Differential Selection on Caste-Associated Genes in a Subterranean Termite

Abstract

Analyzing the information-rich content of RNA can help uncover genetic events associated with social insect castes or other social polymorphisms. Here, we exploit a series of cDNA libraries previously derived from whole-body tissue of different castes as well as from three behaviourally distinct populations of the Eastern subterranean termite Reticulitermes flavipes. We found that the number (~0.5 M) of single nucleotide variants (SNVs) was roughly equal between nymph, worker and soldier caste libraries, but dN/dS (ratio of nonsynonymous to synonymous substitutions) analysis suggested that some of these variants confer a caste-specific advantage. Specifically, the dN/dS ratio was high (~4.3) for genes expressed in the defensively specialized soldier caste, relative to genes expressed by other castes (~1.7−1.8) and regardless of the North American population (Toronto, Raleigh, Boston) from which the castes were sampled. The populations, meanwhile, did show a large difference in SNV count but not in the manner expected from known demographic and behavioural differences; the highly invasive unicolonial population from Toronto was not the least diverse and did not show any other unique substitution patterns, suggesting any past bottleneck associated with invasion or with current unicoloniality has become obscured at the RNA level. Our study raises two important hypotheses relevant to termite sociobiology. First, the positive selection (dN/dS > 1) inferred for soldier-biased genes is presumably indirect and of the type mediated through kin selection, and second, the behavioural changes that accompany some social insect urban invasions (i.e., ‘unicoloniality’) may be detached from the loss-of-diversity expected from invasion bottlenecks.

Keywords: caste differentiation; kin selection; social insect invasions; urban entomology.

Figure 1

A simplified representation of caste developmental pathway in Reticulitermes flavipes [18,27]. Larvae differentiate into non-reproductive (worker and soldier) or reproductive (nymphs, primary queens and kings) castes. Workers can further differentiate into soldiers. These two points of caste differentiation or ‘switches’ sensu [28] are shown by open circles.

Figure 2

A graphical workflow depicting the major steps that we undertook in data processing. Oval shapes represent input/output, and the rectangular shapes represent the tools/processing steps. To begin, we accessed the nine NCBI-SRA library files (FASTQ format) and the corresponding reference assembly (FASTA format). From there, we used these data files in combination to generate the four principal analyses shown at the bottom of the flow chart.

Figure 3

Read depth of high-quality (depth > 10) SNVs extracted from variant call files for three termite castes (worker, soldier, nymph) sampled from three populations (Raleigh, Boston, Toronto). The overall profiles of read depth coverage are similar between libraries, suggesting consistently high qualities.

Figure 4

Counts of SNVs by caste and population. Termite castes are relatively uniform in SNV diversity. A pair-wise post-hoc analysis indicated that workers, soldiers and nymphs all harbour nearly identical levels of diversity in their transcribed sequences (Pairwise post-hoc Dunn Test, p > 0.4 in all cases). Among populations, however, termites collected from Raleigh were more variable than were those collected from Boston (Dunn tests; p = 0.0109), with the Toronto population being intermediate (Dunn test; p > 0.08 in both cases). Error bars are 95% CI. Different letters above boxes indicate a significant difference. Note that for the Toronto termite population, the three caste libraries used had a very similar SNV counts, and the estimate of variance is miniscule.

Figure 5

Heatmap of SBS-96 mutation counts for each of the nine samples. All 96 possible mutations are plotted (format available at https://cancer.sanger.ac.uk/cosmic/signatures/SBS, (accessed on 1 November 2021). Most plots show similar biases in mutation types and counts—for example, an excess of transitions across all samples. An SBS-96 plot with purine as a lead base is very similar qualitatively (not shown), consistent with similar mutational mechanisms across populations and castes. There are nonetheless slight differences in the actual numbers of SBSs between caste and population. Additionally, the scale for the Boston population has a smaller range than for Toronto and Raleigh, which is consistent with fewer variants found there (Figure 4B).

Figure 6

The d_N/d_S ratios of whole transcriptome and a subset of n = 230 caste-specific transcripts are grouped by caste or population. For each graph, mean and standard deviation are displayed as a black circle and grey line, respectively. We observed no transcriptome-wide differences in mean d_N/d_S between castes (shown in (A)). However, when caste-specific transcripts were tested separately (B), we found d_N/d_S of soldier-expressed genes to be significantly higher than d_N/d_S for genes uniquely expressed in nymphs (mean difference = 2.45, Tukey’s post hoc test p = 0.001) or workers (mean difference = 2.599, p = 0.01). We found no difference between populations in mean d_N/d_S at the whole-transcriptome (C) or caste-specific transcriptome (D) level.