Introduction

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is responsible for coronavirus disease 2019 (COVID-19) and the current global pandemic. It has soon gone virus across the world, affecting more than 200 countries/territories [1]. To date, the world has registered more than 24 million individuals contaminated, with more than 5 million deaths [2]. While the disease has mild effects in most individuals, severe COVID-19 is more likely to be observed in the those with comorbidities such as cardiovascular diseases [3]. Additionally, why certain populations are at a higher risk of adverse CHD outcomes post COVID-19 infection is still unclear [4].

Accumulating evidence revealed a bidirectional relationship between coronary heart disease (CHD) and COVID-19 [5,6,7,8], yet consensus has not been achieved regarding their causality. Patients of cardiovascular diseases are at higher risk of severe COVID-19 and death [3, 3). Within this region, the SNP Chr3:45859561 (rs10490770, nearest gene LZTFL1) exhibited the highest maximum PP4 and was considered the putative causal SNP. Additionally, rs17713054, also located in this region, demonstrated suggestive causality concerning the genetic correlation between COVID-19 and CHD. Notably, rs10490770 was previously identified as an IV in MR analyses. Furthermore, as indicated in the KEGG pathway analysis, LZTFL1 was not found to be involved in multiple pathways. Considering the LD relationship between rs10490770 with rs10490770, rs17713054 was not selected as an IV. Assuming that at most one causal SNP exists in the gene/region, the presence of two SNPs with high LD and high PP4 suggests that these two SNPs either both play a causal role in the outcome or are in strong LD with a single causal variant. Interestingly, both rs10490770 and rs17713054 exhibited associations with COVID-19 (P < 5 × 10−8) but not with CHD in their original GWAS (P > 0.05). Consequently, the SNPs identified in this study displayed vertical pleiotropy, impacting CHD solely through their influence on COVID-19 risk.

Table 3 Bayesian colocalization suggested the shared causal SNPs between COVID-19 and CHD susceptibility
Full size table

The cond/conjFDR analysis

To mutually verify the IV validity in MR, and to compensate for the inefficiency in identifying SNPs of horizontal pleiotropy in MR, we further provided a comprehensive, unselected map of shared loci between COVID-19 and CHD with a cond/conjFDR analysis (Fig. 2). With a predefined threshold at conjFDR < 0.05, we identified 5 distinct genetic loci shared between CHD and COVID-19 (Table 4). We observed that 2 SNPs mapped in the ABO gene (rs579459 Chr9:136154168 and rs495828 Chr9:136154867) showed consistent signals across varied COVID-19 severities, which implied the value of the ABO gene for jointly influencing CHD and COVID-19 (P < 5 × 10–4).

Fig. 2
figure 2

Conjunctional false discovery rate (conjFDR) Manhattan plots of conjunctional − log10(FDR) values

Full size image
Table 4 The conjunction false discovery rate (conjFDR) suggested evidence for the shared causal SNPs between COVID-19 and CHD susceptibility
Full size table

To observe increments of SNP enrichment for CHD as a function of the significance of COVID-19, a conditional QQ plot was presented (Additional file 1: Fig. S2). Gradual leftward curves indicated that the proportion of nonnull SNPs varies considerably across different levels of association with CHD, which supports the polygenic overlap between these phenotypes.

Discussion

In the current study, we observed a single-way causal effect of COVID-19 exerted on CHD. Shared genetic variants contributed to the causality, where rs10490770 in LZTFL1 suggested direct causality (SNPs → COVID-19 → CHD), and SNPs in ABO (rs579459, rs495828), ILRUN (rs2744961), and CACFD1(rs4962153, rs3094379) may simultaneously influence their risks.

To date, limited evidence is available in terms of the causality between COVID-19 and CHD. The current study indicated that the genetically determined risk of COVID-19 infection contributed to higher CHD complication risk. Several observational studies reported adverse CHD outcomes after COVID-19 infection [12, 34], which was consistent with our findings. A recent review also reported a considerable proportion of patients who recovered from COVID-19 continued to experience complications including CHD, even in the absence of a detectable viral infection [8]. This condition, which is often referred to as ‘post-acute COVID-19’ or ‘long COVID’, has been the major concern of clinical care for COVID-19 patients [35]. However, results of most observational studies might be confounded and/or influenced by reverse causality [36]. We thus attempted to infer causality with an alternative method, i.e., bidirectional MR [25,26,27]. For the reliability of causality observed in our study, we made multiple attempts to provide a rather robust estimate that was less likely to be false positive, for example, multiple sensitivity analyses, and IV validity check in an integrated framework. To be specific, the SNP (rs10490770) identified in colocalization analysis was suspected of vertical pleiotropy with its biological role validated from KEGG. It is suggested that LZTFL1, where rs10490770 was mapped, was involved in merely one pathway known to date, namely ‘BBSome-mediated cargo-targeting to cilium’ pathway. No additional evidence was shown regarding its direct association with CHD. Meanwhile, we noted that rs10490770 functioned exactly as IV, which strengthened our beliefs regarding its direct causality of SNP → COVID-19 → CHD. Regarding the relatively moderate effect size, the causality estimate was assumed to be attenuated compared to the true causal effect, for the limited number of genetic variants that overlapped between the COVID-19 HG meta-analysis and the datasets that were used for this study. Results should be interpreted with caution in terms of the genetically determined causality between the CHD incidence and COVID-19. In essence, the merit of MR limited its estimation of causality to the genetically determined risk of COVID-19 and CHD risks. Given that the genetic susceptibility of COVID-19 accounted for only a certain proportion of its total phenotypic variation, true causal estimates between these two traits were supposed to be much larger.

As the causality inferred previously may involve an interplay of both genetic and environmental factors, one cannot decide whether the causality inferred previously is contributed from their shared genetic variants. In this context, quantifying the genetic correlation can be thought as a prerequisite for subsequent identification of shared causal SNPs. In the current study, the genetic correlation was estimated at 0.18(COVID-19_C) to 0.23(COVID-19_A), suggesting an increasing trend between CHD and more severe COVID-19. However, it should be noted that the strength of their genetic correlation might be overestimated, as this analysis covers the entire genome. Still, it motivated us to further locate where these shared causal SNPs lied precisely in the genome. Generally, the genetic correlation is thought to be attributable to either vertical pleiotropy or horizontal pleiotropy [37]. Therefore, we applied COLOC coupled with cond/conj FDR to provide a comprehensive view of shared causal SNPs in a genome-wide scale. By integrating corresponding epidemiologic findings from populations with understanding of their biological functions, these SNPs identified may advance current knowledge of underlying mechanisms and may also facilitate clinical care of ‘Long COVID-19’ [35].

Commonly, COVID-19 and CHD are hypothesized to be linked by several biological mechanisms, including immune response, and endothelial dysfunction [21, 38, 39]. Of note, most SNPs identified in the current study were mapped in these pathways. Two SNPs (rs579459 and rs495828) in the ABO gene were assumed to be promising putative causal loci, as they showed consistent signals of pleiotropy when assessing CHD and COVID-19 of the three severity types. Recent studies showed that the ABO gene, located in 9q34.2, which determines blood type, may affect COVID-19 disease severity [40]. Several observational studies further reported a relationship between ABO blood groups and adverse CHD complications post COVID-19 [19, 20, 38]. In a recent GWAS meta-analysis, where investigators sampled 1980 patients with COVID-19-related respiratory failure and analyzed 8582968 SNPs, further cross-replicated the association of rs657152 at locus 9q34 with the COVID-19 severity [41]. In the current study, it is noticed that rs657152 was in linkage disequilibrium with rs579459 (D′ = 0.99), and they both were found to be expression qualitative trait loci (eQTLs) responsible for immune stimulation upon regulatory variant activity (http://pubs.broadinstitute.org/mammals/haploreg/haploreg.php). The activation of the immune response was assumed to be the key for both CHD risks and viral clearance [42]. On the one hand, both rs579459 and rs495828 were repetitively underscored to be associated with CHD risks, where immune response was suspected in the pathology [43,44,45]. On the other hand, both rs579459 and rs495828 functioned as eQTL [46] responsible for immune response activation and thus were hypothesized to influence the COVID-19 outcomes [47]. An alternative hypothesis is Renin-Angiotensin-System (RAS) unbalancing, where RAS-pathway genes, including rs495828 in ABO, was suspected predictive of CHD complications of COVID-19 [48]. Moreover, both rs579459 and rs495828 were found to be associated with Motifs change in Nkx2, where autoimmune mechanism underlying the acute respiratory distress syndrome in SARS-COV-2 was undergirded [47]. Considering Motifs play important regulatory roles, and may bind SARS-CoV-2 spike protein [49], rs579459 and rs495828 identified in the current study are worthy of further replication to fully clarify their roles.

For rs10490770 located near LZTFL1 in the 3p21.31 region, it was in high LD with rs17713054 (D′ = 1), which was identified in COVID-19 GWASs as conferring a twofold increased risk of respiratory failure [5, 41, 50, 51] and an over twofold increased risk of mortality for individuals under 60 years old [52]. Also, rs10490770 was in LD with rs11385942 (D′ = 0.98) which was cross-validated in a recent COVID-19 GWAS meta-analysis [41]. However, despite repetitive statistical significance reported in the 3p21.31 region, its specific role in COVID-19 infection remained unexplained [50, 51]. We coincidentally observed rs17713054 in this region colocalized between COVID-19 and CHD, which might offer insights for further uncovering the biological role in this region by simultaneously taking COVID-19 and its CHD complication into considerations. Recent studies reported that rs17713054-affected enhancer upregulates LZTFL1 [16, 17], which is currently known to be actively involved in the epithelial–mesenchymal transition in the viral response pathway of COVID-19 [50]. We further supplemented eQTL colocalization analysis in the 3p21.31 region (data not shown in the main text). We confirmed that rs10490770 colocalized with eQTL and was highly expressed in the lung tissue (Additional file 1: Figure S2). Meanwhile, no known evidence showed rs10490770 was directly associated with CHD, nor with its belonging pathway ‘BBSome-mediated cargo-targeting to cilium’ pathway. Evidence gathered thus far consistently supported our findings that rs10490770 function only through COVID-19 to CHD. Although statistical signals are not necessarily validated biological evidence, we provided preliminary hints for further studies to uncover the underlying molecular mechanism.

The current study has several clinical implications. First, this study suggested that genetic predisposition to COVID-19 is a causal risk factor for CHD, leading to the hypothesis that reducing the COVID-19 infection risk or alleviating COVID-19 severity among those with specific genotypes might reduce their subsequent CHD adverse outcomes. Initially recognized as a respiratory system disease, COVID-19 has been found to interact with and affect the cardiovascular system leading to myocardial damage and cardiac and endothelial dysfunction [53]. In fact, cardiac damage has been noted even without clinical features of respiratory disease [54], with new-onset cardiac dysfunction common in this subgroup [55, 56]. Specifically, the CHD incidence secondary to COVID-19 infection might be characterized with massive damage in the vascular endothelium and cardiac myocytes due to the systemic inflammatory response in severe COVID-19, which includes the release of high levels of cytokines (known as cytokine release syndrome) [57,58,59]. These patients with subsequent myocardial involvement could suffer from several potentially life-threatening symptoms [1]. Second, these SNPs identified may be of clinical implications for identifying the target population who are more vulnerable to adverse CHD outcomes post COVID-19 and may also advance treatments of ‘Long COVID-19’ [35].

This study also had limitations worthy of noting. First, the validity of the genetic instruments was not fully understood. It is possible that the genetic instruments may have an indirect effect on the outcome via a currently unknown pathway that does not involve the risk factor for interest. Nevertheless, we addressed this issue by adopting the MR-Egger intercept, although it cannot be ruled out unequivocally. Second, as GWAS summary datasets were extracted from Europeans, the generalizability of the study results was limited to Europeans only. Third, study participants included in the COVID-19 were not screened for CHD at baseline and vice versa. The presence of outcomes in the exposure dataset may bias the causal estimates in MR analyses. However, this is a general limitation of two-sample MR analyses and is inevitable without individual-level data. Third, we acknowledge the potential bias introduced by environmental factors, which could not be completely mitigated, despite our efforts to adjust for age and gender in the GWAS summary statistics and our multivariate Mendelian randomization analysis. However, our results indicate that our findings remain robust in the presence of some established environmental factors. Fourth, the methods we adopted that were built upon GWAS summary statistics, including the LDSC method, along with the Mendelian randomization and Bayesian colocalization, required larger sample sizes than methods that use individual genotypes to achieve equivalent standard error. Of note, our analyses were limited by the number of genetic variants that overlapped between the COVID-19 HG meta-analysis and the datasets that were used for this study. Thus, we could not test some genes that may be of importance. It is also possible that, with larger sample sizes, the genetic association of COVID-19 severity and CHD could become more significant, and confidence intervals would narrow around true estimates.

Conclusion

In summary, this study indicated the putative causality of COVID-19 genetic susceptibility on incident CHD. It underlined rs10490770 located near LZTFL1 and SNPs in ABO (rs579459, rs495828), ILRUN (rs2744961), and CACFD1 (rs4962153, rs3094379) may simultaneously influence their risks. Further studies are warranted to clarify their underlying mechanism.