Genome-Wide Scan and Test of Candidate Genes in the Snail Biomphalaria glabrata Reveal New Locus Influencing Resistance to Schistosoma mansoni

Abstract

Background: New strategies to combat the global scourge of schistosomiasis may be revealed by increased understanding of the mechanisms by which the obligate snail host can resist the schistosome parasite. However, few molecular markers linked to resistance have been identified and characterized in snails.

Methodology/principal findings: Here we test six independent genetic loci for their influence on resistance to Schistosoma mansoni strain PR1 in the 13-16-R1 strain of the snail Biomphalaria glabrata. We first identify a genomic region, RADres, showing the highest differentiation between susceptible and resistant inbred lines among 1611 informative restriction-site associated DNA (RAD) markers, and show that it significantly influences resistance in an independent set of 439 outbred snails. The additive effect of each RADres resistance allele is 2-fold, similar to that of the previously identified resistance gene sod1. The data fit a model in which both loci contribute independently and additively to resistance, such that the odds of infection in homozygotes for the resistance alleles at both loci (13% infected) is 16-fold lower than the odds of infection in snails without any resistance alleles (70% infected). Genome-wide linkage disequilibrium is high, with both sod1 and RADres residing on haplotype blocks >2 Mb, and with other markers in each block also showing significant effects on resistance; thus the causal genes within these blocks remain to be demonstrated. Other candidate loci had no effect on resistance, including the Guadeloupe Resistance Complex and three genes (aif, infPhox, and prx1) with immunological roles and expression patterns tied to resistance, which must therefore be trans-regulated.

Conclusions/significance: The loci RADres and sod1 both have strong effects on resistance to S. mansoni. Future approaches to control schistosomiasis may benefit from further efforts to characterize and harness this natural genetic variation.

Fig 1. Distribution of allele frequencies among RAD sites.

We characterized 1611 informative RAD sites in 19 inbred lines by minor allele count (“MAC”) (maximum of 19; i.e. 50% allele frequency) and difference in MAC between susceptible (N = 10) and resistant (N = 9) lines (“MAC difference”, theoretical maximum = 18; i.e. 9 resistant lines fixed for one allele, 10 susceptible lines fixed for another). Circle sizes are proportional to the number of RAD sites showing each pattern. The cumulative percentage of RAD sites, starting with the highest observed MAC difference, is shown on the righthand y-axis. The highest MAC difference was observed for 10 RAD sites in perfect mutual LD, with a MAC of 13 and a MAC difference of 13, which we defined as the RADres region and examined further (red arrow; encompasses scaffolds of subsequently examined markers RADres1 and RADres2). The one remaining RAD site with an equivalent MAC difference was also in high, but not perfect, LD with RADres (pink arrow; scaffold not examined further). The sod1 haplotype block had a MAC difference of 5, which was higher than average but not an outlier (blue arrow).

Fig 2. Influence of RADres1 on susceptibility in 50 inbred lines.

Lines are colored based on genotype (purple = EE, orange = FF) and binned based on susceptibility in windows of 20%. EE lines are more resistant on average. The most striking difference between genotypes occurs in highly resistant lines (susceptibility < 20%), which include no FF lines and 15 EE lines. One additional line (not depicted in figure) was heterozygous with 57% susceptibility.

Fig 3. Extent of linkage disequilibrium (LD) in 13-16-R1.

Among 19 inbred lines, RAD SNPs with identical genotype patterns (LD of 1) were grouped into haplotype blocks. The number of RAD sites represented by each haplotype block is shown. Most (82%) RAD sites show perfect LD with at least one other RAD site, indicating high genome-wide LD. The RADres block shows a typical extent of LD, with 10 RAD sites in perfect mutual LD. RAD sites with SNPs differentiating sod1 B haplotype from the other sod1 haplotypes form the single largest haplotype block in the genome (46 RAD sites). Thus, sod1 appears to be in an unusually large haplotype block of perfect LD.

Fig 4. Effects of single alleles in outbred snails.

We measured the effects of individual alleles using separate simple logistic regression analyses for each allele at all loci. Each allele is plotted according to its multiplicative effect on the odds of susceptibility (x-axis) and the p-value for its effect (not corrected for multiple tests; y-axis). Standard errors are shown with horizontal lines flanking each point. Alleles are colored based on locus. Only alleles at RADres1, RADres2, sod1, and bmplys are significant (uncorrected threshold of 0.05 indicated by dotted line, threshold after Bonferroni correction of 0.002 indicated by dashed line). Alleles are labeled (A-L, P, W, and T) as in Table 2. For non-significant loci, only the most common allele is shown.

Fig 5. Independent influence of RADres1 and sod1 genotypes on resistance.

Among 439 snails, both the E allele at RADres1 and the B allele at sod1 are positively correlated with resistance. For each locus (red genotypes = RADres1; blue genotypes = sod1), all three genotypes are significantly different from each other, indicating additive rather than dominant effects. Standard errors of proportions are indicated by vertical bars.

Fig 6. Joint influence of RADres1 and sod1 genotypes on resistance.

Genotype combinations are indicated along the x and y axes. Empirical resistance values are plotted as black circles along the z axis. Standard errors of proportions for resistance at each genotype combination are shown with a vertical yellow line. Predicted values from an additive 2-locus multiple regression model with no dominance or epistasis are indicated with green squares. The data are consistent with this simple model, although minor non-additive effects may be responsible for small nonsignificant differences between predicted and empirical values.