Brenna R. Forester

Integrating landscape genomics and spatially explicit approaches to detect loci under selection in clinal populations

Uncovering the genetic basis of adaptation hinges on the ability to detect loci under selection. However, population genomics outlier approaches to detect selected loci may be inappropriate for clinal populations or those with unclear population structure because they require that individuals be clustered into populations. An alternate approach, landscape genomics, uses individual‐based approaches to detect loci under selection and reveal potential environmental drivers of selection. We tested four landscape genomics methods on a simulated clinal population to determine their effectiveness at identifying a locus under varying selection strengths along an environmental gradient. We found all methods produced very low type I error rates across all selection strengths, but elevated type II error rates under “weak” selection. We then applied these methods to an AFLP genome scan of an alpine plant, Campanula barbata, and identified five highly supported candidate loci associated with precipitation variables. These loci also showed spatial autocorrelation and cline patterns indicative of selection along a precipitation gradient. Our results suggest that landscape genomics in combination with other spatial analyses provides a powerful approach for identifying loci potentially under selection and explaining spatially complex interactions between species and their environment.

Detecting spatial genetic signatures of local adaptation in heterogeneous landscapes

The spatial structure of the environment (e.g. the configuration of habitat patches) may play an important role in determining the strength of local adaptation. However, previous studies of habitat heterogeneity and local adaptation have largely been limited to simple landscapes, which poorly represent the multiscale habitat structure common in nature. Here, we use simulations to pursue two goals: (i) we explore how landscape heterogeneity, dispersal ability and selection affect the strength of local adaptation, and (ii) we evaluate the performance of several genotype–environment association (GEA) methods for detecting loci involved in local adaptation. We found that the strength of local adaptation increased in spatially aggregated selection regimes, but remained strong in patchy landscapes when selection was moderate to strong. Weak selection resulted in weak local adaptation that was relatively unaffected by landscape heterogeneity. In general, the power of detection methods closely reflected levels of local adaptation. False‐positive rates (FPRs), however, showed distinct differences across GEA methods based on levels of population structure. The univariate GEA approach had high FPRs (up to 55%) under limited dispersal scenarios, due to strong isolation by distance. By contrast, multivariate, ordination‐based methods had uniformly low FPRs (0–2%), suggesting these approaches can effectively control for population structure. Specifically, constrained ordinations had the best balance of high detection and low FPRs and will be a useful addition to the GEA toolkit. Our results provide both theoretical and practical insights into the conditions that shape local adaptation and how these conditions impact our ability to detect selection.

High performance computation of landscape genomic models including local indicators of spatial association.

With the increasing availability of both molecular and topo‐climatic data, the main challenges facing landscape genomics – that is the combination of landscape ecology with population genomics – include processing large numbers of models and distinguishing between selection and demographic processes (e.g. population structure). Several methods address the latter, either by estimating a null model of population history or by simultaneously inferring environmental and demographic effects. Here we present samβada, an approach designed to study signatures of local adaptation, with special emphasis on high performance computing of large‐scale genetic and environmental data sets. samβada identifies candidate loci using genotype–environment associations while also incorporating multivariate analyses to assess the effect of many environmental predictor variables. This enables the inclusion of explanatory variables representing population structure into the models to lower the occurrences of spurious genotype–environment associations. In addition, samβada calculates local indicators of spatial association for candidate loci to provide information on whether similar genotypes tend to cluster in space, which constitutes a useful indication of the possible kinship between individuals. To test the usefulness of this approach, we carried out a simulation study and analysed a data set from Ugandan cattle to detect signatures of local adaptation with samβada, bayenv, lfmm and an FST outlier method (FDIST approach in arlequin) and compare their results. samβada – an open source software for Windows, Linux and Mac OS X available at http://lasig.epfl.ch/sambada – outperforms other approaches and better suits whole‐genome sequence data processing.

Deep Genetic Divergence Between Disjunct Refugia in the Arctic-Alpine King's Crown, Rhodiola integrifolia (Crassulaceae)

Despite the strength of climatic variability at high latitudes and upper elevations, we still do not fully understand how plants in North America that are distributed between Arctic and alpine areas responded to the environmental changes of the Quaternary. To address this question, we set out to resolve the evolutionary history of the King’s Crown, Rhodiola integrifolia using multi-locus population genetic and phylogenetic analyses in combination with ecological niche modeling. Our population genetic analyses of multiple anonymous nuclear loci revealed two major clades within R. integrifolia that diverged from each other ~ 700 kya: one occurring in Beringia to the north (including members of subspecies leedyi and part of subspecies integrifolia), and the other restricted to the Southern Rocky Mountain refugium in the south (including individuals of subspecies neomexicana and part of subspecies integrifolia). Ecological niche models corroborate our hypothesized locations of refugial areas inferred from our phylogeographic analyses and revealed some environmental differences between the regions inhabited by its two subclades. Our study underscores the role of geographic isolation in promoting genetic divergence and the evolution of endemic subspecies in R. integrifolia. Furthermore, our phylogenetic analyses of the plastid spacer region trnL-F demonstrate that among the native North American species, R. integrifolia and R. rhodantha are more closely related to one another than either is to R. rosea. An understanding of these historic processes lies at the heart of making informed management decisions regarding this and other Arctic-alpine species of concern in this increasingly threatened biome.

Comparing methods for detecting multilocus adaptation with multivariate genotype–environment associations

Identifying adaptive loci can provide insight into the mechanisms underlying local adaptation. Genotype–environment association (GEA) methods, which identify these loci based on correlations between genetic and environmental data, are particularly promising. Univariate methods have dominated GEA, despite the high dimensional nature of genotype and environment. Multivariate methods, which analyse many loci simultaneously, may be better suited to these data as they consider how sets of markers covary in response to environment. These methods may also be more effective at detecting adaptive processes that result in weak, multilocus signatures. Here, we evaluate four multivariate methods and five univariate and differentiation‐based approaches, using published simulations of multilocus selection. We found that Random Forest performed poorly for GEA. Univariate GEAs performed better, but had low detection rates for loci under weak selection. Constrained ordinations, particularly redundancy analysis (RDA), showed a superior combination of low false‐positive and high true‐positive rates across all levels of selection. These results were robust across the demographic histories, sampling designs, sample sizes and weak population structure tested here. The value of combining detections from different methods was variable and depended on the study goals and knowledge of the drivers of selection. Re‐analysis of genomic data from grey wolves highlighted the unique, covarying sets of adaptive loci that could be identified using RDA. Although additional testing is needed, this study indicates that RDA is an effective means of detecting adaptation, including signatures of weak, multilocus selection, providing a powerful tool for investigating the genetic basis of local adaptation.

Guidelines for planning genomic assessment and monitoring of locally adaptive variation to inform species conservation

Identifying and monitoring locally adaptive genetic variation can have direct utility for conserving species at risk, especially when management may include actions such as translocations for restoration, genetic rescue, or assisted gene flow. However, genomic studies of local adaptation require careful planning to be successful, and in some cases may not be a worthwhile use of resources. Here, we offer an adaptive management framework to help conservation biologists and managers decide when genomics is likely to be effective in detecting local adaptation, and how to plan assessment and monitoring of adaptive variation to address conservation objectives. Studies of adaptive variation using genomic tools will inform conservation actions in many cases, including applications such as assisted gene flow and identifying conservation units. In others, assessing genetic diversity, inbreeding, and demographics using selectively neutral genetic markers may be most useful. And in some cases, local adaptation may be assessed more efficiently using alternative approaches such as common garden experiments. Here, we identify key considerations of genomics studies of locally adaptive variation, provide a road map for successful collaborations with genomics experts including key issues for study design and data analysis, and offer guidelines for interpreting and using results from genomic assessments to inform monitoring programs and conservation actions.

Coherent synthesis of genomic associations with phenotypes and home environments

Local adaptation is often studied via (i) multiple common garden experiments comparing performance of genotypes in different environments and (ii) sequencing genotypes from multiple locations and characterizing geographic patterns in allele frequency. Both approaches aim to characterize the same pattern (local adaptation), yet the complementary information from each has not yet been coherently integrated. Here, we develop a genome‐wide association model of genotype interactions with continuous environmental gradients (G × E), that is reaction norms. We present an approach to impute relative fitness, allowing us to coherently synthesize evidence from common garden and genome–environment associations. Our approach identifies loci exhibiting environmental clines where alleles are associated with higher fitness in home environments. Simulations show our approach can increase power to detect loci causing local adaptation. In a case study on Arabidopsis thaliana, most identified SNPs exhibited home allele advantage and fitness trade‐offs along climate gradients, suggesting selective gradients can maintain allelic clines. SNPs exhibiting G × E associations with fitness were enriched in genic regions, putative partial selective sweeps and associations with an adaptive phenotype (flowering time plasticity). We discuss extensions for situations where only adaptive phenotypes other than fitness are available. Many types of data may point towards the loci underlying G × E and local adaptation; coherent models of diverse data provide a principled basis for synthesis.

Spatial detection of outlier loci with Moran eigenvector maps

The spatial signature of microevolutionary processes structuring genetic variation may play an important role in the detection of loci under selection. However, the spatial location of samples has not yet been used to quantify this. Here, we present a new two‐step method of spatial outlier detection at the individual and deme levels using the power spectrum of Moran eigenvector maps (MEM). The MEM power spectrum quantifies how the variation in a variable, such as the frequency of an allele at a SNP locus, is distributed across a range of spatial scales defined by MEM spatial eigenvectors. The first step (Moran spectral outlier detection: MSOD) uses genetic and spatial information to identify outlier loci by their unusual power spectrum. The second step uses Moran spectral randomization (MSR) to test the association between outlier loci and environmental predictors, accounting for spatial autocorrelation. Using simulated data from two published papers, we tested this two‐step method in different scenarios of landscape configuration, selection strength, dispersal capacity and sampling design. Under scenarios that included spatial structure, MSOD alone was sufficient to detect outlier loci at the individual and deme levels without the need for incorporating environmental predictors. Follow‐up with MSR generally reduced (already low) false‐positive rates, though in some cases led to a reduction in power. The results were surprisingly robust to differences in sample size and sampling design. Our method represents a new tool for detecting potential loci under selection with individual‐based and population‐based sampling by leveraging spatial information that has hitherto been neglected.

Integrating environmental, molecular, and morphological data to unravel an ice-age radiation of arctic-alpine Campanula in western North America

Many arctic-alpine plant genera have undergone speciation during the Quaternary. The bases for these radiations have been ascribed to geographic isolation, abiotic and biotic differences between populations, and/or hybridization and polyploidization. The Cordilleran Campanula L. (Campanulaceae Juss.), a monophyletic clade of mostly endemic arctic-alpine taxa from western North America, experienced a recent and rapid radiation. We set out to unravel the factors that likely influenced speciation in this group. To do so, we integrated environmental, genetic, and morphological datasets, tested biogeographic hypotheses, and analyzed the potential consequences of the various factors on the evolutionary history of the clade. We created paleodistribution models to identify potential Pleistocene refugia for the clade and estimated niche space for individual taxa using geographic and climatic data. Using 11 nuclear loci, we reconstructed a species tree and tested biogeographic hypotheses derived from the paleodistribution models. Finally, we tested 28 morphological characters, including floral, vegetative, and seed characteristics, for their capacity to differentiate taxa. Our results show that the combined effect of Quaternary climatic variation, isolation among differing environments in the mountains in western North America, and biotic factors influencing floral morphology contributed to speciation in this group during the mid-Pleistocene. Furthermore, our biogeographic analyses uncovered asynchronous consequences of interglacial and glacial periods for the timing of refugial isolation within the southern and northwestern mountains, respectively. These findings have broad implications for understanding the processes promoting speciation in arctic-alpine plants and the rise of numerous endemic taxa across the region.

Using genotype-environment associations to identify multilocus local adaptation

Identifying loci under selection can provide insight into the mechanisms underlying local adaptation and inform management decisions for agricultural, natural resources, and conservation applications. Genotype-environment association (GEA) methods, which identify adaptive loci based on associations between genetic data and environmental variables, are particularly promising for distinguishing these loci. Univariate statistical methods have dominated GEA, despite the high dimensional nature of genomic data sets. Multivariate and machine learning methods, which can analyze many loci simultaneously, may be better suited to these large data sets since they consider how groups of markers covary in response to environmental predictors. These methods may also be more effective at detecting important adaptive processes, such as selection on standing genetic variation, that result in weak, multilocus signatures. Here we evaluate four of these methods, as well as a popular univariate approach, using published simulations of multilocus selection. We found that the machine learning method, Random Forest, performed poorly as a GEA. The univariate approach performed better, but had low detection rates for loci under weak selection. Constrained ordinations showed a superior combination of low false positive and high true positive rates across all levels of selection. These results were robust across demographic history, sampling designs, and sample sizes. Although further testing is needed on more complex genetic architectures, this study indicates that constrained ordinations are an effective means of detecting adaptive processes that result in weak, multilocus molecular signatures, providing a powerful tool for investigating the genetic basis of local adaptation and improving management actions.Identifying adaptive loci can provide insight into the mechanisms underlying local adaptation. Genotype-environment association (GEA) methods, which identify these loci based on correlations between genetic and environmental data, are particularly promising. Univariate methods have dominated GEA, despite the high dimensional nature of genotype and environment. Multivariate methods, which analyze many loci simultaneously, may be better suited to these data since they consider how sets of markers covary in response to environment. These methods may also be more effective at detecting adaptive processes that result in weak, multilocus signatures. Here, we evaluate four multivariate methods, and five univariate and differentiation-based approaches, using published simulations of multilocus selection. We found that Random Forest performed poorly for GEA. Univariate GEAs performed better, but had low detection rates for loci under weak selection. Constrained ordinations showed a superior combination of low false positive and high true positive rates across all levels of selection. These results were robust across the demographic histories, sampling designs, sample sizes, and levels of population structure tested. The value of combining detections from different methods was variable, and depended on study goals and knowledge of the drivers of selection. Reanalysis of genomic data from gray wolves highlighted the unique, covarying sets of adaptive loci that could be identified using redundancy analysis, a constrained ordination. Although additional testing is needed, this study indicates that constrained ordinations are an effective means of detecting adaptation, including signatures of weak, multilocus selection, providing a powerful tool for investigating the genetic basis of local adaptation.