To our knowledge, this is the first study to investigate the relationships of a polygenic predisposition to depressive symptoms, as measured with a single trait and multi-trait approaches, with individual differences in depressive symptoms at baseline and a rate of change in depressive symptoms in the following 14 years among the nationally representative adults aged 50 years old and older. Cumulatively, our results contribute to a better understanding of the role a higher polygenic predisposition to depressive symptoms, independently or in interaction with socio-economic status, which was measured by educational attainment and accumulated wealth, plays in increasing risk for depressive symptoms onset and their longitudinal trajectory in the general population in the UK, for which knowledge is currently lacking.
Our results showed that the risk of having higher depressive symptoms is amplified by a higher loading of common genetic markers associated with depressive symptoms, with a greater genetic liability indicating a greater risk in older adults. Building on previous findings showing that PGS for depression derived from adults generalize to depression outcomes including depressive symptoms severity in youths18 and middle-aged adults14, these results may suggest that individual differences in depressive symptoms are influenced by an individual load of common genetic markers associated with depressive symptoms throughout a lifespan. Even though multi-trait polygenic score maximises the predictive power for depression23, our results showed that a polygenic score for depressive symptoms that encompass genetic information from the related traits, such as subjective well-being, neuroticism, loneliness, and self-rated health in addition to depressive symptoms, was not a stronger predictor of depressive symptoms at baseline and a rate of change in depressive symptoms in the following 14 years when compared to a single-trait PGS. This result is suspiring considering there are evidence suggesting that each of the traits included in the multi-trait polygenic score was shown to associate with depressive symptoms45. This may suggest that the multi-trait associations are due to the inclusion of depressive symptoms trait within the combined score, and that the additional factors make no difference. Nonetheless, this mapping of aetiological sources of cross-disorder overlap can guide future research aiming to identify specific mechanisms contributing to risk of depressive symptoms onset in older adults from the general population.
In agreement with a consensus that a higher educational attainment is protective against onset of depressive symptoms46,47, our results showed that each additional year of completed schooling was associated with a lower score in depressive symptoms at baseline in older adults. It has been hypothesised that a higher educational attainment may protect from the depressive symptom risk via more effective coping strategies or healthier lifestyles47–49. Similar to educational attainment, a lower accumulated wealth, which reflects limited socio-economic resources, low digital literacy, and limited access to participation in cultural activities or reduced social networks50,51, was also highlighted to be an important factor influencing individual levels of depressive symptoms independently from polygenic predisposition to depressive symptoms. We further observed an interaction effect between polygenic predisposition to depressive symptoms and low socio-economic status in association with onset of depressive symptoms in older adults. These results are supported by a twin study highlighting shared genetic risk for low education and major depression49. Therefore, providing psychoeducation about how to reduce stress52 associated with having lower educational attainment and accumulated wealth53, especially among older adults who have a higher polygenic predisposition to depressive symptoms, may prove beneficial in reducing risk for developing depressive symptoms in older adults.
Although an increase in depressive symptoms was observed over the 14-year long follow-up time, common genetic variants associated with depressive symptoms additively were not associated with a greater increase in depressive symptoms during this period in older adults from the general population. We further did not observe a significant interaction between either indicator of lower socio-economic status and polygenic predisposition in influencing the rate of change in depressive symptoms during 14 years of follow-up. These non-significant findings may reflect attrition effects, which are unavoidable in longitudinal cohorts. Similarly, because of the results presented in the study are based on longitudinal study with prospectively collected data, collider bias may have contributed to the non-significant findings54, which might have arisen from selection bias or attrition. However, the proportion of missingness in the present study was comparable to many longitudinal cohorts54–56; we further imputed missing values using robust approaches37,38. Therefore, it is unlikely that attrition or selection bias influenced our results. It is nevertheless possible that only a subset of the genetic factors for depressive symptoms may have an impact on individual differences in rate of change in depressive symptoms, which, due to the nature of the PGS approach, might not have been captured in the present study. Therefore, further analyses, such as pathway-specific polygenic score analyses, genomic structural equation modelling and gene-set enrichment analyses, may be needed before we can draw more conclusions.
Methodological considerations
This study consisted of a large sample size, which was a national cohort of older adults from England followed-up for 14 years. It was representative and included a relatively equal proportion of men and women with diverse backgrounds and different socio-economical positions. The duration of study brings a great advantage of repeated measures for depressive symptoms over 14 years of follow-up. Nonetheless, even though PGS is a good marker for genetic risk and tool for studying gene-environment interactions, it may have poor generalisability across populations because results are mostly based on European participants. Further studies are needed to assess genetic risk, develop PGS models and assess gene-environment interactions for non-white, non-European populations. It is worth mentioning that environmental factors can also have a different impact on non-white populations, with suggestions that education could have a greater protective effect for individuals of white ethnicity50. Similarly, because PGSs are built on GWAS, they may be restricted by the same limiting factors that are inherent to GWASs, such as being unable to capture rare variants, poorly tagged or multiple independent variants, gene-by-gene interactions, and gene-environment correlation51. In the light of the above, PGS evaluation should be treated as a tool for identification of population strata at higher risk of disease rather than accurate predictive diagnosis of individuals57. To minimise chances of collider bias affecting our findings58, covariates in the present study were those that were set at birth; however, we did not adjust the confounding effect of other factors on individual differences in depressive symptoms at baseline and during the follow-up.