Allele-frequency clines
We observed a highly significant allele-frequency cline between 24.9 °S and 4.8 °N for both the DE and SAE alleles (P <0.0001; DE loci: adjusted R 2 = 0.78, SAE loci: adjustedR 2 = 0.77; Figure 2, Table S2). The two slopes were similar, differing by a factor of only 1.2. Predicted DE allele frequencies decreased from 0.75 (95% CI: 0.72-0.77) at 24.9 °S to 0.46 (95% CI: 0.42-0.51) at 4.8 °N (a decrease of 0.0095 per degree latitude). Predicted SAE allele frequencies decreased from 0.40 (95% CI: 0.37-0.43) at 24.9 °S to 0.17 (95% CI: 0.13-0.20) at 4.8 °N (a decrease of 0.0079 per degree latitude). When considered together, the clines were significantly stronger than clines expected under a random selection of alleles (DE loci: P = 0.081, SAE loci: P = 0.043, all loci: P = 0.013), which indicates that they cannot be explained by genetic drift. The cline based on both the DE and SAE alleles explained as much as 87% of the variation (adjustedR 2; Figure 3, Table S2). Predicted allele frequencies decreased from 0.57 (95% CI: 0.55-0.59) at 24.9 °S to 0.29 (95% CI: 0.27-0.32) at 4.8 °N (a decrease of 0.0091 per degree latitude). No clines were observed for the remaining alleles, which were not associated with male-deleterious traits (negative adjustedR 2 values). Frequencies of DE alleles, but not SAE alleles, in East and central Africa were significantly higher than pooled frequencies based on a random selection of alleles (six DE alleles: multiply the 1-sided P = 0.013 by 2, nine SAE alleles: 1-sided P = 0.70).
The populations from HiP and Nairobi NP deviated from the cline, showing relatively low DE and SAE allele frequencies that were close to the frequencies of the remaining microsatellite alleles. In HiP 18 out of 20 and in Nairobi NP 16 out of 20 of the non-pooled male-deleterious-trait-associated alleles at the microsatellites analysed in both populations showed relatively low frequencies compared to nearby populations (HiP: compared with northern and southern KNP, Nairobi NP: compared with other East African populations; Table S3). Low frequencies were observed for most non-pooled alleles, which is not expected under genetic drift (Wilcoxon signed rank exact test; HiP:P = 0.0010; Nairobi NP: P = 0.0083, Holm-Šidák correction based on 12 possible focal populations: P = 0.095, Table S3). In HiP, low frequencies have earlier been attributed to negative selection (van Hooft et al., 2019).