Effects of genotypic diversity on growth and survival of an ecologically important coral

Abstract

Biodiversity loss threatens ecosystems worldwide, but we lack a thorough understanding of how genotypic diversity—as opposed to species diversity—impacts threatened foundation species, such as corals on tropical reefs. Corals of the genus Pocillopora have been important to natural reef recovery and human-assisted restoration in the Pacific. They consistently perform better in polycultures with other coral species than in monocultures, but whether genotypic diversity within a species enhances performance remains untested. We assessed whether increased genotypic diversity of the coral Pocillopora verrucosa (ie the number of genetically distinct source colonies per plot) improved growth, reduced tissue mortality, or increased resistance to invasion of competing seaweeds over 3.5 years in experimental plots on reefs in Mo‘orea, French Polynesia. We could detect no significant differences in any of these performance metrics among plots comprised of one, three, six, or nine genotypes. Moreover, when we quantified mitochondrial genetic divergence among genotypes within each plot using mean pairwise nucleotide distance, performance was unrelated to the degree of genetic differentiation represented. Thus, we detected no effect of genotypic diversity on Pocillopora performance. Pocillopora acclimates rapidly to variable physical and biological conditions, which may make genotypic diversity effects less critical for these corals.

Significance statement.

Theory and previous studies suggest that genotypic diversity can enhance performance of foundation species, but whether such positive effects occur among corals on tropical reefs remains poorly understood. To test this, we manipulated genotypic diversity of the common coral Pocillopora verrucosa in experimental field plots and monitored performance of these assemblages for 3.5 years. Although Pocillopora genotypes differed significantly in growth, we found no overall benefit of genotypic diversity for metrics of population performance—including growth, mortality, or resistance to macroalgal competitors. Our findings indicate that genotypic diversity per se does not enhance the performance of this foundation species, which may inform reef-restoration efforts.

Introduction

Biodiversity plays a vital role in maintaining the function and resilience of ecosystems in the Anthropocene (1). This is especially concerning for tropical reefs, which are rapidly losing reef-building corals that are the foundation species of these ecosystems (2). Recent findings demonstrate that greater coral species diversity can enhance coral growth, survival, and resistance to colonization by macroalgal competitors, suggesting that biodiversity loss may trigger negative feedbacks that further suppress reef biodiversity and ecosystem function (3, 4). However, as coral species are lost to disturbance, the functional and adaptive capacity of reefs may also depend on genotypic variation within the surviving species, which can influence coral growth, disease resistance, spawning strategies, and thermal tolerance (5). In other systems, genotypic diversity has produced effects that are comparable to, or exceed, those of species diversity (6); thus, this dimension of diversity may be an important determinant of coral performance as reefs continue to degrade (5).

A recent review of marine foundation species found that the effects of genotypic diversity on ecosystem function were most often neutral; however, most studies of corals were observational, correlative, and focused on genotypic diversity of symbiotic dinoflagellates living within corals (Symbiodinium) rather than the coral host (7). We are aware of only one study that explicitly tested this notion via manipulative experimentation; using the Caribbean coral Acropora cervicornis, investigators found no overall effect of genotypic diversity per se, despite differences among genotypes in growth and survival (8). Thus, our knowledge of whether coral genotypic diversity impacts population resilience and performance remains limited. This is especially true for biodiverse Indo-Pacific reefs, which unlike Caribbean reefs, still maintain high levels of genetic diversity (9).

In this study, we manipulated genotypic diversity of the coral Pocillopora verrucosa within experimental plots in the lagoon of Mo‘orea, French Polynesia, to test whether genotypic diversity affected population performance. Pocillopora verrucosa represents a genus that is an important reef builder in the Eastern Pacific (10), is commonly used in restoration (11), and is the primary contributor to the recurrent coral recovery in Mo‘orea following disturbances (12)—putatively through diversity-mediated mechanisms (12, 13). In field experiments, both P. verrucosa and congeners (Pocillopora damicornis) were found to grow better in polycultures than in monocultures (3, 4). If positive effects of diversity translate from the species to genotype level, then P. verrucosa might be a species especially likely to benefit. To test this hypothesis, we manipulated the genotypic diversity of P. verrucosa plots in the field for a period of 3.5 years and assessed coral growth, mortality, and colonization by macroalgal competitors.

Methods

Ethical statement

The experiment in French Polynesia was conducted in accordance with relevant permissions (Protocole d'Accueil 2020–2024) from the Délégation à la Recherche and the Haut-commissariat de la République en Polynésie Française (DTRT). Coral samples used for genetic analyses were exported from French Polynesia to the United States under CITES permit FR2098700061-E. Because corals are cnidarian invertebrates, no institutional animal care or ethics approval was required for the experimental work.

Sequencing and identification of P. verrucosa

During April through June of 2020, we tagged 72 colonies visually resembling P. verrucosa within the back reef lagoon on the north coast of Mo‘orea, French Polynesia (Supplementary Document S1, Fig. S1). Small (<1 cm³) tissue samples collected from each colony were placed in an RNA/DNA preservation buffer and frozen upon arrival at the field station. DNA was extracted from these samples by placing a small fraction of the sample (<2 mm³) in 200 µL of phosphate buffered saline and heated to 95 °C for 10 min. We used a mitochondrial marker—an open reading frame (mtORF) of unknown function previously shown to distinguish haplotypes of P. verrucosa using the FATP6.1 and RORF primers (14)—to assess each colony. Using this marker gene and primers, we successfully sequenced 64 of the tagged colonies. From these, we selected 24 colonies initially identified as P. verrucosa that were large enough to generate the numerous branch tips needed for use in our experimental treatments. Pocillopora verrucosa populations have low inbreeding coefficients (15) and physically separated individuals are never, or almost never, discovered to be clones (16, 17). Based on the absence of clones in previous studies and the broad areas over which our experimental colonies were located (Supplementary Document S1, Fig. S1), we considered each colony of P. verrucosa to be a separate genotype. We thus use the term genotypic diversity throughout this manuscript to refer to the number of distinct source colonies per plot in our manipulations.

During the 3.5-year duration of this experiment, molecular studies better clarified species boundaries within this complex genus (16, 18, 19), so we reevaluated our initial sequencing data at the end of the experiment to take advantage of these advances. This reevaluation detected errors in the processing of the initial data. These errors resulted in inclusion of one individual later recognized as Pocillopora tuahiniensis, a species described after the experiment was initiated (see Ref. (20)) and one individual of Pocillopora meandrina in our initial selection of colonies thought to be P. verrucosa (Supplementary Document S1, Fig. S2). These inappropriate colonies were included in one of our single-genotype replicates, three of our three genotype replicates, four of our six genotype replicates, and nine of our nine genotype replicates (n = 12 for each treatment type; see below).

Because previous experiments at this site and others had demonstrated significant positive effects of coral species diversity on growth of Pocillopora species (3, 4), the inclusion of these species could bias in favor of detecting and confusing positive effects of genotypic diversity with the effects of species diversity. However, if no effects occurred (as was the case), then conclusions would seem even more robust. That said, in addition to evaluating the full experiment as planned (ie including the misidentified colonies), we also statistically evaluated the effect of P. verrucosa genotype diversity alone by excluding all replicates that included inappropriate species and comparing only the one-, three-, and six-genotype treatments (there were too few remaining replicates to evaluate the nine-genotype treatment). Previous demonstrations of positive species diversity effects at this location found significant effects at the three species and six species treatments, with positive effects declining at nine species (4). Thus, if the effects of genotypic diversity were similar, we would expect to see positive effects in the three- or six-genotype treatments.

Study site and experimental setup

To assess the effect of genotypic diversity on coral performance, we conducted a manipulative field experiment from 2021 May 18 to 2024 November 11 in the back reef lagoon on the north coast of Mo‘orea, French Polynesia (17°28′37′′S, 149°50′21′′W). We manipulated genotypic diversity of P. verrucosa on the upper surface of 40 cm × 40 cm cement plots to compare performance of assemblages composed of one, three, six, or nine genotypes (n = 12 interspersed plots per treatment). Each plot was affixed to the substrate and elevated ∼30 cm on cinder blocks to prevent scouring by sand or unconsolidated rubble (for plot design, see Ref. (4)) and mimic the typical coral presence on raised bommies around our experimental site (Fig. 1A). The upper surface of each plot contained a six-by-six grid space in which we embedded 18 upturned soda bottle caps per plot within every other space. To create a standardized coral population on each plot, 6- to 8-cm-long branches of P. verrucosa were fragmented in situ from the 24 genotyped colonies discussed above and were individually epoxied (Z-Spar A-788 Splash Zone Epoxy) into the cut-off necks of inverted soda bottles that were individually labeled; corals could then be screwed into bottle caps embedded within the upper surface of their respective plot. Depending on treatment, plots contained either one (18 branches/genotype/plot), three (six branches/genotype/plot), six (three branches/genotype/plot), or nine (two branches/genotype/plot) genotypes, which were designated at random for each plot from the pool of 24 genotypes (18 corals per plot and 864 corals total).

Figure 1.

A) Experimental plots in the field at month 0 (left panel) and 30 (right panel) with either one, three, six, or nine genotypes of P. verrucosa. Photo credits: Cody S. Clements. Percent coral (B) growth and (C) tissue mortality (mean ± SE) at 5 months for each Pocillopora genotype used in plots containing either one, three, six, or nine genotypes. Genotype 10 and 75 were misidentified in our initial sequencing and are actually P. meandrina and P. tuahiniensis, respectively. D) Percent coral growth (mean ± SE) and E) tissue mortality at 5 months for plots with either one, three, six, or nine genotypes of Pocillopora.

At the initiation of the experiment (May 2021), corals and their uniquely numbered epoxy/bottle-top base were wet-weighed in the field using an electronic scale (OHAUS Scout Pro) enclosed within a plastic container mounted to a tripod holding it above the water. Before weighing, each coral was gently shaken 30 times to remove excess water and then weighed, immediately placed back in the water, and attached to their designated plot. This provided a wet mass starting value for each coral and its base.

At 5 months (October 2021), we assessed the percentage of tissue mortality and growth of individual corals in each plot. Each coral was visually examined from all sides, and the percentage of tissue mortality was estimated and assigned in 10% classes (0, 10, and 20%, etc., up to 100%). To assess coral growth, corals and their epoxy/bottle-top base were unscrewed, wet-weighed as described above, and immediately secured back on their treatment plot. Via subtraction, we could determine the percentage growth of corals for each plot. Twenty-four to 48 hours before this weighing, each coral's epoxy/bottle-top base was brushed clean of fouling organisms. At month 36 (May 2024), outplants on most plots had grown too large to be unscrewed and reattached successfully, so we photographed each plot and analyzed area of live coral cover using ImageJ (version 1.8.0_121). This was repeated at 42 months (November 2024), but because this was the endpoint of the experiment, corals were also detached (making sure any fractured branches were retained with their respective individual) from their plots and again assessed for percentage tissue mortality and wet-weighed as described above.

At 42 months, we also assessed macroalgal colonization by collecting all upright macroalgae from each plot, separating to genus, and wet weighing after removing excess water using a salad spinner (15 revolutions per sample). All corals were outplanted back to the reef.

Statistical analyses

We used permutation-based, linear mixed-effects (LME) models in the R package predictmeans (21) to compare differences in the percentage mass change and tissue mortality of corals among treatments and genotypes at 5 months. In each analysis, genotypic diversity (eg one, three, six, or nine genotype plots) and coral genotype were treated as fixed factors, with individual replicate plots treated as a random effect nested within treatments. Due to indiscernible labels preventing identification of individual replicates within a plot at 42 months, all response variables were analyzed at the plot level. Percentage mass change was calculated as the sum of coral mass per plot at 42 months relative to initial (0 month) values. Percent tissue mortality, live coral area (measured at 36 and 42 months), and macroalgal mass (at 42 months) were also averaged or summed per plot. For each response variable, we first tested for normality (Shapiro–Wilk test on residuals) and homogeneity of variance (Levene's test), and when assumptions of normality and homoscedasticity were met, we performed parametric one-way ANOVA using JMP (version 18.2.0). However, to ensure methodological consistency and robustness across all metrics—especially for tissue mortality, algal mass, and 36-month live coral area, which violated assumptions—we conducted ANOVA using a permutation approach (5,000 permutations) in the R package coin (22). Permutation-based results are presented for all metrics; parametric ANOVA results are included for comparison when assumptions were met. In addition, to explore potential nonlinear effects of genotypic diversity on coral performance, we treated genotypic diversity as a continuous variable and fit quadratic regressions between genotypic diversity (treated as a continuous variable) and each response variable at 42 months using the lm function in R (23). Significant curvilinear relationships were interpreted as indicative of nonmonotonic diversity effects.

To further evaluate whether variation in genetic composition among plots influenced coral performance, we quantified the degree of nucleotide divergence among genotypes within each plot using the mitochondrial marker described above. Aligned mtORF sequences (702 bp) from the 24 source colonies were used to calculate pairwise nucleotide distances among all genotypes as uncorrected p-distances (proportion of nucleotide sites differing between sequences). For each plot, we identified the set of unique genotypes present and calculated the mean pairwise nucleotide distance (MPD) among genotypes on that plot. Plots containing a single genotype were assigned an MPD value of zero. This metric provided a continuous estimate of genetic divergence and allowed plots containing the same number of genotypes to differ in the degree of nucleotide differentiation represented. We then evaluated whether plot-level coral performance metrics at 5 months (percentage mass change and percent tissue mortality), 36 months (live coral area), and 42 months (percentage mass change, percent tissue mortality, live coral area, and macroalgal mass) were related to MPD using linear regression models. For each response variable, we additionally tested for nonlinear relationships by including a quadratic MPD term. To provide inference robust to distributional assumptions, we conducted permutation-based tests (5,000 resamples) using the R package coin; permutation-based P-values are reported in the results. Model fit is summarized using R² values from the linear models.

Results

When all replicates were included in the analysis at 5 months (ie including those containing a species other than P. verrucosa), percentage coral growth significantly differed among coral genotypes (LME, F_{(23, 724)} = 4.598, P < 0.001)—ranging from +32.0 to +71.1% (Fig. 1B)—but not as a function of genotypic diversity (LME, F_{(3, 41)} = 0.305, P = 0.831; Fig. 1D). Growth of the two species (P. tuahiniensis and P. meandrina) included in error did not drive the difference in growth rate among genotypes, as growth values of these two individuals fell toward the center of the distribution (genotype 10 and 75, respectively, in Fig. 1B), indicating that variance in growth of P. verrucosa genotypes was considerable and bracketed that of these other species. Across all individual corals at 5 months, growth was negatively associated with tissue mortality (Spearman's ρ = −0.451, P < 0.001), indicating that faster-growing outplants exhibited lower levels of tissue loss. However, percentage tissue mortality at 5 months neither significantly differed among genotypes (LME, F_{(23, 818)} = 1.165, P = 0.279; Fig. 1C) nor as a function of genotypic diversity (LME, F_{(3, 44)} = 0.333, P = 0.800; Fig. 1E). Similarly, at 42 months, percentage coral growth (permutation ANOVA, χ² = 5.63, P = 0.124; parametric ANOVA, F_{(3, 44)} = 2.00, P = 0.128; Fig. 2A) and coral tissue mortality (permutation ANOVA, χ² = 3.78, P = 0.290; Fig. 2B) were statistically indistinguishable among plots of differing genotypic diversity. Area of live coral cover also did not differ among genotypic diversity treatments at 36 (permutation ANOVA, χ² = 0.40, P = 0.943) or 42 months (permutation ANOVA, χ² = 3.92, P = 0.278; parametric ANOVA, F_{(3, 44)} = 1.33, P = 0.276; Fig. 2C).

Figure 2.

A) Percent coral growth (mean ± SE), B) percent tissue mortality, C) area of live coral, and D) mass of algae at 42 months for plots with either one, three, six, or nine genotypes of Pocillopora. P-values are from permutation-based ANOVA (5,000 permutations).

Colonization of algae at 42 months did not differ significantly among diversity treatments (permutation ANOVA, χ² = 5.35, P = 0.137; parametric ANOVA, F_{(3, 44)} = 1.89, P = 0.146; Fig. 2D). However, post hoc regression analyses of the 42-month patterns revealed significant or marginally significant quadratic trends for coral growth (P = 0.028, adjusted R² = 0.067) and algal mass (P = 0.022, adjusted R² = 0.073), and weaker hump-shaped trends for coral cover (P = 0.077, adjusted R² = 0.032) and tissue mortality (P = 0.084, adjusted R² = 0.028), suggesting a potential weak peak in performance at intermediate genotypic diversity.

To evaluate whether genetic divergence among genotypes—rather than genotype richness per se—explained variation in coral performance, we quantified MPD among genotypes within each plot and analyzed MPD as a continuous predictor of performance metrics. MPD was not significantly associated with percentage growth at 5 months (linear model, F_{(1, 46)} = 0.64, P = 0.426, R² = 0.014; permutation test P = 0.428) or at 42 months (F_{(1, 46)} = 0.82, P = 0.370, R² = 0.018; permutation test P = 0.363). Similarly, MPD did not explain variation in tissue mortality at 5 months (F_{(1, 46)} = 0.46, P = 0.502, R² = 0.010; permutation test P = 0.498) or 42 months (F_{(1, 46)} = 2.12, P = 0.152, R² = 0.044; permutation test P = 0.157). Live coral area was not related to MPD at 36 months (F_{(1, 46)} = 1.59, P = 0.214, R² = 0.033; permutation test P = 0.212) or 42 months (F_{(1, 46)} = 1.47, P = 0.232, R² = 0.031; permutation test P = 0.243). Macroalgal mass at 42 months showed a weak negative trend with MPD (F_(1,46) = 3.17, P = 0.082, R² = 0.064; permutation test P = 0.079), but this relationship was not statistically significant. Inclusion of quadratic MPD terms did not improve model fit for any response variable (all P > 0.20). Thus, neither the number of genotypes present nor the degree of genetic divergence among them explained variation in coral performance.

When the above contrasts were reanalyzed excluding all replicates containing the misidentified species (ie removing all replicates containing individuals of colony 10 or 75) and without the nine-genotype treatment where remaining n = 3, none of the metrics assessed differed as a function of genotypic richness per se or MPD (see Supplementary Document S1, Results S1).

Discussion

As tropical reefs continue to degrade globally, there is an urgent need to determine whether and how coral biodiversity at different levels of organization impacts the function and stability of coral populations and communities (3, 4, 24). This includes potential effects of coral genotypic diversity, which may become increasingly important as reefs lose species diversity (2) and restoration efforts focus on a subset of coral species that are frequently outplanted as monocultures (∼28% of documented restoration projects; see Ref. (11)).

In this study, we manipulated genotypic diversity of P. verrucosa corals in the field for a duration of 3.5 years to test how this intraspecific dimension of diversity affected performance of experimental coral assemblages. Our experiment mistakenly included two colonies of other Pocillopora species, but these species performed (in terms of growth) much like P. verrucosa (Fig. 1B), and inclusion or exclusion of replicates containing these additional species produced the same general findings. Although Pocillopora genotypes differed significantly in growth (indicating variable responses to the field conditions experienced), we found no overall effect of genotypic diversity per se for multiple metrics of coral performance. This was also the case when we assessed performance metrics based on mean pairwise mitochondrial nucleotide divergence among genotypes within our plots—rather than genotype richness per se. These findings suggest that ecological benefits that are common at higher-level measures of diversity—such as selection or complementarity effects associated with species diversity (3, 4)—may be limited for this important reef-building coral.

Because genotypic diversity has been demonstrated to have positive effects in previous experiments (6, 25, 26), and because one would assume that a mix of genotypes and physiologies would be more likely to be resilient in the face of various stresses (27, 28, 29) we were surprised at the lack of a detectable effect in our 3.5-year experiment. This absence of an effect was not due to physiological uniformity across genotypes because there was significant variance among genotypes in net growth (Fig. 1B), with the most rapidly growing genotype exceeding the least rapidly growing by ∼120% at 5 months. Substantial variability in growth and tissue mortality was observed among plots within treatments across time points and may reflect differences among genotypes in their responses to environmental conditions experienced during the experiment (eg thermal stress events). Such among-plot variation could reduce the detectability of diversity-level effects when genotypic diversity does not strongly influence mean performance.

Our treatments were assembled using genetically distinct source colonies, encompassing a range of intra- and interhaplotype variation and even introduced a low level of heterospecificity into some higher diversity plots. The absence of any performance benefit—despite this mixture—suggests that the functional trait variation among colonies was too limited to produce biodiversity effects under the conditions of our study. This contrasts with previous findings in which species-level diversity enhanced Pocillopora performance (3, 4) and indicates that positive biodiversity effects observed among species do not always translate to finer taxonomic scales. Our study constitutes a first step toward rigorously testing for genotypic diversity effects among Indo-Pacific corals, which are generally more genetically diverse (9) and have exhibited greater resilience following disturbance than have Caribbean corals (30).

One might expect the advantages of genotypic diversity to be most strongly expressed during periods of stress or disturbance (27, 28, 29), which were recurrent throughout our study. Corals in our manipulations were exposed to several thermal stress events (31; C. Clements personal observation) during the 3.5-year experiment but did not appear to benefit from increased genotype diversity. However, the absence of major stressors (eg large-scale bleaching, predator outbreaks) could have reduced the selective pressure under which genotypic diversity might confer advantages (27, 28), such as differential survival or growth among genotypes. For example, following a marine heatwave and widespread coral bleaching in Mo‘orea in 2019, Burgess et al. (13) found that 86% of dead pocilloporids belonged to a single haplotype from a cryptic lineage—despite sampling 12 haplogroups spanning at least five species, including P. verrucosa. It is also possible that our manipulations did not encompass a sufficiently broad range of genotypic diversity (7), or that genotypic diversity offers limited ecological benefit for P. verrucosa relative to other pocilloporid species more strongly impacted by disturbance. Reciprocal transplants of P. verrucosa between variable habitats indicate that it does acclimate and perform well even when depth, light, competition, and predation are varied experimentally in the field (32). Given the evidence suggesting Pocillopora diversity may promote reef resistance and resilience to disturbance (12), future manipulative experiments exploring how intrageneric diversity influences coral reef ecosystem function would be useful.

Although our primary analyses did not reveal significant differences among treatments, exploratory regression analyses suggested potential, but weak (R² = ∼0.03–0.07), nonlinear trends in coral performance. For example, coral growth and macroalgal colonization both exhibited significant hump-shaped relationships at 42 months, with peak coral growth and minimal macroalgal accumulation occurring at intermediate levels of genotypic diversity. Similar curvilinear trends were observed for mortality and live coral cover but were not statistically significant. Hump-shaped relationships are often observed between species diversity and ecological function (33), but those involving genotypic diversity are less commonly documented and limited to laboratory settings (34). While the explanatory power of our models was modest, the trends observed in this study are reminiscent of previous findings from this system where overall coral community performance (ie coral growth and mortality)—including growth of pocilloporids (P. verrucosa and P. damicornis) specifically—was enhanced at intermediate species diversity (4). Our current findings suggest that intermediate genotypic diversity might also confer subtle functional advantages, but if so, the effect appears weak.

Increased diversity of Pocillopora genotypes within our manipulations did not translate into enhanced coral performance, but our experiment primarily manipulated genotypic diversity of a single Pocillopora species within fixed-area plots, controlling for coral density and spatial arrangement. Despite these and the other potential limitations outlined above, our experiment—which ran for 3.5 years, involved 864 corals, was conducted under natural field conditions that included stress events such as heat-induced bleaching in Mo‘orea, and was carried out in the same location and with the same methods that previously detected significant effects of coral species diversity (4)—found little or no detectable effect of genotypic diversity. Previous work suggests that P. verrucosa may possess a broad physiological range or ability to acclimatize (32, 35) and thus may be less dependent on genotypic diversity than other corals. This could also help explain the disproportionate role Pocillopora has played in the rapid recovery of coral cover on Mo‘orea following a host of severe disturbances in the past (12). These insights could also inform restoration strategies (eg the use of fragments from fewer, higher performance donor colonies) on increasingly degraded reefs.

Supplementary material

Supplementary material is available at PNAS Nexus online.

Competing interests

The authors declare no competing interest.

Funding

This study was funded by the U.S. National Science Foundation Division of Ocean Sciences (grant no. OCE 1947522), the Teasley Endowment to the Georgia Institute of Technology, and the Anna and Harry Teasley Gift Fund. This work represents a contribution of the Moʻorea Coral Reef (MCR) LTER Site supported by the U.S. National Science Foundation (grant no. OCE 16-37396).

Author contributions

Cody S Clements (Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Writing—original draft, Writing—review & editing), Zoe A Pratte (Data curation, Formal analysis, Methodology, Writing—review & editing), Frank J Stewart (Funding acquisition, Methodology, Project administration, Resources, Supervision, Writing—review & editing), and Mark E. Hay (Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Writing—original draft, Writing—review & editing)

Data availability

Experimental datasets used in this study are available from the BCO-DMO data system at https://www.bco-dmo.org/project/837802. Sequence data used for genotyping are available in the National Center for Biotechnology Information database under BioProject PRJNA1339017 at https://www.ncbi.nlm.nih.gov/bioproject/PRJNA1339017.