Alternative breeding times and evolutionary pathways in corals in north Western Australia

Natalie L. Rosser BSc. (Hons)

This thesis is presented for the degree of Doctor of Philosophy of the University of Western Australia School of Biology 2016

Abstract

This thesis explores the influence of asynchronous reproductive timing on patterns of genetic structure in coral populations, at both ecological and evolutionary scales, in the context of understanding the evolution of seasonal breeding patterns in Western

Australia. Most broadcast-spawning corals reproduce once a year, often in highly synchronized events known as “mass spawning”. On some Western Australian reefs there are two mass spawning events each year, a primary event in autumn and a secondary event in spring, but at the outset of this project it was uncertain how widespread the two events were. Reproductive surveys from 12°S to 23°S showed that at all latitudes, a high proportion of species spawned in autumn (82-98%); however, there was a correlation with latitude in the spring spawning event (r2 = 0.72), with a decrease in the proportion of species spawning from 49% at Ashmore Reef (12°S) to

7% at Ningaloo Reef (23°S).

My previous research had shown that in some biannually-spawning species of

Acropora, conspecific colonies spawned in autumn or spring but not both, but it was unknown whether the autumn- and spring-spawning cohorts were reproductively isolated or whether colonies switched spawning time at random. Here, population genetic analysis of microsatellites in sympatric ( samoensis) and allopatric (A. tenuis) autumn- and spring-spawning colonies showed that the reproductive cohorts were genetically differentiated (FST = 0.17 in both species), confirming strong isolation.

Furthermore, in both species the seasonal cohorts had highly divergent lineages of the nuclear intron PaxC that were not present in the other DNA sequence markers. The unexpected finding that PaxC showed a different pattern to the other phylogenetic

i

markers was tested via comparative phylogenetic analysis of 20 Western Australian

Acropora species using the mitochondrial Control Region, PaxC and 10,034 genome- wide SNPs. This analysis confirmed the atypical pattern in PaxC, and suggested a selective connection to timing of reproduction in Acropora, raising the possibility that the PaxC gene or intron might play a role in coral spawning. In addition, the utility of genome-wide markers provided a finer resolution of the Acropora phylogeny than the

CR or PaxC, and a number of cryptic species were discovered.

Phylogeographic and population genetic analyses of Acropora tenuis over 12° of latitude revealed a phylogenetic break between Ashmore Reef and all other WA reefs, suggesting that the post-Pleistocene re-colonization of WA reefs was from two different sources. The integration of biogeographic history, genetics and spawning time lead to the conclusion that rather than being an inherited genetic legacy, seasonal breeding patterns are a result of long-term natural selection.

This study has three important implications for conservation and future research.

First, because the seasonality of breeding seasons in Western Australia is influenced by local selection it is imperative that more research is devoted to understanding exactly which environmental factors drive reproductive schedules and how they will be affected by rapid climate change. Second, coral biodiversity on Western Australian reefs is higher than is typically accounted for, due to the incidence of cryptic species; moreover reproductive timing can contribute to cryptic speciation, so it is vital that reproductive timing is incorporated into population genetic studies of corals, and that cryptic species are identified. Third, PaxC appears to be influenced by some selective connection to timing of reproduction in corals, so it should be used with caution as a phylogenetic marker.

ii

Table of Contents

Abstract ...... i Declaration and Publications...... v Acknowledgements ...... vii Foreword ...... xi 1. General Introduction ...... 1 2. Biannual coral spawning decreases at higher latitudes on Western Australian Reefs ...... 9 Abstract ...... 10 Introduction ...... 10 Methods ...... 12 Results & Discussion ...... 15 3. Asynchronous spawning in sympatric populations of a hard coral reveals cryptic species and ancient genetic lineages ...... 21 Abstract ...... 22 Introduction ...... 22 Methods ...... 25 Results ...... 33 Discussion ...... 43 4. Asynchronous spawning and demographic history shape genetic differentiation among populations of the hard coral Acropora tenuis in Western Australia ...... 49 Abstract ...... 50 Introduction ...... 50 Methods ...... 52 Results ...... 58 Discussion ...... 62

5. Phylogenomics provides new insight into evolutionary relationships and genealogical discordance in the reef-building coral genus Acropora ...... 71 Abstract ...... 72 Introduction ...... 72 Methods ...... 74 iii

Results ...... 78 Discussion ...... 83

6. Synthesis ...... 87 References ...... 98 Appendices ...... 111

iv

Declaration and Publications

This thesis is presented as a series of papers. All parts of this thesis have been designed, executed, and written by Natalie Rosser with advice from my supervisor, Michael S. Johnson. Chapter 5 was written with five co-authors (listed below). I confirm that I made the following contribution to that chapter: 80% of the design, 40% of the data collection, 85% of the data analyses, 85% of the interpretation and 90% of the writing. I have obtained permission from all co-authors of Chapter 5 to include this chapter in my thesis.

The details of publications arising from this thesis are:

Chapter 2: Rosser NL (2013) Biannual coral spawning decreases at higher latitudes on Western Australian reefs. Coral Reefs 32, 455-460.

Chapter 3: Rosser NL (2015) Asynchronous spawning in sympatric populations of a hard coral reveals cryptic species and ancient genetic lineages. Molecular Ecology 24, 5006-5019

Chapter 4: Rosser NL (2016) Asynchronous spawning and demographic history shape genetic differentiation amoung populations of the hard coral Acropora tenuis in Western Australia. Molecular Phylogenetics and Evolution 98, 89-96.

Chapter 5: Rosser NL, Thomas L, Stankowski S, Richards ZT, Kennington WJ, Johnson MS (in review) Phylogenomics provides new insight into evolutionary relationships and genealogical discordance in the reef-building coral genus Acropora. Proceedings of the Royal Society B Biological Sciences

------Natalie Rosser

------Michael Johnson (coordinating supervisor)

v

vi

Acknowledgements

First and foremost I offer my sincere thanks to Mike Johnson, to whom I am overwhelmingly grateful, for his endless patience, wisdom, advice and guidance in the supervision of this study; thank you for taking me on, Mike. I also especially thank Yvette Hitchen for her invaluable help, advice and cheerfulness in the lab, which was absolutely essential to this study. A heartfelt thanks to Brad Burzec, Inger Shimell, Ivor Bruce, Eamon Dorricott and Pete Rosser for helping me in the field, with an extra special thanks to Eamon and Pete for putting up with my morning sickness and ill-humor during that field trip. Many thanks to o amilton for starting me off in the lab including teaching me how to use a pipette; staff at RPS for collecting the Acropora samoensis samples used in Chapter 3; Jim Underwood and the Australian Institute of Marine Science for donating the A. tenuis DNA used in Chapter 4; Zoe Richards and the Western Australian Museum for collecting the Kimberley samples and allowing me to use them in Chapters 4 and 5; Luke Thomas for donating the samples from the Abrolhos Islands used in Chapter 5; Carden Wallace for identifying the samples collected in Chapters 2 and 4; Sean Stankowski for running RAxML in Chapter 5; Stuart Field at DPaW for inviting me on his Montebellos field trip and giving me access to the reproductive samples in Chapter 3; my office and lab mates for sharing the long journey with me (Esther Levy, Frances Leung, Ana Hara, Jamie Tedeschi, Kaori Yokochi, Veronica Philips, Phil Allen, Luke Thomas and Elf); my faithful and loving cat, Scout, for all the hours you kept me company while I wrote; the members of my thesis review panel Jason Kennington and Jane Prince for keeping me on track and reviewing my thesis; and my external examiners David Ayre, Allen Chen and Steve Palumbi for examining my thesis. I gratefully acknowledge the financial support I received from the following people/organizations which made this project possible: the Holsworth Wildlife Endowment, UWA School of Animal Biology, UWA Convocation, the Australian Coral Reef Society, and Prof Mike Johnson. I also received in-kind support from Australian Customs and The Department of Parks and Wildlife. All research was undertaken with the appropriate state and federal permits. To my friends and family, especially Penny Bunning, Mike Forde, Jas Cullen, Aurora Brosnan, Tennille Irvine, Pete Rosser and Sandy Rosser, thank you for your

vii

ongoing support, encouragement and humor at crucial times when I needed it most. My extra special thanks to Penny and mum for all the hours you spent minding Lilly for me, and for always being there on the end of the phone.

viii

I dedicate this thesis to my dad, Andrew Rosser, who sadly passed away during my PhD. He was my greatest inspiration, and he taught me perseverance and grit. I miss him every day.

I also dedicate this thesis to my two most important people: my long-suffering husband, Patrick Hollingworth, whose love, kindness and support I couldn’t be without; and my little Lilly, who makes my life so rich with happiness.

ix

x

Foreword

Why should we care about coral reefs?

Coral reefs are spectacularly beautiful with exceptionally high biodiversity and a wide range of intrinsic, ecological and economic values. South East Asia contains the largest area of coral reefs in the world (34% of the world’s total), and more than 60% of the ~600 million people of South East Asia live within 60km of the coast (Wilkinson 2008). Most of these people are highly dependent on coral reefs for food security, livelihoods, building materials, medicines and trade goods, so for many people the value of coral reefs may be purely economic; the most recent assessment of the potential economic value of coral reefs in south-east Asia was US$12.7 billion (Wilkinson 2008). For other people, the value of coral reefs lies in their recreational use and aesthetic value, with divers, snorkellers, swimmers, surfers, anglers, boating enthusiasts and others sharing in their recreational benefits. Natural environments also provide spiritual values, and large, natural, unmodified seascapes, otherwise known as wilderness, and can be valued for the solitude and limited access by humans. Yet others would argue that coral reefs have intrinsic value; that all life depends on the functioning of natural systems to ensure the supply of energy and nutrients, and that the ecological processes that maintain the integrity of the biosphere should be maintained (McNeely et al. 1990).

“Because after the last open coast of Australia is tamed, polluted and overfished, what’s left except nostalgia and the desert at our backs?” (Tim Winton, Land’s Edge)

xi

xii

Chapter 1

General Introduction

“West coasts tend to be wild coasts, final coasts to be settled, lonelier places for being last. In Australia the east coast is the pretty side, the Establishment side, the civilized side...As in Ireland and America, our west is seen as something of a new frontier, remote and open.” (Tim Winton, Land’s Edge)

1

Western Australia’s coral reefs

Coral reefs in Western Australia span over 17° of latitude, from the tropical offshore reefs near Indonesia, to the subtropical reefs at the Houtman Abrolhos Islands.

Unlike the Great Barrier Reef on the east coast of Australia, which is a series of nearly continuous reefs that stretch for over 2,000 km, the north-west coast is characterized by a series of discontinuous reefs, which occur on the continental slope, on the continental shelf edge and along the coastline. The offshore systems consist of atoll-like reefs on the continental slope that rise from deep ramp settings (e.g. Scott Reef and the Rowley

Shoals), as well as shallower reefs perched on the edge of the continental shelf (e.g.

Ashmore Reef), while the inshore reefs occur along the coastline and around inshore islands (Fig. 1.1).

Tropical marine fauna was established on the Western Australian coast in the early to middle Miocene, when massive evolutionary radiation and latitudinal expansion of the fauna of the Sea of Tethys was occurring (Wilson 2013). Western

Australian coral reefs have a long history of contraction and expansion in response to oscillating glacial cycles, and consequent cooling and warming in the Pliocene and

Quaternary saw the contraction and expansion of the tropical fauna on the north-west coast. The most recent of these was the last Pleistocene glaciation ~20,000 years ago, when sea level was -130 m. At that time, the wide continental shelf of the north-west coast was exposed, and the position of the WA coastline was along the 120 m contour of the Rowley shelf (Yokoyama et al. 2001), leaving the present-day coastal reefs, including Dampier, the Montebello Islands, and the Kimberley coast, on dry land. While the offshore atolls of Scott Reef and the Rowley Shoals would have existed during the

LGM, whether the lower global temperatures and contracted habitats on the atolls supported coral populations has been widely debated (Wilson 2013). 2

Fig 1.1. A bathymetric map of north-west Australia showing the location of the coral reefs included in this study (red stars). The margin of the continental shelf lies just beyond the 200 m contour.

It was thought that the connection between the Indo-West Pacific reefs of the

Indonesian Archipelago and Western Australia’s reefs is maintained in the present day by a series of southward-flowing currents that originate in the Pacific and Indian Oceans

(Nof et al. 2002; Domingues 2006; D'Adamo et al. 2009). The flow of these currents is strongest during the austral autumn (March-May; Holloway & Nye 1985), which corresponds to the time of the major coral spawning season on the west coast, and this was thought to create significant gene flow between northern Indonesian reefs, and

Western Australian reefs (Simpson 1991). Simpson also proposed that the breeding season of corals in Australia is the result of an inherited, endogenous rhythm from northern ancestral populations that influences reproductive seasonality on Western

3

Australian reefs (Simpson 1988, 1991). He argued that, while the adaptive value may have been lost in the descendant populations, the spawning rhythm remains because the new regime does not exert a selective pressure to counteract the genetic legacy

(Babcock et al. 1994), or because the genetic connection between the regions is high enough to inhibit the ability of the population to adapt to new conditions (Simpson

1988). This hypothesis has not been tested; however, a study of connectivity in the broadcast spawning coral Acropora tenuis (Underwood 2009a) showed significant genetic divergence between northern-offshore reefs (Scott Reef and Rowley Shoals) and the southern-coastal reefs (Dampier and Ningaloo), suggesting that the genetic connectivity between Indonesia and Western Australia is probably not high. In addition, the recent discovery of biannual spawning in Western Australia has complicated interpretations of spawning season.

Biannual coral spawning in Western Australia

Most reef building corals are broadcast-spawners that spawn once each year

(Baird et al. 2009a). In some regions, including Australia, a large number of species participate in multi-specific or “mass” spawning events, where 20-30 species have been recorded spawning on the same night during a mass spawning event (Willis et al. 1985;

Babcock et al. 1986). Synchronized spawning within coral populations results in higher concentrations of gametes and better fertilization success (Oliver & Babcock 1992;

Levitan et al. 2004), which increases the probability of successful reproduction among corals that spawn together (Harrison 2011).

While the seasonal timing of coral spawning events varies around the globe, at each location annual coral spawning is highly predictable (Willis et al. 1985; Vize et al. 2005; Levitan et al. 2011). In Western Australia historical reproductive surveys

4

showed that most broadcast-spawning corals have a single annual reproductive period in autumn (Babcock et al. 1994; Simpson 1991), but more recent evidence indicates that on some reefs in north Western Australia there is a secondary reproductive season in spring (Rosser & Gilmour 2008; Gilmour et al. 2009; Rosser & Baird 2009). Most broadcast-spawning corals have an annual gametogenic cycle and spawn once a year, but there are some instances in which corals undergo two gametogenic cycles, culminating in spawning twice a year (Stobart et al. 1992; Guest et al. 2005; Mangubhai

& Harrison 2006). It was initially thought that some corals in Western Australia were spawning twice each year (L. Smith pers. comm.), but research showed that among conspecific colonies that spawned in October and March, each individual had only one gametogenic cycle, spawning in either spring or autumn but not both (Rosser &

Gilmour 2008). This raised the possibility that conspecific spawning populations may be separate reproductive populations, possibly representing two evolutionary lineages, or alternatively, that colonies may sometimes switch their spawning time. At the outset of this project, the degree to which conspecific colonies spawning in spring and autumn on Western Australian reefs were reproductively isolated and genetically differentiated was unknown. High levels of genetic differentiation (and potentially the early stages of speciation) among conspecific colonies of corals that spawn in different months have been recorded at other locations in the Indo-Pacific (Dai et al. 2000; Wolstenholme

2004), and it has been hypothesized that spawning at different times (e.g. different hours on the same night) has been the basis for reproductive isolation and speciation in several sympatric species of Acropora (van Oppen et al. 2001; Fukami et al. 2003).

The control of reproductive timing is complex, and it is thought that multiple environmental cycles control the season, lunar phase and hour that spawning occurs

(Babcock et al. 1986; Vize et al. 2009). The seasonal cycle that controls coral spawning

5

is the least well-understood, but correlative studies suggest it is influenced by solar insolation levels (Penland et al. 2004; van Woesik et al. 2006), water temperature

(Babcock et al. 1986; Fan & Dai 1999; Nozawa 2012) and calm weather (van Woesik

2010). The night of spawning is set by the lunar phase (Babcock et al. 1986; Hunter

1988; Oliver et al. 1988), and the hour (and even minute) of the spawning window is set by sunset time (Knowlton et al. 1997; Levitan et al. 2004; Vize et al. 2009). There is also a genetic component that underlies spawning time (Levitan et al. 2011), but it is unclear how much the timing of coral spawning is entrained by biological rhythms or regulated directly by environmental signals. Corals contain several circadian clock genes that are likely to regulate entrained processes (Levy et al. 2007; Vize 2009;

Shoguchi et al. 2013), but their exact role in regulating reproductive behavior is not well understood.

The importance of a genetic perspective

The use of genetics to understand the origin and consequences of seasonal breeding in corals in Western Australia is a powerful approach that will allow me to test for reproductive isolation between the spring and autumn spawners (hence whether individuals retain their spawning season), and the evolutionary relationships of spring and autumn spawners at different geographic scales. Individuals vary in the composition of their DNA, and the fate of any given genetic variant will be influenced by the biological and ecological circumstances of its life, such as reproductive success, migration, population size, connectivity, natural selection and historical events

(Sunnucks 2000). Hence, by measuring the genetic variation among individuals and applying models, we can make inferences about the biology of organisms and the processes that have shaped their existence. Different genetic markers have different

6

rates of change, and so provide information about population biology over different time scales. Microsatellite markers, for example, can describe small genetic differences between populations, and are useful for discerning information about population structure, gene flow and dispersal (Estoup & Angers 1998), but high mutation rates and constraints on allele size eliminate signature events in the distant past, so they have limited use in inferring evolutionary history (Garza et al. 1995; Paetkau et al. 1997;

Selkoe & Toonen 2006). DNA sequences are useful for inferring evolutionary relationships, and well-resolved phylogenetic trees are useful for understanding evolutionary history and longer term processes such as speciation and selection (Avise

2004). Theoretical models use either gene frequencies (as in the case of microsatellites or single-locus markers) or geneology and genetic distance (as in DNA sequences) to measure the distribution of genetic variation.

The extent to which natural populations become genetically differentiated depends largely upon the amount of gene flow between them (e.g. Johnson & Black

1998, 2006b; Ayre & Hughes 2004). Reproductive isolation many arise when some sort of barrier prevents groups from exchanging gametes and genes, such as asynchronous spawning, habitat specialization, selection, or gamete compatibility (Palumbi 1994).

Such reproductive isolation constrains gene flow between groups, and may eventually lead to genetic divergence and speciation. Hence differences in reproductive timing that result in genetic subdivision within populations, is potentially a powerful component of the evolution of biological diversity and ultimately in the splitting of lineages to form new species.

7

Aims and scope of this thesis

This thesis explores the influence of asynchronous reproductive timing on patterns of genetic structure in coral populations, at both ecological and evolutionary levels, in the context of understanding the evolution of seasonal breeding patterns in

Western Australia. Each chapter focuses on a different set of questions that are described in the introduction of each chapter. Chapter 2 documents the geographical extent of coral spawning in autumn and spring on WA reefs, to understand how the seasonality of spawning varies with latitude. Chapter 3 explores the extent to which spring- and autumn-spawning cohorts of Acropora samoensis are reproductively isolated and genetically diverged in sympatric populations in the Pilbara region. Chapter

4 widens the geographic scope, examining the phylogeography and population genetics of allopatric populations of A. tenuis with different spawning patterns. Finally, Chapter

5 expands on the unexpected results from Chapters 3 and 4, which revealed divergent selection on the PaxC marker, and compares molecular phylogenies from 20 species of

Acropora using the mtDNA control region, PaxC, and 10,034 genome-wide SNPs to test whether PaxC distorts phylogenetic inferences.

The broad questions integrating these chapters are:

(i) Are conspecific colonies that spawn in autumn and spring reproductively isolated and genetically differentiated, or do colonies switch spawning time, allowing genetic mixing?

(ii) Are conspecific autumn- and spring-spawning colonies associated with distinct phylogenetic lineages?

(iii) Are spawning patterns on WA reefs a result of an inherited legacy from northern ancestors, or natural selection?

8

Chapter 2

Biannual coral spawning decreases at higher latitudes on

Western Australian reefs

This chapter has been published in Coral Reefs:

Rosser NL (2013) Biannual coral spawning decreases at higher latitudes on Western

Australian reefs. Coral Reefs 32, 455-460.

9

Abstract

Seasonal differences in the timing of multi-specific coral spawning between the eastern and western coasts of Australia may be the result of a genetic legacy or of adaptation to local conditions. Using estimates of the proportions of coral species that spawned in spring and autumn at Ashmore Reef (12°S) and Ningaloo Reef (23°S) in Western

Australia, in combination with findings of previous surveys, I examined whether reproductive seasonality varied with latitude. A consistently high proportion of species spawned during the main reproductive season in autumn regardless of latitude.

However, there was a clear decrease in the proportion of species spawning in spring, from an average of 49% at Ashmore Reef (12°S) to 7% at Ningaloo Reef (23°S).This suggests that seasonality of coral reproduction in Western Australia reflects environmental gradients and natural selection rather than an inherited genetic legacy.

Introduction

Broadcast spawning corals typically reproduce once a year in highly synchronized events (Baird et al. 2009a). In many locations, the timing of multi-specific spawning is remarkably predictable (Willis et al. 1985; Simpson 1991; Vize et al. 2005;), but the factors that influence reproductive seasonality are still poorly understood. In Australia, synchronized, multi-specific spawning occurs primarily in spring on the east coast and in autumn on the west coast, while there is also a secondary, multi-specific spawning period in the opposite season (in autumn on the east coast and in spring on the west coast; Stobart 1994; Wolstenholme 2004; Rosser & Gilmour 2008).

Two hypotheses have been proposed to explain the difference in spawning seasons on the east and west coasts of Australia. The ‘genetic legacy’ hypothesis

10

(Simpson 1988, 1991) suggests that the reproductive seasonality of corals in Australia is the result of an inherited, endogenous rhythm from northern ancestors that controls spawning seasonality. The different coral spawning seasons on east and west coasts in

Australia result from the prevailing southward-flowing currents on each coast in different seasons (the Leeuwin Current in autumn on the west coast, and the East

Australian Current in spring on the east coast), that selectively disperse coral larvae down the west coast in autumn, and down the east coast in spring. This implies that local environmental variables are not strong enough to overcome the inherited historical constraints (Babcock et al. 1994).

An alternative hypothesis is that reproductive seasonality on each coast is the outcome of long-term natural selection to spawn when conditions favour offspring survival (Oliver et al. 1988). Reproductive seasonality in corals could be influenced by water temperature (Willis et al. 1985), photoperiod (Babcock et al. 1994), monthly rainfall (Mendes & Woodley 2002), solar insolation (van Woesik et al. 2006), and regional wind fields (van Woesik 2010). However, none of these variables can be easily linked to the timing of coral spawning events in Western Australia.

The occurrence of multi-specific coral spawning in spring on some Western

Australian reefs was discovered fairly recently (Rosser & Gilmour 2008; Gilmour et al.

2009), and reproductive surveys have suggested that spring-spawning patterns may vary geographically within Western Australia (Rosser & Baird 2009). This heterogeneity could provide insight into the factors that influence reproductive seasonality, but the available data are limited. Here, I surveyed the reproductive state of coral assemblages in spring and autumn at Ashmore Reef (12ºS) and Ningaloo Reef (23ºS), so extending the latitudinal range both north and south of previous surveys (Rosser & Gilmour 2008;

11

Gilmour et al. 2009; Rosser & Baird 2009), to better determine how the seasonality of spawning varies with latitude on the Western Australian coast.

Methods

Reproductive surveys were made at Ashmore Reef (12°14’ S, 122°58’ E) and Ningaloo

Reef (22-23°10' S, 113°45' E) (Fig. 2.1), extending the latitudinal range of sites studied in WA by 4° of latitude (600 km). Surveys were made at both Ashmore and Ningaloo

Reefs in spring 2010, at Ashmore Reef in autumn 2011, and at Ningaloo Reef in autumn

2012, on and around the full moons of October and February in each case. During each survey every attempt was made to sample five colonies from a minimum of 20 species

(Styan & Rosser 2012) that typically participate in “mass spawning” events (i.e., hermaphroditic broadcast spawners; Harrison & Wallace 1990). If fewer than five colonies per species were sampled, the species was eliminated from the data set unless spawning was detected in at least one colony, in which case the species was included in the data set.

Cumulative binomial probability distributions were calculated in MS Excel using the BINOMDIST function, to estimate the probability that at least one spawning colony would be detected when sampling five colonies per species (Appendix 2.1). This showed that when sampling five colonies per species, spawning was only likely to be detected in a given species when > 40% of colonies were spawning Table 2.1).

Therefore this sampling design may underestimate the proportion of species spawning because it is only likely to detect spawning in species where a high proportion of colonies are spawning, however, this agrees with the original definition of “mass” synchronous spawning (Willis et al. 1985).

12

Surveys were conducted in situ by breaking open a small piece of each colony to reveal developing oocytes (Baird et al. 2009b). The reproductive state of each colony was classified based on the visibility and colour of developing oocytes; colonies were assumed to be going to spawn within three months if pigmented or white oocytes were visible. I assumed that they would not spawn within three months if oocytes were not visible (Babcock 1984; Harrison et al. 1984; Wallace 1985; Kenyon 1992; Hanafy et al. 2010). In corals with small oocytes (e.g., faviids), broken samples were collected and examined with a 10x hand lens out of the water. Colonies were selected haphazardly, with the criteria that they were of reproductive size (colony diameter >

20cm; Wallace 1985) and were not visibly compromised (e.g., bleached or disease- affected, which may affect their ability to reproduce). Corals were identified in-situ, or skeletal samples were collected and sent to the Museum of Tropical Queensland for identification.

115°0'0"E 120°0'0"E 125°0'0"E 130°0'0"E

Timor Sea 10°0'0"S 10°0'0"S

Ashmore Reef

Bonaparte Scott Reef Archipelago

15°0'0"S 15°0'0"S Indian Ocean

20°0'0"S Dampier 20°0'0"S Barrow Island

Ningaloo Reef 0 200 400 Kilometers

115°0'0"E 120°0'0"E 125°0'0"E 130°0'0"E

Fig 2.1. Sampling locations of spring and autumn surveys. Dark circles indicate sites surveyed in this study, light circles indicate sites surveyed previously 13

To test whether the proportion of species spawning in a given season differed between sites, the frequency of species spawning or not spawning in each season at Ashmore and Ningaloo Reefs were compared using Fisher’s exact tests

(SPSS). Species were only included if five or more colonies were sampled at both sites.

To test for a general association with latitude, the proportions of species spawning in spring and autumn in this study were combined with data from three other sites (Fig.

2.2): Scott Reef (Gilmour et al. 2009); the Bonaparte Archipelago (Rosser and Baird

2009); and the Dampier Archipelago (Baird et al. 2011). The previous surveys used the same in-situ reproductive assessment as in this study (i.e., reproductive state was not classified histologically) and similar classification criteria for maturity of oocytes (the difference being that data from Scott Reef and the Bonaparte Archipelago included only the numbers of colonies with pigmented oocytes, and not white oocytes). Colonies were classified as spawning in “spring” if they had mature oocytes in October, November or

December, and as spawning in “autumn” if they had mature oocytes in February, March or April. Permutation tests were used to test for heterogeneity of spawning proportion among sites using the 2 x N Monte Carlo contingency test in P-Stat (Bill Engels, 1993-

1997, http://engels.genetics.wisc.edu/pstat/).

Table 2.1. Relationship between proportion of colonies of a species that have ripe oocytes and the probability of detecting ripe oocytes in a sample of 5 colonies per species

Proportion of colonies spawning within a chance of detecting ≥ 1 spawning colony when species sampling 5 colonies per species 10 % 40 % 20 % 70 % 30 % 80 % 40 % 90 % 50 % 100 %

14

Results & Discussion

There was a significant difference in the proportion of species spawning in spring between Ashmore and Ningaloo Reefs (n=18, p = 0.008, 2-tailed Fisher’s Exact test). At

Ashmore Reef 49% of species (19 of 39) were estimated to spawn in spring, compared with only 7% species (2 of 27) at Ningaloo Reef (Table 2.2, Fig. 2.2). Contrastingly in autumn, there was no significant difference between the proportions of species spawning at Ashmore and Ningaloo Reefs (n=22, p= 0.4): 82% species (41 of 50) at

Ashmore Reef, and 95% species (37 of 39) at Ningaloo Reef (Table 2.2, Fig. 2.2).

Combining the data from Ashmore and Ningaloo Reefs with those from the three previously surveyed WA sites (Fig. 2.2) showed that there is a clear latitudinal component to the level of mass spawning in spring on WA reefs. The proportions of species spawning in spring at different latitudes ranged from 7-57% (p = 0.00005, 2 x N

Monte Carlo contingency test). In contrast, a high proportion of species spawn in the main reproductive season in autumn at all latitudes (range 82-98%), although the proportions did differ among the sites (p = 0.01, 2 x N Monte Carlo contingency test).

If reproductive timing in Australia is the result of an inherited rhythm from ancestral populations to the north of Western Australia (Simpson 1991), then a latitudinal decline in spring spawning may be a function of dispersal distance from source populations, due to the weak flow of the Leeuwin Current in spring (Cresswell

1991). However, at least three species (Acropora. tenuis, A. millepora and A. secale) mostly spawned in spring at Ashmore Reef in the north, but mostly in autumn at

Ningaloo Reef in the south (Table 2.2). Thus, if reproductive seasonality is a result of an inherited rhythm, either the southern Ningaloo populations are not connected to the northern Ashmore populations, or corals that spawn in spring conditions at Ningaloo

15

Reef are selected against and removed from the contemporary population (van Woesik

2010). The predominance of autumn spawning at Ningaloo Reef seems unlikely to be simply the result of historical constraints on which genotypes arrived from the north. In seasonal environments, organisms that reproduce only once a year can be expected to time their reproduction to coincide with the most suitable conditions for success.

Fig. 2.2. The proportion of species spawning in spring (Oct-Dec) and autumn (Feb-Apr) at different latitudes in Western Australia; (a) = this study, (b) = Gilmour et al. 2009, (c) = Rosser and Baird 2009, (d) = Baird et al. 2011, (e) = this study. Error bars are 95% confidence intervals.

If natural selection determines the timing of reproductive seasonality in Western

Australia, then a latitudinal decline in spring spawning may be a result of environmental gradients that limit successful reproduction in spring, i.e., successful gametogenesis, fertilisation or larval survival do not occur in the majority of species above or below certain values. For example, sea temperature may control timing of reproduction because the final maturation of gametes requires a minimum water temperature (Hunter

1988; van Woesik et al. 2006; Baird et al. 2009b). Similarly wind speed can influence fertilisation, so predictably strong winds may limit reproductive success at certain times

16

in particular locations (van Woesik 2010). The latitudinal decline in the secondary spawning season, but not in the main spawning season in this study and others (Baird et al. 2009a) suggests that there is not a simple correlation between a single environmental variable and spawning. It is unlikely that one climatic variable would act alone to influence reproductive seasonality, as shown by plants and some broadcast-spawning invertebrates that respond to changes in the combination of temperature and day length

(Olive & Pillai 1983; Simpson & Dean 2002). Hence, reproductive seasonality is likely to depend on a combination of environmental variables.

The occurrence of multiple environmental variables acting in concert or synergistically to influence reproductive seasonality in corals has not been fully explored in any study. A multivariate analysis on the interactions between the number of species spawning, latitude, and climatic variables was beyond the scope of this study because of the small number of sites/populations (meaning any multivariate analysis of this data would have more independent variables than the sample size and any complex analysis would be invalid). However, this is an interesting avenue for future research that may provide insight into the factors that influence reproductive seasonality, not only on Australian reefs, but on coral reefs worldwide. The patterns revealed in this study offer much scope for testing hypotheses about the factors that influence reproductive seasonality in the hope of furthering our understanding of the evolution of reproductive seasonality in broadcast spawning corals.

17

Table 2.2. The proportion (%) of colonies in each species that spawn in spring and autumn at Ashmore and Ningaloo Reefs as determined by in-situ observations of gamete maturity; * indicate cases where a single colony was sampled and found to have mature gametes, but no inference was made about the proportion of all colonies that were reproductively active.

Ashmore Ningaloo Spring N Autumn N Spring N Autumn N (%) (%) (%) (%) Acropora acueleus 50 2 Acropora anthocercis 71 7 57 7 100 4 Acropora austera 40 5 0 5 0 6 Acropora cerealis 0 5 100 7 0 5 Acropora cytherea 60 5 63 8 * 1 Acropora digitifera 0 5 100 6 0 5 100 8 Acropora divaricata 0 5 57 7 100 4 Acropora florida 67 6 33 9 0 5 83 6 Acropora gemmifera 29 7 40 10 100 4 Acropora grandis 0 5 33 3 0 6 86 7 Acropora hoeksemani 50 2 Acropora humilis 43 7 56 9 * 1 Acropora hyacinthis 20 5 17 6 0 5 80 5 Acropora intermedia 63 8 0 5 57 7 Acropora latistella 0 5 86 7 71 7 Acropora listeri 83 6 0 6 * 1 Acropora loripes 50 10 0 9 Acropora lutkeni 100 5 0 6 Acropora micropthalma 0 5 40 5 Acropora millepora 60 22 0 20 0 17 91 20 Acropora monticulosa 60 5 0 7 Acropora muricata 0 5 55 11 0 5 56 9 Acropora nana 25 4 Acropora nasuta 0 5 100 6 Acropora paniculata 0 5 100 3 Acropora papillare 0 6 75 8 0 9 Acropora robusta 0 5 100 5 0 5 67 6 Acropora samoensis 0 5 29 7 0 9 82 11 Acropora secale 60 5 0 8 0 5 100 6 Acropora spicifera 38 8 83 6 0 5 56 9 Acropora subulata 0 5 40 5 Acropora sukarnoi * 1 67 3 Acropora tenuis 80 20 0 20 0 16 68 15 Acropora valensinesi 83 6 Acropora valida 0 5 83 6 0 5 100 6 Acropora vaughni * 1 Echinophyllia aspera 0 5 100 7 0 5 100 5 Echinopora lamellosa 80 5 0 5 50 6

18

Favia matthai 100 3 67 3 Favia pallida 0 5 100 6 100 7 Favia speciosa 80 5 Favia stelligera * 1 Favites halicora 0 6 71 7 80 5 Galaxea astreata 100 5 Goniastrea favulus 67 3 Goniastrea pectinata 57 7 Gonistrea retiformis 20 5 40 5 0 6 83 6 Hydnophora excesac 0 5 0 6 75 4 Hydnophora rigida 20 5 71 7 Leptoria Phrygia 60 5 Lobophyllia hemprichii 0 5 80 5 Montipora capricornis 0 5 80 5 Montipora spumosa 0 5 80 5 Merulina ampliata 0 5 60 10 0 5 80 5 Merulina scabricula 0 5 78 9 0 5 60 5 Montastrea curta * 1 Oulophyllia crispa 0 5 Pachyseris speciosa 0 5 Platygyra daedalea * 1 0 7 100 6 Platygyra pini 67 3 100 2 Podobacia crustacean 86 7 Total species 19 39 41 50 2 27 37 39

19

20

Chapter 3

Asynchronous spawning in sympatric populations of a

hard coral reveals cryptic species and ancient genetic

lineages

This chapter has been published in Molecular Ecology:

Rosser NL (2015) Asynchronous spawning in sympatric populations of a hard coral reveals cryptic species and ancient genetic lineages. Molecular Ecology 24, 5006-5019.

21

Abstract

Genetic subdivision within a species is a vital component of the evolution of biodiversity. In some species of Acropora corals in Western Australia, con-specific individuals spawn in two seasons six months apart, which has the potential to impede gene flow and result in genetic divergence. Genetic comparison of sympatric spring and autumn spawners of Acropora samoensis was conducted to assess the level of reproductive isolation and genetic divergence between the spawning groups based on multiple loci (13 microsatellite loci, the mitochondrial control region and two nuclear introns). Bayesian clustering and principal co-ordinates analysis of the microsatellite loci showed a high level of genetic differentiation between the spawning groups (F’ST =

0.30; P < 0.001), as did the sequence data from PaxC and Calmodulin (ΦST = 0.97 and

0.31, respectively). At the PaxC locus the autumn and spring spawners were associated with two divergent lineages that were separated by an evolutionary distance of 1.7%, and statistical tests suggest divergent selection in PaxC, suggesting this gene may play a role in coral spawning. This study indicates that the autumn and spring spawners represent two cryptic species, and highlights the importance of asynchronous spawning as a mechanism influencing speciation in corals.

Introduction

Genetic subdivision within populations can ultimately lead to the splitting of lineages and the evolution of new species, and molecular genetic studies of population structure have resulted in the discovery of many cryptic species across both plant and animal kingdoms (Bickford et al. 2007). Genetic differences may arise when some sort of barrier prevents cohorts from exchanging genes, and one possible mechanism

22

influencing within-species diversity is asynchronous reproduction. In sympatric species, differences in the timing of reproduction that prevent gametes from crossing are likely to impede gene flow and result in genetic divergence between reproductive cohorts.

Studies of flowering plants, salmonid fishes, sea urchins and other invertebrates have shown that sympatric populations with different breeding times are often significantly and substantially differentiated at neutral loci (Hendry & Day 2005; Savolainen et al.

2006; Bird et al. 2011; Binks et al. 2012). Ultimately the degree to which populations may become differentiated depends not only on the level of reproductive isolation between them, but also on the nature of selection on particular genes. During the early stages of ecological speciation, genomic regions of functional importance quickly diverge under selection (Turner et al. 2005; Via 2009), so studying populations in which ecotypes or races are not yet completely reproductively isolated can reveal genetic changes that contribute to reproductive isolation before they are confounded by additional genetic differences that accumulate after speciation (Via 2009).

In addition to molecular and morphological differentiation, cryptic species have been identified along other lines of evidence, including habitat specificity (Warner et al. 2015), mode of reproduction (Schmidt-Roach et al. 2012) and symbiont association (Pinzon & LaJeunesse 2010). However, relatively few studies have investigated the association between differences in the timing of reproduction and cryptic speciation in corals. The importance of reproductive timing in maintaining gene flow is evident in broadcast-spawners which spawn gametes into the water where fertilization takes place externally within hours, so individuals that spawn more than several hours apart are unlikely to cross-fertilize. High levels of genetic differentiation among con-specific colonies of corals that spawn at different times have been recorded at various locations in the Indo-Pacific (Dai et al. 2000; Wolstenholme 2004), and even

23

spawning at different hours on the same night is sufficient to prevent colonies from interbreeding, and has resulted in genetic divergence (Knowlton et al. 1997; Fukami et al. 2003).

Genetic differentiation between the reproductive cohorts can occur only when reproductive timing is strongly heritable and not individually flexible. For example, snow buttercups (Ranunculus adoneus) do not show significant differentiation between early and late-flowering sites, because the variation in flowering time is determined by snow melt times (i.e. environmental) and does not have a genetic basis

(Stanton et al. 1997). Spawning time of corals, however, is thought to be strongly heritable, with some studies showing that individuals of Colpophyllia and Montastrea spawn at the same time every year (within minutes) over many years (Vize et al. 2005;

Levitan et al. 2011). Nevertheless, transplant experiments in Echinopora and

Montastrea have also shown that reproductive timing can change in response to new environmental conditions (Fan & Dai 1999; Levitan et al. 2011), suggesting that corals have some degree of reproductive plasticity.

Broadcast-spawning corals typically reproduce in annual, synchronized spawning events (Harrison & Wallace 1990; Baird et al. 2009a), which generally occur in autumn in Western Australia (Simpson 1991). Contrary to this pattern, however, there is also a secondary spawning event in spring on some reefs in north-western

Australia (Rosser 2013). A previous survey of two species (Acropora samoensis and

Acropora cytherea) in which colonies were tagged and sampled over multiple years, showed that adjacent, conspecific colonies spawned in different seasons (autumn and spring), raising the question of whether this asynchrony could lead to reproductive isolation (Rosser & Gilmour 2008). Here I examine genetic differences between individuals of Acropora samoensis on a reef where individuals that spawn in autumn

24

and spring occur adjacent to one another, and are morphologically indistinguishable.

The aim of this study was to determine: (a) the extent to which the autumn-spawning and spring-spawning cohorts are reproductively isolated and genetically differentiated and (b) whether the con-specific spawning groups represent independently evolving lineages.

Methods

Population and reproductive sampling

The reproductive state of Acropora can be gauged in-situ by examining the presence/absence of visible eggs, and colour of developing eggs in broken colonies

(Baird et al. 2002) in these hermaphroditic corals. Mature eggs in Acropora change from white to pink approximately 1-3 weeks prior to spawning (Harrison et al. 1984;

Wallace 1985, 1999), therefore colonies containing visible, pigmented eggs in-situ are assumed to spawn within approximately 3 weeks, while colonies with visible but white

(unpigmented) eggs are assumed to spawn within 1-3 months (Baird et al. 2002).

Reproductive and genetic sampling of Acropora samoensis was conducted on a reef off Barrow Island, Western Australia (20.7861 °S, 115.5067 °E). To locate the two reproductive groups of A. samoensis, 80 colonies were tagged and sampled in

November (spring) 2010, January (late summer) 2011 and October (spring) 2011.

Reproductive maturity was measured in-situ on each occasion and classified as

‘pigmented’ ‘white’ or ‘absent’ (after Baird et al. 2002). All colonies were located along two 100 m transects in the same habitat (patch reef / lagoonal habitat) at the same depth (approximately 6 m).

25

In addition, samples were collected and fixed in 10% formalin in seawater for preservation during each reproductive survey. Samples from a subset of the tagged colonies (n=38) were and then decalcified in a solution of 10% formic acid and dissected to verify the in situ assessments. Five polyps from each specimen were dissected and viewed under a stereomicroscope (Wallace 1985); egg size was measured using a graticule slide and calculated as the geometric mean (Wallace 1985), and the presence/absence of testes was noted. In the Acropora, mean egg size ranges from 200-

600 µm within 8 weeks of spawning, and from 300–945 µm prior to release (Wallace

1985; Szmant 1986; Kenyon 1992; Wallace 1999; Vargas-Angel et al. 2006; Rosser &

Gilmour 2008). Testes develop and become visible in dissected samples only 4–6 weeks prior to spawning, so the presence/absence of testes can also be used to gauge reproductive maturity (Wallace 1985). In each colony, the time of spawning was inferred from (a) the presence/absence of eggs and egg colour in situ, (b) the presence/absence of testes in a subset of dissected samples and (c) egg size in a subset of dissected samples.

Skeletal voucher specimens from sequenced colonies were bleached and examined by a coral taxonomist, Dr Zoe Richards, from the Western Australian

Museum (WAM). All voucher specimens are housed at the WAM (registration numbers

WAMZ84437 to WAMZ84464; corresponding GenBank Accession Numbers listed on

Dryad http://dx.doi.org/10.5061/dryad.20g8r and in Appendix 3.1).

PCR, DNA sequencing and microsatellite genotyping

Tissue samples collected for genetic analysis were preserved in 95% ethanol, which was replaced after 24 hours and again after one week. DNA was extracted from branch tips using DNeasy DNA extraction kits for animal tissue (Qiagen, USA) according to the

26

manufacturer’s instructions. To test for genetic differentiation between spawning groups, genotypes at 13 microsatellite loci were determined for each individual using primers developed by Wang et al. (2009) for Acropora (Appendix 3.2). Loci were amplified using fluorescently labelled primers in 13 single-plex reactions. PCRs of 5µL contained 1 µL of Bioline buffer, 0.5 µL of BSA, 0.1 µL each of forward and reverse primers, 0.1 µL of Bioline Taq, 2.2 µL of water and 1 µL of DNA. PCR amplifications were carried out in an Eppendorf Mastercycler, and consisted of an initial denaturation at 95°C for 3 mins, followed by 35 cycles of 60 sec at 95°C, 60 sec at 49°C, and 60 sec at 72°C, and finally 72°C for 5 mins. Fragments were analysed using an Applied

Biosystems 3730 capillary sequencer, and allele sizes scored using GeneMarker 1.91

(SoftGenetics, LLC). Microsatellite genotypes from all individuals were submitted to

Dryad ( http://dx.doi.org/10.5061/dryad.20g8r ).

DNA sequencing of a mitochondrial and two nuclear markers was used to determine whether seasonal spawning groups represent separate evolutionary lineages.

The mtDNA control region (CR) was amplified in a PCR using primers rns (5’-

GGTTTCTAATACCTCCGAGG-3’) and Cox3 (5’- TACATAACACTGCCCACAGT-

3’;van Oppen et al. 2001). The nuclear PaxC intron was amplified using the primers

PaxC_intron-FP1 (5’- TCCAGAGCAGTTAGAGATGCTGG-3’) and PaxC_intron-

RP1 (5’-GGCGATTTGAGAACCAAACCTGTA-3’; van Oppen et al. 2000), and the nuclear Calmodulin intron was amplified using the primers CalMf (5’-

GAGGTTGATGCTGATGGTGAG-3’) and CalMr2 (5’-

CAGGGAAGTCTATTGTGCC-3’;Vollmer & Palumbi 2002). PCRs contained 1µL

MgCl2 (50nM), 1.2 µL dNTPs (2.5nM), 0.2 µL platinum Taq, 2.5 µL 10 x PCR buffer,

1 µL each of the forward and reverse primers, 2 µL of DNA, and 17.1 µL dH2O in a 25

µL reaction. Positive (known DNA sample) and negative (no DNA) controls were

27

included with each reaction. Thermocycling profiles consisted of an initial denaturation step of 95°C for 3 min, followed by 35 cycles of 94°C for 30 sec, 59°C for 1 min and

72°C for 1 min, and finally 72°C for 10 min. Products were sequenced in both directions at the Australian Genome Research Facility (AGRF) on an Applied

Biosystems 3730xl DNA sequencer using POP7 capillaries gel matrix. Individuals heterozygous for a single nucleotide polymorphism (SNP) were resolved by comparing forward and reverse sequences at variable sites, and individuals heterozygous for multiple SNPs and indels were resolved via cloning (n=8) using TOPO-TA cloning kits

(Invitrogen USA). Sequences were edited manually in Sequencher 4.5 (Gene Codes

Corp., Ann Arbor, MI, USA) and aligned using ClustalW in MEGA6 (Tamura et al.

2013). Unique sequences were submitted to GenBank (Accession Numbers KT447642-

KT447682).

Microsatellite analyses

Potential genotyping errors and spurious allele scores caused by large allele dropout or stuttering were assessed in the program Microchecker (van Oosterhout et al. 2004). The program FreeNa (Chapuis & Estoup 2007) was used to estimate null allele frequencies for each locus and spawning group. This program creates a dataset corrected for null alleles, and uses it to calculate global and pairwise FST values across all loci and for each locus. There was little difference between the corrected FST values and the uncorrected values (see Results), so the original dataset was used for the remaining analyses.

The presence of linkage disequilibrium between all pairs of loci in each spawning group, and departures from Hardy-Weinberg Equilibrium (HWE) among loci and spawning groups were tested using GENEPOP 4.2 (Raymond & Rousset 1995)

28

with sequential Bonferroni correction (Rice 1989) for multiple comparisons. Estimates of molecular diversity and population genetic structure were calculated in Arlequin

3.5.1.2 (Excoffier & Lischer 2010). Diversity indicies included gene diversity (HE) and the effective number of haplotypes/alleles (Ne). FST (Weir & Cockerham 1984) was calculated for each locus with 1000 permutations (α = 0.004 after Bonferonni correction; Rice 1989) to determine how many loci showed differences between the spawning groups. Standardized F’ST (Meirmans 2006) was calculated in GenALEx 6.5

(Peakall & Smouse 2006, 2012) with 1000 permutations, and ΦST (Excoffier et al.

1992) was calculated in Arlequin with 1000 permutations using the infinite allele model

(IAM). Loci were also checked for private alleles associated with spawning groups. To illustrate the patterns of genotypic similarities, Principal Coordinates Analysis (PCoA) was performed on a matrix of pairwise genetic distance (codom-genotypic) in

GenALEx 6.5 (Peakall & Smouse 2006, 2012).

The Bayesian clustering method of Pritchard et al. (2000) was implemented in the program STRUCTURE 2.3.3 to estimate the number of population clusters (K) in the microsatellite data, and to assign individuals to these clusters based on genotypic data. The first analysis, using the simulation method described by Evanno et al. (2005), included all individuals with no a priori information about spawning season. Ten independent runs were performed for each value of K (1-5) using the admixture model with correlated allele frequencies with a burn in of 100,000 followed by 106 MCMC iterations. A subsequent STRUCTURE analysis was performed on the autumn- spawning group to test for sub-structure in this group (but was not done on the spring- spawning group due to the small sample size of this group).

Once the optimal number of populations had been estimated (the result of which was K=2), STRUCTURE was run again using the ‘Locprior’ model to assign

29

individuals to each cluster based on genotypic data. The ‘Locprior’ model is an extension of the basic STRUCTURE model, that has been proposed for data sets with low information content (i.e. relatively few loci or individuals; Hubisz et al. 2009); hence this model was used due to the small sample size of the spring-spawning group.

This model makes use of sampling information (i.e. spawning season) by placing a higher prior weight on clustering outcomes when they are correlated with sampling information, but it does not uncover structure where none exists (Hubisz et al. 2009).

STRUCTURE was also used to identify any potential hybrid individuals by identifying individuals with parental or grandparental ancestry from the other population (Pritchard et al. 2010). STRUCTURE was run using ‘Usepopinfo’ to assign individuals to each spawning group and the ‘Gensback’ option to test whether each individual had an ancestor in the preceding two generations (i.e. parent or grandparent).

The output from this mode also includes posterior probabilities that each individual is correctly assigned to the given population, and this was used as a check that each individual had been assigned to the correct population.

DNA sequence analyses

To determine the genealogical relationships between haplotypes of each sequenced marker, maximum parsimony haplotype networks were created in NETWORK 4.6.1.0.

(Fluxus Technology Ltd) using the median-joining algorithm. The Bayesian Information

Criterion scores calculated in MEGA6 (Tamura et al. 2013) were used to select the best fitting models of sequence evolution, which were used to calculate evolutionary distances between and within spawning groups in MEGA6. Measures of genetic diversity and differentiation, including gene diversity (HE), effective number of

30

alleles/haplotypes (Ne), FST (Weir & Cockerham 1984) and ΦST (Excoffier et al. 1992) with 1000 permutations were calculated in Arlequin 3.5.1.2 (Excoffier & Lischer 2010).

To provide a more direct historical context for the evolution of the A. samoensis spawning groups, phylogenetic analyses were conducted with a range of other Acropora species for PaxC and CR (but not Calmodulin due to a dearth of available sequences) from sequences downloaded from GenBank. Phylogenetic relationships among unique haplotypes were estimated using maximum likelihood methods (with bootstrap values conducted for 1000 replicates) implemented in MEGA6

(Tamura et al. 2013) for each gene separately. The Western Australian PaxC sequences contained two indels (8bp and 43bp), which were coded as single base changes as suggested by Simmons & Ochoterena (2000). The model of best fit for PaxC was the

Tamura 3 parameter model with Gamma distribution, and the alignment was trimmed to

438 bp. The model of best fit for the CR was the HKY model, and the alignment was trimmed to 1,094 bp. In a preliminary analysis, 27 species of Acropora were included, however, the results showed that many of these species provided little context for the evolutionary history of the spawning groups, so the final analysis included species only from the A. humilis group (to which A. samoensis belongs; Wallace 1999) and species that were informative in the preliminary analysis (i.e. shown to be closely related to A. samoensis). A. tenuis and A. intermedia were used as outgroups.

To assess whether selection has affected the loci examined, Tajima’s D

(1989) and Fu & Li’s F* test (Fu & Li 1993) of neutrality were performed in DnaSP

(Librado & Rozas 2009). These tests assume no recombination within loci, so each locus was examined for recombination using IMgc (Woerner et al. 2007), and where recombination was detected (in PaxC and Calmodulin), new files were generated with the largest non-recombining block of DNA sequence which was used to conduct the

31

neutrality tests. Additionally, because the PaxC intron was suspected of being under selection, in a subset of samples (n=10) a portion of the coding region of the PaxC gene was sequenced (GenBank Accession Numbers KT582778-KT582782; methods described in Appendix 3.3) in order to calculate the ratio of nonsynonymous substitutions (dN) to synonymous substitutions (dS).

Combined analyses

To determine whether the genetic variation shared between the spring and autumn spawners is simply a remnant of variation in the common ancestor or if it is due to ongoing gene exchange, Isolation with Migration (IM) coalescent analyses were conducted in IMa2 (Hey & Nielsen 2007). IMa2 estimates a posterior probability distribution for multiple demographic parameters, including divergence time, migration rate, and effective population sizes of two current populations and an ancestral population. The parameters are scaled by the neutral mutation rate and converted to a population migration rate, and time since divergence. The program was run with a total of 9 loci; 7 microsatellites (WGS112, EST016, EST254, EST063, EST097, EST181,

EST196) were run under the SMM, the mitochondrial control region was run under the

HKY model, and Calmodulin was run under the infinite sites model (using the largest non-recombining block of DNA sequence generated in IMgc as above). PaxC was excluded from the analysis because it potentially violates the model assumption of selective neutrality. Several preliminary runs were conducted to optimize model settings and prior parameter distributions, to assess mixing and adjust burn-in periods. Once all settings had been optimized, the final M-mode runs consisted of 80 chains with a geometric heating scheme of ha = 0.99 and hb = 0.75, priors q = 9.95, t = 5.98, m= 2.01, using the J2 model and a generation time of 5 years. Mutation rates for the CR and

32

Calmodulin were estimated following the rates published (% per million years) in Chen et al. (2009). Final simulations were run for six million steps following a burn-in of 1.7 million steps. Each M-mode analysis was repeated three times using different random seed numbers to assure convergence.

Results

Spawning seasonality

The combination of in situ observations and dissections of tagged colonies of Acropora samoensis identified distinct spring and autumn reproductive cohorts. Reproductive patterns were consistent in each colony (autumn or spring spawner) over the two years of this study. In situ observations in November 2010 showed that one cohort had mature, pigmented eggs (indicating spawning was imminent), while at the same time the other cohort had no visible eggs. Dissections showed that one cohort contained large, mature eggs and testes (500-529 µm, Table 3.1), while the other cohort had small, immature eggs (Table 3.1), indicating spawning was not imminent in the second cohort.

In mid-January 2011, one group had no visible eggs and the other had white eggs.

Dissections confirmed that all colonies that had contained large eggs in November 2010 no longer had eggs, indicating that spawning had indeed occurred in this cohort. In the other cohort, eggs were of medium size (200-314 µm, Table 3.1), and testes were present, indicating that spawning was likely to occur within 4-8 weeks of sampling (i.e. around February/March 2011). One colony (LOW_26) that had small, immature eggs in

November 2010 (suggesting it was an autumn spawner) did not appear to have eggs in

January 2011, and was therefore unlikely to have spawned in February/March 2011.

Nevertheless, this colony had small, immature eggs again the following year in October

33

confirming it was not a spring spawner (Table 3.1). Consistent with this assessment, genotypic assignment from the microsatellites and PaxC placed it in the autumn spawning cohort (see below). Consistency between the spring samples in 2010 and 2011 applied to all individuals, whereby the same individuals either had large eggs and testes in both years (spring spawners) or in neither year (autumn spawners).

Microsatellite analyses

Microchecker did not detect large allele dropout or stuttering in the microsatellite data set, but found null alleles at six loci (063, 112, 245, 254, 098, 196). FreeNa indicated the null alleles had a negligible effect on the dataset (corrected FST = 0.186 95%, CI

0.076-0.326, uncorrected FST = 0.181, 95% CI 0.066-0.324). No pairs of loci were found to be in linkage disequilibrium in either spring or autumn groups following

Bonferroni correction. The spring-spawning group was in HWE but the autumn- spawning group was not. Significant departures from Hardy-Weinberg equilibrium were not detected at any loci in the spring-spawning group, but deficits of heterozygotes were detected at three loci in the autumn-spawners (WGS112, EST098, EST196; Table 3.2).

A Wahlund effect might explain the heterozygote deficits in the autumn-spawning group, however, there was no population structure detected in the autumn-spawning group in the STRUCTURE analysis (see below). Two individuals had the same genotype, indicating they could be clonemates, so one was removed from the final dataset. Allele frequencies are provided in Appendix 3.4.

Based on all 13 loci, significant genetic differentiation was evident between the spring and autumn-spawning groups (FST = 0.17, p < 0.0001; F’ST = 0.30, p < 0.0001;

ΦST = 0.21, p < 0.001). The PCoA illustrated two genetic clusters that correspond to the spring and autumn-spawning groups (Fig 3.1). The bar plots in STRUCTURE mirrored

34

this result showing two genetic clusters (Fig. 3.2), and clear peaks in the plot of ∆K and

LnP(K) indicated that the optimal number of genetic clusters was two (K=2). In the subsequent Evanno et al. (2005) simulation searching for substructure within the autumn cluster, no further substructure was found. The initial decline in log probability estimates for K indicated that the probable number of clusters was one, and when K > 1 the proportion of individuals assigned to each cluster was fairly even as expected when the population structure is not real (Pritchard et al. 2010).

Posterior probabilities for assignment tests in STRUCTURE showed that all individuals were correctly assigned to their reproductive cohort as determined from the reproductive assessments; for example there were no individuals that were classified as a ‘spring’ spawner by the reproductive assessments but an ‘autumn’ spawner by the genotypic data (or vice versa). Permutation tests showed that autumn- and spring- spawning groups were significantly different at six of the 13 loci (Table 3.2). Private alleles were found in both the spring and autumn groups, but were more common in the autumn-spawning group (31, compared with 4 in the spring spawners), which had a larger sample. Some “private” alleles were simply uncommon, with 7 occurring as singletons (Table 3.2). Strikingly, at the EST196 locus 11 out of 18 alleles were unique to the autumn-spawning group (Table 3.2; Appendix 3.4). Despite these differences, the two spawning groups shared the same most-common allele at all but four loci (EST016,

EST196, EST234, WGS112).

Geneological analyses

The three DNA sequence markers showed varying levels of differentiation between the reproductive cohorts. Sequences of the PaxC intron (657 bp) from 38 individuals revealed 23 alleles, and showed two divergent lineages/clades with no alleles shared

35

between the autumn- and spring-spawning cohorts (Fig. 3.3a). The two clades were distinguished by 6 fixed differences and two informative indels (151 and 43 bp), and the level of p-distance between the PaxC clades was 0.017 ± 0.005 SE (Table 3.3). This distinctness represented ΦST of 0.97 (p = 0.0001; Table 3.4), indicating that almost all of the variation occurred between seasonal groups (Table 3.4).

A noteworthy exception detected in three of the cloned individuals was one allele which had characteristics of both autumn and spring clades; of the 6 nucleotide differences between the autumn and spring spawners this allele had 3 of those from the spring clade, yet it had the same two indels as the autumn clade, and 12 unique

(different from autumn or spring) differences. All individuals that had this sequence

(n=3) were heterozygotes that also contained an allele from the autumn-spawning clade, and all of these individuals were autumn spawners. Except for the three aforementioned individuals, all other heterozygotes had alleles within clades.

Sequences of the Calmodulin intron (359 bp) from 34 individuals revealed 17 unique alleles, and AMOVA and the haplotype network showed that the reproductive cohorts were genetically differentiated (ΦST = 0.31, p = 0.0001; Table 4; Fig. 3.3b). The level of divergence between sequences was much lower than for PaxC (mean p distance

= 0.0062 ± 0.0026 SE; Table 3.3). Sequences of the mtDNA control region (1236 bp) from 36 individuals revealed seven unique haplotypes, with no differentiation between the autumn and spring spawners (ΦST = -0.01, p = 0.5; Table 3.4; Fig. 3.3c), and the level of divergence between sequences was low (mean p distance = 0.0003 ± 0.0002 SE;

Table 3.3). Haplotype diversity was also low, with 72% of individuals carrying a single haplotype (Fig. 3.3c).

36

Table 3.1. Results of coral reproduction surveys (of sequenced samples) from in situ assessments and dissections. +T = testes were present; dash (-) = fixation did not work; n/a = colony was not sampled

Field Oocyte Field Oocyte Field Oocyte Assess diameter Assess diameter Assess. diameter Inferred Nov. Nov. 2010 Jan. Jan 2011 Oct Oct 2011 spawning PaxC Colony ID 2010 (µm) 2011 (µm) 2011 (µm) season clade LOW_4 Pink 500 + T Absent Empty White 330 + T Spring A LOW_6 Pink 524 + T Absent Empty White 392 + T Spring A LOW_61 n/a n/a n/a n/a White 330 + T Spring A LOW_67 n/a n/a n/a n/a White 346 + T Spring A LOW3_11 Pink 503 + T Absent Empty White 346 + T Spring A LOW3_13 Pink - Absent Empty White 346 + T Spring A LOW3_15 Pink 529 + T Absent Empty White 346 + T Spring A LOW3_17 Pink 529 + T Absent Empty Pink 291 + T Spring A LOW3_18 Pink 529 + T Absent Empty Pink 465 + T Spring A LOW3_3 Pink 524 + T Absent Empty Pink 387 + T Spring A LOW3_4 Pink 503 + T Absent Empty Pink 387 + T Spring A LOW3_5 Pink 500 + T Absent Empty White 387 + T Spring A LOW3_7 Pink 500 + T Absent Empty White 315 + T Spring A LOW3_8 Pink 503 + T Absent Empty White 387 + T Spring A LOW3_G42 n/a n/a n/a n/a White 387 + T Spring A LOW_19 Absent 100 White 314 + T Absent 50 Autumn A/B HET LOW_3 Absent 84 White 310 + T Absent 50 Autumn A/B HET LOW3_20 Absent 71 White 200 + T Absent 74 Autumn A/B HET LOW_1 Absent 141 White 266 + T Absent 50 Autumn B LOW_10 Absent 138 White 305 + T Absent 74 Autumn B LOW_20 Absent 100 White 350 + T Absent 84 Autumn B LOW_23 Absent 85 White 266 + T Absent Empty Autumn B LOW_24 Absent 100 White 314 + T Absent Empty Autumn B LOW_26 Absent 98 Absent Empty Absent 84 Autumn B LOW_28 Absent 100 White 314 + T Absent 50 Autumn B LOW_32 Absent 138 White - Absent 74 Autumn B LOW_33 Absent 90 White 314 + T Absent 60 Autumn B LOW_62 n/a n/a n/a n/a Absent 74 Autumn B LOW_66 n/a n/a n/a n/a Absent 74 Autumn B LOW_68 n/a n/a n/a n/a Absent 50 Autumn B LOW_9 Absent 100 White 309 + T Absent 74 Autumn B LOW3_10 Absent 100 White 309 + T Absent 74 Autumn B LOW3_14 Absent 132 White 309 + T Absent 50 Autumn B LOW3_16 Absent 100 White 266 + T Absent 50 Autumn B LOW3_19 Absent 141 White 300 + T Absent 74 Autumn B LOW3_2 Absent 132 White 314 + T Absent 74 Autumn B LOW3_6 Absent 100 White 266 + T Absent 50 Autumn B LOW3_9 Absent 84 White 266 + T Absent 50 Autumn B

37

Table 3.2. Measures of genetic diversity and variation in spring- and autumn-spawning cohorts: number of alleles/haplotypes (n), number of private alleles (priv), expected heterozygosity (HE), effective number of alleles/haplotypes (Ne) and FST (Weir and Cockerham 1984). Private alleles in bold indicate that at least one had a frequency of > 5%. * = significant results of permutation tests (p < 0.0001); ŧ = significant deviations from HWE (p < 0.01).

Spring Autumn FST

Locus n S priv. HE Ne n A priv. HE Ne Microsatellites EST016 4 0 0.61 2.56 6 2 0.58 2.39 *0.29 EST032 2 0 0.27 1.37 3 1 0.04 1.05 *0.28 EST063 3 1 0.21 1.28 5 3 0.41 1.72 0.03 EST097 3 0 0.53 2.14 4 1 0.63 2.70 *0.11 EST098 4 1 0.61 2.60 5 2 ŧ 0.66 2.92 -0.009 EST149 1 0 0.00 1.00 2 1 0.03 1.03 -0.15 EST181 4 0 0.72 3.62 6 2 0.46 1.86 *0.18 EST196 7 0 0.88 7.87 18 11 ŧ 0.92 12.20 0.02 EST245 3 1 0.16 1.19 4 2 0.23 1.39 -0.01 EST254 2 0 0.41 1.69 2 0 0.11 1.13 *0.65 WGS112 8 1 0.80 5.21 11 4 ŧ 0.70 3.30 *0.22 WGS153 2 0 0.08 1.08 2 0 0.04 1.05 -0.19 WGS211 1 0 0.00 1.00 3 2 0.20 1.26 0.04 PaxC 12 12 0.90 10.00 3 3 0.51 2.04 *0.30 Calmodulin 3 0 0.32 1.47 15 14 0.97 33.33 *0.26 mtDNA 4 0 0.65 2.86 5 0 0.35 1.54 0.06

Fig 3.1. Scores on the first two principal co-ordinates of pairwise genetic distance of microsatellite loci between autumn and spring spawners. The percentage of variation explained by the first three axes were 13%, 9% and 8% respectively. 38

Spring Autumn

Fig 3.2. Bayesian population assignment from the software program STRUCTURE when K=2. Cluster 1 = spring spawners, cluster 2 = autumn spawners. Each column represents a single individual, and the proportions of each individual’s genome that come from each cluster are shown by different colours.

Fig 3.3. Haplotype networks estimated for (a) PaxC (b) Calmodulin and (c) mtDNA control region in NETWORK using the median joining algorithm. Size of circles are proportional to number of individuals in that haplotype. Links represent character differences, and small colourless ovals are ancestral sequences required to connect existing sequences. Dark shading = autumn spawners, light shading = spring spawners.

39

Table 3.3. Average pairwise sequence distances (± SE) within and between spawning groups in mtDNA CR, PaxC and Calmodulin, and Tajima’s D and Fu & Li’s F* neutrality test values; significant values are indicated by * (p < 0.05).

bp Within spring Within autumn Between groups Tajima’s D Fu & Li’s F* mtDNA 1236 0.0002 ± 0.0002 0.0003 ± 0.0002 0.0003 ± 0.0002 -1.25 -0.86

PaxC 515 0.0047 ± 0.0017 0.0018 ± 0.0010 0.1268 ± 0.0197 2.32* 1.81*

Calmodulin 360 0.0023 ± 0.0014 0.0067 ± 0.0025 0.0062 ± 0.0026 -0.87 -0.27

Table 3.4. Analysis of Molecular Variance (AMOVA) from distance-based pairwise differences for all loci.

Source of variation DF SS Variance % variation ΦST p component Microsats Among populations 1 23.1 0.47 21.1 0.21 0.0001 Within populations 154 266.5 1.73 78.8 Total 155 289.6 2.20 PaxC Among populations 1 590.7 29.8 96.7 0.97 0.0001 Within populations 38 38.2 1.0 3.2 Total 39 629.0 30.8 Calmodulin Among populations 1 7.8 0.44 31.1 0.31 0.0001 Within populations 32 31.2 0.98 68.9 Total 33 39.0 1.42 mtDNA Among populations 1 0.47 -0.01 -1.35 -0.01 0.50 Within populations 34 20.8 0.60 101.3 Total 35 21.3

Skeletal examination of the reproductive cohorts revealed that the spring- spawning cohort had flared corallites more similar in appearance to Acropora digitifera than A. samoensis, raising the question of whether the spring-spawners were actually A. digitifera. Phylogenetic analysis of the PaxC intron among the A. humilis group indicates that the spring spawners are indeed closely related A. digitifera (Fig. 3.4a), however, phylogenetic analysis of the Control Region, suggests that the spring- spawning group is not A. digitifera (Fig. 3.4b). This also agrees with observed 40

differences in colony morphology and habitat preference between the spring-spawners and A. digitifera (author’s observations).

The evolutionary history of the spring- and autumn-spawning groups was somewhat difficult to interpret, due to the discordance between the PaxC and CR phylogenetic trees. Phylogenetic analysis of PaxC showed that the autumn spawners of

A. samoensis formed a clade with A. florida while the spring spawners were more closely related to A. digitifera (Fig. 3.4a). Phylogenetic analysis of the CR showed that the autumn and spring spawners combined formed a clade with A. aspera, A. florida, A. sarmentosa, A. humilis and A. gemmifera, and a subclade with A. florida, while the other two members of the A. humilis group, A. digitifera and A. monticulosa, were distinct from the other species (Fig. 3.4b).

A subset of individuals was sequenced for the PaxC protein coding region, and the primers amplified an 882 bp sequence that included a 429 bp-segment of the coding region (including part of the ‘paired domain’) and a 453 bp-segment of an adjacent intron. The 429 bp segment of the coding region was highly conserved and did not differ between the spawning groups, nor did it differ from the published sequence of another species (Acropora millepora Accession number AF053459). However the adjacent intron (a different intron to the previously mentioned PaxC intron) mirrored the other PaxC intron results, revealing two clades with mean p-distance = 0.08 which were separated by 35 point mutations and three indels.

Tests for selection

Tajima’s D and Fu & Li’s F* tests of neutrality were not significant for Calmodulin or the CR but were significant for PaxC (Table 3.3). The positive values of the PaxC test

(Table 3.3) suggest PaxC has been the target of positive selection. In addition, in the

41

second intron sequenced (adjacent to the paired domain, described above) Tajima’s D and Fu & Li’s F* neutrality tests were also significant (2.13 and 1.68 respectively, p <

0.05).

Fig 3.4. Acropora species maximum-likelihood phylogenetic trees for (a) PaxC intron and (b) mtDNA CR. Codes after species names are GenBank Accession numbers. A. samoensis spring and autumn spawners from this study are shown in bold. Bootstrap values from 1000 replicates are shown above branches.

Hybridization and time since divergence

The STRUCTURE analyses for the identification of hybrid individuals found that the highest probability of any individual having a parent from the other spawning group was 0.13, and the highest probability of an individual having a grandparent from the other spawning group was 0.18.

42

The IMa2 analyses confirmed that shared polymorphism of the microsatellites, CR and Calmodulin between the spring and autumn spawners was due to gene flow between the groups and not due to ancestral polymorphism, as the goodness of fit test in IMa2 rejected the null model of zero migration (p < 0.001).

Repeat runs of the IMa2 program revealed marginal posterior probability distribution curves of migration rates with a peak at 0.8 (2Nm = 0.6 migrant gene copies per generation).

While the coalescent analyses in IMa2 consistently revealed a migration rate of 0.8, the analyses were unable to provide a reliable estimate of divergence time. This could be due to either high migration rates obscuring divergence time, or lack of information in the data (J. Hey pers. comm.), and since migration estimates are low, it is assumed that the small sample size of the spring population (n=13) accounts for this lack of resolution (small sample sizes will affect some parameters and not others, so migration rate estimates should still be valid; J Hey pers. comm.).

Discussion

Cryptic speciation

Reproductive assessments confirmed the existence of two reproductive cohorts in the population of Acropora samoensis, one that spawned in autumn and one that spawned in spring. The combined results from the analyses of microsatellites, PaxC and

Calmodulin indicate that the two cohorts are highly genetically differentiated. The

STRUCTURE analyses and PCoA separated the spawning groups into two clearly defined genetic clusters, and the AMOVA showed significant and substantial genetic differentiation between the groups in three of four markers. Most remarkably of all,

43

PaxC showed two genetic lineages associated with spawning time that were separated by an evolutionary distance of 1.7%. Following reproductive and genetic identification of the colonies, small differences in radial corallite shape between spring- and autumn- spawners were observed in skeletal samples. Hence, the combined evidence of genetic, morphological and reproductive differences between the autumn and spring spawners of

Acropora samoensis observed in this study indicates that these reproductive cohorts represent two cryptic species (cryptic species defined as two or more species that have been classified as a single nominal species because they are at least superficially morphologically indistinguishable; Bickford et al. 2007).

An explosion of cryptic species discoveries in scleractinian corals indicates cryptic speciation is widespread across many genera (Knowlton et al. 1992; Flot et al.

2011; Ladner & Palumbi 2012; Schmidt-Roach et al. 2014; Warner et al. 2015), and here is another example of the molecular identification of cryptic species. In this case cryptic species are associated with asynchronous reproduction, and this may well be a common mechanism in other cases (e.g. Ladner & Palumbi 2012), and should be considered more widely in studies of population genetic structure in corals.

Asynchronous reproduction in corals is increasingly recognized, and biannual spawning is widespread across Western Australia (Rosser 2013), Indonesia (Baird et al. 2009a), the Great Barrier Reef (Stobart 1994; Wolstenholme 2004) and some areas of the south

Pacific (Mildner 1991), and it is crucial that possible cryptic species within populations are considered in experimental design, analysis and interpretation of gene flow across these regions (Warner et al. 2015).

Although the mitochondrial CR did not show any differentiation between the autumn and spring spawners, this is not unusual, because mitochondrial differentiation is often absent between both recognized and cryptic coral species

44

(Wolstenholme 2004; Flot et al. 2008; Ladner & Palumbi 2012), mainly due to the slowly evolving mitochondrial genome in corals (Shearer et al. 2002; Hellberg 2006) and introgression between species (van Oppen et al. 2001; Vollmer & Palumbi 2002). A previous phylogenetic study of the Acropora humilis group based on molecular, morphometric and reproductive evidence showed that reproductive timing provided the greatest level of taxonomic resolution between species in this group (Wolstenholme

2003).

In the cryptic species reported here, reproductive isolation between autumn and spring spawners has not been absolute, as indicated by low levels of introgression detected by the IMa2 analysis. Whether this is ongoing is unclear, as the STRUCTURE analysis did not convincingly detect any first- or second-generation hybrids, although this might simply be an artefact of small sample size. The reproductive assessments of

A. samoensis in this study and previously (Rosser & Gilmour 2008) found that spawning season was consistent over a number of years, yet for introgression to occur the timing of reproduction must occasionally be switched in a colony to allow gene flow between the seasonal reproductive cohorts. The high degree of genetic differentiation between the spawning groups and low level of introgression detected here suggests that reproductive times are largely heritable, but are occasionally influenced by environmental or physiological effects.

Evolutionary history

Interpreting the history of the spring- and autumn-spawners in A. samoensis is complicated by the discordance in the species trees between PaxC and the CR, indicating either a very old or a recent origin of the cryptic species, or introgression . A rudimentary estimate of the divergence time between the two highly divergent lineages

45

in the PaxC intron (based on the divergence rate of the PaxC intron in Acropora of

0.138-0.550 % per Ma; Chen et al. 2009) is between 32 and 8 mya. The Acropora humilis group is an old species group with the earliest recorded fossils from the Eocene, around 41 mya (Wallace & Rosen 2006). Such an ancient speciation is inconsistent with the low levels of sequence divergence in Calmodulin and the CR, and the many shared alleles among the microsatellites. Thus, the most parsimonious interpretation of the divergent PaxC lineages is that, rather than indicating ancient cryptic species, the PaxC lineages date to an old, ancestral polymorphism that has been retained in the evolution of A. samoensis. In addition, the occurrence of introgression at some point in the past, perhaps during the initial contact of two allopatrically diverged populations, could also explain the homogeneity in the mitochondrial CR and the discordance between the CR and PaxC gene trees.

In the PaxC intron the average level of sequence divergence between the autumn- and spring-spawning groups (1.7%) is over three times the level of divergence in Calmodulin (0.5%) and the CR (0.03%). The retention of these divergent PaxC clades in the lineage of A. samoensis suggests selection on this gene, which is supported by the results of the neutrality tests. In general, regions of the genome that are under selection must be of functional importance (Nielson 2005), and the function of Pax genes suggest possible mechanisms of selection on this gene in corals. The PaxC gene in cnidarians is a precursor of Pax6 (Miller et al. 2000; de Jong 2005), which plays a central role in eye development in vertebrates and Drosophila (Hill et al. 1991; Quiring et al. 1994). While not all cnidarians have eyes, all sense light, and gametogenesis and spawning are cued by seasonal, lunar and daily changes in light intensity and spectral quality (Hunter 1988; Reitzel et al. 2013). Thus, it may well be that the PaxC gene plays a role in spawning time (alongside other photoreceptor genes e.g. Shoguchi et al.

46

2013). Sequencing of a small part of the protein (31%) showed that it was highly conserved among the spawning groups and among other species, which may indicate that divergent selection is acting upon aspects of the regulation of PaxC, rather than the protein itself. Genomic regions that contain quantitative trait loci quickly diverge under selection and become resistant to gene exchange (Turner et al. 2005; Via 2009), which would explain the much higher level of divergence in PaxC compared to Calmodulin.

Further analysis of other genes (e.g. cryptochromes, which are thought to mediate coral spawning; Levy et al. 2007) may find a set of linked loci diverged among spring and autumn spawners that are under joint selection. Although speculative, this possibility indicates that PaxC is a very interesting gene in cnidarians, and the potential for it to influence spawning in corals is an exciting avenue for future research.

Divergence between the PaxC clades could have begun in sympatry or allopatry during the Miocene. Allopatric divergence among populations is the most common explanation of the origin of large differences in breeding time (Coyne & Orr

2004), but allochronic divergence (a mode of sympatric divergence resulting from a phenological shift; Alexander & Bigelow 1960) is considered responsible for genetic diversification in other broadcast-spawning marine invertebrates (Bird et al. 2011).

Regardless of whether divergence began in sympatry or allopatry, it appears that divergent selection has resulted in genetically-based differences in spawning season.

This study highlights the importance of incorporating observations of reproductive timing into studies of genetic structure in corals, because at both the population level and the species level, differences in reproductive timing play a significant role in their evolution.

47

48

Chapter 4

Asynchronous spawning and demographic history shape genetic differentiation among populations of the hard coral

Acropora tenuis in Western Australia

This chapter has been published in Molecular Phylogenetics and Evolution:

Rosser NL (2016) Demographic history and asynchronous spawning shape genetic differentiation among populations of the hard coral Acropora tenuis in Western

Australia. Molecular Phylogenetics and Evolution 98, 89-96.

49

Abstract

Reproductive isolation can facilitate genetic subdivision between populations, and a significant facilitator of isolation is reproductive asynchrony, such as seen in broadcast- spawning corals. However, the factors that shape genetic variation in marine systems are complex and ambiguous, and ecological genetic structure may be influenced by the overriding signature of past demographic events. Here, the relative roles of the timing of reproduction and historical geography on the partitioning of genetic variation were examined in seven populations of the broadcast-spawning coral Acropora tenuis over

12° of latitude. The analysis of multiple loci (mitochondrial control region, two nuclear introns and six microsatellites) revealed significant genetic division between the most northern reef and all other reefs, suggesting that WA reefs were re-colonized from two different sources after the Pleistocene glaciation. Accompanying this pattern was significant genetic differentiation associated with different breeding seasons (spring and autumn), which was greatest in the PaxC intron, in which there were two divergent lineages (ΦST = 0.98). This is the second study to find divergent clades of PaxC associated with spring and autumn spawners, strengthening the suggestion of some selective connection to timing of reproduction in corals. This study reiterates the need to incorporate reproductive timing into population genetic studies of corals because it facilitates genetic differentiation; however, careful analysis of population genetic data is required to separate ecological and evolutionary processes.

Introduction

A fundamental component of population genetics is the level of connectivity among populations of a species, because local adaptation and eventually speciation

50

depend upon patterns of gene flow. In populations of broadcast-spawners a potential facilitator of genetic subdivision is asynchronous reproduction, as opportunities for fertilization and gene exchange are extremely low between individuals that spawn more than several hours apart (Levitan et al. 2011), so differenes in reproductive timing can enforce assortative mating and genetic divergence (Knowlton et al. 1997; Dai et al.

2000; Fukami et al. 2003; Hendry & Day 2005; Bird et al. 2011; Binks et al. 2012). In contrast to most other regions, north-western Australian and Indonesian reefs have two major coral spawning events each year, one in autumn and one in spring (Rosser &

Gilmour 2008; Permata et al. 2012). A recent study of sympatric populations of autumn- and spring-spawning cohorts of Acropora samoensis in Western Australia showed that the reproductive cohorts were genetically differentiated on an ecological scale (F’ST = 0.3), and had divergent lineages of the phylogenetic marker PaxC which were not present in other DNA sequences markers, raising the possibility that PaxC may be under selection (Rosser 2015).

The factors that shape genetic variation in marine systems, however, are complex and ambiguous, and one difficulty in interpreting patterns of genetic subdivision is that ecological patterns may be confounded by the overriding signature of past demographic events, so it is necessary to separate population history from population structure. This is particularly relevant in the marine realm, where high larval dispersal has the potential to connect populations over large distances (Palumbi 1992;

Ayre & Hughes 2000; Johnson & Black 2006a), and geographic distance alone can be a poor predictor of genetic structure (Johnson & Black 2006b).

Climatic oscillations in glacial and non-glacial cycles cause expansions and retractions of species’ ranges, involving local extinction, migration, drift and adaptation, and these processes leave different genetic signatures on populations (Hewitt 1996). The

51

coral reefs of Western Australia have a long history of expansion and contraction of their geographic range in response to oscillating glacial cycles. The north-west coast is characterized by a series of discontinuous reefs, which occur on the continental slope, on the continental shelf edge, and along the coastline. The offshore reef systems consist of atoll-like reefs on the continental slope that rise from deep-ramp settings (e.g. Scott

Reef and the Rowley Shoals), as well as shallower reefs perched on the edge of the continental shelf (e.g. Ashmore Reef), while the inshore/coastal reefs occur along the modern-day coastline and around inshore islands. At the height of the Last Glacial

Maximum ~ 18,000 years ago when sea level was -120 m, the present-day coastal reefs, including the Kimberley coast, the Montebello Islands, Dampier and Ningaloo Reef, were on dry land (Yokoyama et al. 2001). These coastal reefs were recolonized in the

Holocene transgression, but where they were recolonized from is uncertain; while the offshore atolls of Scott Reef and the Rowley Shoals would have existed during the

LGM, whether refuge coral populations survived lower temperatures and contracted habitats during the LGM has been widely debated (Wilson 2013).

To investigate the roles of historical geography and the timing of reproduction on the partitioning of genetic variation, this study examines phylogeographic and population genetic variation over a large geographic range in the scleractinian coral

Acropora tenuis, which reproduces in spring and autumn in Western Australia.

Methods

Study sites and spawning patterns of A. tenuis

Genetic samples of Acropora tenuis were collected from seven locations spanning

1600km and 12° of latitude, from Ashmore Reef to Ningaloo Reef, over which there is a

52

latitudinal gradient of spawning time. Spawning is in spring at the most northern location of Ashmore Reef (Rosser 2013), in autumn at the most southern locations of

Dampier and Ningaloo Reef (Baird et al. 2011; Rosser 2013), and in both spring and autumn at the geographically intermediate Scott Reef (Gilmour et al. 2009). Tissue samples were collected by snapping off small branches (1-3 cm) from individual colonies, which were sampled at least 5m apart to reduce the likelihood of collecting clonemates. Samples were preserved in 95% ethanol, which was replaced after 24 hours and again after one week. Sample sizes at each location ranged from 3 to 49 (Table 4.1).

Colonies of A. tenuis at Ashmore Reef were visibly different from other locations and difficult to identify, so skeletal samples from each colony were sent to the Museum of

Tropical Queensland for species verification.

DNA extraction, PCR and DNA sequencing

DNA sequencing of a mitochondrial gene and two nuclear genes was used to explore phylogeographic patterns. DNA was extracted from branch tips using DNeasy DNA extraction kits for animal tissue (Qiagen, USA) according to the manufacturer’s instructions. The mtDNA control region, the nuclear PaxC intron and a microsatellite flanking region were amplified in polymerase chain reactions (PCR). The mtDNA control region was amplified using primers rns (5’-GGTTTCTAATACCTCCGAGG-

3’) and Cox3 (5’- TACATAACACTGCCCACAGT-3’) after van Oppen et al. (2001).

The PaxC intron was amplified using the primers PaxC_intron-FP1 (5’-

TCCAGAGCAGTTAGAGATGCTGG-3’) and PaxC_intron-RP1 (5’-

GGCGATTTGAGAACCAAACCTGTA-3’) after van Oppen et al. (2000).

Microsatellite flanking regions have been shown to be phylogenetically informative

(Chatrou et al. 2009), so primers were developed for the flanking region of one of the

53

microsatellites used in this study, Amil02_018 (sequence downloaded from Genbank

Accession No.: EF989161.1). The software OLIGO was used to develop the primers

FL-FP (5’-GGAAAGCCCTCCTTAGGTGAT-3’) and FL-RP (5’-

CAAATTGAAGGCAAATGTCGG-3’). PCR reactions contained 1µL MgCl2 (50nM),

1.2 µL dNTPs (2.5nM), 0.2 µL platinum Taq, 2.5 µL 10 x PCR buffer, 1 µL each of the forward and reverse primers, 2 µL of DNA, and 17.1 µL dH2O in a 25 µL reaction.

Positive (known DNA sample) and negative (no DNA) controls were included with each reaction. Thermocycling profiles consisted of an initial denaturation step of 95°C for 3 min, followed by 35 cycles of 94°C for 30 sec, 50°C for 1 min and 72°C for 1 min and a final cycle of 72°C for 10 min. Products were sequenced in both directions at the

Australian Genome Research Facility in Perth. Individuals heterozygous for a single nucleotide polymorphism (SNP) were resolved by comparing forward and reverse sequences at variable sites. Individuals with multiple SNPs, or individuals that were heterozygous for two informative indels in the PaxC gene, were resolved via cloning using TOPO-TA cloning kits (Invitrogen, USA) and standard cloning procedures.

Sequences were edited manually in Sequencher v 4.5 (Gene Codes Corp., Ann Arbor,

MI, USA) and aligned using ClustalW in MEGA version 5 (Tamura et al. 2011).

Microsatellite genotyping

New samples from Ashmore Reef and the Montebello Islands were combined with a subset of published microsatellite data (Underwood 2009a; Underwood 2009b) to test for population genetic structure. One population from each of the four geographic locations in Underwood’s dataset (Scott Reef, Rowley Shoals, Dampier and Ningaloo) was randomly selected from his complete dataset for inclusion in this study (Table 4.1).

To standardize the microsatellite data from Ashmore Reef and the Montebello Islands

54

with Underwood’s (2009a) data, 10 tissue samples that were amplified and scored by

Underwood (2009a) were also amplified and scored in the present study (using

GeneMarker v 1.9; SoftGenetics LLC), to provide direct calibration. Underwood

(2009b) amplified seven loci, all of which were repeated in this study (Appendix 4.1), but one locus (Amil2_006) could not be successfully calibrated so was excluded from this study.

Table 4.1. Location information and number of individuals used in analyses of the Control Region, PaxC and microsatellites. * denotes microsatellite data used from Underwood (2009a,b). Location CR PaxC Flank Msats Ashmore Reef 12 11 10 40 Scott Reef 10 9 9 *47 Rowley Shoals 8 11 9 *49 Kimberley 7 4 4 - Dampier 10 10 9 *42 Montebello Is 12 11 8 27 Ningaloo 11 9 10 *48 Total 66 65 58 253

Microsatellites from Ashmore Reef and the Montebellos were amplified in 13 µL containing 11µL of Platinum PCR mix (Invitrogen: 22 U/mL complexed recombinant

Taq DNA polymerase with Platinum Taq antibody, 22 mM Tris-HCl, 55 mM KCl, i.65 mM MgCl2, 220 µm dGTP, 220 µm dATP, 220 µm dTTP , 220 µm dCTP), 1 µL DNA and 0.25 – 1 µL of 3.3 µmol florescent-tagged forward primer for each locus. PCR amplifications consisted of an initial denaturation step at 94 °C for 2 min followed by

30 cycles of 45 s at 94°C, 45 s at annealing temperature (50°C), 45 s at 72°C and finally

72°C for 5 min. Each individual was genotyped for six microsatellite loci. Any

55

individuals that failed to amplify at ≥5 loci were excluded from analyses (3% of individuals).

Phylogeographic analyses of DNA sequences

Maximum parsimony haplotype networks were created in NETWORK v 4.6.1.0.

(Fluxus Technology Ltd) using the median-joining algorithm. Each locus was examined for recombination using IMgc (Woerner et al. 2007), and where recombination was detected (in the microsatellite flanking region sequences, but not PaxC), new files were generated with the largest non-recombining block of DNA sequence, which was used to construct haplotype networks. In some individuals in which significant recombination was detected, IMgc was unable to generate a non-recombining block, so these individuals were excluded from all further analyses (n= 9, distributed across all sites).

Analysis of PaxC revealed two divergent clades (labeled A and B), and the frequency of clade A appeared to be associated with latitude, and hence with spawning season, so the Spearman rank correlation coefficient (rs) was used to test this association. To decouple PaxC-associated variation from geography-associated variation, a series of AMOVA tests was performed in Arlequin (Excoffier & Lischer

2010) to determine where the greatest level of genetic differentiation occurred. For this analysis individuals were grouped according to: (a) the main groups identified in the

STRUCTURE analysis of the microsatellites (see below), which corresponded to

“Ashmore” and “non-Ashmore” populations (ΦRT), (b) PaxC clades A and B (ΦCT), and

(c) individual populations (ΦST). Evolutionary distances between populations and clades were calculated in MEGA5 (Tamura et al. 2011) using the most appropriate model of

DNA evolution, as determined by the Bayesian Information Criterion. Diversity indices

56

including gene diversity (HE), nucleotide diversity (π) and number of effective alleles

(Neff) were calculated for each population using Arlequin (Excoffier & Lischer 2010).

Population genetic analyses of microsatellites

Linkage disequilibrium and departures from Hardy-Weinberg Equilibrium were tested within sites using FSTAT (with sequential Bonferroni correction applied; Rice

1989). The dataset was checked for individuals that had the same multi-locus genotypes, and replicates were removed. The program FreeNa was used to estimate the frequency of null alleles and to generate a dataset corrected for null alleles to determine pairwise

FST values (Weir & Cockerham 1984) across all loci and for each locus. The FreeNa dataset yielded similar levels of pairwise FST and overall FST as the original dataset

(uncorrected FST = 0.117, 95% CI 0.05-0.21; corrected FST = 0.114, 95% CI 0.06-0.20), so the original dataset was used for the remaining analyses. FST (Weir & Cockerham

1984) and standardized F’ST (Meirmans 2006) were calculated in GenAlEx 6.5 (Peakall

& Smouse 2012).

The Bayesian clustering method of Pritchard et al. (2000) was implemented in the program STRUCTURE v2.3.3 to estimate the number of population clusters (K) in the microsatellite data, using the simulation method described by Evanno et al. (2005).

For datasets with relatively few loci, the ‘LocPrior’ model uses the sampling locations, and places a higher prior weight on clustering outcomes when they are correlated with sampling locations. This helps to find subtle genetic structure, without tending to uncover structure where none exists (Hubisz et al. 2009). This model was used here, due to the low number of loci in the microsatellite dataset. Simulations were based on the admixture model with correlated allele frequencies, and sampling included 500,000 repetitions following a burn-in of 200,000. Simulations were performed for K=1-10

57

genotypic clusters, running five replicates for each scenario. Results were processed in the program STRUCTURE HARVESTER (Earl & vonHoldt 2012) to determine the most likely value of K. Results of the five STRUCTURE runs were merged with

CLUMPP (Jakobsson & Rosenberg 2007) and visualized with DISTRUCT (Rosenberg

2004). Estimates of K differed between plots of ∆K and mean L(K), suggesting hierarchical population structure was present, so subsequent STRUCTURE analyses were also conducted to test for substructure.

As with the DNA sequence analyses, a series of AMOVA tests was performed in

Arlequin (Excoffier & Lischer 2010) to determine where the greatest level of genetic differentiation occurred, and individuals were grouped in the same way (see above). In addition, a Principal Co-odinates Analysis (PCoA) was conducted in GenAlEx 6.5

(Peakall & Smouse 2012), separating individuals into their PaxC clades, to examine the relationships visually. The correlation between log-transformed pairwise FST and log- transformed geographic distance between populations was assessed using regression analyses in Arlequin, to test for patterns of isolation by distance (IBD), with significance assessed using a Mantel test (10,000 permutations). The same diversity indices mentioned above were calculated in Arlequin, and a two-way t-test comparing

HE (Appendix 4.2) between Ashmore Reef, Scott Reef and Rowley Shoals vs

Montebello, Dampier and Ningaloo was conducted in MS Excel.

Results

Phylogeographic structure

Sequences of the PaxC intron (509 bp) showed the greatest amount of genetic differentiation among the DNA sequence markers. Twelve unique haplotypes were detected, which formed two distinct clades (Fig. 4.1). The two clades were distinguished 58

by 15 fixed differences and two indels (4bp and 13bp), which co-segregated with the fixed SNPs. The average evolutionary divergence between the clades using the Tamura

3-parameter model was 0.014 SE ± 0.005 (maximum between haplotypes = 0.019 SE ±

0.006), and this distinctness was represented by ΦCT of 0.98. While most individuals were homozygous for either clade A or clade B, cloning revealed that six individuals were heterozygous for both clades (9% of individuals sequenced). These heterozygotes occurred at Ashmore Reef and the Kimberley.

The distribution of the PaxC clades varied with latitude (rs = 0.957, P < 0.01;

Fig. 4.2), paralleling variation of breeding seasons in A. tenuis. On Ashmore Reef, where spawning occurs in spring, clade A predominated, while at Dampier and

Ningaloo, where spawning occurs in autumn, clade B predominated, while at Scott

Reef, where spawning occurs in both seasons, both clades occurred (Fig. 4.2).

Sequences of the mtDNA control region (1077 bp) revealed 20 unique haplotypes, and the haplotype network had a ‘star-like’ form (Fig. 4.1). The central haplotype was most common, and was found in every population, while the majority of the derived haplotypes were singletons linked by just one mutational step. The average evolutionary divergence among all sequence pairs using the Tamura-Nei model was 0.0018 ± 0.0005

SE (maximum between haplotypes = 0.0047 ± 0.0019 SE). Sequences of the microsatellite flanking region (565 bp) revealed 15 unique haplotypes, also with no large gaps in the haplotype network (Fig. 4.1). The average evolutionary divergence among all sequence pairs using the Kimura 2-parameter model was 0.007 ± 0.0017

(maximum between haplotypes = 0.018 ± 0.005 SE).

59

CR FL Ashmore Reef

Scott Reef

Rowley Shoals

Kimberley Coast

Montebello Is.

Dampier

Ningaloo Reef

PaxC

1 5 10

40

CladeA CladeB

Fig 4.1. Haplotype networks for the mtDNA control region (CR), Flanking Region (FL) and PaxC intron (PaxC). Each circle represents a different haplotype, and its relative size is proportional to its frequency as indicated by circular diagram. Colours represent geographic locations. Small cross- hatched lines represent the number of additional mutational changes.

Fig 4.2. Map of sampling locations, with pie-charts showing the frequencies of PaxC clades A (white) and B (black) in each population of A. tenuis. The dotted line represents the 100m contour and the approximate location of the Australian coastline during the LGM. The bar graph shows the proportion of colonies (%) spawning in spring (grey) and autumn (black) in A. tenuis documented in previous studies (Gilmour et al. 2009; Baird et al. 2011; Rosser 2013). 60

Demographic structure

All populations had deficits of heterozygotes at the microsatellite loci (Appendix

4.2), which were not accompanied by linkage disequilibrium among loci. FreeNa indicated null alleles had a negligible effect on the dataset, suggesting that heterozygote deficits are more likely to be due to admixture or inbreeding. FST values indicated significant genetic structure in the microsatellites (FST = 0.11; F’ST = 0.23).

Plots of ∆K (Evanno et al. 2005) from STRUCTURE indicated that the most likely number of genetic clusters (K) estimated from the microsatellite loci was two.

Subsequent STRUCTURE analyses of the ‘Ashmore’ group found no further substructure, while in cluster analyses of the ‘other’ group the greatest ∆K occurred at K

= 4. The four genetic clusters identified in this ‘non-Ashmore’ sub-group were the same clusters that occurred in the whole data set at K = 4 (Fig. 4.3). At K = 2 the greatest level of genetic differentiation was between Ashmore Reef and all other localities (Fig.

4.3). Scott Reef showed a high level of admixture between the Ashmore cluster and the

‘other’ cluster. At K = 4 the genetic clusters were distributed between (i) Ashmore and

Scott Reefs (ii) Scott Reef and Rowley Shoals (iii) Ningaloo and the Montebello Is. and

(iv) Dampier (Fig. 4.3).

In line with the Bayesian clustering results, population pairwise F’ST values were greatest between Ashmore and other populations (Table 4.2). A Mantel test found no significant association between genetic and geographic distance in the microsatellite loci (r2 = 0.09, P = 0.6), ruling out IBD over the entire region.

A series of AMOVAs was conducted to determine where the greatest level of genetic variation occurred, to disentangle geographic genetic structure from potential structure associated with spawning season. In both the microsatellites and the flanking region sequences, genetic variation was highest between the STRUCTURE-identified 61

groups (“Ashmore” and “non-Ashmore” populations), followed by the PaxC-clade groups, and then by populations, although the difference was small for the microsatellites (0.20 compared with 0.17; Table 4.3). As expected in PaxC, genetic variation was higher between the clades than between STRUCTURE-identified groups

(Table 4.3). The control region was the least informative marker, and did not detect any significant variation at any level (Table 4.3). A PCoA conducted on the microsatellite loci also suggested that genetic variation was higher between the Ashmore and non-

Ashmore populations than between PaxC clades A and B, because the individuals with

Clade A from Scott Reef and Rowley Shoals did not cluster together with Ashmore

Reef, regardless of clade (Fig. 4.4).

Patterns of genetic diversity varied across markers, but generally, Ashmore

Reef, Scott Reef and Rowley Shoals had higher genetic diversity than the Montebello

Islands, Dampier and Ningaloo Reef (P = 0.01, 2 tailed t-test; Appendix 4.2).

Discussion

The combination of molecular markers used in this study provided insight into patterns of genetic variation on both an ecological and an evolutionary scale, and presented two key findings. First, all genetic markers (except CR, which was uninformative) revealed a phylogeographic break between Ashmore Reef and all other reefs. Second, there were two clades in the PaxC gene tree separated by an evolutionary distance of 1.4% that are not present in the other genetic markers, and these clades are geographically associated with timing of reproduction.

62

Fig. 4.3. Results of Bayesian cluster analysis of A. tenuis populations in Western Australia as identified by STRUCTURE showing K = 2 to K = 4 in descending order. Each individual is shown as a vertical bar and indicates the relative membership proportion to each genetic cluster (blue, yellow, pink and purple).

Fig. 4.4 Principal Co-ordinates Analysis (PCoA) of A. tenuis microsatellite loci. Individuals are grouped into (i) PaxC clades (clade A = large circles; clade B = small circles), and (ii) geographic region (Ashmore = white; non-Ashmore = black). Heterozygous individuals for clades A and B are represented by large, grey circles. 63

Table 4.2. Pairwise F ’ST values (Meirmans 2006) for microsatellite loci (below diagonal) and geographic distance (km) (above diagonal, shaded). All F ’ST values were significant (p < 0.01).

Ashmore Scott Rowley Montebello Dampier Ningaloo Ashmore 0 219 683 1150 1163 1667 Scott 0.27 0 468 898 975 1300 Rowley 0.49 0.07 0 435 525 893 Montebello 0.43 0.10 0.09 0 137 474 Dampier 0.58 0.22 0.15 0.07 0 369 Ningaloo 0.40 0.10 0.09 0.07 0.16 0

Table 4.3. Summary of results from Analysis of Molecular Variance (AMOVA) when variation was partitioned among groups (‘Ashmore’ and ‘non-Ashmore’ as per the Bayesian cluster analysis), among PaxC clades, and among populations. Tests were conducted in Arlequin on pairwise ΦST for the DNA sequence loci, and FST for the microsatellites. The highest values for each marker are shown in bold. Only significant results (P < 0.001) from 1000 permutations are presented (ns = non-significant).

Among groups Among PaxC clades Among populations

Locus (Φ/FRT ) (Φ/FCT ) (Φ/FST ) mtDNA CR ns ns ns PaxC 0.65 0.98 0.39 Flank 0.34 0.18 0.21 Microsats 0.20 0.17 0.12

Phylogeographic structure and reproductive timing

The Bayesian cluster analysis of the microsatellites, the pairwise FST values, and the AMOVAs of the microsatellites and the flanking region all showed that the greatest genetic split is between Ashmore Reef and all other populations, indicative of a phylogeographic break in this region. This pattern is a likely result of allopatric divergence in two glacial refugia during the Pleistocene, from which WA reefs were subsequently re-colonized in the Holocene transgression. At the height of the last

64

Pleistocene glaciation the wide continental shelf was exposed, the coastline was at the edge of the Sahul shelf along the 130 m isobaths (Yokoyama et al. 2001), and the present-day coastal reefs were on dry land (Fig 4.2). During this time there was possibly a glacial refugium along the convoluted Sahul shelf, from which Ashmore reef may have been re-colonized during the Holocene transgression, while the other possible source was Scott Reef. The higher genetic diversity of Scott Reef compared to the inshore populations suggests Scott Reef is an older population, in which more time has elapsed for mutations to accumulate (Hewitt 1996; Grant & Bowen 1998).

Alternatively, if the Sahul shelf refugium was south of Ashmore Reef, it could have recolonized Scott Reef and the other WA reefs, while Ashmore Reef was recolonized from a northern source, most likely Indonesia.

Given the discontinuous, stepping-stone nature of coral reefs in Western

Australia, and the fact that most coral recruitment occurs over 10s to 100s of km (Treml et al. 2008), a pattern of IBD might be expected. However, IBD was not detected, as illustrated by the much stronger genetic divergence between Ashmore and Scott Reefs

(F ’ST = 0.27, 219 km) compared to Scott Reef and Ningaloo Reef (F ’ST = 0.10, 1300 km), despite the much larger geographic distance between the latter pair. This finding further supports the view that WA reefs were re-colonized from two different sources in the Holocene transgression.

Phylogeographic breaks can provide a starting point for speciation (Hewitt 2000;

Avise 2004), and several studies have revealed cryptic species among marine taxa that have experienced population fragmentation during glacial periods (reviewed in Provan

& Bennett 2008). Differentiation between Ashmore Reef and the reefs south of Scott

Reef was substantial (F’ST of 0.40 to 0.58), and was accompanied by subtle differences in colony morphology (author’s observations), suggesting they may even be cryptic

65

species according to the ‘genotypic cluster’ definition of species (Mallet 1995). Cryptic species are increasingly being discovered in scleractinian corals (reviewed in Warner et al. 2015), and their detection is important for accurately estimating biodiversity and population connectivity. If cryptic diversity is uncharacterized, multiple species may be unwittingly pooled, which can result in the incorrect interpretation of population connectivity, gene flow, dispersal, biodiversity and ecological patterns (Ladner &

Palumbi 2012; Warner et al. 2015).

Surprisingly, gene flow in contemporary populations has not homogenized the population and erased this signature of colonization. A likely explanation is that reproductive isolation maintains the genetic difference between Ashmore Reef and the other WA reefs. Acropora tenuis spawns predominantly in spring at Ashmore Reef

(Rosser, 2013) and predominantly in autumn on other WA reefs (Gilmour et al., 2009;

Baird et al., 2011; Rosser, 2013). Because spawning time is largely heritable and consistent amoung years (Vize et al. 2005; Levitan et al. 2011; Rosser 2015) spawning in different seasons would inhibit gene flow between the populations and promote genetic divergence (e.g. Dai et al. 2000; Wolstenholme 2004; Rosser 2015).

In both the microsatellites and flanking region sequences the AMOVA tests indicated that genetic differentiation between Ashmore and non-Ashmore reefs was greater than the differentiation between PaxC clades. However, the absence of spawning data on specific individuals means I was unable to test the classification as a

‘spring’ or ‘autumn’ spawner based on PaxC clade, and some evidence suggests that it is not absolute. A previous study found evidence of introgression between spring and autumn-spawning cohorts of A. samoensis (Rosser 2015) implying that occasionally a colony must switch spawning time to facilitate gene flow between the cohorts, so that a spring-spawner harboring PaxC-clade A becomes an autumn-spawner harboring PaxC-

66

clade A (or vice versa). Assignment tests in STRUCTURE based on microsatellites suggested that three individuals that harbored PaxC clade A and were classified as

‘spring’ spawners may have been miss-classified; in other words while they harbored

PaxC clade A, they clustered more closely with the ‘autumn’ spawners. The

STRUCTURE plots suggest there are two cohorts at both Ashmore and Scott (blue and red; Fig. 4.3) which are likely to correspond to spring spawners (blue) and autumn spawners (red), but reproductive data are required to confirm this.

Divergent PaxC lineages

Phylogeographic analysis of PaxC revealed two divergent lineages with a divergence distance of 1.4%, in comparison to the absence of distinctive phylogenetic lineages and much less differentiation in the flanking region (average distance = 0.7%) and CR (average distance = 0.2%). A genealogical discordant pattern such as this could be due to a number of possibilities such as a recent selective sweep, differential introgression, incomplete lineage sorting or selection. The distribution of the PaxC clades in A. tenuis mirrors the geographic spawning patterns in this species, whereby on reefs where spawning occurs in spring clade A predominates, on reefs where spawning occurs in spring and autumn both clades occur, and on reefs where spawning occurs in autumn clade B predominates (Fig 4.2). This observation, together with the same finding of divergent PaxC clades associated with spring- and autumn-spawning cohorts in Acropora samoensis (Rosser 2015), suggests an association between PaxC and spawning seasonality and a role for natural selection. PaxC is an ancestor of Pax6 which plays a central role in eye specification in vertebrates (Catmull et al. 1998), and it has been hypothesized that PaxC may function as a type of photoreceptor in corals that cues spawning (Rosser 2015). However, the PaxC marker examined here is an intron,

67

an untranslated, non-coding region of the genome, which raises the question of why an intron would be associated with spawning seasonality and /or under divergent selection.

One explanation is that the PaxC intron (or gene) is simply a hitchhiker that is closely linked to some other speciation gene that is contributing to reproductive isolation. A second explanation is that the PaxC intron plays a functional role; introns have a variety of functions such as transcription initiation, transcription elongation and transcription termination (reviewed in Chorev & Carmel 2012). Some introns also have an important role in increasing gene expression through intron-mediated enhancement (IME)

(Mascarenhas et al. 1990), a phenomenon that has been observed in plants, invertebrates and vertebrates (reviewed in Rose et al. 2008). PaxC is a homeobox gene that encodes transcription factors (Miller et al. 2000), and interestingly, IME has been observed in another homeobox gene in Drosophila (Haerry & Gehring 1996). These ideas are speculative, of course, and further studies of transcriptome sequencing, and gene expression and regulation are required to test whether the PaxC coding region, or intron, play a role in reproductive timing in Acropora.

Conclusions

The combination of molecular evidence presented here illustrates the complexity of genetic structure that can result from a combination of factors such as selection, asynchronous reproduction and demographic history. Here, genetic differentiation was confounded by different spawning seasons and different demographic histories.

Unraveling intricate genetic structure from multiple sources is required in the analysis and interpretation of population genetic structure in corals, because management agencies rely on accurate estimates of population connectivity and genetic diversity for adequate conservation planning. This study reiterates the need to incorporate

68

observations of reproductive timing into population genetic studies of corals, because biological, ecological and evolutionary processes all play significant roles in the genetic structuring of coral populations.

69

70

Chapter 5

Phylogenomics provides new insight into evolutionary

relationships and genealogical discordance in the reef-building

coral genus Acropora

This chapter is in review at Proceedings of the Royal Society B Biological Sciences:

Rosser NL, Thomas L, Stankowski S, Richards ZT, Kennington WJ, Johnson MS

(2016) Phylogenomics provides new insight into evolutionary relationships and genealogical discordance in the reef-building coral genus Acropora

71

Abstract

Understanding the genetic basis of reproductive isolation is a long-standing goal of speciation research. In recently diverged populations, genealogical discordance may reveal genes and genomic regions that contribute to the speciation process. Previous work has shown that conspecific colonies of Acropora that spawn in different seasons

(spring and autumn) are associated with highly diverged lineages of the phylogenetic marker PaxC. Here, we used 10,034 single nucleotide polymorphisms (SNPs) to generate a genome-wide phylogeny and compared it to gene geneologies from the PaxC intron and the mtDNA Control Region (CR) in 20 species of Acropora, including three species with spring- and autumn-spawning cohorts. The PaxC phylogeny separated conspecific autumn and spring spawners into different genetic clusters in all three species; however this pattern was not supported in two of the three species at the genome-level, suggesting a selective connection between PaxC and reproductive timing in Acropora corals. This genome-wide phylogeny provides an improved foundation for resolving phylogenetic relationships in Acropora, and combined with PaxC, provides a fascinating platform for future research into regions of the genome that influence reproductive isolation and speciation in corals.

Introduction

Molecular phylogenies are archival road maps of biodiversity, providing the fundamental framework for interpreting evolutionary history and adaptation.

Accurately inferring evolutionary relationships, however, is made complicated by discordant trees from different markers, as a result of gene duplication, incomplete lineage sorting, selective sweeps, and introgression/hybridization (Maddison 1997;

72

Hahn & Nakhleh 2015; Mallet et al. 2015). Conversely, genealogical incongruence can also provide important insight into the regions of the genome that contribute to adaptation, reproductive isolation and speciation (Fontaine et al. 2015; Lamichhaney et al. 2015; Stankowski & Streisfield 2015). In the initial stages of speciation, regions of the genome that generate reproductive isolation may diverge quickly, as they experience reduced effective recombination compared to other regions (Wu 2001; Dopman et al.

2005; Via 2009). Thus, in recently diverged populations where reproductive isolation is incomplete, genealogical discordance may reveal genes and genomic regions that contribute to speciation.

An important trait affecting reproductive isolation in scleractinian corals is the timing of reproduction (Knowlton et al. 1997; van Oppen et al. 2001; Fukami et al.

2003; Wolstenholme 2004; Nakajima et al. 2012). Timing of reproduction in broadcast spawners is particularly important because gametes are viable for only a few hours, so individuals that spawn more than a few hours apart are unlikely to cross-fertilize

(Levitan et al. 2011). In Western Australia there are two spawning seasons (spring and autumn), and in some species there are two seasonal reproductive cohorts in the population (Gilmour et al. 2016; Rosser 2015; Rosser & Gilmour 2008).

To date, phylogenetic studies of Acropora have focused on the mitochondrial

DNA Control Region (CR) and the nuclear PaxC intron (van Oppen et al. 2000; van

Oppen et al. 2001; Marquez et al. 2002; van Oppen et al. 2004; Vollmer & Palumbi

2007; Richards et al. 2013). Incongruence between these regions has been attributed primarily to introgressive hybridization (van Oppen et al. 2001; Marquez et al. 2002; van Oppen et al. 2004; Richards et al. 2008), but recent evidence suggests that PaxC is under selection, and is associated with differences in spawning time (Rosser 2015,

73

2016). Specifically, PaxC revealed two highly diverged lineages (ΦST = 0.98) for different seasonal spawning cohorts in A. samoensis and A. tenuis.

Next generation sequencing technologies provide affordable, genome-wide resolution to resolve phylogenetic discordance (e.g. Nadeau & Jiggins 2010; Rubin et al. 2012; McCormack et al. 2013; Escudero et al. 2014; Rivers et al. 2016), avoiding the earlier restriction to a few genes that could be reliably amplified with sufficient variation to resolve phylogenetic relationships. Here, we construct the first genome- wide phylogeny for the scleractinian coral genus Acropora using a genotyping by sequencing (GBS) approach, and compare it to molecular phylogenies from the PaxC

46/47 intron and the mtDNA Control Region. Tests of congruence among these phylogenies provide a clearer understanding of the evolution of PaxC and an improved foundation for resolving phylogenetic relationships and patterns of speciation in corals.

Materials and Methods

Sample collection

Specimens of twenty species of Acropora were collected from a wide latitudinal range in Western Australia (Fig 5.1). Three individuals were sampled for each of 17 species (Appendix 5.1; except A. stoddarti and A. gemmifera with n=2) for which accompanying reproductive data were not collected. In the other three species, reproductive data were collected in the field by examining the size and colour of oocytes in broken branches and classifying the colonies as autumn or spring spawners

(following protocols in Rosser 2013, 2015). These included eight colonies of A. millepora (4 autumn-spawners from Ningaloo Reef and 4 spring-spawners from

Ashmore Reef; two of each were included in the GBS dataset), and 16 colonies (eight

74

spring-spawners and eight autumn-spawners) each of A. samoensis (from Rosser 2015) and A. tenuis (from Rosser 2016). Voucher specimens were retained for most samples and are housed at the Western Australian Museum (Appendix 5.1).

Fig 5.1. Geographical locations in Western Australia from where the 20 Acropora species were collected (blue boxes) for this study.

Sequencing

DNA used for Sanger sequencing was extracted from branch tips using DNeasy extraction kits for animal tissue (Qiagen, USA). Partial sequences of the mtDNA CR and the PaxC 46/47 intron were amplified using the primers and protocols described in

(Rosser 2015). We attempted to include three replicates from each species, but some samples could not be amplified across both genes (see Appendix 5.1). DNA fragments 75

were sequenced in both directions at BGI Hong Kong, edited manually in Sequencher v

4.5 (Gene Codes Corp., Ann Arbor, MI, USA), and aligned using ClustalW and Muscle in MEGA6 (Tamura et al. 2013). Heterozygotes were identified in the PaxC sequences, and IUPAC nucleotide ambiguity codes were assigned to heterozygous bases.

Genome-wide single nucleotide polymorphism (SNP) data were generated at

Diversity Arrays Technology (DArT P/L http://www.diversityarrays.com). DArTseq™ is genotyping-by-sequencing technology which represents a combination of a DArT complexity reduction methods and next generation sequencing platforms and is similar to the widely applied RADseq methodology (see Appendix 5.2 for details of DArTseq marker development). Briefly, genomic DNA was processed in digestion/ligation reactions principally as per (Kilian et al. 2012) but replacing a single PstI-compatible adaptor with two different adaptors corresponding to two different restriction enzyme overhangs. Sequencing was carried out on a single lane of an Illumina Hiseq2500 and processed using proprietary DArT analytical pipelines.

Data analysis

Phylogenetic relationships were estimated for PaxC and the CR using a

Bayesian statistical framework implemented in MrBayes 3.1.2 (Ronquist &

Huelsenbeck 2003), and Maximum Likelihood (ML) analyses in PhyML 3.0 (Guindon et al. 2010) (detail in Appendix 5.2). Both genes contained numerous indels, which were coded as single base changes, and trees were rooted with the sister genus Isopora as an outgroup, using sequences of I. cuneata obtained from GenBank (Accession No.s

EU918925 and AY026429). P-distances between autumn- and spring-spawning cohorts were calculated in MEGA6 (Tamura et al. 2013). For the SNP analyses, the SNP markers were extracted from the sequences (read length ~100 bp) and concatenated into

76

supermatrices using IUPAC codes for heterozygous loci. ML analyses were conducted in RAxML (Stamatakis 2014) using the GTR + gamma model of sequence evolution and support for each node was assessed with 100 bootstrap replicates. Because the concatenation of variable SNPs artificially inflates branch lengths (Leache et al. 2015), we used the acquisition bias correction implemented in RAxML to generate the final topology. In addition to the tree-based methods, we also conducted a Principal

Coordinate Analysis (PCoA) in GenAlEx (Peakall & Smouse 2006, 2012) for both the entire dataset and on subsets of the data comprising sympatric spring and autumn A. samoensis colonies (n = 16) and allopatric A. tenuis colonies (n=16).

Incongruence between gene trees can occur simply because the phylogenetic signal is weak, and the tree topologies differ by chance, or because the genes have not shared the same evolutionary history (Planet 2006). To eliminate the first possibility, we used two likelihood-based tests, a Shimodaira-Hasegawa test (SH) (Shimodaira &

Hasegawa 1999) and a one-tailed Kishino-Hasegawa test (KH) (Goldman et al. 2000) to compare tree topologies among the SNP, CR and PaxC trees and assess whether all trees were equally good explanations of the data. Only samples that were successfully amplified across all three of the datasets were included in the tests, which were conducted in Tree-Puzzle (Schmidt et al. 2002) using nonparametric bootstrap with re- estimated log likelihoods (RELL) approximation for re-sampling the log likelihood with

1,000 replicates.

77

Results

Sequencing

In total, 85 individuals were sequenced for PaxC and 84 were sequenced for CR.

After the indels were coded as single base changes and sequences were aligned and trimmed, segments of the CR and PaxC sequences consisted of 1,036 bp and 357 bp respectively from 20 species of Acropora. Eighty-six individuals were genotyped using

DArT-seq methodologies across a total of 44,356 loci. We filtered DArT loci for a minimum genotype call rate of 0.70 and minimum 8X coverage, leaving 10,034 (23%) loci remaining for phylogenetic analyses. Within each species, approximately 18% of loci were polymorphic, and the average frequency of homozygotes for the reference allele was 0.716 (+/- 004) (Appendix 5.3). The resolution of the inferred tree topology substantially increased as the SNP data matrix increased in size; using genotype call rate of 1.00 and 0.90 produced topologies with much lower bootstrap support in the internal branches than when using a call rate of 0.70 (Appendix 5.4).

Data analysis

Tests of congruence between the trees of the CR, PaxC and the SNP datasets showed that the SNP tree was the optimal tree (Table 5.1), and the results of the KH and

SH tests indicated that the CR and PaxC trees were significantly worse representations of the data than the SNP tree (P < 0.005, Table 5.1). High posterior probabilities extended to finer relationships in the SNP tree, offering greater resolution of evolutionary relationships than the CR or PaxC (Fig. 5.2). All methods of analysis recovered three of the four phylogenetically discrete Acropora clades from the original

Acropora phylogeny in van Oppen et al. (2001) (the monotypic fourth clade was not

78

recovered because A. latistella was not included in this study), although clade III in

PaxC and clade IV in CR were weakly supported by bootstrap values (Fig. 5.2). PaxC was a better fit to the SNP tree than the CR (PaxC -ln L = 3831.7 compared to CR –ln L

= 4183.72; Table 5.1) and contained three discrepancies with the SNP tree, while the

CR contained seven discrepancies with the SNP tree (Appendix 5.5). Three species were polyphyletic with colonies split between clades III and IV in all of the CR, PaxC and the SNP trees (A. digitifera, A. aspera and A. lutkeni; Fig 5.2). Many species in the

SNP tree were polyphyletic with colonies split within clades (A. subulata, A. pulchra, A. stoddarti, A. gemmifera, A. muricata, A. tenuis, A. selago, A. florida, A. samoensis and

A. divaricata; Fig. 5.2), and one species that was monophyletic in the SNP tree was polyphyletic in the CR tree (A. spicifera; Fig. 5.2).

Table 5.1. Test results obtained from the likelihood-based Shimodaira-Hasegawa (SH) and Kishino-Hasegawa (KH) tests for the three phylogenetic trees (SNP, PaxC and CR). P values (P) were calculated from 1,000 permutations using the RELL method and were significant at 0.05*.

Trees -ln[L] ∆ln[L] KH (P) SH (P)

SNP 3586.08 - 1.0000 1.0000

PaxC 3831.70 245.61 0.0000* 0.0010*

CR 4183.72 597.63 0.0000* 0.0000*

The PaxC tree placed the spring- and autumn-spawners of all three species (A. millepora, A. samoensis and A. tenuis) into different clusters within the major clades

(Fig. 5.2). Contrastingly, the SNP tree split the spring- and autumn-spawners of A. samoensis but not of A. millepora or A. tenuis (Fig. 2). The spring- and autumn- spawners were not separated in any species in the CR tree, which had low phylogenetic signal. The differences in PaxC sequences between spring- and autumn-spawners in

79

Figure 5.2. Comparison of phylogenetic trees in the genus Acropora using 10,034 SNPs, mtDNA CR and PaxC, with collapsed nodes to illustrate major patters (see Appendix 5.4 for uncollapsed SNP trees). Branch support values are maximum likelihood bootstrap values and Bayesian posterior probabilities (outgroups are not shown). Major Acropora clades are indicated by Roman numerals; spring and autumn spawners are shown in blue and red; symbols after species indicate polyphyletic lineages. 80

A. millepora, A. samoensis and A. tenuis were characterized by multiple fixed differences and phylogenetically informative indels (Fig. 5.3). The p-distance between the autumn and spring spawners was highest in A. samoensis (p = 0.017; Fig. 5.3a) and lowest in A. tenuis (p = 0.011; Fig. 5.3c).

Figure 5.3. Comparisons between haplotypes of PaxC in autumn- and spring-spawners in (a) A. samoensis, (b) A. millepora and (c) A. tenuis. Each box represents a nucleotide base; the top row represents the most common haplotype in each species (dark shading), and light shading illustrates a different nucleotide. Only unique haplotypes within each reproductive cohort are shown. Fixed differences between the spring- and autumn- spawners are shown by black circles; the numbers indicate the presence/absence of an indel and detail the length (number of base pairs) in fixed indels. p = mean p distance between autumn- and spring- spawners in each species.

81

The PCoA based on the SNP dataset showed a very similar pattern to the phylogenetic analyses; the genus was split into three clusters that corresponded to the three phylogenetic clades, but none of the species showed a clear split between the autumn and spring spawners (Fig. 5.4a). Limiting the SNP dataset to A. samoensis and

A. tenuis with genotype call rates of 1.0 (3,554 SNPs for A. samoensis and 1,458 SNPs for A. tenuis), revealed clear clustering between sympatric spring and autumn spawners in A. samoensis (Fig. 5.4c), but not a clear split between the autumn and spring spawners in A. tenuis (Fig. 5.4b).

(a) III

I

2 (10%) 2 Coord

IV

Coord 1 (28%) (b) (c)

Figure 5.4. Plots of the first two axes from Principal Co-ordinates Analyses (PCoA); (a) all Acropora species examined (10,034 loci), (b) A. tenuis seasonal cohorts (1,458 loci) and (c) A. samoensis seasonal cohorts (3,554 loci). Dashed eclipses enclose species within the major Acropora clades; triangles = A. tenuis, circles = A. samoensis, diamonds = all other Acropora species; within A. tenuis and A. samoensis white shading = autumn-spawners, grey shading = spring-spawners.

82

Discussion

This study represents a significant step forward in unravelling the evolutionary relationships among Acropora corals, and provides insight into the influence of reproductive timing on the evolutionary patterns in this group of important reef-building species. There are three possible explanations for incongruence between the phylogenies presented in this study. First, PaxC could be a rapidly evolving gene that presents a highly resolved phylogeny, with which other loci will eventually become phylogenetically concordant. However published mutation rates of the PaxC intron are less than those of the CR (Chen et al. 2009), making this explanation unlikely.

Furthermore, the SH test suggests lower phylogenetic signal in the PaxC tree, and there is less resolution at branch tips in the PaxC tree than the SNP tree, indicating that the

PaxC phylogeny is not more highly resolved. A second possible explanation is that the

PaxC intron and/or coding region are under selection associated with coral spawning season. The PaxC gene in anthozoans is closely related to Pax6 in higher order , which is involved in developing eyes (Callaerts et al. 1997; de Jong 2005; Matus et al.

2007; Miller et al. 2000). While anthozoans do not have eyes, they are sensitive to light, and they use light cues to control spawning on a range of time scales (Brady et al. 2009;

Reitzel et al. 2013; Sweeney et al. 2011; van Woesik et al. 2006); hence PaxC might function as a type of photoreceptor that cues spawning. A third possible explanation is that PaxC is simply a hitchhiker linked to other aspects of reproductive isolation, as quantitative trait loci for different traits under divergent selection can co-localize on the genetic linkage map (Hawthorne & Via 2001). Further investigation is required to separate these two possibilities, and provides a fascinating avenue for future research into the genes that influence reproductive isolation and speciation in corals.

83

Irrespective of whether it is functional or a hitchhiker, our analysis suggests that

PaxC is located in a genomic region that contributes to reproductive isolation in

Acropora corals. Identifying genes and genomic regions that confer isolation is a major goal of speciation research, because they can provide insight into ecological settings, evolutionary forces and molecular mechanisms that drive the divergence of populations

(Orr et al. 2004; Rieseberg & Blackman 2010). Genes that confer reproductive isolation may very well be leading indicators of evolutionary relationships and define the branches of what will ultimately become the species tree. In A. samoensis, the PaxC lineages are clearly separated in the SNP phylogeny, but in A. tenuis and A. millepora they are not. The level of genetic differentiation between PaxC sequences associated with autumn- and spring-spawning cohorts is lower in A. tenuis and A. millepora than in

A. samoensis (Fig. 5.3), suggesting incomplete reproductive barriers or recent polymorphism in these species. Greater divergence in PaxC in A. samoensis could reflect a longer period of temporal isolation between autumn and spring spawners in this species. The A. humilis group is an old species group, with fossil discoveries in

Indonesia dating A. samoensis to 9.4-9.8 MY old (Santodomingo 2014), and fossils of

A. slovenica (also in the A. humilis group and very similar to A. samoensis) to the

Oligocene – approximately 28.1 to 33.9 MY old (Wallace & Bosellini 2014). If so, these patterns suggest that genetic divergence associated with the timing of reproduction may take a long time to evolve to the stage of reciprocal monophyly.

Despite the incongruence associated with PaxC and reproduction, all three gene trees revealed numerous polyphyletic species split across the major clades. In three polyphyletic species in particular (A. aspera, A. digitifera and A. lutkeni), conspecifics occur in the same position in all phylogenetic analyses, they were well supported in all trees, and patterns consistent between the mitochondrial and nuclear genes. This

84

combined evidence suggests that these are indeed phylogenetic lineages, and represent morphologically cryptic species. Advances in molecular techniques in the past two decades have revealed a plethora of cryptic species that are widespread across scleractinian coral genera (Flot et al. 2011; Ladner & Palumbi 2012; Schmidt-Roach et al. 2013; Warner et al. 2015; Richards et al. 2016), and their continued identification is critical for the successful conservation and management of these ecologically important and globally threatened group of corals.

85

86

Chapter 6

Synthesis

“Because we have much more landscape and coastline than people, our shores

and shallows are still rich in life, diversity and strangeness. We have perhaps

more than our fair share of shoreline miracles, of visitations and wonders…”

(Tim Winton, on Western Australia, in Land’s Edge)

87

This thesis explored the influence of asynchronous reproductive timing on patterns of genetic structure in coral populations, at both ecological and evolutionary levels, in the context of understanding the evolution of seasonal breeding patterns in

Western Australia. Following documentation of the geographical pattern of seasonal breeding in Acropora, this thesis aimed to answer three questions:

(i) Are conspecific colonies that spawn in autumn and spring reproductively isolated and genetically differentiated, or do colonies switch spawning time, allowing genetic mixing?

(ii) Are conspecific autumn- and spring-spawning colonies associated with distinct phylogenetic lineages?

(iii) Are spawning patterns on WA reefs a result of an inherited legacy from northern ancestors, or natural selection?

Here, I integrate the results from each chapter to draw together the key findings and answer these questions. I also review the limitations of this study, and outline the main implications, including areas for future research.

Asynchronous spawning influences reproductive isolation

Population genetic analysis of microsatellites in sympatric (Acropora samoensis) and allopatric (A. tenuis) populations in Chapters 3 and 4 showed that autumn- and spring-spawning cohorts are genetically differentiated, (FST = 0.17 in both species), confirming strong isolation. In A. samoensis, this differentiation was accompanied by subtle morphological differences, hence the combination of genetic, phylogenetic, morphological and reproductive differences between the autumn- and spring-spawners indicate that the reproductive cohorts represent cryptic species (i.e.

88

distinct genotypic clusters; Mallet 1995). However, in A. tenuis, while the autumn- and spring-spawning cohorts were genetically differentiated, this differentiation was confounded by geographic separation, and the lack of seasonal spawning data for specific genotypes made it difficult to isolate the contribution of spawning-related variation from geographic variation. Nevertheless, FST between populations at Ashmore

Reef and populations south of Scott Reef was substantial (range 0.40-0.58), and was accompanied by subtle morphological differences and differences in reproductive timing, suggesting the population of A. tenuis at Ashmore Reef may be also be a cryptic or incipient species. The identification of cryptic species is important not only for estimates of biodiversity, but also for correct interpretation of population connectivity, gene flow, dispersal and ecological patterns (Bickford et al. 2007). Conservation management relies on accurate estimates of these processes, and the recognition that asynchronous reproduction can be associated with cryptic speciation indicates the importance of incorporating observations of reproductive timing into population genetic studies of corals and that cryptic species are identified.

The level of reproductive isolation between the sympatric reproductive cohorts of A. samoensis reflects the findings from other coral species, that seasonal spawning time is strongly heritable (Levitan et al. 2011; Vize et al. 2005), and does not routinely switch between seasons. Nevertheless, spawning time is not etched in stone, as shown by the IMa2 analysis of A. samoensis in Chapter 3, which detected some introgression between the reproductive cohorts, indicating some individuals have switched spawning season in the past. Experimental studies suggest that, while there is a strong genetic component to reproductive timing, it is also influenced by local environmental variation

(Fan & Dai 1999; Levitan et al. 2011).

89

The origin of autumn and spring spawners on WA reefs

Phylogeographic analysis of A. tenuis in Chapter 4 provides a clue to the evolutionary origin of autumn and spring spawners on Western Australian reefs. That chapter showed that in terms of the distribution of the PaxC clades, Clade B (associated with autumn spawners) was ubiquitous, but Clade A (associated with spring spawners) was found only on northern reefs. This clinal variation of PaxC clade A suggests that (i) its origin was most likely north of the study area, and (ii) that a gene-environment interaction may have evolved in allopatric populations with different climatic conditions, where selection pressure resulted in successful breeding in different seasons.

Repeated glacial cycles have resulted in the emergence and subsidence of the Sunda

Shelf in the Indonesian Archipelago (Voris 2000), potentially isolating coral populations on either side of the shelf during periods of low sea level, and phylogeographic breaks along the Sunda Shelf are common among invertebrates in this region (reviewed in Carpenter et al. 2011). Hence the two PaxC clades (and associated spring and autumn spawners) on WA reefs are likely to be in secondary contact following allopatric divergence in Indonesia. However, a fuller understanding of the phylogenetic origin of the lineages of A. tenuis in Western Australia will require examination of samples from Indonesia.

While distinct phylogenetic lineages of PaxC were evident in autumn- and spring-spawning cohorts of A. samoensis, A. tenuis and A. millepora, this did not translate into distinct phylogenetic lineages within each species (except perhaps in A. samoensis). This suggests that the autumn- and spring-spawning cohorts have not been reproductively isolated long enough or completely enough for genome-wide evolutionary divergence to occur. Conversely at the PaxC locus, divergent selection means that this locus is resistant to gene exchange (Turner et al. 2005; Via 2009). 90

Studying barriers to gene flow in populations that are not yet completely reproductively isolated can reveal important aspects of the evolutionary process that have never been seen before (Via 2009), as was the case with PaxC.

The evolution of seasonal breeding times on WA reefs is a result of natural selection

The spawning surveys in Chapter 2 showed that the major coral spawning season is in autumn at all locations in Western Australia, but the magnitude of coral spawning in spring is correlated with latitude, with a decrease in the proportion of species spawning in spring from 49% at Ashmore Reef (12°S) to 7% at Ningaloo Reef

(23°S). Two possibilities could explain the lack of spring spawners on high latitude reefs: (a) seasonal filtering of larval recruits from northern reefs, or (b) localized natural selection. The results of this study suggest that spawning patterns on WA reefs are a result of local selection, and not the seasonal filtering of recruits from northern reefs.

Chapter 2 showed that some species that were not found to spawn in autumn at

Ashmore Reef do spawn in autumn at Ningaloo Reef (e.g. A. tenuis, A. millepora, A. secale), suggesting that the populations at Ningaloo Reef have not inherited their spawning time from their northern relatives. Nevertheless, Ashmore Reef has a different evolutionary history to other WA reefs, and was probably re-colonized from a different source during the Holocene, so perhaps corals on southern reefs inherited their spawning time from Scott Reef and not Ashmore. However, these same patterns occur on other reefs: A. millepora spawns predominantly in spring at Scott Reef but in autumn at Ningaloo (Gilmour et al. 2009; Rosser 2013); A. samoensis and A. valida spawn in spring at the Montebello Islands but only in autumn at Ningaloo (Rosser unpublished data). Therefore, this pattern suggests it is highly unlikely that these species do not spawn in spring at Ningaloo on account of the filtering of recruits, but rather, that

91

spawning patterns on WA reefs are a result of local selection that drives corals to reproduce at a time that coincides with the most suitable conditions for fertilization success and offspring survival.

Reproductive surveys on the east coast of Australia have shown latitudinal variation in spawning season among conspecific populations in numerous Acropora species, with the date of mass spawning ranging from October at 15°S to January at

30°S, in association with the timing of the maximum sea surface temperature in each region (Babcock et al. 1986; Baird et al. 2015; Wilson & Harrison 2003). It is an old paradigm that variation in seasonal water temperatures accounts for latitudinal variation in the breeding season of marine benthic invertebrates (Orton 1920). In Western

Australia, however, mass spawning occurs in the same month (March) at both tropical and subtropical latitudes, despite marked differences in the temperature regimes, and this was one of the arguments in support of an inherited genetic legacy determining coral spawning seasonality (Babcock et al. 1994; Simpson 1991). Nonetheless, this merely suggests that while temperature is undoubtedly important in determining the time of coral spawning (reviewed in Nozawa 2012), temperature is not the only factor involved, and multiple environmental variables are likely to play an important role in influencing the timing of coral spawning, but these require further investigation.

The association between PaxC and spawning time

Perhaps the most unexpected and interesting finding of this study was the association between divergent PaxC lineages and different spawning seasons, that was unique to this marker and was not present in other DNA sequence markers (CR,

Calmodulin, flanking region or 10,034 SNPs). This result suggests there is a selective connection of some sort between PaxC and reproductive timing. All of the PaxC

92

markers that showed an association with spawning in this study were introns, which are non-coding regions of the genome, raising the question of why an intron would be associated with spawning seasonality and /or under divergent selection. However, some introns are functional and play a role in enhancing gene expression, or initiating/terminating transcription (Mascarenhas et al. 1990; Haerry & Gehring 1996;

Rose et al. 2008; Chorev & Carmel 2012). The introns in PaxC are very ancient, predating the cnidarian/triploblast split at least 543 million years ago (Grotzinger et al.

1995; Miller et al. 2000) indicating they are highly conserved, and evolutionary conservation is often indicative of biological function (Chorev & Carmel 2012).

Alternatively, the PaxC intron, or gene, may simply be a hitchhiker, and it may be that reproductive timing is linked to other aspects of selection and that both evolve sympatrically. Further studies of transcriptome sequencing, and gene expression and regulation are required to test these ideas, and provide intriguing avenues for future research.

If PaxC does have a role in reproductive timing in corals, it is particularly fascinating owing to the relationship of this gene to eye development in higher order animals. The control of reproductive timing in corals is complex, and is influenced by multiple environmental cycles that control the season, lunar phase and hour that spawning occurs, but the extent to which each of these cycles is regulated by environmental signals, or entrained and regulated by circadian, circalunar or circanual rhythms is unknown. Acroporids have several circadian clock genes (e.g. clock, timeless, cry2) which are orthologs to mammalian clock genes (Shoguchi et al. 2013;

Vize 2009), and several photoreceptor genes (e.g. opsins, melanopsin and cryptochromes; Levy et al. 2007; Shoguchi et al. 2013), and exhibit photoreception in the blue region of the light spectrum, with particular sensitivity to blue moonlight

93

irradiance levels (Gorbunov & Falkowski 2003; Levy et al. 2003). Photo-responsive cells of corals detect and respond to light by altering cytoplasmic calcium levels, similarly to the way transduction pathways occur in complex invertebrate eyes (Hilton et al. 2012). If PaxC is acting as a photoreceptor, it raises the question of why a photoreceptor gene would be associated with seasonal spawning. One explanation is that it could have a dual role in light perception and endogenous timekeeping in a similar manner to cryptochromes. Cryptochromes are blue-light activated proteins that transduce the light input into the clock mechanism via signaling cascades and interaction with other clock genes (Oliveri et al. 2014); so they effectively play a role in both photoreception and circadian regulation in corals. If this were also the case in

PaxC it could explain the link between photoreception (as in Pax6) and the association with seasonal reproductive timing. Seasonal cycles are often entrained and controlled by long-term circannual rhythms, such as plant flowering and bird migration (Dunlap et al.

2004), but the genetic basis of circannual rhythms and the genes that contribute to the circannual clock are currently unknown (Visser et al. 2010). These ideas are highly speculative, of course, and require much further investigation to determine whether

PaxC plays any role in photoreception, circadian regulation or reproductive timing in

Acropora.

Limitations of this study

The main limitations of this study were the small sample sizes, and the absence of detailed spawning data for all of the populations surveyed. Western Australian reefs are remote, and many are hard to access, and most of the field work in this study was piggy-backed onto other expeditions with different agendas, which often precluded collecting ideal data. For example, the A. samoensis colonies studied in Chapter 3 were

94

tagged and monitored as part of a coral spawning monitoring program for another project, and the population was randomly chosen with no preference for spring or autumn spawners. This resulted in only a small sample of spring-spawners (n=15) of A. samoensis, which limited the capacity of some advanced statistical analyses, such as estimating the divergence time between the spring and autumn cohorts in the IMa2 analysis.

In Chapter 4, most of the populations of A. tenuis were sampled by a former

PhD student, who was studying dispersal and not coral reproduction, so data on spawning season were not collected for those samples. The absence of spawning data for specific individuals of A. tenuis meant that hypotheses about spawning time and

PaxC clade could only be inferred, rather than specifically tested. For example, analyses in Chapter 3 showed that introgression has occurred between the spring and autumn cohorts in A. samoensis, implying that occasionally a colony must switch spawning time to facilitate gene flow between the cohorts, so that a spring-spawner harboring PaxC- clade A becomes an autumn-spawner harboring PaxC-clade A (or vice versa). While I did not detect any individuals where this was the case in A. samoensis, the same cannot be said for A. tenuis. When individuals were assigned to spring and autumn spawning groups according to the PaxC clade they harbored, assignment tests in STRUCTURE based on the microsatellites suggested that several individuals were misclassified.

Therefore, the type of PaxC clade could not be used to classify colonies as spring or autumn spawners with certainty, so I could not determine the level of differentiation between the non-Ashmore spring and autumn spawners, or compare the similarity between the spring-spawners at Ashmore with the spring-spawners at non-Ashmore

Reefs. These comparisons may have provided further information about the age and

95

colonization history of Ashmore and Scott Reef populations, and the extent of local adaptation in these populations.

Implications and future directions for research

The results of this study have three important implications. First, the seasonality of breeding seasons in Western Australia is not due to an inherited “genetic legacy”, but rather is influenced by local natural selection, presumably to spawn when conditions favour offspring survival. Therefore, in the face of rapid climate change more research is needed to understand exactly which environmental factors drive reproductive schedules, and how these will be influenced by changing climatic conditions. The finding that introgression occurs between spring and autumn reproductive cohorts in A. samoensis indicates that spawning time is not fixed in stone, and occasionally a colony switches from spawning in autumn to spawning in spring, presumably due to environmental effects. But what are these environmental effects, and how important are they? This line of enquiry needs further research.

The second major implication is that coral diversity on Western Australian reefs is higher than is typically accounted for, due to the incidence of cryptic species. The phylogenetic analysis of Western Australian acroporids identified three species that were polyphyletic and contained cryptic species. Spring- and autumn-spawning cohorts of A. samoensis are also cryptic species, which takes the number of new cryptic species identified in this study alone to four. These findings have obvious implications for estimates of biodiversity on Western Australia’s coral reefs, but moreover, the identification of cryptic species is also important for correct interpretation of population connectivity, gene flow, dispersal and ecological patterns to inform conservation and management planning. This study illustrated that differences in reproductive timing can

96

contribute to genetic subdivision within populations, therefore it is vital that observations of reproductive timing is incorporated into the population genetic studies of corals.

The third major implication is that there is an association between PaxC and spawning time, so it should be used with caution as a phylogenetic marker. More importantly this finding opens up exciting avenues of research into the genes that influence coral spawning, and this field of knowledge is in its infancy. If PaxC is somehow involved in coral reproduction, the clinal variation in the distribution of PaxC clades in A. tenuis suggests potential genetic-environmental interactions, and the extent to which coral reproductive schedules are influenced by genetics or environmental factors is another fascinating and important avenue for future research.

97

References

Alexander RD, Bigelow RS (1960) Allochronic speciation in field crickets, and a new species, Acheta veletis. Evolution 14, 334-346. Avise JC (2004) Molecular markers, natural history, and evolution Second edn. Sinauer Associates Inc. Publishers, Massachusetts. Ayre DJ, Hughes TP (2000) Genotypic diversity and gene flow in brooding and spawning corals along the Great Barrier Reef, Australia. Evolution 54, 1590- 1605. Ayre DJ, Hughes TP (2004) Climate change, genotypic diversity and gene flow in reef corals. Ecology Letters 7, 273-278. Babcock RC (1984) Reproduction and distribution of two species of Goniastrea () from the Great Barrier Reef Province. Coral Reefs 2, 187-195. Babcock RC, Bull GD, Harrison PL, et al. (1986) Synchronous spawnings of 105 scleractinian coral species on the Great Barrier Reef. Marine Biology 90, 379- 394. Babcock RC, Wills BL, Simpson CJ (1994) Mass spawning of corals on a high latitude coral reef. Coral Reefs 13, 161-169. Baird AH, Birrel CL, Hughes TP, et al. (2009b) Latitudinal variation in reproductive synchrony in Acropora assemblages: Japan vs Australia. Galaxea 11, 101-108. Baird AH, Blakeway DR, Hurley TJ, Stoddart JA (2011) Seasonality of coral reproduction in the Dampier Archipelago, northern Western Australia. Marine Biology 158, 275-285. Baird AH, Cumbo VR, Gudge S, et al. (2015) Coral reproduction on the world’s southernmost reef at Lord Howe Island, Australia. Aquatic Biology 23, 275-284. Baird AH, Guest JR, Willis BL (2009a) Systematic and biogeographical patterns in the reproductive biology of scleractinian corals. Annual Review of Ecology, Evolution and Systematics 40, 551-571. Baird AH, Marshall PA, Wolstenholme JK (2002) Latitudinal variation in the reproduction of Acropora in the Coral Sea 1, 385-389. Bickford D, Lohman DJ, Sodhi NS, et al. (2007) Cryptic species as a window on diversity and conservation. Trends in Ecology & Evolution 22, 148-155. Binks R, Prince J, Evans JP, Kennington JW (2012) More than Bindin divergence: reproductive isolation between sympatric subspecies of a sea urchin by asynchronous spawning. Evolution 66, 3545-3557. Bird CE, Holland BS, Bowen BW, Toonen RJ (2011) Diversification of sympatric broadcast-spawning limpets (Cellana spp.) within the Hawaiian archipelago. Molecular Ecology 20, 2128-2141. Brady AK, Hilton JD, Vize PD (2009) Coral spawn timing is a direct response to solar light cycles and is not an entrained circadian response. Coral Reefs 28, 677-680. Brawand D, Wagner CE, Yang IL, et al. (2014) The genomic substrate for adaptive radiation in African cichlid fish. Nature 513, 375-381. Callaerts P, Halder G, Gehring W (1997) PAX-6 in development and evolution. Annual Review of Neuroscience 20, 483-532. Carpenter KE, Barber PH, Crandall ED, et al. (2011) Comparative Phylogeography of the Coral Triangle and Implications for Marine Management. Journal of Marine Biology 2011, 1-14.

98

Catmull J, Hayward DC, McIntyre NE, et al. (1998) Pax6 origins - implications from the structure of two coral Pax genes Dev Genes Evol 208, 352-356. Chapuis MP, Estoup A (2007) Microsatellite null alleles and estimation of population differentiation. Molecular Biology and Evolution 24, 621-631. Chatrou LW, Escribano MP, Viruel MA, et al. (2009) Flanking regions of monomorphic microsatellite loci provide a new source of data for plant species- level phylogenetics. Molecular Phylogenetics and Evolution 53, 726-733. Chen I-P, Tang C-Y, Chiou C-Y, et al. (2009) Comparitive analyses of coding and noncoding DNA regions indicate that Acropora (: Scleractinia) posses a similar evolutionary tempo of nuclear vs. mitochondrial genomes as in plants. Marine Biotechnology 11, 141-152. Chorev M, Carmel L (2012) The function of introns. Frontiers in Genetics 3, 55, 1-15. Creer S (2007) Choosing and Using Introns in Molecular Phylogenetics. Evolutionary Bioinformatics 3, 99-108. Coyne JA, Orr HA (2004) Speciation Sinauer Associates, Sunderland, M.A. Cresswell GR (1991) The Leeuwin Current - observations and recent models. Journal of the Royal Society of Western Australia 74, 1-14. D'Adamo N, Fandry C, Buchan S, Domingues C (2009) Northern sources of the Leeuwin Current and the "Holloway Current" on the North West Shelf. Journal of the Royal Society of Western Australia 92, 53-66. Dai CF, Fan TY, Yu JK (2000) Reproductive isolation and genetic differentation of a scleractinian coral Mycedium elephantotus. Marine Ecology Progress Series 201, 179-187. de Jong D (2005) Characterisation of homologs of nervous system patterning genes in the staghorn coral, Acropora millepora (; Anthozoa; Scleractinia). PhD thesis, James Cook University. Domingues CM (2006) Kinematics and heat budget of the Leeuwin Current, Flinders University. Dopman EB, Perez L, Bogdanowicz SM, Harrison RG (2005) Consequences of reproductive barriers for geneological discordance in the European corn borer. Proceedings of the National Academy of Sciences of the United States of America 102, 14706-147011. Dunlap JC, Loros JL, DeCoursey PJ (2004) Chronobiology: biological timekeeping Sinauer Associates, Sunderland, MA. Earl DA, vonHoldt BM (2012) STRUCTURE HARVESTER: a website and program for visualizing STRUCTURE output and implementing the Evanno method. Conservation Genetics Resources 4, 2359-2361. Eaton DAR, Hipp AL, Gonzalez-Rodriguez A, Cavender-Bares J (2015) Historical introgression among the American live oaks and the comparitive nature of tests for introgression. Evolution 69, 2587-2601. Escudero M, Eaton DAR, Hahn M, Hipp AL (2014) Genotyping-by-sequencing as a tool to infer phylogeny and ancestral hybridization: a case study in Carex (Cyperaceae). Estoup A, Angers A (1998) Microsatellites and minisatellites for molecular ecology: theoretical and empirical considerations In: Advances in Molecular Ecology (ed. Carvalho GR). IOS Press. Evanno G, Regnault S, Goudet J (2005) Detecting the number of clusters of individuals using the software STRUCTURE: a simulation study. Molecular Ecology 14, 2611-2620.

99

Excoffier L, Lischer HEL (2010) Arlequin suite ver 3.5: A new series of programs to perform population genetics analyses under Linux and Windows. Molecular Ecology Resources 10, 564-567. Excoffier L, Smouse P, Quattro J (1992) Analysis of molecular variance inferred from metric distances among DNA haplotypes: Application to human mitochondrial DNA restriction data. Genetics 131, 479-491. Fan TY, Dai CF (1999) Reproductive plasticity in the reef coral Echinopora lamellosa. Marine Ecology Progress Series 190, 297-301. Flot J-F, Blanchot J, Charpy L, et al. (2011) Incongruence between morphotypes and genetically delimited species in the coral genus Stylophora: phenotypic plasticity, morphological convergence, morphological stasis or interspecific hybridization? BMC Ecology 11, 22. Flot J-F, Magalon H, Cruaud C, Couloux A, Tillier S (2008) Patterns of genetic structure among Hawaiian corals of the genus Pocillopora yield clusters of individuals that are compatable with morphology. Comptes Rendus Biologies 331, 239-247. Fontaine MC, Pease JB, Steele A, et al. (2015) Extensive introgression in a malaria vector species complex revealed by phylogenomics. Science Fu Y, Li W (1993) Statistical tests of neutrality of mutations. Genetics, 693-709. Fukami H, Omori M, Shimoike K, Hayashibara T, Hatta M (2003) Ecological and genetic aspects of reproductive isolation by different spawning times in Acropora corals. Marine Biology 142, 679-684. Garza J, Slatkin M, Freimer N (1995) Microsatellite allele frequencies in humans and chimpanzees, with implications for constraints on allele size. Molecular Biology and Evolution 12, 594-603. Gilmour JP, Smith LD, Brinkman RM (2009) Biannual spawning, rapid larval development and evidence of self-seeding for scleractinian corals at an isolated system of reefs. Marine Biology 156, 1297-1309. Gilmour JP, Underwood JN, Howells EJ, Gates E, Heyward AJ (2016) Biannual spawning and temporal reproductive isolation in Acropora corals. PLoS ONE 11, e0150916 Goldman N, Ansderson JP, Rodrigo AG (2000) Likelihood-based tests of topologies in phylogenetics. Systematic Biology 49, 652-670. Gorbunov MY, Falkowski PG (2003) Photoreceptors in the cnidarian hosts allow symbiotic corals to sense blue moonlight. Limnology and Oceanography 47, 309-315. Grant WS, Bowen BW (1998) Shallow population histories in deep evolutionary lineages of marine fishes: insights from sardines and anchovies and lessons for conservation. Journal of Heredity 89, 415-426. Grotzinger JP, Bowring SA, Saylor BZ, Kaufman AJ (1995) Biostratigraphic and geochronologic constraints on early animal evolution. Science (Washington) 270, 598-604. Guest JR, Baird AH, Goh BPL, Chou LM (2005) Reproductive seasonality in an equatorial assemblage of scleractinan corals. Coral Reefs 24, 112-116. Guindon S, Dufayard JF, Lefort V, Anisimova M, Hordijk W, Gascuel O (2010) New algorithms and methods to estimate maximum-likelihood phylogenies: assessing the performance of PhyML 3.0. Systematic Biology 59, 307-321. Haerry TE, Gehring WJ (1996) Intron of the mouse Hoxa-7 gene contains conserved homeodomain binding sites that can functionas an enhancer element

100

in Drosophila. Proceedings of the National Academy of Sciences of the United States of America 93, 13884-13889. Hahn MW, Nakhleh L (2015) Irrational exuberance for resolved species trees. Evolution 70, 7-17 Hanafy MH, Aamer MA, Habib M, Rouphael AB, Baird AH (2010) Synchronous reproduction of corals in the Red Sea. Coral Reefs 29, 119-124. Harrison PL (2011) Sexual reproduction of scleractinian corals. In: Coral reefs: an ecosystem in transition (eds. Dubinsky Z, Stambler N). Springer. Harrison PL, Babcock RC, Bull GD, Oliver JK, Wallace CC, Willis BL (1984) Mass spawning in tropical reef corals. Science (Washington) 223, 1186-1189. Harrison PL, Wallace CC (1990) Reproduction, dispersal and recruitment of scleractinian corals. In: Ecosystems of the world, Vol 25: Coral Reefs (ed. Dubinsky Z), pp. 133-207. Elsevier Science Publishers, Amsterdam. Hawthorne DJ, Via S (2001) Genetic linkage facilitates ecological specialization and reproductive isolation in pea aphids. Nature 412, 904-907. Hellberg ME (2006) No variation and low synonymous substitution rates in coral mtDNA despite high nuclear variation. BMC Evolutionary Biology 6. Hendry AP, Day T (2005) Population structure attributable to reproductive time: isolation by time and adaptation by time. Molecular Ecology 14, 901-916. Hewitt G (2000) The genetic legacy of the Quaternary ice ages. Nature 405, 907-913. Hewitt GM (1996) Some genetic consequences of ice ages, and their role, in divergence and speciation. Biological Journal of the Linnean Society 58, 247-276. Hey J, Nielsen R (2007) Integration within the Felsenstein equation for improved Markov Chain Monte Carlo methods in population genetics. Proceedings of the National Academy of Sciences of the United States of America 104, 2785-2790. Hill R, Favor J, Hogan B, et al. (1991) Mouse small eye results from mutations in a paired-like homeobox-containing gene. Nature 354, 522-525. Hilton JD, Brady AK, Spaho A, Vize PD (2012) Photoreception and signal transduction in corals: proteomic and behavioural evidence for cytoplasmic calcium as a mediator of light responsivity. Biological Bulletin 223, 291-299. Hillis DM, Bull JJ (1991) Of genes and genomes. Science (Washington) 254, 528-558. Holloway PE, Nye HC (1985) Leeuwin Current and wind distributions on the southern part of the Australian North West shelf between January 1982 and July 1983. Australian Journal of Marine and Freshwater Research. 36, 123-137. Huelsenbeck JP, Bull JJ (1996) A likelihood ratio test to detect conflicting phylogenetic signal. Systematic Biology 45, 92-98. Hubisz MJ, Falush D, Stephens M, Pritchard JK (2009) Inferring weak population structure with the assistance of sample group information. Molecular Ecology Resources 9, 1322-1332. Hunter CL (1988) Environmental cues controlling spawning in two Hawaiian corals, Montipora verrucosa and M. dilatata. In: Proceedings of the 6th International Coral Reef Symposium, pp. 727-732, Townsville. Irwin DE (2012) Local adaptation along smooth ecological gradients causes phylogeographic breaks and phenotypic clustering. The American Naturalist 180, 35-49. Jakobsson M, Rosenberg NA (2007) CLUMPP: a cluster matching and permutation program for dealing with label switching and multimodality in analysis of population structure. Bioinformatics 23, 1801-1806.

101

Johnson MS, Black R (1998) Effects of isolation by distance and geographical discontinuity on genetic subdivision of Littoraria cingulata. Marine Biology 132, 295-303. Johnson MS, Black R (2006a) Effects of mode of reproduction on genetic divergence over large spatial and temporal scales in intertidal snails of the genus Bembicium Philippi (Gastropoda: Littorinidae). Biological Journal of the Linnean Society 89, 689-704. Johnson MS, Black R (2006b) Islands increase genetic subdivision and disrupt patterns of connectivity of intertidal snails in a complex archipelago. Evolution 60, 2498- 2506. Kenyon JC (1992) Sexual reproduction in Hawaiian Acropora. Coral Reefs 11, 37-43. Kenyon JC (1997) Models of reticulate evolution in the coral genus Acropora based on chromosome numbers: Parallels with plants. Evolution 51, 756-767. Kilian A, Wenzl P, Huttner E, et al. (2012) Diversity Arrays Technology: A generic genome profiling technology on open platforms. In: Data Production and Analysis in Population Genomics (eds. Pompanon F, Bonin A). Humana Press. Knowlton N (2000) Molecular genetic analyses of species boundaries in the sea. Hydrobiologia 420, 73-90. Knowlton N, Mate JL, Guzman HM, Rowan R, Jara J (1997) Direct evidence for reproductive isolation among the three species of the Montastrea annularis complex in Central America (Panama and Honduras). Marine Biology 127, 705- 711. Knowlton N, Weil E, Weigt LA, Guzman HM (1992) Sibling species in Montastrea annularis, coral bleaching, and the coral climate record. Science (Washington) 255, 330-333. Ladner JT, Palumbi SR (2012) Extensive sympatry, cryptic diversity and introgression throughout the geographic distribution of two coral species complexes. Molecular Ecology 21, 2224-2238. Lamichhaney S, Berglund J, Sallman Almen M, et al. (2015) Evolution of Darwin’s finches and their beaks revealed by genome sequencing. Nature 518, 371-375 Leache AD, Banbury BL, Felsenstein J, Nieto-Montes de Oca A, Stamatakis A (2015) Short tree, long tree, right tree, wrong tree: new acquisition bias corrections for inferring SNP phylogenies. Systematic Biology 64, 1032-1047 Levitan DR, Fukami H, Jara Z, et al. (2004) Mechanisms of reproductive isolation among sympatric broadcast spawning corals of the Montastraea annularis species complex. Evolution 58, 308-323. Levitan DR, Fogarty ND, Jara J, Lotterhos KE, Knowlton N (2011) Genetic, spatial, and temporal components of precise spawning synchrony in reef building corals of the Montastraea annularis species complex. Evolution 65, 1254-1270. Levy O, Appelbaum L, Leggat W, et al. (2007) Light-responsive cryptochromes from a simple muticellular animal, the coral Acropora millepora. Science (Washington) 318, 467-470. Levy O, Dubinsky Z, Achituv Y (2003) Photobehaviour of stony corals: responses to light spectra and intensity. Journal of Experimental Biology 206, 4041-4049. Librado P, Rozas J (2009) DnaSP v5: A software for comprehensive analysis of DNA polymorphism data. Bioinformatics 25, 1451-1452. Liu L, Yu L, Kubatko L, Pearl DK, Edwards SV (2009) Coalescent methods for estimating phylogenetic trees Molecular Phylogenetics and Evolution 53, 320- 328.

102

Maddison WP (1997) Gene trees in species trees. Systematic Biology 46, 523-536 Mallet J (1995) A species definition for the modern synthesis. Trends in Ecology & Evolution 10, 294-299. Mangubhai S, Harrison PL (2006) Seasonal patterns of coral reproduction on equatorial reefs in Mombassa, Kenya. In: Proceedings of the 10th International Coral Reef Symposium, pp. 106-114, Okinawa. Mascarenhas D, Mettler IJ, Pierce DA, Lowe HW (1990) Intron-mediated enhancement of heterologous gene expression in maize. Plant Molecular Biology 15, 913-920. Marquez LM, van Oppen MJH, Willis BL, Reyes A, Miller DJ (2002) The highly cross- fertile coral species, Acropora hyacinthus and Acropora cytherea, constitute statistically distinguishable lineages. Molecular Ecology 11, 1339-1349. Matus DQ, Pang K, Daly M, Martindale MQ (2007) Expression of Pax gene family members in the anthozoan cnidarian, Nematostella vectensis. Evolution and Development 9, 25-38. McCormack JE, Hird SM, Zellmer AJ, Carstens BC, Brumfield RT (2013) Applications of next-generation sequencing to phylogeography and phylogenetics. Molecular Phylogenetics and Evolution 66, 526-538. McNeeley JA, Miller KR, Reid RA, Mittermeier WV, Werner TB (1990) Conserving the world's biological diversity. World Resources Institute, Washington, DC. Mendes JM, Woodley JD (2002) Timing of reproduction in Montastraea annularis : relationship to environmental variables. Marine Ecology Progress Series 227, 241-251. Meirmans PG (2006) Using the AMOVA framework to estimate a standardized genetic differentiation measure. Evolution 60, 2399-2402. Mildner S (1991) Aspects of the reproductive biology of selected scleractinian corals on Western Somoan and Fijian reefs Masters, James Cook University. Miller DJ, Hayward DC, Reece-Hoyes JS, et al. (2000) Pax gene diversity in the basal cnidarian Acropora millepora (Cnidaria, Anthozoa): Implications for the evolution of the Pax gene family. Proceedings of the National Academy of Sciences of the United States of America 97, 4475-4480. Nadeau NJ, Jiggins CD (2010) A golden age for evolutionary genetics? Genomic studies of adaptation in natural populations. Trends in Genetics 26, 484-492. Nakajima Y, Nishikawa A, Iguchi A, Sakai K (2012) The population genetic approach delineates the species boundary of reproductively isolated corymbose acroporid corals. Molecular Phylogenetics and Evolution 63, 527-531 Nater A, Burri R, Kawakami T, Smeds L, Ellegren H (2015) Resolving evolutionary relationships in closely related species with whole-genome sequencing data. Systematic Biology 64, 1000-1017. Nielson R (2005) Molecular signatures of natural selection. Annual Review of Genetics 39, 197-218. Nof D, Pichevin T, Sprintall J (2002) "Teddies" and the origin of the Leeuwin Current. Journal of Physical Oceanography 32, 2571-2573. Nosil P, Schluter D (2011) The genes underlying the process of speciation. Trends in Ecology and Evolution 26, 160-167. Nozawa Y (2012) Annual variation in the timing of coral spawning in a high-latitude environment: Influence of temperature. Biological Bulletin 222, 192-202. Olive PJW, Pillai G (1983) Reproductive biology of the polychaete Kefersteinia cirrata Keferstein (Hesionidae). II. The gametogenic cycle and evidence for

103

photoperiodic control of oogenesis. Invertebrate Reproduction & Development 6, 307-315. Oliver J, Babcock R (1992) Aspects of the fertilisation ecology of broadcast spawning corals: sperm dilution affects and in situ measurments of fertilisation. Biological Bulletin 183, 409-417. Oliver JK, Babcock RC, Harrison PL, Willis BL (1988) Geographic extent of mass coral spawning: Clues to ultimate causal factors, 803-810. Oliveri P, Fortunato AE, Petrone L, et al. (2014) The cryptochrome/photolyase family in aquatic organisms. Marine Genomics 14, 23-37. Orr HA, Masly JP, Presgraves DC (2004) Speciation genes. Current Opinion in Genetics and Development 14, 675-679 Orr HA (2005) The genetic basis of reproductive isolation: insights from Drosophila. Proceedings of the National Academy of Sciences of the United States of America 102, 6522-6526. Orton JH (1920) Sea temperature, breeding and distribution in marine animals. Journal of Marine Biology Association U.K. 12, 339-366. Paetkau D, Waits L, Clarkson P, Craighead L, Strobeck C (1997) An empirical evaluation of genetic distance statistics using microsatellite data from bear (Ursidae) populations. Genetics 147, 1943-1957. Palumbi SR (1992) Marine speciation on a small planet. Trends in Ecology & Evolution 7, 114-118. Palumbi SR (1994) Genetic divergence, reproductive isolation, and marine speciation Annual Review of Ecology, Evolution and Systematics 25, 547-572. Peakall R, Smouse PE (2006) GENALEX 6: genetic analysis in Excel. Population genetic software for teaching and research. Molecular Ecology Notes 6, 288- 295. Peakall R, Smouse PE (2012) GenAlEx 6.5: genetic analysis in Excel. Population genetic software for teaching and research-an update. Bioinformatics 28, 2537- 2539. Penland L, Kloulechad J, Idip D, van Woesik R (2004) Coral spawning in the western Pacific Ocean is related to solar insolation: evidence of multiple spawning events in Palau. Coral Reefs 23, 133-140. Permata D, Indrayanti E, Haryanti D, et al. (2012) Biannual multispecific spawning in Karimunjawa Archipelago, Indonesia. Coral Reefs DOI 10.1007/s00338-012- 0909-9. Pinzon JH, LaJeunesse TC (2010) Species delimitation of common reef corals in the genus Pocillopora using nucleotide sequence phylogenies, population genetics and symbiosis ecology. Molecular Ecology 20, 311-325. Planet PJ (2006) Tree disagreement: Measuring and testing incongruence in phylogenies. Journal of Biomedical Informatics 39, 86-102 Pritchard JK, Stephens M, Donnelly P (2000) Inference of population structure using multilocus genotype data. Genetics 155, 945-959. Pritchard JK, Wen X, Falush D (2010) Documentation for structure software: Version 2.3.http://pritchardlab.stanford.edu/structure_software/release_versions/v2.3.4/st ructure_doc.pdf. Provan J, Bennett KD (2008) Phylogeographic insights into cryptic glacial refugia. Trends in Ecology and Evolution 23, 564-571.

104

Quiring R, Walldorf U, Kloter U, Gehring W (1994) Homology of the eyeless gene of Drosophila to the Small eye gene in mice and Aniridia in humans. Science (Washington) 265, 785-789. Raymond M, Rousset F (1995) Population genetics software for exact tests and ecumenicism. Journal of Heredity 86, 248-249. Reitzel AM, Tarrant AM, Levy O (2013) Circadian clocks in the Cnidaria: Environmental entrainment, molecular regulation and organismal outputs. Intergrative and Comparative Biology 53, 118-130. Rieseberg LH, Blackman BK (2010) Speciation genes in plants Annals of Botany.10.1093/aob/mcq126 Rice W (1989) Analyzing tables of statistical tests. Evolution 43, 223-225. Richards ZT, van Oppen MJH, Wallace CC, Willis BL, Miller DJ (2008) Some rare Indo-Pacific coral species are probable hybrids. PLoS ONE 3, e3240. Richards ZT, Miller DJ, Wallace CC (2013) Molecular phylogenetics of geographically restricted Acropora species: implications for threatened species conservation. Molecular Phylogenetics and Evolution 69, 837-851 Richards ZT, Berry O, van Oppen MJH (2016) Cryptic genetic divergence within threatened species of Acropora coral from the Indian and Pacific Oceans. Conservation Genetics doi:10.1007/s10592-015-0807-0 Ronquist F, Huelsenbeck JP (2003) MrBayes 3: Bayesian phylogenetic inference under mixed models. Bioinformatics 19, 1572-1574. Rose AB, Elfersi T, Parra G, Korf I (2008) Promoter-proximal introns in Arabidopsis thaliana are enriched in dispersed signals that elevate gene expression. Plant Cell 20, 543-555. Rosenberg NA (2004) Distruct: a program for the graphical display of population structure. Molecular Ecology Notes 4, 137-138. Rosser NL (2013) Biannual coral spawning decreases at higher latitudes on Western Australian reefs. Coral Reefs 32, 455-460. Rosser NL (2015) Asynchronous spawning in sympatric populations of a hard coral reveals cryptic species and ancient genetic lineages. Molecular Ecology 24, 5006-5019. Rosser NL (2016) Asynchronous spawning and demographic history shape genetic differentiation amoung populations of the hard coral Acropora tenuis in Western Australia. Molecular Phylogenetics and Evolution 98, 89-96 Rosser NL, Baird AH (2009) Multi-specific coral spawning in spring and autumn in far north-western Australia. In: Proceedings of the 11th International Coral Reef Symposium, pp. 366-370, Florida. Rosser NL, Gilmour JP (2008) New insights into patterns of coral spawning on Western Australian reefs. Coral Reefs 27, 345-349. Rubin BER, Ree RH, Moreau CS (2012) Inferring phylogenies from RAD sequence data. PLoS ONE 7, e33394. Santodomingo N (2014) Miocene reef-coral diversity of Indonesia: unlocking the murky origins of the Coral Triangle PhD thesis, Utrecht University. Savolainen V, Anstett MC, Lexer C, et al. (2006) Sympatric speciation in palms on an oceanic island. Nature 441, 210-213. Schmidt HA, Strimmer K, Vingron M, von Haeseler A (2002) TREE-PUZZLE: maximum likelihood phylogenetic analysis using quartets and parallel computing. Bioinformatics, 502-504.

105

Schmidt-Roach S, Lundgren P, Miller KJ, Gerlach G, Noreen AME, Andreakis N (2013) Assessing hidden species diversity in the coral Pocillopora damicornis from Eastern Australia. Coral Reefs 32, 161-172. Schmidt-Roach S, Miller KJ, Lundgren P, Andreakis N (2014) With eyes wide open: a revision of species within and closely related to the Pocillopora damicornis species complex (Scleractinia; Pocilloporidae) using morphology and genetics. Zoological Journal of the Linnean Society 170, 1-33. Schmidt-Roach S, Miller K, Woolsey E, Gerlach G, Baird AH (2012) Broadcast spawning by Pocillopora species on the Great Barrier Reef. PLoS ONE 7, e50847. Selkoe KA, Toonen RJ (2006) Microsatellites for ecologists: a practical guide to using and evaluating microsatellite markers. Ecology Letters 9, 615-629. Shearer TL, van Oppen MJH, Romano SL, Worheide G (2002) Slow mitochondrial DNA sequence evolution in the Anthozoa (Cnidaria). Molecular Ecology 11, 2475-2487. Shimodaira H, Hasegawa M (1999) Multiple comparisons of log-likelihoods with applications to phylogenetic inference. Molecular Biology and Evolution 16, 114-1116. Shoguchi E, Tanaka M, Shinzato C, Kawashima T, Satoh N (2013) A genome-wide survey of photoreceptor and circadian genes in the coral, Acropora digitifera. Gene 515, 426-431. Simmons MP, Ochoterena H (2000) Gaps as characters in sequence based phylogenetic analyses. Systematic Biology 49, 369-381. Simpson CJ (1988) Coral reproduction and mass spawning. In: Ecology of scleractinian corals in the Dampier archipelago, Western Australia. Environmental Protection Authority, Perth, Western Australia. Simpson CJ (1991) Mass spawning of corals on Western Australian reefs and comparisons with the Great Barrier Reef. Journal of the Royal Society of Western Australia 74, 85-91. Simpson GG, Dean C (2002) Arabidopsis, the rosetta stone of flowering time? Science (Washington) 296, 285-289. Stamatakis A (2014) RAxML version 8: a tool for phylogenetic analysis and post- analysis of large phylogenies. Bioinformatics 30, 1312-1313 Stankowski S, Streisfield MA (2015) Introgressive hybridization facilitates adaptive divergence in a recent radiation of monkeyflowers. Proceedings of the Royal Society of London, Series B: Biological Sciences 282 doi:10.1098/rspb.2015.1666 Stanton ML, Galen C, Shore J (1997) Population structure along a steep environmental gradient: consequences of flowering time and habitat variation in the snow buttercup, Ranunculus adoneus. Evolution 51, 79-94. Stobart B (1994) Delimiting coral species using alternative techniques: Montipora digitata (Dana, 1846), a case study PhD thesis, James Cook University. Stobart B, Babcock RC, Willis BL (1992) Biannual spawning of three species of scleractinian coral from the Great Barrier Reef. In: Proceedings of the 7th international coral reef symposium, pp. 494-499, Guam. Styan CA, Rosser NL (2012) Is monitoring for mass spawning events in coral assemblages in north Western Australia likely to detect spawning? Marine Pollution Bulletin 64, 2523-2527.

106

Sunnucks P (2000) Efficient genetic markers for population biology. Trends in Ecology and Evolution 15, 199-203. Sweeney AM, Boch CA, Johnsen S, Morse DE (2011) Twilight spectral dynamics and the coral reef invertebrate spawning response. The Journal of Experimental Biology 214, 770-777. Szmant AM (1986) Reproductive ecology of Caribbean reef corals. Coral Reefs 5, 43- 54. Tajima F (1989) Statistical method for testing the neutral mutation hypothesis by DNA polymorphism. Genetics 123, 585-595. Tamura K, Peterson D, Peterson N, Stecher G, Nei M, Kumar S (2011) MEGA5: Molecular evolutionary genetics analysis using maximum likelihood, evolutionary distance, and maximum parsimony methods. Molecular Biology and Evolution 28, 2731-2739. Tamura K, Stecher G, Peterson D, Filipski A, Kumar S (2013) MEGA6: Molecular Evolutionary Genetics Analysis version 6.0. Molecular Biology and Evolution 30, 2725-2729. Todd PA (2008) Morphological plasticity in scleractinian corals Biological Reviews 83, 315-337. Treml EA, Halpin PN, Urban DL, Pratson LF (2008) Modeling population connectivity by ocean currents, a graphic-theoretic approach for marine conservation. Landscape Ecology 2008, 19-36. Turner TL, Hahn MW, Nuzhdin SV (2005) Genomic islands of speciation in Anopheles gambiae. PLoS Biology 3, 1572-1578. Underwood JN (2009a) Data from: Genetic diversity and divergence among coastal and offshore reefs in a hard coral depend on geographic discontinuity and oceanic currents.doi:10.5061/dryad.kr5j71n0 Underwood JN (2009b) Genetic diversity and divergence among coastal and offshore reefs in a hard coral depend on geographic discontinuity and oceanic currents. Evolutionary Applications 2, 222-233. van Oosterhout C, Hutchinson WF, Wills DPM, Shipley P (2004) Micro-checker: software for identifying and correcting genotyping errors in microsatellite data. Molecular Ecology Notes 4, 535-538. van Oppen MJH, Willis BL, van Vugt HWJA, Miller DJ (2000) Examination of species boundaries in the Acropora cervicornis group (Scleractinia, Cnidaria) using nuclear DNA sequence analyses. Molecular Ecology 9, 1363-1373. van Oppen MJH, McDonald BJ, Willis BL, Miller DJ (2001) The evolutionary history of the coral genus Acropora (Scleractinia, Cnidaria) based on mitochondrial and a nuclear marker: reticulation, incomplete lineage sorting or morphological convergence? Molecular Biology and Evolution 18, 1315-1329. van Oppen MJH, Koolmees EM, Veron JEN (2004) Patterns of evolution in the scleractinian coral genus Montipora (). Marine Biology 144, 9-18. van Woesik R (2010) Calm before the spawn: global coral spawning patterns are explained by regional wind fields. Proceedings of the Royal Society of London, Series B: Biological Sciences 277, 715-722. van Woesik R, Lacharmoise F, Koksal S (2006) Annual cycles of solar insolation predict spawning times of Carribbean corals. Ecology Letters 9, 390-398. Vargas-Angel B, Colley SB, Hoke SM, Thomas JD (2006) The reproductive seasonality and gametogenic cycle of Acropora cervicornis off Broward County, Florida, USA. Coral Reefs 25, 110-122.

107

Veron JEN (1995) Corals in space and time: the biogeography and evolution of the Scleractinia Cornell University Press, Ithaca, London. Via S (2009) Natural selection in action during speciation. Proceedings of the National Academy of Sciences of the United States of America 106, 9939-9946. Visser ME, Caro SP, van Oers K, Helm B (2010) Phenology, seasonal timing and circannual rhythms: towards a unified framework. Philosophical Transactions of the Royal Society of London, Series B: Biological Sciences 365, 3113-3127. Vize PD (2009) Transcriptome analysis of the circadian regulatory network in the coral Acropora millepora. Biological Bulletin 216, 131-137. Vize PD, Hilton JD, Brady AK, Davies SW (2009) Light sensing and the coordination of coral broadcast spawning behaviour. In: Proceedings of the 11th international coral reef symposium, pp. 378-381, Ft. Lauderdale, Florida. Vize PD, Embesi JA, Nickell M, Brown DP, Hagman DK (2005) Tight temporal consistency of coral mass spawning at the Flower Garden Banks, Gulf of Mexico, from 1997–2003. Gulf of Mexico Science 23, 107-114. Vollmer S, Palumbi S (2002) Hybridization and the evolution of reef coral diversity. Science (Washington) 296, 2023-2025. Vollmer SV, Palumbi SR (2007) Restricted gene flow in the Caribbean staghorn coral Acropora cervicornis: implications for the recovery of endangered reefs. Journal of Heredity 98, 40-50. Voris HK (2000) Maps of Pleistocene sea levels in SoutheastAsia: shorelines, river systems and time durations. Journal of Biogeography 27, 1153-1167. Wallace CC (1985) Reproduction, recruitment and fragmentation in nine sympatric species of the coral genus Acropora. Marine Biology 88, 217-233. Wallace CC (1999) Staghorn corals of the world A revision of the genus Acropora CSIRO Publishing, Melbourne. Wallace CC, Bosellini FR (2014) Acropora (Scleractinia) from the Oligocene and Miocene of Europe: species longevity, origination and turnover following the Eocene-Oligocene transition. Journal of Systematic Palaeontology 12, 447-469. Wallace CC, Rosen BR (2006) Diverse staghorn corals (Acropora) in high-latitude Eocene assemblages: implications for the evolution of modern diversity patterns of reef corals. Proceedings of the Royal Society of London, Series B: Biological Sciences 273, 975-982. Wang S, Zhang L, Matz M (2009) Microsatellite characterization and marker development from public EST and WGS databases in the reef building coral Acropora millepora (Cnidaria, Anthozoa, Scleractinia). Journal of Heredity 100, 329-337. Warner PA, Van Oppen MJH, Willis BL (2015) Unexpected cryptic species diversity in the widespread coral Seriatopora hystrix masks spatial-genetic patterns of connectivity. Molecular Ecology 24, 2993-3008. Weir BS, Cockerham CC (1984) Estimating F-statistics for the analysis of population structure. Evolution 38, 1358-1370. Wilkinson C (2008) Status of Coral Reefs of the World: 2008 Global Coral Reef Monitoring Network and Reef and Rainforest Research Centre, Townsville, Australia Willis BL, Babcock RC, Harrison PL, Oliver JK (1985) Patterns in the mass spawning of corals on the Great Barrier Reef from 1981 to 1984 4, 343-348.

108

Willis BL, Babcock RC, Harrison PL, Wallace CC (1997) Hybridization and breeding incompatibilities within the mating systems of mass spawning reef corals. In: Proceedings of the 8th International Coral Reef Symposium, pp. 81-90, Panama. Willis BL, van Oppen MJH, Miller DJ, Vollmer SV, Ayre DJ (2007) The role of hybridization in the evolution of reef corals. Annual Review of Ecology, Evolution and Systematics 37, 489-517. Wilson BR (2013) The biogeography of the Australian north west shelf: environmental change and life's response. Elsevier. Wilson JR, Harrison PL (2003) Spawning patterns of scleractinian corals at the Solitary Islands - a high latitude coral community in eastern Australia. Marine Ecology Progress Series 260, 115-123. Woerner AE, Cox MP, Hammer MF (2007) Recombination-filtered genomic datasets by information maximization. Bioinformatics 23, 1851-1853. Wolstenholme JK (2003) Species boundaries in scleractinian corals: a case study of the Acropora humilis group, PhD Thesis, James Cook University. Wolstenholme JK (2004) Temporal reproductive isolation and gametic compatibility are evolutionary mechanisms in the Acropora humilis species group (Cnidaria; Scleractinia). Marine Biology 144, 567-582. Wu C (2001) The genic view of the process of speciation. Journal of Evolutionary Biology 14, 851-865 Yokoyama Y, Purcell A, Lambeck K, Johnston P (2001) Shore-line reconstruction around Australia during the Last Glacial Maximum and Late Glacial Stage. Quaternary International 83-85, 9-18.

109

110

Appendices

Appendix 2.1

Explanation of the Cumulative Binomial Probability Distribution

I used the binomial probability distribution because I was most interested in the probability of failing to detect spawning when sampling 5 colonies, to evaluate the error rate in my sampling design. As this probability will depend upon the proportion of colonies in each species that are actually spawning (which could be anywhere between 1 and 99%), I used the binomial probability distribution (see formula below). However, because I was interested in the probability of failing to detect spawning in at least one colony when sampling five colonies (rather than exactly one), I instead used the cumulative binomial distribution mathematical model (see formula below). This calculation was performed in MS Excel using the BINOMDIST function.

The formula for the binomial probability distribution from Fowler et al. (1999) is:

Where P = probability of outcome k = number of events (in this case 5) x = the probability of a stated outcome (in this case 1 colony detected) p = probability of a particular outcome (based on the proportion of colonies in the population that are spawning)

While the formula I used to calculate the cumulative binomial distribution was:

111

Appendix 3.1

Registration numbers for skeletal voucher specimens at the Western Australian Museum and corresponding GenBank Accession numbers for sequences. Only unique sequences were uploaded to GenBank, so repeat accession numbers indicate the same haplotype/allele.

Colony ID WAM Inferred PaxC clade Access. Access. Access. voucher spawning Number Number Number number season PaxC- Control Calmod. intron Region LOW_4 Z84437 Spring A KT447648 KT447675 KT447661 LOW_6 Z84438 Spring A KT447651 KT447677 KT447664 LOW_61 n/a Spring A KT447643 KT447675 KT447660 LOW_67 n/a Spring A KT447644 KT447675 KT447665 LOW3_11 Z84443 Spring A KT447652 KT447675 KT447665 LOW3_13 Z84444 Spring A KT447642 KT447675 KT447665 LOW3_15 Z84445 Spring A KT447648 KT447675 KT447665 LOW3_17 Z84446 Spring A KT447645 n/a KT447665 LOW3_18 Z84447 Spring A KT447645 KT447675 KT447665 LOW3_3 n/a Spring A KT447645 KT447675 KT447665 LOW3_4 Z84439 Spring A KT447649 KT447677 KT447665 LOW3_5 Z84440 Spring A KT447645 KT447675 KT447661 LOW3_7 Z84441 Spring A KT447645 KT447678 n/a LOW3_8 Z84442 Spring A KT447653 KT447678 KT447665 LOW3_42 n/a Spring A KT447647 KT447682 KT447664 LOW_19 n/a Autumn A/B HET KT447657 KT447675 n/a LOW_3 Z84449 Autumn A/B HET KT447657 KT447675 n/a LOW3_20 Z84460 Autumn A/B HET KT447657 KT447675 n/a LOW_1 Z84448 Autumn B KT447654 n/a KT447658 LOW_10 Z84450 Autumn B KT447655 KT447681 KT447669 LOW_20 Z84451 Autumn B KT447655 KT447681 KT447670 LOW_23 n/a Autumn B KT447655 KT447678 KT447670 LOW_24 Z84461 Autumn B KT447654 KT447675 KT447665 LOW_26 Z84452 Autumn B KT447655 KT447679 n/a LOW_28 Z84463 Autumn B KT447656 KT447675 n/a LOW_32 Z84453 Autumn B KT447655 KT447675 KT447665 LOW_33 Z84462 Autumn B KT447655 KT447676 KT447674 LOW_62 n/a Autumn B KT447655 KT447675 KT447662 LOW_66 n/a Autumn B KT447655 KT447675 KT447667 LOW_68 Z84464 Autumn B KT447655 KT447675 KT447667 LOW_9 n/a Autumn B KT447655 KT447675 KT447665 LOW3_10 Z84456 Autumn B KT447655 KT447675 KT447671 LOW3_14 Z84457 Autumn B KT447655 KT447675 KT447672 LOW3_16 Z84458 Autumn B KT447655 KT447675 KT447671 LOW3_19 Z84459 Autumn B KT447655 n/a KT447672 LOW3_2 n/a Autumn B KT447655 KT447675 KT447662 LOW3_6 Z84454 Autumn B KT447655 KT447675 KT447674 LOW3_9 Z84455 Autumn B KT447655 KT447675 KT447669 112

Appendix 3.2

Characteristics of 13 microsatellite primers used in this study, developed by Wang et al. 2009. All primers annealed at 49°C

Reference Fluorescent No. Product Locus Primer sequence (5'-3') repeat motif label alleles size (bp) EST_016 CTATCTGTGTATGATCAGGACTA (AAC)7 FAM 2 97-111 TCCATCTGTTGTGGAAACTGGT

EST_032 AGGCACAAGAAAGTGGAAAACAA (TAA)21 FAM 3 119-125 TGAAGGGATGTGAAGCATGGT

EST_063 TATTGTAGTCGTTACGTAGGCT (TC)8 VIC 6 97-114 AACAATCGTGCATACTAGCTCA

EST_097 TGACAACGACATCAATCATGGT (TGA)7 PET 4 127-136 ACAGCAGGAGCTGTCAGCACT

EST_098 ACAAATTGCGCTCAAGTTGATG (TG)12 VIC 2 96-129 ACGGCTGCGAAGGAGTCTAGT

EST_149 ACGTCAAATGGATTTTCACATGA (GAT)9 PET 2 117-123 AGGTGCTTCTCTTTCCTCAGA

EST_181 TGATTGCTGAGAAAGCTAGAGAT (ATG)10 VIC 6 152-170 GCCTCACCTTGCCTTGTACA

EST_196 GTGTTGGCTATCTCATGTATAGT (TAA)9 VIC 19 131-179 ACAACACATCATCAACAACAGCA

EST_245 CAGAATGATATTTCTGCAGCACT (CA)10 FAM 5 116-123 CGCAATCGAGATTATAGGAAGA

EST_254 GGTGACCAATCAGAGTCTTGA (CA)12 PET 2 92-94 TACACTTGCTATAGTAACTTGCT

WGS_112 ACTCCACTCAGTCCTATTACCA (AAT)9 VIC 12 164-197 ACACTTCCAAGAGTCCCTACA

WGS_153 TTTCCAAGTTGCTGTGAGTACA (AATC)7 VIC 2 103-110 CGGGTGCTAAGCTTGCTCAA

WGS_211 TGACGACGAAACGTTGGCTAT (TAA)8 PET 3 181-191 AGACCGTTTCCTTTAACCAGAA

113

Appendix 3.3

Methods for primer design and sequencing of the protein-coding region of PaxC

A sequence of the PaxC transcription factor from Acropora millepora published on GenBank (Accession No. AF053459.2) was used to design forward and reverse primers in the program OLIGO version 7. Two sets of primers were trialed in the PCR process and the second set was used for final sequencing (Table 1). PCRs contained 21 µL Platinum PCR Supermix (Invitrogen), 1 µL each of the forward and reverse primers and 2 µL of DNA. Thermocycling profiles consisted of an initial denaturation step of 95°C for 3 min, followed by 35 cycles of 94°C for 30 sec, 49°C for 1 min and 72°C for 1 min, and finally 72°C for 10 min. Products were sequenced in both directions at BGI Hong Kong. Sequences were edited manually in Sequencher 4.5 and aligned using ClustalW in MEGA6. Unique sequences were uploaded to GenBank (Accession Numbers KT582778-KT582782).

Table 1. Primer sequences used for PCR and sequencing Primer 5’ to 3’ sequence PaxC-cds2_FP TTG CTG CAT GTG GCG TAA G PaxC-cds2_RF TTC CTG GAT TGC CTC GGT C

114

Appendix 3.4

Allele frequencies for 13 microsatellite loci in the spring- and autumn-spawning populations of Acropora samoensis. Sample size for each locus/population is given in brackets.

Locus-allele Spring Autumn

EST245 (12) (64) 116 0 0.031 117 0.042 0 120 0.917 0.844 121 0.042 0.094 123 0 0.031 WGS153 (13) (65) 103 0.962 0.977 110 0.038 0.023 WGS112 (13) (65) 164 0.115 0.015 167 0.115 0.038 170 0 0.015 173 0.192 0.023 176 0.385 0.046 179 0 0.454 182 0.038 0.308 185 0 0.015 188 0.077 0.008 191 0.038 0.069 194 0 0.008 197 0.038 0 EST149 (12) (63) 117 0 0.016 123 1 0.984 EST016 (13) (64) 96 0.115 0.031 99 0 0.586 102 0.577 0.273 105 0.077 0.047 108 0.231 0.039 111 0 0.023 EST098 (13) (64) 96 0 0.016 99 0.231 0.195 114 0.154 0.172 118 0.577 0.523 128 0 0.086 129 0.038 0.008 115

EST254 (13) (65) 92 0.269 0.938 94 0.731 0.062 EST196 (11) (64) 132 0 0.07 135 0.227 0.047 138 0.045 0.008 141 0.091 0.055 144 0 0.016 147 0.091 0.117 150 0.091 0.039 153 0.227 0.156 156 0.182 0.109 159 0 0.109 161 0 0.094 164 0 0.023 167 0 0.078 170 0 0.008 173 0 0.016 179 0 0.031 185 0 0.008 188 0 0.016 WGS211 (13) (53) 181 0 0.094 187 0 0.019 191 1 0.887 EST032 (13) (63) 119 0.154 0.008 122 0.846 0.976 125 0 0.016 EST063 (13) (63) 97 0 0.024 102 0 0.008 104 0.885 0.73 106 0.077 0.23 110 0.038 0 114 0 0.008 EST181 (13) (61) 152 0 0.09 158 0.269 0.041 161 0.077 0.016 164 0.385 0.721 167 0.269 0.025 170 0 0.107 EST097 (13) (65)

116

127 0.038 0.377 130 0.577 0.462 133 0.385 0.138 136 0 0.023

117

Appendix 4.1

Characteristics of microsatellite primers used in this study after Underwood 2009a

Fluoro Product size Locus Primer sequence (5'-3') Reference repeat motif label (bp) Amil2_006 CTTGACCTAAAAAACTGTCGTACAA (CA)4TA(CA)4 Vic - GTTATTACTAAAAAGGACGAGAATAACTTT Amil2_010 CAGCGATTAATATTTTAGAACAGTTTT TA(TG) 11 Fam 147 -151 CGTATAAACAAATTCCATGGTCTG Amil2_011 CACTCCTTACGCTGCTAGAT (CA) 2GA(CA)6CT Ned 147 -155 CTCGCTAAAATGAGAGACCA Amil2_012 TTTTAAAATGTGAAATGCATATGACA GA(CA) 6GA(CA)2 Pet 106 -111 TCACCTGGGTCCCATTTCT Amil2_018 GCCCTCCTTAGGTGATTTAC (CA) 9 Fam 355 -370 ATCGTTTTGAGCAATCAGAC Amil2_022 CTGTGGCCTTGTTAGATAGC (AC) 10 Vic 161 -180 AGATTTGTGTTGTCCTGCTT

Amil5_028 GGTCGAAAAATTGAAAAGTG (TCACA)7TCAC(TCA Ned 101-117 ATCACGAGTCCTTTTGACTG CA)4

118

Appendix 4.2

Summary of molecular diversity across loci showing sample size (n), gene diversity (HE), nucleotide diversity (π) and number of effective alleles (Neff). Heterozygote deficits in the microsatellite loci are indicated by *

Population n HE ± SD π ± SD Neff Control Region Ashmore 12 0.44 ± 0.16 0.0004 ± 0.0004 1.79 Scott Reef 10 0.87 ± 0.09 0.0012 ± 0.0009 7.69 Rowley Shoals 8 0.89 ± 0.11 0.0020 ± 0.0014 9.09 Kimberley 7 0.71 ± 0.18 0.0013 ± 0.0011 3.45 Montebello Is 12 0.76 ± 0.12 0.0026 ± 0.0016 4.17 Dampier 10 0.35 ± 0.16 0.0003 ± 0.0004 1.54 Ningaloo 11 0.49 ± 0.18 0.0006 ± 0.0006 1.96 PaxC Ashmore 11 0.89 ± 0.09 0.0069 ± 0.0043 9.09 Scott Reef 9 0.56 ± 0.17 0.0064 ± 0.0042 2.27 Rowley Shoals 11 0.60 ± 0.15 0.0047 ± 0.0031 2.50 Montebello Is 11 0.18 ± 0.14 0.0004 ± 0.0005 1.22 Dampier 10 0.20 ± 0.15 0.0004 ± 0.0006 1.25 Ningaloo 9 0.00 ± 0.00 0.0000 ± 0.0000 1.00 Flank Ashmore 8 0.86 ± 0.11 0.0070 ± 0.0045 7.14 Scott Reef 7 0.86 ± 0.14 0.0056 ± 0.0037 7.14 Rowley Shoals 8 0.54 ± 0.12 0.0020 ± 0.0016 2.17 Montebello Is 7 0.60 ± 0.18 0.0011 ± 0.0012 2.50 Dampier 7 0.67 ± 0.16 0.0030 ± 0.0023 3.03 Ningaloo 9 0.78 ± 0.11 0.0039 ± 0.0027 4.55 Microsats Ashmore 40 0.51 ± 0.21 n/a 2.04* Scott Reef 47 0.61 ± 0.10 n/a 2.56* Rowley Shoals 49 0.52 ± 0.16 n/a 2.08* Montebello Is 27 0.43 ± 0.14 n/a 1.75* Dampier 42 0.39 ± 0.15 n/a 1.64* Ningaloo 48 0.47 ± 0.14 n/a 1.89*

119

Appendix 5.1

GPS co-ordinates of sites from which each sample was collected in this study, and registration numbers for the Western Australian Museum where samples are housed. Samples annotated with _A = autumn spawner and _S = spring spawner. * = sample couldn’t be amplified in CR; ^ = sample couldn’t be amplified in PaxC; # = sample not included in SNP analysis.

Sample Species Sample Region Site name Lat Long Registration name code Number asp1 A. aspera K11 #29 Kimberley Mavis Reef S15.50517 E123.6082 Z65722 asp2 A. aspera K11 #75 Kimberley Fraser Is S16.05467 E123.3504 Z65735 asp3 A. aspera 49 Abrolhos Easter Group S28.681521 E113.8607 cyt1 A. cytherea K10 #2 Kimberley Cassini Island S13.95603 E125.6231 Z65641 cyt2 A. cytherea K12 #39 Kimberley Browse Is S14.11754 E123.5389 Z65792 cyt3 A. cytherea 363 Montebello S17 S20.7861 E115.5067 dig1 A. digitifera K11 #33a Kimberley Wild Cat Reef S15.28229 E124.105 Z65717a dig2 A. digitifera K11 # 33b Kimberley Wild Cat Reef S15.28229 E124.105 Z65717a dig3 A. digitifera K10 #73 Kimberley Cassini Island S13.95603 E125.6231 Z65671 div1 A. divaricata K10 #13 Kimberley Cassini Island S13.95603 E125.6231 Z65637 div2 A. divaricata K9 # 61 Kimberley Adele Is S15.32.429 E123.0766 Z65609 div3 A. divaricata K10 #80 Kimberley Cassini Island S13.95603 E125.6231 Z65673 don1 A. donei K10 #12 Kimberley Cassini Island S13.95603 E125.6231 Z65636 don2 A. donei K10 #253 Kimberley Long Reef S13.85676 E125.8248 Z65697 don3 A. donei K12 #386 Kimberley Browse Is S14.11754 E123.5389 Z65759 flo1 A. florida K12 #7 Kimberley Jameison Reef S14.06194 E125.3667 Z65765 flo2 A. florida K13 #78 Kimberley Ashmore S12.23728 E123.1600 Z66287 flo3 A. florida K11 #36 Kimberley Black Rocks S15.03889 E124.4278 Z65712 gem1 A. gemmifera K12 #33 Kimberley Browse Is S14.11754 E123.5389 Z65750 gem2 A. gemmifera K12 #381 Kimberley Browse Is S14.11754 E123.5389 Z65756 hum1 A. humilis 601 Montebello S6 S20.40527 E115.5818 hum2*^ A. humilis K12 #23 Kimberley Browse Is S14.11754 E123.5389 Z65748 hum3 A. humilis 388 Montebello S28 Dugong S20.9077 E115.4627 int1 A. intermedia K10 #134 Kimberley Cassini Island S13.95603 E125.6231 Z65689 int2*^ A. intermedia 27 Abrolhos Pelsaert Group S28.85259 E114.0120 int3 A. intermedia 161 Abrolhos Pelsaert Group S28.85259 E114.0120 los1* A. loisetteae 169 Abrolhos Pelsaert Group S28.85259 E114.0120 los2 A. loisetteae 63 Abrolhos Easter Group S28.681521 E113.8607 los3 A. loisetteae 78 Abrolhos Easter Group S28.681521 E113.8607 lut1 A. lutkeni K10 #44 Kimberley Cassini Island S13.95603 E125.6231 Z65647 lut2 A. lutkeni K10 #46 Kimberley Cassini Island S13.95603 E125.6231 Z65649

120

lut3*^ A. lutkeni K9 #160 Kimberley Montgomery Reef S15.52588 E124.1773 Z65624 mil1 A. millepora_A 501 Ningaloo North Ningaloo S22.168611 E113.865 mil2 A. millepora_A 502 Ningaloo North Ningaloo S22.168611 E113.865 mil3# A. millepora_A 500 Ningaloo North Ningaloo S22.1686 E113.865 mil4# A. millepora_A 503 Ningaloo North Ningaloo S22.1686 E113.865 mil5 A. millepora_S 215 Ashmore S2 S12.2455 E122.9867 mil6 A. millepora_S 216 Ashmore S2 S12.2455 E122.9867 mil7# A. millepora_S 201 Ashmore S2 S12.2455 E122.9867 mil8# A. millepora_S 207 Ashmore S2 S12.2455 E122.9867 mur1 A. muricata K10 #45 Kimberley Cassini Island S13.9560 E125.6231 Z65648 mur2 A. muricata K12 #439 Kimberley Heritage Reef S14.2545 E125.1596 Z65775 mur3 A. muricata K10 #129 Kimberley Cassini Island S13.9560 E125.6231 Z65684 pul1 A. pulchra K10 #79 Kimberley Cassini Island S13.9560 E125.6231 Z65654 pul2*^ A. pulchra K9 #121 Kimberley Montgomery Reef S15.5514 E124.1773 Z65621 pul3 A. pulchra K10 #128 Kimberley Cassini Island S13.9560 E125.6231 Z65683 sam1 A. samoensis_S LOW_61 Barrow Lowendal Shelf S20.7861 E115.5067 sam2 A. samoensis_S LOW_67 Barrow Lowendal Shelf S20.7861 E115.5067 sam3 A. samoensis_S LOW3_5 Barrow Lowendal Shelf S20.7861 E115.5067 Z84440 sam4 A. samoensis_S LOW3_7 Barrow Lowendal Shelf S20.7861 E115.5067 Z84441 sam5 A. samoensis_S LOW3_8 Barrow Lowendal Shelf S20.7861 E115.5067 Z84442 sam6 A. samoensis_S LOW3_11 Barrow Lowendal Shelf S20.7861 E115.5067 Z84443 sam7 A. samoensis_S LOW3_17 Barrow Lowendal Shelf S20.7861 E115.5067 Z84446 sam8 A. samoensis_S LOW3_18 Barrow Lowendal Shelf S20.7861 E115.5067 Z84447 sam9 A. samoensis_A LOW_20 Barrow Lowendal Shelf S20.7861 E115.5067 Z84451 sam10 A. samoensis_A LOW_23 Barrow Lowendal Shelf S20.7861 E115.5067 sam11 A. samoensis_A LOW_24 Barrow Lowendal Shelf S20.7861 E115.5067 Z84461 sam12 A. samoensis_A LOW_26 Barrow Lowendal Shelf S20.7861 E115.5067 Z84452 sam13 A. samoensis_A LOW_28 Barrow Lowendal Shelf S20.7861 E115.5067 Z84463 sam14 A. samoensis_A LOW_32 Barrow Lowendal Shelf S20.7861 E115.5067 Z84453 sam15 A. samoensis_A LOW3_6 Barrow Lowendal Shelf S20.7861 E115.5067 Z84454 sam16 A. samoensis_A LOW3_9 Barrow Lowendal Shelf S20.7861 E115.5067 Z84455 sel1 A. selago K10 #57 Kimberley Cassini Island S13.9560 E125.6231 Z65665 sel2 A. selago K10 #255 Kimberley Long Reef S13.8567 E125.8248 Z65698 sel3 A. selago 366 Montebello S31 S21.0444 E115.4702 spi1 A. spicifera K09 #90 Kimberley Adele Is S15.3242 E123.0766 Z65612 spi2 A. spicifera 70 Abrolhos Pelsaert Group S 28.8525 E114.0120 spi3 A. spicifera 343 Montebello S19 S20.5161 E115.4666 sto1*^ A. stoddarti K9 #113 Kimberley Adele Is S15.3242 E123.0766 Z65616 sto2 A. stoddarti K10 #158 Kimberley Cassini Island S13.9560 E125.6231 Z65677 sto3 A. stoddarti K10 #295 Kimberley Long Reef S13.8567 E125.8248 Z65706 sub1 A. subulata K10 #127 Kimberley Cassini Island S13.9560 E125.6231 Z65682 121

sub2 A. subulata K11 #215 Kimberley Fraser Is S16.0546 E123.3504 Z65738 sub3 A. subulata K10 #257 Kimberley Long Reef S13.8567 E125.8248 Z65700 ten1 A. tenuis_A 309a Montebello S19 S20.5161 E115.4666 ten2 A. tenuis_A 389a Montebello S19 S20.516111 E115.4666 ten3 A. tenuis_A 448a Montebello S19 S20.516111 E115.4666 ten4 A. tenuis_A 445a Montebello S19 S20.516111 E115.4666 ten5 A. tenuis_A Z327 Kimberley East Cassini Island S13.95603 E125.6231 Z65663 ten6 A. tenuis_A Z212 Kimberley Long Reef S13.85676 E125.8248 Z65695 ten7 A. tenuis_S Z59 Kimberley West Cassini Island S13.95603 E125.6231 Z65667 ten8 A. tenuis_S Z144 Kimberley SW Cassini Island S13.95603 E125.6231 Z65655 ten9 A. tenuis_A 427a Montebello S19 S20.516111 E115.4666 ten10 A. tenuis_S E6 Ashmore Lagoon S12.240833 E122.9805 ten11 A. tenuis_S F43 Ashmore Lagoon S12.240833 E122.9805 ten12 A. tenuis_S E48 Ashmore Lagoon S12.240833 E122.9805 ten13 A. tenuis_S E33 Ashmore Lagoon S12.240833 E122.9805 ten14 A. tenuis_A A32 Ashmore NE corner S12.184444 E123.1091 ten15 A. tenuis_S E16 Ashmore Lagoon S12.240833 E122.9805 ten16 A. tenuis_S 169 Ashmore Lagoon S12.240833 E122.9805

122

Appendix 5.2

Additional detail on methods described in Chapter 5

DArTseq

Genome-wide single nucleotide polymorphism (SNP) data was generated at Diversity

Arrays Technology (DArT P/L http://www.diversityarrays.com). DArTseq™ represents a combination of a DArT complexity reduction methods and next generation sequencing platforms. DArTseq methods are optimized for different organisms and applications by selecting the most appropriate complexity reduction method (size of the representation and the fraction of a genome selected for the assays). Four methods of complexity reduction were tested and the PstI-HpaII method was selected. Genomic DNA was processed in digestion/ligation reactions principally as per Kilian et al. (2012) but replacing a single PstI-compatible adaptor with two different adaptors corresponding to two different Restriction Enzyme (RE) overhangs. The PstI-compatible adapter was designed to include Illumina flowcell attachment sequence, sequencing primer sequence and “staggered”, varying length barcode region, similar to the sequence reported by

Elshire et al. (2011). Reverse adapter contained the flowcell attachment region and

HpaII-compatible overhang sequence. Sequencing was carried out on a single lane of an

Illumina Hiseq2500 and processed using proprietary DArT analytical pipelines. In the primary pipeline, the FASTQ files were first processed to filter away poor quality sequences. Approximately 2,500,000 sequences per barcode/sample were identified and used in marker calling. Identical sequences were collapsed into fastqcall files and these files were used in the secondary pipeline for DArT’s proprietary SNP calling algorithms

(DArTsoft14). Sequences were blasted against a Symbiodinium reference genome to

123

ensure that only sequences belonging to the coral host and not the symbiont were included in the dataset.

DaRTseq generates two types of data: (a) “Silico DaRT” which comprises presence/absence dominant markers based on a range of DNA variation types such as

SNPs, indels and methylation variation, and (b) SNPs in fragments of approximately

100 bp. In this study only the SNP data in fragments was used, and SNPs were extracted from each fragment and concatenated into supermatrices using IUPAC codes for heterozygous loci.

Phylogenetic analyses

For the CR and PaxC, the most appropriate model of DNA substitution was determined in MEGA6 (Tamura et al. 2013) using the Bayesian Information Criterion

(CR = HKY model, PaxC = K80), and these models were used in phylogenetic analyses run in PhyML 3.0 (Guindon et al. 2010). Support for each node was based upon 1000 bootstrap replicates. For the Bayesian analyses, MCMC chains were run for 3,000,000 generations (for each gene/SNP matrix) and sampled every 100th generation, with the first 7,500 runs discarded as burn-in. The PaxC alignment contained 14 indels

(including several large indels up to 390 bp), and the CR contained three large indels, and many of the indels were phylogenetically informative. Nevertheless, to take a conservative approach, each indel was coded as a single base change. The CR and PaxC phylogenetic trees were rooted with sequences from the sister genus Isopora, with sequences of I. cuneata obtained from GenBank (Accession No.s EU918925 and

AY026429).

124

References

Elshire RJ, Glaubitz JC, Sun Q, Poland JA, Kawamoto K, Buckler ES, Mitchell SE 2011. A Robust, Simple Genotyping-by-Sequencing (GBS) Approach for High Diversity Species. PLoS One, 6 , 19379-10.1371/journal.pone.0019379.

Kilian A, Wenz lP, Huttner E, Carling J, Xia L, et al.. 2012 Diversity Arrays Technology (DArT) - a generic genome profiling technology on open platforms. Methods in Molecular Biology Edited by Francois Pompanon and Aurelie Bonin, Humana Press: 67–91

Tamura K, Stecher G, Peterson D, Filipski A, Kumar S 2013 MEGA6: Molecular Evolutionary Genetics Analysis version 6.0. . Mol Biol Evol 30, 2725-2729.

Guindon S., Dufayard J.F., Lefort V., Anisimova M., Hordijk W., Gascuel O. 2010 New algorithms and methods to estimate maximum-likelihood phylogenies: assessing the performance of PhyML 3.0. Syst Biol 59(3), 307-321.

125

Appendix 5.3

Descriptive statistics of DArT single nucleotide polymorphism dataset for each Acropora species sampled (± SE).

Freq. Freq. Freq. Freq. Species N heterozygotes homozygote ref. missing data polymorphic loci A. aspera 3 0.028 0.736 0.160 0.182 A. cytherea 3 0.040 0.719 0.172 0.138 A. digitifera 3 0.037 0.731 0.118 0.163 A. divaricata 3 0.048 0.702 0.206 0.169 A. donei 3 0.036 0.721 0.171 0.155 A. florida 3 0.077 0.702 0.060 0.203 A. gemmifera 2 0.052 0.729 0.127 0.078 A. humilis 3 0.075 0.709 0.041 0.150 A. intermedia 3 0.047 0.719 0.094 0.134 A. loisetteae 3 0.026 0.735 0.151 0.078 A. lutkeni 3 0.023 0.732 0.166 0.166 A. millepora 4 0.044 0.725 0.104 0.115 A. muricata 3 0.020 0.727 0.211 0.180 A. pulchra 3 0.025 0.734 0.187 0.226 A. samoensis 16 0.049 0.716 0.088 0.226 A. selago 3 0.024 0.711 0.378 0.226 A. spicifera 3 0.029 0.723 0.150 0.226 A. stoddarti 3 0.100 0.681 0.071 0.226 A. subulata 3 0.042 0.720 0.128 0.226 A. tenuis 16 0.045 0.694 0.387 0.226 Min 0.020 0.681 0.041 0.078 Max 0.100 0.736 0.387 0.226

Mean 0.046 (± 0.006) 0.716 (± 0.004) 0.160 (± 0.019) 0.179 (± 0.010)

126

Appendix 5.4

Maximum Likelihood trees generated in RAxML from SNP matrices with genotype call rate of 100%, 90% and 70% and a minimum coverage of 8X. Blue stars represent spring spawners and red stars represent autumn spawners.

ML - 100% call rate

127

ML - 90% call rate

128

ML - 70% call rate

129

Appendix 5.5

Comparison of trees showing the placement of species within major clades (clades I, III or IV) and the placement of individuals within species (tight or split; MS = major split between clades; ms = minor split within a clade), and where discrepancies in topologies between CR or PaxC and the SNP tree occurred. The criteria for a discrepancy is where there is support for different topologies on the two trees, but does not apply where one tree is more highly resolved than another.

Discrepancy Species PaxC CR SNPs in topology A. aspera III+IV: split = MS+ms III+IV: split = MS+ms III+IV: split = MS+ms - A. cytherea III: tight IV: tight III: tight CR ≠ SNP A. digitifera III+IV: split = MS III+IV: split = MS III+IV: split = MS - A. divaricata III: split = ms III: split = ms III: split = ms CR ≠ SNP A. donei I+III: split = MS I+II+IV: split = MS I+III: split = MS CR ≠ SNP A. florida IV: split = ms IV: tight IV: split = ms - A. gemmifera IV: split = ms IV: tight IV: split = ms - A. humilis IV: tight IV: tight IV: tight - A. intermedia IV: tight IV: tight IV: tight - A. loisetteae III: tight III: tight III: tight - A. lutkeni III+IV: split = MS III+IV: split = MS III+IV: split = MS - A. millepora III: split = ms III: tight III: tight PaxC ≠ SNP A. muricata III: split = ms III+IV: split = MS III: split = ms CR ≠ SNP A. pulchra III: tight III+IV: split = MS III: tight CR ≠ SNP A. samoensis IV: split = ms IV: tight IV: split = ms - A. selago I: split = ms I: tight I: split = ms PaxC ≠ SNP A. spicifera III: tight III+IV: split = MS III: split = ms CR ≠ SNP A. stoddarti III: tight III: tight III: split = ms - A. subulata III: split = ms III+IV: split = MS III: split = ms CR ≠ SNP A. tenuis I: split = ms I: tight I: split = ms PaxC ≠ SNP

130