Genomic Signatures of Local Adaptation in Clam Shrimp (Eulimnadia texana) from Natural Vernal Pools

Abstract

Vernal pools are unique in their isolation and the strong selection acting on their resident species. Vernal pool clam shrimp (Eulimnadia texana) are a promising model due to ease of culturing, short generation time, small genomes, and obligate desiccated diapaused eggs. Clam shrimp are also androdioecious (sexes include males and hermaphrodites), and here we use population-scaled recombination rates to support the hypothesis that the heterogametic sex is recombination free in these shrimp. We collected short-read sequence data from pooled samples from different vernal pools to gain insights into local adaptation. We identify genomic regions in which some populations have allele frequencies that differ significantly from the metapopulation. BayPass (Gautier M. 2015. Genome-wide scan for adaptive divergence and association with population-specific covariates. Genetics 201(4):1555-1579.) detected 19 such genomic regions showing an excess of population subdivision. These regions on average are 550 bp in size and had 2.5 genes within 5 kb of them. Genes located near these regions are involved in Malpighian tubule function and osmoregulation, an essential function in vernal pools. It is likely that salinity profiles vary between pools and over time, and variants at these genes are adapted to local salinity conditions.

Keywords: adaptation; ecological genetics; genomics/proteomics; landscape genetics; natural selection and contemporary evolution; population genetics: empirical.

Fig. 1.

A map of the sampling locations for the 11 study populations and a maximum likelihood tree generated by TreeMix depicting the relatedness of the populations based on genome-wide allele frequency estimates. All populations were taken as soil samples from field sites in New Mexico and Arizona. The “EE_Ancestor” strain is a laboratory strain descended from WAL.

Fig. 2.

Isolation by distance. This plot depicts the pairwise F_ST of all natural populations versus the geographical distance between them. The blue line is a linear regression of F_ST on distance. The red line is the same regression, but excluding the more distant WAL population from the analysis. The slopes of the two regressions are nearly identical.

Fig. 3.

Manhattan plots comparing F_ST and X^TX (A and C), demonstrating that peaks of differentiation are identified by both methods but are more clearly resolved using X^TX. Locations of peaks identified by a hidden Markov model tuned to identify outliers of population differentiation (B and D). Horizontal lines indicate the threshold for the 0.01% most significant loci.

Fig. 4.

Manhattan plots of single SNP X^TX values indicating excess differentiation among the 11 populations for the regions 6, 12, and 13. The plots indicate that the signal is highly localized, often suggesting a single gene (polycystin-1 for locus 6, Dh44 for locus 12, and nephrin for locus 13). The red rectangle in each plot indicates the region identified as significant by the hidden Markov model. The “R” indicators in the titles indicate the region number. “Freq” refers to the per-population allele frequency at each locus. “Cov” indicates the sequencing depth per population, after normalizing each population-specific coverage to the genome-wide average coverage calculated from 550-bp tiled windows. A coverage of 1 here indicates average coverage, 2 indicates double, etc. The lower plot is a zoomed figure indicating just the region identified as differentiated by the hidden Markov model.

Fig. 5.

Manhattan plots of single SNP Bayes factors for different environmental variables. All plots show the 11-population Bayes factor associated with the given environmental variable, with environmental variables indicated. Horizontal lines indicate the threshold for the 0.01% most significant loci.

Fig. 6.

Downsampling of SNPs demonstrates that reduced marker density dramatically reduces power to detect differentiated regions.