The role of candidate genetic polymorphisms in covid-19 susceptibility and outcomes

Abstract

Background

This study aims to investigate the association between candidate host genetic polymorphisms and COVID-19 susceptibility, severity, hospitalization, hypoxia, and their combined effect, measured by the polygenic risk score (PRS).

Methods

Three hundred and seventy-six Lebanese participants, comprising 151 controls and 225 cases, were included. Clinical data were obtained from questionnaires and medical records. DNA isolated from peripheral blood was genotyped for ACE1 rs1799752, ACE2 rs2074192, TMPRSS2 rs75603675 and OAS1 rs107746771 using TaqMan assays, and for TMPRSS2 rs35074065 using Sanger Sequencing. Candidate genetic variants were analyzed in association with COVID-19 susceptibility, severity, hospitalization and hypoxia, using univariate and multivariate models. PRS constructed from the weighted sum of variants was evaluated in association with COVID-19 outcomes.

Results

In this study, there were no statistically significant differences in the frequencies of candidate variant alleles between cases, controls and within disease outcomes subgroups, after adjustment for confounders. PRS was not associated with COVID-19 susceptibility and hospitalization, it however significantly predicted COVID-19 severity (P = 0.01).

Conclusion

This study highlights the importance of genetic testing for key host genes involved in COVID-19 life cycle and eventually measuring the PRS which proves to be an important tool for prognosis assessment in vulnerable individuals, potentially enhancing patient care.

Introduction

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), a member of the Coronaviridae family of viruses, is a positive single stranded RNA virus whose outbreak in Wuhan, China on December 2019 progressed to the Coronavirus disease 2019 (COVID-19) global pandemic [1]. As of today, SARS-CoV-2 has infected over 774 million individuals and has taken the lives of more than 7 million worldwide [2]. Although the pandemic has ended, many countries are still experiencing COVID-19 outbreaks due to emerging mutant variants [3]. Epidemiological studies have shown that although SARS-CoV-2 has spread worldwide, some countries have witnessed higher incidence and mortality rates compared to others. For instance, France had the highest mortality rate (13.38%) whereas Germany had the lowest (1.94%) [4]. This variability in susceptibility and severity is due to an interplay between environmental, social and genetic factors.

Many genetic polymorphisms were found to be associated with COVID-19 severity and susceptibility [5]. These polymorphisms are particularly those of genes involved in the pathogenesis mechanism of SARS-CoV-2 [6]. ACE2 and TMPRSS2 are examples of such genes as their expression leads to the formation of cell surface proteins angiotensin-converting enzyme 2 and transmembrane serine protease, respectively, that together aid in SARS-CoV-2 entry into host cells [7]. Multiple ACE2 variants were found to distort the protein’s structure, suggesting a role in altering SARS-CoV-2’s entry into cells and subsequently COVID-19 susceptibility and severity [8]. The ACE2 rs2074192 variant (C > T) was shown to be more prevalent in symptomatic patients and those with more severe outcomes [9, 10]. As for TMPRSS2, Garcia et al.’s study [11] on 817 patients with COVID-19 showed that the CC and CA genotypes of the TMPRSS2 rs75603675 variant were associated with increased COVID-19 severity compared to the AA genotype. Another variant is the TMPRSS2 rs35074065 (C>-) polymorphism which, based on Russo et al.’s in-silico study [12], is suggested to be associated with increased susceptibility and severity of COVID-19 due to its role in increasing this gene’s expression.

Invariably, genes involved in mounting an immune response against the infection were also found to be related to COVID-19 susceptibility and severity [13]. For instance, and in addition to variants related to cytokines and interleukins which are beyond the scope of this manuscript, the OAS1 variant (rs10774671), via large Biobank and in-silico studies [13, 14], has been shown to affect both the susceptibility to SARS-CoV-2 and the severity of its infection. OAS1 is an immunity-related gene that codes for oligoadenylate synthetase 1 (OAS1) protein, and this G > A variant in particular lacks the ability to target the OAS1 protein to the endomembrane system where it can detect the RNA of SARS-CoV-2 and mount an immune response against it [15].

Additionally, genes involved in the normal physiology, such as ACE1, could also lead to a higher predisposition to disease if aberrant [16]. Of particular interest, an insertion/deletion (I/D) allele in ACE1 (rs1799752), a gene involved in the renin-angiotensin system (RAS), was previously investigated and found to be associated with COVID-19 susceptibility and severity [17]. Nevertheless, studies establishing the role of genetic polymorphisms in COVID-19 infections, particularly those common in frequency in the European and Middle Eastern regions, are conflicting in results and remain particularly limited. Additionally, only a few take into consideration the combined effect of these polymorphisms outlined by the polygenic risk score (PRS), a risk prediction tool.

The aim of this study is to hence determine whether these previously described relatively common host candidate genetic polymorphisms, independently or combined, are associated with COVID-19 susceptibility, severity, risk of hospitalization and hypoxia in our population.

Methods

Human subjects

This study was approved by the Institutional Review Board (IRB) of the American University of Beirut (AUB) that is compliant with the Helsinki Declaration, and builds on a previously described cohort that was recruited during the peak of the COVID-19 pandemic to examine the effect of the ACE1 I/D polymorphism on COVID-19 susceptibility, severity and prognosis [17]. All participants signed an informed consent.

A total of 387 Lebanese participants, aged 18 and above, were initially recruited upon visiting the American University of Beirut Medical Center (AUBMC) for COVID-19 related reasons such as PCR testing (regardless of the outcome), COVID-19 hospitalization, or follow-up for persistent COVID-19 symptoms. Recruitment entailed a one-time participation that comprised informed consent, data collection and peripheral blood withdrawal for DNA isolation and genotyping. Out of the initial 387 participants, 376 agreed that their DNA be used for additional genotyping and are hence included in the current study. All participants were recruited prior to the availability of the COVID-19 vaccine.

Data collection

The collected data mainly included demographics, comorbidities, medications intake, PCR testing results, presentation of participants upon recruitment and progression of COVID-19 during their stay when applicable. These were obtained from a questionnaire administered to the participants and via access to the electronic health records on EPIC, the electronic health information system used at AUBMC.

Genotyping

Peripheral blood was withdrawn from the participants and collected into EDTA containing tubes then later processed into aliquots to be stored at -80 °C. DNA isolation was performed using FlexiGene^Ⓡ DNA Kit (QIAGEN^Ⓡ, Germany) as per the manufacturer’s guidelines. Subsequently, the DNA was quantified and assessed for purity using the DS-11 Spectrophotometer (DeNovix^Ⓡ, USA) then stored at -20 °C.

ACE1 I/D (rs1799752) was previously genotyped as per the methods described in the aforementioned study [17]. Genotyping of ACE2 (rs2074192), TMPRSS2 (rs75603675) and OAS1 (rs107746771) was carried out on Bio-Rad’s CFX384 real-time polymerase chain reaction (PCR) machine using TaqMan Single Nucleotide Polymorphism (SNP) allele discrimination assays C__16163821_10, C_102710002_10, and C___2567433_10, respectively (Thermo Fisher Scientific, USA) following manufacturer’s recommendations.

Genotyping of TMPRSS2 (rs35074065), a nucleotide (C) deletion polymorphism, was performed by Sanger sequencing. The DNA was first amplified through PCR on CFX 96 machine (Biorad, USA) using the following primers: forward: 5’-GGGCCCCCAAAGTAACCAATGGA-3’ and reverse: 5’-ATGGACCATTGAGCCAGTGCTTATGT-3’ at a final concentration of 0.1µM in the presence of Redtaq PCR reaction mix (Sigma-Aldrich, Germany), and according to the following protocol: 94 ◦C/30 s, 55 ◦C/30 s, and 72 ◦C/30 s per cycle for 30 cycles [18]. The amplified DNA was then purified using GenElute PCR Clean-Up kit (Sigma-Aldrich, Germany) as per the manufacturer’s protocol. Subsequently, 10% of the samples were loaded on 1% agarose gel and verified for the presence of the 252 bp band. The purified PCR products were then sequenced on the ABI 3500^Ⓡ Sanger sequencer (Life Technologies, USA) with the forward primer above and the BigDye Terminator v3.1 Cycle Sequencing kit (Applied Biosystems, USA).

10% of the samples were genotyped twice and exhibited 100% reproducibility.

Statistical analysis

Data were initially entered on Microsoft Excel then exported to IBM SPSS Statistics for data description and analysis.

Baseline characteristics included age, body mass index (BMI), sex, comorbidities, and intake of ACE inhibitor (ACEI) or angiotensin receptor blocker (ARB). The outcomes of interest were COVID-19 susceptibility, severity, hospitalization and hypoxia as described in the previous cohort [17]. COVID-19 cases were participants who tested positive on the COVID-19 polymerase chain reaction (PCR) testing whereas controls were those who tested negative at the time of recruitment. COVID-19 severity was categorized as per the WHO clinical progression scale into three stages: stage I (mild), stage II (moderate), and stage III (severe) [19]. The mild stage included any combination of the following: fever and/or chills, cough, shortness of breath, sore throat, congestion and/or rhinorrhea, fatigue, myalgias, headache, nausea and/or vomiting, diarrhea, anosmia, and ageusia. The moderate stage included symptomatic patients who had shortness of breath but a blood oxygen saturation (SpO2) ≥ 94% with minimal or no oxygen therapy required, or those with abnormal chest imaging results [19]. The severe disease stage included critically ill patients in respiratory distress who required hospitalization [19]. In our cohort, all of those categorized as moderate or severe were hospitalized, yet the sample size is not exactly the same to the hospitalized category as two had mild disease but still hospitalized for a short period. We also could not categorize few subjects as mild or moderate due to missing data.

The association analyses of baseline characteristics and genetic factors with the four outcomes were conducted using Pearson Chi-Square or Fisher’s Exact test for categorical variables as applicable and independent sample t-test or ANOVA for continuous variables. In addition, binary or multinomial regressions, as applicable, were initially carried out at the univariate followed by multivariate regressions while adjusting for all statistically significant confounders at the univariate level. The results are shown as odds ratios (ORs) with 99% confidence intervals. Since five SNPs were tested, we applied a Bonferroni correction for multiple testing and considered P-value ≤ 0.01 to be statistically significant.

To evaluate the cumulative effect of the variants, we computed PRS using individual data and adjusted odds ratios of the variants in association with the investigated outcomes. Based on a published guide for performing PRS analysis with a binomial outcome, the recommended target or effective sample size is at least 100 individuals [20]. To adjust for case-control imbalance in our clinical outcomes, we calculated the effective sample size, and PRS analysis was performed for clinical outcomes with effective sample size over 100. Of note that for disease severity, we lumped moderate and severe disease together into one category.

First, genotype data were prepared in binary form and tested for quality control (MAF > 5%, genotyping rate > 1%, Hardy-Weinberg equilibrium P > 10^− 6) with PLINK version 1.07. After that, PRS was computed with PRSice version 2.3.3 using default clumping parameters. All variants passed quality control and clumping parameters and were included in the PRS calculation (Supp Methods), except for rs2074192 since it is located on a sex chromosome. The best-fit PRS was determined, and logistic regression was performed to assess the association between the best-fit PRS and each outcome after adjustment for corresponding confounders. Empirical P-values were generated using the permutation method [20, 21]. PRS analysis was performed using Ubuntu version 22.04.3 and R version 4.2.3.

Genotype analysis was performed for both the additive and allele models with frequencies presented for both. Yet, logistic regression and PRS results are presented for the allele model only since results were essentially similar.

Results

Three hundred and seventy-six Lebanese adult participants were included in this study, out of whom 225 were cases and 151 were controls. The variant alleles were generally the least common as compared to wild type carriers with the following allele frequencies in the controls: ACE1 rs1799752 I allele (30.8%), ACE2 rs2074192 T allele (35.0%), TMPRSS2 rs35074065 C allele deletion (40.9%), and TMPRSS2 rs75603675 A allele (38.1%). The only exception was OAS1 rs10774671 A allele that had a frequency of 86.4%.

Susceptibility

Baseline characteristics were compared between cases and controls and showed similar results to the previous cohort [17] and to what is already known in the literature. For instance, cases were mainly males, significantly older, had a higher body mass index, and had more comorbidities when compared to controls (Supp Table 1).

As for the genetic association analysis, there were no statistically significant differences in terms of variant allele frequencies between cases and controls as shown in Supp Table 2. Additionally, as shown in Supp Table 3, multivariate analysis correcting for age, BMI, sex and statistically significant confounders (hypertension, diabetes and cancer; Supp Table 1) revealed no statistically significant associations between the candidate allele variants and susceptibility. In addition, PRS computed from the weighted sum of the four SNPs (rs1799752, rs35074065, rs75603675, rs10774671) was not associated with susceptibility, whereby mean (SD) was 0.09 (0.02) in the cases versus 0.08 (0.02) in the controls (P = 0.10) (Fig. 1, Supp Table 4).

Fig. 1

PRS in association with hospitalization, susceptibility and severity. PRS (mean + SD) in each category of the investigated outcomes: (A) Susceptibility; (B) Severity; (C) Hospitalization. Logistic regression was performed for the association between PRS and outcome, with adjustment of corresponding confounders, and empirical P-value was calculated using the permutation method [21]

Severity

Among the 225 cases, 9 patients were lost to follow-up whereas the remaining 216 were categorized based on the severity of their symptoms ranging from mild to moderate to severe. One-hundred and thirty-one (60.7%) patients had a mild infection, 25 (11.6%) had a moderate severity, and 60 (27.7%) suffered from a severe COVID-19 infection (Supp Table 1).

Comparison of baseline characteristics among the different severity levels also showed similar results to the previous cohort [17]. As such, males with multiple comorbidities, taking ACEIs or ARBs, and who are significantly older with a higher body mass index were more likely to experience moderate to severe COVID-19 infection (Supp Table 1).

As for the genetic association analysis, there were no statistically significant differences in terms of variant allele frequencies between mild, moderate and severe cases as shown in Supp Table 2. Additionally, as shown in Supp Table 3, multivariate analysis correcting for age, BMI, sex and statistically significant confounders (dyslipidemia, hypertension, diabetes, heart disease, kidney disease, coagulation disorders, cancer, ACEI/ARB intake; Supp Table 1) revealed no statistically significant associations between the candidate allele variants and severity. PRS computed from the weighted sum of the four SNPs (rs1799752, rs35074065, rs75603675, rs10774671) was positivity associated with severity, whereby mean (SD) was 0.08 (0.07) in the cases versus 0.05 (0.08) in the controls (P = 0.01) (Fig. 1, Supp Table 4).

Hospitalization

Among the 225 cases, 87 patients (38.7%) required hospitalization for additional management and monitoring (Supp Table 1). Comparing baseline characteristics between hospitalized and non-hospitalized patients showed similar results to the previous cohort as well [17]. Briefly, the patient population more likely to be hospitalized was similar to that having severe prognosis of COVID-19 (Supp Table 1).

As for the genetic association analysis, as shown in Supp Table 3, multivariate analysis correcting for age, BMI, sex and statistically significant confounders (dyslipidemia, hypertension, diabetes, heart disease, kidney disease, cancer, ACEI/ARB intake; Supp Table 1) revealed no statistically significant associations between the candidate allele variants and hospitalization. Of note that, although significantly more of the hospitalized cases were homozygous TT for the ACE2 rs2074192 allele when compared to the non-hospitalized cases (32.2% vs. 15.2%, respectively; P = 0.002)) (Supp Table 2), this significance did not reach the cut off P value of 0.01 with the univariate and multivariate logistic regression using the allele model (Supp Table 3). In addition, no statistically significant results were shown with the additive model in the multivariate logistic regression after adjustment for confounders (Data not shown).

In addition, PRS computed from the weighted sum of the four SNPs (rs1799752, rs35074065, rs75603675, rs10774671) was not associated with hospitalization, whereby mean (SD) was 0.11 (0.06) in the cases versus 0.10 (0.07) in the controls (P = 0.05) (Fig. 1, Supp Table 4).

Hypoxia

Among the 87 hospitalized patients, 61 (70.1%) suffered from hypoxia as a result of their COVID-19 infection (Supp Table 1).

The baseline characteristics of hospitalized hypoxic and non-hypoxic patients were also similar to the previous cohort [17] and to what is found in the literature. As a matter of fact, hypoxic patients were more likely to be obese compared to the non-hypoxic group with a mean BMI of 31.07 kg/m2 vs. 27.28 kg/m2; P = 0.006) (Supp Table 1).

As for the genetic association analysis, there were no statistically significant differences in terms of variant allele frequencies between hospitalized and non-hospitalized cases as shown in Supp Table 2. Additionally, as shown in Supp Table 3, multivariate analysis correcting for age, BMI and sex (Supp Table 1) revealed no statistically significant associations between the candidate allele variants and hypoxia. PRS analysis was not performed for hypoxia, since it lacked the sample size recommended for this analysis of at least 100 individuals [20].

Discussion

With the on-going COVID-19 outbreaks, it is still quite clear that individuals are affected differently from this virus in terms of disease susceptibility and outcomes. Multiple studies have explored the cause of these different outcomes, some linking it to comorbidities and others to genetic predisposition [5, 22]. Genetic studies in the Middle East are limited and we are the first to explore polygenic risk score associations in this region. We have adopted the candidate gene variants approach and evaluated some of the alleles with the most compelling evidence and that were common among Europeans and Middle Easterns assuming that they would be so in the Lebanese. In fact, ACE1 I/D allele frequency in our controls is somewhat similar to what is found in the literature in the Middle Eastern population whereby the D allele is more common compared to the I allele [23]. ACE2 rs2074192 allele frequency was also similar to Italian controls whereby the C allele was more common than T [9]. As for TMPRSS2 rs35074065 and TMPRSS2 rs75603675, the C alleles of both SNPs in our samples more common among our controls, similar to what is found in Italian and Iranian populations, respectively [18, 24]. Lastly OAS1 rs10774671 A allele was also the most common in our patient controls, in line with what is described in the literature for the European population [15].

In our previous study, the ACE1 II genotype was found to be associated with increased risk of COVID-19 susceptibility and decreased risk of hypoxia [17]. In the current study however, although the results are of the same trend as we previously reported, no statistically significant results were revealed after correcting for multiple testing. In addition, the numbers and P values are not exactly the same as we had to exclude the few participants who did not consent that we use their samples for additional analyses.

The TMPRSS2 rs75603675 AA genotype is expected to affect the structure of the transmembrane serine protease, thus, altering SARS-CoV-2 entry into cells, whether by facilitation or hindrance [25]. Studies on this particular SNP are conflicting. For instance, the Garcia et al.’s study [11] found that the C/C and C/A diplotype lead to more severe outcomes as compared to the A/A diplotype. Additionally, Rokni et al. [24] found that the AA genotype is protective from the severe form of COVID-19 infection requiring ICU admission. On the other hand, Minashkin et al.’s study [26] on 319 COVID-19 patients found that carriers of the A allele were at a higher risk of developing moderate to severe COVID-19 infection. These incongruent results highlight the importance of further exploring the clinical effect of this SNP. Nevertheless, in our study, none of the candidate genetic polymorphisms, including the rs75603675 A allele, were found to have an association with hospitalization.

We also did not find an association between the ACE2 rs2074192 T allele with COVID-19 susceptibility and disease outcomes despite the fact that this SNP alters the ACE2 protein structure [27] and was previously found to be associated with cardiovascular comorbidities such as hypertension and left ventricular hypertrophy [9]. Our results are consistent with Jalaleddine et al.’s study [28] on 82 COVID-19 patients which found that T allele of the ACE2 rs2074192 polymorphism is not associated with COVID-19 severity. Nonetheless, Sheikhian et al.’s study [27] showed an association between mortality and the TT genotype for 3 different SARS-CoV-2 variants; this was also in line with the UK biobank study of Hamet et al. [10] that found an association between the T allele of this genetic polymorphism and worse outcomes from COVID-19. Therefore, results differ widely between studies emphasizing the need for additional data to verify the role of this SNP.

Regarding the TMPRSS2 rs35074065 polymorphism, we found no associations with neither disease susceptibility nor outcomes, despite the fact that the -- genotype increases the expression of the serine protease on host cells [12, 29]. This effect would hypothetically facilitate SARS-CoV-2 entry into cells by providing additional points of entry on the cell’s surface. Nevertheless, our results are consistent with Vitello et al.’s study [18] on 102 COVID-19 infected and non-infected patients that showed no association between this polymorphism and COVID-19 susceptibility or severity.

Lastly, the OAS1 A allele of the rs10774671 variant was not found to be associated with COVID-19 susceptibility or disease outcomes either. This allele encodes the p42 isoform of the OAS1 that is predominantly targeted to the cytosol rather than the trans-Golgi compartment like the p46 isoform. This cytosolic targeting decreases the protein’s ability to detect viral RNA although its enzymatic function is not affected [30], thus we would expect that the A allele should pose a higher risk of COVID-19 severity rather than having no effect. In fact, Soveg et al.’s [13] showed that the OAS1 p46 isoform encoded by the G allele was found to be protective of severe COVID-19, hence conferring a harm risk to the A allele. It is important to note, however, that clinical studies involving this polymorphism are scarce and additional studies are required to ascertain its clinical effect and relevance.

Combining the effect of the genetic polymorphisms together into a PRS revealed a positive association between PRS and severity. This reflects the magnitude of the effect of the remaining polymorphisms on these variables and highlights the polygenicity of disease-related phenotypes. In addition, and since there is overlap between the various severity categories, a similar PRS trend was revealed with hospitalization (P = 0.05) when compared to severity but may not have reached significance due to the minor differences in category allocations. It is valuable to note that multiple other genetic polymorphisms have been described in the literature to be associated with COVID-19 susceptibility and severity such as VDR, IFITM3, ABO, TYK2 and others [5, 31]. All these additional variants require evaluation with extensive PRS to determine their combined effect, which are still scant. An example of such is Farooqi et al.’s study [32] that found a significant association between multi-ethnic PRS based on leading risk-variants and COVID-19 susceptibility and severity, allowing for prediction of patient outcomes and their stratification. Likewise, Crossfield et al. [33] found an association between the PRS of SNPs involved in the immune-related pathways and severe COVID-19 infection, allowing for better biological understanding of the disease’s outcomes in terms of host factors, specifically genetics.

Limitations

In addition to our previously described limitations for this study’s cohort [17], it is important to note that multiple other polymorphisms described in the literature play a role in determining COVID-19 susceptibility and disease outcomes. These SNPs were not taken into account for the PRS calculation in our study, therefore, limiting its result and interpretation. Additional studies from large samples and Biobanks that include multiple genetic polymorphisms are needed to address these limitations.

Conclusion

In conclusion, despite the aforementioned limitations, this study offers valuable insights into the influence of candidate genetic polymorphisms, whether considered individually or in combination, on COVID-19 susceptibility and outcomes. Our findings highlight the significance of genotyping for these candidate genes to determine their combined effect. As such, by calculating PRS, we can improve the precision of prognosis assessment for individuals vulnerable to the disease. Utilizing this quantification as a means of stratification of COVID-19 prognosis could notably enhance patient care and management. However, further research is imperative to delve deeper into the role of genetics in disease-related phenotypes and to advance our comprehension of this intricate association.