Comparative phylogeography of five widespread tree species: Insights into the history of western Amazonia

Abstract Various historical processes have been put forth as drivers of patterns in the spatial distribution of Amazonian trees and their population genetic variation. We tested whether five widespread tree species show congruent phylogeographic breaks and similar patterns of demographic expansion, which could be related to proposed Pleistocene refugia or the presence of geological arches in western Amazonia. We sampled Otoba parvifolia/glycycarpa (Myristicaceae), Clarisia biflora, Poulsenia armata, Ficus insipida (all Moraceae), and Jacaratia digitata (Caricaceae) across the western Amazon Basin. Plastid DNA (trnH–psbA; 674 individuals from 34 populations) and nuclear ribosomal internal transcribed spacers (ITS; 214 individuals from 30 populations) were sequenced to assess genetic diversity, genetic differentiation, population genetic structure, and demographic patterns. Overall genetic diversity for both markers varied among species, with higher values in populations of shade‐tolerant species than in pioneer species. Spatial analysis of molecular variance (SAMOVA) identified three genetically differentiated groups for the plastid marker for each species, but the areas of genetic differentiation were not concordant among species. Fewer SAMOVA groups were found for ITS, with no detectable genetic differentiation among populations in pioneers. The lack of spatially congruent phylogeographic breaks across species suggests no common biogeographic history of these Amazonian tree species. The idiosyncratic phylogeographic patterns of species could be due instead to species‐specific responses to geological and climatic changes. Population genetic patterns were similar among species with similar biological features, indicating that the ecological characteristics of species impact large‐scale phylogeography.


| INTRODUC TI ON
Comparative phylogeography has the potential to elucidate the demographic history of co-occurring species at large spatial scales since different organisms are expected to accumulate similar geographic genetic signatures in response to common historical biogeographic drivers (Avise et al., 1987). For example, congruent geographic patterns of genetic diversity and divergence among populations in different tree species have been related to the occurrence of Pleistocene refugia and postglacial colonization events, in both temperate vegetation in the northern hemisphere (Petit et al., 2003) and tropical areas in Central Africa . In the Neotropics, comparative phylogeographic studies of tree species have focused on relating common phylogeographic breaks to geological events, such as the rise of the Andes (Dick, Lewis, Maslin, & Bermingham, 2013;Jones, Cerón-Souza, Hardesty, & Dick, 2013) or the boundary of Pliocene islands in Central America (Poelchau & Hamrick, 2013). In western Amazonia, common patterns of genetic differentiation among communities of Inga species (Fabaceae) were reported across a 250-km transect, which were used to argue for a zone of secondary contact between two historically isolated communities that were fragmented by a common event (Dexter, Terborgh, & Cunningham, 2012).

Early phylogeographic studies of animals in the western
Amazon basin showed that phylogeographic breaks coincided with geological paleoarches such as the Iquitos Arch in the Jurua River (Lougheed, Gascon, Jones, Bogart, & Boag, 1999;da Silva & Patton, 1998). In plants, paleoarches such as the Iquitos and Fitzcarrald arches also correlated well with some of the genetic groups detected in microsatellite markers for Theobroma cacao (Malvaceae), a widespread tree species in the Amazon (Motamayor et al., 2008), and the Fitzcarrald Arch coincided with the common phylogeographic break identified for a range of Inga tree species by Dexter et al. (2012). Paleoarches in western Amazonia, such as the Fitzcarrald Arch, formed because of uplift of the Andean Cordillera during the Late Miocene/Pliocene (Räsänen, Salo, Jungnert, & Pittman, 1990).
These geological features may have not attained sufficient elevation to directly isolate populations of lowland species (de Fátima Rossetti, Mann de Toledo, & Góes, 2005). Rather, the uplift of these paleoarches caused significant changes in drainage systems creating new habitats and environmental conditions (Espurt et al., 2010;Higgins et al., 2011;Regard et al., 2009), which may have played a role in shifting the distribution of species and causing population isolation.
Pleistocene climatic oscillations and contraction of the Amazon rain forest into refuges is another process that could cause common phylogeographic patterns across rain forest tree species (Prance, 1982). Past fragmentation of the rain forest is expected to leave a congruent genetic signature across lowland rain forest species, with refugial areas maintaining old populations with high levels of intraspecific genetic diversity, while populations in recently colonized areas would have low levels of genetic diversity Petit et al., 2003). Although the existence of such refuges for lowland species in Amazonia has been suggested to be controversial (Colinvaux, De Oliveira, & Bush, 2000), more recent studies indicate considerable impacts of glacial climates on Amazonia (Anhuf et al., 2006;Arruda, Fernandes-Filho, Solar, & Schaefer, 2017;Cheng et al., 2013), for example, suggesting expansion of drier formations in eastern Amazonia. These studies indicate greater stability of forest in northwestern Amazonia, differing from earlier work (e.g., Prance, 1982), which suggested multiple separate refugia in northwestern Amazonia.
If rain forest tree species experienced shifts in their distribution during Pleistocene climatic oscillations, then we may also expect them to show similar patterns of population expansion in the same areas. The reconstruction of tropical rain forest distribution during the Last Glacial Maximum (LGM) based on paleoenvironmental data suggests reduction and fragmentation of humid forest in 54% of the current area in Amazonia (Anhuf et al., 2006). This study showed differences of climatic conditions across western Amazonia during the LGM, with more stable climates in northwestern Amazonia (NWA), and the expansion of vegetation adapted to dry conditions (dry forest/savanna) in peripheral areas, including southwestern Amazonia (SWA). Palynological studies also suggest the recent expansion of the Amazon rain forest in SWA (Mayle, Burbridge, & Killeen, 2000), which would be consistent with Amazonian expansion during the current, wetter, interglacial.
The potential of paleoarches and Pleistocene refugia to drive common phylogeographic patterns across multiple tree species has not been tested at a large spatial scale across the western Amazon basin. Given the independent nature of the sampling schemes and different spatial scales of previous studies (Dexter et al., 2012;Dick et al., 2013;Jones et al., 2013;Motamayor et al., 2008), they are hard to compare. Here, we sought to apply a standardized approach for multiple Amazon tree species to facilitate more robust comparisons among different species and populations to explore the importance of historical events in shaping common phylogeographic patterns. By employing uniform and thorough geographic sampling of populations across widespread western Amazonian tree species, including three late successional species and two pioneers, we make a comparative study of their genetic diversity and population genetic structure. In particular, using plastid and nuclear markers sampled in 34 locations across 2,500 km of geographic distance, we test: 1. If there are concordant areas of genetic differentiation across populations of the five tree species, which would suggest common underlying reasons for phylogeographic congruence, due either to a) geological paleoarches such as the Fitzcarrald Arch or b) the contraction of populations into refugia during Pleistocene climate cycles.
2. If species show a genetic signature of population expansion in the same geographic areas, and particularly in southwestern Amazonia.

| The study species
We studied five widespread tree species that have overlapping ranges in western Amazonia. Otoba parvifolia/glycycarpa Pollen of these species is dispersed by different insects, and seeds are dispersed by animals (Table 1)  Our sampling spans the Fitzcarrald Arch (FA), a geological feature that divides the Amazonia into a northern and southern foreland basin (Espurt et al., 2010). Sampling on either side of the FA was equivalent in terms of number of populations sampled, and their geographic extent (107,233 km 2 north of FA vs. 130,554 km 2 south of FA), and the mean geographic distance between sampled populations did not differ significantly between the two regions (t test: t = 1.51, df = 443.4, p = 0.13). We therefore chose the FA to divide the sampled populations into two regional groups for some of the downstream analyses. The area north of the FA is generally termed the northwest Amazon (NWA) and is thought to have experienced a stable climate through Pleistocene climatic oscillations (Anhuf et al., 2006), while the area south of the FA is generally termed the southwest Amazon (SWA) and is thought to have experienced recent rain forest expansion (Mayle et al., 2000).

| Genetic diversity and haplotype definition
Genetic variation in each species was recorded as the number of nucleotide substitutions, indels, and inversions, the number of trnH-psbA haplotypes and ITS ribotypes, and the number of private haplotypes and ribotypes (i.e., the number of haplotypes or ribotypes F I G U R E 1 Map with the location of sampled localities, the Fitzcarrald Arch and the two regional groups (NWA, northwest Amazon and SWA, southwest Amazon). Areas proposed as refugia for plants by Prance (1982) are indicated in dotted lines. The Andean Cordillera is indicated by low altitudes in light gray (400-1200 m.a.s.l.) and higher altitudes in dark gray unique to each population). Genetic diversity was computed independently for each species using the trnH-psbA and ITS sequences by estimating the probability that two individuals carry distinct haplotypes or ribotypes (h) and the mean genetic distance between haplotypes or ribotypes of two distinct individuals (v) averaged over

populations (h S and v S ) and over all population pairs (h T and v T ) in
Permut CpSSR 2.0 (Pons & Petit, 1996). Measures were obtained by a permutation procedure set to 10,000 replicates.
For each species, haplotypes and ribotypes and their genealogical relationships were estimated independently for trnH-psbA and ITS sequences using statistical parsimony in the package "pegas" (Paradis, 2010) (Pons & Petit, 1996).

| Differentiation and population genetic structure
A spatial analysis of molecular variance (SAMOVA) was performed for each species for trnH-psbA and ITS sequences using Samova v. 1.0 (Dupanloup, Schneider, & Excoffier, 2002) to determine the position of genetic breaks among populations.
Several runs were performed using increasing numbers of groups (k = 1−20) and 100 annealing simulations for each k. In each, populations were clustered into genetically and geographically homogenous groups (Dupanloup et al., 2002). The number of groups was chosen so as to maximize genetic differentiation among them (Φ CT ). Genetic structure using pairwise nucleotide differences was further examined by analysis of molecular variance in Arlequin v.
3.5 (Excoffier & Lischer, 2010) for all populations and for groups of populations defined by SAMOVA. Significance of genetic structure indices was tested using a nonparametric randomization procedure.

| Historical demographic analysis
We tested for patterns of population expansion independently for each species using trnH-psbA and ITS sequences for all populations.
We used the neutrality test, Fu´s Fs (Fu, 1997), which is based on TA B L E 2 Number of samples, number of haplotypes or ribotypes, genetic diversity (h T and v T ), mantel test, genetic differentiation (G ST and N ST ), and demographic analyses (Fu´s Fs and HRI) for the trnH-psbA and ITS sequences in five western Amazonia tree species. Note that p-values for Fu's Fs are only considered significant at the 95% level if the p-value is lower than 0.02 (Excoffier & Lischer, 2010) Taxon a sudden demographic expansion model (Schneider & Excoffier, 1999). Multimodal distributions generally indicate that populations are at demographic equilibrium and unimodal ones suggest a recent demographic expansion (Rogers & Harpending, 1992). Statistics to compare the observed and simulated distributions were based on Harpending's raggedness index (HRI) using 1,000 parametric bootstrap replicates (Harpending, 1994

| Regional patterns
The number of haplotypes/ribotypes and private haplotypes/ribotypes, overall genetic diversity (v T ), and demographic analyses were estimated for each geographic region by independently analyzing populations located in the NWA and SWA regions. The regional analyses were only performed for trnH-psbA sequences, because of data availability.  Figure 2b).

| Genetic diversity and haplotype definition
50% of ribotypes were private to a given population in C. biflora and F. insipida (Table 2). Overall diversity values for ITS sequences, estimated by h T and v T , were high in C. biflora and P. armata (over 0.80), much lower in F. insipida (0.14), while measures were not available for TA B L E 3 Analyses of molecular variance based on pairwise nucleotide differences for the trnH-psbA and ITS sequences in five western Amazonia tree species. The analysis to examine genetic structure was run using pairwise nucleotide differences and independently for all populations and for populations grouped by SAMOVA groups

| Differentiation and population genetic structure
For trnH-psbA, the Mantel test between genetic and geographic distances showed significant isolation by distance for only two species, C. biflora and J. digitata (Table 2). Otoba parvifolia/glycycarpa had highly divergent haplotypes and significant phylogeographic structure ( Table 2). The most divergent populations were located in northern (haplotype 3) and southern Peru (haplotypes 10 and 14). Clarisia biflora had a group of closely related haplotypes, and most populations had single haplotypes. Poulsenia armata showed significant phylogeographic structure (Table 2) with the most divergent region occurring in southern Peru and Bolivia (H5, H6, and H7). Ficus insipida showed significant phylogeographic structure (Table 2)  For ITS, the Mantel test between genetic and geographic distances did not show significant isolation by distance for any species (Table 2). Clarisia biflora had significant phylogeographic structure (Table 2) Genetic differentiation among populations for the trnH-psbA sequences was generally above 91% in all species, except for F. insipida where only 67% of differentiation was explained by interpopulation nucleotide differences (Table 3). In the case of ITS, only C. biflora (55%) and P. armata (96%) showed significant genetic differentiation among populations (Table 3)

| Historical demographic analysis
Neutrality tests for the trnH-psbA and ITS sequences were not significant in most species (p > 0.02; Table 2), except for C. biflora and marginally for F. insipida, both for ITS. A unimodal mismatch distribution was only observed in C. biflora for trnH-psbA and ITS.
However, based on the HRI statistics, the recent demographic expansion model was consistently accepted for all species and markers, except for O. parvifolia/glycycarpa (p > 0.05; Table 2).

| Regional patterns
For trnH-psbA, NWA and SWA regions had haplotypes private to each region in all species, but other haplotypes were present in both regions such as H10 in O. parvifolia/glycycarpa, H4 in C. biflora, and H1-H3 in F. insipida (Figure 2a). Overall genetic diversity was similar in both regions for C. biflora and P. armata, while diversity in NWA was much greater than in SWA in J. digitata and slightly greater in O. parvifolia/glycycarpa and F. insipida (Table 4). We had only marginally significant negative values of Fu´s Fs for J. digitata in the SWA region. Within each regional group, recent expansion was consistently accepted for the SWA region for all species and for C. biflora, P. armata, and F. insipida in NWA (Table 4).

| No concordant areas of genetic differentiation
Across five Amazonian rain forest tree species, we did not find evidence for shared phylogeographic break that could have been congruent with either geological paleoarches or putative Pleistocene refugia. Of the various geological paleoarches in the Amazon (Motamayor et al., 2008), our sampling was best suited to testing the influence of the Fitzcarrald Arch on phylogeographic patterns, but we failed to find congruent genetic differentiation associated with this feature. This contrasts with the study of Inga species by Dexter et al. (2012), but it is important to note that only 10 of the 20 Inga species in Dexter et al. (2012) showed this genetic break.
The Fitzcarrald Arch does not seem to represent a consistent historical or present-day barrier to seed dispersal or gene flow for the Amazonian tree species in our study.

Recent work suggests stability of forest across northwestern
Amazonia at the Last Glacial Maximum (Anhuf et al., 2006;Arruda et al., 2017;Cheng et al., 2013), which contrasts with earlier suggestions (see various chapters in Prance, 1982) of multiple refugia in this area. If there have been such multiple refugia, we would expect species to have congruent phylogeographic breaks in the areas between them. We found no such evidence for congruent phylogeographic breaks across the western Amazon basin. Our results contrast with those of a comparative phylogeographic study in tropical areas of Atlantic Central Africa that found congruent genetic patterns in five of eight tree study species supporting a scenario of past forest fragmentation caused by glacial climate change . By using four densely sampled species, Heuertz, Duminil, Dauby, Savolainen, and Hardy (2014) also found a common genetic break with differentiation of northern and southern populations in the same region of Central Africa.
Our contrasting results may indicate that there has been greater stability of Pleistocene climates in western Amazonia than in tropical Africa.
For each of the five tree species in our study, we found significant genetic differentiation among populations for the plastid marker, with clear geographic breaks between more genetically homogeneous sets of populations, but these breaks are not geographically coincident.

| Common genetic patterns
A regional signal of low overall genetic diversity in SWA for the plastid marker was observed in J. digitata, and to a lesser degree for O. parvifolia/glycycarpa and F. insipida. This regional pattern is consistent with the postglacial rain forest expansion reported on the southern margin of the Amazon rain forest in Bolivia, based on pollen and charcoal records (Burbridge, Mayle, & Killeen, 2004;Mayle et al., 2000).  (Nason, Herre, & Hamrick, 1998) and over more than 150 km in riparian vegetation of the Namibian desert (Ahmed, Compton, Butlin, & Gilmartin, 2009). In these two species, pollen dispersal may succeed even among geographically distant individuals, indicating high effective population sizes and gene flow over large distances (Loveless & Hamrick, 1984). Very low genetic diversity in ITS was also characteristic of other pioneer species such as C. pentandra (Dick et al., 2007), C. alliodora (Rymer et al., 2013), and Schizolobium parahyba (Turchetto-Zolet et al., 2012), in which pollen is dispersed by bats, moths, or bees, respectively (Bawa, Bullock, et al., 1985;Bawa, Perry, & Beach, 1985), and in the shade-tolerant species Symphonia globulifera (Dick & Heuertz, 2008), in which pollen is dispersed by hummingbirds (Bawa, Bullock, et al., 1985). Further studies on the effectiveness of gene flow will require sampling of species with different growth strategies.

| CON CLUS IONS
This comparative phylogeographic study of trees in western Amazonia demonstrates that genetic breaks are not congruent among species.
We found no evidence consistent with the effects of Pleistocene refugia or paleoarches. These results imply that Amazon forest trees did not evolve, disperse, and successfully occupy the Amazon environment en masse, but rather each individual lineage did so idiosyncratically. The one signature of genetic concordance that we uncovered is that of recent, rapid population expansion in southwestern Amazonia, which is consistent with paleoecological evidence for the expansion of the southern fringe of the Amazonian rain forest since the Last Glacial Maximum (Burbridge et al., 2004;Mayle et al., 2000).
The lack of congruent patterns of genetic structure is consistent with studies that suggest a key role for idiosyncratic historical dispersal in shaping the biogeography of Amazonian tree species (e.g., Dexter et al., 2017). In addition, genetic patterns shown by pioneer and shade-tolerant species suggest that understanding the contemporary ecology of species is critical to explaining their phylogeography, although our limited sample size in terms of numbers of species precludes formal tests of this (cf. Lowe et al., 2018).
In sum, our results suggest that idiosyncratic dispersal and ecological differences among species together played key roles in shaping patterns of genetic diversity in neotropical rain forest tree species. This would explain why this comparative phylogeographic study, and most prior studies of individual species (e.g., Dick et al., 2007;Dick & Heuertz, 2008;Rymer et al., 2013), did not detect congruent patterns. Like Amazonian birds (Smith et al., 2014), Amazonian tree species appear to have originated and occupied environments individualistically rather than as congruent components of a coherent Amazonian flora.