications intended for sequencing were performed in at least three separate reactions 42 and then combined before sequencing to reduce the possibility of PCR errors. DNA 43 sequencing was performed by the University of Arizona Genetics Core using Applied 44 Biosystems 3730 DNA Analyzers (Waltham, MA). 45 46 TREE CONSTRUCTION

4 86

1 We estimated a new phylogeny in order to include novel strains and species (Nozaki et al. 2 2006, 2014, 2015a,b; Hayama et al. 2010; Nakada et al. 2010; Nozaki and Coleman 3 2011; Isaka et al. 2012) not included in previous analyses. A total of 19 of the 97 ingroup 4 terminal taxa in this study were not included in previous analyses. The ingroup was 5 defined as the smallest monophyletic clade containing Chlamydomonas reinhardtii and 6 Volvox carteri, a group commonly referred to as the “volvocine green algae”. We 7 generated a concatenated phylogeny (Fig. 2) using Bayesian Markov chain Monte Carlo 8 implemented in MrBayes version 3.2.2 (Ronquist et al. 2012) using default parameters 9 except as described below. The data matrix included sequences for 97 volvocine OTUs 10 and seven outgroup taxa. The outgroup consisted of non-volvocine algae, including two 11 taxa from the immediate sister group and one taxa from each of five other major groups 12 (Herron and Michod 2008). The sequence data consisted of five chloroplast genes (ATP 13 synthase beta-subunit, atpB; P700 chlorophyll a-apoprotein A1, psaA; P700 chlorophyll 14 a-apoprotein A2, psaB; photosystem II CP43 apoprotein, psbC; and the large subunit of 15 Rubisco, rbcL; Table S2). The best partitioning scheme and nucleotide substitution 16 models were determined using PartitionFinder version 2.1.1 (Lanfear et al. 2016) using 17 AICc and a greedy search algorithm with branch lengths linked. Eleven data blocks were 18 determined (3 codon positions for 5 protein-coding chloroplast genes, Table S3). Four 19 independent Bayesian runs of four chains each (three heated chains and one cold chain) 20 were run for 2×107 generations with a burn-in of 5×106 generations. Trees were sampled 21 every 100 generations. We considered the runs to have adequately sampled the solution 22 space when the standard deviation of split frequencies was below 5×10-3. Post burn-in 23 trees were combined and assembled to construct a majority-rule consensus phylogram. 24 Posterior probabilities for nodes were calculated using the pooled set of all post burn-in 25 trees. An ultrametric tree, necessary for maximum likelihood ancestral state 26 reconstruction using the diversitree R package (FitzJohn 2012), was calculated using a 27 penalized likelihood function in the ape package (Paradis et al. 2004). A correlated model 28 was used without age constraints. Branch lengths were qualitatively similar those 29 previously estimated (Herron et al. 2009). 30 Sequence data for all five chloroplast genes were not available for all taxa (16.5% 31 of cells in the data matrix are were missing). Previous analyses suggest that this low level 32 of missing data should be inconsequential (Wiens and Morrill 2011). Nevertheless, we 33 constructed an additional Bayesian concatenated tree including only taxa for which 34 sequence data for all five chloroplast genes are available (72 of the 97 ingroup OTUs 35 remained, Table S2). PartitionFinder was independently run on this dataset resulting in 36 the same best partition scheme (Table S3). This tree had no strongly supported 37 topological differences from the tree generated from the full dataset. 38 39 IDENTIFICATION OF CHARACTER STATES 40 Trait data for each species and strain were compiled from published reports 41 (Table S4), including sexual traits ((1) isogamy/anisogamy/oogamy/, (2) number of 42 meiotic products hatching (gone cells), (3) external/internal fertilization, (4) presence of 43 dwarf males, and (5) presence of extra-fertile females) and continuous metrics ((1) 44 maximum observed number of rounds of cell division (log2[cell number]), (2) maximum 45 observed percent of somatic cells, (3) maximum observed body length, (4) maximum 46 observed zygote diameter, (5) anisogamy ratio). Three metrics of multicellular

5 87

1 complexity were included to capture the variation in morphological complexity observed 2 in the volvocine algae through increased cell number, cellular differentiation, and 3 secretion of extracellular matrix. The maximum observed and reported values of these 4 metrics was used as many volvocine species can plastically vary the number of rounds of 5 cell division, thus also affecting body size, depending on environmental conditions 6 (Coleman 2012). We calculated the anisogamy ratio (a measure of gamete dimorphism), 7 defined by the macrogamete volume divided by microgamete volume, following 8 Randerson and Hurst (2001), assuming investment in a sperm packet and an egg is 9 approximately equal. The analysis found in Randerson and Hurst (2001) was repeated 10 since previous statistical significance depended on which of two tree topologies was 11 used. Furthermore, both trees were based on internal transcribed spacer (ITS) data and 12 were quite different from subsequent trees based on chloroplast DNA data (Nozaki et al. 13 2000; Herron and Michod 2008). 14 To avoid possible incorrect assignment of character state data due to erroneous 15 species identification or intraspecific variation, care was taken to ensure data were strain 16 specific, rather than species or genus specific. That is, character data were assigned to 17 individual strains based on published data for each strain, rather than based on 18 generalizations made about an entire species or genus. 19 For Gonium pectorale Russia, the maximum reproductive cell number was 20 measured using a Benchtop B3 Series FlowCam model VS-IV (Fluid Imaging 21 Technologies, Scarborough ME). A 20X objective was used, on AutoImage Mode, to 22 image over 500 colonies of exponential phase G. pectorale Russia. Gonium pectorale 23 Russia has a maximum of 16 cells, all of which are reproductive. The maximum colony 24 length, 85 µm, was evaluated based on 25 images using a Nikon Eclipse Ti-E (Nikon, 25 Tokyo, Japan). These values are consistent with other strains of G. pectorale (Table S4). 26 This strain does not show sexual activity in nitrogen-deficient medium (SVM with urea 27 omitted and 0.5 M CaCl2 replacing Ca(NO3)2), possibly caused by a decline in mating 28 efficiency due to long-term culture (Hamaji et al. 2013). 29 Cultures of Volvox perglobator Tucson were isolated from a water fountain in 30 Reid Park, Tucson, Arizona in October 2012 and September 2014 (GPS coordinates: 32° 31 12' 35.6" N, 110° 55' 21.2" W). This population was identified as Volvox perglobator 32 based on heterothallic mating and straight, blunt-tipped zygotes (Isaka et al. 2012). Single 33 colonies were isolated and grown until sexuality was naturally induced. Every isolate 34 produced either sperm or eggs, never both (heterothallic). Male and female strains were 35 observed using a Nikon SMZ800 stereomicroscope (Nikon, Tokyo, Japan) throughout 36 their lifecycle. When male and female strains were mixed, sex was oogamous with 37 internal fertilization, producing zygotes. The number of somatic cells present in a sample 38 of 20 asexual colonies was evaluated by counting the number of cells on the 39 circumference (n) and inferring the total number of somatic cells (N) by assuming cells 40 are hexagonally arranged (Smith 1944). Setting the surface area of a sphere equal to the 41 surface area of N hexagonal cells and using the equivalent formula for circumference to 42 solve for N (Janet 1912), ! = ! !!. We observed an average of 3,800 somatic cells and ! ! 43 a maximum of 7,500 somatic cells (cell numbers are rounded following convention and 44 reflecting the approximate nature of this approach). The maximum reproductive cell 45 number was measured using a Benchtop FlowCam model VS-IV (Fluid Imaging 46 Technologies, Scarborough ME). A 4X objective was used, on Trigger Mode, to image

6 88

1 250 colonies of exponential phase V. perglobator Tucson. Asexual colonies have 1–11 2 gonidia (reproductive cells/structures), with 4 gonidia on average. A maximum colony 3 length of 657 µm, measured using a Benchtop FlowCam (Fluid Imaging Technologies, 4 Scarborough, ME), was determined from a sample of 750 mature adult colonies. The 5 female strain spontaneously differentiated in older cultures with 18–79 eggs, with 53 6 eggs on average. 7 8 MEASURING PHYLOGENETIC SIGNAL 9 We first tested the level of phylogenetic signal in each trait to assess its lability across the 10 tree. For continuous traits, Blomberg’s K (2003) and Pagel’s λ (1999) were measured 11 using the phytools R package (Revell 2012; R Core Team 2013). For discrete traits, the D 12 value (Fritz and Purvis 2010) was measured using the caper R package (Orme et al. 13 2012). All traits had statistically significant phylogenetic signal (Table S5). 14 15 ANCESTRAL STATE RECONSTRUCTION 16 Ancestral states were reconstructed using maximum likelihood and Bayesian methods. 17 For the maximum likelihood analysis, several models of character evolution were 18 evaluated for each ancestral state reconstruction, including equal rates of change between 19 states (ER), symmetric rates of gain and loss between states (SYM, only relevant when 20 three states are considered), and all rates different between states (ARD). Models were 21 compared using the AICc (Akaike information criterion with a small sample size 22 correction, Akaike 1974; Burnham and Anderson 2002), which should identify the best- 23 fitting model without including unnecessary parameters (Table S5). When no model of 24 character evolution was strongly preferred (ΔAICc < 2), reconstructions were performed 25 using both models; in these cases, all reconstructions were robust to the model of 26 character evolution used. The root prior was weighted by the observed frequency of each 27 state (FitzJohn et al. 2009), in all cases using alternative root priors (flat prior, two given 28 priors) did not affect reconstructions. 29 The state with the highest probability was considered the most likely for a given 30 node. However, a state was considered to be significantly supported at a given node only 31 if it was at least 7.39 times (if the natural logarithm of the ratio of two likelihoods is 32 greater than 2) more likely than the alternative state (Pagel 1999). 33 Statistical support for estimated character states at internal nodes was further 34 evaluated using Bayesian hypothesis testing implemented in BayesTraits version 2 (Pagel 35 et al. 2004). Phylogenetic uncertainty was explicitly taken into account by analyzing a 36 sample of trees. Every 1,000th post-burnin tree from the four runs was included, for a 37 total of 600 trees. Outgroups and taxa without available character states (e.g., the number 38 of meiotic products hatching has never been observed in Basichlamys sacculifera) were 39 trimmed from these trees. A new ultrametric tree was calculated for each trimmed tree 40 using a penalized likelihood function with a correlated model without age constraints. As 41 above, three models of character evolution (ER, SYM when applicable, ARD) were 42 analyzed. A Bayes factor (BF) was estimated based on twice the difference between the 43 highest harmonic mean log likelihood from five independent Markov chain Monte Carlo 44 (MCMC) runs for each model (Table S5). The most supported model of evolution was 45 used for hypothesis tests of specific ancestral nodes of interest. When no model of 46 character evolution was strongly preferred (BF < 2), nodes of interest were also tested

7 89

1 under the alternative model. For specific nodes of interest, the ancestral character state 2 was tested by estimating a Bayes factor from five independent MCMC runs in which the 3 node in question was constrained to one state or the other. Uniform priors and gamma- 4 distributed hyperpriors seeded from a uniform distribution were used to seed all rate 5 parameters. The root state was not given but rather inferred by each MCMC run. All 6 MCMC runs included 5,500,000 generations with a burn-in period of 500,000 7 generations. 8 9 PHYLOGENETIC STATISTICAL TESTS 10 To test the hypothesis predicting a correlation between sexual characters and the size and 11 complexity of asexual colonies (Parker et al. 1972), we used phylogenetic t-tests, 12 implemented in the phytools R package version 0.5-64 (Revell 2012). For all 13 phylogenetic t-tests regarding multicellularity, three covarying metrics were used (log 14 body length, percent of cells that are somatic, and maximum rounds of cell divisions, Fig. 15 2). Pagel’s test (1999) was implemented in phytools in R (Revell 2012) to test for 16 correlations between discrete traits. The likelihood ratio test was used to compare 17 alternative maximum likelihood models, i.e., when testing for evolutionary irreversibility. 18 To test whether larger species had more derived sex characters, we performed a 19 phylogenetic generalized least squares (PGLS) analysis. In these analyses, we regressed 20 the number of derived sex traits in each species with our three metrics of multicellular 21 complexity. PGLS was implemented using the caper R package (Orme et al. 2012). Only 22 species with available data for all six characters ((1) reduced meiotic products, (2) 23 internal fertilization, (3) anisogamy, (4) oogamy, (5) extra fertile females, (6) dwarf 24 males) were included (n = 65 species). We repeated with this analysis including sperm 25 packets as a seventh trait, which did not affect statistical results (phylogenetic generalized 26 least squares, adjusted p < 3.69×10-7). 27 28 MULTIPLE COMPARISON ADJUSTMENTS 29 When multiple p-values were inferred in testing whether a discrete sexual trait correlated 30 with three metrics of multicellularity, p-values were adjusted for multiple comparisons 31 following Benjamini and Hochberg (1995). 32 33 Results 34 PHYLOGENETIC TREE 35 Given the lack of a phylogenetic tree which includes recently described species (Nozaki 36 et al. 2006, 2014; Isaka et al. 2012), a Bayesian tree was constructed using five 37 chloroplast genes (Table S2). This tree includes the novel taxa G. pectorale Russia and 38 rediscovered V. perglobator Tucson. The resulting tree (Fig. 2) is concordant with 39 recently published phylogenetic analyses (Nozaki et al. 2006, 2014; Herron and Michod 40 2008); any differences between our topology and other published tree topologies had low 41 support. Volvox perglobator Tucson is nested within Volvox sect. Volvox (a.k.a. 42 Euvolvox), whereas G. pectorale Russia is nested within G. pectorale (Fig. 2). 43 44 EVOLUTION OF VOLVOCINE SEXUAL TRAITS 45 To understand the evolutionary history of the sexual traits studied, ancestral state 46 reconstructions were performed using likelihood and Bayesian analyses. We found a

8 90

1 complex evolutionary history for all sexual traits, including multiple gains and losses 2 (Fig. 3). Based on phylogenetic distribution (Fig. 3), the first sexual trait to evolve was a 3 reduction in the number of meiotic products hatching from a desiccation-resistant diploid 4 spore (Fig. 1G). This reduction in surviving meiotic products, along with the evolution of 5 polar bodies, evolved twice: once in the ancestor of the genus Astrephomene and again in 6 the family Volvocaceae (Fig. 3A). A relationship between a reduced number of surviving 7 meiotic products and three metrics of multicellularity is not significantly supported 8 (phylogenetic t-test, adjusted p = 0.216, Table 1). However, a four-fold reduction in 9 surviving meiotic products likely has important evolutionary consequences through a 10 reduction in both the number and genetic variation of sexual offspring. The repeated 11 origins of polar bodies in the volvocine algae parallels the independent evolution of polar 12 bodies in plants and animals (Schmerler and Wessel 2011). The function of polar bodies 13 in the volvocine algae is unknown; they may facilitate increased cytoplasmic 14 provisioning to progeny, similar to the endosperm in angiosperms (Schmerler and Wessel 15 2011). 16 In the volvocine algae, anisogamy evolved from isogamy twice independently 17 (Fig. 1H, 3B). Therefore, the volvocine algae provide support for the claim that 18 anisogamous species are “almost certainly derived from isogamous ancestors” (Lehtonen 19 et al. 2016a). The evolution of oogamy occurred at least three times, likely with one 20 reversion to non-oogamous anisogamy (Fig. 3B). There are no inferred reversions from 21 anisogamy to isogamy (Fig. 3B). To test for evolutionary reversal of anisogamy, two 22 models were compared, one constraining the rate of loss of anisogamy to zero, the other 23 constraining the rate of loss to be equal to the rate of gain. Using the likelihood ratio test, 24 the model preventing reversals in anisogamy is not significantly supported (χ2 = 1.65, p = 25 0.199). 26 Did anisogamy evolve in a unicellular or multicellular ancestor? This is not a 27 foregone conclusion, since several unicellular species formerly classified as 28 Chlamydomonas (now Oogamochlamys and Lobochlamys) are anisogamous (Ettl 1983; 29 Pröschold et al. 2001). Herron and Michod (2008) previously reconstructed 12 asexual 30 morphological traits underlying complexity in the volvocine algae (Kirk 2005). By 31 comparing these reconstructions (Fig. 3B, Herron and Michod (2008)), 7 of these 12 32 traits were already present in the ancestor that evolved anisogamy; therefore, anisogamy 33 evolved in a multicellular volvocine ancestor. 34 Consistent with previous theory (Parker et al. 1972; Bulmer and Parker 2002) and 35 previous analyses (Knowlton 1974; Bell 1985; Randerson and Hurst 2001), anisogamous 36 algae are significantly larger than isogamous algae (without treating oogamy as a separate 37 character, phylogenetic t-test, adjusted p < 0.048, Table 1). Additionally, zygotes of 38 anisogamous species are significantly larger than zygotes of isogamous species 39 (phylogenetic t-test, p = 0.002, Table 1). Lastly, the anisogamy ratio (a measure of 40 gamete dimorphism) significantly increases with multicellular complexity (phylogenetic 41 generalized least squares, adjusted p < 3.3×10-5, Fig. S1). 42 Internal fertilization (Fig. 1I) differs from gametes simply finding and fertilizing 43 each other (external fertilization). Instead, it requires a multicellular sperm packet to 44 penetrate the multicellular female colony wall and dissociate into individual sperm. 45 Individual sperm must then swim through the extracellular matrix to find and fertilize a 46 unicellular egg. Internal fertilization evolved twice (Fig. 3C), broadly corresponding to

9 91

1 the evolution of anisogamy (though there are anisogamous species with external 2 fertilization (Nozaki et al. 2014)). Internal fertilization significantly correlates with all 3 three metrics of multicellular complexity (phylogenetic t-test, adjusted p < 0.034, Table 4 1) and correlates with both anisogamy (Pagel’s test, adjusted p = 0.026) and oogamy 5 (Pagel’s test, adjusted p = 0.026, Table S6) as may be expected given that internal 6 fertilization maintains gamete fusion with the evolution of large, less motile gametes 7 (anisogamy and oogamy). 8 Secondary sexual dimorphism in the volvocine algae includes extra-fertile 9 females, multicellular sperm packets, and dwarf males (Fig. 1J,K). Extra-fertile females, 10 which can increase sexual reproductive output up to 15 times that of asexual reproduction 11 (Coleman 2012), have evolved three times independently (Fig. 3D) and were lost once 12 (V. tertius). Since all known anisogamous species of volvocine algae produce 13 multicellular sperm packets, the inferred evolutionary history of this trait is identical to 14 that of anisogamy, i.e. it has evolved twice independently. Dwarf males have evolved 15 once and were subsequently lost in V. dissipatrix (Fig. 3E). The presence of extra-fertile 16 females and dwarf males (both of which only occur in the genus Volvox (Nozaki 2003)) 17 are each significantly correlated with all three metrics of multicellularity (phylogenetic t- 18 test, adjusted p < 0.026, Table 1). As with anisogamy itself, species that produce 19 multicellular sperm packets are significantly larger than those that do not (phylogenetic t- 20 test, adjusted p < 0.048). These results are consistent with the hypothesis that the 21 morphological complexity present in Volvox species allows for anisogamy to drive the 22 secondary evolution of exaggerated sexual dimorphism. 23 24 MULTICELLULARITY AND SEXUAL TRAITS 25 By identifying which states are derived (Fig. 3), we were able to test whether sexual traits 26 evolve with the evolution of multicellular complexity. We performed phylogenetic 27 regressions of the number of derived sexual traits present in each species and three 28 metrics of multicellularity (Fig. 4). Species with higher metrics of multicellular 29 complexity have significantly more derived sexual characters (phylogenetic generalized 30 least squares, adjusted p < 2.11×10-7, Fig. 4). Combined with the observation that all six 31 sexual traits evolved in a multicellular ancestor (Fig. 2-3), this suggests multicellularity 32 drove the evolution of sexual traits. 33 Conversely, does the evolution of sex reciprocally drive the evolution of 34 multicellular complexity as previously predicted (Woodland 2016)? The evolution of 35 anisogamy is thought to subsequently require increased complexity (cellular 36 differentiation), to support increasing egg size. This is predicted to lead to a correlation 37 between anisogamy and cellular differentiation, with anisogamy evolving before cellular 38 differentiation (Woodland 2016). However, anisogamy and cellular differentiation are not 39 significantly correlated in the volvocine algae (Pagel’s test, likelihood-ratio = 7.14, p = 40 0.129). Furthermore, cellular differentiation has evolved in isogamous Astrephomene 41 (Brooks 1966; Nozaki 1983), suggesting the evolution of anisogamy does not correlate 42 with the evolution of cellular differentiation as predicted. The evolution of 43 multicellularity appears to drive the evolution of sexual traits, but the opposite 44 relationship does not seem to appear in the volvocine algae. 45 46 Discussion

10 92

1 Using ancestral state reconstruction and phylogenetic analysis, we have shown that sex 2 and multicellularity are fundamentally linked in the volvocine green algae. Volvocine 3 sexual traits have a complex evolutionary history, with multiple gains and/or losses in 4 every trait studied, resulting in both evolutionary convergence and substantial sexual 5 diversity among species. Using evolutionary repeatability or convergence as an indicator 6 of adaptive value (Simpson 1953; Mayr 1966; Bonner 2000; Schluter 2000), the multiple 7 gains of the sexual traits studied here suggest that they have an inherent adaptive value, 8 despite their apparent high costs (the four-fold reduction in meiotic products, the two-fold 9 cost of anisogamy). 10 Previous theory has predicted that multicellularity drives the evolution of sexual 11 traits, particularly anisogamy (Parker et al. 1972; Bulmer and Parker 2002). Consistent 12 with these predictions, we have provided phylogenetically controlled and statistically 13 robust support for the predictions that anisogamy evolved from isogamy (Fig. 3B), that 14 anisogamy evolved in a multicellular ancestor (Figs. 2-3), that anisogamous species are 15 larger than isogamous species (Table 1), and that anisogamous species produce larger 16 zygotes than isogamous species (Table 1). While a reduced number of meiotic products 17 hatching did not significantly correlate with multicellular complexity (Table 1), this trait 18 may interact with the effect of larger zygotes in anisogamous species, further increasing 19 provisioning per sexual offspring as zygote size increases (Fig. 3A,B). Taken together, 20 these results are strongly consistent with previous theoretical predictions that 21 multicellularity drives the evolution of costly anisogamy (Bulmer and Parker 2002). 22 The ‘anisogamy gateway’ hypothesis attempts to explain the prevalence of 23 anisogamy despite its apparent costs by proposing that anisogamy acts as a ‘gateway’ for 24 the diversity of sexual forms we observe today. Under this hypothesis, anisogamy 25 evolves in lineages where previously evolved constraints make asexual invasion unlikely 26 (Lehtonen et al. 2016a). Briefly, invasion of an isogamous population by obligately 27 asexual individuals is prevented by previously evolved constraints (Lehtonen et al. 28 2016a). Then, a need to produce larger bodied offspring drives the evolution of 29 anisogamy (Parker et al. 1972; Bulmer and Parker 2002). Isogamous individuals do not 30 subsequently invade because sperm competition maintains anisogamy (Parker 1982). In 31 the volvocine algae, these previously evolved constraints may include ecological 32 variation, where only sexually produced diploid zygospores are resistant to detrimental 33 environmental change such as heat and desiccation. Thus, obligately asexual individuals 34 are unlikely to invade a volvocine population. As predicted, we find that anisogamy is 35 strongly correlated with increased body size, increased zygote size, and multicellularity, 36 consistent with the anisogamy gateway hypothesis. There were no observed reversions 37 from anisogamy to isogamy, although there was not strong statistical support for the 38 irreversibility of anisogamy. However, the anisogamy gateway hypothesis predicts that 39 anisogamy evolves in obligately sexual species with subsequently derived facultative 40 sexual reproduction (Lehtonen et al. 2016a). The volvocine algae disprove this 41 prediction; all known volvocine species are facultatively sexual, including the ancestor in 42 which anisogamy evolved. This apparent contradiction may be explained by the temporal 43 nature of the evolved constraints preventing reversion to asexuality in the volvocine 44 green algae. Periods of heat or desiccation prevent invasion of obligate asexuality, but 45 otherwise, facultatively asexual reproduction may be most advantageous. Thus, obligate 46 sexuality may not be a necessary predecessor of anisogamy, as long as obligately asexual

11 93

1 individuals cannot invade sexual populations. 2 Ecology might also play an important role in the evolution of anisogamy (Madsen 3 and Waller 1983). Under the “fast start” hypothesis, species living in stressful 4 environments such as ephemeral pools exhibit anisogamy or oogamy because the larger 5 zygospore allows for more rapid growth when necessary. Species living in more stable 6 environments, such as soil, do not require this fast start (Madsen and Waller 1983). 7 Unfortunately, insufficient information about volvocine habitat is available to further 8 evaluate this hypothesis. Nevertheless, some observations are consistent with this 9 hypothesis. For example, isogamous Chlamydomonas is a soil alga and larger 10 anisogamous/oogamous genera such as Pleodorina and Volvox do live in ephemeral 11 pools and lakes (Kirk 1998; Harris 2009). 12 13 Conclusions 14 Using ancestral state reconstruction and phylogenetic regression, we have shown how the 15 evolutionary history of six derived sexual characters has resulted in not only convergent 16 evolution but also a stable diversity in forms of sexual reproduction. Taken together, 17 these sexual traits demonstrate the relationship between the evolution of sexual traits and 18 the evolution of multicellularity. These sexual traits are shared amongst other 19 multicellular taxa, highlighting the evolutionary convergence amongst multicellular taxa, 20 and suggesting the evolution of traits related to multicellularity and sexuality in the 21 volvocine algae may offer a window into their origins and evolution across the Tree of 22 Life. 23 24 LITERATURE CITED 25 Akaike, H. 1974. A new look at the statistical model identification. IEEE Trans. 26 Automat. Contr. 19:716–723. 27 Bateman, A. J. 1948. Intra-sexual selection in Drosophila. Heredity (Edinb). 2:349–368. 28 Bell, G. 1978. The evolution of anisogamy. J. Theor. Biol. 73:247–270. 29 Bell, G. 1982. The materpiece of nature: the evolution and genetics of sexuality. Croom 30 Helm, London. 31 Bell, G. 1985. The origin and evolution of germ cells as illustrated by the Volvocales. Pp. 32 221–256 in H. O. Halvorson and A. Monroy, eds. The origin and evolution of sex. 33 Alan R. Liss, New York. 34 Benjamini, Y., and Y. Hochberg. 1995. Controlling the false discovery rate: a practical 35 and powerful approach to multiple testing. J. R. Stat. Soc. B 57:289–300. 36 Beukeboom, L., and N. Perrin. 2014. The evolution of sex determination. Oxford 37 University Press, Oxford. 38 Blomberg, S. P., T. Garland, and A. R. Ives. 2003. Testing for phylogenetic signal in 39 comparative data: behavioral traits are more labile. Evolution 57:717–745. 40 Bonner, J. T. 2000. First Signals: The Evolution of Multicellular Development. Princeton 41 University Press, Princeton, New Jersey. 42 Brooks, A. 1966. The sexual cycle and intercrossing in the genus Astrephomene. J. 43 Eukaryot. Microbiol. 13:367–375. 44 Bulmer, M. G., and G. A. Parker. 2002. The evolution of anisogamy: a game-theoretic 45 approach. Proc. R. Soc. B Biol. Sci. 269:2381–2388. 46 Burnham, K. P., and D. R. Anderson. 2002. Model selection and multi-model inference: a

12 94

1 practical information-theoretic approach. Springer-Verlag, New York. 2 Clarke, E. 2013. The multiple realizability of biological individuals. J. Philos. 110:413– 3 435. 4 Coleman, A. W. 2012. A comparative analysis of the Volvocaceae (Chlorophyta). J. 5 Phycol. 48:491–513. 6 Ettl, H. 1983. Chlorophyta I - Phytomonidia. Pp. 1–807 in H. Ettl, J. Gerloff, H. Heynig, 7 and D. Mollenhauer, eds. Süsswasserflora von Mitteleuropa. Gustav Fischer, 8 Stuttgart. 9 FitzJohn, R. G. 2012. Diversitree: Comparative phylogenetic analyses of diversification 10 in R. Methods Ecol. Evol. 3:1084–1092. 11 FitzJohn, R. G., W. P. Maddison, and S. P. Otto. 2009. Estimating trait-dependent 12 speciation and extinction rates from incompletely resolved phylogenies. Syst. 13 Biol. 58:595–611. 14 Fritz, S. A., and A. Purvis. 2010. Selectivity in mammalian extinction risk and threat 15 types: A new measure of phylogenetic signal strength in binary traits. Conserv. 16 Biol. 24:1042–1051. 17 Hamaji, T., P. J. Ferris, I. Nishii, Y. Nishimura, and H. Nozaki. 2013. Distribution of the 18 sex-determining gene MID and molecular correspondence of mating types within 19 the isogamous genus Gonium (Volvocales, Chlorophyta). PLoS One 8:e64385. 20 Hanschen, E. R., D. R. Davison, Z. I. Grochau-Wright, and R. E. Michod. 2017. 21 Evolution of individuality: a case study in the volvocine green algae. Philos. 22 Theory, Pract. Biol. 9:3. 23 Hanschen, E. R., T. N. Marriage, P. J. Ferris, T. Hamaji, A. Toyoda, A. Fujiyama, R. 24 Neme, H. Noguchi, Y. Minakuchi, M. Suzuki, H. Kawai-Toyooka, D. R. Smith, 25 V. Luria, A. Karger, M. W. Kirschner, H. Sparks, J. Anderson, R. Bakaric, P. M. 26 Durand, R. E. Michod, H. Nozaki, and B. J. S. C. Olson. 2016. The Gonium 27 pectorale genome demostrates co-option of cell cycle regulation during the 28 evolution of multicellularity. Nat. Commun. 7:11370. 29 Harris, E. H. 2009. The Chlamydomonas Sourcebook (Volume 1). Second. Acad, San 30 Diego, California. 31 Hayama, M., T. Nakada, T. Hamaji, and H. Nozaki. 2010. Morphology, molecular 32 phylogeny and taxonomy of Gonium maiaprilis sp. nov. (Goniaceae, 33 Chlorophyta) from Japan. Phycologia 49:221–234. 34 Herron, M. D., J. D. Hackett, F. O. Aylward, and R. E. Michod. 2009. Triassic origin and 35 early radiation of multicellular volvocine algae. Proc. Natl. Acad. Sci. USA 36 106:3254–3258. 37 Herron, M. D., and R. E. Michod. 2008. Evolution of complexity in the volvocine algae: 38 transitions in individuality through Darwin’s eye. Evolution 62:436–451. 39 Isaka, N., H. Kawai-Toyooka, R. Matsuzaki, T. Nakada, and H. Nozaki. 2012. 40 Description of two new monoecious species of Volvox sect. Volvox 41 (Volvocaceae, Chlorophyceae), based on comparative morphology and molecular 42 phylogeny of cultured material. J. Phycol. 48:759–767. 43 Janet, C. 1912. Le Volvox. Imprimerie-librairie Ducourtieux et Gout, Limoges, France. 44 Kass, R. E., and A. E. Raftery. 1995. Bayes Factors. J. Am. Stat. Assoc. 90:773–795. 45 Kirk, D. L. 2005. A twelve-step program for evolving multicellularity and a division of 46 labor. BioEssays 27:299–310.

13 95

1 Kirk, D. L. 1998. Volvox: Molecular-genetic origins of multicellularity and cellular 2 differentiation. Cambridge University Press, Cambridge. 3 Knowlton, N. 1974. A note of the evolution of gamete dimorphism. J. Theor. Biol. 4 45:283–285. 5 Lanfear, R., P. B. Frandsen, A. M. Wright, T. Senfeld, and B. Calcott. 2016. 6 PartitionFinder 2: New methods for selecting partitioned models of evolution for 7 molecular and morphological phylogenetic analyses. Mol. Biol. Evol. 8 34:msw260. 9 Lehtonen, J., H. Kokko, and G. A. Parker. 2016a. What do isogamous organisms teach us 10 about sex and the two sexes? Philos. Trans. R. Soc. B Biol. Sci. 371:20150532. 11 Lehtonen, J., G. A. Parker, and L. Schärer. 2016b. Why anisogamy drives ancestral sex 12 roles. Evolution 70:1129–1135. 13 Madsen, J. D., and D. M. Waller. 1983. A note on the evolution of gamete dimorphism in 14 algae. Am. Nat. 121:443–447. 15 Maynard Smith, J. 1982. Evolution and the theory of games. Cambridge University Press, 16 Cambridge, UK. 17 Maynard Smith, J. 1978. The evolution of sex. Cambridge University Press, Cambridge. 18 Mayr, E. 1966. Animal Species and Evolution. Harvard University Press, Cambridge, 19 Massachusetts. 20 Merchant, S. S., S. E. Prochnik, O. Vallon, E. H. Harris, S. J. Karpowicz, G. B. Witman, 21 A. Terry, A. Salamov, L. K. Fritz-Laylin, L. Maréchal-Drouard, W. F. Marshall, 22 L.-H. Qu, D. R. Nelson, A. a Sanderfoot, M. H. Spalding, V. V Kapitonov, Q. 23 Ren, P. Ferris, E. Lindquist, H. Shapiro, S. M. Lucas, J. Grimwood, J. Schmutz, 24 P. Cardol, H. Cerutti, G. Chanfreau, C.-L. Chen, V. Cognat, M. T. Croft, R. Dent, 25 S. Dutcher, E. Fernández, H. Fukuzawa, D. González-Ballester, D. González- 26 Halphen, A. Hallmann, M. Hanikenne, M. Hippler, W. Inwood, K. Jabbari, M. 27 Kalanon, R. Kuras, P. A. Lefebvre, S. D. Lemaire, A. V Lobanov, M. Lohr, A. 28 Manuell, I. Meier, L. Mets, M. Mittag, T. Mittelmeier, J. V Moroney, J. Moseley, 29 C. Napoli, A. M. Nedelcu, K. Niyogi, S. V Novoselov, I. T. Paulsen, G. Pazour, 30 S. Purton, J.-P. Ral, D. M. Riaño-Pachón, W. Riekhof, L. Rymarquis, M. Schroda, 31 D. Stern, J. Umen, R. Willows, N. Wilson, S. L. Zimmer, J. Allmer, J. Balk, K. 32 Bisova, C.-J. Chen, M. Elias, K. Gendler, C. Hauser, M. R. Lamb, H. Ledford, J. 33 C. Long, J. Minagawa, M. D. Page, J. Pan, W. Pootakham, S. Roje, A. Rose, E. 34 Stahlberg, A. M. Terauchi, P. Yang, S. Ball, C. Bowler, C. L. Dieckmann, V. N. 35 Gladyshev, P. Green, R. Jorgensen, S. Mayfield, B. Mueller-Roeber, S. Rajamani, 36 R. T. Sayre, P. Brokstein, I. Dubchak, D. Goodstein, L. Hornick, Y. W. Huang, J. 37 Jhaveri, Y. Luo, D. Martínez, W. C. A. Ngau, B. Otillar, A. Poliakov, A. Porter, 38 L. Szajkowski, G. Werner, K. Zhou, I. V Grigoriev, D. S. Rokhsar, and A. R. 39 Grossman. 2007. The Chlamydomonas genome reveals the evolution of key 40 animal and plant functions. Science 318:245–50. 41 Miller, S. M., and D. L. Kirk. 1999. glsA, a Volvox gene required for asymmetric division 42 and germ cell specification, encodes a chaperone-like protein. Development 43 126:649–58. 44 Nakada, T., M. Tomita, and H. Nozaki. 2010. Volvulina compacta (Volvocaceae, 45 Chlorophyeae), new to Japan, and its phylogenetic position. J. Japanese Bot. 46 85:364–369.

14 96

1 Nozaki, H. 2003. Flagellated green algae. Pp. 225–252 in J. D. Wehr and R. G. Sheath, 2 eds. Freshwater Algae of North America. Academic Press, San Diego, California. 3 Nozaki, H. 1983. Morphology and taxonomy of two species of Astrephomene 4 (Chlorophyta) in Japan. Journ. Jap. Bot. 58:345–352. 5 Nozaki, H., and A. W. Coleman. 2011. A new species of Volvox sect. Merrillosphaera 6 (Volvocaceae, Chlorophyceae) from Texas. J. Phycol. 47:673–679. 7 Nozaki, H., and M. Itoh. 1994. Phylogenetic relationships within the colonial Volvocales 8 (Chlorophyta) inferred from cladistic analysis based on morphological data. J. 9 Phycol. 30:353–365. 10 Nozaki, H., R. Matsuzaki, K. Yamamoto, M. Kawachi, and F. Takahashi. 2015a. 11 Delineating a new heterothallic species of Volvox (Volvocaceae, Chlorophyceae) 12 using new strains of “Volvox africanus.” PLoS One 10:1–16. 13 Nozaki, H., K. Misawa, T. Kajita, M. Kato, S. Nohara, and M. M. Watanabe. 2000. 14 Origin and evolution of the colonial volvocales (Chlorophyceae) as inferred from 15 multiple, chloroplast gene sequences. Mol. Phylogenet. Evol. 17:256–268. 16 Nozaki, H., F. D. Ott, and A. W. Coleman. 2006. Morphology, molecular phylogeny and 17 taxonomy of two new species of Pleodorina (Volvoceae, Chlorophyceae). J. 18 Phycol. 42:1072–1080. 19 Nozaki, H., N. Ueki, O. Misumi, K. Yamamoto, S. Yamashita, M. D. Herron, and F. 20 Rosenzweig. 2015b. Morphology and reproduction of Volvox capensis 21 (Volvocales, Chlorophyceae) from Montana, USA. Phycologia 54:316–320. 22 Nozaki, H., T. K. Yamada, F. Takahashi, R. Matsuzaki, and T. Nakada. 2014. New 23 “missing link” genus of the colonial volvocine green algae gives insights into the 24 evolution of oogamy. BMC Evol. Biol. 14:37. BMC Evolutionary Biology. 25 Orme, C. D. L., R. P. Freckleton, G. H. Thomas, T. Petzoldt, S. A. Fritz, N. J. B. Isaac, 26 and W. Pearse. 2012. The caper package: comparative analysis of phylogenetics 27 and evolution in R. 28 Pagel, M. 1999. The maximum likelihood approach to reconstructing ancestral character 29 states of discrete characters on phylogenies. Syst. Biol. 48:612–622. 30 Pagel, M., A. Meade, and D. Barker. 2004. Bayesian estimation of ancestral character 31 states on phylogenies. Syst. Biol. 53:673–684. 32 Paradis, E., J. Claude, and K. Strimmer. 2004. APE: Analyses of phylogenetics and 33 evolution in R language. Bioinformatics 20:289–290. 34 Parker, G. A. 1982. Why are there so many tiny sperm? Sperm competition and the 35 maintenance of two sexes. J. Theor. Biol. 96:281–294. 36 Parker, G. A., R. R. Baker, and V. G. F. Smith. 1972. The origin and evolution of gamete 37 dimorphism and the male-female phenomenon. J. Theor. Biol. 36:529–553. 38 Prochnik, S. E., J. Umen, A. M. Nedelcu, A. Hallmann, S. M. Miller, I. Nishii, P. J. 39 Ferris, A. Kuo, T. Mitros, L. K. Fritz-Laylin, U. Hellsten, J. Chapman, O. 40 Simakov, S. A. Rensing, A. Terry, J. Pangilinan, V. Kapitonov, J. Jurka, A. 41 Salamov, H. Shapiro, J. Schmutz, J. Grimwood, E. Lindquist, S. Lucas, I. V. 42 Grigoriev, R. Schmitt, D. L. Kirk, and D. S. Rokhsar. 2010. Genomic analysis of 43 organismal complexity in the multicellular green alga Volvox carteri. Science 44 329:223–6. 45 Pröschold, T., B. Marin, U. G. Schlösser, and M. Melkonian. 2001. Molecular 46 phylogency and taxonomic revision of Chlamydomonas (Chlorophyta). I.

15 97

1 Emendation of Chlamydomonas ehrenberg and Chloromonas gobi, and 2 description of Oogamochlamys gen. nov. and Lobochlamys gen. nov. Protist 3 152:265–300. 4 R Core Team. 2013. R: A language and environment for statistical computing. Vienna. 5 Randerson, J. P., and L. D. Hurst. 2001. A comparative test of a theory for the evolution 6 of anisogamy. Proc. R. Soc. B Biol. Sci. 268:879–884. 7 Revell, L. J. 2012. phytools: An R package for phylogenetic comparative biology (and 8 other things). Methods Ecol. Evol. 3:217–223. 9 Ronquist, F., M. Teslenko, P. van der Mark, D. L. Ayres, A. Darling, S. Hohna, B. 10 Larget, L. Liu, M. A. Suchard, and J. P. Huelsenbeck. 2012. MrBayes 3.2: 11 Efficient Bayesian phylogenetic inference and model choice across a large model 12 space. Syst. Biol. 61:539–542. 13 Santelices, B. 1999. How many kinds of individual are there? Trends Ecol. Evol. 14:152– 14 155. 15 Schluter, D. 2000. The Ecology of Adaptive Radiation. Oxford University Press, Oxford. 16 Schmerler, S., and G. Wessel. 2011. Polar Bodies- more a lack of understanding than a 17 lack of respect. Mol. Reprod. Dev. 78:3–8. 18 Simpson, G. G. 1953. The Major Features of Evolution. Columbia University Press, New 19 York. 20 Smith, G. M. 1944. A comparative study of the species of Volvox. Trans. Am. Microsc. 21 Soc. 63:265–310. 22 Togashi, T., and P. A. Cox. 2011. The evolution of anisogamy. Cambridge University 23 Press, Cambridge, UK. 24 Wiens, J. J., and M. C. Morrill. 2011. Missing data in phylogenetic analysis: Reconciling 25 results from simulations and empirical data. Syst. Biol. 60:719–731. 26 Woodland, H. R. 2016. The Birth of Animal Development: Multicellularity and the 27 Germline. Pp. 609–630 in P. M. Wassarman, ed. Current Topics in 28 Developmental Biology. Academic Press, Burlington. 29 30

16 98

1 Figure Legends 2 Figure 1. Exemplar species and diversity of sexual systems in the volvocine algae. A. 3 unicellular Chlamydomonas reinhardtii, B. undifferentiated Gonium pectorale Russia, C. 4 undifferentiated Pandorina morum, D. undifferentiated Eudorina elegans, E. soma 5 differentiated Pleodorina californica, F. germ-soma differentiated Volvox carteri f. 6 nagariensis. G. A desiccation-resistant diploid zygospore produces four meiotic products 7 (left) or a reduced number of meiotic products (right). H. Isogamy (equal-sized gametes), 8 anisogamy (unequal sized gametes, with smaller sperm produced by males and larger, 9 flagellated eggs produced by females), and oogamy (unequal sized gametes, unflagellated 10 eggs are much larger than sperm). I. Fertilization external to a multicellular organism 11 versus internal fertilization. J. Female colonies versus extra-fertile female colonies, which 12 have at least a two-fold increase in reproductive cell number. K. Male colonies with 13 sperm packets versus dwarf male colonies with sperm packets. Cartoons in panels I-K are 14 shown with Volvox-like morphology for illustrative purposes only.

15

17 99

1 Figure 2. Phylogenetic tree of the volvocine green algae. Numbers indicate Bayesian 2 posterior probabilities (PP). Unlabeled nodes are supported with PP=1.00. Outgroup taxa 3 have been trimmed. Dots show relative metrics of multicellular complexity (body size; 4 number of rounds of cell division, percent somatic cells, and body length). A key 5 showing minimum and maximum values of metrics is included. The scale bar (for branch 6 lengths) shows expected substitutions per site.

7

18 100

1 Figure 3. Ancestral state reconstructions of six sexual characters. A. The evolution of all 2 (green) and reduced number of (black) meiotic products germinating from a diploid 3 zygospore. B. The evolution of isogamy (green), anisogamy (blue), and oogamy (black). 4 C. The evolution of external (green) and internal (black) fertilization. D. The evolution of 5 normal females (green) and extra-fertile females (black). E. The evolution of normal 6 males (green) and dwarf males (black). For all panels branch color refers to the most 7 likely state inferred by maximum likelihood reconstruction (ML). Pie charts at nodes 8 represent scaled marginal likelihoods from ML reconstruction. Dashed branches indicate 9 a statistically ambiguous reconstruction. Numbers at select nodes (relevant to the trait in 10 question) indicate Bayes factors (support for that character state against the next most 11 likely state), colored by which state is most strongly supported. Interpretation of Bayes 12 factors (Kass and Raftery 1995): 0 to 2 barely worth mentioning, 2 to 6 positive, 6 to 10 13 strong, >10 very strong.

14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

19 101

1 Figure 4. Relationships between the number of derived sexual traits in each species (not 2 including multicellular sperm packets, which are perfectly correlated with anisogamy) 3 and three traits related to multicellular complexity. A. The number of rounds of cell 4 division in that species; B. the percent of somatic cells; and C. the natural logarithm- 5 transformed body length. Relative sizes of the data points (tips of the phylogeny) are 6 scaled to indicate the number of species at that coordinate (from 1 to 15). Phylogenetic 7 regression line and statistics (red) and simple linear regression and statistics (blue) are 8 shown.

9

20 102

1 Table 1. Phylogenetic t-test for sexual traits, using three continuous metrics of 2 multicellular complexity (rounds of cell division, percent somatic cells, natural 3 logarithm-transformed body length). A phylogenetic t-test comparing natural logarithm- 4 transformed zygote diameter of isogamous and anisogamous algae is included. When 5 multiple metrics were used, p-values were adjusted for multiple comparisons following 6 Benjamini and Hochberg (1995). 7 Phylogenetic, Sample, p- adj.,p- t-test, size, Continuous,variable, t, value, value, All#vs.# Rounds#of#cell# 6.01# 0.234# 0.306# reduced# division# n=65# meiotic# Percent#somatic#cells# 4.51# 0.306# 0.306# products# Ln#body#length# 13.31# 0.072# 0.216# Rounds#of#cell# 50.85# 0.045# 0.048# Isogamy#vs.# division# n=86# anisogamy# Percent#somatic#cells# 49.32# 0.048# 0.048# Ln#body#length# 99.74# 0.006# 0.018# Isogamy#vs.# n=72# Ln#zygote#diameter# 88.81# 0.002# H# anisogamy# Rounds#of#cell# External#vs.# 65.90# 0.029# 0.033# division# internal# n=86# Percent#somatic#cells# 63.67# 0.033# 0.033# fertilization# Ln#body#length# 117.30# 0.009# 0.033# Rounds#of#cell# Females#vs.# 235.14# 0.001# 0.001# division# extraHfertile# n=85# Percent#somatic#cells# 161.66# 0.001# 0.001# females# Ln#body#length# 86.61# 0.001# 0.001# Rounds#of#cell# 24.11# 0.006# 0.009# Males#vs.# division# n=84# dwarf#males# Percent#somatic#cells# 33.63# 0.001# 0.003# Ln#body#length# 14.93# 0.026# 0.026# 8 9

21