bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
1 ARTICLE
2 Running title: Coexistence in annual killifish
3 Title: Trait evolution and historical biogeography shape assemblages of annual killifish
4
5 Andrew J. Helmstetter1, 2, Tom J. M. Van Dooren3, 4, 5, Alexander S. T. Papadopulos6, Javier Igea7, Armand M.
6 Leroi1, and Vincent Savolainen1
7
8 1 Imperial College London, Department of Life Sciences, Silwood Park Campus, Ascot, Berkshire SL5 7PY,
9 UK
10 2 Institut de Recherche pour le Développement (IRD), UMR-DIADE, 911 Avenue Agropolis, BP 64501, 34394
11 Montpellier, France.
12 3Sorbonne University, UMR 7618, Institute of Ecology and Environmental Sciences Paris, 4 Place Jussieu,
13 75005 Paris, France
14 4CNRS, CEREEP Ecotron IleDeFrance (UMS 3194), École Normale Supérieure, 78 rue du Château, 77140 St-
15 Pierre-lès-Nemours, France
16 5Naturalis Biodiversity Center, Darwinweg 2, Leiden 2333 CR, The Netherlands
17 6 Molecular Ecology and Fisheries Genetics Laboratory, Environment Centre Wales, School of Natural
18 Sciences, Bangor University, Bangor, LL57 2UW, UK
19 7 University of Cambridge, Department of Plant Sciences, Downing Street, Cambridge, CB2 3EA, UK
20 Keywords: Annual fish, phylogenetics, trait evolution, biogeography, body size, geography
21 of speciation, Austrolebias
22 Word count: 7671
23 Corresponding Authors: [email protected], [email protected],
25 Data archival location: Genbank and Dryad
26 Elements: title page, abstract, text, literature cited, figure legends & figures and
27 supplementary materials
28
1 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
29 ABSTRACT (199)
30 Reconstructions of evolutionary and historical biogeographic processes can improve our
31 understanding of how species assemblages developed and permit inference of ecological
32 drivers affecting coexistence. We explore this approach in Austrolebias, a genus of annual
33 fishes possessing a wide range of body sizes. Regional assemblages composed of different
34 species with similar size distributions are found in four areas of eastern South America.
35 Using phylogenetic trees, species distribution models and size data we show how trait
36 evolution and historical biogeography have affected the composition of species assemblages.
37 We extend age-range correlations to improve estimates of local historical biogeography. We
38 find that size variation principally arose in a single area and infer that ecological interactions
39 drove size divergence. This large-size lineage spread to two other areas. One of these
40 assemblages was likely shaped by adaptation to a new environment, but this was not
41 associated with additional size divergence. We found only weak evidence that environmental
42 filtering has been important in the construction of the remaining assemblage with the smallest
43 range of sizes. The repeated assemblage structures were the result of different evolutionary
44 and historical processes. Our approach sheds light on how species assemblages were built
45 when typical clustering approaches may fall short.
46
47
2 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
48 INTRODUCTION
49 Coexistence of related species is shaped by the effects of different ecological, evolutionary
50 and historical processes. These include speciation (Nosil 2012; Warren et al. 2014;
51 Mittelbach and Schemske 2015) and extinction, local ecological processes such as
52 competition (Hardin 1960; Pigot and Tobias 2012) and environmental filtering (Mouillot et
53 al. 2007; Lebrija-Trejos et al. 2010) and processes independent of phenotype, e.g. random
54 dispersal (Gotelli and McGill 2006). Their effects on coexistence differ across spatial and
55 temporal scales (Webb et al. 2002). Ecological processes acting over short time scales (e.g.,
56 resource competition) can contribute to the selective pressures that drive trait evolution and
57 ecological speciation over longer time scales (Langerhans and Riesch 2013), e.g. character
58 displacement (Brown and Wilson 1956; Losos 1990; Schluter and McPhail 1992).
59 Understanding the interplay between ecological and evolutionary forces is essential in order
60 to determine the processes affecting species coexistence.
61
62 The importance of evolutionary and historical biogeographic processes in community
63 assembly is increasingly acknowledged (Gerhold et al. 2018) but they remain relatively
64 understudied in community ecology when compared to local and recent ecological processes
65 (Warren et al. 2014; Mittelbach and Schemske 2015). In community phylogenetics (Webb et
66 al. 2002), phylogenetic reconstructions are used to characterise assemblages and to predict
67 the ecological processes at work in them. An assemblage that is phylogenetically
68 overdispersed is usually inferred to be structured by competition, while phylogenetically
69 clustered assemblages are thought be shaped by environmental filtering. However, a single
70 ecological process can have variable effects on phylogenetic relatedness and trait variation in
71 an assemblage (Cavender-Bares et al. 2009). Typical community phylogenetic approaches
72 based on phylogenetic relatedness are unable to discriminate between processes effectively as
3 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
73 they implicitly assume simple evolutionary models (Weber et al. 2017). They are also
74 susceptible to interpreting historical effects on current geographic distributions as evidence
75 for ecological and evolutionary processes (Warren et al. 2014).
76
77 Trait evolution is an essential component in the formation of species assemblages (Webb et
78 al. 2002, Cavender-Bares et al. 2004; Kraft et al. 2007). However, when models of trait
79 evolution are applied, they are often relatively simple: convergence in traits is often not
80 modelled while conserved traits are modeled using a Brownian motion (BM) model of
81 evolution (e.g. Kraft et al. 2007). Tools that permit inference of interactions between
82 ecological and evolutionary processes are still in development (Weber et al. 2017), but it is
83 possible to investigate variability of evolutionary processes across a phylogenetic tree
84 without having to resort to heuristic sampling algorithms (Kraft et al. 2007). Methods can be
85 used to detect whether species traits in an assemblage are attracted to more than a single
86 phenotypic optimum, by identifying selection regime shifts (Butler and King 2004). They can
87 also be used to infer whether traits of different species converge towards the same optimum
88 (Ingram and Mahler 2013; Oke et al. 2017; Speed and Arbuckle 2017). Taking advantage of
89 these approaches is important because trait convergence can also be the result of independent,
90 random divergence from distant starting points (Webb et al. 2002; Stayton 2008). By
91 separating the history of selection regimes from evolutionary random walks we can improve
92 our understanding of the driving forces behind trait evolution and coexistence (Oke et al.
93 2017).
94
95 Historical biogeography is also intrinsic to how species assemblages form but is often
96 neglected in empirical studies (Warren et al. 2014; Mittelbach & Schemske 2015). For
97 example speciation in sympatry or parapatry (i.e. non-allopatric) produces co-occurring sister
4 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
98 species while speciation in allopatry does not, although this patterns will change over time
99 due to post-speciation range shifts. The most common way biogeography is included in
100 phylogenetic studies is through ancestral range estimation (ARE) models, albeit at a coarse
101 level. Species are assigned to predefined areas, which are used to estimate different types of
102 ‘cladogenetic events’ that imply varying levels of geographic proximity during divergence.
103 For example, ‘founder-event speciation’ (long-distance dispersal followed by isolation and
104 speciation; Matzke et al. 2014) lies on one end of the geographic continuum while ‘within-
105 area speciation’ is as close as these models can get to the other end. Identifying speciation in
106 sympatry is particularly difficult (Papadopulos et al. 2011; Igea et al. 2015; Martin et al.
107 2015) and requires the examination of evidence beyond the scope of phylogenetic
108 comparative methods (Coyne and Orr 2000). Nevertheless, ARE can identify instances where
109 species ranges may have overlapped during divergence, which can be useful to develop
110 further hypotheses. We can attempt to refine phylogenetic predictions of geographic overlap
111 during divergence by using methods that make use of species range maps such as the age-
112 range correlation approach (ARC; Barraclough and Vogler 2000). Making use of ARC
113 alongside ARE can help us obtain a more complete picture of the biogeographic context and
114 the processes that shaped species assemblages.
115
116 To better understand how trait evolution and historical biogeography have shaped species
117 assemblages we need a study group with some natural replication, i.e., multiple assemblages
118 with comparable traits but different sets of related species. One such group is a genus of
119 freshwater fishes, Austrolebias (Costa 1998; 2006). These annual killifishes live in seasonal
120 freshwater pools and wetlands on the South American grasslands and floodplains. During the
121 reproductive season, a mating pair will dive into the muddy substrate of the pond, oviposit
122 and fertilise their eggs. As the dry season commences, ponds dry and the adults die. Eggs
5 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
123 persist within the soil, going through a single or several stages of diapause until hatching is
124 triggered by the wet season rains (Wourms 1972). Once hatched, fish grow rapidly to
125 adulthood. The annual life-cycle of Austrolebias restricts them to a specific habitat where
126 interactions are typically between congenerics. This is useful because their close relatedness
127 may allow us to link history, ecology and evolution more easily.
128
129 Austrolebias are principally found in the seasonal ponds distributed throughout basins of the
130 La Plata and Paraguay Rivers, the Negro River drainage of the Uruguay river basin and the
131 Patos-Merin lagoon drainage system (Fig. 1). These areas are referred to as their areas of
132 endemism (Costa 2010) and we consider assemblages as the species that are found in each of
133 these areas. The ranges of Austrolebias species can be small or large within each area of
134 endemism and each range can be characterised by environmental variables. These distinguish
135 species by what we could call their realised environmental niche. We refer to quantitative
136 representations of species differences in these variables simply as environmental niches. The
137 life-cycle of Austrolebias relies heavily on the seasonal environment and the extent to which
138 their environmental niche can vary and have changed in the past is unknown. It is possible
139 that this has influenced the formation of current assemblages and the trait variation within.
140 We will apply a commonly used heuristic and study changes in environmental niche using
141 the same methods as for regular traits (Münkemüller et al. 2015), while we are aware that
142 these abstract traits are not defined at the organismal level and can change without evolution.
143
144 Within each area of endemism species vary substantially in adult body size, resulting in an
145 approximate replication of trait distributions (Fig. 1). The largest species can grow to more
146 than 150mm in length while the smallest can be less than 25mm when mature (Costa 2006)
147 and differently-sized species are known to locally coexist in each area (Fig. 1). Variation in
6 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
148 body size is particularly interesting because size influences resource use in freshwater
149 systems (Woodward and Hildrew 2002) and stronger competition is expected between those
150 species with similar body sizes because they use the same resources (MacArthur and Levins
151 1967; May and MacArthur 1972). Diet has already been shown to be linked to body size in
152 Austrolebias (Laufer et al. 2009; Arim et al. 2010; Ortiz and Arim 2016) and the largest
153 species are known to prey upon the smaller species (Costa 2009) indicating that differently
154 sized species occupy distinct niches in a pond. An ecological study within a single area found
155 that local community structure was determined more by body size than species identity
156 (Canavero et al. 2014). A reconstruction of body size evolution is fundamental for
157 understanding how Austrolebias assemblages were built.
158
159 We use models of trait evolution and historical biogeography to disentangle the processes
160 that led to the replicated body size distributions across Austrolebias assemblages. When we
161 detect a selection regime shift for the optimal size of a species relative to its ancestor, models
162 of historical biogeography and a decision tree can help determine the process that has
163 generated this change (Fig. 2). We suggest three types of scenario that can be detected when
164 using this approach. Scenario (A1) occurs when biogeographic analyses at the regional and
165 local scales predict that a particular speciation event was allopatric. Concurring regime shifts
166 for size and environmental niche following divergence will indicate that size evolution is
167 linked to a substantial change in the environmental niche. Therefore, the size change can be
168 interpreted as an adaptation to a new environment. Scenario (B1) is similar to (A1) but
169 without the shift in environmental niche. Here a number of processes (e.g., dispersal
170 syndromes (Stevens et al. 2014); accelerated evolution upon colonisation (Aubret 2015)) are
171 compatible with the scenario (B1) so we leave the inferred process open. Scenarios (A2) and
172 (B2) are identical to (A1) and (B1) but account for the possibility that speciation events
7 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
173 classified as within-area at the regional scale may be revealed as allopatric at the local scale.
174 Scenario (C) involves a selection regime shift for size and a shift for environmental niche
175 while speciation is non-allopatric. Competition and adaptation to a new environment can
176 contribute to this scenario however their presence is not required such that we cannot reliably
177 infer either process. Scenario (D) occurs when speciation is non-allopatric. A regime shift for
178 body size alone may indicate that trait divergence was driven by divergent or disruptive
179 selection towards different size optima through ecological interactions (Fig. 2).
180
181 Identifying the occurrence of these scenarios can shed light on how trait variation arose but
182 may not fully explain how similar trait distributions were composed in each assemblage. We
183 can again use historical biogeographic approaches to understand how size variation generated
184 in one assemblage became distributed over others. For example, size-diverged species that
185 originated through competition (D) can emigrate from their original assemblage to occupy
186 the same ecological niche in a different assemblage and speciate in allopatry. This would pre-
187 empt the scope for scenarios (C) or (D) in the colonised assemblage (Rueffler et al. 2006).
188 We can rank assemblages according the contribution of dispersal to size variation (Fig. 2). At
189 one extreme all major size variation in an assemblage came from shifts within or upon
190 colonisation of the assemblage area. At the opposite extreme major size variation appeared
191 exclusively by dispersal.
192
193 Here we sequence nuclear (nDNA) and mitochondrial DNA (mtDNA) markers in 26 species
194 to reconstruct the evolutionary history of Austrolebias and address four main questions. In
195 conjunction with new phylogenetic trees we perform traditional community phylogenetics to
196 ask whether we find phylogenetic clustering or overdispersion in the assemblages. We collate
197 size and occurrence data to model species distributions, estimate environmental niches and
8 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
198 identify shifts in selection optima to ask whether selection regime shifts in body size and
199 shifts in environmental niche occurred. We reconstruct regional biogeographic events and
200 complement this with an extension of the age-range correlation (Barraclough and Vogler
201 2000) to test if we can make predictions of local historical biogeography. Finally, when
202 selection regime shifts are detected, we can attempt to uncover how size variation arose,
203 determine the sources of size variation per assemblage (Fig. 2) and answer our last question:
204 Have the similar size patterns across Austrolebias assemblages been generated in the same
205 way? We found that although trait distributions are similar, the major processes that shaped
206 size variation in each assemblage were not.
207
208 MATERIAL AND METHODS
209
210 DNA sequencing
211 We used the Qiagen DNeasy Blood & Tissue kit to extract genomic DNA from 60
212 individuals of 26 Austrolebias species obtained from field sampling and populations
213 maintained at CEREEP in Nemours, France. For each extraction, 15mg of fin tissue was
214 used, except for small fish where we complemented with muscle tissue. All individuals were
215 sequenced on an ABI 3130xL Genetic Analyzer for fragments of ectoderm-neural cortex
216 protein 1 gene (enc1), recombination activating gene 1 (rag1), SH3 and PX domain
217 containing 3 (snx33), rhodopsin (rh1), and three fragments of 28S ribosomal DNA (28S-
218 rRNA). Three mitochondrial genes were sequenced: 12S ribosomal DNA (12S-rRNA), 16S
219 ribosomal DNA (16S-rRNA) and cytochrome b (cytB). Primers for the amplification and
220 sequencing of three fragments of 28S-rRNA and the mitochondrial genes 12S-rRNA, 16S-
221 rRNA and cytB were taken from Van Dooren et al. (2018). For the enc1 gene primers from
9 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
222 (Li et al. 2007) were used. We designed new primers for snx33, rag1 and rh1 using Primer3
223 (Untergasser et al. 2012; Table S1).
224
225 To improve taxonomic coverage, our dataset for tree inference was supplemented with
226 sequence data from Van Dooren et al. (2018). When sequences from different sources were
227 combined, the sequences were of same source population (Table S2). Additional sequences
228 for Pterolebias longipinnis (GenBank accession EF455709, KC702007, KC702072) and
229 Hypsolebias magnificus (KC701989, KC702056, KC702122) were downloaded from
230 GenBank to be used as outgroups. Sequences were aligned in Geneious (v6.1, Biomatters)
231 using the MAFFT alignment plugin (MAFFT v7.017; Katoh et al. 2002) and low quality ends
232 were trimmed. The combined nuclear matrix is composed of 65 individuals and 3,128 aligned
233 characters (enc1: 706, rag1: 651, snx33: 543, 28S-rRNA: 724, rh1: 504). The combined
234 mitochondrial matrix comprises 65 individuals and 1,563 characters (cytB: 720, 12S-rRNA:
235 345, 16S-rRNA: 498).
236
237 Phylogenetic inference
238 Sequences representing 63 Austrolebias individuals and two outgroups were used to build
239 nDNA and mtDNA-based trees as well as a tree constructed using nDNA and assuming a
240 multispecies coalescent with *BEAST. Phylogenetic tree inference was conducted using
241 BEAST2 v2.4.3 (Bouckaert et al. 2014). We initially used a model testing approach to choose
242 the appropriate substitution model for each partition but BEAST2 analyses failed to
243 converge. Therefore, we used the relatively simple substitution model HKY+G as suggested
244 in Drummond and Bouckaert (2015) and empirical base frequencies for each gene/partition.
245 For mtDNA and nDNA trees we used linked trees with separate substitution models for each
246 locus and a single uncorrelated relaxed clock model. The exception were the 28S rRNA
10 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
247 fragments, which were run with linked substitution models across fragments. In our
248 *BEAST analyses for the coalescent, substitution, clock and tree models were unlinked and
249 the 28s fragments were run as a concatenated alignment. Yule tree priors were used for all
250 analyses.
251
252 Phylogenetic trees were calibrated using divergence times of Austrolebias with two
253 outgroups and normal priors; Pterolebias longpinnis (mean = 46.79 million years ago (Ma);
254 s.d. = 3.45) and Hypsolebias magnificus (mean = 16.54 Ma; s.d.= 2.75). Both secondary
255 calibrations were taken from Helmstetter et al. (2016). We constrained clades containing (i)
256 Austrolebias and Hypsolebias and (ii) only Austrolebias to be monophyletic to match the
257 topology in (Helmstetter et al. 2016). BEAST2 and *BEAST were run for 5.0 x 107
258 generations and trees were sampled every 5,000 generations. We repeated runs three times
259 for each dataset. Tracer v1.6 (Drummond and Rambaut 2007) was used to identify at which
260 point stationarity had been reached and LogCombiner (Bouckaert et al. 2014) to combine
261 trees from separate runs. The relevant percentage of trees was discarded as burn-in to ensure
262 that effective sample size was greater than 200 for all parameters. Finally, TreeAnnotator
263 (Bouckaert et al. 2014) was used to generate the maximum clade credibility (MCC) tree. For
264 downstream analyses we converted mtDNA and nDNA trees to species trees with the GLASS
265 algorithm (Liu et al. 2009) using the speciesTree function in R library ‘ape’ (Paradis et al.
266 2004). This created a pair of species trees we could use along the *BEAST species tree for
267 further analyses.
268
269 Phylogenetic clustering
270 We tested for the presence of phylogenetic clustering or overdispersion in the assemblages of
271 each of the four areas of endemism (detailed below). We used mean pairwise phylogenetic
11 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
272 distance (MPD, Webb et al. 2002) and simulated null distributions for this statistic per
273 assemblage (R library picante; Kembel et al. 2010).
274
275 Location data, species ranges & species distribution modelling
276 We aggregated coordinates of ponds where Austrolebias have been observed using primary
277 publications, the Global Biodiversity Information Facility (https://www.gbif.org/),
278 information on amateur egg trading websites and data shared by hobbyist collectors or from
279 our own collection trips (Table S3, updated until March 2017). Using Google Earth, we
280 obtained coordinates for all locations, identified duplicates and verified pond locations when
281 images of their locations were available. At each location in the dataset, we examined species
282 co-occurrence and added matrices in Figure 1 showing which species pairs co-occur in our
283 data complemented with reports in Loureiro et al. (2015).
284
285 We characterised species ranges in two ways. First, we used the four areas of endemism that
286 distinguish the assemblages in our analysis (Costa 2010). These are: the western region of the
287 Paraguay River basin (Western Paraguay or W), the lower La Plata River basin and the
288 middle-lower Uruguay River basin (La Plata or L), the Negro River drainage of the Uruguay
289 River basin and the upper/middle parts of the Jacuí, Santa Maria, Jahuarão and Quaraí river
290 drainages (Negro River or N) and the Patos-Merin lagoon system including the southern
291 coastal plains (Patos Lagoon or P) (Fig. 1). A fifth area of endemism is found the middle
292 section of the Iguaçu River basin. It is home to a single species, A. carvalhoi, for which we
293 do not have data, so it was not included.
294
295 Second, we built species distribution models (SDMs) using the above data and MaxEnt
296 (version 3.3.3k; Phillips et al. 2006; Phillips and Dudík 2008) to more accurately determine
12 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
297 species ranges. We used the current global climate data layers of 19 bioclimatic variables and
298 altitude taken from WorldClim (www.worldclim.org) at a spatial resolution of 30 arcseconds
299 (approximately 1 km2), 10 variables characterising soil composition and two variables with
300 river basin characteristics (Table S4). As most locations where Austrolebias have been
301 sampled are at roadsides, we sampled background points from a raster file of roads at the
302 same resolution (CIESIN 2013). We ran a PCA on the unstandardised environmental
303 variables and used the principal component scores (PC) as covariates in MaxEnt. This
304 produced ranges with fewer spurious predicted occurrences at a large distance from the
305 known capture sites than PC from standardised environmental variables. Range sizes were
306 calculated by applying a threshold to the logistic output of MaxEnt, at which the sum of the
307 sensitivity (true positive rate) and specificity (true negative rate) is highest. When a cell
308 possessed a habitat suitability score above this threshold, the species was counted as present
309 in that cell.
310
311 Environmental niches
312 We characterised the environmental niche of each species with statistics of the environmental
313 variables listed above. We followed an approach similar to the outlying mean index (OMI)
314 analysis proposed by Dolédec et al. (2000). First, environmental variables were all
315 standardised to zero mean and standard deviation equal to one. Then their values at the
316 known locations per species were averaged. We weighed locations by presence/absence as
317 local abundances are unknown and the different ecological roles in the genus might affect
318 abundances independent of environmental preferences (e.g. large piscivore species are less
319 numerous in ponds). On the resulting values a PCA was carried out which did not account for
320 phylogenetic relatedness among species as we did not want to impose an evolutionary model
321 a priori.
13 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
322
323 Body size data
324 We represented species size by a measure for asymptotic length. Growth is indeterminate in
325 fish and age information is often not available with size measurements. Body size data were
326 obtained by taking the largest known field measurements of adult male standard length (SL)
327 for each species from the literature and our own field records (full dataset in Table S5).
328
329 Prediction of ancestral ranges
330 Ancestral range estimation (ARE) was conducted using the R package ‘BioGeoBEARS’
331 (Matzke 2014). The biogeographic events this approach assesses are within-area speciation,
332 within-area subset speciation, range expansion, range contraction, vicariance and founder
333 event speciation (summarised in Dupin et al. 2016). Some events occur within an area
334 defining an assemblage, others imply a dispersal or vicariance event. ‘BioGeoBEARS’ fits
335 models similar to the Dispersal-Extinction-Cladogenesis (DEC) model (Ree and Smith 2008),
336 the BayArea (Bayesian Inference of Historical Biogeography for Discrete Areas) model
337 (Landis et al. 2013) and the Dispersal-Vicarance (DIVA) model (Ronquist 1997), with and
338 without founder-event speciation (indicated by +j). We ran separate sets of analyses where
339 the maximum number of areas composing ancestral ranges was restricted to two (maximum
340 observed) or four (maximum possible). We performed model selection based on Akaike
341 Information Criterion corrected for sample size (AICc), to identify which models best fit the
342 data. We ran analyses where ranges were restricted to include only adjacent areas. For
343 example P could only share a range with N, while N could be part of a range with both L and
344 P (Fig. 1). Finally, Costa (2010) suggested that marine transgressions during the middle to
345 late Miocene may have connected W, N and P, allowing for dispersal among these areas. We
346 therefore ran a separate set of models that classified these areas as adjacent between 15 and
14 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
347 11 Ma to determine whether including this transgression led to a better fit. We implemented
348 Biogeographical Stochastic Mapping (BSM; Dupin et al. 2016) to predict the frequencies of
349 different biogeographic events, using the best models for each tree (500 simulations).
350
351 Selection regime shifts for size and environmental niche shifts
352 We used SURFACE (Ingram and Mahler 2013) to identify convergent evolution in body size
353 and convergent environmental niche shifts. This program fits Ornstein–Uhlenbeck (OU)
354 models with one or several optima and applies adapted Information Criterion comparisons in
355 a stepwise manner to identify regime shifts on a phylogenetic tree and determines whether
356 these tend towards the same optima. We limited our analysis of the species environmental
357 niche to the first two PC of environmental variables. We also ran l1OU (Khabbazian et al.
358 2016) to see whether an alternative method of identifying evolutionary shifts corroborated
359 our SURFACE results.
360
361 Ancestral range correlations
362 Inferring historical biogeography using areas of endemism and ARE can oversimplify spatial
363 patterning of the actual species ranges and their overlaps. We used the age-range correlation
364 (ARC) approach of Barraclough and Vogler (2000) which can handle detailed range data.
365 Simulations have shown that it has some power to discriminate geographic modes of
366 speciation (Barraclough and Vogler 2000). ARCs are regression models of range overlap on
367 phylogenetic node age. Descendant ranges are pooled to calculate range overlap per node and
368 overlap is expected to increase (decrease) with node age for nodes that originated by
369 allopatric (sympatric) speciation. A major fault in the current inference of the probabilities of
370 different modes is that these are inferred from the estimated intercept of the regression across
371 pooled data. However, there are many combinations of geographical speciation modes that
15 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
372 can produce the same intercept. We used mixture regressions in the ‘flexmix’ R package
373 (Grün and Leisch 2008) to analyse age-range overlap data. Nodes of a phylogenetic tree can
374 be assigned to different clusters of events representing different geographical speciation
375 modes and with the intercept and node age slope estimated per cluster. Mixtures fitted
376 consisted of 1-3 component regression models.
377
378 Range overlaps were calculated as fractional overlap of the smallest range size following
379 Barraclough and Vogler (2000). For one species, A. paucisquama, we were unable to infer a
380 reasonable SDM so we could not include this species in our ARC analyses. After an
381 exploration of logit and other models, we analysed the data as mixtures of gaussian random
382 variables with linear regressions. These produced fewer spurious results and an absence of
383 strong effects of parameterisation details. We constrained the residual variances of all
384 mixture components to be equal. We otherwise obtained regressions with very small error
385 variances, which were not observed in process simulations (Barraclough and Vogler 2000).
386 We inferred range overlap intercepts by bootstrapping the preferred mixture model, which
387 was selected on the basis of comparing ICL (Integrated Completed Likelihood criterion)
388 between models.
389
390 RESULTS
391
392 Phylogenetic inference
393 The topologies of nDNA (Fig. 3A) and mtDNA (Fig. 3C) trees were well supported with
394 posterior probabilities (PP) > 0.9 in 74% & 69% of nodes, respectively. Support in the
395 coalescent tree (Fig. 3B) was markedly lower (38% of nodes with PP > 0.9, 58% of nodes
396 with PP > 0.75). The topology of our mtDNA tree is similar to two other recent mtDNA trees
16 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
397 (García et al. 2014; Van Dooren et al. 2018). Estimates of the age of the genus overlapped
398 between the trees. It was calculated to be 18.37 Ma (95% HPD = 14.03, 22.94) in the nDNA
399 tree, slightly older in the mtDNA tree at 19.50 Ma (95% HPD = 14.63, 23.53) and 12.48 Ma
400 (95% HPD = 8.47, 16.61) for the coalescent tree. These estimates are similar to a study of the
401 Cynolebiini (Costa et al. 2017) which dated the crown age of Austrolebias at approximately
402 10-15 Ma. García et al. (2014) did not use secondary calibrations and estimated a more recent
403 crown age of approximately 8 Ma.
404
405 We consistently recovered three major groups within Austrolebias in all trees (Fig. 3), which
406 did not include all species. Several species were non-monophyletic in our nDNA and mtDNA
407 trees including A. apaii, A. bellottii, A. cinereus, A. juanlangi & A. viarius. Austrolebias
408 bellottii and A. apaii were treated as the same species because A. apaii is a junior synonym of
409 A. bellottii (García et al. 2012). In some cases A. cinereus individuals were more closely
410 related to A. vazferreirai than to their conspecifics. These species are also morphologically
411 similar (Costa 2006) and A. cinereus is known from only a single population. We decided to
412 merge A. vazferreirai and A. cinereus for the comparative analyses in this paper including the
413 *BEAST analysis. Our mtDNA topology differed extensively from our nrDNA tree (Fig. 3),
414 which had substantial effects on downstream inference. There are many reasons why the
415 mtDNA history may not accurately reflect the species tree (Shaw 2002; Ballard and Whitlock
416 2004; see discussion). Similarly, the poor support in the coalescent tree means that it is likely
417 unreliable for inference. Therefore, while we performed each analysis using all three trees,
418 we focus on summarising the results based on the nDNA tree.
419
420 Phylogenetic clustering
17 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
421 We found significant phylogenetic clustering for the Negro River assemblage (MPD score
422 22.5, null expectation 28.4, z = -2.50, p = 0.02). Classically, this would be taken as evidence
423 that environmental filtering was the major process involved in building that assemblage. For
424 the other assemblages, there is no significant phylogenetic clustering or overdispersion.
425
426 Ancestral range estimation
427 The best-fitting ARE models were DIVA+j for all trees. The parameters and likelihoods of
428 each model are summarised in Table S6. These models revealed that jump-dispersal of
429 lineages between areas was common (Fig. 4, Fig. S1). Accordingly, biogeographic stochastic
430 mapping found frequencies in the range of 47-60% of events (Fig. S2). Likelihood ratio tests
431 revealed that models with the jump-dispersal parameter conferred significantly higher
432 likelihood to the data than those without (Table S7) indicating that founder-event speciation
433 was important in shaping the distribution of Austrolebias species. One striking observation
434 was that the most northern area of endemism, Western Paraguay (W), was colonised by
435 Austrolebias in three (coalescent & mtDNA) (Fig. S1) or four (nDNA) independent instances
436 (Fig. 4).
437
438 Speciation was also estimated within areas of endemism. Expected numbers of within-area
439 speciation events were particularly high in Patos Lagoon (an average of 4.4 events per
440 simulation) and Negro River (2.7) while the other areas had averages below one. Vicariance
441 events were rare, probably due to the rarity of ancestral ranges with more than one area.
442 Within-area subset speciation is not included in the DIVA model. Neither adjacency matrices
443 nor changing the maximum number of areas significantly improved the likelihood (< 2
444 difference in AICc) when compared to the simplest set of models (Table S6).
445
18 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
446 Size evolution
447 Our ancestral state reconstructions (Fig. 5A) show substantial divergence in size. The four
448 largest Austrolebias species form a single clade. Models consistently found a shift in
449 selection optimum towards this clade (Group II in Fig. 3) in all trees, just after the divergence
450 of the smaller A. luteoflammulatus. (Fig. 5A, Fig. S3, AICc decrease relative to single
451 optimum model mtDNA:13.3, nDNA: 13.6 coalescent: 11.1). An additional shift towards an
452 intermediate optimum was inferred in the mtDNA tree (AICc decrease 0.9), in a clade
453 containing A. bellottii, A. melanoorus, A. robustus, A. vandenbergi and A. vazferreirai (Fig.
454 S3C). The remaining species in each tree tended towards another, smaller optimum and were
455 not subjected to a shifted selection optimum for size throughout their history. Model
456 comparisons using l1OU supported a shift in the clade with the largest species. In agreement
457 with the pattern in AICc values observed, it found no evidence of any further shift (Fig. S4).
458 Therefore selection regime shifts with convergence of optimal sizes have not occurred.
459
460 Environmental niches
461 The first two PCs of the environmental niche variables explained 60% of species variation.
462 We found that species cluster to some degree by area of endemism (Fig. 5B, C), and along
463 two axes. Patos Lagoon species and Western Paraguay occupy two extremes for PC2, and all
464 other species and ancestors group along an axis mostly determined by PC1. We assessed the
465 scores of the two PCs. PC1 consists largely of soil information and temperature seasonality,
466 plus information on the driest and coldest season. Western Paraguay and La Plata are similar
467 for this PC and have less acidic soil with more silt, a larger seasonal range in temperatures
468 and less annual precipitation. PC2 contains information relating to average temperatures,
469 precipitation seasonality and precipitation/temperature during the wettest month and the
470 warmest season. This axis primarily distinguishes Western Paraguay from the other areas,
19 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
471 with warmer temperatures, increased precipitation seasonality and more rain in the warmest
472 season.
473
474 We assessed whether there were shifts and convergent changes among the environmental
475 niche changes. SURFACE analyses using PC1 & PC2 revealed two regime shifts (Fig 5, Fig.
476 S3; AICc differences nDNA: 12.5 mtDNA: 11.8 coalescent: 17.0). All four species from
477 Western Paraguay were subject to shifts associated with their dispersal into the area (Panels
478 B & C in Fig. 5). According to the models based on the nDNA tree, convergent shifts
479 towards a new environmental niche occurred independently when A. vandenbergi and A.
480 monstrosus colonised Western Paraguay. A group containing A. patriciae and A. toba tended
481 towards a second different environmental niche at a less decreased value of PC2 (Fig. 5B, C).
482 Four out of seven species in the Negro assemblage converge to that same niche regime. l1OU
483 analyses inferred just two shifts, for A. monstrosus and A. vandenbergi (Fig. S5). None of the
484 regime shifts for environmental niche are associated with selection regime shifts for size,
485 therefore scenarios (A1) and (A2) where size evolves due to adaptation to a new environment
486 (Fig. 2) did not occur.
487
488 Age-range correlations
489 Ranges varied considerably in size from the very large ones of A. bellottii and A. nigripinnis
490 to the extremely small of A. toba and A. affinis (Fig. S6). In general, the area under the curve
491 (AUC) values for the training data were above 0.95. Among mixture models fitted to range
492 overlaps, models were preferred with node age effects (Table S8 contains ICL values of fitted
493 models). For the nDNA tree (Fig. 6), a mixture with three components representing different
494 modes of speciation was preferred. The intercepts of these three components are -0.02 (s.e.
495 0.03), 0.24 (s.e. 0.04) and 0.59 (s.e. 0.09). The first value corresponds to allopatric speciation,
20 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
496 the other two to varying degrees of non-allopatric speciation. For the coalescent tree, a
497 mixture with three components had lowest ICL, which was only slightly smaller than the ICL
498 of a single-component mixture. We therefore infer for the coalescent and mtDNA tree (Fig.
499 S7) a model with a single mixture component, each time with an intercept not significantly
500 different from zero (coalescent 0.16 (s.e. 0.11), mtDNA 0.071 (s.e. 0.101)).
501
502 When we compare the regional and more local analyses (Fig. 4, 6A), 17 of 23 nodes showed
503 agreement between analyses. There are three events where the regional analysis inferred non-
504 allopatric speciation, while ARC did not. For three allopatric speciation events at the regional
505 scale, ARC inferred non-allopatry. This combination is an impossible geography, showing
506 limitations of historical biogeographic approaches. Most importantly, the selection regime
507 shift for size occurred at a node where non-allopatric speciation is inferred. We can conclude
508 that ecological interactions (C) drove this size divergence.
509
510
511 DISCUSSION
512
513 A new phylogenetic hypothesis for Austrolebias
514 The trees presented in this study are the most comprehensive inferences in Austrolebias to
515 date. Our nDNA tree is the first multi-locus tree based on nuclear DNA for this genus and
516 provides novel insight into species relationships. We find major differences between nDNA
517 and mtDNA trees. Incomplete lineage sorting and introgression are possible causes of
518 incongruence among different molecular markers and the maternal inheritance of mtDNA
519 may lead to a different evolutionary history (Maddison 1997; Funk and Omland 2003;
520 Ballard and Whitlock 2004). As an example we take A. luteoflammulatus, a species with
21 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
521 different yet strongly supported placements in nDNA and mtDNA trees. It is most closely
522 related to Austrolebias gymnoventris and A. wolterstorffi in the mtDNA tree, two species it
523 co-occurs with. Conversely, it is sister to the clade of the largest species in the nDNA and
524 coalescent trees. This might be evidence of mitochondrial capture - where the mtDNA of a
525 species is completely replaced by that of another and known to occur in fish (Chan and Levin
526 2005; Willis et al. 2013). Other cases of discordance can be explained by low posterior
527 probabilities in one or more phylogenetic trees. We found lower support in the coalescent tree
528 when compared to the others and difficulty in resolving non-monophyletic species such as A.
529 viarius and A. juanlangi (Fig. 3A) may have contributed to this.
530
531 We used sequence data and/or geographic information from 26 Austrolebias species, while
532 more than 40 have been described (Costa 2006; García et al. 2014). Many species not
533 included were discovered recently and have few known sites of occurrence. Some may be
534 junior synonyms like A. apaii (García et al. 2012). Furthermore, it is not possible to build
535 accurate species distribution models with very rare species, as evidenced by A. paucisquama.
536 Nevertheless, our data covered variation among Austrolebias species well: all extant species
537 in the La Plata and Western Paraguay regions were included, as well as all of the largest
538 species (Costa 2006).
539
540 Evidence for phylogenetic clustering in one assemblage
541 Assessing our assemblages using a traditional community phylogenetic approach revealed
542 that only one of the four areas deviated from null expectations. The Negro River assemblage
543 was significantly more clustered than expected, which is usually interpreted as evidence for
544 environmental filtering. Environmental niches of species in Negro River did not cluster
545 strongly (Fig. 5C) and we found no robust evidence for regime shifts, perhaps because their
22 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
546 environmental niches were similar to species in Patos Lagoon and La Plata in most cases.
547 Alternatively, we can explain the clustering by a relatively recent diversification of several
548 species within the assemblage and limited scope for jump-dispersal within that time frame
549 (Warren et al. 2014). Lack of significant results in the other three areas left the processes
550 shaping them undefined. This justifies taking a closer look at evolutionary and historical
551 processes to try to understand how the observed size patterns came to be.
552
553 A single selection regime shift for body size
554 Body size was a priori expected to be an important factor shaping species assemblages due to
555 the similar variation within each area of endemism (Fig. 1) and the strong size effects found
556 in Patos Lagoon communities (Canavero et al. 2014). Our selection regime analyses
557 consistently recovered a single shift towards the largest species. We found that this shift
558 occurred in Patos Lagoon just after A. luteoflammulatus (Fig. 4, 5A) diverged. Austrolebias
559 luteoflammulatus is a typical small species, A. cheradophilus is a large omnivore and A.
560 prognathus is a very large piscivore. We speculate that a change in diet (see Laufer et al.
561 2009; Arim et al. 2010; Ortiz and Arim 2016) may have been associated with the regime shift
562 for size and that predation by a piscivore facilitates coexistence of competitors. In the La
563 Plata & Patos Lagoon areas small, similar-sized Austrolebias species co-occur (Fig. 1), which
564 is unexpected if competition among Austrolebias would be the main factor structuring
565 communities there, so other important processes are likely at play as well. Furthermore, in
566 Western Paraguay Austrolebias are found in ponds together with other annual killifishes such
567 as Trigonectes aplocheiloides, Papiliolebias bitteri and Neofundulus paraguayensis (Alonso
568 et al. 2016) and with Cynopoecilus melanotaenia (Canavero et al. 2014) in Patos Lagoon.
569 These other species may have played a role in how species assemblages are constructed in
23 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
570 some areas e.g. by preventing the establishment of additional small Austrolebias species in
571 Western Paraguay.
572
573
574 Convergent shifts for environmental niche
575 The seasonal life cycle of Austrolebias is strongly climate and habitat-dependent. We did not
576 expect to find substantial variation nor shifts in environmental niche among species. In the
577 African annual killifish genus Nothobranchius, only moderate environmental niche
578 differences were found among five species (Reichard et al. 2017). In Nothobranchius,
579 altitude was found to be the most important factor explaining the distribution of different
580 clades and species whereas in Austrolebias environmental niche differences between species
581 were predominantly due to differences in soil characteristics, temperature and precipitation.
582
583 Contrary to our expectations we found that environmental niches of species from Western
584 Paraguay were markedly differentiated from the others. SURFACE analyses revealed that the
585 colonisation of Western Paraguay coincided with substantial and convergent shifts towards
586 new environmental niches. These seem instrumental in how the Western Paraguay
587 assemblage was constructed. It should be investigated whether evolutionary changes are
588 associated with them. Western Paraguay is in the far north and west of the range of
589 Austrolebias species and another distinguishing characteristic seems to be that ponds fill in
590 winter (Schalk et al. 2016), not in summer as is typical for the other assemblages (Errea and
591 Danulat 2001). We found two shifts of different magnitudes for species inhabiting Western
592 Paraguay, but the less extreme found by SURFACE (Fig. 5) was not recovered in l1OU
593 analyses (Fig. S5). While the less extreme shift seems plausible, SURFACE is known to
24 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
594 overestimate the number of shifts (Ho and Ané 2014). Observed shifts for environmental
595 niche were not associated with selection regime shifts for size.
596
597 Improved prediction of local historical biogeography
598 To improve the prediction of historical biogeography we have extended a method developed
599 to determine the primary geographical mode of speciation in a clade (Barraclough and Vogler
600 2000). We used mixture regression models to cluster data into groups of nodes fitted by a
601 shared regression. This allowed us to estimate the intercepts of the regression lines fitted to
602 each cluster, which approximate the range overlap at cladogenesis per group. We used the
603 assignment of nodes to mixture components to assign speciation modes to each divergence
604 event, without actually reconstructing ancestral ranges and without pre-defined discrete
605 biogeographic areas.
606
607 The ARC approach has been criticised in the past (Fitzpatrick and Turelli 2006) and these
608 criticisms still apply to our extended methodology. The power of the original method is low
609 and the discriminatory power between alternative modes of speciation disappears at older
610 nodes in a phylogenetic tree (Barraclough and Vogler 2000). A heuristic node weighting
611 method has been proposed (Fitzpatrick and Turelli 2006) but in trials we did, these issues
612 seemed aggravated by the suggested modifications. Despite these caveats we have shown that
613 ARC can be useful for examining historical biogeography at a finer scale than ARE
614 approaches. It is worth noting that we had to evaluate properties of several model
615 specifications to find inference results that seemed adequate. This calls for extensive
616 simulation studies to produce guidelines for application of the approach in different
617 taxonomic groups.
618
25 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
619 While patterns in Austrolebias revealed an appreciable fraction of non-allopatric speciation
620 events, this did not reflect the pattern in Nothobranchius where speciation is thought to be
621 exclusively allopatric and triggered by geological events (Dorn et al. 2014; Bartáková et al.
622 2015). Our ARC results corroborated our ARE analyses (Fig. 6) for most of the speciation
623 events in the tree. Results suggest that by using ARC we could improve upon estimates of
624 geographic proximity of incipient species during divergence. For example, the divergence of
625 A. duraznensis and A. affinis was estimated as within-area speciation using ARE (Fig. 4) but
626 in allopatry using ARC (Fig. 6). Additional simulations would be useful to understand factors
627 affecting error rates of the approach, as we found three cases where ARE and ARC predicted
628 incompatible biogeographic events.
629
630 Repeated patterns are the result of different processes
631 In Figure S8 we provide a visual summary of our reconstructions per assemblage.
632 We observed a single selection regime shift in size, indicating that there was a single source
633 of major size variation in the genus. Historical biogeographic approaches estimated that the
634 speciation event that preceded this shift was within-area speciation at the regional level and
635 non-allopatric speciation at the local level, taking place in the Patos Lagoon area. We
636 therefore predict that scenario (C) took place - size divergence was generated through
637 ecological interactions while incipient species were in contact. It is difficult to determine
638 exactly what kind of interaction was behind size divergence; competition is a first possibility
639 and Van Dooren et al. (2018) suggested that cannibalism may have led to the emergence of
640 large predator species. The species involved in this shift, (A. cheradophilus, A.
641 luteoflammulatus & A. prognathus) are currently known to coexist within the same ponds
642 (Fig. 1), which lends support to the idea that they coexisted during divergence. Estimating
643 historical patterns of gene flow among these species could investigate this further.
26 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
644
645 Our best fitting ARE model (Fig. 4) reveals two major events that were key in distributing
646 size variation across assemblages. First, a jump-dispersal event from Patos Lagoon to La
647 Plata led to A. elongatus, followed by an event from La Plata to Western Paraguay which led
648 to A. monstrosus. Therefore Patos Lagoon represents one extreme of size redistribution by
649 dispersal (Fig. 2) because major size variation originated within, and Western Paraguay and
650 La Plata another since major size variation originated in other areas. Conspicuously absent is
651 the Negro River assemblage, which harbours no size variation generated by a selection
652 regime shift. Austrolebias vazferreirai is the largest species in this assemblage and is thought
653 to be a generalist (Costa 2006) rather than a piscivore. Negro River is the youngest
654 assemblage (colonised a maximum of ~7.3 Ma ago compared to ~11.8 Ma in the next
655 youngest), so there may not have been enough time for a large predator to colonise or arise.
656
657 Perspective
658 Understanding the processes that lead to current species assemblages is central in the study of
659 ecology and evolution (Hutchinson 1959), and our study adds to increasing knowledge of the
660 importance of species interactions (Cornell and Lawton 1992) and size differences
661 (Simberloff and Boecklen 1981). We find similarities between our results and those from
662 other study groups. Leitao et al. (2016) showed how rare species had a disproportionate role
663 in the functional structure of assemblages. This may also be the case in Austrolebias, where
664 large species are typically rarer than smaller species (Lanés and Maltchik 2010; Lanés et al.
665 2014) and occupy different functional niches, therefore widening the functional richness of
666 assemblages. Dispersal has not led to monopolisation (whereby the first colonist gains an
667 advantage over the later immigrants by adapting to local conditions; De Meester et al 2016)
668 because different species probably have different ecological roles due to size and functional
27 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
669 niche divergence. Analyses of the co-occurrence of related species of hummingbirds
670 highlighted the importance of multiple mechanisms including competition, dispersal and the
671 environment, in shaping assemblages (Weinstein et al. 2017). Although the methodological
672 approaches were different, we also found that these mechanisms, as well as historical
673 biogeography, were important in generating assemblages of Austrolebias.
674
675 If we had taken only a traditional community phylogenetic approach we would have
676 dismissed Patos Lagoon as uninteresting because of no significant phylogenetic clustering.
677 Instead our approach has revealed evidence for ecological interactions that were key to the
678 way in which Austrolebias assemblages develop. Using models of trait evolution beyond
679 Brownian motion and assessing historical biogeography at multiple levels allowed us to
680 identify that, despite the apparent similarity of Austrolebias assemblages, each was generated
681 under a different process. Having characterised each area of endemism, we can use our
682 results to make predictions on how ponds in different areas are structured.
683
684 Body size was found to be important in structuring pond assemblages in Patos Lagoon
685 (Canavero et al. 2014), motivating our study. In Western Paraguay and La Plata, large
686 piscivores are also present and we would expect results similar to what was found in Patos.
687 However, the two niche regime shifts in Western Paraguay (Fig. 5) make us predict that
688 coexistence there is structured more by environment than in Patos. Species with more
689 extreme environmental niches such as A. monstrosus and A. vandenbergi will occupy
690 different ponds. We predict that body size will be less important for structuring pond
691 assemblages in the Negro River area. Inherently, there is then more scope for weak
692 environmental filtering effects or neutral coexistence. Confirmation of our predictions would
693 require careful assessment of co-occurrence across these areas.
28 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
694
695 CONCLUSION
696 Despite their apparent similarities, Austrolebias assemblage compositions and histories vary
697 greatly. Our analysis demonstrates that reconstructed selection regimes and historical
698 biogeography can inform on ecological processes operating in assemblages. A single
699 selection regime shift towards large body size was critical for generating the similar trait
700 distributions observed in Austrolebias assemblages. It occurred during a series of non-
701 allopatric speciation events in a single assemblage. Historical biogeographic reconstructions
702 revealed that dispersal and subsequent allopatric speciation redistributed major size variation
703 to other areas. Our approach can be used in other studies examining the processes composing
704 trait-structured assemblages. We support calls for a wider application of a historical and
705 evolutionary view on species assemblages.
706
707 REFERENCES
708 Alonso, F., P. A. Calviño, G. E. Terán, and I. García. 2016. Geographical distribution of
709 Austrolebias monstrosus (Huber, 1995), A. elongatus (Steindachner, 1881) and A.
710 vandenbergi (Huber, 1995) (Teleostei: Cyprinodontiformes), with comments on the
711 biogeography and ecology of Rivulidae in Pampasic and Chaco floodplains. Check List
712 12:1945. http://dx.doi.org/10.15560/12.4.1945
713 Arim, M., S. R. Abades, G. Laufer, M. Loureiro, and P. A. Marquet. 2010. Food web
714 structure and body size: trophic position and resource acquisition. Oikos 119:147–153.
715 Aubret, F. 2015. Island colonisation and the evolutionary rates of body size in insular neonate
716 snakes. Heredity 115:349-356.
717 Barraclough, T. G., and A. P. Vogler. 2000. Detecting the geographical pattern of speciation
718 from species-level phylogenies. The American Naturalist 155:419–434.
29 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
719 Ballard, J. W., and M. C. Whitlock. 2004. The incomplete natural history of mitochondria.
720 Molecular Ecology 13:729-744.
721 Bartáková, V., M. Reichard, R. Blažek, M. Polačik, and J. Bryja. 2015. Terrestrial fishes:
722 rivers are barriers to gene flow in annual fishes from the African savanna. Journal of
723 Biogeography 42:1832-1844.
724 Bouckaert, R., J. Heled, D. Kühnert, T. Vaughan, C.-H. Wu, D. Xie, M. A. Suchard, A.
725 Rambaut, and A. J. Drummond. 2014. BEAST 2: a software platform for Bayesian
726 evolutionary analysis. PLoS Computational Biology 10:e1003537.
727 Brown, W. L., and E. O. Wilson. 1956. Character displacement. Systematic Biology 5:49-64.
728 Canavero, A., D. Hernández, M. Zarucki, and M. Arim. 2014. Patterns of co-occurrences in a
729 killifish metacommunity are more related with body size than with species identity. Austral
730 Ecology 39:455–461.
731 Cavender-Bares, J., D. D. Ackerly, D. A. Baum, and F. A. Bazzaz. 2004. Phylogenetic
732 overdispersion in Floridian oak communities. American Naturalist 163:823–843.
733 Cavender-Bares, J., K. H. Kozak, P. V. A. Fine, and S. W. Kembel. 2009. The merging of
734 community ecology and phylogenetic biology. Ecology Letters 12:693–715.
735 Chan, K. M. A., and S. A. Levin. 2005. Leaky prezygotic isolation and porous genomes:
736 rapid introgression of maternally inherited DNA. Evolution 59:720–729.
737 Center for International Earth Science Information Network - CIESIN - Columbia University,
738 and Information Technology Outreach Services - ITOS - University of Georgia. 2013. Global
739 Roads Open Access Data Set, Version 1 (gROADSv1). Palisades, NY: NASA
740 Socioeconomic Data and Applications Center (SEDAC).
741 http://dx.doi.org/10.7927/H4VD6WCT
30 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
742 Cornell, H. V., and J. H. Lawton. 1992. Species interactions, local and regional processes,
743 and limits to the richness of ecological communities: a theoretical perspective. The Journal of
744 Animal Ecology 61:1-12.
745 Costa, W. J. E. M. 1998. Phylogeny and classification of Rivulidae revisited: origin and
746 evolution of annualism and miniaturization in rivulid fishes (Cyprinodontiformes:
747 Aplocheiloidei). Journal of Comparative Biology 3:33–92.
748 Costa, W. J. E. M. 2006. The South American annual killifish genus Austrolebias (Teleostei:
749 Cyprinodontiformes: Rivulidae): phylogenetic relationships, descriptive morphology and
750 taxonomic revision. Zootaxa 1213:1-162.
751 Costa, W. J. E. M. 2009. Trophic radiation in the South American annual killifish genus
752 Austrolebias (Cyprinodontiformes: Rivulidae). Ichthyological Exploration of Freshwaters
753 20:179–191.
754 Costa, W. J. E. M. 2010. Historical biogeography of cynolebiasine annual killifishes inferred
755 from dispersal-vicariance analysis. Journal of Biogeography 37:1995–2004.
756 Costa, W. J. E. M., P. F. Amorim, and J. L. O. Mattos. 2017. Molecular phylogeny and
757 timing of diversification in South American Cynolebiini seasonal killifishes. Molecular
758 Phylogenetics and Evolution 116:61-68. https://doi.org/10.1016/j.ympev.2017.07.020
759 Coyne, J. A., and H. A. Orr. 2004. Speciation. Sinaeur Associates Massachusetts.
760 Davies, T. J., S. Meiri, T. G. Barraclough, and J. L. Gittleman. 2007. Species co-existence
761 and character divergence across carnivores. Ecology Letters 10:146–152.
762 De Meester, L., Vanoverbeke, J., Kilsdonk, L. J., and Urban, M. C. 2016. Evolving
763 perspectives on monopolization and priority effects. Trends in Ecology & Evolution 31:136-
764 146.
765 Doledec, S., D. Chessel, and C. Gimaret-Carpentier. 2000. Niche separation in community
766 analysis: a new method. Ecology 81:2914-2927.
31 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
767 Dorn, A., Z. Musilová, M. Platzer, K. Reichwald and A. Cellerino. 2014. The strange case of
768 East African annual fishes: aridification correlates with diversification for a savannah aquatic
769 group? BMC Evolutionary Biology 14:210
770 Drummond, A. J., and A. Rambaut. 2007. BEAST: Bayesian evolutionary analysis by
771 sampling trees. BMC Evolutionary Biology 7:214.
772 Drummond, A. J., and R. R. Bouckaert. 2015. Bayesian Evolutionary Analysis with BEAST.
773 Cambridge University Press, Cambridge.
774 Dupin, J., N. J. Matzke, T. Särkinen, S. Knapp, R. G. Olmstead, L. Bohs, and S. D. Smith.
775 2016. Bayesian estimation of the global biogeographical history of the Solanaceae. Journal of
776 Biogeography 44:887–899.
777 Elliot, M. G., and A. Ø. Mooers. 2014. Inferring ancestral states without assuming neutrality
778 or gradualism using a stable model of continuous character evolution. BMC Evolutionary
779 Biology 14:226.
780 Errea, A., and E. Danulat. 2001. Growth of the annual fish, Cynolebias viarius
781 (Cyprinodontiformes), in the natural habitat compared to laboratory conditions.
782 Environmental Biology of Fishes 61:261-268.
783 Fitzpatrick, B. M., and M. Turelli. 2006. The geography of mammalian speciation: mixed
784 signals from phylogenies and range maps. Evolution 60:601–616.
785 Funk, D. J., and K. E. Omland. 2003. Species-level paraphyly and polyphyly: frequency,
786 causes, and consequences, with insights from animal mitochondrial DNA. Annual Review of
787 Ecology, Evolution, and Systematics 34:397-423.
788 García, G., V. Gutiérrez, J. Vergara, P. Calviño, A. Duarte, and M. Loureiro. 2012. Patterns
789 of population differentiation in annual killifishes from the Paraná-Uruguay-La Plata Basin:
790 the role of vicariance and dispersal. Journal of Biogeography 39:1707–1719.
32 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
791 García, G., V. Gutiérrez, N. Ríos, B. Turner, F. Santiñaque, B. López-Carro, and G. Folle.
792 2014. Burst speciation processes and genomic expansion in the neotropical annual killifish
793 genus Austrolebias (Cyprinodontiformes, Rivulidae). Genetica 142:87-98.
794 Gerhold, P., M.B. Carlucci, Ş. Procheş, and A. Prinzing. 2018. The deep past controls the
795 phylogenetic structure of present, local communities. Annual Review of Ecology, Evolution,
796 and Systematics 49:1.
797 Gotelli, N. J., and B. J. McGill. 2006. Null versus neutral models: what’s the difference?
798 Ecography 29:793-800.
799 Grün, B., and F. Leisch. 2008. FlexMix Version 2: Finite mixtures with concomitant
800 variables and varying and constant parameters. Journal of Statistical Software 28:1–35.
801 Hardin, G. 1960. The competitive exclusion principle. Science 131:1292–1297.
802 Helmstetter, A. J., A. S. T. Papadopulos, J. Igea, T. J. M. Van Dooren, A. M. Leroi, and V.
803 Savolainen. 2016. Viviparity stimulates diversification in an order of fish. Nature
804 Communications 7:11271.
805 Ho, L. S. T., and C. Ané. 2014. Intrinsic inference difficulties for trait evolution with
806 Ornstein-Uhlenbeck models. Methods in Ecology and Evolution 5:1133–1146.
807 Hutchinson, G. E. 1959. Homage to Santa Rosalia or why are there so many kinds of
808 animals? The American Naturalist 93:145–159.
809 Igea, J., D. Bogarín, A. S. T. Papadopulos, and V. Savolainen. 2015. A comparative analysis
810 of island floras challenges taxonomy‐based biogeographical models of speciation. Evolution
811 69: 482-491.
812 Ingram, T., and D. L. Mahler. 2013. SURFACE: detecting convergent evolution from
813 comparative data by fitting Ornstein-Uhlenbeck models with stepwise Akaike Information
814 Criterion. Methods in Ecology and Evolution 4:416–425.
33 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
815 Katoh, K., K. Misawa, K. I. Kuma, and T. Miyata. 2002. MAFFT: a novel method for rapid
816 multiple sequence alignment based on fast Fourier transform. Nucleic Acids Research
817 30:3059–3066.
818 Kembel, S. W., P. D. Cowan, M. R. Helmus, W. K. Cornwell, H. Morlon, D. D. Ackerly, S.
819 P. Blomberg, and C.O. Webb. 2010. Picante: R tools for integrating phylogenies and ecology.
820 Bioinformatics 26:1463-1464.
821 Khabbazian, M., R. Kriebel, K. Rohe, and C. Ané. 2016. Fast and accurate detection of
822 evolutionary shifts in Ornstein-Uhlenbeck models. Methods in Ecology and Evolution 7:811–
823 824.
824 Kraft, N. J. B., W. K. Cornwell, C. O. Webb, and D. D. Ackerly. Trait evolution, community
825 assembly, and the phylogenetic structure of ecological communities. The American
826 Naturalist. 170: 271-283.
827 Landis, M. J., N. J. Matzke, B. R. Moore, and J. P. Huelsenbeck. 2013. Bayesian analysis of
828 biogeography when the number of areas is large. Systematic Biology 62:789–804.
829 Lanés, L., and L. Maltchik. 2010. Discovery of the critically endangered annual killifish,
830 Austrolebias wolterstorffi (Ahl, 1924) (Rivulidae: Cyprinodontiformes) in Lagoa do Peixe
831 National Park, Rio Grande do Sul, southern Brazil. Journal of Threatened Taxa 2:1282–1285.
832 Lanés, L., Â. Gonçalves, and M. Volcan. 2014. Discovery of endangered annual killifish
833 Austrolebias cheradophilus (Aplocheiloidei: Rivulidae) in Brazil, with comments on habitat,
834 population structure and conservation status. Neotropical Ichthyology 12:117–124.
835 Langerhans, R. B., and R. Riesch. 2013. Speciation by selection: A framework for
836 understanding ecology’s role in speciation. Current Zoology 59:31–52.
837 Laufer, G., M. Arim, M. Loureiro, J. M. Piñeiro-Guerra, S. Clavijo-Baquet, and C. Fagúndez.
838 2009. Diet of four annual killifishes: an intra and interspecific comparison. Neotropical
839 Ichthyology 7:77–86.
34 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
840 Lebrija-Trejos, E., E. A. Pérez-García, and J. A. Meave. 2010. Functional traits and
841 environmental filtering drive community assembly in a species-rich tropical system. Ecology
842 91:386–398.
843 Leitão, R. P., J. Zuanon, S. Villéger, S. E. Williams, C. Baraloto, C. Fortunel, F. P.
844 Mendonça, and D. Mouillot. 2016. Rare species contribute disproportionately to the
845 functional structure of species assemblages. Proceedings of the Royal Society B: Biological
846 Sciences 283:20160084.
847 Li, C., G. Ortí, G. Zhang, and G. Lu. 2007. A practical approach to phylogenomics: the
848 phylogeny of ray-finned fish (Actinopterygii) as a case study. BMC Evolutionary Biology
849 7:44.
850 Liu, L., L. Yu, and D. K. Pearl. 2009. Maximum tree: a consistent estimator of the species
851 tree. Journal of Mathematical Biology 60:95–106.
852 Losos, J. B. 1990. A phylogenetic analysis of character displacement in Caribbean Anolis
853 lizards. Evolution 44:558-569.
854 Loureiro, M., A. Borthagaray, D. Hernández, A. Duarte, V. Pinelli, and M. Arim. 2015.
855 Austrolebias in space: scaling from ponds to biogeographical regions. Pages 111–132 in
856 Berois, N., G. Garcia and R. O. de Sá, ed. Annual Fishes: Life history strategy, diversity, and
857 evolution. CRC Press, Boca Raton FL, USA
858 MacArthur, R., and R. Levins. 1967. The limiting similarity, convergence, and divergence of
859 coexisting species. The American Naturalist 101:377–385.
860 Maddison, W. P. 1997. Gene trees in species trees. Systematic Biology 46:523-536.
861 Martin, C. H., J. S. Cutler, J. P. Friel, C. Dening Touokong, G. Coop and P. C. Wainwright.
862 2015. Complex histories of repeated gene flow in Cameroon crater lake cichlids cast doubt on
863 one of the clearest examples of sympatric speciation. Evolution 69:1406–1422.
35 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
864 Matzke, N. J. 2014. Model selection in historical biogeography reveals that founder-event
865 speciation is a crucial process in island clades. Systematic Biology 63:951–970.
866 May, R. M., and R. H. MacArthur. 1972. Niche overlap as a function of environmental
867 variability. Proceedings of the Natural Academy of Sciences of the United States of America
868 69:1109–1113.
869 Mittelbach, G. G., and D. W. Schemske. 2015. Ecological and evolutionary perspectives on
870 community assembly. Trends in Ecology & Evolution 30:241–247.
871 Mouillot, D., O. Dumay, and J. A. Tomasini. 2007. Limiting similarity, niche filtering and
872 functional diversity in coastal lagoon fish communities. Estuarine, Coastal and Shelf Science
873 71:443–456.
874 Münkemüller, T., F. C. Boucher, W. Thuiller and S. Lavergne. 2015. Phylogenetic niche
875 conservatism–common pitfalls and ways forward. Functional ecology 29: 627-639.
876 Nosil, P. 2012. Ecological speciation. Oxford University Press.
877 Oke, K. B., G. Rolshausen, C. LeBlond, and A. P. Hendry. 2017. How parallel is parallel
878 evolution? A comparative analysis in fishes. The American Naturalist 190:1–16.
879 Ortiz, E., and M. Arim. 2016. Hypotheses and trends on how body size affects trophic
880 interactions in a guild of South American killifishes. Austral Ecology 41:976–982.
881 Oviedo Alcoba, S., M. Rovira, and G. Garcia. 2016. Genomic isolated regions: Linkage
882 groups in parental and laboratory hybrids between Austrolebias adloffi species group. Pages
883 295–307 in Berois, N., G. Garcia and R. O. de Sá, ed. Annual Fishes: Life history strategy,
884 diversity, and evolution. CRC Press, Boca Raton FL, USA
885 Papadopulos, A. S. T., W. J. Baker, D. Crayn, R. K. Butlin, R. G. Kynast, I. Hutton, and V.
886 Savolainen. 2011. Speciation with gene flow on Lord Howe Island. Proceedings of the
887 National Academy of Sciences 108:13188-13193.
36 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
888 Paradis, E., J. Claude, and K. Strimmer. 2004. APE: Analyses of Phylogenetics and
889 Evolution in R language. Bioinformatics 20:289–290.
890 Phillips, S. J., R. P. Anderson, and R. E. Schapire. 2006. Maximum entropy modeling of
891 species geographic distributions. Ecological Modelling 190:231–259.
892 Phillips, S. J., and M. Dudík. 2008. Modeling of species distributions with Maxent: new
893 extensions and a comprehensive evaluation. Ecography 31:161–175.
894 Pigot, A. L., and J. A. Tobias. 2012. Species interactions constrain geographic range
895 expansion over evolutionary time. Ecology Letters 16:330–338.
896 Ree, R. H., and S. A. Smith. 2008. Maximum likelihood inference of geographic range
897 evolution by dispersal, local extinction, and cladogenesis. Systematic Biology 57:4–14.
898 Reichard, M., M. Janáč, M. Polačik, R. Blažek, and M. Vrtílek. 2017. Community assembly
899 in Nothobranchius annual fishes: Nested patterns, environmental niche and biogeographic
900 history. Ecology and Evolution 7: 2294–2306.
901 Rueffler, C., T. J. M. Van Dooren, O. Leimar, and P. A. Abrams. 2006. Disruptive selection
902 and then what? Trends in Ecology & Evolution 21:238-245.
903 Ronquist, F. 1997. Dispersal-Vicariance analysis: a new approach to the quantification of
904 historical biogeography. Systematic Biology 46:195–203.
905 Schalk, C. M., C. G. Montaña, K. O. Winemiller, and L. A. Fitzgerald. 2016. Trophic
906 plasticity, environmental gradients and food-web structure of tropical pond communities.
907 Freshwater Biology 62:519–529.
908 Schluter, D., and J. D. McPhail. 1992. Ecological character displacement and speciation in
909 sticklebacks. The American Naturalist 140:85–108.
910 Simberloff, D., and W. Boecklen. 1981. Santa Rosalia reconsidered: size ratios and
911 competition. Evolution 35:1206–1228.
37 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
912 Shaw, K. L. 2002. Conflict between nuclear and mitochondrial DNA phylogenies of a recent
913 species radiation: What mtDNA reveals and conceals about modes of speciation in Hawaiian
914 crickets. Proceedings of the Natural Academy of Sciences of the United States of America
915 99:16122-16127.
916 Speed, M. P., and K. Arbuckle 2017. Quantification provides a conceptual basis for
917 convergent evolution. Biological Reviews 92:815-829.
918 Stayton, C. T. 2008. Is convergence surprising? An examination of the frequency of
919 convergence in simulated datasets. Journal of Theoretical Biology 252:1-14.
920 Stevens, V. M., S. Whitmee, J. F.Le Galliard, J. Clobert, K. Böhning‐Gaese., D. Bonte, M.
921 Brändle, D. M. Dehling, , C. Hof, A. Trochet, and M. Baguette. 2014. A comparative analysis
922 of dispersal syndromes in terrestrial and semi‐terrestrial animals. Ecology Letters 17: 1039-
923 1052.
924 Untergasser, A., I. Cutcutache, T. Koressaar, J. Ye, B. C. Faircloth, M. Remm, and S. G.
925 Rozen. 2012. Primer3--new capabilities and interfaces. Nucleic Acids Research 40:e115.
926 Van Dooren, T. J. M., H. A. Thomassen, F. Smit, A. J. Helmstetter, and V. Savolainen. 2018.
927 Scope for sympatric giant-dwarf speciation driven by cannibalism in South-American annual
928 killifish (Austrolebias). bioRxiv, doi: https://doi.org/10.1101/121806.
929 Warren, D. L., M. Cardillo, D. F. Rosauer, and D. I. Bolnick. 2014. Mistaking geography for
930 biology: inferring processes from species distributions. Trends in Ecology & Evolution
931 29:572–580.
932 Webb, C. O., D. D. Ackerly, M. A. McPeek, and M. J. Donoghue. 2002. Phylogenies and
933 community ecology. Annual Review of Ecology and Systematics 33:475–505.
934 Weber, M. G., C. E. Wagner, R. J. Best, L. J. Harmon, and B. Matthews. 2017. Evolution in a
935 community context: On integrating ecological interactions and macroevolution. Trends in
936 Ecology & Evolution 32:291–304.
38 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
937 Weinstein, B. G., C. H. Graham, and J. L. Parra. 2017. The role of environment, dispersal
938 and competition in explaining reduced co-occurrence among related species. PLoS ONE
939 12:e0185493.
940 Willis, S. C., I. P. Farias, and G. Ortí. 2013. Testing mitochondrial capture and deep
941 coalescence in Amazonian cichlid fishes (Cichlidae: Cichla). Evolution 68:256–268.
942 Woodward, G., and A. G. Hildrew. 2002. Body-size determinants of niche overlap and
943 intraguild predation within a complex food web. Journal of Animal Ecology 71:1063–1074.
944 Wourms, J. P. 1972. The developmental biology of annual fishes. III. Pre-embryonic and
945 embryonic diapause of variable duration in the eggs of annual fishes. Journal of Experimental
946 Zoology 182:389–414.
947
948 References cited only in the supplementary materials:
949 Calviño, P. 2003. Redescubriendo Austrolebias robustus (Gunther, 1883). Suplemento
950 especial, Boletin del Killi Club Argentino, Enero 2003 1-13
951 Calviño, P. 2005. Austrolebias toba (Cyprinodontiformes: Rivulidae), una especie nueva de
952 pez anual de la Argentina. Revista del Museo Argentino de Ciencias Naturales 7:183-190.
953 Ferrer, J., L. R. Malabarba and W. J. E. M. Costa. 2008. Austrolebias paucisquama
954 (Cyprinodontiformes: Rivulidae), a new species of annual killifish from southern Brazil.
955 Neotropical Ichthyology 6:175-180.
956 Osinaga, K. 2006. Nuevo registro para Bolivia de Austrolebias monstrosus (Huber, 1995
957 Rivulidae). Kempffiana 2:60-62.
958 Perujo, E., P.A. Calviño, H. Salvia, and F. Prieto. 2005. Austrolebias luzardoi
959 (Cyprinodontiformes: Rivulidae), una especie nueva de pez anual de la cuenca del río
960 Cuareim,
961 República Oriental del Uruguay. Revista del Museo de La Plata 17:1-12
39 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
962
963
964
965
966 Figure 1 A map depicting the four main areas of endemism in Austrolebias, inferred by
967 drawing shapes around sampling locales for species in each region. The most northern region,
968 in purple, is the area west of the River Paraguay (W). The basin of the La Plata river and its
969 delta are in green (L). The drainage of the Negro River, part of the River Uruguay basin is
970 represented in blue (N). The Patos-Merin Lagoon region is highlighted in red (P). The species
971 used in this study are shown on the right side of this figure, grouped by area. Austrolebias
972 apaii and A. cinereus are not shown. Body size measurements are depicted as bars alongside
973 species names. Local patterns of species co-occurrence are shown in the pairwise table to the
974 left of species names. Coloured dots indicate that a record of a pond with both species was
975 found. Images of fishes are approximately to scale. All fish images are of males except A.
976 vazferreirai.
977
40 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
978
979 Figure 2 A decision tree showing how we can combine historical biogeography and trait
980 evolution with selection regime shifts to understand how replicated size distributions across
981 assemblages were generated. We examine the historical biogeography at each cladogenetic
982 event in the tree. Initially we do this at the regional level, classifying divergence events based
983 on results from BioGeoBEARS analysis (ARE). We subsequently use ancestral range
984 correlations (ARC) to see if this finer-scale approach can improve upon ARE. When a
985 selection regime shift in size or a shift in environmental niche ("environment" in the diagram)
986 is inferred just after divergence, we can determine whether the pattern corresponds to
987 adaptation to a new environment or to ecological interactions (e.g. competition). Several
988 patterns of shifts do not allow clear inference of processes. Once we have established how
989 major size variation generated by selection regime shifts has arisen we can then examine if
990 and how it has been redistributed throughout the assemblages.
991
992
993
41 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
994
995 Figure 3 Maximum clade credibility trees from Bayesian analyses of (A) nDNA, (B)
996 coalescent nDNA and (C) mtDNA. Black circles indicate a posterior probability (PP) from
997 0.90 – 1.00 and grey circles indicate a PP from 0.75 to 0.90. PP < 0.75 is depicted as a white
998 node. Highlighted, colour-coded sections represent three major clades that are recovered in
999 all trees. Relationships within clades may vary between trees. Branches units are in millions
1000 of years.
42 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
1001
1002 Figure 4 A biogeographic estimation of ancestral ranges using the trimmed nDNA tree and
1003 the best fitting model (DIVA+j with a maximum of two areas) in BioGeoBEARS. Most
1004 probable states are shown on each node and corner of the tree, current ranges are shown on
1005 the tips. Colours correspond to regions as depicted in the map in the top left and Figure 1.
1006 The current range of A. vazferreirai consists of two areas - Negro River and La Plata
1007 (coloured as teal). The body lengths of each species are shown as silhouettes to scale to the
1008 right of tree.
1009
43 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
1010
1011 Figure 5 (A) Ancestral state reconstruction of body size across Austrolebias using stabletraits
1012 (Elliot and Mooers 2014). Ancestral values shown are the medians of the posterior
1013 distributions of size. Colours correspond to the selective optima species are tending towards
1014 as inferred by SURFACE. Bars on the right show the 95% confidence intervals of these
1015 optima. The interval for the larger size (blue) is truncated. (B) The nDNA tree arranged by
1016 the species values for principal component two (PC2) of the environmental niche. Coloured
1017 branches show the clades in which niche shifts have taken place. (C) A scatterplot of the
1018 species averages of the first two principal components. Points are tips and nodes of the
1019 nDNA tree, connected by branches representing the inferred topology. Tip values are
1020 coloured based on current areas of endemism as shown in Figure 1. Black points represent
1021 median ancestral values reconstructed from the bivariate species averages using stabletraits.
1022 95% confidence ellipses for the combination of environmental niche values characterising
1023 each regime estimated by SURFACE are shown in their respective colours.
44 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
1024
1025 Figure 6 Assessments of patterns of past range overlap at cladogenetic events. (A) The
1026 nDNA tree with pie charts per node representing probabilities of assignment to mixture
1027 components. A model with three components was favoured. The component with the smallest
1028 value of overlap at the intercept is labelled red (allopatric), with increasing levels of overlap
1029 magenta and blue respectively (both non-allopatric). Nodes where we found concurrent
1030 results in our ARC and BioGeoBEARS analyses are plotted larger. (B) Scatterplot of range
1031 overlap per node on node age. Range overlap is estimated as in Barraclough and Vogler
1032 (2000) and fitted regression lines are shown for each cluster. A rug and density plot of
1033 bootstrapped intercept values is added at node age zero.
1034
45 bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
1 SUPPLEMENTARY INFORMATION
2
3 Running title: Coexistence in annual killifish
4 Title: Trait evolution and historical biogeography shape assemblages of annual killifish
5
6 Andrew J Helmstetter, Tom JM Van Dooren, Alexander ST Papadopulos, Javier Igea, Armand M Leroi, and Vincent Savolainen
7
8 Version 04 October 2018
9
10
11 8 FIGURES
12 S1-S8
13
14 8 TABLES
15 S1-S8
16
17 1
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
18
19
20 Figure S1 Results from ancestral range estimation using BioGeoBEARS. Most probable states under the best-fitting model for each tree are
21 depicted on nodes with letters indicating areas of endemism. This model was DIVA+j with a maximum of two areas for the coalescent tree (A)
22 and for the mtDNA tree (B). 2
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
23
24
25 Figure S2 Results from 500 biogeographic stochastic mapping simulations showing frequency histograms of event counts for each event type.
26 Results are shown for the (A) nDNA tree, (B) coalescent tree and (C) mtDNA tree. 3
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
27
28
29 Figure S3 SURFACE analyses showing shifts in selective optima and trait values for body size and two environmental niche PCs using the (A) nDNA tree (B) coalescent
30 tree and (C) mtDNA tree. Black branches represent branches where the trait is attracted to the first selective optimum for each trait, blue branches the second and red
31 branches the third, if detected. Sizes of circles next to tips indicate magnitudes of trait values, with dark colours indicating negative magnitudes and light colours positive
32 values. Colours correspond to relevant shifts traits are involved in.
4
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
33
34
35 Figure S4 L1OU results for body size for nDNA (A), coalescent (B) and mtDNA (C) tree, where shifts are indicated by an asterisk and a change
36 in branch colour. Shift magnitudes are shown at the tree edge of the shift. The sample data are shown in the bar chart to the right of each tree,
37 with the data normalized.
5
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
38
A duraznensis B wolterstorffi C charrua affinis gymnoventris arachan
periodicus luteoflammulatus reicherti
alexandri cheradophilus viarius
3.05,−27.42 juanlangi * monstrosus vazferreirai toba prognathus robustus
paucisquama elongatus bellottii
patriciae vazferreirai melanoorus
nigripinnis viarius 2.85,−19.79vandenbergi
3.06,−14.36 * * monstrosus charrua luteoflammulatus elongatus arachan gymnoventris
prognathus reicherti wolterstorffi
cheradophilus robustus elongatus
luteoflammulatus bellottii cheradophilus
2.83,−18.93 bellottii * vandenbergi prognathus melanoorus melanoorus 3.08,−23.18monstrosus
2.83,−12.86 * * vandenbergi nigripinnis patriciae robustus patriciae alexandri
reicherti paucisquama affinis
arachan toba duraznensis
viarius affinis periodicus
charrua duraznensis juanlangi
vazferreirai juanlangi paucisquama
gymnoventris periodicus toba
wolterstorffi alexandri nigripinnis
39 −2.5 PC1 1.89 −2.46 PC2 1.11 −2.5 PC1 1.89 −2.46 PC2 1.11 −2.5 PC1 1.89 −2.46 PC2 1.11
40 Figure S5 L1OU results for two environmental niche PCs for nDNA (A), coalescent (B) and mtDNA (C) tree, where shifts are shown by an
41 asterisk and affected tree branches by a change in branch colour. The shift magnitudes are shown at the edge of the shift. The sample data for
42 each PC are shown in the two bar charts to the right of each tree. The niche traits were normalized for these bar charts.
43
6
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
44
45
7
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
46
8
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
47
9
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
48
10
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
49
50
51 Figure S6 Modelled species ranges. The raster area used in our MaxEnt analyses is shaded
52 grey. Outlines of river basins where species occur in are shown in black (obtained from a
53 global river basin shapefile, with vertices smoothed with a 500m threshold;
54 http://www.waterbase.org/download_data.html. Species occurrences are depicted as black
55 points. Modelled suitability was subjected to a threshold as detailed in the methods section
56 and the resulting predicted ranges are coloured green. The inference for A. paucisquama
57 failed.
11
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
58
59
60 Figure S7 Age-range correlation results depicted for the (A) coalescent tree with pie charts
61 showing predicted cluster membership at each node. Red indicates the cluster with the lowest
62 range overlap, purple intermediate level and blue the highest level of overlap. A scatterplot of
63 node age against range overlap (B) is shown with points coloured by cluster membership.
64 Regression lines fitted to each cluster and bootstrapped intercepts shown as a rug plot along
65 the y axis. A density plot of the inferred intercepts is shown in (C). The same is shown for the
66 mtDNA tree (D), with corresponding scatterplot (E) and density plot (F).
12
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
67
68
69 Figure S8 A graphical representation of four species assemblages in Austrolebias. Each box
70 represents an area of endemism; age (Ma) is shown on the y axis and maximum standard
71 length (mm) is on the x axis. The nuclear DNA phylogenetic tree is pruned to the species and
72 ancestors that existed in each area based on the results of our ancestral range estimation
73 analysis. Immigration events and source areas are represented on the left of the box and
74 emigrations are shown on the right. Nodes where age-range correlation (ARC) and ancestral
75 range estimation agreed on non-allopatry are coloured based on the cluster they belonged to
13
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
76 in our ARC mixture model (Fig. 6A). Shifts estimated by SURFACE are shown blue for
77 environmental niche (see Fig. 5B) and thick branches show the shift in body size optimum
78 (Fig. 5A), or a combination of the two in the case that both shifts were present on the same
79 branch.
80
14
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
81 Table S1 DNA sequences of primers for each locus
82
Locus Direction Sequence (5' - 3')
enc1 F GACATGCTGGAGTTTCAGGA
enc1 R ACTTGTTRGCMACTGGGTCAAA
rag1 F AGCTTCTCCCTGGCTTTCAC
rag1 R GAACGGGTTGGTTCTCCAGA
28s c1c2 F ACCCGCTGAATTTAAGCAT
28s c1c2 R TGAACTCTCTCTTCAAAGTTCTTTTC
28s c6c7 F TCACCTGCCGAATCAACTAGC
28s c6c7 R ACTACCACCAAGATCTGCAC
28s c12d12 F TTATGACTGAACGCCTCTAAG
28s c12d12 R TGACTTTCAATAGATCGCAG
rh1 F TGTCAACCCAGCAGCCTATG
rh1 R TGGTCTCAGACTCCTGCTGA
sh3px3 F TGCTCCATTGAAGACCCCAC
sh3px3 R TGTCGTCCATCTTCTTGGCA
16s SARL CGCCTGTTTATCAAAAACAT
16s SBRH CCGGTCTGAACTCAGATCACGT
12s F AAAAAGCTTCAAACTGGGATTAGATACCCCACTAT
12s R TGACTGCAGAGGGTGACGGGCGGTGTGT
cytb F GGCAAATAGGAARTATCATTC
cytb R TGACTTGAARAACCAYCGTTG
83
84
85
15
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
86 Table S2 A list of the data taken from Van Dooren et al. (2018) and used for phylogenetic
87 tree inference.
Tip Name Locus Van Dooren et al. 2018
alexandrii_sanjavier 16S alexandrii_sanjavier 2
arachan_parquerivera2 16S arachan_parquerivera 1
bellottii_sol 16S bellottii_sol 2
charrua_ca2 16S charrua_ca 1
cheradophilus_castillos 16S cheradophilus_castillos 4
cheradophilus_castillos2 16S cheradophilus_castillos 2
cheradophilus_lp 16S cheradophilus_lp2
cinereus_1 16S cinereus
cinereus_arroyoviboras 16S cinereus_arroyoviboras 2
elongatus_villasoriano 16S elongatus_villasoriano 2
gymnoventris_salamanca4 16S gymnoventris_salamanca 1
gymnoventris_velasquez 16S gymnoventris_velasquez 2
luteoflammulatus_1 16S luteoflammulatus_1
nigripinnis_ceibas 16S nigripinnis_ceibas 3
periodicus_1 16S periodicus
pterolebias_longipinnis 16S pterolebias_longipinnis 2
robustus_1 16S Robustus_1
s_magnificus 16S s_magnificus 2
alexandrii_sanjavier 12S alexandrii_sanjavier 2
arachan_parquerivera2 12S arachan_parquerivera2
bellottii_sol 12S bellottii_sol 2
charrua_ca 12S charrua_ca 3
charrua_ca2 12S charrua_ca 2
cheradophilus_castillos 12S cheradophilus_castillos 2
cheradophilus_castillos2 12S cheradophilus_castillos 3
16
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
cheradophilus_lp 12S cheradophilus_lp 2
cinereus_1 12S cinereus
cinereus_arroyoviboras 12S cinereus_arroyoviboras
elongatus_villasoriano 12S elongatus_villasoriano 2
gymnoventris_salamanca4 12S gymnoventris_salamanca 1
gymnoventris_velasquez 12S gymnoventris_velasquez 2
luteoflammulatus_1 12S Luteoflammulatus 1
nigripinnis_ceibas 12S nigripinnis_ceibas 2
periodicus_1 12S periodicus
prognathus_salamanca 12S prognathus_salamanca 2
pterolebias_longipinnis 12S pterolebias_longipinnis
robustus_1 12S Robustus 1
s_magnificus 12S s_magnificus 2
periodicus_1 c1c2 periodicus
alexandrii_sanjavier c1c2 alexandrii_sanjavier
arachan_parquerivera2 c1c2 arachan_parquerivera 2
charrua_ca c1c2 charrua_ca 1
charrua_ca2 c1c2 charrua_ca 2
cheradophilus_castillos c1c2 cheradophilus_castillos 1
cheradophilus_castillos2 c1c2 cheradophilus_castillos 2
cinereus_1 c1c2 cinereus
elongatus_villasoriano c1c2 elongatus_villasoriano 1
gymnoventris_salamanca4 c1c2 gymnoventris_salamanca 1
gymnoventris_velasquez c1c2 gymnoventris_velasquez
luteoflammulatus_1 c1c2 Luteoflammulatus 1
nigripinnis_ceibas c1c2 nigripinnis_ceibas
pterolebias_longipinnis c1c2 pterolebias_longipinnis
robustus_1 c1c2 Robustus 1
17
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
bellottii_sol c1c2 bellottii_sol
cinereus_arroyoviboras c1c2 cinereus_arroyoviboras
melanoorus_1 c1c2 melanoorus_r5km399 1
periodicus_1 c6c7 periodicus
alexandrii_sanjavier c6c7 alexandrii_sanjavier
arachan_parquerivera2 c6c7 arachan_parquerivera 1
charrua_ca c6c7 charrua_ca 1
cheradophilus_castillos2 c6c7 cheradophilus_castillos 2
cinereus_1 c6c7 Cinereus
gymnoventris_velasquez c6c7 gymnoventris_velasquez
luteoflammulatus_1 c6c7 luteoflammulatus_2
nigripinnis_ceibas c6c7 nigripinnis_ceibas 1
pterolebias_longipinnis c6c7 pterolebias_longipinnis
bellottii_sol c6c7 bellottii_sol
cinereus_arroyoviboras c6c7 cinereus_arroyoviboras
melanoorus_1 c6c7 melanoorus_r5km399
arachan_parquerivera2 cytb arachan_parquerivera 2
bellottii_sol cytb bellottii_sol 2
charrua_ca2 cytb charrua_ca 2
cheradophilus_castillos cytb cheradophilus_castillos
cheradophilus_castillos2 cytb cheradophilus_castillos 2
cinereus_1 cytb Cinereus 1
cinereus_2 cytb Cinereus 2
cinereus_arroyoviboras cytb cinereus_arroyoviboras 2
elongatus_gc cytb elongatus_ezeiza
elongatus_villasoriano cytb elongatus_villasoriano 2
gymnoventris_salamanca4 cytb gymnoventris_salamanca 4
gymnoventris_velasquez cytb gymnoventris_velasquez
18
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
luteoflammulatus_velasquez cytb Luteoflammulatus 2
nigripinnis_ceibas cytb nigripinnis_ceibas 2
periodicus_1 cytb periodicus
prognathus_salamanca cytb prognathus_salamanca 2
pterolebias_longipinnis cytb pterolebias_longipinnis 2
robustus_1 cytb Robustus
s_magnificus cytb s_magnificus 2
salviai_pdd cytb salviai_pdd 3
wolterstorffi_elbagre2 cytb wolterstorffi_elbagre
periodicus_1 c12d12 periodicus
alexandrii_sanjavier c12d12 alexandrii_sanjavier
arachan_parquerivera2 c12d12 arachan_parquerivera 2
charrua_ca c12d12 charrua_ca 3
charrua_ca2 c12d12 charrua_ca 4
cheradophilus_castillos c12d12 cheradophilus_castillos 1
cheradophilus_castillos2 c12d12 cheradophilus_castillos 2
elongatus_villasoriano c12d12 elongatus_villasoriano 1
gymnoventris_salamanca4 c12d12 gymnoventris_salamanca 1
gymnoventris_velasquez c12d12 gymnoventris_velasquez
luteoflammulatus_1 c12d12 luteoflammulatus
nigripinnis_ceibas c12d12 nigripinnis_ceibas 2
pterolebias_longipinnis c12d12 pterolebias_longipinnis
robustus_1 c12d12 Robustus 1
bellottii_sol c12d12 bellottii_sol
cinereus_arroyoviboras c12d12 cinereus_arroyoviboras
cheradophilus_lp c12d12 cheradophilus_lp
melanoorus_1 c12d12 melanoorus_r5km399 1
88
19
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
89 Note: Shown in the table are the tip labels in this study, the locus and the corresponding
90 sequence in Van Dooren et al. 2018.
91
20
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
92 Table S3 List of Austrolebias occurrence locations. Due to the sensitive nature of these data,
93 an excel file of all locations is available upon request.
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110 \
111
112
113
21
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
114 Table S4 Variables used for MaxEnt and PCA analysis of environmental variables at locations where Austrolebias have been observed.
115
Parameters Loadings PC1 Loadings PC2 Description
River Basin
SLOPE_MEAN -0.01664 -0.19114
ASPECT_MEAN -0.15439 0.01187 Mean compass direction slopes face
Soil
BLDFIE_M_sl2_1km_ll 0.14784 0.06689 Bulk density (fine earth) in kg / cubic–meter at depth 0.05 m
CLYPPT_M_sl2_1km_ll -0.17428 -0.10597 Clay content (0–2 micro meter) mass fraction in % at depth 0.05 m
CLYPPT_M_sl6_1km_ll -0.23710 0.02142 Clay content (0–2 micro meter) mass fraction in % at depth 1.00 m
CRFVOL_M_sl2_1km_ll 0. 19497 0.09239 Coarse fragments volumetric in % at depth 0.05 m
ORCDRC_M_sl2_1km_ll -0.12776 0.08128 Soil organic carbon content (fine earth fraction) in g per kg at depth 0.05 m
PHIHOX_M_sl2_1km_ll 0.28586 -0. 04331 Soil pH x 10 in H2O at depth 0.05 m
SLTPPT_M_sl2_1km_ll 0.18569 0.12186 Silt content (2–50 micro meter) mass fraction in % at depth 0.05 m
SLTPPT_M_sl6_1km_ll 0.23183 0.03450 Silt content (2–50 micro meter) mass fraction in % at depth 1.00 m
SNDPPT_M_sl2_1km_ll 0.02064 0.00988 Sand content (50–2000 micro meter) mass fraction in % at depth 0.05 m
SNDPPT_M_sl6_1km_ll 0.06806 -0.04879 Sand content (50–2000 micro meter) mass fraction in % at depth 1.00 m
22
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
Worldclim
Altitude -0.14227 -0.11809
Annual Mean Temperature -0.03943 -0.31807
Mean Diurnal Range 0.15978 -0.19825
Isothermality -0.08156 -0.24060
Temperature Seasonality 0.23948 0.03849
Max Temperature of Warmest Month 0.09716 -0.30108
Min Temperature of Coldest Month -0.15338 -0.22317
Temperature Annual Range 0.23235 -0.07502
Mean Temperature of Wettest Quarter 0.09391 -0.29189
Mean Temperature of Driest Quarter -0.17140 0.00490
Mean Temperature of Warmest Quarter 0.04840 -0.31611
Mean Temperature of Coldest Quarter -0.09713 -0.29070
Annual Precipitation -0.27622 -0.01402
Precipitation of Wettest Month -0.15459 -0.21615
Precipitation of Driest Month -0.25412 0.14519
Precipitation Seasonality 0.16840 -0.27028
23
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
Precipitation of Wettest Quarter -0.19086 -0.20828
Precipitation of Driest Quarter -0.26069 0.14317
Precipitation of Warmest Quarter -0.16828 -0.22459
Precipitation of Coldest Quarter -0.25299 0.16338
116
117 Note: Each variable name used is shown as well as a short description in cases where an explanation is needed. Loadings list the contribution of
118 each variable to the scores of the first two principal components used for niche trait analysis.
119
120
121
122
123
124
125
126
127 24
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
128 Table S5 Maximum male body size measurements (standard length) for each species and sources of each measurement.
129
Species SL (mm) Source
affinis 32.3 Costa 2006
alexandri 43 Costa 2006
arachan 45.8 Costa 2006
bellottii 70 Costa 2006
charrua 47.2 Costa 2006
cheradophilus 95.07 TVD Field Experiment 2008
durazensis 28.1 Costa 2006
elongatus 151.9 Costa 2006
gymnoventris 30.8 Costa 2006
juanlangi 33.7 Costa 2006
luteoflammulatus 48.3 Costa 2006
melanoorus 50 Costa 2006
monstrosus 150 Osinaga 2006
nigripinnis 32 Costa 2006
25
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
patriciae 41 Costa 2006
paucisquama 34.2 Ferrer et al. 2008
periodicus 40.5 Costa 2006, Perujo et al. 2005
prognathus 130 Photograph by M. Volcan
robustus 76 Calviño 2003
reicherti 45.9 Costa 2006
toba 45.5 Calviño 2005
vandenbergi 76 Photo, Halbluetzel, www.fishbase.org date accessed 01/07/2015
vazferreirai 83.2 Costa 2006
viarius 61 Costa 2006
wolterstorffi 77.5 Costa 2006
130
131
132
133
134
135 26
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
136 Table S6 Parameters and statistics from BioGeoBEARS analyses.
137
tree areas constraint model LnL numparams d e j AICc AICc_wt
mtDNA 2 none DEC -55.71 2 0.025 0.026 0 116 3.70E-09
mtDNA 2 none DEC+J -36.78 3 0.0019 1.00E-12 0.097 80.7 0.17
mtDNA 2 none DIVA -49.5 2 0.019 1.40E-10 0 103.5 1.80E-06
mtDNA 2 none DIVA+J -35.36 3 0.0022 1.00E-12 0.16 77.86 0.69
mtDNA 2 none BAYAREA -65.71 2 0.039 0.086 0 136 1.70E-13
mtDNA 2 none BAYAREA+J -36.88 3 0.0016 1.00E-07 0.16 80.9 0.15
mtDNA 2 adjacency DEC -64.86 2 0.11 0.063 0 134.3 1.80E-13
mtDNA 2 adjacency DEC+J -35.34 3 0.0041 1.00E-12 0.17 77.83 0.32
mtDNA 2 adjacency DIVA -62.57 2 0.11 0.048 0 129.7 1.80E-12
mtDNA 2 adjacency DIVA+J -34.85 3 0.0057 1.00E-12 0.13 76.84 0.52
mtDNA 2 adjacency BAYAREA -71.38 2 0.15 0.12 0 147.3 2.60E-16
mtDNA 2 adjacency BAYAREA+J -36.06 3 0.0036 1.00E-07 0.17 79.26 0.16
mtDNA 2 transgression DEC -62.9 2 0.065 0.029 0 130.3 1.20E-12
mtDNA 2 transgression DEC+J -35.4 3 0.0036 1.00E-12 0.17 77.94 0.3
27
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
mtDNA 2 transgression DIVA -57.7 2 0.066 0.02 0 119.9 2.30E-10
mtDNA 2 transgression DIVA+J -34.77 3 0.0049 1.00E-12 0.15 76.69 0.56
mtDNA 2 transgression BAYAREA -71.75 2 0.16 0.12 0 148.1 1.80E-16
mtDNA 2 transgression BAYAREA+J -36.14 3 0.0033 1.00E-07 0.17 79.43 0.14
mtDNA 4 none DEC -55.41 2 0.016 0.014 0 115.4 3.50E-09
mtDNA 4 none DEC+J -35.83 3 0.0017 1.00E-12 0.14 78.81 0.31
mtDNA 4 none DIVA -49.9 2 0.017 1.00E-12 0 104.3 8.70E-07
mtDNA 4 none DIVA+J -35.1 3 0.0021 1.00E-12 0.15 77.34 0.64
mtDNA 4 none BAYAREA -65.57 2 0.018 0.1 0 135.7 1.40E-13
mtDNA 4 none BAYAREA+J -37.55 3 0.0015 1.00E-07 0.27 82.25 0.055
mtDNA 4 adjacency DEC -64.86 2 0.11 0.063 0 134.3 1.7e-13
mtDNA 4 adjacency DEC+J -35.34 3 0.0041 1.0e-12 0.17 77.83 0.30
mtDNA 4 adjacency DIVA -62.57 2 0.11 0.048 0 129.7 1.7e-12
mtDNA 4 adjacency DIVA+J -34.76 3 0.0050 1.0e-12 0.15 76.67 0.55
mtDNA 4 adjacency BAYAREA -71.38 2 0.15 0.12 0 147.3 2.5e-16
mtDNA 4 adjacency BAYAREA+J -36.06 3 0.0036 1.0e-07 0.17 79.26 0.15
mtDNA 4 transgression DEC -63.17 2 0.061 0.029 0 130.9 9.70E-13
28
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
mtDNA 4 transgression DEC+J -35.45 3 0.0038 1.00E-12 0.15 78.04 0.29
mtDNA 4 transgression DIVA -57.92 2 0.062 0.02 0 120.4 1.80E-10
mtDNA 4 transgression DIVA+J -34.78 3 0.0051 1.00E-12 0.14 76.71 0.56
mtDNA 4 transgression BAYAREA -73.95 2 0.12 0.12 0 152.4 2.00E-17
mtDNA 4 transgression BAYAREA+J -36.14 3 0.0035 1.00E-07 0.17 79.43 0.14
nDNA 2 none DEC -59.99 2 0.03 0.033 0 124.5 1.20E-10
nDNA 2 none DEC+J -37.12 3 0.0019 1.00E-12 0.17 81.38 0.27
nDNA 2 none DIVA -55.9 2 0.026 0.012 0 116.3 7.00E-09
nDNA 2 none DIVA+J -36.68 3 0.0019 1.00E-12 0.26 80.51 0.42
nDNA 2 none BAYAREA -66.64 2 0.054 0.094 0 137.8 1.50E-13
nDNA 2 none BAYAREA+J -37 3 0.0013 1.00E-07 0.27 81.14 0.31
nDNA 2 adjacency DEC -65.81 2 0.12 0.064 0 136.2 3.10E-13
nDNA 2 adjacency DEC+J -37.39 3 0.0027 1.00E-12 0.71 81.93 0.18
nDNA 2 adjacency DIVA -63.87 2 0.096 0.051 0 132.3 2.10E-12
nDNA 2 adjacency DIVA+J -37.02 3 0.0047 1.00E-12 0.59 81.19 0.27
nDNA 2 adjacency BAYAREA -72.59 2 0.14 0.097 0 149.7 3.50E-16
nDNA 2 adjacency BAYAREA+J -36.29 3 0.0032 1.00E-07 0.23 79.72 0.55
29
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
nDNA 2 transgression DEC -66.81 2 0.096 0.053 0 138.2 7.00E-14
nDNA 2 transgression DEC+J -36.62 3 0.0036 1.00E-12 0.43 80.38 0.25
nDNA 2 transgression DIVA -64.45 2 0.1 0.047 0 133.5 7.40E-13
nDNA 2 transgression DIVA+J -36.08 3 0.0046 1.00E-12 0.24 79.3 0.43
nDNA 2 transgression BAYAREA -73.69 2 0.11 0.11 0 151.9 7.30E-17
nDNA 2 transgression BAYAREA+J -36.36 3 0.003 1.00E-07 0.23 79.86 0.32
nDNA 4 none DEC -59.01 2 0.017 0.013 0 122.6 2.60E-10
nDNA 4 none DEC+J -36.58 3 0.0015 1.00E-12 0.27 80.3 0.39
nDNA 4 none DIVA -55.31 2 0.02 1.00E-12 0 115.2 1.10E-08
nDNA 4 none DIVA+J -36.65 3 0.0017 1.00E-12 0.24 80.44 0.37
nDNA 4 none BAYAREA -65.95 2 0.021 0.11 0 136.5 2.50E-13
nDNA 4 none BAYAREA+J -37.08 3 0.0009 1.00E-07 0.31 81.3 0.24
nDNA 4 adjacency DEC -65.81 2 0.12 0.064 0 136.2 3.10E-13
nDNA 4 adjacency DEC+J -37.39 3 0.0027 1.00E-12 0.71 81.93 0.18
nDNA 4 adjacency DIVA -63.87 2 0.096 0.051 0 132.3 2.10E-12
nDNA 4 adjacency DIVA+J -37.02 3 0.0047 1.00E-12 0.59 81.19 0.27
nDNA 4 adjacency BAYAREA -72.59 2 0.14 0.097 0 149.7 3.50E-16
30
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
nDNA 4 adjacency BAYAREA+J -36.29 3 0.0032 1.00E-07 0.23 79.72 0.55
nDNA 4 transgression DEC -68.2 2 0.091 0.059 0 140.9 2.00E-14
nDNA 4 transgression DEC+J -36.84 3 0.0035 1.00E-12 0.51 80.83 0.22
nDNA 4 transgression DIVA -65.56 2 0.088 0.045 0 135.7 2.70E-13
nDNA 4 transgression DIVA+J -36.21 3 0.0059 1.00E-12 0.31 79.57 0.42
nDNA 4 transgression BAYAREA -76.3 2 0.11 0.12 0 157.1 5.90E-18
nDNA 4 transgression BAYAREA+J -36.36 3 0.003 1.00E-07 0.23 79.87 0.36
coalescent 2 none DEC -56.96 2 0.041 0.038 0 118.5 1.10E-09
coalescent 2 none DEC+J -36.3 3 0.0039 1.00E-04 0.2 79.75 0.28
coalescent 2 none DIVA -51.71 2 0.037 0.018 0 108 2.10E-07
coalescent 2 none DIVA+J -35.61 3 0.0032 1.00E-12 0.21 78.36 0.56
coalescent 2 none BAYAREA -72.32 2 0.069 0.13 0 149.2 2.30E-16
coalescent 2 none BAYAREA+J -36.89 3 0.0023 1.00E-07 0.22 80.92 0.16
coalescent 2 adjacency DEC -65.88 2 0.15 0.12 0 136.3 9.80E-14
coalescent 2 adjacency DEC+J -35.82 3 0.0059 1.00E-12 0.25 78.79 0.3
coalescent 2 adjacency DIVA -62.05 2 0.18 0.074 0 128.6 4.50E-12
coalescent 2 adjacency DIVA+J -35.37 3 0.0069 1.00E-12 0.22 77.89 0.47
31
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
coalescent 2 adjacency BAYAREA -79.08 2 0.24 0.18 0 162.7 1.80E-19
coalescent 2 adjacency BAYAREA+J -36.12 3 0.0051 1.00E-07 0.21 79.39 0.22
coalescent 2 transgression DEC -65.2 2 0.18 0.092 0 134.9 1.70E-13
coalescent 2 transgression DEC+J -35.69 3 0.0059 1.00E-12 0.24 78.52 0.3
coalescent 2 transgression DIVA -62.07 2 0.17 0.075 0 128.7 3.80E-12
coalescent 2 transgression DIVA+J -35.14 3 0.0068 1.00E-12 0.22 77.42 0.51
coalescent 2 transgression BAYAREA -79.59 2 0.24 0.16 0 163.7 9.30E-20
coalescent 2 transgression BAYAREA+J -36.14 3 0.005 1.00E-07 0.21 79.42 0.19
coalescent 4 none DEC -55.27 2 0.023 0.014 0 115.1 4.30E-09
coalescent 4 none DEC+J -35.88 3 0.0023 1.00E-12 0.22 78.9 0.31
coalescent 4 none DIVA -50.15 2 0.026 0.0043 0 104.9 7.10E-07
coalescent 4 none DIVA+J -35.24 3 0.003 1.00E-12 0.19 77.63 0.58
coalescent 4 none BAYAREA -72 2 0.03 0.15 0 148.5 2.30E-16
coalescent 4 none BAYAREA+J -36.93 3 0.0021 1.00E-07 0.22 81.01 0.11
coalescent 4 adjacency DEC -65.88 2 0.15 0.12 0 136.3 9.80E-14
coalescent 4 adjacency DEC+J -35.82 3 0.0059 1.00E-12 0.25 78.79 0.3
coalescent 4 adjacency DIVA -62.05 2 0.18 0.074 0 128.6 4.50E-12
32
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
coalescent 4 adjacency DIVA+J -35.37 3 0.0069 1.00E-12 0.22 77.89 0.47
coalescent 4 adjacency BAYAREA -79.08 2 0.24 0.18 0 162.7 1.80E-19
coalescent 4 adjacency BAYAREA+J -36.12 3 0.0051 1.00E-07 0.21 79.39 0.22
coalescent 4 transgression DEC -65.72 2 0.13 0.077 0 135.4 7.0e-14
coalescent 4 transgression DEC+J -35.65 3 0.0058 1.0e-12 0.24 77.31 0.30
coalescent 4 transgression DIVA -62.02 2 0.17 0.073 0 128 2.9e-12
coalescent 4 transgression DIVA+J -35.09 3 0.0068 1.0e-12 0.22 76.17 0.52
coalescent 4 transgression BAYAREA -82.72 2 0.13 0.13 0 169.4 2.9e-21
coalescent 4 transgression BAYAREA+J -36.16 3 0.0049 1.0e-07 0.20 78.32 0.18
138
139 Note: Results from the six models per tree, with an adjacency-matrix (adjacency) or an adjacency matrix including marine transgressions
140 (transgression) and a maximum of two or four areas. Model fit was compared using log-likelihood and AICc.
141
142
143
144
33
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
145 Table S7 Comparisons of the best-fitting models with and without the jump-dispersal parameter using likelihood ratio tests.
146
data alt null LnLalt LnLnull DFalt DFnull DF Dstatistic pval
nDNA DIVA+J DIVA -36.68 -55.9 3 2 1 38.43 5.70E-10
coalescent DIVA+J DIVA -35.24 -50.15 3 2 1 29.82 4.70E-08
mtDNA DIVA+J DIVA -35.36 -49.5 3 2 1 28.28 1.00E-07
147
148 Note: Significant p-values indicate that the model with jump dispersal fits better.
149
150
151
152
153
154
155
156
157
34
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
158 Table S8 ICL values mixture regressions for ancestral range correlations.
159
Model Coalescent mtDNA nDNA
~ intercept + age + sizepic + nichepic 1: 16.4 1: 7.9 1: 22.2
2: 36.5 2: 3.6 2: 27.1
3: 35.0 3: 6.1 3: 13.4
~ intercept + age + sizepic 1: 16.5 1: 15.0 1: 24.2
2: 27.9 2: 26.7 2: 29.1
3: 34.3 3: 36.8 3: 16.2
~ intercept + age + nichepic 1: 22.0 1: 5.8 1: 19.9
2: 35.6 2: 5.5 2: 20.6
3: 27.5 3: 2.6 3: 13.6
~ intercept + age 1: 19.5 1: 11.7 1: 21.3
2: 23.6 2: 19.8 2: 20.2
3: 19.2 3: 24.5 3: 7.0
~ intercept + sizepic 1: 19.5 1: 18.7 1: 27.3
2: 27.8 2: 21.4 2: 20.6
35
bioRxiv preprint doi: https://doi.org/10.1101/436808; this version posted October 5, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.
3: 28.9 3: 30.2 3: 43.0
~ intercept + nichepic 1: 21.7 1: 4.1 1: 18.6
2: 27.4 2: 7.3 2: 9.8
3: 45.5 3: -5.5 3: 13.9
~ intercept 1: 21.0 1: 15.5 1: 24.1
2: 24.8 2: 17.4 2: 14.7
3: 48.1 3: 20.7 3: 37.1
160 Note: Model equations with explanatory variables are given and ICL values for mixtures with different components. Explanatory variables:
161 "age": node age; "sizepic": phylogenetic independent contrast for size; "nichepic": euclidean niche differences based on bivariate phylogenetic
162 independent contrasts for environmental niche PC1 and PC2. If mixture components had different regression equations these are given in the
163 "Model" column. The model with lowest ICL per tree is indicated in bold. In case the model with lowest ICL among the mixtures with only
164 intercepts and node age effects is a different model, it is indicated in red.
165
36