An experimental test of temperature‐dependent selection on mitochondrial haplotypes in Callosobruchus maculatus seed beetles

Abstract Mitochondrial DNA (mtDNA) consists of few but vital maternally inherited genes that interact closely with nuclear genes to produce cellular energy. How important mtDNA polymorphism is for adaptation is still unclear. The assumption in population genetic studies is often that segregating mtDNA variation is selectively neutral. This contrasts with empirical observations of mtDNA haplotypes affecting fitness‐related traits and thermal sensitivity, and latitudinal clines in mtDNA haplotype frequencies. Here, we experimentally test whether ambient temperature affects selection on mtDNA variation, and whether such thermal effects are influenced by intergenomic epistasis due to interactions between mitochondrial and nuclear genes, using replicated experimental evolution in Callosobruchus maculatus seed beetle populations seeded with a mixture of different mtDNA haplotypes. We also test for sex‐specific consequences of mtDNA evolution on reproductive success, given that mtDNA mutations can have sexually antagonistic fitness effects. Our results demonstrate natural selection on mtDNA haplotypes, with some support for thermal environment influencing mtDNA evolution through mitonuclear epistasis. The changes in male and female reproductive fitness were both aligned with changes in mtDNA haplotype frequencies, suggesting that natural selection on mtDNA is sexually concordant in stressful thermal environments. We discuss the implications of our findings for the evolution of mtDNA.


DNA extraction protocol
For each sample, we randomly selected 100 adult male beetles and extracted high-quality DNA using a salt-ethanol precipitation protocol. Beetles were first gently macerated and placed in preparation buffer (100 mM NaCl, 10 mM Tris-HCl, pH = 8.0, 0.5% SDS) together with proteinase K, vortexed and incubated at 50 o C overnight.
Samples were then frozen overnight. To precipitate DNA, we added saturated NaCl several times before adding 95% ethanol, and then spun the DNA into a pellet. The DNA pellet was suspended in TE buffer (pH = 7.6). DNA quality and quantity were assessed using NanoDrop, Qubit and Bioanalyzer.

Molecular divergence across haplotypes
Using the MutPred2 (Li et al. 2009) algorithm we assessed putative pathogenicity of the nonsynonymous substitutions found between the three mtDNA haplotypes. This analysis identified three SNPs across two genes as potentially pathogenic: in cob a change from V to I (in BRA, at a position 296) alters the transmembrane protein (probability score = 0.25, P < 0.0001, affected ELM motifs: ELME000333, 000336), and a change from N to S (in YEM, at a position 173. Affected ELM and PROSITE motifs: ELME000053, ELME000070, ELME000085, ELME000239, PS00001) causes altered ordered interface (prob. score = 0.25, P = 0.02), loss of N-linked glycosylation (prob. score = 0.17, P = 0.0053), altered transmembrane protein (prob. score = 0.02, P = 0.0083) and a loss of GPI-anchor amidation. In the gene nad1 a substitution from F to S (in YEM, at the position 310, ELME000053, ELME000328) changes the ordered interface of the protein (prob. score = 0.24, P = 0.04).
Four SNPs were found in tRNAs, of which three were unique to YEM (Met, Cys, Asn) and one to CAL haplotype (Thr). There was also a deletion of a single base pair in tRNA(Asp) in the BRA haplotype. Five SNPs were detected in both long and short rRNAs (long: 3 SNPs unique to CAL and 1 in BRA and YEM; short: 3 SNPs unique to BRA and 2 to YEM). In addition to genes, the three haplotypes also differ in the two non-coding long intergenic regions recently identified in the mitogenome of seed beetles (see Sayadi et al. 2017).

A comparison of within and between population haplotype divergence
A previous global analysis of C. maculatus mtDNA haplotype divergence (Kébé et al. 2017) showed that the total mitochondrial variance is dominated by withinpopulation variance (38.6%) and among-populations-within-region (48.7%) variance, with only a small fraction of the total genetic variance accounted by among-continent variance (12.7%). Nevertheless, to assay sympatric haplotype diversity in C. maculatus, we sequenced (forward and reverse) a 625 bp segment of COI from in total 43 isofemale lines from west Africa. Of these, 41 derive from a focal single population collected in a crop field in Lome (Togo) and 2 from a single population named Ofuya (Nigeria). For a description of DNA extraction, primers, PCR-conditions and sequencing, we refer in full to Tuda et al. (2006). In addition, we added sequence data for the same segment of COI from four different reference populations to our analyses. These were the three haplotypes used in our experimental evolution lines (i.e., California, Yemen, Brazil) as well as the South India (SI) genome reference population (Sayadi et al. 2017(Sayadi et al. , 2019. Overall, the west African sample showed a very high level of mtDNA haplotype diversity. There were 42 variables sites and 22 out of the 43 haplotypes were unique. Synonymous within-population nucleotide diversity was pS = 0.041, which is considerably higher than that between any of the three haplotypes used in our experiment. In addition, there were two non-synonymous polymorphic substitutions in this region (pN = 0.001). One involves a substitution from isoleucine to valine, shown by 8 of our 20 unique haplotypes from the Lome population. One out of our four reference sequences (California) also carries this particular non-synonymous substitution and the 8 Lome haplotypes carrying this substitution cluster together with the California haplotype ( Figure SI1). The other non-synonymous substitution is from methionine to isoleucine and is carried by a single of our 20 haplotypes from Lome.
A close analysis of this region of COI shows that west African within-population mtDNA variation in C. maculatus is comparable to between-population variation (see also Kébé et al. 2017). For example, Sayadi et al. (2017) reported a betweenpopulation pS = 0.062 for COI and pS ranged from 0.014 to 0.071 for other protein coding genes of the mitogenome. The relationship among the haplotypes analyzed here is visualized in Figure SI1. These analyses show that two major haplotype groups occur sympatrically in both the Lome and the Ofuya population.
We conclude that although two of the three particular mtDNA haplotypes used in our experimental evolution lines were distinct from those occurring in the Lome population ( Figure SI1), the three haplotypes are not more divergent than are haplotypes occurring sympatrically within the Lome population.  Tables   Table S2. Tests of sex-specific interaction effects between the line type (i.e. mtDNA variation or not), nuclear genetic background, and generation (i.e. 3 or 33 generations in warmer thermal environment), on the lifetime reproductive success.      CAL or YEM) where placed together in exactly equal proportions (1/3 each) to create new lines with mtDNA variation ("mtDNA-mix"). These lines where then assigned to either colder or warmer thermal conditions. Each mtDNA-mix line was replicated twice using separate introgression lines (i.e. with mtDNA from separate "mitochondrial Eves"), and each of these were split into two replicates per temperature, resulting in a total of eight mtDNA-mix lines per nuclear genetic background. For the reproductive success assays we also used control lines that included the original combination of mtDNA haplotype and nuclear genetic background, created through the same introgression scheme with the same replication.  Figure S3. Observed mtDNA haplotype frequency changes in all 23 experimental evolution lines kept under cold (23°C; blue symbols) or hot (35°C; red symbols) conditions. All lines were started with a frequency of 0.33 for each of the three haplotypes at generation 1 (indicated by a yellow circle). The CAL haplotype is denoted by triangles, the YEM by squares and the BRA by circles. Note that haplotype frequencies were estimated at generation 23 (cold) and 36 (warm) but generations are slightly off-set in the figure, to avoid overlap and facilitate visual interpretation. Figure S4. The relative lifetime reproductive success (mean LRS +/-SE, vertical and horizontal lines) of the 'mtDNA-mix' lines harboring mtDNA variation for females and males at the generations 3 and 33. The three different nuclear genetic backgrounds are highlighted with separate symbols. The mean LRS data is plotted relative to the fittest CAL mtDNA haplotype (on its native background). This figure shows how the relative reproductive success of the lines remains sexually concordant across the time points where we verified a significant increase in frequency of the CAL. The horizontal and vertical dashed lines indicate equal fitness in the sexes.  Figure S5. Neighbourhood-Joining based haplotype network of the three haplotypes (closed circles) based on data of all mitochondrial protein coding genes, tRNAs and rRNAs. The number of single nucleotide substitutions is highlighted in parenthesis.