Fig. S1 Predictions for the distribution of continuous rainforest (refugia) at the LGM, providing evidence for multiple refugia in Cameroon consistent with paleoecological data (Maley 1996). Lower and upper bounds for the prediction were estimated by comparison of model predictions for the present to percent tree cover data from the MODIS satellite.

Fig. S2 Habitats at sampling sites in Cameroon delineated with PCA on WORLDCLIM climate data, MODIS data (LAI maximum, % tree cover [VCF], NDVI maximum, and NDVI during greening up period), and radar (Quick Scatterometer mean and SD of backscatter) remote sensing data. PC1 and PC2 explain 54% and 21% of habitat variation among sampling sites, respectively.

Fig. S3 Distribution of localities for Trachylepis affinis, including literature, museum records, and our field sampling, used to build ecological niche models. Localities not included in models because of missing environmental data are not shown.

Fig. S4 (a) Minimum spanning tree of ND4 haplotypes showing lower haplotypic diversity in the ecotone, and haplotype sharing between the two rainforest genetic clusters identified with Structure. White plus signs and asterisks indicate divergent clusters of haplotypes that are found sympatrically at Kribi (site 1). (b) GDM of ND4 FST distance matrix, with environmental variables and geographic distance as predictors. Percent variation explained are for the full model (D+E), and geographic distance alone (D).

Fig. S5 GDM spatial predictions of genetic turnover in Trachylepis affinis for all 16 AFLP data sets (neutral pooled and individual outlier loci under selection), and variables selected in each GDM model that allows for environmental variables, geographic distance (and for neutral data, resistance surfaces accounting for the Sanaga River barrier and the effects of habitat suitability on movement) as possible predictors. Colors between panels are not comparable, while within panels, areas with similar shading along the color scale are predicted to be genetically more similar.

Fig. S6 Means of PC1 (body size) and PC2 (body shape) on morphological traits in T. affinis, showing morphological divergence between rainforest and ecotone habitats: rainforest (FOR); montane (MONT); lower ecotone (LOEC); upper ecotone (UPEC). Error bars represent ± 1SE. Habitats with the same letter (a or b) are not significantly different in morphology for the component in question. Habitat types were determined using climate variables, remote sensing of vegetation, and elevation (see Table S2 for details on variables).

Fig. S7 (a) Last Glacial Maximum and (b) present distributions of T. affinis, generated with an ecological niche model (Maxent), indicate postglacial range expansion. These models are based upon a restricted set of climate variables. LGM predictions employ climate data generated by PMIP’s CCSM3 general circulation model (see methods for details).

Fig. S8 ND4 mismatch distributions for sets of T. affinis populations comprising genetic clusters ascertained from Structure analysis of neutral AFLP loci (filled circles), and those simulated under a demographic expansion (solid line), and under a spatial expansion with constant population size (dashed line). All populations showed evidence for demographic expansion based upon differences between observed and simulated estimates of Ѳ0, Ѳ1, and τ (Schneider and Excoffier 1999); similar tests based upon the raggedness index (Harpending 1994) also failed to reject demographic expansion except for South of Sanaga River (SSAN). All genetic clusters but SSAN showed evidence for spatial expansion (Excoffier 2004). Populations showing substantial admixture (3 and 4) at AFLP loci are excluded.

Fig. S9 Mantel correlations between pairwise FST and geographic distance, produced by adding an additional geographic distance increment to forestecotone population pairs, showing negligible increase in correlation (neutral AFLPs) or decrease (ND4) in r when distance is added.

Table S1 Sampling localities, sample sizes by sex and molecular data type, and habitat for Trachylepis affinis collected in Cameroon.

Table S2 Bioclimatic, vegetation, and elevation variables included as possible predictor variables for generalized dissimilarity modeling of neutral and adaptive genetic variation in Trachylepis affinis.

Table S3 Localities for Trachylepis affinis used to build ecological niche models, with associated voucher specimen repository and catalogue number (if available), and literature source.

Table S4 Overall and population-level observed and estimated allele frequencies for AFLP loci sampled from 12 populations of Trachylepis affinis in Cameroon. Loci identified as under selection are based upon Bayes Factors (BF) denoting “very strong” evidence for selection (BF =31). Bolface BF are for loci with “definitive” evidence for selection (BF ≥ 99), while those in parentheses are for loci for “very strong” evidence for selection, followed up with a selection scan based only upon 8 forest populations.

Table S5 Diversity statistics by population for 191 repeatable AFLP markers, using Bayesian estimation of allele frequencies with non-uniform priors (Zhivotovsky 1999). Numbers in parenthesis are values for the 99 neutral loci.

Table S6 For 99 neutral AFLP loci, likelihood of K genetic clusters, probability of the data given the number of clustersa, and Evanno’s statistic (ΔK) for estimating the number of genetic clusters, based upon 10 Structure runs for each K with an admixture model.

Table S7 Loadings of morphological traits from Trachylepis affinis on the first two principle components.

Table S8 Mantel correlations between environmental variables and geographic distance at sampling localities. Significant correlations are in boldface. See Table S2 for definitions of environmental variables.

Table S9 Band frequencies for FST outlier loci showing divergence, based upon GDM, along the rainforest-ecotone gradient, showing parallel changes in band frequencies across ecotone populations in response to selection.

Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

MEC_4684_sm_SupportingText.pdf109KSupporting info item
MEC_4684_sm_Table_S4.xls205KSupporting info item
MEC_4684_sm_Tables-Figures.pdf5281KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.