Socio-economic status is a social construct with heritable components and genetic consequences

Abstract

In civilizations, individuals are born into or sorted into different levels of socio-economic status (SES). SES clusters in families and geographically, and is robustly associated with genetic effects. Here we first review the history of scientific research on the relationship between SES and heredity. We then discuss recent findings in genomics research in light of the hypothesis that SES is a dynamic social construct that involves genetically influenced traits that help in achieving or retaining a socio-economic position, and can affect the distribution of genes associated with such traits. Social stratification results in people with differing traits being sorted into strata with different environmental exposures, which can result in evolutionary selection pressures through differences in mortality, reproduction and non-random mating. Genomics research is revealing previously concealed genetic consequences of the way society is organized, yielding insights that should be approached with caution in pursuit of a fair and functional society.

Figure 1. Changes in the heritability of educational attainment over time in Europe.

The Figures show the percentage of variation in educational attainment explained by genetic and environmental influences with 95% confidence intervals, as estimated in a meta-analysis of 28 European twin cohorts. Shared environmental influences refer to environmental factors that make siblings more similar to each other, while unique environmental influences are factors that do not, and also include measurement error. In twin studies, shared environmental influences can be overestimated at the expense of genetic influences when assortative mating occurs, as it increases the genetic similarity between siblings, mimicking shared environmental influences. The Figure is based on data from Table 2 from Silventoinen et al (2020).

Figure 2. Traits that show higher genetic correlations with educational attainment tend to show stronger regional differences.

Genetic correlations (r_g) can vary between -1 (100% shared variance due to the same genetic effects in the opposite direction) through 0 (no overlap in genetic effects) to 1 (100% of shared variance due to the same genetic effects in the same direction). Y-axis indicates the absolute genetic correlation of the trait with educational attainment (EA) (Lee et al, 2018), excluding all British. Genetic correlations were computed with LD Score Regression. X-axis shows the Moran’s I, a measure of geographic clustering, of 31 polygenic scores in ∼320k individuals in Great Britain. The Moran’s I of the educational attainment polygenic score is 0.6 (not shown). This Figure was adapted from Supplementary Figure 5 in Abdellaoui et al (2019).

Figure 3. Polygenic prediction of average phenotypes per region likely captures environmental influences.

Polygenic scores for educational attainment (EA) capture environmental effects on a regional level that are not visible when examining individual-level data. Panel (a) shows this based on ∼320k unrelated UK Biobank participants of European descent. The polygenic score for height explains 21% of individual differences in height, while the polygenic score for EA explains only 1% of individual differences in height. When we consider the average scores per region, however, the polygenic score for EA explains more regional differences in height (64%) than the polygenic score for height does (52%), presumably because the regional average of the EA polygenic score better captures regional differences in poor versus rich environments, and these affect height. Panel (b) displays the geographic distributions of regional averages of polygenic scores for height and EA, alongside the regional average for phenotypic height. Black lines denote coal mining regions, where environmental circumstances associated with socio-economic deprivation tend to cluster with polygenic scores for EA. Without such clustering, coal mining regions would be among the taller regions of the country, which phenotypically they are not. For statistical analyses regarding these regional differences and the migration herein, see Abdellaoui et al. (2019). The hypothesized causal diagram in panel (c) illustrates environmental influences on height (E) that cluster regionally with genetic influences associated with SES (G(SES)), making the regional average of those genes ${\overline{(G(SES)}}_{regional})$ predictive of the regional average of phenotypic height $\left({\overline{Height}}_{regional}\right)$. G(Height) denotes the genetic influences on height on an individual level (Height). For details on the data and QC for panels (a) and (b), see Abdellaoui et al 2022; for details on polygenic score computation and geographic regions, see Abdellaoui et al, 2019.

Figure 4. Genetic correlations show that Covid-19 infections and deaths in England originate in higher SES regions and spread more widely in lower SES regions.

Panels a and b show data for Covid-19 cases and deaths respectively. The panels show results from a total of 2,924 regional GWASs (RGWASs; four per day - cases and deaths, cumulative and weekly - for 2 years, i.e., 731 days) performed on 1.2 million common single nucleotide polymorphisms (SNPs) from 396,042 individuals of European descent living in England. As opposed to a traditional GWAS, in an RGWAS the subjects are given the phenotype of the region they live in (315 regions), which often results in genetic signals associated with socio-economic outcomes due to their geographic clustering. Covid-19 data on the 315 regions were obtained from Public Health England. Each dot is one RGWAS (one day), for which either the weekly or the cumulative cases (a) or deaths (b) were analyzed as the phenotype. The upper panel shows the variation explained by all 1.2 million SNPs (the SNP-based heritability or SNP-h²). The large middle panel shows the genetic correlation (r_g) of the genetic signal with the educational attainment (EA) GWAS from Lee et al (2018). The gray shaded areas around the points indicate 95% confidence intervals for both the SNP-h² and genetic correlations. These genetic correlations shows several positive peaks, including at the start of the pandemic in March 2020 and around the start of the spread of the new and more contagious B1.1.7 variant in December 2020, both reflecting more infections in richer regions of the country (in or near London), after which the genetic correlation with education becomes negative again. The two bottom panels show the total number of cumulative and weekly Covid-19 cases (a) or deaths (b). For more information on quality control and statistical approaches used, see Abdellaoui (2020), where the same results were reported for cases only for the first ∼6 weeks of the Covid-19 pandemic.