Habitat Loss Does Not Always Entail Negative Genetic Consequences

Although habitat loss has large, consistently negative effects on biodiversity, its genetic consequences are not yet fully understood. This is because measuring the genetic consequences of habitat loss requires accounting for major methodological limitations like the confounding effect of habitat fragmentation, historical processes underpinning genetic differentiation, time-lags between the onset of disturbances and genetic outcomes, and the need for large numbers of samples, genetic markers, and replicated landscapes to ensure sufficient statistical power. In this paper we overcame all these challenges to assess the genetic consequences of extreme habitat loss driven by mining in two herbs endemic to Amazonian savannas. Relying on genotyping-by-sequencing of hundreds of individuals collected across two mining landscapes, we identified thousands of neutral and independent single-nucleotide polymorphisms (SNPs) in each species and used these to evaluate population structure, genetic diversity, and gene flow. Since open-pit mining in our study region rarely involves habitat fragmentation, we were able to assess the independent effect of habitat loss. We also accounted for the underlying population structure when assessing landscape effects on genetic diversity and gene flow, examined the sensitivity of our analyses to the resolution of spatial data, and used annual species and cross-year analyses to minimize and quantify possible time-lag effects. We found that both species are remarkably resilient, as genetic diversity and gene flow patterns were unaffected by habitat loss. Whereas historical habitat amount was found to influence inbreeding; heterozygosity and inbreeding were not affected by habitat loss in either species, and gene flow was mainly influenced by geographic distance, pre-mining land cover, and local climate. Our study demonstrates that it is not possible to generalize about the genetic consequences of habitat loss, and implies that future conservation efforts need to consider species-specific genetic information.


Supplementary Figures
: Habitat loss with fragmentation (A) and without fragmentation (B). In the later case habitat loss does not result in decreased structural connectivity between habitat patches, since these were already separated. This is the dominant pattern of habitat loss resulting from open-pit mining in our study region (Fig. 1).

Figure S4:
Plots showing the optimal number of genetic clusters (k) for Brasilianthus carajensis (A and C) and Monogereion carajensis (B and D) based on the Admixture program (upper panels) and on the discriminant analysis of principal components -DAPC (lower panels). In the Admixture program the optimal k is based on cross-validation errors. In the DAPC the minimum value of Bayesian Information Criterion (BIC) indicate best-supported number of genetic cluster.

Figure S7
: Transformation applied to each of continuous (left panels) and categorical (right panels) variables to generate resistance surfaces for Brasilianthus carajensis from Serra Norte. For continuous variables, original data values are represented in x-axes, while transformed data values are shown in y-axes, and histograms present the distribution of original untransformed and transformed variables. For categorical variables, land cover classes were averaged across ten replicates and standard errors (SE) are provided. Categorical resistance surfaces were built for different years (1979, 2011, 2014 and 2016). Bio06 represents the minimum temperature of the coldest month, and Bio16 and Bio19 represent precipitation of wettest and coldest quarter, respectively.

Figure S8
: Transformation applied to each of continuous (left panels) and categorical (right panels) variables to generate resistance surfaces for Brasilianthus carajensis from Serra Sul. For continuous variables, original data values are represented in x-axes, while transformed data values are shown in y-axes, and histograms present the distribution of original untransformed and transformed variables. For categorical variables, land cover classes were averaged across ten replicates and standard errors (SE) are provided. Categorical resistance surfaces were built for different years (1979, 2014 and 2016). Bio06 represents the minimum temperature of the coldest month, and Bio16 and Bio19 represent precipitation of wettest and coldest quarter, respectively.

Figure S9
: Transformation applied to each of continuous (left panels) and categorical (right panels) variables to generate resistance surfaces for Monogereion carajensis from Serra Norte. For continuous variables, original data values are represented in x-axes, while transformed data values are shown in y-axes, and histograms present the distribution of original untransformed and transformed variables. For categorical variables, land cover classes were averaged across ten replicates and standard errors (SE) are provided. Categorical resistance surfaces were built for different years (1979, 2011, 2014 and 2016). Bio04 and Bio06 represent temperature seasonality and the minimum temperature of the coldest month, respectively, while Bio16 signifies precipitation of wettest quarter. Figure S10: Transformation applied to each of continuous (left panels) and categorical (right panels) variables to generate resistance surfaces for Monogereion carajensis from Serra Sul. For continuous variables, original data values are represented in x-axes, while transformed data values are shown in y-axes, and histograms present the distribution of original untransformed and transformed variables. For categorical variables, land cover classes were averaged across ten replicates and standard errors (SE) are provided. Categorical resistance surfaces were built for different years (1979, 2014 and 2016). Bio04 and Bio06 represent temperature seasonality and the minimum temperature of the coldest month, respectively, while Bio16 signifies precipitation of wettest quarter. Figure S12: Germination rate (A, B) and germination speed (C, D) of Brasilianthus carajensis (left panels) and Monogereion carajensis (right panels) in four different substrates: Whatman® paper, Canga topsoil, Forest topsoil and Mining waste. Mean values of the cumulative germination rate (percentage of seeds germinated at 33 th day) carrying the same letters indicate no significant differences between the treatments after a post hoc Tukey HSD test at P < 0.05. The error bars represent the mean standard deviation (n=5).