Selective breeding produces increased occurrence of alleles with intermediate frequencies on which natural selection can act
Selective breeding is one genetic intervention strategy that may quickly increase adaptive genetic variation in corals, as such facilitating adaptation to increasing sea surface temperatures (Chan, Hoffmann, & van Oppen, 2019; van Oppen et al., 2015). This approach has been used across a range of aquaculture and mariculture species to improve a number of commercially important traits. Commercially important species like oysters and mussels share similar life-history attributes with corals, including the production of numerous, potentially low-quality gametes, and high levels of genetic diversity (e.g. R-strategist) (Ellegren & Galtier, 2016). In those systems, only a few generations of selective breeding has resulted in 30% higher growth under elevated pCO2 conditions compared to wild populations (Parker et al., 2012) and 50% higher growth in redclaw crayfish (Stevenson, Jerry, & Owens, 2013) without significantly eroding genetic diversity compared to wild populations after pooling breeding combinations (O’Connor, Dove, & Knibb, 2016). Selective breeding over only one generation in corals has shown that significant increases in heat tolerance of larvae (Dixon et al., 2015) and juveniles (Quigley et al., 2020) are possible using these methods. Lessons learnt here may therefore provide insights into the mechanisms underpinning the success of these techniques in a wild marine context.
We found that alleles at intermediate frequencies were at greater abundances than expected under HWE in WC and CW crosses relative to the on average distribution of loci in WW crosses, suggesting an overall increase in genetic diversity resulting from selective breeding of corals from different regions of the GBR. Distributions of allele frequencies typically follow that most loci nearly reach fixation at either 0 or 1, where alleles at intermediate frequencies are less common but are important given an increased abundance of intermediate alleles are the raw material for selection (Jombart & Collins, 2015). Hence, genetic material on which selective adaptive processes can operate likely exists at greater abundances in the WC and CW relative to the WW crosses. Hybridisation may lead to increased performance/fitness relative to the parental generation, with this increased performance having been linked to dominant, regulatory, single quantitative trait nucleotides (Jakobson & Jarosz, 2019). Increases in genetic diversity, hybrid vigour and genetic rescue are well-known but inconsistent features of intra- and inter-specific hybridization (Chan, Hoffmann & van Oppen, 2019; Flowers et al., 2019; Hazzouri et al., 2019; Weeks et al., 2011) but has not yet been demonstrated in the selective breeding of corals. For example, it was unknown whether the breeding of divergent populations would cause a decrease in genetic diversity due to processes like genome incompatibility (Hogenboom, 1975). These results support that even in a small number of crosses and contributing parents, genetic diversity improves. Observed heterozygosity and alpha Diversity (0Dα and2Dα at q = 0 and 2) was highest in the interpopulation crosses (WC, CW). The0Dα metric is sensitive to the presence of rare alleles, which is important given breeding across populations may initially introduce novel mutants (Sherwin et al., 2017), suggesting that WC, CW, and WW2 had an increased occurrence of rare variants. 1Hα, also known as the Shannon Information criteria, is informative as a natural measure of evolvability, and was highest in WC and WW2. Finally, preliminary analysis of adult colonies shows Tijou corals had on average higher observed heterozygosity compared to Backnumbers corals (Ho = 0.065 vs. 0.0217, unpublished data), suggesting that there could have been a risk of reduced diversity in offspring. Importantly, we did not observe a loss of genetic diversity by crossing these two divergent populations.
Finally, we found that interpopulation selective breeding significantly changed the resulting allele frequencies of offspring compared to modelled HWE distributions. This is perhaps unsurprising given assumptions of random mating, no gene flow, and infinite population sizes were not met. However, this suggests the influence of breeding on the genetic architecture of the resulting F1 generation. Therefore, the interbreeding of even a small number of corals from different reefs across the GBR may result in extensive introgression and therefore accelerate the potential for adaptation to warming, although the production of F2 generations would be needed to confirm this.
We also demonstrate that variants associated with immune responses, growth and cellular operating may be re-arranged during breeding but are maintained within the next generation. This suggests that the important functional diversity (i.e., at stress tolerance genes) associated with focal populations can be maintained in the breeding process. In response to heat stress, corals alter their gene and protein expression patterns, as reflected in changes in their structural lipids, metabolism, and immune responses (Barshis et al., 2013; Sogin et al., 2016). Differences in proteins associated with collagen production and sodium bicarbonate transport were important in differentiating the five families produced in our study. Collagen is important for the production of the extracellular matrix, required for multicellularity and the spatial organization of functional units of cells (Helman et al., 2008) whereas bicarbonate transporters are pivotal for coral calcification and hence growth (Zoccola et al., 2015). Basic cellular functioning also potentially varied through the differences in dTDP-glucose 4,6-dehydratase-like proteins detected and their involvement in the non-oxidative pentose phosphate pathway (Buerger, Wood-Charlson, Weynberg, Willis, & van Oppen, 2016; Yuyama, Watanabe, & Takei, 2011), critical for glucose utilization. Differences between crosses in these foundational processes like cellular organization and biomineralization therefore suggests that even breeding across relatively few individuals has the potential to substantially create distinct genetic combinations.
Protein NLRC3-like, which has been previously implicated in acroporid immune suppression in response to heat stress by acting in Toll-like receptor modification (Zhou et al., 2017), also varied significantly across families. Other proteins involved in immunity and stress were also detected (lysosomal-trafficking regulator-like proteins), and these have also been linked to the mounting of innate immune responses through Toll-like receptor activity in mice (Westphal et al., 2017). Additional immunity related proteins were detected, including CEPU-1-like protein (Spaltmann & Brummendorf, 1996), E3 ubiquitin-protein ligase RNF213-like (Iguchi et al., 2019), NACHT (Hamada et al., 2013), and spondin (Palmer & Traylor-Knowles, 2012). Interestingly, NFX1-type zinc finger-containing protein was found here and downregulated in resistant corals exposed to disease (Polato et al., 2010). Expression of nascent polypeptide-associated complex subunit alpha protein (Bellantuono, Hoegh-Guldberg, & Rodriguez-Lanetty, 2012) was similarly downregulated in resistant corals, suggesting that these proteins provide an important role in protective immune responses.
Genotyping individual aposymbiotic larva from the five families reared under ambient conditions (27.5°C) provides foundational knowledge as to how selective breeding influences underlying genetic architecture. It also sheds light on the underlying molecular origins and mechanisms of heritability, a long-standing goal of quantitative genetics (Jakobson & Jarosz, 2019). Broad and narrow-sense heritability has been quantified for a range of traits, but the underlying mechanisms have rarely been described (Dixon et al., 2015; Dziedzic, Elder, & Meyer, 2017). The majority of alleles fixed at the extremes of allele frequency distributions (“U-shaped”sensu Hill, Goddard, & Visscher, 2008) are likely driven by the small number of parents from which the larvae in each cross were derived (5 unique parental colonies used across the families, Supplementary Table 1), but may also suggest selection against heterozygotes during the aquarium rearing period. Interpopulation hybrids displayed greater genetic diversity relative to within population purebreds, which may be the result of the varying effects of selection on purebreds versus hybrids in the aquarium environment. U-shaped distributions may arise under strong cases of artificial selection (e.g. aquariums) combined with rare mutations (Hill et al., 2008).
Functional variation associated with selectively bred families
How does genomic variation lead to phenotypic differences between corals? The location and effect size of the SNP difference are important to determining its eventual phenotypic effect, and simplistically, differences in coding vs. non-coding regions are predicted to cause phenotypic effects through a variety of mechanisms (Cavallo & Martin, 2005). Assigning function to SNP differences is challenging given the majority of SNPs detected by association studies are non-coding (Nishizaki & Boyle, 2017), whereas the majority of key changes are coding (Cavallo & Martin, 2005), setting up a situation of difficult detection and classification. Analysis of population structure using DAPC links the genomic patterns seen in the multilocus genotypes with the underlying biological processes quantified in the heritability models. Using this approach, we identified alleles contributing to the separation of these selectively bred families, revealing that breeding of the selected populations targets changes to the immunity and stress responses and growth, likely important processes in survival generally. Assigning differences in SNPs to phenotypic differences between individuals will be key to understanding and increasing thermal tolerance for intervention methods.
Selective breeding influences the level of admixture and population discontinuity
Admixture events affect members of species, populations and individuals differently (Lawson, Van Dorp, & Falush, 2018). We saw this in the extent of admixture and its associated variance across the five families, in which some families exhibited very little admixture relative to others (especially CW). Although this can be somewhat dependent on the k structure of the model used, both patterns were explored independently using two techniques (DAPC and PCoA) and in conjunction with AIC, confirmed statistically the likelihood of population differentiation. Therefore, this may suggest that the shared ancestry of the colony sourced from Backnumbers2 may be limited between the other Backnumbers and Tijou corals or whose origins were from few or divergent founders (Lawson et al., 2018). This would suggest that adult colony Backnumbers2 is not highly related to other Backnumbers or Tijou colonies. Furthermore, PCoA and DAPC both demonstrated that the five families separate out in multidimensional space given the magnitude of allelic covariance between individuals. Irrespective of population labels, DAPC analysis also recapitulated the number of selectively bred families produced, although interestingly, the purebred families were not assigned to single population clusters but instead retained hybrid clustering structure in which the proportion of ancestry was shared between multiple two to three clusters simultaneously. The discontinuity between populations was also surprising, as demonstrated by the reduced spread of individuals between clusters, especially in WW1, suggestive that selective breeding produces discrete differences in the underlying genomic architecture of F1 individuals, even in populations likely undergoing some underlying level of gene flow (Lukoschek, Riginos, & van Oppen, 2016).