Genome-wide association study of post COVID-19 syndrome in a population-based cohort in Germany, 2025, Russ+

[Nightsong]

Abstract:
If health impairments due to coronavirus disease 2019 (COVID-19) persist for 12 weeks or longer, patients are diagnosed with Post-COVID Syndrome (PCS), or Long-COVID. Although the COVID-19 pandemic has largely subsided in 2024, PCS is still a major health burden worldwide, and identifying potential genetic modifiers of PCS remains of great clinical and scientific interest.

We therefore performed a case-control type genome-wide association study (GWAS) of three recently developed PCS (severity) scores in 2,247 participants of COVIDOM, a prospective, multi-centre, population-based cohort study of SARS-CoV-2-infected individuals in Germany. Each PCS score originally represented the weighted sum of the binary indicators of all, or a subset, of 12 PCS symptom complexes, assessed six months or later after the PCR test-confirmed SARS-CoV-2 infection of a participant.

For various methodical reasons, however, the PCS scores were dichotomized along their respective median values in the present study, prior to the GWAS. Of the 6,383,167 single nucleotide polymorphisms included, various variants were found to be associated with at least one of the PCS scores, although not at the stringent genome-wide statistical significance level of 5 × 10− 8. With p = 6.6 × 10− 8, however, the genotype-phenotype association of SNP rs9792535 at position chr9:127,166,653 narrowly missed this threshold.

The SNP is located in a region including the NEK6, PSMB7 and ADGRD2 genes which, however, does not immediately suggest an etiological connection to PCS. As regards functional plausibility, variants of a possible effect mapped to the olfactory receptor gene region (lead SNP rs10893121 at position chr11:123,854,744; p = 2.5 × 10− 6). Impairment of smell and taste is a pathognomonic feature of both, acute COVID-19 and PCS, and our results suggest that this connection may have a genetic basis.

Three other genotype-phenotype associations pointed towards a possible etiological role in PCS of cellular virus repression (CHD6 gene region), activation of macrophages (SLC7A2) and the release of virus particles from infected cells (ARHGAP44). All other gene regions highlighted by our GWAS did not relate to pathophysiological processes currently discussed for PCS. Therefore, and because the genotype-phenotype associations observed in our GWAS were generally not very strong, the complexity of the genetic background of PCS appears to be as high as that of most other multifactorial traits in humans.

Link | PDF (Nature Scientific Reports, May 2025, open access)

 

[Yann04]

Although the COVID-19 pandemic has largely subsided in 2024,

Has it? Maybe we’re switching from Pandemic to Endemic but the virus is still there and a major risk for many.

All other gene regions highlighted by our GWAS did not relate to pathophysiological processes currently discussed for PCS

That’s more of a reason to look into it no? If we only accept evidence that bolsters our pet theories and ignore the ones that don’t, progress will be difficult.

 

[Yann04]

The most significant association, with SNP rs10893121 (p = 2.5 × 10− 6) at position chr11:123,854,744, is located in a region harbouring genes from olfactory receptor families 4, 6 and 10. The minor allele of rs10893121 increases the risk of developing more severe PCS by approximately 50% (OR = 1.49). The strongest association (OR = 2.47) was observed for SNP rs61739314 located at position chr20:39,990,377 in a region that contains the ZHX3 (zinc fingers and homeoboxes 3), LPIN3 (lipin 3), EMILIN3(elastin microfibril interfacer 3) and CHD6(chromodomain helicase DNA-binding protein 6) genes. Notably, the region around the FOXP4 gene (highlighted in green in Fig. 1A), the site of the sole genome-wide significant association with PCS reported so far, by Lammi et al.8, was not found to be associated (at the ‘suggestive’ significance level of p < 10− 5) with the dichotomized original PCS score in our data. Moreover, none of the associations highlighted in the present GWAS achieved nominal significance in the study by Lammi et al.8.

 

[EndME]

So in this GWAS none of the things found to be significant in the other GWAS by Lammi et al are significant here (and obviously vice versa). It does seem that this study has some strengths that Lammi et al doesn't have (for instance 6 months vs 3 months are considered to be LC).

They use 3 different PCS scores here which are developed in:
-Subdomains of Post-COVID-Syndrome (PCS) -- A Population-Based Study, Bahmer, Scheibenbogen et al
-Severity, predictors and clinical correlates of Post-COVID syndrome (PCS) in Germany: A prospective, multi-centre..., 2022, Bahmer et al

Since they are getting different results per score they used also that seems to indicate that the choice of definition is influencial in driving their results to some extent, so what associations are they picking up on or are they all rather arbitrary?

One should expect that those that had a severe acute infection should have genes that have been found to be associated to more severe acute infections (and if not the results could be rather interesting)? Perhaps a nice sanity check to do and to see how those results would line up with the results of Lammi et al who found signals that were predominantly also related to a severe acute infection (the sample size however would be very small).

 

[Andy]

2,247 participants

Probably insufficiently powered (not large enough) to provide accurate results.

 

[wigglethemouse]

A very low number of controls were used for GWAS. That seems to be a red flag for me.

A PCS score-based median split (at 15.5 points) of the combined Kiel and Würzburg/Berlin dataset yielded 1,048 cases and 1,067 controls, with no evidence of systematic genetic confounding (λG = 1.016). With this sample size, only unrealistically strong genotype-phenotype associations would have been detectable at the so-called ‘genome-wide significance level’ of 5 × 10− 8, which in turn would have led to an unacceptably high false negative rate. Based upon preceding power calculations, we therefore set the p value threshold for further consideration of an association as ‘suggestively’ significant to 10− 5.

 

[Adrian]

Probably insufficiently powered (not large enough) to provide accurate results.

If health impairments due to coronavirus disease 2019 (COVID-19) persist for 12 weeks or longer, patients are diagnosed with Post-COVID Syndrome (PCS), or Long-COVID.

Given it is probably quite a diverse group (esp with 12 weeks or longer effects) then this does feel underpowered say compared to the numbers in Decode.