Coronavirus disease 2019 (COVID-19) clinical presentation is heterogeneous, ranging from entirely asymptomatic up to severe cases and death. Another level of heterogeneity is observed regarding persistent symptoms: one study has estimated that the median proportion of individuals who experienced at least one persistent symptom was 73% (Nasserie et al., 2021). Uncovering biomarkers linking patients with distinct prognosis subgroups would be beneficial. Different strategies have been employed to uncover molecular markers predicting odds for better prognosis and severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection susceptibility. Proteins, lipids, and metabolites have already been examined (Praissman and Wells, 2021; Samprathi and Jayashree, 2021). Genetic variability has been shown to be a valuable source for biomarker research. COVID-19 prognosis and infection susceptibility are multifactorial traits determined by the complex interaction of environmental factors and multiple genes. Thus, significant single-gene results may lead to substantial predictors such as the C–C chemokine receptor type five (CCR5) gene association with HIV susceptibility and prognosis (Liu et al., 2012), or ABO blood type and dengue severity (Hashan et al., 2021).

Genetic association studies can be designed within prespecified genes of interest (candidate gene approach) or with a broader strategy characterizing diversity across large genomic areas (genome-wide association studies, whole exome and genome sequencing). Angiotensin-converting enzyme 2 (ACE2), transmembrane serine protease 2 (TMPRSS2), human leukocyte antigen (HLA), interferon-induced transmembrane protein 3 (IFITM3), tumor necrosis factor-alpha (TNF-α), FURIN, and angiotensin I-converting enzyme (ACE1) were the most studied genes using the candidate gene approach in 2020 (Araújo et al., 2021). They all present strong biological plausibility since they act on viral cell entry or human immune response to SARS-CoV-2.

Findings from single association studies must always be considered carefully because of the likelihood of producing spurious outcomes (Sullivan, 2007). Replication is essential before considering using genetic markers in the clinical setting. Although that has been proved hard, inconsistency frequently can be attributed to shortfalls in study design, implementation, and interpretation, with inadequately powered sample groups being of significant concern (Hattersley and McCarthy, 2005). A systematic meta-analytic approach may support estimating population-wide effects of genetic risk factors in human diseases (Ioannidis et al., 2001). The PROSPERO (Moher et al., 2014) database, indicating protocols for systematic reviews, has already been presented for HLA (CRD42021251670) (Deb et al., 2022), ACE2, and TMPRSS2 (CRD42021229963) contribution with COVID-19 outcomes. Therefore, we focused our systematic review on IFITM3, FURIN, ACE1, and TNF-α genetic variants and their association with COVID-19 susceptibility and prognosis to reduce unnecessary duplication.

IFITM3 (MIM 605579; 11p15.5) is a protein-coding gene that disturbs cell entry by inhibiting viral fusion with cholesterol-depleted endosomes (Amini-Bavil-Olyaee et al., 2013); a mechanism also described during SARS-CoV-2 infection (Prelli Bozzo et al., 2021). The IFITM3 rs12252 polymorphism has been associated with influenza severity (Prabhu et al., 2018). The TNF (MIM 191160; 6p21.33) gene encodes a multifunctional proinflammatory cytokine. Although TNF-α is not as relevant as interleukin-6 on the cytokine storm presented in severe COVID-19 patients (Karki and Kanneganti, 2021), anti-TNF-α drug repositioning for COVID-19 has been proposed (Stebbing et al., 2020). FURIN is coded by the FURIN (MIM 136950; 15q26.1) gene. It regulates constitutive exocytic and endocytic pathways and has a central role in SARS-CoV-2 transmission (Peacock et al., 2021). The ACE1 (MIM 106180; 17q23.3) gene produces a protein related to blood pressure regulation and electrolyte balance, and ACE1/ACE2 balance has been suggested to play a pivotal role in the pathobiology and treatment of COVID-19 (Sriram and Insel, 2020). The ACE1 rs4646994 variant is a 287-bp Alu repeat insertion/deletion (indel) on intron 16 known to alter ACE-1 levels and influence several clinical traits (Castellon and Hamdi, 2007). Here, we present the result of a systematic review and, whenever possible, a meta-analysis of IFITM3, FURIN, ACE1, and TNF-α genetic association with susceptibility to SARS-CoV-2 infection and COVID-19 severity.

The systematic review protocol was submitted to PROSPERO (CRD42021268578). Preferred Reporting Items for Systematic reviews and Meta-analysis (PRISMA) was adopted as a guideline for reporting this systematic review (Page et al., 2021). The study selection was carried out in three phases: identification, screening, and eligibility. Search on the PubMed and Scielo databases led to article identification. The PECO question for prognosis was Participants (P) = subjects with COVID-19, Exposition (E) = minor alleles, Control (C) = major alleles of genetic variants, and Outcomes (O) = COVID-19 severity (asymptomatic or severe presentation); while the PECO question for susceptibility was P = overall population, E = minor alleles, C = major alleles of genetic variants, and Outcomes (O) = COVID-19 positive diagnosis. The bibliographic search included all studies published until February 8, 2022, with no language restriction, using the search arguments listed in Supplementary Material SI. Two independent researchers conducted article screening. Inclusion criteria were primary articles covering genetic association of COVID-19 susceptibility or prognosis with IFITM3, FURIN, ACE1, and TNF-α variants, comprising four separate searches. Exclusion criteria were review articles or primary articles evaluating the association of COVID-19 susceptibility or prognosis with other genes.

We assessed study quality using the Q-Genie tool (Sohani et al., 2016) performed by two independent researchers. This instrument contains 11 questions to be marked on a seven-point Likert scale examining several aspects of a genetic association study: scientific basis for the development of the research question, ascertainment of comparison groups (e.g., cases and controls), technical and nontechnical classification of tested genetic variants (e.g., genotyping call rates, blinded experiments), classification of the outcome (e.g., sampling strategy, definition criteria), discussion of sources of bias, appropriateness of sample size, description of planned statistical analyses, statistical methods applied, test of assumptions in the genetic studies (e.g., Hardy–Weinberg equilibrium), and appropriate interpretation of the results. Proposed cutoffs for understanding are ≤35 poor, > 35 moderate, and ≥45 good quality, with the total score ranging from 7 to 77 points.

Meta-analysis was conducted whenever three or more studies were included for the same polymorphism and outcome. We carried out single meta-analyses for each polymorphism considering allelic and genotypic effects (under both allele recessive model assumptions). Heterogeneity between studies was assessed using the chi-square test. We used the metabin function coded on meta package in R (version 4.1.0) (R Core Team, 2014) to estimate overall odds ratios (ORs) and its 95% confidence interval (CI). Original ORs were combined using the Mantel–Haenszel weighted means method under both fixed- and random-effect models. The significance level was set at 0.05.

Twenty-seven studies were included in the systematic review: five with IFITM3 (Zhang et al., 2020; Alghamdi et al., 2021; Cuesta-Llavona et al., 2021; Gómez et al., 2021; Schönfelder et al., 2021), two with Furin (Latini et al., 2020; Torre-Fuentes et al., 2021), three with TNF-α (Saleh et al., 2020; Fishchuk et al., 2021; Heidari Nia et al., 2021), and 17 with ACE1 (Gómez et al., 2020; Aladag et al., 2021; Annunziata et al., 2021; Cafiero et al., 2021; Gunal et al., 2021; Hubacek et al., 2021; Karakaş Çelik et al., 2021; Kouhpayeh et al., 2021; Mir et al., 2021; Möhlendick et al., 2021; Saad et al., 2021; Verma et al., 2021; Akbari et al., 2022; Gong et al., 2022; Mahmood et al., 2022; Papadopoulou et al., 2022). (Figure 1). Inconsistencies in reported frequencies were found in two studies (Gómez et al., 2021; Karakaş Çelik et al., 2021).

All manuscripts but one reached moderate or good quality scores in the Q-Genie analysis (Supplementary Material S1). Among the 11 questions, it is clear that all studies had the worst performance for questions number 5 and 10. While question 5 examines reported information regarding how genotyping was conducted (e.g., blinded experiments, batch effects), question 10 evaluated whether genetic relationships among subjects were tested, and sex and ethnicity were stated.

Five meta-analyses were carried out, including 22 studies evaluating three genes (IFITM3 n = 3, TNF-α n = 3, and ACE1 n = 16). Twelve studies, including 2,318 control subjects and 5,194 COVID-19 positives, evaluated the ACE1 rs4646994 association with COVID-19 susceptibility (Table 1). Significant heterogeneity was observed for all genetic models with no significant association under the random model (Figure 2). Similar findings were detected in the meta-analysis of the ACE1 rs4646994 variant with asymptomatic presentation (Table 2), indicating no significant effect pooled from three studies (Figure 3). We observed high heterogeneity in the sampling places and reported ethnic backgrounds.

We were able to conduct a meta-analysis investigating whether ACE1 rs4646994 polymorphism could predict COVID-19 severity. Eleven studies were included reaching a total of 692 individuals with severe COVID-19 and 1,433 with nonsevere manifestation (Table 2). The allelic association was observed with increased odds for deletion (D) allele compared with I-allele (pooled OR: 1.45; 95% CI: 1.26–1.66) (Figure 4A). Homozygous deletion (D/D) carriers showed 49% increased odds to present severe COVID-19 compared with heterozygous (D/I) and homozygous insertion allele (I/I) carriers combined (pooled OR: 1.49, 95% CI: 1.22–1.83) (Figure 4B). On the other hand, the I/I genotype was protective against severe COVID-19 (pooled OR: 0.57, 95% CI: 0.45–0.74) (Figure 4C).

The IFITM3 rs12252 meta-analysis with severity included three studies totaling 308 individuals admitted to an intensive care unit and 726 who were not admitted (Table 3). No significant association was observed under any genetic model (Figure 5). Meta-analysis for other outcomes with the IFITM3 rs12252 could not be conducted. The TNF-α rs1800629 association with death was analyzed in three studies (Table 4), including 111 subjects who died and 1,095 survivors. No significant association was observed under the random-effect models (Figure 6). FURIN (Table 5) genetic variants had less than three studies; therefore, no meta-analyses were carried out.

We conducted a systematic review followed by meta-analysis including studies covering genetic association of COVID-19 susceptibility or prognosis with IFITM3, FURIN, ACE1, and TNF-α variants. Four studies included in the meta-analyses did not report the sample collection date, which is of particular interest in COVID-19 studies due to the emergence of variants of concern (VOCs) in the last part of 2020 (Konings et al., 2021). Some VOCs have been associated with higher viral load, worse prognosis, and lethality (Davies et al., 2021; Faria et al., 2021), thus, confounding factors when evaluating genetic effects. Age can also be a confounding factor for COVID-19 association analysis (Fernández Villalobos et al., 2021). Most studies failed to conduct age-corrected estimation or even describe age separately for case and control groups. The same trend was observed for comorbidities (data now shown).

Ancestrality could also contribute to COVID-19 outcomes. Several studies do not present the ethnic background or, at least, the place of birth of the included subjects. Although heterogeneity was seen in parameters associated with ancestrality, the literature fails on genetic background diversity, an issue already raised for genomic data before (Popejoy and Fullerton, 2016). Another literature issue that needs attention is the selective reporting biases leading to the more likely publication of positive findings (Munafò et al., 2009; Sagoo et al., 2009).

We did not find an association of IFITM3 rs12252 with Covid-19 severity. Our results corroborate the most extensive association study published to date since no significance was reported on any of chromosome 11 loci (Niemi et al., 2021). However, the second evaluated polymorphism, the ACE1 rs4646994, showed significant effects with homozygous D carriers presenting higher odds of developing severe COVID-19. Several hits on the large arm of chromosome 17 have been previously reported (Niemi et al., 2021), although their genomic location is too far to hypothesize linkage disequilibrium. It is important to note that genome-wide data may find hits on loci that not necessarily are the ones harboring the causative variants because of its experimental design (Spencer et al., 2009). Furthermore, candidate-gene, whole-exome or whole-genome sequencing studies are more suitable in exploring large indel variants.

The ACE1 rs4646994 has been associated with several clinical phenotypes, including COVID-19 (Castellon and Hamdi, 2007; Li et al., 2021). Most previous findings report associations with COVID-19 outcomes on a population level, indicating high variability on allelic frequencies across different populations (Delanghe et al., 2020; Pati et al., 2020; Yamamoto et al., 2020). On a molecular level, expression results indicate increased levels of ACE1 in D-allele carriers (Suehiro et al., 2004) with increased angiotensin II production (Hamdi and Castellon, 2004) and decreased ACE2 protein levels in lung tissue, thereby potentially affecting infectivity by SARS-CoV-2 (Jacobs et al., 2021). Our group has previously indicated that lower ACE2 levels may increase the risk of COVID-19 respiratory distress (Rossi et al., 2021). Although there is a robust biological hypothesis linking ACE1 rs4646994 with COVID-19, further reports are needed to understand better whether ACE1 variants could contribute to COVID-19 severity. Moreover, studies are still required to adequately evaluate IFITM3, FURIN, and TNF-α genetic variants’ role in COVID-19 susceptibility and outcomes.