Determinants of genetic diversity and species richness of North American amphibians

Ecological limits on the population sizes and number of species a region is capable of supporting are thought to simultaneously produce spatial patterns in genetic diversity and species richness. However, we do not know the extent to which resource-based environmental limits jointly determine these patterns of biodiversity in ectotherms because of their low energy requirements compared to endotherms. Here, we adapt a framework for the ways ecological limits may shape genetic diversity and species richness previously tested in mammals for amphibians to determine whether similar processes produce continental patterns of biodiversity across both taxa. Repurposing open, raw microsatellite data from 19 species sampled at 554 sites in North America we found that spatial patterns of genetic diversity run opposite to patterns of species richness and genetic differentiation. However, while measures of resource availability and niche heterogeneity predict 89% of the variation in species richness, these landscape metrics are poor predictors of genetic diversity. Although heterogeneity appears to be an important driver of genetic and species biodiversity patterns in both amphibians and mammals, our results suggest that variation in genetic diversity both within and across species makes it difficult to infer general processes producing spatial patterns of amphibian genetic diversity.

Abstract: Ecological limits on the population sizes and number of species a region is capable of 26 supporting are thought to simultaneously produce spatial patterns in genetic diversity and species 27 richness. However, we do not know the extent to which resource-based environmental limits 28 jointly determine these patterns of biodiversity in ectotherms because of their low energy 29 requirements compared to endotherms. Here, we adapt a framework for the ways ecological 30 limits may shape genetic diversity and species richness previously tested in mammals for 31 amphibians to determine whether similar processes produce continental patterns of biodiversity 32 across both taxa. Repurposing open, raw microsatellite data from 19 species sampled at 554 sites 33 in North America we found that spatial patterns of genetic diversity run opposite to patterns of 34 species richness and genetic differentiation. However, while measures of resource availability 35 and niche heterogeneity predict 89% of the variation in species richness, these landscape metrics Introduction 45 Although species richness is higher in the tropics for most taxa, the details of diversity patterns 46 differ among species groups. In North America for instance, vertebrate richness generally 47 increases with resource availability, but mammals and birds tend to have higher species richness 48 in dry, mountainous areas, and reptiles and amphibians are more diverse in wet, lower elevation 49 regions (Currie 1991). This pattern suggests that while richness increases with resource 50 availability, taxon-specific traits may cause richness patterns to diverge from a strictly latitudinal analyses of mammals suggest that environments can simultaneously shape species richness and 57 genetic diversity on continental scales (Schmidt et al. 2020). Whether this is also true in 58 ectothermic taxa is unknown. Understanding whether common processes underlie variation in 59 biogeographic patterns of diversity across taxa with different environmental requirements can 60 help us move toward a general understanding of the drivers of biodiversity at multiple levels. 61 In mammals, continental scale multi-species patterns of nuclear genetic diversity and species 62 richness could be inferred with estimates of resource and niche availability, or ecological 63 opportunity (Schmidt et al. 2020). This suggested that ecological limits placed on the number of 64 individuals and species an environment can support are important drivers of broad-scale relationships between genetic diversity, species richness, and environments are consistent across 90 endothermic and ectothermic taxa. 91 The determinants of species richness across all terrestrial vertebrates are generally related to 92 resource availability as estimated by energy (e.g., potential evapotranspiration, primary 93 productivity), water-energy balance (e.g., actual evapotranspiration, precipitation), and 94 heterogeneity (e.g., elevation variability, land cover diversity) (Currie 1991; Kerr and Packer  Indeed in Europe, the best predictors of species richness in mammals and birds shift from energy 100 to water availability at decreasing latitudes, but amphibian species richness remains strongly 101 related to water-energy balance regardless of latitude (Whittaker et al. 2007).

102
The causes of population genetic diversity are rarely studied at the same time or scale as patterns 103 of species richness (but see Marshall and Camp 2006; Schmidt et al. 2020), yet the presumed 104 mechanisms related to the more-individuals hypothesis and heterogeneity are closely related to 6 reduces population size and limits gene flow due to increased niche specialization.

113
Heterogeneous environments also facilitate population differentiation due to spatially varying 114 selection. In mammals, evolutionary processes acting on the population level scaled up and 115 interacted with resource availability and heterogeneity to produce genetic diversity and species 116 richness patterns (Schmidt et al. 2020).

117
Whether mechanisms related to ecological limits and niche availability predict patterns of 118 species richness and genetic diversity in ectotherms is unclear. To test this prediction, we 119 analyzed previously-published microsatellite genotype data from 19 North American amphibian 120 species (8 frogs, 11 salamanders), with >13000 individuals sampled at 554 sites. Our first 121 objective was to identify existing spatial patterns in genetic diversity and differentiation and 122 quantify the extent to which genetic diversity and species richness covary. We then tested 123 whether limits on resources and niche availability jointly determined genetic diversity and 124 species richness using structural equation models, which allowed us to evaluate multiple 125 hypotheses at genetic and species level biodiversity simultaneously. We based our conceptual 126 framework on previous findings in mammals (Schmidt et al. 2020) and adapted it for 127 amphibians. First, we excluded human presence because previous investigation shows it did not 128 have a clear effect on amphibian genetic diversity (Schmidt and Garroway 2021a). Second, we 129 included a measure of water availability as an additional indicator of resource availability. Third, 130 body size was used as a proxy for whole species census size in mammals, however we do not 131 know the extent to which body size is correlated with effective population size or genetic 132 diversity within amphibians. We therefore first tested for a relationship between body size and 133 genetic diversity for species in our sample to determine whether it should be incorporated into 134 the model. Finally, we compare our results to previous results in mammals (Schmidt et  American amphibians compiled by (Schmidt and Garroway 2021a). This data set was assembled 142 from raw microsatellite datasets publicly archived in Dryad (DataDryad.org). To identify data 143 sets we conducted a systematic search of the Dryad data repository with the following keywords: 144 species name (e.g. Plethodon cinereus), "microsat*", "short tandem*", and "single tandem*". 145 We used the IUCN Red List database to obtain a list of amphibian species native to North 146 America for the search. We excluded datasets that lacked spatial reference, were not located in 147 North America, did not sample neutral microsatellite loci, or had study designs that may have Species richness. We estimated species richness at each of our genetic diversity sample sites 158 using amphibian range extent data from the IUCN RedList (IUCN 2019), applying filters for 159 native, extant species ranges. We measured species richness as the number of species' ranges 160 overlapping each genetic sample site.

162
Diversity maps and spatial variation partitioning 163 We used distance-based Moran's eigenvector maps (MEMs) to detect spatial patterns in genetic 164 diversity and differentiation and compare these to patterns of species richness. MEMs are 165 orthogonal spatial eigenvectors with eigenvalues that are directly proportional to Moran's I. 166 They measure spatial autocorrelation at all scales present in the data. We computed MEMs in the  Next, we determined the extent to which spatial patterns in genetic diversity and species richness 177 were shared using variation partitioning. Because our MEM analysis for both levels of 178 biodiversity had the same input distance matrix, the resulting spatial MEMs were directly 9 comparable. This was not the case for genetic differentiation, which had fewer sample sites. We 180 therefore did not partition variation in genetic differentiation because these MEMs are not the 181 same as those computed for genetic diversity and species richness. 182 We determined the fraction of total variation explained by spatial structure, shared spatial 183 structure, and non-spatial variation using variation partitioning as follows. We ran a series of  with species as a random effect allowing intercepts to vary and predicted a negative relationship. 230 We found no relationship between body size and genetic diversity in our data (Fig. S2) and positive effects on genetic diversity and species richness, respectively. We hypothesize that 235 these direct effects would act through population size were such data available.  Finally, we tested whether landscape heterogeneity was related to increased population 267 differentiation. We regressed heterogeneity on population-specific FST using a hierarchical model 268 with a random effect for species accounting for differences in mean FST (intercepts) while 269 allowing the strength and direction of the effect of heterogeneity (slopes) to vary across species.

270
To account for spatial variation we included MEMs describing spatial patterns in FST as 271 covariates. We performed these analyses in parallel across all four heterogeneity buffers.

274
Spatial patterns in genetic variation. We detected spatial patterns across genetic diversity, 275 genetic differentiation, and species richness (Fig. 1). The major axis of broad-scale variation is   Genetic diversity and differentiation in the western samples were in the mid-range of values 285 across sample sites (Fig. 1).

286
In general, species richness was more spatially structured than genetic diversity, with 85% and 287 23% of variation explained by spatial patterns, respectively (Fig. 1). We detected shared spatial 288 patterns between both levels of biodiversity, however, while shared patterns accounted for the 289 entirety of the spatial variation in genetic diversity, they explained less of the variation in species 290 richness (18%).

291
Common causes of genetic diversity and species richness. Our conceptual model (Fig 2a) fit the 292 data well (Fisher's C = 0.49, p = 0.78, 2 degrees of freedom) with no additional links suggested 293 at any scale. Note that for SEM, p > 0.05 means that our conceptual model is not rejected. We 294 present results from the 40 km buffer in the main text; results from all models can be found in 295 tables S1 -4. Species richness was well explained (R 2 = 0.89) and increased with water 296 availability, environmental heterogeneity, and species body size (Fig 2, Table S3). Water 297 availability had the strongest effect on species richness. Genetic diversity was not well predicted 298 by any variables in our model (R 2 = 0.04; Fig. 2). Residuals from genetic diversity models did 299 not exhibit spatial autocorrelation. Species richness residuals were spatially autocorrelated at 300 local scales (Moran's I = 0.06). In general, the environmental covariates in our models captured 301 broad spatial patterns well, and we did not incorporate fine-scale spatial structure into our 302 models as this was likely due to clustered sampling of some species. Lastly, genetic 303 differentiation within species decreased with heterogeneity (β = -0.32 ± 0.10 SE) at the most 304 local spatial scale we tested (10 km buffer), but this relationship disappeared at larger buffer 305 sizes.

308
We found spatial variation shared across genetic diversity, differentiation, and species richness at 309 broad spatial scales (Fig. 1). In general, areas with high species richness tended to have 310 genetically differentiated populations with relatively low genetic diversity. There was a 311 latitudinal gradient in genetic diversity across eastern sites which varied in the opposite direction 312 of the gradient in species richness and genetic differentiation. These patterns are consistent with 313 our predicted effects of heterogeneity, however, these relationships were not well-reflected by our structural equation model (Fig. 2). Our variation partitioning suggests that all spatial 315 variation in genetic diversity was shared with species richness, but the environments that 316 predicted species richness did not predict genetic diversity well. This finding suggests that other 317 processes drive the genetic diversity patterns we detect. These processes are likely species-318 specific, and not generalizable at continental scales as they were in mammals (Schmidt et al.   Interestingly, although genetic diversity was not affected by environmental heterogeneity at any 334 scale, genetic differentiation decreased with heterogeneity at the most local scale we tested; 335 genetic differentiation was relatively low in the northeast where landscape heterogeneity was 336 higher (Fig. 1). If heterogeneity increases niche availability and creates opportunities for 337 specialization and divergence, we predicted that it would increase genetic differentiation. 338 However, the pattern we detect here may be expected if species that were capable of recolonizing 339 northern regions following glaciation tend to be widely distributed generalists that maintain  Despite limitations modeling population size, it appears that spatial patterns in genetic diversity 382 and species richness in amphibians are driven by processes similar to mammals. Interestingly, 383 similar proportions of variation in genetic diversity (~ 25%) could be attributed to spatial 384 processes across both taxa. Additionally, it was the case in both groups that spatial variation in 385 genetic diversity was primarily due to factors that also shaped species richness, but other factors        Tables S1-S4). Regression coefficients with standard errors are shown along each path. Paths between variables where no effect was detected are colored in gray (see Table S3 for a complete summary of all paths). R 2 , the proportion of variation explained by the model, is given for genetic diversity and species richness. For genetic diversity, R 2 m is the variation explained by fixed effects and R 2 c is the variation explained by both fixed effects and the random species effect.

Determinants of genetic diversity and species richness of North American amphibians Supplementary Information
Figures S1-3 Tables S1-4 Figure S1. Maps of raw data for genetic diversity, species richness, and genetic differentiation. Genetic diversity is Nei's gene diversity; species richness is the number of amphibian species' ranges overlapping each site; genetic differentiation is site-specific FST. Figure S2. Relationship between genetic diversity and body size (snout-vent length). We detected no relationship using a mixed-effects model including a random intercept for species (β = -0.03 ± 0.2 SE). Figure S3. Spatial patterns (MEMs) for genetic diversity and species richness are correlated with environments: energy availability (potential evapotranspiration; PET), water availability (actual evapotranspiration; AET), and heterogeneity (land cover diversity). MEMs are ordered along the x-axis according to spatial scale explained, from broadest (MEM1) to finest (MEM18). MEMs to the left of the dashed lines indicate the broadest-scale patterns with Moran's I > 0.25 used to produce maps of genetic diversity and species richness. Light blue bars are MEMs explaining spatial patterns of species richness, and dark blue bars are MEMs explaining spatial patterns shared by genetic diversity and species richness. No MEMs explained spatial patterns in genetic diversity only.