Iván F. Calixto Botía

Testing Ecological Speciation in the Caribbean Octocoral Complex bipinnata-kallos (: Octocorallia): An Integrative Approach

TESTING ECOLOGICAL SPECIATION IN THE CARIBBEAN OCTOCORAL COMPLEX -kallos (CNIDARIA: OCTOCORALLIA): AN INTEGRATIVE APPROACH

IVÁN FERNANDO CALIXTO BOTÍA, M.Sc.

A Doctoral dissertation submitted to the Department of Biological Sciences, Universidad de Los Andes, Colombia

as a requirement to obtain the degree of Doctor of Philosophy in Biological Sciences

Advisor University of Los Andes JUAN ARMANDO SÁNCHEZ, Ph.D.

Advisor University of Giessen THOMAS WILKE, Ph.D.

UNIVERSITY OF LOS ANDES 2018

Faculty Dean:

Prof. Dr. Ferney J. Rodríguez (University of Los Andes)

Advisors:

Prof. Dr. Juan A. Sánchez (University of Los Andes)

Prof. Dr. Thomas Wilke (Justus Liebig Universität)

Evaluators:

Prof. Dr. Oscar Puebla (University of Kiel)

Prof. Dr. Andrew Crawford (University of Los Andes)

Iván F. Calixto-Botía. (2018). Testing Ecological Speciation in the Caribbean Octocoral Complex Antillogorgia bipinnata-kallos (Cnidaria: Octocorallia): An integrative approach. This dissertation has been submitted as a requirement to obtain the degree of Doctor in Philosophy (Ph.D.) Biological Sciences at the Universidad de Los Andes, Colombia, advised by Professor Juan A. Sánchez (University of Los Andes, Colombia) and Professor Thomas Wilke (Justus Liebig Universität, Germany).

Table of content

Introduction ………………………………………………………………………………………………………………. 1

Chapter 1. A case of modular phenotypic plasticity in the depth gradient for the gorgonian coral Antillogorgia bipinnata (Cnidaria: Octocorallia) …………………………………………………… 9

Chapter 2. Testing Adaptive genetic divergence in parallel across the depth cline: Population genomics in the coral complex Antillogorgia bipinnata-kallos (Cnidaria: Octocorallia)…….………………………………………………………………………………………………………….17

Chapter 3. Coevolution in an ecological speciation scenario: fine-scale population genomics in the gorgonian coral complex Antillogorgia bipinnata-kallos and its symbiont…...…………………………………………...... 47

TESTING ECOLOGICAL SPECIATION IN THE CARIBBEAN OCTOCORAL COMPLEX Antillogorgia bipinnata-kallos (CNIDARIA: OCTOCORALLIA): AN INTEGRATIVE APPROACH

ABSTRACT

The ecological speciation is a central concept in evolution to differentiate one of the two big processes by which natural selection can produce new . Defined as the evolution of reproductive isolation by divergent natural selection in populations adapting to different ecological environments, the ecological speciation serves as a framework to test explicit predictions on the extent of natural selection to explain the diversification patterns we observe in the living forms. The present research aimed to test a putative ecological speciation scenario for two close related species where the phenotypic divergence of colonial forms overlaps between species along the depth cline. Antillogorgia bipinnata (Verrill 1864) and A. kallos (Bielschowsky 1918), conform a group of Caribbean corals where species pairs are distributed in sympatry along broad environmental ranges. We addressed this hypothesis by an integrative approach including a reciprocal transplants experiment, finding an adaptive plasticity response between depths, high survival rates and a genetic component explaining the variance of the traits assessed. Population genomic analyses with a pooling strategy detected an association between levels of genetic differentiation and habitats in four locations, implying parallel events of genetic divergence. Additionally, observations of reproductive asynchrony provided a potential mechanism for gene flow reduction. Finally, a fine-scale population genomic analysis remarkably supports the taxonomical status of the two species. Deepening in the role of plasticity, environmental mechanisms for gene flow reduction, macro- and micro-spatial genetic structure and the functional background of divergence, this research provides substantial elements to propose an ecological speciation scenario compatible with the diversification patterns for other marine organisms where the species can arise without evident barriers for the evolution of the adaptive genetic divergence.

RESUMEN

La especiación ecológica es un concepto central en teoría evolutiva al diferenciar uno de los dos grandes procesos por los cuáles la selección natural puede producir nuevas especies. Definida como la evolución del aislamiento reproductivo por selección natural divergente en poblaciones adaptadas a diferentes ambientes ecológicos, la especiación ecológica sirve como marco de trabajo para evaluar predicciones explícitas sobre el alcance de la selección natural para explicar los patrones de diversificación que observamos en las formas 1

vivientes. La presente investigación buscó evaluar un escenario putativo de especiación ecológica para dos especies cercanamente relacionadas donde la divergencia fenotípica de las formas coloniales se superpone entre las especies a lo largo de la clina de profundidad. Antillogorgia bipinnata (Verrill 1864) y A. kallos (Bielschowsky 1918) pertenecientes a la familia (Octocorallia), conforman un grupo de corales del Caribe donde los pares de especies se distribuyen en simpatría frente a amplios rangos ambientales. Nosotros direccionamos esta hipótesis mediante una aproximación integrativa incluyendo un experimento de trasplantes recíprocos, encontrando una respuesta plástica adaptativa entre profundidades, altas tasas de supervivencia y un componente genético explicando la varianza de los rasgos evaluados. Análisis en genómica poblacional con una estrategia de pooles detectaron asociación entre los niveles de diferenciación genética y los hábitats para cuatro regiones, implicando eventos paralelos de divergencia genética. Adicionalmente, observaciones de asincronía reproductiva proporcionaron un mecanismo potencial para la reducción del flujo génico. Finalmente, un análisis genómico a escala fina fuertemente sustenta el estatus taxonómico de las dos especies. Profundizando en el rol de la plasticidad, los mecanismos ambientales para la reducción del flujo génico, la estructura genética a escala macro y micro y el trasfondo funcional de la divergencia, esta investigación provee elementos sustanciales para proponer un escenario de especiación ecológica compatible con patrones de diversificación para otros organismos marinos, donde las especies pueden surgir sin barreras evidentes para la evolución de la divergencia genética adaptativa.

INTRODUCTION

A large number of studies have now emerged searching in its more intriguing details, the most inclusive theory in the field of Biological Sciences, the Natural Selection (Nielsen, Hellmann, Hubisz, Bustamante, & Clark, 2007; Rundle & Schluter, 2004). Developed by Charles Darwin and Russell Wallace, the relative role of the natural selection in the evolutionary process came to be recognized by the scientific community until the early 30 and 40’s with the evolutionary synthesis, however, testing the power of selection in speciation on in situ systems was limited in the following decades (Rundle & Schluter, 2004; Via & West, 2008). The renewed interest in recent years to assess in natural models the selective processes generating speciation is largely the result of molecular tools. Particularly, the fast advances in genomic sequencing approaches allow the recognition of heredity footprints left by selection in the evolution of species and allow to test explicit predictions about the evolution of reproductive isolation (Lee & Mitchell-Olds, 2006; Patrik Nosil, 2012; Wright & Andolfatto, 2008).

Two general kinds of natural selection can be distinguished: mutation-order speciation and ecological speciation. Mutation-order speciation refers to the evolution of reproductive isolation as a consequence of different mutations fixed in separate

2

populations adapting to similar environmental pressures (Mani & Clarke, 1990; Dolph Schluter, 2009). Ecological speciation can be defined as the process by which barriers to gene flow evolve between populations as a result of ecologically based divergent selection (D Schluter & Nagel, 1995; Dolph Schluter, 2009). In this scenario, the generation of new species is a final consequence of differentiated ecological pressures, where selection directly favors the evolution of reproductive isolation (Patrik Nosil, 2012). Therefore, it has been proposed that ecological speciation can occur in both sympatric and allopatric populations, where the reproductive isolation (pre or postzygotic) finally emerges as a byproduct of divergent populations (U Dieckmann, Metz, Doebeli, & Tautz, 2004; Rundle & Schluter, 2004). The ecological speciation extending on the contemporary patterns of diversification, mechanisms and rhythm, remains unknown. Particularly, the role of phenotypic plasticity as a force of adaptive divergence promoting speciation, given the different ecological pressures, is a key question in Evolutionary biology (Aubin-Horth & Renn, S., 2009; Moczek et al., 2011).

The present thesis aims to identify the ecological speciation process and propose as a research model the coral species complex Antillogorgia (=) bipinnata- kallos (Octocorallia: Gorgoniidae) (Williams & Chen, 2012) which, together other groups of Caribbean gorgonian corals with broad environmental status and sympatric distribution ranges, has been considered a case of marked phenotypic plasticity in an incipient ecological speciation process (Sánchez, Aguilar, Dorado, & Manrique, 2007). The complex A. bipinnata-kallos shows a bathymetric distribution from approximately 1 to 45m in the restricted locations where it is distributed in the Caribbean, with main populations in Panama, Belize, Bahamas, and Colombia (Figure 1 and Figure 2). In this depth gradient the species is under significant environmental variations as nutrient concentration, light intensity and water motion, exhibiting trait variation in coloration, form of sclerite (skeletal elements used as diagnostic character for Alcyonacean octocorals), height and the branching pattern. Particularly this last trait allows the differentiation of three basic morphotypes with distributions closely associated to depth ranges, the “deep morphotype”, “typical morphotype” and the “bushy morphotype” or “kallos”. The last was described as a distinct species, Antillogorgia (=Pseudopterogorgia) kallos (Bielschowsky 1918) and it has held this status (Bayer, 1961) even with the conspicuous similarity to A. bipinnata. The diffuse boundaries between the species at the taxonomical level were supported by molecular analyses showing doubtful genetic differentiation between the two species (Sánchez et al., 2007). Based on these observations, in the present study we refer to the A. bipinnata complex as the sensu stricto species plus the A. kallos as the shallower morphotype.

In some coral reefs of Colombia and Panama, in just 7 m of difference in the water column, the three morphotypes can be present when there is an abrupt change in depth

3

and reef slope, suggesting that this physiological challenge generates an adaptive morphological response. Genetic analyses for the A. bipinnata-kallos clade preliminary showed a complex pattern closer to geographic distance than to the kind of morphotype. Using mtDNA and ITS2 markers (Sánchez et al., 2007), a phylogeographic pattern was recognized for populations from Bahamas and Belize. An analysis with a larger dataset using ITS2 sequence data identified differences between kallos form populations from Belize and Panama, and differences between kallos morphotypes in Belize but not in Panama (Sánchez unpublished data).

As a surface brooder coral, the A. bipinnata complex has a potential dispersion of hundreds of meters in the water column, mediated by the interplay between gamete, larvae and ocean currents. Overall, the spatial scale between the different ecotypes of the A. bipinnata complex depict a system of overlapping geographical ranges, fitting to a sympatric distribution of the phenotypic divergence (Coyne & Orr, 2004; Knowlton, 1993). Considering the relation between dispersal abilities and spatial scales for depth distributions, our model has the potential to reduce the effect of the distance as a barrier for gene flow reduction between habitats, where the high gene flow across depth gradients is expected to homogenize the genetic variation, precluding local adaptation. At the same time, the spatial distribution of main populations across the Caribbean depict a highly limited dispersal potential (due hundreds of kilometers of distances involved) fitting more to an allopatry distribution mediated by distance. The global habitat/geographic scenario recognized for the complex of A. bipinnata resembles a parallel ecological speciation, a particular case made up of multiple distinct speciation events (D Schluter & Nagel, 1995). The replicated model represents an elucidatory component of the ecological speciation due it is unlikely that the same ecological barriers present in the different localities could independently produce a reduced gene flow just by chance (Luikart, England, Tallmon, Jordan, & Taberlet, 2003; P Nosil, Funk, & Ortiz-Barrientos, 2009; D Schluter & Nagel, 1995).

4

Fig 1. Morphotypes of the Antillogorgia bipinnata complex. For each morphotype the differences in branch pattern and sclerite form are displayed. Kallos morphotype (A), Intermedial (B) and Deep morphotype (C). Distribution of A. bipinnata morphotypes in four locations of the Caribbean (D). Modified from Sanchez et al 2007.

This thesis project seeks to carry out an assessment of ecological speciation for a parallel case in the A. bipinnata complex from an integrative approach. This involves the assessing of plastic components in the phenotypic divergence, potential mechanisms for gene flow reduction between habitats, the generation of the first draft genome for the species, a genomic survey of SNPs across localities in the Caribbean and on a fine scale including intermedial forms and the symbiont. Additionally, this integrative proposal has the potential to associate molecular results with traits of ecological and evolutionary interest, providing a better understanding of the functional roles and processes (Aubin-Horth & Renn, S., 2009).

Figure 2. Distribution of A. bipinnata complex in the three main areas of the Caribbean. There are available samples from 5 locations (red stars).

As first chapter of the thesis, we have tested the phenotypic plasticity role in the morphological variation related to the environment heterogeneity characterized by the depth (Calixto-Botía & Sánchez, 2017). A reciprocal transplant experiment was conducted, which has the power to define the degree of plasticity and the genetic substrate for selection. Reciprocal transplant is a method to infer ecological speciation via a direct measurement of divergent ecological selection on parental genotypes (e.g., immigrant inviability) in the different environments (Ostevik, Moyers, Owens, & Rieseberg, 2012).

5

Our results gave a clear signal of plasticity, and adaptive, as traits variated in correspondence with the new habitat. Particularly, deep colonies in shallow habitats had a conspicuous plastic response, resembling the kallos morphotype and supporting A. kallos as the extreme form in the continuum of the phenotypic divergence. Furthermore, the analysis of variance in the modular traits pointed to a genetic background, it means, the colonial traits defining morphotypes show a genetic control in an interaction with an environmental cue. This results plus high survival rates for foreign transplants seems to rule out immigrant inviability and support an intermediate scenario between adaptive plasticity and local adaptation to depth in the A. bipinnata complex.

Genomics is allowing to experimentally test different evolutionary models that leads to insight of the molecular processes underlying speciation, even those where is speculated that genetic divergence is in its first states (P Nosil et al., 2009). As second chapter of the thesis, a high throughput SNP screening derived from a reduced representation technique was performed to resolve the genetic structure between morphotypes and localities where the A. bipinnata complex is distributed. By an explicit test of Isolation By Adaptation (IBA), the focus of the chapter was to detect association between gene flow reduction and habitats for the kallos and deep morphotypes and to provide statistical robustness, a parallel mode of the process was tested including four locations in the Caribbean. We found genetic differentiation between morphotypes for each of the four localities, and non-monophyly of the same morphotype between environments, representing independent origins of the process. Additionally, a spawning monitoring evidenced an asynchronous gamete maturation between habitats, indicating a potential mechanism for gene flow reduction associated to depth. In the face of high potential interbreeding, the nonrandom gene flow between morphotypes could be explained by habitat adaptation acting in a concerted mode with isolation by allochrony. Therefore, based on a noticeable parallel process of IBA in the face of gene flow with an observed depth-associated spawning, we putatively infer an ecological speciation process promoted by the depth for the Antillogorgia bipinnata complex (Patrik Nosil, 2012; D Schluter & Nagel, 1995).

Finally, as third chapter of the thesis, we are performing a more in detail genetic structure analysis including the intermedial space of the divergence, different populations from previous study, individual screening with a high-resolution genomic technique and particularly, a differentiated analysis of the zooxanthellae genetic structure. Preliminary results remarkably confirm the IBA between habitats for the first of the new locations analyzed in San Andres Island, Colombia, with effective larval dispersal within meters of distance from the mother colony. Thus, the thesis provides important evidence to the ecological speciation theory via the evaluation of a mechanism where the species can arise without geographic barriers. In turn, this study is deepening in the role of plasticity,

6

environmental gradients, reinforcement mechanisms for gene flow reduction, micro spatial genetic structure for the coral and its symbiont and the functional meaning of genetic divergence (Funk, 1998; Funk, Filchak, & Feder, 2002; Patrik Nosil, 2007; D Schluter & Nagel, 1995).

References Aubin-Horth & Renn, S., N. (2009). Genomic reaction norms: using integrative biology to understand molecular mechanisms of phenotypic plasticity. Molecular Ecology, 18(18), 3763–3780. Bayer, F. M. (1961). The shalow water Octocorallia of the West Indian Region. In M. Nijhoff (Ed.), A manual for marine biologists (p. 373). The Hague. Calixto-Botía, I., & Sánchez, J. A. (2017). A case of modular phenotypic plasticity in the depth gradient for the gorgonian coral Antillogorgia bipinnata (Cnidaria: Octocorallia). BMC Evolutionary Biology, 17(1), 55. https://doi.org/10.1186/s12862- 017-0900-8 Coyne, J. A., & Orr, H. A. (2004). Speciation. Sunderland, Massachusetts, USA: Sinauer Associates, Inc. Dieckmann, U., Metz, J., Doebeli, M., & Tautz, D. (2004). Adaptative Speciation. (U. Dieckmann, M. Doebeli, J. A. J. Metz, & D. Tautz, Eds.), Cambridge Studies in Adaptative Dinamics (Vol. 1). London: CAMBRIDGE UNIVERSITY PRESS. Funk, D. J. (1998). Isolating a role for natural selection in speciation: Host adaptation and sexual isolation in Neochlamisus bebbianae leaf beetles. Evolution, 52, 1744–1759. Funk, D. J., Filchak, K. E., & Feder, J. L. (2002). Herbivorous insects: model systems for the comparative study of speciation ecology. Genetica, 116(2–3), 251–267. Knowlton, N. (1993). Sibling Species in the Sea. Annual Review of Ecology and Systematics, 24(1), 189–216. https://doi.org/10.1146/annurev.es.24.110193.001201 Lee, C. E., & Mitchell-Olds, T. (2006). Ecological and evolutionary genomics of populations in nature. Mol. Ecol., 15, 1103–1196. Luikart, G., England, P. R., Tallmon, D., Jordan, S., & Taberlet, P. (2003). The power and promise of population genomics: From genotyping to genome typing. Nat Rev Genet, 4, 981–994. Mani, G. S., & Clarke, B. C. (1990). Mutation order — A major stochastic process in evolution. Proceedings of the Royal Society B: Biological Sciences, 240, 29–37. Moczek, A. P., Sultan, S., Foster, S., Ledon-Rettig, C., Dworkin, I., Nijhout, H. F., … Pfennig, D. W. (2011). The role of developmental plasticity in evolutionary innovation. Proceedings of the Royal Society B: Biological Sciences, 278, 2705–2713. https://doi.org/10.1098/rspb.2011.0971 Nielsen, R., Hellmann, I., Hubisz, M., Bustamante, C., & Clark, A. G. (2007). Recent and ongoing selection in the human genome. Nature, 8, 857–868. Nosil, P. (2007). Divergent Host Plant Adaptation and Reproductive Isolation between

7

Ecotypes of Timema cristinae Walking Sticks. The American Naturalist, 169(2), 151– 162. https://doi.org/10.1086/510634 Nosil, P. (2012). Ecological Speciation. Oxford University Press. Nosil, P., Funk, D. J., & Ortiz-Barrientos, D. (2009). Divergent selection and heterogeneous genomic divergence. Mol. Ecol., 18(3), 375–402. https://doi.org/10.1111/j.1365- 294X.2008.03946.x Ostevik, K. L., Moyers, B. T., Owens, G. L., & Rieseberg, L. H. (2012). Parallel ecological speciation in plants? International Journal of Ecology. https://doi.org/10.1155/2012/939862 Rundle, H., & Schluter, D. (2004). Natural selection and ecological speciation in sticklebacks. In: Dieckmann U, Doebeli M, Metz JAJ, Tautz D (Eds) Cambridge University Press: Cambridge, 192–209. Sánchez, J., Aguilar, C., Dorado, D., & Manrique, N. (2007). Phenotypic plasticity and morphological integration in a marine modular invertebrate. BMC Evolutionary Biology, 7(1), 122. Retrieved from http://www.biomedcentral.com/1471-2148/7/122 Schluter, D. (2009). Evidence for Ecological Speciation and Its Alternative. Science, 323(5915), 737–741. https://doi.org/10.1126/science.1160006 Schluter, D., & Nagel, L. (1995). Parallel speciation by natural selection. The American Naturalist, 146(2), 292– 301. Via, S., & West, J. (2008). The genetic mosaic suggests a new role for hitchhiking in ecological speciation. Mol. Ecol., 17, 4334–4345. Williams, G. C., & Chen, J.-Y. (2012). Resurrection of the octocorallian Antillogorgia for Caribbean species previously assigned to Pseudopterogorgia, and a taxonomic assesment of the relationship of these genera with Leptogorgia (Cnidaria, , Gorgoniidae). ZOOTAXA, 3505, 39–52. Wright, S., & Andolfatto, P. (2008). The Impact of Natural Selection on the Genome: Emerging Patterns in Drosophila and Arabidopsis. Annu. Rev. Ecol. Evol. Syst, 39, 193– 213.

8

Calixto-Botía and Sánchez BMC Evolutionary Biology (2017) 17:55 DOI 10.1186/s12862-017-0900-8

RESEARCH ARTICLE Open Access A case of modular phenotypic plasticity in the depth gradient for the gorgonian coral Antillogorgia bipinnata (Cnidaria: Octocorallia) Iván Calixto-Botía1,2* and Juan A. Sánchez2,3

Abstract Background: Phenotypic plasticity, as a phenotypic response induced by the environment, has been proposed as a key factor in the evolutionary history of corals. A significant number of octocoral species show high phenotypic variation, exhibiting a strong overlap in intra- and inter-specific morphologic variation. This is the case of the gorgonian octocoral Antillogorgia bipinnata (Verrill 1864), which shows three polyphyletic morphotypes along abathymetricgradient.ThisresearchtestedthephenotypicplasticityofmodulartraitsinA. bipinnata with a reciprocal transplant experiment involving 256 explants from two morphotypes in two locations and at two depths. Vertical and horizontal length and number of new branches were compared 13 weeks following transplant. The data were analysed with a linear mixed-effects model and a graphic approach by reaction norms. Results: At the end of the experiment, 91.8% of explants survived. Lower vertical and horizontal growth rates and lower branch promotion were found for deep environments compared to shallow environments. The overall variation behaved similarly to the performance of native transplants. In particular, promotion of new branches showed variance mainly due to a phenotypic plastic effect. Conclusions: Globally, environmental and genotypic effects explain the variation of the assessed traits. Survival rates besides plastic responses suggest an intermediate scenario between adaptive plasticity and local adaptation that may drive a potential process of adaptive divergence along depth cline in A. bipinnata. Keywords: Phenotypic plasticity, Antillogorgia bipinnata, Reaction norm, Octocoral, Depth cline

Background will be hindered [2, 3]. However, the potential role of Phenotypic plasticity has been defined as the natural adaptive plasticity in promoting speciation has been capacity of an organism to react to a phenotypic change suggested in some cases where it can contribute to niche in form, state, movement or activity rate in response to diversification and further evolutionary change [1, 4–6]. environmental variation [1]. In the past century, pheno- Phenotypic and genetic accommodation, the Baldwin typic plasticity was largely considered a barrier to speci- effect, and the Waddington’s genetic assimilation have ation: if there is no need for genetic change to adapt to been proposed to explain environmental-induced changes the environment (masking the genotype for negative se- fixed in the genome and susceptible to promote a spe- lection), then the process of adaptive genetic divergence ciation process [7, 8]. Reciprocal transplant experiments consist of transfer- ring phenotypic variants to the opposite environments * Correspondence: [email protected] 1Department of Ecology and Systematics, Justus Liebig Universität, and are a practical and cost-effective approach to testing Heinrich-Buff-Ring 26-32 IFZ D-35392, Giessen, Germany phenotypic plasticity. In marine systems, variation re- 2 Laboratory of Biología Molecular Marina-Biommar, Department of Biological lated to environmental heterogeneity has been specially Sciences-Faculty of Sciences, Universidad de los Andes, Carrera 1E No 18A – 10, P.O. Box 4976, Bogotá, Colombia studied along depth gradients, which can vary in light Full list of author information is available at the end of the article

© The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Calixto-Botía and Sánchez BMC Evolutionary Biology (2017) 17:55 Page 2 of 8

intensity, wave exposure and nutrient concentration. In reassigned from the genus Pseudopterogorgia [19]. Popu- corals, most studies assessing plasticity and genetic lations of the complex A. bipinnata-kallos are clustered adaptation have found that adaptive divergence corre- in reefs in the Southern Caribbean, the Mesoamerican sponds to depth gradients. One of the first experiments Reef, with Panama and Belize populations closely related using a common garden approach on corals was done [20], and the Florida-Bahamas region. Colonies of the with Orbicella annularis and Siderastrea siderea, in four ‘deep’ morphotype are larger, with longer principal axes, reef habitats and detected high phenotypic plasticity in secondary branches, and internodal lengths, but fewer response to the transplanted habitat [9]. In another secondary branches in contrast to the ‘bushy’ morpho- study, shallow and deep morphotypes of Eunicea flex- type. These traits emerge from their modular organisa- uosa, a Caribbean gorgonian, exhibited low phenotypic tion, with the polyp as an iterative unit (sensu stricto) plasticity of sclerites and a strong genetic divergence sig- and branch as derived modular units [21, 22]. Thus, nal [10]. Finally, using reciprocal transplants of Seriato- differences in the architectural pattern between morpho- pora hystrix, a case of adaptive plasticity was detected in types shape the distribution and polyp density across the response to light conditions and adaptive divergence colony. Colony architecture has a feasible role in nutri- along the depth gradient [11, 12]. ent capture, overall photosynthetic rate, physical stress, Some Caribbean corals, including the feather-like and can be directly related to adaptive responses to en- gorgonian coral Antillogorgia bipinnata (Verrill 1864), vironmental variables in the depth cline. possess broad environmental preferences, sympatric dis- tributions, and contain highly plastic species complexes Study Area that are likely undergoing incipient ecological speciation The experiment was conducted at two localities in the processes [13–15]. A. bipinnata is distributed along northern Panamanian Caribbean, Hospital Point and coral reefs from 1 to 45 m deep in Panama, Belize, Crawl Key, Bocas del Toro (Fig. 1a). These locations ex- Bahamas, Florida, Colombia, but is absent on the hibit some environmental differences; Hospital Point is a Eastern side of the Western Atlantic. Along with this protected reef with abundant suspended particles and bathymetric gradient, the species varies phenotypically, low water motion compared to Crawl Key. 15 km away, including variation in size, coloration, sclerite form and Crawl Key is an exposed reef flat, characterised by branching pattern. Three basic morphotypes can be eroded coral skeletons and greater light penetration. recognised over a depth cline, the ‘deep morphotype’, Shallow habitats are characterised by coarser sand ‘typical morphotype’ and ‘bushy morphotype’ or ‘kallos’ with corals typically established on hard surfaces, such [15], the latter (A. kallos Bayer) described as a distinct as rocks. In contrast, in deep habitats, the slope is species [16, 17]. In some coral reefs (such as Panama) lower with greater sediment perturbation. Data for including at depths as shallow as 7 m, all three morpho- temperature and illuminance (total luminous flux per types can be present where there is an abrupt change in unit area) were collected using HOBO temperature depth and reef slope, suggesting that this physiological and light data loggers (Hobo Water Temp Pro, Onset challenge promotes an adaptive morphological response. Computer Corp., Bourne, Mass). In the shallow habi- To identify the contribution of plasticity to trait vari- tat at both localities, the temperature ranged between ation in corals emerging from the colonial structure, a 26–31 °C in comparison to 26–30 °C for deep habitats. reciprocal transplant experiment was carried out be- Illuminance strongly varied between habitats with the tween the deep and bushy morphotypes of A. bipinnata greatest variation in shallow habitats of Crawl Key (0– from two locations in Bocas del Toro, Panama. With this 35800 lux), and Hospital Point (0–30300 lux). In deep research, we measured modular traits related to habitats, illuminance varied from 0–14400 lux for bathymetric adaptation to detect the genotypic and en- Crawl Key and 0–2900 lux at Hospital Point. vironmental components involved in colonial structure variation between morphotypes of A. bipinnata. These Establishing reciprocal transplants and data collection data deepen our understanding of the evolutionary To assess the response of modular traits after transplant- mechanisms and patterns of diversification for a remark- ation, we took advantage of the overcompensation able number of species from this subclass that show phenomenon in A. bipinnata, a fast apical branching re- marked phenotypic variation related to environmental sponse following injury [23]. Fragments approximately gradients [17, 18]. 30 cm long were cut from 32 healthy colonies of the bushy and deep morphotypes at Crawl Key and Hospital Methods Point. The 16 colonies collected from each locality were Antillogorgia bipinnata at least 12 m apart from each other, a total of 21 bushy Based on the molecular and morphological evidence, and 11 deep morphotypes. In a fully factorial design, we Antillogorgia bipinnata (Verrill) has been recently transplanted 256 coral segments across shallow (2–3 m) Calixto-Botía and Sánchez BMC Evolutionary Biology (2017) 17:55 Page 3 of 8

ab

Fig. 1 Locations and design of the reciprocal transplant experiment. a Map of Bocas del Toro, Panama, signalling the localities of Hospital point (black star) and Crawl key (red circle) (basemap from https://google.com/maps/). b Experiment design with arrows indicating the direction of transplants between habitats and localities with curved arrow for native controls. Eight segments per colony were used to get a fully crossed and replicated design

and deep (7–8.5 m) habitats in both locations, one we assessed the effect of the response variables both control kept in the native habitat and three experi- separately and jointly over trait performances. Re- mental units in two replicate groups (Fig. 1b). Seg- sponse variables (vertical length, horizontal length, ments from the colony source were randomly and new branch promotion) were regressed on the assigned to destinations. To firmly attach the ex- predictors morphotype (two levels, bushy and deep) plants, each was fixed to a piece of PVC-pipe with and target habitat (two levels, shallow and deep) as PVC-clamps [23]. To collect data on focal colonial well as on their interactions. Location (Hospital Point traits, we took pictures (PowerShot G12, Canon®) fol- and Crawl Key) was included as a random effect to lowing initial transplantation, on July 13–14, 2011, account for environmental differences, as well as the and at the end of the experiment, on October 17, internal variance of morphotypes (genotypes nested 2011, 13 weeks later. Segments were considered dead into morphotypes). The significance of the model if high tissue loss was detected (>50%). A background terms was assessed using Akaike information criterion grid (white acrylic board) with the scale was used for (AIC) and calculated with the “dredge” function [25]. image correction and latermeasuringinthedigital AIC value below two suggests substantial evidence for processing. the model, values between 3 and 7 means that the model has considerably less support, and values over Digital processing and statistical analysis ten indicates that the model is very unlikely [26]. We used Adobe Photoshop® software (Adobe Sys- Fixed variables from all models with AIC values tems, San Jose, CA) to set pictures to a single op- below two, were examined for Beta, t-student and tical plane for perspective correction. ImageJ® p-values to test for significant relationships. The sig- software [24] was used for measuring the traits, by nificance values of Beta, t-student and p-values sup- transforming the scale from pixels to metric units ported the analysis of AIC test for a better [21]. We measured modular traits: (1) the vertical interpretation across each predictor assessed. In length variation by measuring the variation in length addition, the normality of data was examined with ex- of the main axis, (2) horizontal length variation by ploratory graphics of quantile-quantile plots. haphazardly selecting secondary branches (1688 A statistical significant morphotype effect indicates branches, x 9:4persegment)andmeasuringthe that genotype differences may explain the response. A ¼ variation in length at both time points, and (3) new significant target habitat effect on trait variation implies branches promotion generated after injury, in the plasticity, indicating consistent variation in the trait with growing apical segment as well as branches on old the environment. A significant morphotype by target secondary branches. habitat interaction indicates a genotype by environment Linear mixed-effects models (LMMs) were con- effect on the response (G X E), where variation in the structed to explain the variance in the traits measured degree of plasticity between morphs is detected. A between morphotypes and habitats. With this model, graphical approach to joint statistical analysis was Calixto-Botía and Sánchez BMC Evolutionary Biology (2017) 17:55 Page 4 of 8

performed using reaction norms of the three assessed indicated that length differed between deep and shallow traits using the median values (M) and Median Absolute habitats in the destination variable (β = −2.52 [SE = 1.09], Deviation (MAD) keeping in mind the asymmetric dis- t (d.f. = 71.9) = −2.30, p-value = 0.023). Reaction norms tributions of the resulting variance. Reaction norms for for vertical length variation (Fig. 2) show that higher new branches promotions were constructed using a values were common at shallow habitats, for example, density of branches (number of branches over a length the deep morphotype in shallow habitat at Crawl Key of the main axis) to get appropriate slopes in cases of (M = 18.23, MAD = 3.51). Except for segments from null promotion. All statistical analyses were performed deep habitat at Hospital Point that were transplanted to in R version 3.2.3 [27] using the lme4 package [28]. Crawl Key shallow habitat (Fig. 2e), the slopes between habitats were in the same direction and in some cases, Results close to parallel (Fig. 2a, d and f). Sixty one of the initial 256 segments were lost during The highest ranked model for horizontal length vari- photo recovery or analysis (e.g. it was not possible to ation based on AIC only contained the terms target identify the code of the colony, the metric references or habitat and morphotype. Horizontal length differed to recognise some of the traits). From the remaining 195 between deep and shallow destination (β = −2.54 [0.79], explants, 16 died (see Additional file 1), giving a 91.8% p = 0.001) along with to the type of morphotype (β = of survival and providing a large enough sample size to 3.24 [0.83], p = 0.0006). Compared to the positive values assess fixed factors. for vertical lengths, in most cases, horizontal lengths LMM, significance test and reaction norms assessed were more variable and reached negative values even for signs to target habitat together with morphotype as the native explants, indicating loss of secondary branch tis- main explanatory variables of the variance for the three sue (Fig. 3). Higher horizontal growth values were evi- modular traits of A. bipinnata, i.e. plasticity pattern, as dent in shallow habitats, with peaks in shallow natives well as a genetic component, explains the variance from Hospital Point (M = 2.77; MAD = 2.34) compared 13 weeks after transplantation. Q-Q plots did not show to deep natives from the Crawl Key locality (M = −6.35, overdispersion of data for any of the variables. Ranking MAD = 0.67). Most slopes were in the same direction of LMM highligted in each of the three variables target with the exception of Fig. 3f and nearly parallel as in habitat and morphotype as general explanators of vari- Fig. 3b and e (involving transplants between habitats ance for AIC values below 2 (Table 1). Therefore, for the and locations). response variable vertical length variation, predictable Promotion of new branches was higher in shallow hab- terms based on AIC contained all terms except morpho- itats compared to deep ones, as such the highest ranked type alone. Nevertheless, the significance test only model based on AIC contained the term target habitat

Table 1 Ranking of Linear Mixed-effects Model for three modular traits Trait Model AIC Δi d.f. Weight Vertical length variation TgHt, Mrph 673.3 0.00* 6 0.343 TgHt 673.5 0.19* 5 0.311 TgHt, Mrph, TgHt:Mrph 674.3 1.02* 7 0.205 Mrph 676.4 3.07 5 0.074 676.5 3.25 4 0.067 Horizontal length variation TgHt, Mrph 580.1 0.00* 6 0.742 TgHt, Mrph, TgHt:Mrph 582.4 2.25 7 0.240 Mrph 587.8 7.66 5 0.016 TgHt 592.3 12.15 5 0.002 598.6 18.50 4 0.000 New branches promotion TgHt 855.6 0.00* 5 0.427 TgHt, Mrph 856.2 0.52* 6 0.329 TgHt, Mrph, TgHt:Mrph 856.9 1.22* 7 0.232 863.9 8.30 4 0.007 Mrph 864.5 8.84 5 0.005 Each model incorporates both fixed- and random—effects terms in the linear predictor expression, from which the conditional mean of the response can be evaluated. AIC Akaike information criterion, Δi delta in AIC score with respect to the best model, d.f. degrees of freedom TgHt Target Habitat; Mrph, Morphotype. AIC values below two are marked with asterisks Calixto-Botía and Sánchez BMC Evolutionary Biology (2017) 17:55 Page 5 of 8

Fig. 2 Reaction norms for vertical length variation in A. bipinnata. a-f, on the Y-axis is the vertical length variation in mm and on the X-axis the environment. Colours encode source locality of colonies: black for Hospital Point and red for Crawl Key. Data are laterally offset for im- proved visualisation. Dots represent the median magnitudes and bars the + − MAD (Median Absolute Deviation). HPs = Hospital Point shallow, HPd = Hospital Point deep, CKs = Crawl Key shallow, CKd = Crawl Key deep

(β = −2.52 [1.09], p = 0.023). Reaction norms using model of phenotypic response [1, 29]. This is congruent branch density (Fig. 4) show that a large number of with a recent evaluation of a reciprocal transplant asses- branches were generated in the explants from Crawl Key sing the sclerite trait on the same species [30]. The deep to Crawl Key shallow (Fig. 4d; M = 0.86; MAD = general trend of variation in the three traits tested re- 0.35) and to Hospital Point shallow (Fig. 4b; M = 0.93; sembled the native transplants performance. The sur- MAD = 0.43) in comparison to deep controls of Crawl vival rate of foreign transplants and the trend towards Key (M = 0.51; MAD = 0.03) and Hospital point (M = 0.37; native values in the reaction norms indicated some grade MAD = 0.01) and even the shallow natives. of adaptive response [29, 31]. For bushy segments transplanted into deep habitats, a lower vertical and Discussion horizontal growth and fewer new branches were found, The LMM tests and reaction norms graphs for each of compared to the general response of foreign deep seg- the traits indicate adaptive phenotypic plasticity and ments in shallow habitats at both localities. Even when genetic variance in a classical genotype and environment the standard errors were high, which could be typical for

Fig. 3 Reaction norms for horizontal length variation in A. bipinnata. a-f, on the Y-axis is the horizontal length in mm and on the X-axis the environ- ment. Colours encode source locality of colonies: black for Hospital Point and red for Crawl Key. Data are laterally offset for improved visualisa- tion. Dots represent the median magnitudes and bars the + − MAD. HPs = Hospital Point shallow, HPd = Hospital Point deep, CKs = Crawl Key shallow, CKd = Crawl Key deep Calixto-Botía and Sánchez BMC Evolutionary Biology (2017) 17:55 Page 6 of 8

Fig. 4 Reaction norms for branch density in A. bipinnata. a-f, on the Y-axis the branch density calculated over the length of the main axis and on the X-axis the environment. Colours represent source locality of colonies: black for Hospital Point and red for Crawl Key. Data are laterally offset for improved visualisation. Black and red dots represent the median magnitudes and bars the + − MAD. HPs = Hospital Point shallow, CKs = Crawl Key shallow, CKd = Crawl Key deep this kind of architectural organisation [21] the trait data habitats, which decrease the light intensity and may is suggestive of adaptive phenotypic plasticity, i.e., slopes compromise the zooxanthellae photosynthetic process. trend similarity [32, 33]. Non-parallel reaction norms This may be especially true at Hospital Point, which within localities indicate even more strongly that plastic differed from Crawl Key by 11500 lux. Consequently, the responses were similar but not identical, suggesting that lower abundance of colonies in deep environments ecological and genetic divergence involved in the differ- support a range edge of morphotype performance [35], ent responses, where there is a large habitat influence where it is possible that a controlled variable such as in- but the phenotypic responses occur in different ways jury could act as a stressor in differential grades. over the bushy and deep morphotypes [1, 34]. Different magnitudes of genotypic and environmental The vertical and horizontal length variation for most effects appear to occur across the three variables, which transplants was positive in the shallow habitats, indicat- could be explained by different environmental drivers af- ing greater tissue growth, and negative in deep habitats, fecting each modular trait. The most conspicuous plastic indicating lower tissue growth. In deep habitats, this was response was new branch promotion, as statistics and particularly true for horizontal branches, which experi- reaction norms indicate. Reaction norms resembled the enced null growth and even tissue loss. Greater growth native ones from shallow habitats, where bushy morpho- in native controls in shallow habitats compared to na- types have the higher density of secondary branches on tives from deep habitats could be counterintuitive since shorter main axes. In particular, the shallow habitat at longer main and secondary branches characterise deep Crawl Key had the most positive values for branch pro- morphotypes. However, colony growth analysis in A. motion, reaching as high as 127 new branches in only elisabethae, a closely related species with the same colo- one deep morphotype segment (Fig. 4b, d). By contrast, nial architecture, have shown that initial branching there was almost no branch promotion from the two lo- growth is greater in small colonies than in larger ones calities in the Crawl Key deep habitat. Since number of with a drastic reversion in time [21]. Thus, it is possible branches is a trait positively correlated to wave exposure that growth in the A. bipinnata transplants could be ex- and currents in gorgonians [21], high physical disturb- plained by an initial difference in response to habitat ance in shallow versus deep habitats at Crawl Key com- after injury with an insufficient time to display such a re- pared to low water motion at both depths at Hospital version in branch growth behaviours. Point may explain this result. Similarly, in this group of At the same time, lost tissue (i.e. dead polyps) was corals, the light intensity is positively correlated with in- most common on horizontal branches of segments situ- vestment in branching. Therefore, in shallow habitats ated in deep habitats, which could indicate that environ- branching focused on secondary branches while in deep mental conditions in these locations are stronger drivers habitats the growth is focused on height, in order to in- for adaptation than shallow environmental conditions, as crease light exposure [36]. supported by the positive controls. One of these condi- These types of responses point to plastic properties in tions is the higher level of sedimentation in deep colonial organisms such as gorgonians, which could be Calixto-Botía and Sánchez BMC Evolutionary Biology (2017) 17:55 Page 7 of 8

expected to be more flexible in traits at a modular scale Availability of data and materials level [22, 37]. To understand the impact of modular All data generated during this study are included in this published article and its supplementary information. Additional file 1. properties on species evolability requires in addition to assessing canalization capacity, identifying other compo- Authors’contributions nents, such as the genetic fixation mechanisms and cost- ICB participated in collected field data, carried out data processing, benefit trade-offs. At the same time, it could be advisable analysis and wrote the manuscript; JAS conceived and designed the to assess life history attributes, such as game to genesis, study, participated in collectedfielddataandhelpeddraftthe manuscript. Both authors read and approved the final manuscript. number of reproductive cycles/years or larval produc- tion, which can serve as better proxies for fitness, even Competing interests in challenging systems, such as corals [38]. The authors declare that they have no competing interests.

Consent for publication Conclusions Not applicable. Immigrant inviability, proposed as a key element in re- productive isolation between morphotypes in the depth Ethics approval and consent to participate cline [39, 40], does not appear to be the main force in The experiment complies with the current laws of the country in which the evolution of phenotypic divergence in A. bipinnata. it was performed. Study permit provided under the project “Plasticidad fenotípica como promotora de especiación ecológica en un coral Instead, due to 83.3% of survival of transplanted segments gorgonáceo del Caribe” by the Smithsonian Tropical Research Institute in a previous transplant experiment [30] and 91.8% sur- (STRI) (Senior Latin American Fellow to J.A. Sánchez). vival in addition to environmental and genotype effects Author details found in this study, an intermediate scenario between 1Department of Animal Ecology and Systematics, Justus Liebig Universität, adaptive plasticity and local adaptation seems to be being Heinrich-Buff-Ring 26-32 IFZ D-35392, Giessen, Germany. 2Laboratory of carried out in this species. Additional molecular evalu- Biología Molecular Marina-Biommar, Department of Biological Sciences-Faculty of Sciences, Universidad de los Andes, Carrera 1E No 18A – ation of the genetic basis of modular traits divergence 10, P.O. Box 4976, Bogotá, Colombia. 3Marine Sciences, International Giessen could be enlightening, keeping in mind that preliminary Graduate Centre for the Life Sciences (GGL), Justus Liebig Universität, neutral genetic variation among A. bipinnata morpho- Giessen, Germany. types in Panama seems to be negligible [20]. Assessing Received: 18 August 2016 Accepted: 2 February 2017 adaptive genetic divergence coupled with a functional background of positive selection signals could provide an integral perspective to test the current hypothesis of an References incipient ecological speciation mediated by modular 1. West-Eberhard MJ. Developmental plasticity and evolution. Oxford: Oxford University Press; 2003. plasticity mechanisms in A. bipinnata. 2. Wright S. Evolution in mendelian populations. Genetics. 1931;16:97–159. 3. Schlichting C. The evolution of phenotypic plasticity in plants. Annu Rev Ecol Evol Syst. 1986;17:667–93. Additional file 4. Baldwin JM. Development and evolution. New York: MacMillan Co.; 1902. 5. Price TD, Qvarnström A, Irwin DE. The role of phenotypic plasticity in driving genetic evolution. Proc R Soc Lond B Biol Sci. 2003;270:1433–40. Additional file 1: Raw data of the reciprocal transplant for two locations 6. Gomez-Mestre I, Roger J. A heuristic model on the role of plasticity in and two depths in Antillogorgia bipinnata (Cnidaria: Octocorallia). Growth adaptive evolution: plasticity increases adaptation, population viability data for each branch in the two times are in the same row. Photo_Code and genetic variation. Proc R Soc B. 2013;280:1471–2954. description follows the next example for 3459_AJ: 3459 unique number; 7. Waddington CH. Canalization of development and the inheritance of A is the replicate (A or B); J (June) or O (October). (XLS 223 kb) acquired characters. Nature. 1942;150:563–5. 8. Whitman DW, Agrawal AA. “What Is Phenotypic Plasticity and Why Is It Important?”. In: Whitman DW, Ananthakrishnan TN, editors. Phenotypic Acknowledgements plasticity of insects: Mechanisms and consequences. Enfield: Science Thanks to Oscar Ramos, Sergej Sereda, Adriana Grismaldo and Courtenay Ray Publishers; 2009. p. 1–63. for statistical support and advice. Thanks to Diana Ballesteros Luisa Dueñas, 9. Foster A. Phenotypic plasticity in the reef corals Montastrea annularis and Adriana Sarmiento, Rocio Acuña and Cindy Gonzalez for field assistance. Siderastrea siderea. J Exp Mar Biol Ecol. 1979;39:25–54. Comments and informal discussion with Howard R. Lasker, Harilaos Lessios, 10. Prada C, Schizas N, Yoshioka P. Phenotypic plasticity or speciation? A case Carlos Prada, Nick Schizas, Zaira Garavito and two anonymous reviewers are from a clonal marine organism. BMC Evol Biol. 2008;8:47. greatly appreciated. Thanks to Colciencias (Programa Doctorados Nacionales 11. Bongaerts P, Riginos C, Ridgway T, Sampayo EM, van Oppen MJH, Englebert to I.F. Calixto-Botía). The Smithsonian Tropical Research Institute (STRI) (Senior N, et al. Genetic divergence across habitats in the widespread coral Latin American Fellow to J.A. Sánchez), Universidad de los Andes, Colombia seriatopora hystrix and its associated symbiodinium. PLoS One. 2010;5: (STAI, Vicerrectoría de Investigaciones and Facultad de Ciencias) and e10871. CEMarin (Center of Excellence in Marine Sciences) sponsored this research. 12. Bongaerts P, Riginos C, Hay K, VanOppen M, Hoegh-Guldberg O, Dove S. Adaptive divergence in a scleractinian coral: physiological adaptation of Funding Seriatopora hystrix to shallow and deep reef habitats. BMC Evol Biol. 2011;11:303. This research was sponsored by Colciencias (Programa Doctorados 13. Preston KA, Ackerly DD. Allometry and evolution in modular organisms. In: Nacionales to I.F. Calixto-Botía), the Smithsonian Tropical Research Institute Pigliucci M, Preston KA, editors. Modul. Phenotypic Complex. Oxford: Oxford (STRI) (Senior Latin American Fellow to J.A. Sánchez), Universidad de los Univ. Press; 2004. p. 80–106. Andes, Colombia (STAI, Vicerrectoría de Investigaciones and Facultad de 14. Sánchez JA, Zea S, Diaz JM. Gorgonian communities of two contrasting Ciencias) and CEMarin (Center of Excellence in Marine Sciences). environments from oceanic caribbean atolls. Bull Mar Sci. 1997;61:61–72. Calixto-Botía and Sánchez BMC Evolutionary Biology (2017) 17:55 Page 8 of 8

15. Sánchez J, Aguilar C, Dorado D, Manrique N. Phenotypic plasticity and morphological integration in a marine modular invertebrate. BMC Evol Biol. 2007;7:122. 16. Bayer FM. The shallow water Octocorallia of the West Indian Region: A manual for marine biologists. In: Nijhoff M, editor. Stud. The Hague: Fauna Curacao other Caribb. Islands; 1961. p. 373. 17. Gutierrez-Rodriguez C, Barbeitos MS, Sánchez JA, Lasker HR. Phylogeography and morphological variation of the branching octocoral Pseudopterogorgia elisabethae. Mol Phylogenet Evol. 2009;50:1–15. 18. Sánchez J, Wirshing H. A field key to the identification of zooxanthellate octocorals from the Caribbean and western atlantic. Caribb J Sci. 2005;41: 508–22. 19. Williams GC, Chen J-Y. Resurrection of the octocorallian genus Antillogorgia for Caribbean species previously assigned to Pseudopterogorgia, and a taxonomic assesment of the relationship of these genera with leptogorgia (Cnidaria, Anthozoa, Gorgoniidae). Zootaxa. 2012;3505:39–52. 20. Dorado D, Sánchez JA. Internal transcribed spacer 2 (its2) variation in the gorgonian coral Pseudopterogorgia bipinnata in Belize and Panama, Smithson. Contrib Mar Sci. 2009;38:173–9. 21. Lasker HR, Boller MA, Castanaro J, Sánchez JA. Determinate growth and modularity in a gorgonian octocoral. Biol Bull. 2003;205:319–30. 22. Sánchez JA, Lasker HR. Patterns of morphological integration in marine modular organisms: supra-module organizationin branching octocoral colonies. Proc R Soc Lond B Biol Sci. 2003;270:2039–44. 23. Sánchez JA, Lasker H. Do multi-branched colonial organisms exceed normal growth after partial mortality? Proc R Soc Biol Sci. 2004;271:S117–20. 24. Schneider CA, Rasband WS, Eliceiri KW. NIH Image to imageJ: 25 years of image analysis. Nat Methods. 2012;9:671–5. 25. Bartoń K. MuM. In: Multi-Model Inference. R package version 1.15.6. 2016. Available from: https://cran.r-project.org/package=MuMIn 26. Burnham KP, Anderson DR. Model selection and inference: a practical information-theoretic approach. secondth ed. New York: Springer; 2002. 27. R Development Core Team. R: A language and Computing, statistical computing. Vienna, Austria: R Foundation for Statistical Computing. 2008. Available from: http://www.r-project.org. 28. Bates D, Maechler M, Bolker B, Walker S. Fitting linear mixed-effects models using lme4. J Stat Softw. 2015;67:48. 29. Sultan S, Stearns S. Environmentally contingent variation: phenotypic plasticity and norms of reaction. In: Hallgrimsson B, Hall BK, editors. Var. A Hierarchical Exam. a Cent. Concept Biol. Burlington: Academic Press; 2005. p. 303–32. 30. Joseph EO, Carlo JM, Lasker HR. Plasticity and conservatism in sclerites of a Caribbean octocoral. Hydrobiologia. 2014;1–10. 31. de Jong G. Evolution of phenotypic plasticity: patterns of plasticity and the emergence of ecotypes. New Phytol. 2005;166:101–18. 32. Pigliucci M. Evolution of phenotypic plasticity: where are we going now? Trends Ecol Evol. 2005;20:481–6. 33. Pigliucci M, Murren CJ, Schlichting CD. Phenotypic plasticity and evolution by genetic assimilation. J Exp Biol. 2006;209:2362–7. 34. Crispo E. Modifying effects of phenotypic plasticity on interactions among natural selection, adaptation and gene flow. J Evol Biol. 2008;21:1460–9. 35. Hengeveld R, Jaap H. The distribution of abundance. I. measurements. J Biogeogr. 1982;9:303–16. 36. Cadena NJ, Sánchez JA. Colony growth in the harvested octocoral Pseudopterogorgia acerosa in a Caribbean coral reef. Mar Ecol. 2010;31(4): 566–73. 37. de Kroon H, Huber H, Stuefer JF, van Groenendael JM, Sánchez JA. A modular concept of phenotypic plasticity in plants. New Phytol. 2005;166:73–82. 38. Beiring EA, Lasker HR. Egg productionbycoloniesofagorgoniancoral. Submit your next manuscript to BioMed Central Mar Ecol Prog Ser. 2000;196:169–77. 39. Prada C, Hellberg M. Long prereproductive selection and divergence and we will help you at every step: by depth in a caribbean candelabrum coral. Proc Natl Acad Sci U S A. • We accept pre-submission inquiries 2013;110:3961–6. 40. Prada C, Hellberg M. Strong natural selection on juveniles maintains a narrow • Our selector tool helps you to find the most relevant journal adult hybrid zone in a broadcast spawner. Am Nat. 2014;184:702–13. • We provide round the clock customer support • Convenient online submission • Thorough peer review • Inclusion in PubMed and all major indexing services • Maximum visibility for your research

Submit your manuscript at www.biomedcentral.com/submit 1 TESTING ADAPTIVE GENETIC DIVERGENCE IN PARALLEL ACROSS THE DEPTH CLINE: 2 POPULATION GENOMICS IN THE CORAL COMPLEX 3 Antillogorgia bipinnata-kallos (Cnidaria:Octocorallia) 4 5 Iván Calixto-Botía1,2, Jorge Duitama3, Andrea Gonzalez4, Thomas Wilke2 & Juan A. 6 Sánchez1,5 7 8 1. Laboratory of Biología Molecular Marina-Biommar, Department of Biological Sciences-Faculty of Sciences, 9 Universidad de Los Andes, Carrera 1E No 18A – 10, P.O. Box 4976, Bogotá, Colombia. 10 2. Department of Animal Ecology and Systematics. Justus Liebig Universität. Heinrich-Buff-Ring 26-32 IFZ D- 11 35392, Giessen, Germany. 12 3. Department of Systems and Computing Engineering-Faculty of Engineering, Universidad de Los Andes, 13 Carrera 1E No 18A – 10, P.O. Box 4976, Bogotá, Colombia. 14 4. BIOS, Centro de Biología Computacional y Bioinformática, Ecoparque Los Yarumos, Manizales, Caldas, 15 Colombia. 16 5. Marine Sciences, International Giessen Graduate Centre for the Life Sciences (GGL), Justus-Liebig- 17 Universität-Gießen, Germany 18 19 20 ABSTRACT 21 Depth has been proposed as a main factor to explain the strong phenotypic variation and 22 diversification patterns in a broad range of marine organisms. In Caribbean octocorals, 23 most species exhibit a strong overlap in morphologic variation across sympatric cline 24 depths, which complicates the study of their diversity and evolutionary history. This is the 25 case of the gorgonian Antillogorgia bipinnata (Verrill 1864), which exhibits phenotypic 26 divergence related to adaptation to bathymetric gradients, where relationship between 27 forms and populations across the Caribbean is unclear. The purpose of the present 28 research was to assess the association between gene flow, environmental dissimilarity, 29 and biogeography to test a parallel case of Isolation By Adaptation of the A. bipinnata 30 complex promoted by depth cline. This case represents an optimal model to understand 31 the extent of ecological speciation in the diversification patterns of sessile organisms. The 32 ezRAD, a reduced-genome representation methodology, was performed to genotype 33 pools of kallos and deep colonies across four locations of the A. bipinnata complex. A first 34 reference genome for the species was generated to align the ezRAD reads. Strong 35 geographic isolation, typical of brooders, showed a parallel process of divergence. An 36 Isolation by Adaptation test revealed reduced gene flow between morphotypes, indicating 37 an adaptive genetic process. Based on spawning records, asynchrony in gamete 38 maturation between morphotypes is proposed as a potential mechanism for gene flow 39 reduction. These data create an emergent view of sympatric adaptive divergence in a 40 putative scenario of parallel ecological speciation for the Antillogorgia bipinnata complex. 41 42 Keywords: Adaptive genetic divergence, Ecological Speciation, Octocoral, Depth Cline, 43 Population Genomics, Pooling. 44 45 46

17

47 INTRODUCTION 48 49 Depth is a key ecological variable for the promotion of conspicuous adaptations in marine 50 organisms, particularly for sessile species. The environmental heterogeneity across 51 bathymetric gradients involves mainly light, but also other correlated selective forces such 52 as sediment concentration, temperature, oxygen concentration, nutrient concentration 53 and composition, as well as physical stress due to water movement and the differential 54 distribution of mutualists and predators. The consistency of these environmental variables 55 over evolutionary time and the spatial scale across interbreeding morphotypes are two 56 essential elements to assess the explanatory power of depth clines, not only for 57 promoting phenotypic variants, but its extend in the formation of new species. 58 59 Antillogorgia bipinnata (Verrill 1864) is a gorgonian octocoral showing phenotypic 60 variation at the colonial level where the distribution of extreme morphotypes depends on 61 habitat, particularly water characteristics along the bathymetric gradients. Based on 62 molecular and morphological evidence, A. bipinnata has been recently reassigned from 63 the genus Pseudopterogorgia (Williams and Chen 2012). A. bipinnata is a shallow water 64 Caribbean coral in close association with the photosynthetic endosymbiont Symbiodinium 65 antillogorgium clade B1 (Parkinson, Coffroth, and LaJeunesse 2015). Similar to its sister 66 species, A. elisabethae, A. bipinnata is a surface brooder with restricted larvae dispersion, 67 a philopatric pattern, and a suspected strong geographic structure of populations. 68 Populations of the complex A. bipinnata are grouped in clusters of reefs over the southern 69 Caribbean, the Mesoamerican Reef (with Panamanian and Belizean populations closely 70 related) and the Florida-Bahamas region (Figure 1). 71 72 Antillogorgia kallos (Bielschowsky 1918), the sister species of A. bipinnata, strikingly 73 resembles a bushy morphotype of A. bipinnata (Sánchez et al. 2007). The little available 74 genetic data for the clade indicate a complex scenario where A. kallos is sometimes 75 nested inside A. bipinnata, even though resembling a geographic pattern (Sánchez, 76 unpublished). Assessed ITS2 and mtDNA markers show a phylogeographic pattern for 77 populations from the Bahamas and Belize, where the morphotypes are polyphyletic 78 (Sánchez et al. 2007). A larger dataset analysis using ITS2 sequence data, identified 79 differences between the kallos forms in Belize and Panama (Sánchez, unpublished). Based 80 on the diffuse boundaries between the species at the taxonomical and molecular levels, in 81 the present study we refer to the A. bipinnata complex as the sensu stricto species plus 82 the A. kallos as the shallower morphotype. 83 84 As a modular organism, with polyps and branches as a repetitive unit, A. bipinnata is 85 equipped with a flexibility and versatility of parts generating the colonial architecture. The 86 ‘deep’ morphotype is characteristically bigger, with longer main and secondary branches 87 and internodal lengths, but fewer secondary branches compared to the ‘bushy’ (=A. 88 kallos) morphotype. Based on their inherent modular organization, the polyp represents 89 an iterative unit (sensu stricto) and the branches derived modular units (Lasker et al. 2003; 90 Sánchez and Lasker 2004). Thus, differences between morphotypes (involving distribution

18

91 and density of polyps) can be directly related to environmental variables in the depth cline 92 as feasible adaptive responses to nutrient capture, overall photosynthetic rate and 93 physical stress tolerance (Calixto-Botía and Sánchez 2017). 94 95 Previous work focused on the plasticity of modular traits that define morphotypes was 96 performed via a reciprocal transplant experiment between shallow and deep depths. 97 (Calixto-Botía and Sánchez 2017). This research found high survival rates for foreign 98 transplants and adaptive plasticity (supporting A. kallos as a morphotype), but also a 99 genotype component interacting for the resulting variance of the assessed traits. This 100 scenario could represent an incipient phase of ecological speciation related to depth cline 101 with a genetic component of plastic variance susceptible to selection (Pigliucci, Murren, 102 and Schlichting 2006; Thibert-Plante and Hendry 2011). The present research contributes 103 to answering the hypothesis of an ecological speciation process by a test of Isolation By 104 Adaptation (IBA). In this approach, ecological speciation can be inferred by reduced gene 105 flow between morphotypes due to several reasons such as selection against hybrids, 106 immigrants, or assortative mating (Feder, Egan, and Nosil 2012; Nosil 2007; Schluter 107 2009). Neutral SNPs can provide a measure of genetic differentiation where it is expected 108 that during divergent selection, at least one of the localities will show reduced gene flow 109 between morphotypes of the complex. Additionally, due to philopatry and focal 110 populations separated by long distances across the Caribbean, it is expected that the 111 process could take place in a replicated mode. Parallel ecological speciation is a particular 112 case composed of several independent speciation events (Schluter and Nagel 1995). 113 114 In recent years, population studies have migrated from classical markers to reduced- 115 genome representation by Next Generation Sequencing technologies (NGS) techniques. 116 NGS technologies have led to improvements in the analysis of phylogeographic 117 relationships, where otherwise the use of traditional markers can result in low resolution, 118 estimation biases in population differentiation (due to loci under genetic hitchhiking or 119 selection), branch lengths, and topology of phylogenies (Landry, Koskinen, and Primmer 120 2002; Luikart et al. 2003). For Octocorallia this is particularly important since 121 mitochondrial markers are rarely informative due to DNA correction by the enzyme 122 mtMutS (Bilewitch and Degnan 2011). In addition, mitochondrial markers tend to be 123 conserved in Hexacorallia (Hellberg 2006) and show intraspecific variation and multicopy 124 for ITS (Van Oppen et al. 2000; Vollmer and Palumbi 2004). In the present work, we 125 applied a population genomics approach based on pooling of samples and the ezRAD 126 (Restriction Associated to DNA) methodology (Toonen et al. 2013), a variant of RADSeq 127 (Davey and Blaxter 2010) successfully implemented in studies of population 128 differentiation and phylogeography in both model and non-model species (Emerson et al. 129 2010; Hohenlohe, Catchen, and Cresko 2012; Puritz, Addison, and Toonen 2012). 130 131 With the current undetermined relationship between the different forms of the complex 132 and between the populations across the Caribbean, the present research relies on how 133 the genetic diversity is partitioned for the A. bipinnata complex. To resolve the question of 134 genetic divergence of the complex at a fine-scale, markers derived from reduced-genome

19

135 representation methodologies were assessed for intra-local allele frequency differences 136 across the phenotypic variants and contrasted with differences between the Caribbean 137 localities of the survey. Our research assesses a parallel ecological speciation scenario for 138 the A. bipinnata complex, representing an elucidatory component since it is unlikely that 139 the same ecological barriers present in the different localities could independently 140 produce a reduced gene flow just by chance (Luikart et al. 2003; Nosil, Funk, and Ortiz- 141 Barrientos 2009; Schluter and Nagel 1995). Therefore, we predict that if a speciation 142 process in parallel mediated by ecological pressures occurs in the A. bipinnata complex, 143 reduced gene flow between the extreme morphotypes will be found for at least two of 144 the considered localities. 145

146 147 Figure 1. Distribution of main Antillogorgia bipinnata populations. (A) Three main populations of 148 A. bipinnata are distributed in the Caribbean. (B) Distribution of A. bipinnata morphotypes along 149 the depth cline for four locations from genomic libraries were constructed: Bocas del Toro 150 (Panama), Cartagena (Colombia), Carrie Bow Cay (Belize) and Providencia (Colombia). 151 152 153 MATERIALS AND METHODS 154 155 Reference genome assembly and annotation 156 The Antillogorgia bipinnata genome was sequenced and used as reference for mapping 157 the ezRAD reads. This genome is derived from a pool of 50 larvae (to reduce zooxanthellae 158 contamination), from a controlled cross in aquarium between two colonies separated by 159 12Km of distance and located at 6m of depth in Bocas del Toro, Panama (Smithsonian 160 Tropical Research Institute, November 2012). Larvae were collected from the mother 161 colony surface and grown in artificial and filtered seawater for two weeks until DNA 162 extraction. Larvae were pooled to obtain the minimum DNA quantity for the construction 163 of a paired-end library of 500bp (overlapping reads, 380bp insert, 120bp adapters). The 164 library ran on one lane of the Illumina Hiseq2000 platform. Library construction and

20

165 sequencing were accomplished in the facilities of the Lewis-Sigler Institute for Integrative 166 Genomics at Princeton University. 167 168 Raw paired-end (PE) sequencing reads were pre-processed to filter adapter, low quality, 169 and contamination reads. Quality control was assessed with FastQC (Andrews 2010), 170 adapter and low quality reads were removed using Cutadapt (Martin 2011), and 171 contamination was assessed using the Blobology pipeline (Kumar et al. 2013). After 172 applying Blobology, we still detected contamination present in reads that did not 173 assemble into contigs. Therefore, we built local sequence databases of the main 174 contaminating organisms, downloaded from Genbank, and performed local BLAST 175 searches of the total reads against the local databases. Reads showing BLAST hits with e- 176 value and sequence identity thresholds of 1 x 10-50 and ≥95%, respectively, were removed. 177 In addition, as high sequencing coverages have been reported to not provide any 178 additional benefits to assemblies (Desai et al. 2013), we performed various assembly 179 attempts using different read coverages (86X, 75X,and 70X) in order to determine the 180 coverage that provided the best assembly metrics. For this, different percentages of reads 181 were randomly discarded to reduce coverage from the original coverage of 86X to 75X and 182 70X, using the DownsampleSam option of Picard Tools version 2.8 183 (http://broadinstitute.github.io/picard.). 184 185 De novo assembly was carried out using a two-step approach with MaSuRCA version 3.2.2 186 (Zimin et al. 2013) and ABySS version 1.9.0 (Simpson et al. 2009) assemblers. First, a 187 preliminary contig assembly was performed with MaSuRCA. Second, a scaffold assembly 188 was performed with ABySS using the paired end reads combined with MaSuRCA- 189 assembled contigs greater than 500 bp as ‘long reads’ for re-scaffolding. The scaffold 190 assembly was further extended and gap-closed using SSPACE version 3.0 (Boetzer et al. 191 2011) and GapFiller version 1.10 (Nadalin, Vezzi, and Policriti 2012), resulting in the final 192 genome assembly used for ezRAD mapping. Genome assembly statistics were generated 193 using Quast version 4.5 (Gurevich et al. 2013) and genome completeness was assessed 194 using Benchmarking Universal Single-Copy Orthologs (BUSCO) (Simão et al. 2015) tool 195 version 2.0, by searching for orthologs gene content against the Metazoa odb9 dataset. 196 Finally, scaffolds greater than or equal to 1000 bp were retained for the downstream 197 analyses. An assessment of ploidy level was performed over the assembled reference 198 genome of A. bipinnata. Raw reads were aligned to the assembly using bowtie2 and the 199 distribution of relative allele frequencies at heterozygous sites was calculated using the 200 command RelativeAlleleCounts of NGSEP (Supplementary Table S2). 201 202 Genome annotation was carried out based on gene predictions using the MAKER2 (Holt 203 and Yandell 2011) pipeline; specifically, ab initio predictions were generated with SNAP 204 (Bromberg and Rost 2007) and Augustus (Stanke et al. 2006) programs, as well as with 205 protein homology evidence-based predictions against RefSeq protein sequences for 206 Anthozoa. Ciona intestinalis gene models were used for running SNAP predictions and 207 Amphimedon gene models were used for Augustus predictions. Contamination was once 208 again detected and removed from the protein set based on a blast search against a local

21

209 database of Papilio, Bombyx, Serratia, and Symbiodinium RefSeq protein sequences. The 210 filtered proteins were annotated using Blast2GO (Conesa and Gotz 2008), based on blastp 211 searches against the Genbank non-redundant (nr) protein database, Interpro searches, 212 and GO mapping and annotation. 213 214 ezRAD library constructions and sequencing 215 To achieve an accurate picture of the gene flow process, a genomic approach between 216 extreme morphotypes of A. bipinnata from Bocas del Toro (Panama), Carrie Bow Cay 217 (Belize), Providencia Island (Colombia), and Cartagena (Colombia) was conducted. A cost- 218 effective pooling strategy was followed by selecting the extreme kallos and deep 219 morphotypes of A. bipinnata over each location to obtain eight genomic libraries in total 220 (Figure 1). These libraries were constructed following the ezRAD methodology (Toonen et 221 al. 2013), which involves a step of enzymatic digestion with the isoschizomers MobI and 222 Sau3AI (allowing minimization of any potential impacts of DNA methylation). 223 224 DNA was extracted from dry, ethanol or DMSO preserved tissue with a modified phenol- 225 chloroform protocol (Coffroth et al. 1992) and DNeasy Blood & Tissue Kit (Qiagen®). DNA 226 from the zooxanthellae symbiont present in the samples was reduced by these extraction 227 protocols as well as an additional first step of centrifugation. DNA quality and 228 concentrations were measured through electrophoresis on 1% agarose gels, absorbance 229 readings using a NanoDrop spectrophotometer (NanoDrop Technologies, Wilmington, DE), 230 and by an AccuBlue® High Sensitivity fluorescence assay on a SpectraMax M2 plate reader 231 (BiotiumTM). Most of the samples displayed degradation with fragments below 500bp. In 232 those cases, a second round of elutions using a 1:1.4 ratio (DNA:beads) was performed for 233 cleaning/size selection according to Agencourt AMPure XP Purification system (Beckman 234 coulter®). Based on the final quality obtained, equal amounts of DNA were pooled for each 235 habitat in each location: 28 for kallos and 28 for deep in Bocas del Toro, 22 for kallos and 236 22 for deep in Cartagena, 12 for kallos and 12 for deep in Carrie Bow Cay, and 40 for kallos 237 and 40 for deep in Providencia. 238 239 Double digestions were performed with MobI and Sau3AI (NEB). For Bocas del Toro, Carrie 240 Bow Cay, and Cartagena libraries, cleaned digestions were end-repaired following the 241 TruSeq DNA v2 kit (Illumina). FlashGel methodology (Lonza®) allowed for a high 242 purification of ligate products and size selection (300-500bp insert size). An enrichment 243 step was performed using 15 cycles. After a final cleaning, libraries were validated by 244 visualization on an Agilent 2100 BioAnalyzer and quantified by qPCR. Libraries were 245 sequenced (150bp paired-ends) with an Illumina GAIIx platform on 1/12 lane each. Library 246 construction and sequencing were performed in the Hawai’i Institute of Marine Biology 247 EPSCoR Core sequencing facility of the University of Hawai’i, USA. The TruSeq PCR-free kit 248 (Illumina) was used to construct the two libraries from Providencia (350bp insert size) in 249 the facilities of BIOMMAR Lab, Los Andes University and sequenced (150bp paired-ends) 250 on one lane of Illumina HiSeq2500 by GENEWIZ, NJ, USA. 251 252 ezRAD bioinformatic analyses

22

253 Two methodological pipelines were executed to obtain higher robustness of the analysis 254 of the sequencing data obtained following the ezRAD protocol. In the first pipeline, raw 255 reads were aligned to the reference genome with Bowtie2 (Langmead and Salzberg 2012). 256 Then, raw SNPs were identified running the FindVariants command of the NGSEP software 257 pipeline (Duitama et al. 2014) with recommended parameters for genotype by sequencing 258 data (maxBaseQS 30, minQuality 40, maxAlnsPerStartPos 100). 259 260 In the second pipeline, raw reads were first filtered by quality control performed with 261 FastQC (Andrews 2010) and cutadapt (Martin 2011) for adapters, minimum quality base of 262 28 and trimmed to a length of 100bp. Following, a mixed read analysis was performed 263 with MEGAN (Huson et al. 2011), Deconseq (Schmieder and Edwards 2011a), and Prinseq 264 (Schmieder and Edwards 2011b) software to detect and clean sequencing contaminants. 265 For this purpose, a random sampling of the reads was aligned against a reference 266 database using BLASTn. Reads from bacteria, viruses and the zooxanthellae were detected 267 and eliminated using NCBI RefSeq as source for a local database of representative 268 genomes, and the mitochondrial genome was separated using a previously published 269 genome (Medina et al. 2006). Orphan reads were removed with the “repair.sh” script 270 from BBmap (Bushnell 2016). This analysis detected relatively few hits for A. bipinnata 271 mitochondrion, Symbiodinium, bacteria and viruses, eliminating between 1% and 7% of 272 the reads at this step. Processed reads were aligned to the reference genome using the 273 software tool BWA (Li and Durbin 2009). Then, the NGSEP software (Duitama et al. 2014) 274 was used to identify raw SNPs with the same parameters used for the first approach. 275 276 To assess the population structure of the kallos and deep morphotypes of A. bipinnata 277 from Bocas del Toro, Cartagena, Carrie Bow Cay, and Providencia, the following 278 downstream analysis was performed from the merged VCF files provided by NGSEP. First, 279 a VCF file was generated for each location using the FilterVCF command from NGSEP, 280 selecting the deep and kallos samples for each population. This produced four VCF files 281 with two samples each (hereafter called pairwise VCF files). For each entry of each 282 pairwise VCF file, nucleotide counts were recovered and filtered running a local script. This 283 script removes nucleotide counts smaller than a minimum threshold calculated as the 284 maximum between two numbers: a) five times the smallest count for a nucleotide plus 1 285 and b) the sum of nucleotide counts divided by the number of haplotypes in a pool (40 by 286 default). The rationale for these filters is that minor alleles should, on one hand, 287 differentiate clearly from sequencing errors and, on the other hand, show an allele 288 frequency of at least one divided by the number of haplotypes in the sample. The script 289 then filters variants keeping only those having the two samples with read depth larger 290 than 40 and at least one sample with two observed alleles. The entire procedure was 291 performed for the VCF files generated using both bwa and bowtie2. 292 293 Allele counts from the pairwise VCF files were exported to the input format of the tool 294 popoolation2 (Kofler, Vinay-Pandey, and Schloetterer 2011) using a local script. The 295 “subsample-synchronized.pl” script of popoolation was then executed with parameters “-- 296 target-coverage 40”, “--max-coverage 2%" and “--method withoutreplace”. Both the

23

297 “fisher-test.pl” and the “fst-sliding.pl” scripts were executed to calculate fisher exact tests 298 for differences in allele frequencies and F statistics between populations, respectively. 299 Finally, read counts per allele, instead of genotypes prediction, were used to calculate 300 distance matrices with the “VCFDistanceMatrixCalculator” command from NGSEP, with 301 parameters “-s 3 -p 40” and then a distance tree was generated with SplitsTree software 302 (Huson and Bryant 2006). 303 304 To consider different locations as independent events, a VCF file with the two locations 305 with better coverage (Bocas del Toro and Providencia, four samples) was generated 306 following the same procedure described above for pairwise VCFs. After exporting and 307 subsampling allele counts, the CMH (Cochran-Mantel-Haenszel) test for differences 308 between allele frequencies with replicates was calculated for each variant using the 309 “CMH-test.pl” script of popoolation. The “population” parameter was first set to “1-3,2-4” 310 to compare deep vs kallos populations using locations as replicates. Then, it was set to “1- 311 2,3-4” to compare allele frequencies between locations using depths as replicates. 312 313 314 RESULTS 315 316 A draft assembly of the Antillogorgia bipinnata genome 317 We generated a draft genome assembly for Antillogorgia bipinnata performing whole 318 genome shotgun (WGS) Illumina sequencing following the paired-end protocol with 500bp 319 insert length fragments and read lengths between 215 to 230 bp. A total of 234,565,510 320 raw sequencing reads were obtained from this experiment. After quality control and 321 removal of contamination reads, 180,223,329 high quality reads with lengths between 70 322 to 220 bp (~37.85 Gbp) were used for the assembly process. Expectedly, the preprocessing 323 step identified 12.82% contamination reads matching bacterial genomes, mainly the coral 324 pathogen Serratia marcescens (Patterson et al. 2002), and only 0.0003% reads matching 325 the coral symbiont Symbiodinium minutum, observed from the blobplots obtained by the 326 Blobology pipeline (See supplementary Figure S1). This analysis also identified 2.53% 327 reads that matched Lepidoptera sequences (mainly Papilio spp. and Bombyx spp.), 328 product of some type of contamination during the library preparation process. In total, 329 32,814,507 contamination reads were detected and removed. In addition, 5,652 read 330 sequences corresponding to the mitochondrial genome were further filtered based on 331 matches to the Pseudopterogorgia bipinnata mitochondrial genome sequence (NCBI 332 accession NC_008157.1). 333 334 De-novo analysis of the remaining reads yielded an assembly of 128,219 gap-filled 335 scaffolds with a total length of 284 Mbp. This represents 94.7% of an estimated genome 336 size of 300 Mbp (Dueñas et al. unpublished). Fifty percent of the total genome draft 337 assembly is contained in 10,438 scaffolds of at least 7.5 Kbp in length (N50). Overall 338 statistics of the draft assembly of A. bipinnata genome are shown in Table 1. Additionally, 339 raw reads were realigned to the draft assembly and variant calling was performed on the 340 aligned reads. This yielded a total of 259,632,610 predicted heterozygous variants over

24

341 the genome. A relative allele frequency distribution based on allele read counts on these 342 SNPs showed the characteristic pattern for a diploid organism (See Supplementary Figure 343 S2). 344 345 Table 1. Statistics of the coral Antillogorgia bipinnata genome draft assembly. BUSCO results are 346 based on 978 total BUSCO groups searched in the Metazoa odb9 dataset. 347 Metric Total length (bp) 284,441,007 Number of scaffolds 128,219 Number of scaffolds ≥1 Kbp 36,503 Number of scaffolds ≥10 Kbp 7,882 Maximum scaffold length (bp) 149,494 N50 10,438 N50 length (bp) 7,494 GC (%) 35.71 Number of gaps (N) 9,094,289 Number of gaps per 100 Kbp 3,197.25 Complete and single-copy BUSCOs 684 Complete and duplicated BUSCOs 4 Fragmented BUSCOs 181 348 349 Gene prediction and annotation were carried out on contigs or scaffolds of 1 Kbp in length 350 or greater, using the MAKER2 pipeline. We found 5171 genes predicted in 36503 scaffolds 351 over 1 Kbp, based on C. intestinalis and Amphimedon gene models in addition to Anthozoa 352 protein evidence. The blast search against Papilio, Bombyx, Serratia, and Symbiodinium 353 protein sequences showed 82% hits against Papilio and Bombyx sequences, 7% against 354 Serratia sequences, and 0.0001% hits against Symbiodinium. In total, 4990 proteins were 355 removed from the set due to detected contamination, while the remaining 181 proteins 356 were annotated with Blast2GO. Gene Ontology-based annotations for biological 357 processes, molecular function, and cellular component for the A. bipinnata protein set are 358 shown in Figure 2. 359

25

360 361 Figure 2. Gene Ontology (GO) Level 2 annotations for the 181 proteins set predicted from the 362 Antillogorgia bipinnata draft genome assembly. Pie charts are shown for Biological Processes (A), 363 Cellular component categories B) and Molecular Function (C). 364 365 366 Bioinformatic analysis of ezRAD pooled sequencing data 367 Following the ezRAD protocol for cost-effective discovery of single nucleotide 368 polymorphisms (SNPs) within and between populations, we sequenced eight pools 369 corresponding to the deep and kallos morphotypes in for different localities (Figure 1). 370 This sequencing effort resulted in a total of 35,472,836 paired-end reads (5.9 Gbp). 371 Number of raw and processed reads per pool are shown in Figure 3A. We tried two 372 different bioinformatic approaches to discover SNPs and to infer allele frequencies for 373 each SNP within each population. After read preprocessing and alignment using the BWA 374 software, 3,981,435 reads (12.12%) aligned to the A. bipinnata assembly. Pooled variants 375 detection over the aligned reads identified 55,808 SNPs over the 8 samples. Conversely, 376 3,396,779 (9.57%) reads were aligned using the bowtie2 software. Although this number 377 is lower than that obtained with BWA, 66,616 SNPs were identified performing pooled 378 variants detection over the reads aligned with Bowtie2. Allele frequencies for each SNP 379 identified using each methodology were estimated based on read counts. The distribution 380 of the differences between predicted allele frequencies for kallos and deep pools within 381 each locality is shown in Figure 3B. Consistent with the total amount of aligned reads per 382 pool, kallos vs. deep allele frequencies from Bocas del Toro and Providencia could be 383 compared for a larger number of SNPs, particularly from reads aligned using BWA. 384 However, in the locations sequenced at lower depth (Cartagena and Carrie Bow Cay) more 385 SNPs were obtained from reads aligned with Bowtie2. Bocas del Toro and Providencia had 386 the lowest differences in allele frequencies with a maximum value of 0.4, whereas the 387 highest differences were observed for Carrie bow Cay, with values as high as 0.9 for 3 388 SNPs, showing almost complete segregation between kallos and deep morphotypes. 389

26

390 391 Figure 3. Number of raw and processed reads and allele frequencies generated. (A) Number of 392 reads sequenced and aligned for the eight libraries. (B) Kallos vs Deep allele frequencies for the 393 four locations. Blue: Raw reads. Turquoise: BWA bioinformatic pipeline. Red: Bowtie2 394 bioinformatic pipeline. BDT: Bocas del Toro. CAR: Cartagena. CBC: Carrie Bow Cay. PRV: 395 Providencia. B and D letters at the end of previous abbreviations mean Bushy (=Kallos ) and Deep, 396 respectively. 397 398 399 Genetic structure among localities and morphotypes 400 Based on the larger number and more wide distribution of SNPs and allele frequencies 401 obtained from reads aligned with Bowtie2, this dataset was selected for further 402 investigation of signatures of genetic structure among localities and morphotypes. 403 However, analyses performed from the reads aligned using BWA are generally consistent 404 with the main findings presented in this section (Supplementary figures S3 to S6). To 405 assess genetic differentiation between locations, a Cochran–Mantel–Haenszel (CMH) test 406 was performed for the two locations with the highest coverages (Bocas del toro and 407 Providencia), looking for significant differences between allele frequencies and assuming 408 that deep and kallos pools for one locality could be treated as replicates. The distribution 409 of p-values for 560 SNPs for which reliable allele frequencies could be predicted within the 410 4 pools are shown in Figure 4A. The mean p-value for this analysis was 0.366 (95% C.I.: 411 0.335, 0.397) and 13 SNPs were reported as significantly differentiated after correction for 412 multiple testing. 413 414 The same procedure was applied to look for significant differences in allele frequencies 415 between morphotypes, using different localities as replicates. Unfortunately, an attempt 416 to perform this analysis using the four localities was not successful because the number of 417 SNPs with reliable estimations of allele frequencies over the eight pools was very small. 418 The analysis using Bocas del toro and Providencia as replicates is shown in Figure 4B. The 419 mean p-value in this case was 0.511 (95% C.I.: 0.481, 0.519) and none of the SNPs showed

27

420 a significant difference between morphotypes after correction for multiple testing, 421 although 2 SNPs were close to achieve significance. Quantile-quantile plots (QQ plots) 422 related to this analysis show in both cases deviation from the uniform distribution of p- 423 values expected in absence of genetic differentiation. This deviation is a usual signature of 424 genetic differentiation between populations. Consistent with the Manhattan plots, the 425 comparison between localities shows a noticeable larger deviation from the expected 426 distribution. 427

428 429 Figure 4. Manhattan and Q-Q plots of the CMH test between Panama and Providencia libraries. (A) 430 Plots for geography, with colonies from the same location as replicates. (B) Plots for habitat, with 431 colonies from the same habitat as replicates. The y axis shows -log10 of CMH test and the x axis the 432 position of SNPs on scaffolds. The blue line and red lines in the Manhattan plot indicate the 433 thresholds for significant SNPs at p < 1 × 10−5 and p < 5 × 10−8, respectively. 434 435 Additional analyses were performed looking for signatures of differentiation between 436 morphotypes. The Fst statistic and a fisher exact test for differentiation of allele 437 frequencies between morphotypes was performed independently for the 4 localities using 438 the SNPs for which reliable allele frequencies were predicted within the two pools of each 439 locality (Figure 5A and 5B). Pairwise Fst between habitats were performed following a 440 sliding window approach in which the length of the window was defined dynamically by 441 the length of the scaffold where each SNP is located. In total 40 SNPs had Fst values over 442 0.1. Bocas del Toro with 101 SNPs analyzed, had a mean Fst = 0.031 (95% C.I.: 0.025, 443 0.036), Cartagena with 23 SNPs, a mean Fst = 0.04 (95% C.I.: 0.009, 0.071), Carrie Bow Cay 444 with 40 SNPs, a mean Fst = 0.063 (95% C.I.: 0.024, 0.103), and Providencia with 2056 SNPs 445 and a mean Fst = 0.024 (95% C.I.: 0.023, 0.025). 446

28

447 448 Figure 5. Fst and Fisher Exact Test for the four locations. (A) Fst pairwise average between kallos 449 and deep morphotypes for the four locations. (B) Fisher exact test between kallos and deep 450 morphotypes for the four locations with the y axis showing -log10 of Fisher exact test. The x axis 451 shows the position of SNPs on scaffolds. 452 453 The fisher exact test showed 26 SNPs significantly differentiated (p < 1 x 10-5), with the 454 smallest p-values observed for Carrie Bow Cay, in concordance with the Fst test. Bocas del 455 Toro with 362 SNPs, had a mean -log10 p-value of 0.745 (95% C.I.: 0.664, 0.826). Cartagena 456 with 110 SNPs, 0.829 (95% C.I.: 0.552, 1.105). Carrie Bow Cay with 177 SNPs, 1.301 (95% 457 C.I.: 0.909, 1.693) had the lowest values, with an extreme 18.940. Providencia with 20,704 458 SNPs, had the highest Fisher exact test values, with a mean -log10 p-value of 0.546 (95% 459 C.I.: 0.537, 0.556) yet showed outliers as low as 14.153. Finally, the unweighted pair group 460 method with arithmetic mean (UPGMA) was implemented to generate a dendrogram for 461 the eight pools (Figure 6). The tree showed the correspondent eight groups clustered just 462 as the pairs of each of the four localities, with a closer genetic similarity between Bocas 463 del Toro – Providencia and Carrie Bow Cay – Cartagena. 464 465

29

466 467 468 Figure 6 Hierarchical cluster for the eight pools. Dendrogram was generated by UPGMA method, 469 using the read counts per allele. The bar length represents 0.02 substitutions per nucleotide 470 position and values of nodes their correspondent heights. BDT: Bocas del Toro. CAR: Cartagena. 471 CBC: Carrie Bow Cay. PRV: Providencia. 472 473 474 DISCUSSION 475 476 We performed a genome wide analysis to detect the genetic background between 477 extreme morphotypes of an octocoral exhibiting a phenotypic divergence across the 478 depth cline. Four locations were analysed for genetic structure to elucidate if a gene flow 479 reduction is taking place between habitats and to test if it could be occurring in a parallel 480 fashion. A reference genome from the species was generated to align the SNPs from the 481 population screening. This sampling gives us a broad outlook of the relationship between 482 phenotypically divergent forms defined as one species across geographic scales in the 483 Caribbean. At the same time, a parallel process of gene flow reduction between 484 morphotypes, related to an environmental selective pressure, give us statistical power to 485 infer an isolation by adaptation scenario. 486 487 Genome assembly and annotation 488 The preliminary draft assembly for A. bipinnata generated in our study represents the first 489 published for the species and one of the few attempts for the Octocorallia subclass, 490 including the one from Renilla reniformis (Kayal et al. 2018), Pacifigorgia irene (J. Sanchez 491 unpublished) and Paragorgia stephencairnsi (S. Herrera, unpublished). Already, the state 492 of the released draft is helpful for addressing many questions like reconstructing the 493 phylogeny for Cnidaria (Quattrini et al. 2017). Even with a 75X of coverage, our draft is still 494 fragmented as a consequence of a lack of longer reads to gap-fill repetitive regions that 495 seems to be common in other Cnidaria genomes (Putnam et al. 2007; Wang et al. 2017 496 and personal observation for Pacifigorgia irene). Even so, a significant improvement was

30

497 achieved by implementing a two-step assembly approach, along with extension and 498 gapfilling methods, compared to attempts based on only one assembler or without 499 gapfilling. 500 501 A common obstacle found in sequencing projects is cross-species contamination in raw 502 data introduced during sample collection, library preparation, and/or sequencing 503 (Merchant, Wood, and Salzberg 2014), which must be detected and removed to ensure 504 the quality and fidelity of the genome assembly. We detected contamination sources in 505 the sequencing data, e.g. those originating from the coral symbiont Symbiodinium and the 506 bacterial coral pathogen Serratia marcescens. Additionally, unexpected contamination 507 corresponding to Lepidoptera was identified, a model used in the laboratory where the 508 coral was sequenced. While these contamination sources were detected through the 509 Blobology pipeline, these could not be completely removed by the contig-based 510 methodology, since assembly fragmentation resulted in many reads not assembling and 511 being omitted from the contig-based read mapping and cleaning process. Therefore, 512 additional efforts were implemented in order to remove non-coral reads, based on blast 513 matches with strict e-value and identity parameters and in-house scripts for filtering. This 514 highlights the importance of applying mixed-read analyses in genomic projects and 515 implementing both contig-based and read-based cleaning methods for removing 516 contamination when fragmented assemblies are an issue. 517 518 The gene content prediction and annotation process were hindered by the fragmented 519 state of the assembly. In addition, the lack of species-specific gene models and EST or 520 protein evidence for proper training of ab-initio prediction programs highlight the 521 challenges faced during gene finding and annotation in non-model organisms. Finally, 522 since contamination sources were still detected in the predicted protein set, despite 523 cleaning efforts carried out on the reads, the number of proteins that could be annotated 524 provided a limited knowledge of the gene content of the genome. Further sequencing 525 efforts yielding longer reads should provide better results for genome assembly and 526 annotation. 527 528 Comparison between methodologies 529 The choice between the different kinds of reduced-representation techniques depends on 530 the particular model, questions, and resources to ensure that the same homologous 531 regions are assessed between all individuals (Hohenlohe et al. 2010). This strategy coupled 532 with sample pooling is a high throughput and cost effective method to analyse the A. 533 bipinnata complex, and it has been shown to improve SNP discovery, leading to better 534 estimates of population allele frequencies (Futschik and Schlötterer 2010). Good 535 determination of the morphotypes is crucial for the pooling strategy. In the case of A. 536 bipinnata, colonial traits are relatively continuous across the depth cline. This was a 537 reason to select only the extremes of the phenotype array in pools, thereby reducing any 538 possible translocations of genotypes in the pooling analysis. 539

31

540 Following the bioinformatic pipelines we found no large differences that changed the final 541 conclusion for all the analyses. The bowtie2 pipeline, mapping directly the ezRAD raw 542 reads, had similar percentages compared with the cleaned reads. This is due to the mixed- 543 read analysis over the genome that allowed us to discard non-coral reads directly in this 544 step. The low percentage of reads mapping in both pipelines is explained mainly by the 545 mentioned fragmented state of the reference, though it still provided us with SNPs 546 belonging to the coral in a complex holobiont system. Even when the BWA mem aligner 547 produced a slightly higher sensibility with 2.55% (584,656) more reads mapped, SNP 548 calling by NGSEP produced 10,808 more SNPs with the Bowtie2 pipeline than with the 549 BWA pipeline. This is at least partially explained by the longer longitudes of reads available 550 starting from raw sequences. 551 552 Geographic isolation between localities represent independent events 553 The Cochran–Mantel–Haenszel plus Fst test and hierarchical clustering showed genetic 554 differences between localities, indicating data independence. The CMH test was 555 performed over Panama and Providencia data, since they produced the greatest quantity 556 of SNPs for better estimation. The CMH-test, which has greater power to identify 557 causative sites compared to Fst (Kofler and Schlotterer 2014) produced a consistent signal 558 of allele differentiation between the two regions assessed. Q-Q plots for observed p- 559 values between localities (figure 4A) showed a stronger deviation from the expected 560 uniform distribution in comparison with Q-Q plot for habitats (Figure 4B) implying a 561 statistically significant correlation between genetic differentiation and geography. Cluster 562 analysis for UPGMA grouped the morphotypes according to their location. Genetic 563 distances depicting relationships among correspond well with the pair of morphotypes 564 and the four localities populations, reflecting independence of the process. 565 566 Geographic isolation between localities was clearly detected even for the closest two 567 localities sampled, Providencia and Bocas del Toro (457 Km apart) that might account for 568 the closer grouping between these localities in the distances tree (Figure 6). A strong 569 pattern of genetic differentiation is facilitated by extensive distances between the 570 different locations and limited larvae dispersion with a conspicuous philopatry. As a 571 surface brooder, female colonies of A. bipinnata retain eggs on the colony surface until 572 they are fertilized by the sperm (Kahng, Benayahu, and Lasker 2011). Then, the embryos 573 develop for a couple of days on the colony surface until they are release into the water 574 column (Harrison and Wallace 1990; Kahng et al. 2011). This strategy limits the potential 575 migration of larvae to kilometers. In summary, based on the reproductive strategy, 576 distances between locations and the genetic structure observed in the present study, we 577 consider the colonies sampled across geographic locations to be isolated populations 578 through allopatry by distance. In this case, it is less likely that gene flow between localities 579 could support a single origin of the divergence between morphotypes (Nosil 2012; 580 Quesada et al. 2007). The genetic structure pattern shown by the four locations in our 581 study indicates that gene flow reduction is ongoing and that it evolved in parallel between 582 the kallos and deep morphotypes of A. bipinnata. 583

32

584 Disruptive natural selection between habitats 585 Allele frequencies, CMH, Fst, Fisher exact tests and distance matrix over the neutral 586 markers in our genomic survey show a pattern of gene flow reduction between kallos and 587 deep morphotypes in a sympatry distribution across the depth. Statistically significant 588 signals were particularly clear for Carrie Bow Cay in Belize, where allele frequencies were 589 as large as 0.9 and Fisher exact test detected the most of SNPs with p-values < 0.005. In 590 contrast, weaker signals were found for Bocas del Toro, though still significant for a 591 distance as small as 7 m. 592 593 Providencia provided the best quality data, with 2,057 SNP-windows for the Fst test and 594 20,704 SNPs for the Fisher exact test. This locality shows greater genetic variation, where 595 the gene flow reduction between habitats can be additionally promoted by the distances 596 between morphotypes. The reduction in sympatric gene flow is exemplified by this 597 locality, highlighting that the high potential larval dispersal is largely restricted to only few 598 meters around mother colonies. With gaps in the knowledge of natural history and 599 genetics of the species, micro spatial analysis in phylogenetically related species provides 600 key information about larval dispersal dynamics. Analysis of the dispersion distance for 601 the sister species A. elisabethae showed a fine scale structure with significant genetic 602 structure at distances as short as <1-5 m (Smilansky and Lasker 2014). Localized larval 603 dispersal seems to be a common pattern for some brooding octocorals, like A. bipinnata, 604 and our results are consistent with the life history of closely related species (Gutiérrez- 605 Rodríguez and Lasker 2004; Lasker et al. 2008; Smilansky and Lasker 2014). 606 607 Based on the basic principle that gene flow levels are a key force affecting genetic 608 differentiation at neutral alleles (Nichols 1996; Nosil 2012; Wright 1931), we can infer 609 explicit predictions for our model (Foll and Gaggiotti 2008; Schluter 2009). In this way, the 610 correlation between reduced gene flow and ecological divergence highlight that depth is a 611 strong ecological variable promoting disruptive natural selection. Our data strongly 612 suggest that environmental variation related to depth gradients (e.g. pressure, light 613 intensity, temperature, oxygen concentration and nutrient concentrations) generate the 614 colonial phenotypic forms we see in nature, as well as promote the gene flow reduction 615 observed via our genomic screening and promoting adaptive genetic divergence. 616 617 Allochrony as a prezygotic barrier to gene flow between morphotypes 618 Research on coral spawning is constrained by the restricted amount of time per year 619 available for study. Even though, testing the synchrony of the process could be a step 620 towards identifying gene flow barriers where differences in the brooding time between 621 morphotypes can be related to the asynchronous timing of gamete maturation (Knowlton 622 et al. 1997). Antillogorgia bipinnata gametes are released and fertilized near to the new 623 moon during the months of October-December (personal observations). With the goal of 624 assessing the timing of spawning between morphotypes, daily observations were made 625 during November-December, 2012 in the locality of Bocas del Toro at two sites, Crawl Key 626 and Hospital Point, where we performed the ezRAD sampling and the phenotypic 627 plasticity experiment (Calixto-Botía and Sánchez 2017). Monitoring across habitats and

33

628 localities (Figure 7) showed that eight days after a new moon, kallos morphotypes from 629 Crawl Key were the first releasing eggs to the surface. Daily records revealed that A. 630 bipinnata spawned progressively every 2-3 m in depth, with the first eggs released from 631 the deep female colonies at 8 m four days after. After 3 days, spawning had completely 632 ended in the shallow habitat. At Hospital Point, surface brooding was first noticeable in 633 deep colonies at 8m, followed by a fast spawn in shallower colonies starting at 3 m depth 634 three days after and for three more days. Additionally, personal observations during 2001, 635 2003 and 2011 showed that, while the onset of surface brooding in A. bipinnata is not 636 fully predictable by lunar phase and varies across sites and colonies, the timing varied 637 consistently with depth. 638

639 640 Figure 7. Colonies of A. bipinnata with surface brooding. Colonies from Crawl Key during spawning 641 in November, 2012. (A) Freshly released eggs (November 21th) at 3 m depth; (B) Five-day old 642 larvae (November 22th) at 2.5 m depth; (C) detail of five-day old larvae on November 23th, 2.5 m 643 depth; (D) Freshly released eggs, November 25th, 6 m depth. 121x97mm (300 x 300 DPI). 644 645 Although across localities the order of spawning by depth was not the same, it is a 646 remarkable that spawning is synchronous with colonies and habitats, and that kallos and 647 deep morphotypes are asynchronous at different sites. Some mechanisms might 648 contribute to the synchrony of neighboring colonies, as pheromone stimuli, and in the 649 same manner to the temporal separation between morphotypes distributed in the depth 650 gradient (Knowlton et al. 1997). Temporal reproductive mismatch has the capacity to 651 reduce population connectivity and could promote rapid evolutionary processes (Marshall 652 et al. 2010). In corals, mismatch of only a couple of hours seems to be enough to promote 653 speciation processes as gamete viability time and optimal sperm concentration for 654 fertilization occur within narrow limits in the water column (Fukami et al. 2003; Taylor and 655 Friesen 2017). Thus, a potential assortative mating leading to prezygotic isolation could be 656 plausible for the A. bipinnata complex. In this scenario, depth-associated spawning can 657 speed up the adaptive genetic divergence for particular genotypes, a mechanism driving 658 ecological speciation. Allochrony seems to be common during the ecological speciation

34

659 continuum because the adaptations can itself generate selection favoring individuals with 660 appropriate habitat preferences and developmental schedules (Nosil 2012, pag. 86). 661 662 CONCLUSIONS 663 Our research resolves the genetic structure for the A. bipinnata morphs, which was 664 challenged in former studies with classical markers, depicting a putative case of adaptive 665 genetic divergence triggered by the bathymetric gradient. The genetic test performed 666 between the four locations shows a signal of geographic isolation, expected for a 667 noticeable philopatric species and the assessed distances. Based on the four confirmed 668 independent events, at every location we were able to detect a clear association between 669 neutral SNPs differentiation and habitats where A. bipinnata morphotypes are distributed, 670 even at bathymetric distances as short as 6 meters. 671 672 Observations of asynchrony between morphotypes might represent a mechanism of 673 assortative mating. In this case, habitat preferences seem to depict a prezygotic barrier 674 reducing gene flow at an early stage of the speciation continuum. Allochrony could be a 675 more plausible mechanism for gene flow reduction in the case of the A. bipinnata 676 divergence than other mechanisms as immigrant inviability, based on previous data 677 showing remarkable survival rates and partial adaptive plastic responses for transplanted 678 morphotypes. Finally, the results of this study will be helpful for future researches that 679 can include other interesting elements to the putative ecological speciation scenario not 680 covered here. The functional background of outliers in adaptive colonial traits, the 681 genome architecture, the genetic structure for intermedial morphotypes and the analysis 682 of the holobiont system will contribute to the understanding of the putative parallel 683 ecological speciation in sympatric populations of the Antillogorgia bipinnata complex. 684 685 686 AKNOWLEDGMENTS 687 Pauley program summer course genomics, Hawai’i University, ToBo lab particularly to Rob 688 Toonen, Zac Forzman and Ingrid Knapp. Marco Cristancho, Adriana Sarmiento, Luisa 689 Dueñas, Diana Vergara, Elena Quintanilla, Catalina Ramirez, Stefanie Colmenares. Thanks 690 to Peter Andolfatto for the genome sequencing advices and services. Special thanks to 691 Sirius Dive, Providencia and Courtenay Ray. Informal discussion with Robert Kofler and 692 Björn Stelbrink was greatly appreciated. Thanks to Colciencias (Programa Doctorados 693 Nacionales to I.F. Calixto-Botía). The Smithsonian Tropical Research Institute (STRI) (Senior 694 Latin American Fellow to J.A. Sánchez), Universidad de Los Andes, Colombia (STAI, 695 Vicerrectoría de Investigaciones and Facultad de Ciencias) and CEMarin (Center of 696 Excellence in Marine Sciences) sponsored this research. 697 698 FUNDING 699 This research was sponsored by Colciencias (Programa Doctorados Nacionales to I.F. 700 Calixto-Botía), the Smithsonian Tropical Research Institute (STRI) (Senior Latin American 701 Fellow to J.A. Sánchez), Universidad de Los Andes, Colombia (STAI, Vicerrectoría de

35

702 Investigaciones and Facultad de Ciencias), Proyecto Semilla and CEMarin (Center of 703 Excellence in Marine Sciences). 704 705 COMPETING INTERESTS 706 The authors declare that they have no competing interests. 707 708 CONSENT FOR PUBLICATION 709 Not applicable. 710 711 ETHICS APPROVAL AND CONSENT TO PARTICIPATE 712 The study complies with the current laws of the country in which it was performed. Study 713 permit provided under the project “Plasticidad fenotípica como promotora de especiación 714 ecológica en un coral gorgonáceo del Caribe” by the Smithsonian Tropical Research 715 Institute (STRI) (Senior Latin American Fellow to J.A. Sánchez). 716 717 718 REFERENCES 719 720 Andrews, Simon. 2010. “FastQC: A Quality Control Tool for High Throughput Sequence Data.” 721 Bilewitch, Jaret P. and Sandie M. Degnan. 2011. “A Unique Horizontal Gene Transfer Event Has 722 Provided the Octocoral Mitochondrial Genome with an Active Mismatch Repair Gene That 723 Has Potential for an Unusual Self-Contained Function.” BMC Evolutionary Biology 11:228. 724 Retrieved (http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3166940/). 725 Boetzer, Marten, Christiaan V Henkel, Hans J. Jansen, Derek Butler, and Walter Pirovano. 2011. 726 “Scaffolding Pre-Assembled Contigs Using SSPACE.” Bioinformatics 27(4):578–79. 727 Bromberg, Yana and Burkhard Rost. 2007. “SNAP: Predict Effect of Non-Synonymous 728 Polymorphisms on Function.” Nucleic Acids Research 35(11):3823–35. 729 Bushnell, B. 2016. “BBMap Short Read Aligner.” University of California, Berkeley, California. URL 730 Http://Sourceforge. Net/Projects/Bbmap. 731 Calixto-Botía, Iván and Juan A. Sánchez. 2017. “A Case of Modular Phenotypic Plasticity in the 732 Depth Gradient for the Gorgonian Coral Antillogorgia Bipinnata (Cnidaria: Octocorallia).” 733 BMC Evolutionary Biology 17(1):55. Retrieved (https://doi.org/10.1186/s12862-017-0900-8). 734 Coffroth, Mary Alice, Howard R. Lasker, Margaret E. Diamond, Jeremy A. Bruenn, and Eldredge 735 Bermingham. 1992. “DNA Fingerprints of a Gorgonian Coral: A Method for Detecting Clonal 736 Structure in a Vegetative Species.” Mar. Biol. 114(2):317–25. Retrieved 737 (http://dx.doi.org/10.1007/BF00349534). 738 Conesa, Ana and Stefan Gotz. 2008. “Blast2GO: A Comprehensive Suite for Functional Analysis in 739 Plant Genomics.” International Journal of Plant Genomics 2008. Retrieved 740 (http://dx.doi.org/10.1155/2008/619832). 741 Davey, J. W. and M. L. Blaxter. 2010. “RADSeq: Next-Generation Population Genetics.” Brief Funct 742 Genomics 9:416–423. 743 Desai, Aarti et al. 2013. “Identification of Optimum Sequencing Depth Especially for De Novo 744 Genome Assembly of Small Genomes Using Next Generation Sequencing Data.” PLOS ONE 745 8(4):1–11. Retrieved (https://doi.org/10.1371/journal.pone.0060204). 746 Duitama, Jorge et al. 2014. “An Integrated Framework for Discovery and Genotyping of Genomic 747 Variants from High-Throughput Sequencing Experiments.” Nucleic Acids Research 42(6):e44.

36

748 Retrieved (+). 749 Emerson, K. J. et al. 2010. “Resolving Postglacial Phylogeography Using High-Throughput 750 Sequencing.” Proceedings of the National Academy of Sciences 107(37):16196–16200. 751 Feder, J., S. Egan, and P. Nosil. 2012. “The Genomics of Speciation with Gene Flow.” Trends in 752 Genetics 28(7):342–50. 753 Foll, Matthieu and Oscar Gaggiotti. 2008. “A Genome-Scan Method to Identify Selected Loci 754 Appropriate for Both Dominant and Codominant Markers: A Bayesian Perspective.” Genetics 755 180(2):977–93. Retrieved (http://www.genetics.org/content/180/2/977). 756 Fukami, H., M. Omori, K. Shimoike, T. Hayashibara, and M. Hatta. 2003. “Ecological and Genetic 757 Aspects of Reproductive Isolation by Different Spawning Times in Acropora Corals.” Marine 758 Biology 142(4):679–84. Retrieved (https://doi.org/10.1007/s00227-002-1001-8). 759 Futschik, A. and C. Schlötterer. 2010. “The next Generation of Molecular Markers from Massively 760 Parallel Sequencing of Pooled DNA Samples.” Genetics 186(1):207–18. 761 Gurevich, Alexey, Vladislav Saveliev, Nikolay Vyahhi, and Glenn Tesler. 2013. “QUAST: Quality 762 Assessment Tool for Genome Assemblies.” Bioinformatics (Oxford, England) 29(8):1072–75. 763 Gutiérrez-Rodríguez, C. and H. R. Lasker. 2004. “Reproductive Biology, Development, and Planula 764 Behavior in the Caribbean Gorgonian Pseudopterogorgia Elisabethae.” Invertebr. Biol. 765 123:54–67. 766 Harrison, P. L. and C. C. Wallace. 1990. Reproduction, Dispersal and Recruitment of Scleractinian 767 Corals. Coral Reefs, Ecosystems of the World. edited by Z. Dubinsky. Elsevier, Amsterdam, 768 Netherlands. 769 Hellberg, Michael E. 2006. “No Variation and Low Synonymous Substitution Rates in Coral MtDNA 770 despite High Nuclear Variation.” BMC Evolutionary Biology 6:24. Retrieved 771 (http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1431588/). 772 Hohenlohe, P. A., J. Catchen, and W. A. Cresko. 2012. “Population Genomic Analysis of Model and 773 Nonmodel Organisms Using Sequenced RAD Tags.” Methods in Molecular Biology 888:235– 774 60. 775 Hohenlohe, Paul A. et al. 2010. “Population Genomics of Parallel Adaptation in Threespine 776 Stickleback Using Sequenced RAD Tags.” PLoS Genetics 6(2). 777 Holt, Carson and Mark Yandell. 2011. “MAKER2: An Annotation Pipeline and Genome-Database 778 Management Tool for Second-Generation Genome Projects.” BMC Bioinformatics 12:491. 779 Huson, D. H., S. Mitra, N. Weber, H. Ruscheweyh, and S. C. Schuster. 2011. “Integrative Analysis of 780 Environmental Sequences Using MEGAN4.” Genome Research 21:1552–60. 781 Huson, Daniel H. and David Bryant. 2006. “Application of Phylogenetic Networks in Evolutionary 782 Studies.” Molecular Biology and Evolution 23(2):254–67. 783 Kahng, Samuel, Yehuda Benayahu, and Howard Lasker. 2011. “Sexual Reproduction in Octocorals.” 784 Marine Ecology Progress Series 443:265–83. Retrieved 785 (https://www.researchgate.net/publication/272863701_Sexual_reproduction_in_octocorals) 786 . 787 Kayal, Ehsan et al. 2018. “Phylogenomics Provides a Robust Topology of the Major Cnidarian 788 Lineages and Insights on the Origins of Key Organismal Traits.” BMC Evol Biol 18(68). 789 Knowlton, N., J. Mate, H. Guzmán, R. Rowan, and J. Jara. 1997. “Direct Evidence for Reproductive 790 Isolation among the Three Species of the Montastraea Annularis Complex in Central America 791 (Panama´ and Honduras).” Marine Biology 127:705–11. 792 Kofler, R., R. Vinay-Pandey, and C. Schloetterer. 2011. “PoPoolation2: Identifying Differentiation 793 between Populations Using Sequencing of Pooled DNA Samples (Pool-Seq);” Bioinformatics 794 27(24):3435–3436. Retrieved 795 (http://bioinformatics.oxfordjournals.org/content/27/24/3435).

37

796 Kofler, Robert and Christian Schlotterer. 2014. “A Guide for the Design of Evolve and Resequencing 797 Studies.” Molecular Biology and Evolution 31(2):474–83. 798 Kumar, Sujai, Martin Jones, Georgios Koutsovoulos, Michael Clarke, and Mark Blaxter. 2013. 799 “Blobology: Exploring Raw Genome Data for Contaminants, Symbionts and Parasites Using 800 Taxon-Annotated GC-Coverage Plots.” Frontiers in Genetics 4:237. 801 Landry, P. A., M. T. Koskinen, and C. R. Primmer. 2002. “Deriving Evolutionary Relationships among 802 Populations Using Microsatellites and (Δμ)2: All Loci Are Equal, but Some Are More Equal 803 than Others. .” Genetics 161(1339–134). 804 Langmead, Ben and Steven L. Salzberg. 2012. “Fast Gapped-Read Alignment with Bowtie 2.” 805 Nature Methods 9:357. Retrieved (http://dx.doi.org/10.1038/nmeth.1923). 806 Lasker, H., C. Gutierrez-Rodriguez, K. Bala, A. Hannes, and J. Bilewitch. 2008. “Male Reproductive 807 Success during Spawning Events of the Octocoral Pseudopterogorgia Elisabethae.” Marine 808 Ecology Progress Series 367:153–61. 809 Lasker, H. R., M. A. Boller, J. Castanaro, and J. A. Sánchez. 2003. “Determinate Growth and 810 Modularity in a Gorgonian Octocoral.” Biol Bull 205:319–30. 811 Li, H. and R. Durbin. 2009. “Fast and Accurate Short Read Alignment with Burrows-Wheeler 812 Transform.” Bioinformatics (25):1754–60. 813 Luikart, G., P. R. England, D. Tallmon, S. Jordan, and P. Taberlet. 2003. “The Power and Promise of 814 Population Genomics: From Genotyping to Genome Typing.” Nat Rev Genet 4:981–994. 815 Marshall, J. D., K. Monro, M. Bode, M. J. Keough, and S. Swearer. 2010. “Phenotype–Environment 816 Mismatches Reduce Connectivity in the Sea.” Ecology Letters 13(1):128–40. 817 Martin, Marcel. 2011. “Cutadapt Removes Adapter Sequences from High-Throughput Sequencing 818 Reads.” EMBnet.Journal; Vol 17, No 1: Next Generation Sequencing Data Analysis. 819 Medina, M. et al. 2006. “Naked Corals: Skeleton Loss in Scleractinia.” Proceedings of the National 820 Academy of Sciences 103(24):9096–9100. 821 Merchant, Samier, Derrick E. Wood, and Steven L. Salzberg. 2014. “Unexpected Cross-Species 822 Contamination in Genome Sequencing Projects.” PeerJ 2:e675. Retrieved 823 (https://doi.org/10.7717/peerj.675). 824 Nadalin, Francesca, Francesco Vezzi, and Alberto Policriti. 2012. “GapFiller: A de Novo Assembly 825 Approach to Fill the Gap within Paired Reads.” BMC Bioinformatics 13(Suppl 14):S8–S8. 826 Nichols, Mark A. Beaumont and Richard A. 1996. “Evaluating Loci for Use in the Genetic Analysis of 827 Population Structure.” Proc. R. Soc. Lond 263:1619–26. 828 Nosil, P., D. J. Funk, and D. Ortiz-Barrientos. 2009. “Divergent Selection and Heterogeneous 829 Genomic Divergence.” Mol. Ecol. 18(3):375–402. Retrieved 830 (http://www.scopus.com/inward/record.url?eid=2-s2.0- 831 58749092473&partnerID=40&md5=1b720212f4aae295326a0ea04f387c59). 832 Nosil, Patrik. 2007. “Divergent Host Plant Adaptation and Reproductive Isolation between 833 Ecotypes of Timema Cristinae Walking Sticks.” The American Naturalist 169(2):151–62. 834 Retrieved (https://doi.org/10.1086/510634). 835 Nosil, Patrik. 2012. Ecological Speciation. Oxford University Press. 836 Van Oppen, M. J. H., B. L. Willis, H. W. J. Van Vugt, and D. J. Miller. 2000. “Examination of Species 837 Boundaries in the Acropora Cervicornis Group (Scleractinia, Cnidaria) Using Nuclear DNA 838 Sequence Analyses.” Molecular Ecology 9(9):1363–73. 839 Parkinson, John Everett, Mary Alice Coffroth, and Todd C. LaJeunesse. 2015. “New Species of Clade 840 B Symbiodinium (Dinophyceae) from the Greater Caribbean Belong to Different Functional 841 Guilds: S. Aenigmaticum Sp. Nov., S. Antillogorgium Sp. Nov., S. Endomadracis Sp. Nov., and 842 S. Pseudominutum Sp. Nov.” Journal of Phycology 51(5):850–58. Retrieved 843 (https://onlinelibrary.wiley.com/doi/abs/10.1111/jpy.12340).

38

844 Patterson, Kathryn L. et al. 2002. “The Etiology of White Pox, a Lethal Disease of the Caribbean 845 Elkhorn Coral, Acropora Palmata.” Proceedings of the National Academy of Sciences of the 846 United States of America 99(13):8725–30. Retrieved 847 (http://www.ncbi.nlm.nih.gov/pmc/articles/PMC124366/). 848 Pigliucci, M., C. J. Murren, and C. D. Schlichting. 2006. “Phenotypic Plasticity and Evolution by 849 Genetic Assimilation.” J. Exp. Biol. 209:2362–67. 850 Puritz, J. B., J. A. Addison, and R. J. Toonen. 2012. “Next-Generation Phylogeography: A Targeted 851 Approach for Multilocus Sequencing of Non-Model Organisms.” PLoS ONE 7(3):e34241. 852 Putnam, Nicholas H. et al. 2007. “Sea Anemone Genome Reveals Ancestral Eumetazoan Gene 853 Repertoire and Genomic Organization.” Science (New York, N.Y.) 317(5834):86–94. 854 Quattrini, Andrea M. et al. 2017. “Universal Target-Enrichment Baits for Anthozoan (Cnidaria) 855 Phylogenomics: New Approaches to Long-Standing Problems.” Molecular Ecology Resources. 856 Quesada, Humberto, David Posada, Armando Caballero, Paloma Morán, and Emilio Rolán-Alvarez. 857 2007. “PHYLOGENETIC EVIDENCE FOR MULTIPLE SYMPATRIC ECOLOGICAL DIVERSIFICATION 858 IN A MARINE SNAIL.” Evolution 61(7):1600–1612. Retrieved 859 (https://onlinelibrary.wiley.com/doi/abs/10.1111/j.1558-5646.2007.00135.x). 860 Sánchez, J. A. and H. Lasker. 2004. “Do Multi-Branched Colonial Organisms Exceed Normal Growth 861 after Partial Mortality?” Proceedings of the Royal Society -Biological Sciences 271:S117–20. 862 Sánchez, Juan, Catalina Aguilar, Daniel Dorado, and Nelson Manrique. 2007. “Phenotypic Plasticity 863 and Morphological Integration in a Marine Modular Invertebrate.” BMC Evolutionary Biology 864 7(1):122. Retrieved (http://www.biomedcentral.com/1471-2148/7/122). 865 Savolainen, O., M. Lascoux, and J. Merilä. 2013. “Ecological Genomics of Local Adaptation.” Nature 866 Reviews Genetics 14(11:807–20. 867 Schluter, D. and L. Nagel. 1995. “Parallel Speciation by Natural Selection.” The American Naturalist 868 146(2):292– 301. 869 Schluter, Dolph. 2009. “Evidence for Ecological Speciation and Its Alternative.” Science 870 323(5915):737–41. Retrieved (http://www.sciencemag.org/content/323/5915/737.abstract). 871 Schmieder, R. and R. Edwards. 2011a. “Fast Identification and Removal of Sequence 872 Contamination from Genomic and Metagenomic Datasets.” PLoS ONE 6(3). 873 Schmieder, R. and R. Edwards. 2011b. “Quality Control and Preprocessing of Metagenomic 874 Datasets.” Bioinformatics 27:863–64. 875 Shendure, J., R. Mitra, C. Varma, and G. Church. 2004. “Advanced Sequencing Technologies: 876 Methods and Goals.” Nature Reviews Genetics 5(5):335–44. 877 Simão, Felipe A., Robert M. Waterhouse, Panagiotis Ioannidis, Evgenia V Kriventseva, and Evgeny 878 M. Zdobnov. 2015. “BUSCO: Assessing Genome Assembly and Annotation Completeness with 879 Single-Copy Orthologs.” Bioinformatics 31(19):3210–12. 880 Simpson, Jared T. et al. 2009. “ABySS: A Parallel Assembler for Short Read Sequence Data.” 881 Genome Research 19(6):1117–23. 882 Smilansky, V. and Howard R. Lasker. 2014. “Fine-scale Genetic Structure in the Surface Brooding 883 Caribbean Octocoral, Antillogorgia Elisabethae.” Mar Biol 161(4):853–61. 884 Stanke, Mario et al. 2006. “AUGUSTUS: Ab Initio Prediction of Alternative Transcripts.” Nucleic 885 Acids Research 34(Web Server issue):W435–39. 886 Taylor, Rebecca and Vicki Friesen. 2017. “The Role of Allochrony in Speciation.” Molecular Ecology 887 26(13):3330–42. Retrieved (https://doi.org/10.1111/mec.14126). 888 Thibert-Plante, X. and A. P. Hendry. 2011. “The Consequences of Phenotypic Plasticity for 889 Ecological Speciation.” Journal of Evolutionary Biology 24(2):326–342. 890 Toonen, R. J. et al. 2013. “EzRAD: A Simplified Method for Genomic Genotyping in Non-Model 891 Organisms .” PeerJ 1:e203. Retrieved (http://dx.doi.org/10.7717/peerj.203).

39

892 Vollmer, S. V. and S. R. Palumbi. 2004. “Testing the Utility of Internally Transcribed Spacer.” 893 Wang, Xin et al. 2017. “Draft Genomes of the Corallimorpharians Amplexidiscus Fenestrafer and 894 Discosoma Sp.” Molecular Ecology Resources 17(6):e187–95. 895 Williams, Gary C. and JEI-YING Chen. 2012. “Resurrection of the Octocorallian Genus Antillogorgia 896 for Caribbean Species Previously Assigned to Pseudopterogorgia, and a Taxonomic 897 Assesment of the Relationship of These Genera with Leptogorgia (Cnidaria, Anthozoa, 898 Gorgoniidae).” ZOOTAXA 3505:39–52. 899 Wright, S. 1931. “Evolution in Mendelian Populations.” Genetics 16:97–159. 900 Zimin, Aleksey V et al. 2013. “The MaSuRCA Genome Assembler.” Bioinformatics 29(21):2669–77. 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936

40

937 -Supplementary- 938 TESTING ADAPTIVE GENETIC DIVERGENCE IN PARALLEL ACROSS THE DEPTH CLINE: 939 POPULATION GENOMICS IN THE GORGONIAN CORAL 940 Antillogorgia bipinnata-kallos (Cnidaria: Octocorallia) 941 942 943 GENOME 944 Contamination in sequencing data 945 During quality control, read contamination was suspected based on a minor peak 946 observed at 63% GC on the ‘per sequence GC content’ graph of the FastQC report, in 947 addition to the expected peak at 35% GC. Contamination was further confirmed using the 948 Blobology pipeline, where reads belonging to the order Lepidoptera (mainly Papilio spp. 949 and Bombyx spp.), Enterobacteriales (mainly Serratia marcescens), and Suessiales 950 (Symbiodinium minutum, belonging to the clade B1 of S. antillogorgium), among others 951 were detected based on contig matches to the Genbank non-redundant nucleotide 952 sequence database (Figure S1). 953

954 955 Figure S1. Blobplot showing the presence of contamination sources in a preliminary contig 956 assembly (n=499598 contigs) of the reads sequenced from the genomic paired end library of A. 957 bipinnata. The main contamination source detected was Lepidoptera (n=1986, 0.4% of total 958 contigs) and matches to other Arthropoda orders were found by association. Contig sequences are 959 plotted based on their GC content (X-axis) and coverage (Y-axis) and colored by taxonomic group 960 (Order) as specified by the color key. 961 962 Genome Quality Statistics 963 The A. bipinnata genome draft assembly was performed with a two-step assembly 964 approach based on MaSuRCA and ABySS short read assemblers. ABYSS-rescaffolding was 965 performed using a preliminary contig assembly generated by MaSuRCA. Assembly 966 optimization was done by reducing read coverage from the original 86X after read filtering

41

967 to 70X and 75X coverages. The best result of the two attempts was found with a 968 subsample of the cleaned and processed reads at a coverage of 75X over the genome 969 (estimated at 300Mbp by flow cytometry, Dueñas et al. unpublished). A scaffold assembly 970 was obtained using ABySS, with an optimized k parameter of 25, followed by an extended 971 and gapfilled assembly obtained with SSPACE and GapFiller (gaps reduced from 972 44,224,603 N’s to 9,094,289 N’s). Table S1 compares the statistics for both assembly 973 versions. 974 975 Table S1. Comparison of the statistics of the ‘scaffold’ and ‘gapfilled’ genome draft assemblies for 976 A. bipinnata. The best values for each statistic between the two assemblies are shown in bold. The 977 best value for total length (bp) is based on an estimated genome size of 300 Mbp for A. bipinnata. 978 Metric Scaffold Gapfilled assembly assembly Total length (bp) 263,012,593 284,441,007 Number of scaffolds 128,314 128,219 Number of scaffolds ≥1 Kbp 35,901 36,503 Number of scaffolds ≥10 Kbp 6,905 7,882 Maximum scaffold length (bp) 142,300 149,494 N50 9,217 10,438 N50 length (bp) 7,680 7,494 GC (%) 35.61 35.71 Number of gaps (N) 44,224,603 9,094,289 Number of gaps per 100 Kbp 16,815 3,197.25 979 980 Genome ploidy 981 A ploidy level test based on relative allele frequencies in heterozygous sites was 982 performed over the assembled reference genome draft of A. bipinnata relevant for 983 further analysis. The majority of the alleles have a frequency close to 0.5 for both BWA 984 and Bowtie2 approaches, as expected for a diploid genome pattern of heterozygosity.

985

42

986 Figure S2. Ploidy estimation by NGSEP showing a typical diploid genome. (A) reads screened for 987 mixed-reads and mapped with BWA. (B) raw reads mapped with Bowtie2. 988 989 Genome annotation 990 Genome annotation was carried out based on gene predictions using the MAKER2 (Holt 991 and Yandell 2011) pipeline; specifically, ab initio predictions were generated with SNAP 992 (Bromberg and Rost 2007) and Augustus (Stanke et al. 2006) programs, as well as with 993 protein homology evidence-based predictions against RefSeq protein sequences for 994 Anthozoa. Ciona intestinalis gene models were used for running SNAP predictions and 995 Amphimedon gene models were used for Augustus predictions. The predicted proteins 996 were annotated based on local BLASTP searches against the Genbank non-redundant (nr) 997 protein database, as well as through protein domain prediction based on pfam-hmmer 998 searches. 999 1000 Although gene prediction and annotation were limited by the fragmentation of this 1001 assembly, the annotation process predicted 2,521 genes within the 7,882 scaffolds with 1002 lengths above 10 Kbp. The gene models of C. intestinalis and Amphimedon and protein 1003 evidence for Anthozoa were used to support this process. To validate the predicted gene 1004 models, 978 unique copy orthologs from the Metazoa odb9 dataset were searched using 1005 the software tool BUSCO (http://busco.ezlab.org/). Interestingly, 684 (69.93%) A. 1006 bipinnata orthologs were identified in this analysis, suggesting that gene-rich regions were 1007 assembled in the largest contigs. 1008 1009 The gene content prediction and annotation process were hindered by the fragmented 1010 state of the assembly. For gene prediction, scaffold sequences under 10 Kbp in length 1011 were filtered to avoid incomplete and/or incorrect gene predictions, which reduced the 1012 number of scaffolds by 99.8%. In addition, the lack of species-specific gene models and 1013 EST or protein evidence for proper training of ab-initio prediction programs highlight the 1014 challenges faced during gene finding and annotation in non-model organisms, such as A. 1015 bipinnata. 1016 1017 ezRAD analyses 1018 CMH test with the BWA pipeline 1019 1913 SNPs were identified for CMH following the BWA pipeline. The test between 1020 localities gave a median p-value for BWA of 0.341 (95% C.I.: 0.326, 0.357) and the test 1021 between habitats gave a mean p-value of 0.535 (95% C.I.: 0.520, 0.551). Figure S3 and S4 1022 show Q-Q and Manhattan plots by geography and habitats, respectively, with similar 1023 patterns of differentiation in allele frequencies in comparison with Bowtie2 results. 1024

43

1025 1026 Figure S3. Manhattan and Q-Q plots of CMH test for geography between Panama and Providencia 1027 locations, with colonies from the same location as replicates. BWA bioinformatic pipeline. Y axis 1028 shows -log10 of CMH test and x axis the position of SNPs on scaffolds. Blue line and red lines in 1029 Manhattan plot indicates the threshold for significant SNP at p < 1 × 10−5 and p < 5 × 10−8 1030 respectively. 1031 1032 1033

1034 1035 Figure S4. Manhattan and Q-Q plots of CMH test for habitat between Panama and Providencia 1036 locations, with colonies from the same habitat as replicates. BWA bioinformatic pipeline. Y axis 1037 shows -log10 of CMH test and x axis the position of SNPs on scaffolds. Blue line and red lines in 1038 Manhattan plot indicates the threshold for significant SNP at p < 1 × 10−5 and p < 5 × 10−8 1039 respectively. 1040 1041 Fst test for the BWA pipeline 1042 Pairwise Fst analysis following the BWA pipeline between habitats, for Bocas del Toro a 1043 mean Fst = 0.023 (95% C.I.: 0.020, 0.025); Cartagena a mean Fst = 0.034 (95% C.I.: 0.011, 1044 0.056); Carrie Bow Cay a mean Fst = 0.049 (95% C.I.: 0.031, 0.068) and Providencia, very 1045 similar to Bowtie2, with a mean Fst = 0.024 (95% C.I.: 0.023, 0.025).

44

1046

1047 1048 Figure S5. Fst pairwise average between kallos and deep morphotypes for the four locations with 1049 BWA pipeline. Y axis shows Fst values and x axis the position of SNPs on scaffolds. 1050 1051 1052 Fisher Exact Test for the BWA pipeline 1053 Fst pairwise values between habitats for each location were tested for significance in 1054 allele frequencies using the Fisher Exact Test (Figure S5). Fisher exact test showed 6 SNPs -5 1055 under a p < 1 x 10 . Bocas del Toro had a mean -log10 p-value of 0.474 (95% C.I.: 0.446, 1056 0.502), Cartagena with 0.624 (95% C.I.: 0.421, 0.826), Carrie Bow Cay 1.015 (95% C.I.: 1057 0.793, 1.236), and Providencia a mean -log10 p-value of 0.540 (95% C.I.: 0.530, 0.550). 1058

1059 1060 Figure S6. Fisher exact test between kallos and deep morphotypes for the four locations following 1061 the BWA pipeline. Y axis shows -log10 of Fisher exact test and x axis the position of SNPs on 1062 scaffolds. 1063 1064 Scaffold annotation for outliers 1065 Scaffolds of interest were assessed for biological background. 449 scaffolds with lengths 1066 between 200 and 18461 bp gave outliers signals in CMH and a Fst test. MAKER predictions 1067 were run for BLASTP against the nr database of NCBI and HMMER-pfam for protein 1068 domains, annotating 58 scaffolds for CDS or putative genes. After filtering for hypothetical 1069 proteins, uncharacterized proteins and transposon genes, the annotation process 1070 produced only 7 genes, reflecting the scarce representation of Cnidaria information in 1071 databases (Table S2). 1072 1073 Table S2. Putative genes associated to outliers between shallow and deep habitats. 1074

45

SCAFFOLD SPECIE IDENTIFIER DESCRIPTION

11769 - PRKCSH-like Glucosidase II beta subunit-like Stylophora 25311 pistillata DUF1758 Putative peptidase, likely act as aspartic peptidases Exaiptasia 28498 pallida Lipase_GDSL_2 GDSL-like Lipase/Acylhydrolase family Acropora 36592 digitifera COX2 Cytochrome C oxidase subunit II, periplasmic domain Nematostella 37897 vectensis DDE_1 Superfamily endonuclease likely responsible for coordinating metal ions Montastraea 38342 faveolata COX1 Cytochrome C and Quinol oxidase polypeptide I Antillogorgia 38991 bipinnata Spond_N Axonal growth trajectory controlling adhesion of embryonic nerve cells Stylophora 39158 pistillata DUF1891 oxidation-reduction process related to alkylated DNA repair proteins 1075

46

1 2 COEVOLUTION IN AN ECOLOGICAL SPECIATION SCENARIO: 3 FINE-SCALE POPULATION GENOMICS IN THE GORGONIAN CORAL COMPLEX 4 Antillogorgia bipinnata-kallos AND ITS SYMBIONT 5 -Preliminary results- 6 7 Iván Calixto-Botía1,2, Matías Gómez-Corrales1, Jorge Duitama3, Thomas Wilke2, Juan A. 8 Sánchez1,4 9 1. Laboratory of Biología Molecular Marina-Biommar, Department of Biological Sciences-Faculty of Sciences, 10 Universidad de Los Andes, Carrera 1E No 18A – 10, P.O. Box 4976, Bogotá, Colombia. 11 2. Department of Animal Ecology and Systematics. Justus-Liebig-Universität. Heinrich-Buff-Ring 26-32 IFZ D- 12 35392, Giessen, Germany. CEMarin student. 13 3. Department of Systems and Computing Engineering-Faculty of Engineering, Universidad de Los Andes, 14 Carrera 1E No 18A – 10, P.O. Box 4976, Bogotá, Colombia. 15 4. Marine Sciences, International Giessen Graduate Centre for the Life Sciences (GGL), Justus-Liebig- 16 Universität-Gießen, Germany 17 18 19 ABSTRACT 20 Integrative studies on ecological speciation are filling the gap between molecular data and 21 the ecological variables related to adaptive genetic divergence. The present research 22 performed a genome wide analysis of the gorgonian coral complex Antillogorgia 23 bipinnata-kallos, who display a gradual phenotypic divergence across a bathymetric cline. 24 We assessed 142 individuals in two close locations for genetic structure of the coral and 25 the symbiont, including the intermedial space of the phenotypic divergence, and to 26 characterize potential outliers associated to adaptive traits. A new genome for the species 27 was generated, producing a higher number of SNPs than in a previous study. The 28 preliminary results depict a remarkable genetic differentiation between the extreme 29 forms of the complex. These results support the taxonomical differentiation between A. 30 kallos and A. bipinnata in a recent process of genetic divergence promoted by depth, 31 shaping a putative scenario of ecological speciation. 32 33 34 INTRODUCTION 35 36 the ecological speciation, coined by Schluter (2009), is a main concept in evolution to 37 differentiate one of the two big processes by which natural selection can produce new 38 species and can be defined as the evolution of reproductive isolation by divergent natural 39 selection in populations adapting to different ecological environments (Nosil, 2012). 40 Ecological speciation serves as a framework to text explicit predictions over our study 41 models to understand the extent of natural selection to explain the diversification patters 42 we observe in the living forms, where its mechanisms and rhythm, remains unknown 43 (Schluter, 2009). For marine organisms, one of the key environmental variables proposed 44 to understand contemporary speciation patterns, particularly in sympatric scenarios, is the 45 depth gradient. In marine benthic communities, depth has been widely recognized as a

47

46 diversification through marine habitats. However, sound evidence of depth as an 47 ecological speciation force has remained scant for marine invertebrates. 48 49 The present research aims to understand the diversification pattern for two close related 50 species in a sympatric distribution, where the divergence of colonial forms overlaps 51 between species across a depth cline. Antillogorgia bipinnata (Verrill 1864) and A. kallos 52 (Bielschowsky 1918) (=Pseudopterogorgia, Williams & Chen, 2012) are sister species of 53 the Gorgoniidae family (Octocorallia) and conforms a group of Caribbean corals with 54 broad environmental and sympatric distribution ranges, (Sánchez, Aguilar, Dorado, & 55 Manrique, 2007). Even with the conspicuous similarity in colonial traits between species, 56 Bayer (1961) sustained the separation between the species, based on the scaphoid 57 sclerites (specific forms of microscopic calcite structures along tissues) as the definitive 58 diagnostic character (Bayer, 1961). Based on classical markers, A. kallos has been 59 proposed as the shallower morphotype of A. bipinnata (Sánchez et al., 2007), including 60 intermedial and deep morphotypes and illustrating the complex panorama in evolutionary 61 studies for this species. Having in mind these observations, in the present study we refer 62 to the A. bipinnata complex as the sensu stricto species plus the A. kallos as the shallower 63 morphotype. 64 65 Previous research exploring the phenotypic divergence associate to depth gradients for 66 the Antillogorgia bipinnata-kallos complex, exposed key elements to putatively infer that 67 the phenotypic divergence of the complex is the result of a recent phase of genetic 68 divergence promoted by depth. Reciprocal transplants performed for the extreme forms 69 exposed adaptive plasticity, high survival rates for morphotypes in the foreign habitat and 70 a genetic component explaining the variance of the traits assessed, pointing out to a 71 genomic substrate for natural selection (Calixto-Botía & Sánchez, 2017). Population 72 genomic analyses with a pooling strategy for extreme morphotypes detected association 73 between levels of genetic differentiation and habitats (Calixto-Botía et al. unpublished). 74 Furthermore, the process was detected to occur at four locations hundreds of kilometers 75 apart in the Caribbean, supporting a scenario of parallel genetic divergence mediated by 76 the depth. Additionally, by observations of reproductive asynchrony between habitats, a 77 mechanism for gene flow reduction and reinforcement in an early stage of the divergence 78 was suggested. 79 80 Here, we part from these foundations to deepen in the genetic differentiation in the A. 81 bipinnata-kallos complex by the implementation of Next Generation Sequencing 82 technologies of DNA (NGS). This strategy represents an exponential leap in the resolution 83 and uniformity of genetic markers across the coral genome, being able to simultaneously 84 genotype thousands of neutral and adaptive loci outliers from the genomic background. 85 Therefore, the present research takes advantage of the nextRAD (Russello, Waterhouse, 86 Etter, & Johnson, 2015), one of the several methodologies of genome reduced- 87 representation by NGS techniques, to scan thousands of SNP markers across the coral 88 genome and to assess the microscale genetic differentiation along the environmental 89 gradient. These new techniques have led to improvements in the analysis of

48

90 phylogeographic relationships, where the use of traditional genetic markers can produces 91 estimation biases (due loci under genetic hitchhiking or selection) in population 92 differentiation, branch lengths and topology of phylogenies (Landry, Koskinen, & Primmer, 93 2002; Luikart, England, Tallmon, Jordan, & Taberlet, 2003). Providing a much denser 94 genome-wide sample of genotype data, genome reduced-representation techniques are 95 diminished the effect of outlier loci and increased the accuracy over the same population 96 questions (Emerson et al., 2010). 97 98 For this purpose, we individually screened 142 colonies from the kallos, intermedial and 99 deep morphotypes for 2 locations separated by 2 km in San Andres Island, Colombia. We 100 are performing a detailed genetic structure analysis including the intermedial space of the 101 divergence, different populations from previous studies and particularly, assessing rates 102 and directionality of larvae migration and the genetic structure of the algal symbiont. We 103 provide new components to expand our knowledge in the role of depth clines to promote 104 adaptive phenotypes and genetic divergence and its extent in the diversification patterns 105 for sessile cnidarians. 106 107 108 MATERIALS AND METHODS 109 110 Sampling and DNA isolation 111 Two localities in the west side of San Andrés Island, Colombia, were sampled of individual 112 SNP screening, Buconos and West View, with a 2 km distance each other. These localities 113 present a reef slope where the three morphotypes of A. bipinnata are distributed across 114 the depth gradient. For Buconos 24 kallos colonies from 5-10 m of depth were sampled. 115 26 intermedial colonies were sampled at a depth between 15-20 m and 23 colonies from 116 deep morphotype were collected at a depth between 35-40 m. For West View 22 kallos 117 colonies were sampled, corresponding to 5-7,5 m of depth. 22 intermedial colonies were 118 sampled at a depth between 15-17 m and 25 colonies from deep morphotype at a depth 119 between 25-30 m. 120 121 All segments of colonies were collected in DMSO and total genomic DNA was extracted by 122 the CTAB and CIA/FCIA method modified from Coffroth et al., 1992 (Coffroth, Lasker, 123 Diamond, Bruenn, & Bermingham, 1992), and treated with PureLink TM RNase A (10 124 mg/ml, Invitrogen). A Qubit Fluorometer (Thermo- Fisher, Dietikon, Switzerland) was used 125 to measure DNA concentration, and gel (0.7 %) electrophoresis was used to measure 126 quality and purity. 127 128 Genomic data collection and SNP screening 129 NextRAD genotyping-by-sequencing libraries (SNPsaurus, LLC) were constructed from 130 genomic DNA following Russello et al., 2015 (Russello et al., 2015). Genomic DNA was first 131 fragmented with Nextera reagent (Illumina, Inc), which also ligates short adapter 132 sequences to the ends of the fragments. The Nextera reaction was scaled for fragmenting 133 7 ng of genomic DNA, although 10.5 ng of genomic DNA was used for input to compensate

49

134 for the amount of degraded DNA in the samples and to increase fragment sizes. 135 Fragmented DNA was then amplified for 26 cycles at 73 °C, with one of the primers 136 matching the adapter and extending 9 nucleotides into the genomic DNA with the 137 selective sequence GTGTAGAGG. Thus, only fragments starting with a sequence that can 138 be hybridized by the selective sequence of the primer will be efficiently amplified. The 139 nextRAD libraries were sequenced on an Illumina HiSeq 4000 with two partial lanes of 150 140 bp reads (University of Oregon). The genotyping analysis used custom scripts (SNPsaurus, 141 LLC) that trimmed the reads using bbduk, BBMap tools (Bushnell, 2016) with the following 142 parameters: bbmap/bbduk.sh in=$file out=$outfile ktrim=r k=17 hdist=1 mink=8 143 ref=bbmap/resources/nextera.fa.gz minlen=100 ow=t qtrim=r trimq=10. 144 145 Reads were also cleaned of contaminating species using bbduk and the reference 146 genomes of bacteria such as Nocardia farcinica (Ishikawa et al., 2004) and Nocardia 147 nova(Luo, Hiessl, Poehlein, Daniel, & Steinbuchel, 2014), and the reference genome of 148 Symbiodinium minutum (Shoguchi et al., 2013), as a mean to remove reads from the algal 149 symbiont. Next, we mapped the reads to a new genome of A. bipinnata corresponding to 150 a mature deep colony and sequenced as PEX250bp and assembled with abyss-pe (Simpson 151 et al., 2009) after read trimming with bbduk, from BBMap tools (Bushnell, 2016). The 152 remaining loci were then aligned to each other to identify allelic loci and collapse allelic 153 haplotypes to a single representative. All reads were mapped to the reference with an 154 alignment identity threshold of 87% using bbmap (BBMap tools). Genotype calling was 155 done using Samtools and bcftools (samtools mpileup -gu -Q 12 -t DP, DPR -f ref.fasta -b 156 samples.txt | bcftools call -cv - > genotypes.vcf). The vcf was filtered to remove alleles 157 with a population frequency of less than 3%. Heterozygous loci in all samples or that had 158 more than 2 alleles in a sample (suggesting collapsed paralogs), were removed. Artifact 159 absence was checked by counting SNPs at each read nucleotide position and determining 160 that SNP number did not increase with reduced base quality at the end of the reads. 161 162 Population genomic analysis 163 Data quality was maximized by removing all indels, retaining only SNPs that were 164 genotyped in more than 50 % of individuals, had a minor allele frequency of 0.05 and a 165 minimum coverage of 5x. These analyses were executed using the R package “vcfR”. 166 (Knaus & Grünwald, 2017). Clone detection was performed calculating the genetic 167 distance/similarity between all pair of individuals in the R package “poppr”(Kamvar, 168 Brooks, & Grünwald, 2015). The remaining SNPs were then evaluated for significant 169 deviations from the Hardy-Weinberg equilibrium (HWE) in kallos and deep morphotypes 170 using an exact test based on Monte Carlo permutations (1000) of alleles as implemented 171 in the R package “adegenet” (Jombart, 2017). All loci were tested for signals of selection in 172 order to remove those loci putatively violating neutrality assumptions. Population 173 structure was assessed in Structure (Pritchard, Stephens, & Donnelly, 2000) using both the 174 admixture model with correlated allele frequencies and not considering priors (burn-in of 175 100,000 and 50,000 reps), and considering priors with an alpha of 0.5 as recommended by 176 Wang, 2017 (Wang, 2017) for unbalanced size samples. Numbers of cluster (K) varied from 177 1-4, with 10 replicates for each value of K. The optimal value of K was determined using

50

178 the ΔK method (Evanno, Regnaut, & Goudet, 2005) through the web-based software 179 Structure Harvester (Earl, 2012), which results were plotted in Structure Plot (Ramasamy, 180 Ramasamy, Bindroo, & Naik, 2014). 181 182 Population structure was explored via principal component analysis (PCA) and 183 discriminant analysis of principal components (DAPC) in “adegenet”. In the latter method, 184 the number of clusters inferred was estimated by 1000 iterations of the K-means 185 clustering between K=1 and 4 after retaining all PCs, and selecting the optimal number of 186 PCs by 1000 replicates of the a-score (Jombart, 2017). The overall loci dataset was used to 187 calculate observed (Ho) and expected (He) heterozygosity, and inbreeding coefficient (Fis) 188 for deep and kallos morphotypes in the R package “hierfsat”(Goudet, 2005). Population 189 fixation index (Fst) was estimated between depth groups using the above-mentioned 190 package. 191 192 Maximum Likelihood (ML) phylogenetic analyses were performed in the CIPRES web 193 server using the tool RAxML-HPC2 8.1.24 (Stamatakis, 2014). The analysis had 100 194 searches for best tree, and bootstrap (BS) values were calculated under the GTRCAT 195 model (Bray & Bocak, 2016). Finally, phylogenetic trees were reconstructed from the 196 concatenated NextRAD sequence libraries to resolve species, producing high bootstrap 197 support (Brawand et al., 2014). 198 199 Symbiodinium analysis 200 To separate the genetic information of the zooxanthellae from the coral, we mapped the 201 nextRAD reads to a reference, combining the three Symbiodinium genomes reported to 202 date: S. minutum (Shoguchi et al., 2013), S. microadriaticum (Aranda et al., 2016) and S. 203 kawagutii (Lin et al., 2015). Read were aligned to the reference with BWA mem (Li & 204 Durbin, 2009) and raw SNPs were identified running the FindVariants command of the 205 NGSEP software pipeline (Duitama et al., 2014) with recommended parameters for 206 genotype by sequencing data and considering haploid information (maxBaseQS 30, 207 minQuality 40, maxAlnsPerStartPos 100, -ploidy 1). For population genomics analyses we 208 implemented model-free methods based on genetic distance and principal components 209 due the clonality of Symbiodinium. 210 211 212 RESULTS 213 214 SNP screening 215 For the Symbiodinium analysis, from the initial 141 colonies sampled, 137 produced 4902 216 SNPs shared between them and filling the criteria for quality filtering (Supplementary 217 Figure S2). Preliminary analysis for Minimum spanning network did not show clustering for 218 habitats where samples were collected, while a dendrogram with UPGMA method 219 indicated a signal of clustering by depth in a few colonies, particularly from deep 220 environment (Supplementary Figures S2-S3). 221

51

222 For coral data, after quality filtering 2 colonies out of 47 colonies involving kallos and deep 223 morphotypes were removed from the analysis due to their low read coverage. Clone 224 detection of the 45 remaining colonies identified one clone (> 95 % of similarity), retaining 225 the sample with the least missing data, resulting in a total of 44 colonies: 21 belonging to 226 kallos and 23 to deep morphotypes. A total of 356 SNPs that met the minimum 227 parameters for recovering genotypes were identified, from which 962 independent loci 228 were retained for downstream analyses. Genetic structure by depth was identified with 229 the principal component analysis (PCA), revealing two well defined groups with the 95% of 230 data explained inside ellipses (Figure 1a). Also, the discriminant analysis of principal 231 components (DAPC), clearly showed two genetic clusters corresponding with the habitats 232 where colonies where sampled (Figure 1b). The Bayesian clustering analyses, both 233 considering and not considering priors, also inferred the existence of two genetic cluster 234 or subpopulations (ΔK2 = 172.1141, ΔK2 = 9.9716, respectively) (Figure 1c and see 235 Supplementary Figure S1). Only two deep colonies exhibited a kallos genotype, while four 236 kallos ones displayed a deep origin, reflecting a strong reduction of gene flow between 237 habitats. 238

239 240 Figure 1. Genetic structuring between A. bipinnata-kallos. (a) Principal component analysis (PCA) 241 for overall data set, ellipses represent 95 % of data. (b) Discriminant analysis of principal 242 component (DAPC) depth morphotypes. (c) Structure diagram (K=2) depicting two inferred genetic 243 clusters. 244 245 Observed (Ho), expected heterozygosity (He) and inbreeding coefficient (Fis) were 246 statistically similar in both depths (Table 1). The observed heterozygosity was lightly 247 inferior from the expected (gene diversity), and in consequence we obtained positive 248 values for Fis, as an evidence of endogamy inside habitats, although close to cero. It 249 means, there is a soft deviation form Hardy-Weinberg equilibrium inside habitats as 250 colonies are more related between them than expected under random mating. Lastly, the

52

251 global genetic differentiation between depths, measured by the fixation index (Fst), 252 exhibited a value of 0.0987. Then, having in mind the biallelic nature of SNP marker, 253 gonochorism and dispersal potential, the Fst value indicates a moderated strength of 254 population structure exposed by principal components and structure diagram. Overall 255 congruence of analyses lends sound evidence to genetic divergence between the kallos 256 and deep colonies. 257 258 Table 1. Observed heterozygosity (Ho), expected heterozygosity (He), inbreeding coefficient (Fis) 259 and fixation index (Fst) with 95 % CI [±] values for overall data set. Morphotype Ho He Fis Fst Deep 0.2184 0.2272 0.0607

[0.2002 ± [0.2120 ± Overall [0.0156 ± 0.1040] 0.0987 0.2385] 0.2422] loci Kallos 0.1891 0.2025 0,073 [0.0738 ± 0.1276] 356 SNPs [0.1683 ± [0.1878 ± [0.0438 ± 0.1378] 0.2113] 0.2175] 260 261 262 The A. bipinnata-kallos phylogeny using RAxML showed two well supported and 263 differentiated phylogenetic lineages (Figure 2), in congruence with the principal 264 component analysis and structure diagram. One lineage had a support of 58, containing 19 265 colonies belonging to the kallos morphotype and two individuals from the deep 266 morphotype. The second lineage had a support of 90 for bootstrapping, including the 267 remaining 19 deep colonies and the exception of four colonies belonging to the kallos 268 morph. None of the exceptions had a particular colonial trait that would resemble the 269 form of the opposite environment.

270 Deep Shallow Tree scale: 0.01 SAI1121 SAI1218 SAI1225 Morphotype SAI1215 63 SAI1211 85 Deep SAI1210 SAI1230 Shallow SAI1228 SAI1206 SAI1232 SAI1224 58 SAI1209 SAI1125 54 SAI1220 SAI1214 62 SAI1231 SAI1205 SAI1213 SAI1221 SAI1219 SAI1216 SAI1004 SAI1208 SAI1126 SAI1133 SAI1137 90 SAI1107 SAI1111 SAI1124 SAI1134 SAI1109 SAI1130 SAI1113 SAI1112 60 SAI1131 SAI1129 SAI1223 100 SAI1212 SAI1229 65 SAI1128 68 SAI1136 100 SAI1110 SAI1123 271 92 SAI1108 272 Figure 2. Maximum likelihood phylogenetic tree inferred using RAxML for A. bipinnata-kallos. The 273 nodes present their bootstrap support for 100 replicates, and the tip labels are colored accordingly 274 to depth.

53

275 276 DISCUSSION 277 278 In previous work, the genetic differentiation between A. bipinnata and A. kallos, elusive 279 with classical markers (Dorado & Sánchez, 2009; Sánchez et al., 2007), was implied by 280 analysis of the variance for phenotypic plasticity in an experiment of reciprocal transplants 281 (Calixto-Botía & Sánchez, 2017) and later recognized by a pooling strategy of population 282 genomics across four localities exposing a parallel mode of divergence (Calixto-Botía, et al. 283 unpublished). In the present study, we provide new elements to understand the 284 divergence process for the Antillogorgia bipinnata-kallos complex by performing a 285 population genomic analysis over two close locations with a narrow distribution scale 286 across a depth gradient between 5 m to 45 m. Adding individual information and the 287 intermedial forms of the phenotypic divergence, we obtained a final resolution for genetic 288 differentiation between A. bipinnata and A. kallos, pointing out the advantage of high- 289 resolution SNP screening beside classical markers. Also, here we included an assessment 290 of the genetic diversity for the zooxanthellae as an important element from the holobiont 291 perspective to understand adaptive divergence processes mediated by depth. Preliminary 292 results for this data seem to indicate a genetic clustering associated to depth in some 293 zooxanthellae samples. 294 295 This sampling let us to recognize the gene flow dynamics between the kallos, the 296 intermedial and the deep morphotypes in the continuum of the bathymetric profile. A 297 reference genome with a higher quality from the previously produced, due longer contigs 298 and the elimination of contaminants, was generated to align the nextRAD reads. This new 299 assembly produced a higher number of SNPs variants, providing stronger statistics for the 300 downstream tests. Preliminary analyses from principal components, structure, fixation 301 index and phylogenies show a consistent data structuration corresponding with the depth 302 where colonies were sampled. Having in mind the molecular marker used is biallelic, 303 potential for larvae dispersion and the gonochorism for a brooder coral, the global genetic 304 differentiation between depths with a Fst value of 0.0987 exhibit a moderated strength of 305 population structure (Meirmans & Hedrick, 2011; Wright, 1978). This value supports the 306 previous study (Calixto-Botía, et al. unpublished) where pools of deep and kallos colonies 307 from four different regions of the Caribbean presented moderated and significant levels of 308 genetic differentiation between kallos and deep colonies (from 0.024 for Providencia to 309 0.063 for Crawl Bow Cay). Therefore, the major result of the present study is the 310 confirmation of a remarkable genetic structure between A. kallos and A. bipinnata as two 311 genetically distinct entities. 312 313 Interestingly, for each analysis we found a few colonies corresponding to a genetic profile 314 from the opposite habitat, it means colonies from shallow habitat with colonial traits 315 corresponding to kallos morphotypes but with a genetic profile corresponding to the deep 316 morphotype of A. bipinnata, and vice versa for a few colonies from deep habitats. This 317 finding exposes the dispersion potential of colonies to overcome depth gradients and the

54

318 capacity of adaption to foreign habitats (Kahng, Benayahu, & Lasker, 2011). This is 319 consistent with the adaptive phenotypic plasticity plus survival rates found in a reciprocal 320 transplant experiment previously reported (Calixto-Botía & Sánchez, 2017). Therefore, the 321 genotypes found in alternative habitats could be the result of the larvae potential for 322 migration and the capacity to response with a plastic and adaptive phenotype in the 323 modular traits to the new condition. Also, the few numbers of colonies with this pattern 324 seems to support a non-complete process of the adaptive divergence, in a transition for 325 genetic fixation of adaptive traits in the genomic background (Nosil & Feder, J., 2012; 326 Nosil, 2012). 327 328 Current data point to an intermedial phase of Isolation By Adaptation between habitats, 329 with effective larval dispersal within meters of distance from the maternal colony. 330 Assessing the genetic structure for kallos and deep colonies across different locations in 331 the Caribbean we detected association between genetic differentiation and habitats, at 332 geographic distances covered by dispersion potential. Based on these results, it is inferred 333 that A. kallos and A. bipinnata are genetically distinct, where habitat preferences seem to 334 depict a prezygotic barrier reducing gene flow by an asynchrony of spawning, representing 335 a mechanism of assortative mating (Calixto-Botía et al. unpublished). These results 336 support the taxonomical differentiation between A. kallos and A. bipinnata as sister 337 species in a recent process of genetic divergence promoted by depth, shaping a putative 338 scenario of ecological speciation. 339 340 341 AKNOWLEDGMENTS 342 We want to thank to Erik Johnson and Paul Etter from SNPsaurus company for the 343 nextRAD services. Also, they provided us an improved A. bipinnata genome. Thanks to 344 Adriana Sarmiento, Luisa Dueñas, Diana Vergara, Trigal Velasquez, Elena Quintanilla and 345 Catalina Ramirez for field trip assistance and advices. Thanks to Colciencias (Programa 346 Doctorados Nacionales to I.F. Calixto-Botía). Universidad de Los Andes, Colombia (STAI, 347 Vicerrectoría de Investigaciones and Facultad de Ciencias) and CEMarin (Center of 348 Excellence in Marine Sciences) sponsored this research. 349 350 FUNDING 351 This research was sponsored by Colciencias (Programa Doctorados Nacionales to I. Calixto- 352 Botía), Universidad de Los Andes (STAI, Vicerrectoría de Investigaciones and Facultad de 353 Ciencias), Proyecto Semilla and CEMarin (Center of Excellence in Marine Sciences). 354 355 COMPETING INTERESTS 356 The authors declare that they have no competing interests. 357 358 CONSENT FOR PUBLICATION 359 Not applicable. 360 361 ETHICS APPROVAL AND CONSENT TO PARTICIPATE

55

362 The study complies with the current laws of the country in which it was performed. Colony 363 tissue from colonies sampled were deposited in the Invertebrate collection of the 364 Museum of Natural History from Universidad de Los Andes, Colombia. 365 366 REFERENCES 367 Aranda, M., Li, Y., Liew, Y. J., Baumgarten, S., Simakov, O., Wilson, M. C., … Voolstra, C. R. 368 (2016). Genomes of coral symbionts highlight evolutionary adaptations 369 conducive to a symbiotic lifestyle. Scientific Reports, 6, 39734. 370 https://doi.org/10.1038/srep39734 371 Bayer, F. M. (1961). The shalow water Octocorallia of the West Indian Region. In M. 372 Nijhoff (Ed.), A manual for marine biologists (p. 373). The Hague. 373 Brawand, D., Wagner, C. E., Li, Y. I., Malinsky, M., Keller, I., Fan, S., … Di Palma, F. (2014). 374 The genomic substrate for adaptive radiation in African cichlid fish. Nature, 375 513(7518), 375–381. https://doi.org/10.1038/nature13726 376 Bray, T. C., & Bocak, L. (2016). Slowly dispersing neotenic beetles can speciate on a penny 377 coin and generate space-limited diversity in the tropical mountains. Scientific 378 Reports, 6(1), 33579. https://doi.org/10.1038/srep33579 379 Bushnell, B. (2016). BBMap short read aligner. University of California, Berkeley, California. 380 URL Http://Sourceforge. Net/Projects/Bbmap. 381 Calixto-Botía, I., & Sánchez, J. A. (2017). A case of modular phenotypic plasticity in the 382 depth gradient for the gorgonian coral Antillogorgia bipinnata (Cnidaria: 383 Octocorallia). BMC Evolutionary Biology, 17(1), 55. https://doi.org/10.1186/s12862- 384 017-0900-8 385 Coffroth, M. A., Lasker, H. R., Diamond, M. E., Bruenn, J. A., & Bermingham, E. (1992). DNA 386 fingerprints of a gorgonian coral: a method for detecting clonal structure in a 387 vegetative species. Mar. Biol., 114(2), 317–325. https://doi.org/10.1007/bf00349534 388 Dorado, D., & Sánchez, J. A. (2009). Internal transcribed spacer 2 (its2) variation in the 389 gorgonian coral Pseudopterogorgia bipinnata in Belize and Panama. Smithson. 390 Contrib. Mar. Sci., 38, 173–179. 391 Duitama, J., Quintero, J. C., Cruz, D. F., Quintero, C., Hubmann, G., Foulquié-Moreno, M. 392 R., … Tohme, J. (2014). An integrated framework for discovery and genotyping of 393 genomic variants from high-throughput sequencing experiments. Nucleic Acids 394 Research, 42(6), e44. https://doi.org/10.1093/nar/gkt1381 395 Earl, D. A. (2012). STRUCTURE HARVESTER: a website and program for visualizing 396 STRUCTURE output and implementing the Evanno method. Conservation Genetics 397 Resources, 4(2), 359–361. 398 Emerson, K. J., Merz, C. R., Catchen, J. M., Hohenlohe, P. A., Cresko, W. A., Bradshaw, W. 399 E., & Holzapfel, C. M. (2010). Resolving postglacial phylogeography using high- 400 throughput sequencing. Proceedings of the National Academy of Sciences, 107(37), 401 16196–16200. 402 Evanno, G., Regnaut, S., & Goudet, J. (2005). Detecting the number of clusters of 403 individuals using the software structure: a simulation study. Molecular Ecology, 14(8), 404 2611–2620. https://doi.org/10.1111/j.1365-294X.2005.02553.x 405 Goudet, J. (2005). Hierfstat, a package for R to compute and test hierarchical F-statistics.

56

406 Molecular Ecology Resources, 5(1), 184–186. 407 Ishikawa, J., Yamashita, A., Mikami, Y., Hoshino, Y., Kurita, H., Hotta, K., … Hattori, M. 408 (2004). The complete genomic sequence of Nocardia farcinica IFM 10152. 409 Proceedings of the National Academy of Sciences of the United States of America, 410 101(41), 14925–14930. 411 Jombart, T. (2017). adegenet: a R package for the multivariate analysis of genetic markers. 412 R. 413 Kahng, S., Benayahu, Y., & Lasker, H. (2011). Sexual reproduction in octocorals. Marine 414 Ecology Progress Series, 443, 265–283. https://doi.org/10.3354/meps09414 415 Kamvar, Z. N., Brooks, J. C., & Grünwald, N. J. (2015). Novel R tools for analysis of 416 genome-wide population genetic data with emphasis on clonality. Frontiers in 417 Genetics, 6. https://doi.org/10.3389/fgene.2015.00208 418 Knaus, B. J., & Grünwald, N. J. (2017). VCFR: a package to manipulate and visualize variant 419 call format data in R. Molecular Ecology Resources, 17(1), 44–53. 420 Landry, P. A., Koskinen, M. T., & Primmer, C. R. (2002). Deriving evolutionary relationships 421 among populations using microsatellites and (δμ)2: All loci are equal, but some are 422 more equal than others. . Genetics, 161(1339–134). 423 Li, H., & Durbin, R. (2009). Fast and accurate short read alignment with Burrows-Wheeler 424 transform. Bioinformatics, (25), 1754–60. 425 Lin, S., Cheng, S., Song, B., Zhong, X., Lin, X., Li, W., … Morse, D. (2015). The Symbiodinium 426 kawagutii genome illuminates dinoflagellate gene expression and coral symbiosis. 427 Science, 350(6261), 691–694. https://doi.org/10.1126/science.aad0408 428 Luikart, G., England, P. R., Tallmon, D., Jordan, S., & Taberlet, P. (2003). The power and 429 promise of population genomics: From genotyping to genome typing. Nat Rev Genet, 430 4, 981–994. 431 Luo, Q., Hiessl, S., Poehlein, A., Daniel, R., & Steinbuchel, A. (2014). Insights into the 432 Microbial Degradation of Rubber and Gutta-Percha by Analysis of the Complete 433 Genome of Nocardia nova SH22a. Applied and Environmental Microbiology, 80(13), 434 3895–3907. https://doi.org/10.1128/AEM.00473-14 435 Meirmans, P., & Hedrick, P. (2011). Assessing population structure: FST and related 436 measures. Molecular Ecology Resources, 11(1), 5–18. https://doi.org/10.1111/j.1755- 437 0998.2010.02927.x 438 Nosil & Feder, J., P. (2012). Genomic divergence during speciation: causes and 439 consequences. . Philosophical Transactions of the Royal Society B: Biological Sciences, 440 367(1587), 332–342. 441 Nosil, P. (2012). Ecological Speciation. Oxford University Press. 442 Pritchard, J. K., Stephens, M., & Donnelly, P. (2000). Inference of population structure 443 using multilocus genotype data. Genetics, 155(2), 945–959. 444 Ramasamy, R. K., Ramasamy, S., Bindroo, B. B., & Naik, V. G. (2014). STRUCTURE PLOT: a 445 program for drawing elegant STRUCTURE bar plots in user friendly interface. 446 SpringerPlus, 3(1), 431. 447 Russello, M. A., Waterhouse, M. D., Etter, P. D., & Johnson, E. A. (2015). From promise to 448 practice: pairing non-invasive sampling with genomics in conservation. PeerJ, 3, 449 e1106. https://doi.org/10.7717/peerj.1106

57

450 Sánchez, J., Aguilar, C., Dorado, D., & Manrique, N. (2007). Phenotypic plasticity and 451 morphological integration in a marine modular invertebrate. BMC Evolutionary 452 Biology, 7(1), 122. Retrieved from http://www.biomedcentral.com/1471-2148/7/122 453 Schluter, D. (2009). Evidence for Ecological Speciation and Its Alternative. Science, 454 323(5915), 737–741. https://doi.org/10.1126/science.1160006 455 Shoguchi, E., Shinzato, C., Kawashima, T., Gyoja, F., Mungpakdee, S., Koyanagi, R., … 456 Satoh, N. (2013). Draft Assembly of the Symbiodinium minutum Nuclear Genome 457 Reveals Dinoflagellate Gene Structure. Current Biology, 23(15), 1399–1408. 458 https://doi.org/10.1016/j.cub.2013.05.062 459 Simpson, J. T., Wong, K., Jackman, S. D., Schein, J. E., Jones, S. J. M., & Birol, İ. (2009). 460 ABySS: A parallel assembler for short read sequence data. Genome Research, 19(6), 461 1117–1123. https://doi.org/10.1101/gr.089532.108 462 Stamatakis, A. (2014). RAxML version 8: A tool for phylogenetic analysis and post-analysis 463 of large phylogenies. Bioinformatics, 30(9), 1312–1313. 464 https://doi.org/10.1093/bioinformatics/btu033 465 Wang, J. (2017). The computer program STRUCTURE for assigning individuals to populations: 466 easy to use but easier to misuse. Molecular Ecology Resources, 17(5), 981–990. 467 https://doi.org/10.1111/1755-0998.12650 468 Williams, G. C., & Chen, J.-Y. (2012). Resurrection of the octocorallian genus Antillogorgia 469 for Caribbean species previously assigned to Pseudopterogorgia, and a taxonomic 470 assesment of the relationship of these genera with Leptogorgia (Cnidaria, Anthozoa, 471 Gorgoniidae). ZOOTAXA, 3505, 39–52. 472 Wright, S. (1978). Evolution and the Genetics of Populations. Variability within and among 473 Natural Populations (Vol. 4.). University of Chicago Press, Chicago. 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493

58

494 -Supplementary- 495 COEVOLUTION IN AN ECOLOGICAL SPECIATION SCENARIO: 496 FINE-SCALE POPULATION GENOMICS IN THE GORGONIAN CORAL COMPLEX 497 Antillogorgia bipinnata-kallos AND ITS SYMBIONT 498 -Preliminary results- 499 500

DeltaK = mean (|L’’(K)|) / sd(L(K))

Delta k Delta 501 502 Figure S1. Structure diagram. (a) using a k value of 3. (b) using a k value of 4. 503 504

505 506 Figure S2. Genotype accumulative curve for Symbiodinium analysis. The graph indicates 507 that with 130 loci randomly sampled 1000 times, it is reached the plateau to discriminate 508 between unique individuals. 509 510 511

59

512 513 514 Figure S3. Minimum spanning network for Symbiodinium analysis. Each cluster of 515 multilocus genotypes correspond to one node. The nodes are connected by the minimum 516 genetic distance between samples. 517 518 519

520 521 Figure S4. Hierarchical cluster for the Symbiodinium analysis. Dendogram was generated by 522 UPGMA method. (a) Dendrogram for 137 colonies. (b) zoom to the last nodes showing a clustering 523 for 13 individuals associated to habitats (blue=deep, green=intermedial and yellow=kallos). 524

60