*2.2. Population-Level Differentiation between Spatially Disparate B. atrox Groups*

First, we identified the population-level differentiation among venoms of snakes collected at the different habitats, based on the individual venom composition of each group of snakes. The PERMANOVA analysis (Table 1) showed that the environment explained 22.5% of the total variation in venom peak abundances in the sampled snakes (df = 3.33, F = 3.2, *p* < 0.0001). Further exploration of these differences with compositionally robust PCA (Figure 2, score plots) showed that PC1 largely differentiates floodplain from pasture venoms. Forest and degraded venoms occupying an intermediate space on the PC1 axis. Components with the most prominent loadings are fraction 10 and the fractions eluted after 85 min. The highest negative values of PC1, responsible for clustering the pasture venoms, were fraction 23 (PC1 = −0.7308, PC2 = −0.6035) and fraction 21 (PC1 = −0.5028, PC2 = −0.6410). These fractions were characterized in our previous study [15] as PIII-class and PI-class SVMP isoforms, respectively. The highest positive values of PC1, responsible for clustering the floodplain venoms, were fraction 20 (PC1 = 0.2958, PC2 = −0.3732), fraction 24 (PC1 = 0.1241, PC2 = 0.2006), and fraction 10 (PC1 = 0.2284, PC2 = −0.0353), which in our previous study have been characterized as different isoforms of the PIII-class and PI-class SVMPs, with minor proportions of CTL isoforms, and an SVSP, respectively (Figure 2, loading plots).

**Table 1.** Posthoc pairwise PERMANOVA analyses were used to test for significant venom compositional differentiation among all pairs of populations.


Values shown are *p*-values (above diagonal) and R2 values (below the diagonal) for each comparison.

**Figure 2.** Principal component analysis based on peak areas of *B. atrox* individual venoms according to their chromatographic profiles by RP-HPLC, using a C-18 column. Score (**A**) and loading (**B**) plots in 2D graphs of the principal components axes (PC1 = 36% and PC2 = 22.5%) of venoms from *B. atrox* snakes captured in the forest ( ), pasture ( ), floodplain ( ), and degraded ( ) habitats.

In addition, posthoc comparisons of individual population pairs using PERMANOVA showed that the spatially proximate degraded and forest habitats were the only locations between which venom composition did not significantly differ. All other population pairs differed significantly, with the greatest degree of differentiation between the spatially disparate floodplain and pasture populations.

Next, peak-by-peak comparisons showed that several chromatographic peaks contributed to population-level differentiation among venoms. After FDR-correction, nine peaks showed significant variation: Peaks 1, 3, 5, 7, 8, 12, 20, 21, and 23 (Figure 3).

**Figure 3.** Variation in the chromatographic fractions in *B. atrox* snakes from different habitats. The centered log-ratio mean abundance for each reversed-phase high-performance liquid chromatography peak was plotted. Negative numbers correspond to low-abundance peaks, whereas positive numbers correspond to high abundance peaks. (\*) Asterisks indicate peaks that show significant populationlevel variation. X-axis labels correspond to RP-HPLC peak numbers (1 to 26).

In terms of comparisons between habitats, venoms from the forest or recently degraded area were similar in relative peak abundances. Pasture venom was distinct with a higher abundance of Peaks 5 and 7, which correspond to the acidic PLA2s, Peak 12, with CTL as the major toxin, and Peaks 21 and 23, which include the SVMPs. Significantly lower abundances were observed in pasture venoms for Peaks 8 and 20, which have as major toxins PLA2s and PI-SVMPs, respectively. Peak 20 was practically absent in pasture venoms (Figure 1). Floodplain venoms were distinct in peaks related to SVMPs and PLA2s: Peaks 1, 21, and 23, which contain SVMPs and disintegrins, were present at a lower abundance, while Peak 20, which contains mostly PI-SVMPs, is at a comparatively higher abundance. In peaks containing PLA2s, Peak 3, which contains K-49 basic PLA2s, is less abundant, while Peak 7, which contains acidic PLA2s, and Peak 8, rich in D-49 basic PLA2s, were proportionally higher in the venoms from the floodplain.
