COVID-19: The Effect of Host Genetic Variations on Host-Virus Interactions.

Spurred into action by the COVID-19 pandemic, the global scientific community has, in a short of period of time, made astonishing progress in understanding and combating COVID-19. Given the known human protein machinery for (a) SARS-CoV-2 entry, (b) the host innate immune response, and (c) virus-host interactions (protein-protein and RNA-protein), the potential effects of human genetic variation in this machinery, which may contribute to clinical differences in SARS-CoV-2 pathogenesis and help determine individual risk for COVID-19 infection, are explored. The Genome Aggregation Database (gnomAD) was used to show that several rare germline exome variants of proteins in these pathways occur in the human population, suggesting that carriers of these rare variants (especially for proteins of innate immunity pathways) are at risk for severe symptoms (like the severe symptoms in patients who are known to be rare variant carriers), whereas carriers of other variants could have a protective advantage against infection. The occurrence of genetic variation is thus expected to motivate the experimental probing of natural variants to understand the mechanistic differences in SARS-CoV-2 pathogenesis from one individual to another.

Introduction

The worldwide spread of the coronavirus disease 2019 (n class="Disease">COVID-19),[1] caused by the novel virus severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2),[2] has resulted in 1.33 million deaths globally, having advanced at a staggering rate of ∼22 000 new infections and ∼385 deaths every hour (WHO, November 18 situation report). During this health crisis, a knowledge of the risk factors of COVID-19 infection is valuable for determining the most appropriate measures, especially given our limited resources, to mitigate the threat. That is, in addition to efforts toward the discovery of a cure (e.g., vaccines or therapeutics), the identification of individual risk for separating low-risk from high-risk individuals is also important (e.g., in maintaining a minimal workforce to prevent the global economy from plunging into a depression). Determining the relationship between the genetic background of an individual and the susceptibility to and progression of the disease remains an important task in understanding and determining an individual’s risk for COVID-19. Understanding the molecular basis of susceptibility is of vital interest and has been the subject of a number of important studies.[3−10] The emerging picture on susceptibility from these studies suggests that multiple genetic factors could contribute to the risk of SARS-CoV-2 infection and the severity of COVID-19 disease.[11,12] In general, proteins engaged in pathways related to the viral life cycle and host defense are important components of the genetic factors.[11,12] Polymorphic variants of human proteins engaged in these pathways are thus expected to make important contributions toward COVID-19 susceptibility.[11,12] Therefore, a discussion is presented here on polymorphic variants (carried by an individual) of pathway-specific proteins that have the potential to link the susceptibility to the disease to the genetic background of an individual.

This discussion has only been possible due to the availability of results from a number of molecular studies (Table ) carried out in the last 8 months at a very rapid pace imical">n several different discin class="Gene">plines, providing critical advances in understanding the molecular mechanisms of COVID-19 pathogenesis. Despite the health crisis, the efforts of scientists in unraveling the molecular pathology to combat the crisis through breakthroughs (Table ) in a short time have been unparalleled. These results make possible an exploration of the possible relationship between the genetic background of an individual and the susceptibility to and the progression of the disease, which would be valuable to better understand COVID-19 pathogenesis. Genome-wide association study (GWAS) in the human population[13−16] has been the method of choice for identifying genetic risk factors associated with infectious diseases (e.g., mapping of HLA peptide binding cleft amino acid variants for disease susceptibility[14]). Although powerful, GWAS typically maps genetic loci (i.e., not genes necessarily), and probing the molecular mechanism may not be straightforward for a list of genetic loci. GWAS generally also fails to identify rare alleles. In general, GWAS of infectious disease susceptibility has been used for pathogens that have coexisted with human civilization since ancient times.[17] On an evolutionary time scale, the fixation of alleles providing selective advantage[18,19] by natural selection has influenced host susceptibility. COVID-19, in comparison, is a very recent threat, and the possibility of rare alleles providing selective advantage demands exploration. Consistent with the well-known fact that most human genes have several natural variants, the genes identified as interacting with SARS-CoV-2 also have several variants. Thus the question arises of whether the genotype of an individual could have a bearing on COVID-19 infection and prognosis. This question stems from previous observations of natural variants found to afford the host protection against specific pathogens. For example: (a) HBB missense variant against Plasmodium falciparum,[20] (b) CCR5 deletion variant against HIV-1,[21,22] (c) FUT2 variants against the Norwalk virus,[23] (d) SLC4A1 variants against Plasmodium falciparum,[24] and so on. CCR5Δ32 (i.e., CCR5 coreceptor deletion variant) impairs the T-cell entry of HIV-1, just as FUT2 and SLC4A1 variants impair the host-cell entry by the respective pathogens. In other words, certain natural variants of a host protein have the potential to influence the clinical outcome of a specific host–pathogen interaction if the host protein is essential to the pathogen life cycle. A recent in silico study on ACE2 variants found in Iranian ethnic groups[25] suggests that ACE2 variants could possibly alter the receptor’s interaction affinity, implying that subpopulations could have differences in intrinsic susceptibility to COVID-19.[25] Whereas the above example variants are known to provide protection against the pathogens, examples of variants responsible for severe infectious disease symptoms have also been reported[12] (e.g., herpes simplex virus 1 encephalitis,[26] viral respiratory infection[27] and influenza pneumonitis[28,29] in children, etc.). Many of the variants responsible for severe infectious disease symptoms are rare,[12,26−29] and the catalogues of rare variants, now available from next-generation sequencing studies of large populations (e.g., gnomAD[30]), therefore demand careful discussion in the context of disease susceptibility. In addition, recent pharmacogenomic studies[31,32] of receptor–drug interactions also emphasized the importance of rare variants, and like receptor–drug interactions, host-protein rare variants demand attention in the context of virus–host-protein–protein interactions as well. With several human proteins identified in the reports in Table as being likely to play roles in the SARS-CoV-2 life cycle, the analysis of the variants of these proteins thus holds promise for relating the genetic backgrounds of patients to the clinical outcomes, complementing GWAS.[9] On the basis of the observations of some of the reports in Table , the discussion on variants is presented for host-protein machineries that are engaged in (a) viral entry, (b) innate immunity, and (c) virus–host-protein–protein and RNA–protein interactions to complement other discussions[11,12,33,34] on genetic variants in relation to SARS-CoV-2 infections.

Table 1

Breakthrough Reports on Molecular Mechanisms of SARS-CoV-2 Pathogenesis

study topic	examples of specific reports
SARS-CoV-2 spike (S protein) function, structure, and stabilization	(a) Discovery of the proteolysis of the S protein prior to cell fusion[35−38]
	(b) Structure determination of the S protein[39] and in situ structural analysis[40] of profusion spikes of intact[41] or inactivated virus[42]
	(c) Design of profusion stabilized,[43] thermostabilized,[44] conformationally locked,[45] or labeled[46] S protein
	(d) Deep mutational scanning[47] and analysis of S-protein variants for infectivity and antigenicity[48−51]
human ACE2 function and structure	(a) Identification of the ACE2 as the cellular receptor for the S protein[36,52,53] for membrane fusion
	(b) Determination of the tissue-level expression of the receptor and fusion-associated human proteins[54−56] and population level expression of proteins[10,11]
	(c) Structure elucidation of the interactions of the S protein with ACE2 (also chaperon for B0AT1) in complex with[57] or without[58−60] the client protein B0AT1
convalescent patient anti-S protein antibodies	(a) Identification of neutralizing antibodies (using single-cell RNA and VDJ sequencing of antigen-binding B cells[61]) from patients[61−64] and development of novel virus neutralizations tests (VNTs)[65,66]
	(b) Structure determination of these antibody–S-protein complexes[61,62,67−71]
high-affinity proteins targeting S-protein RBD	(a) Design of decoy ACE2 receptor[72] and de novo design of small proteins with picomolar affinity[73]
	(b) Design of nanobodies,[74] bivalent VH domains,[75] antibody,[76] and IgA monoclonal antibody[77]
structure and function of other SARS-CoV-2 proteins	(a) Structure determination of RNA polymerase[78−80] and 2′-O-methyltransferase[81]
	(b) Structure determination of Nsp1[82,83] and M protease[84,85]
drug discovery	(a) Drug discovery through compound repurposing[86,87]
	(b) Small-molecule targeting of SARS-CoV-2 proteins[84,85,88,89]
proteomics	(a) Development of system-level virion proteomics[90] and phospho-proteomics survey of infected cells[91]
	(b) High-throughput experimental analysis of virus–host-protein–protein[92] and RNA–protein[93,94] interactions
animal models	(a) Creation of transgenic mice expressing human ACE2[95,96] or mouse-adapted redesigned RBD[97]
	(b) Use of golden hamster[98] and other model animals[99] to study transmission and pathogenesis
analyses of patients	(a) Clinical analyses of focused patients[100−103] for serological signatures[104]
	(b) Analysis of adaptive immune response,[105,106] host factors,[107] and innate immunity[3−8]
vaccines	(a) Lipid nanoparticle-encapsulated mRNA vaccine encoding RBD[108,109] or stabilized S protein[110]
	(b) Adenovirus vector-based vaccine expressing S protein[111,112] and recombinant S protein as vaccine[113]

Human Protein Variants with Potential to Influence Viral Entry

Just as several of the above examples concern pathogen entry, variants related to SARS-CoV-2 entry are of great imical">nterest imical">n understandimical">ng differences imical">n pathogenesis. n class="Species">Human angiotensin-converting enzyme 2 (ACE2) is the cellular receptor of the SARS-CoV-2 Spike (S) protein,[36,52,53] and the ACE2 protein serves as the critical site for viral attachment. ACE2, a membrane-bound, counter-regulatory carboxypeptidase (responsible for the proteolysis of angiotensin I/II, neurotensin, kinetensin, and apelins),[114,115] is an essential component of the renin–angiotensin hormone system, playing a critical role in cardiovascular homeostasis.[116] Understandably, the probability of a loss-of-function-intolerant (pLI[117]) score for ACE2 is 0.99 (Figure A), indicating that ACE2 truncation variants are less likely to be tolerated and that the virus utilizes an essential human protein for entry. In contrast, the pLI scores for CCR5 and FUT2 are both 0.0, suggesting higher tolerance for deletion/truncation, which is consistent with the presence of deletion/truncation variants of CCR5 and FUT2, even as homozygotes, in the human population being known to provide protective advantages against the respective pathogens. For ACE2, we therefore predominantly focus on its missense variants, in particular, those distinct from the eight variants discussed in the Iranian population study.[25]ACE2 is located on the X chromosome, and thus men are hemizygous for ACE2 variants. The hemizygous nature of ACE2 variants demands attention in the context of SARS-CoV-2 interactions. ACE2 is also included among genes that escape X chromosome inactivation (XCI)[118] (i.e., expressed from both the active and inactive X chromosomes in females), and the contributions of ACE2 variants toward SARS-CoV-2 interactions in females are therefore less likely to depend on XCI. Figure A summarizes the gnomAD variants of ACE2 (along with those of other proteins likely engaged in viral entry). The variants of the proteins are discussed in this section.

Figure 2

Variants of protein machinery engaged in viral entry: (A) List of genes, their chromosomal location, the UniProt accession number of the corresponding protein, pLI and pNull probabilities, and the number of different gnomAD variants observed for each gene. Because the sum (pLI + pNull + pRec = 1) is one, only pLI and pNull probabilities are provided. (B) ACE2 nonmissense variants, stop-gained (red), frameshift (blue), and inframe deletion/insertion, are listed (top) and mapped onto the protein (bottom) showing the peptidase domain (PD) and the transmembrane (TM) regions. The listed variants (top) includes nucleotide position and substitution, mutation type, allele frequency, dbSNP IDs, and the subpopulation in which the variant is identified (NF, non-Finnish). (C) Stop-gained variants (red) of other proteins, B0AT1 (blue), TMPRSS2 (red), ADAM17 (purple). The dbSNP IDs are in parentheses. (D) Cartoon representing how B0AT1 could compete/interfere with TMPRSS2 and ADAM17 proteolysis of ACE2. The cartoon was inspired from the study of Brest et al.,[11] and for convenience, the proteins are colored in the same way as in panels B and C.

ACE2 Variants

Topologically, a single transmembrane helix of ACE2 separates the extracellular peptidase domain (PD) from the cytoplasmic domain.[57] The PD is generally referred to as the receptor for coronaviruses (CoVs), and the C-terminal domain of S protein, referred to as the receptor-binding domain (RBD), physically interacts with the PD.[57−60] Recent structures of the PD–RBD complex[57−60] highlight the similarities and differences in receptor recognition between SARS-CoV-2 and other CoVs.[58,59] To appreciate the possible consequences of human PD sequence variants, it is important to look at how changes in mammalian ACE2 sequence influence SARS-CoV-2 RBD recognition.[119,120] Genetic variants are discussed in the same context as the PD sequence.[119,120] SARS-CoV-2 infects humans, bats, pigs, civets,[52] and golden hamsters[98] but not mice.[52] This observation was the key reason for creating a transgenic mouse (a cost-effective animal model[95]) expressing human ACE2[95,96] to study COVID-19 pathogenesis. A comparison of the mouse PD sequence with those of SARS-CoV-2 hosts (Figure A) sheds light on how the infection may be mediated by PD–RBD complexation. Although sequences of human, bat, pig, hamster, and civet PDs do differ from one another, they still interact with the SARS-CoV-2 RBD, whereas subtle changes in the mouse PD sequence likely compromise the interaction (Figure A). On the basis of the structure of the human ACE2 PD–RBD complex,[57−60] the mouse PD has conservative (with respect to the human) substitutions (D30N, Y83F, and K353H) at three hydrogen-bonding interfacial sites (Figure ), which together are likely sufficient to negatively perturb the interaction. This observation suggests that subtle changes in hydrogen bonding and charged residues at interfacial sites could impact the receptor–S-protein interaction. This is also consistent with adapting mouse ACE2 for the SARS-CoV-2 RBD interaction by a few point mutations.[97] The predicted loss of binding free energy (ΔΔG) for the three mouse PD substitutions is 0.34, 0.73, and 1.04 kcal/mol (using BeAtMuSiC[121,122]). Given the behavior of PD interfacial residues, Genome Aggregation Database (gnomAD,[30] an unbiased large population database) missense variants that map to the PD–RBD interface are of great interest. Six human ACE2 missense variants, including p.Ser19Pro (PolyPhen score,[123] 0.767), p.Ile20Val (0.000), p.Thr27Ala (0.000), p.Glu37Lys (0.712), p.Gly326Glu (0.090), and p.Glu329Gly (0.027), map to the interface (Figure B). The low PolyPhen scores (<0.9) indicate that these variants are unlikely to be deleterious (i.e., unlikely to perturb the essential peptidase function, being distal from the catalytic site). Ser19 is engaged in hydrogen bonding with the SARS-CoV-2 RBD backbone atoms, whereas the hydrogen bonding of Glu329 is prominent in the SARS-CoV RBD (Figure B). In addition, p.Glu37Lys and p.Glu329Gly result in interfacial charge alteration (Figure B). Ser19, Thr27, and Glu37 also engage in intramolecular hydrogen bonds, and variants at these positions could perturb PD–RBD interactions. The predicted ΔΔG for these variants (0.04, 0.18, 0.92, 0.34, 0.62, and 0.33 kcal/mol) is also within the range of the mouse PD substitutions. Given the known role of hydrogen bonds and charged residues in the mouse PD, these variants could contribute to altering the PD–RBD affinity.

Figure 1

Missense variants at ACE2 interfaces: (A) Sequence comparison of the ACE2 peptidase domain (PD) for various mammalian species in the interfacial region with the SARS-CoV-2 S-protein receptor-binding domain (RBD). Human ACE2 PD positions in contact with the SARS-CoV-2 RBD (gray filled circle) showing that the mouse PD sequence differs at three hydrogen-bonding positions (red). Positions of gnomAD missense variants in contact with SARS-CoV-2 (green) and SARS-CoV (blue) are shown. gnomAD variants (e.g., pSer19Pro) are above the green/blue bars. (B) Structure (PDB 6M17) of the human ACE2 dimer (black and gray) in complex with the SARS-CoV-2 receptor-binding domain (RBD, green) and B0AT1 protein (blue), with the membrane represented by dotted lines. The distinct interfaces (numbered 1–3, right) are for PD–RBD (heterodimer), ACE2–ACE2 (homodimer), and ACE2–B0AT1 (heterodimer), respectively. (C) Positions (pink) of variants (e.g., p.Glu37Lys) at the three interfaces in panel A are shown.

Missense variants at ACE2 imical">nterfaces: (A) Sequence comparison of the ACE2 peptidase domain (PD) for various mammalian species in the interfacial region with the SARS-CoV-2 S-protein receptor-binding domain (RBD). Human ACE2 PD positions in contact with the SARS-CoV-2 RBD (gray filled circle) showing that the mouse PD sequence differs at three hydrogen-bonding positions (red). Positions of gnomAD missense variants in contact with SARS-CoV-2 (green) and SARS-CoV (blue) are shown. gnomAD variants (e.g., pSer19Pro) are above the green/blue bars. (B) Structure (PDB 6M17) of the human ACE2 dimer (black and gray) in complex with the SARS-CoV-2 receptor-binding domain (RBD, green) and B0AT1 protein (blue), with the membrane represented by dotted lines. The distinct interfaces (numbered 1–3, right) are for PD–RBD (heterodimer), ACE2–ACE2 (homodimer), and ACE2–B0AT1 (heterodimer), respectively. (C) Positions (pink) of variants (e.g., p.Glu37Lys) at the three interfaces in panel A are shown.

Variants of protein machinery engaged in viral entry: (A) List of genes, their chromosomal location, the UniProt accession number of the corresponding protein, pLI and pn class="Chemical">Null probabilities, and the number of different gnomAD variants observed for each gene. Because the sum (pLI + pNull + pRec = 1) is one, only pLI and pNull probabilities are provided. (B) ACE2 nonmissense variants, stop-gained (red), frameshift (blue), and inframe deletion/insertion, are listed (top) and mapped onto the protein (bottom) showing the peptidase domain (PD) and the transmembrane (TM) regions. The listed variants (top) includes nucleotide position and substitution, mutation type, allele frequency, dbSNP IDs, and the subpopulation in which the variant is identified (NF, non-Finnish). (C) Stop-gained variants (red) of other proteins, B0AT1 (blue), TMPRSS2 (red), ADAM17 (purple). The dbSNP IDs are in parentheses. (D) Cartoon representing how B0AT1 could compete/interfere with TMPRSS2 and ADAM17 proteolysis of ACE2. The cartoon was inspired from the study of Brest et al.,[11] and for convenience, the proteins are colored in the same way as in panels B and C.

Position-specific amino acid enrichment/depletiomical">n imical">n the n class="Gene">ACE2 sequence observed by Chan et al.[72] in the designed decoy ACE2 receptor by deep mutagenesis[72] is also an important reference for predicting the expected behavior of ACE2 missense variants. In the deep mutagenesis study, the substitutions corresponding to three gnomAD missense variants (p.Ser19Pro, p.Thr27Ala, and p.Thr92Ile) were observed to be highly enriched[72] (i.e., Pro for Ser19, Ile for Thr92), suggesting that these variants are anticipated to enhance PD–RBD affinity. Asn90 and Thr92 constitute the N-glycosylation motif of the ACE2 protein, and substitutions at both positions were highly enriched, suggesting that N-glycans at these positions likely interfere with RBD binding,[72] whereas the proline substitution at Ser19 (the first turn of the interfacial helix) likely entropically stabilizes the helix.[72] Substitutions associated with a number of missense variants showing medium-range depletion (p.Glu35Lys, p.Glu37Lys, p.Asn51Ser, p.Lys68Glu, p.Phe72Val, p.Gly326Glu, p.Glu329Gly, p.Gly352Val, and p.Gln388Leu) or enrichment (e.g., p.Lys26Arg, p.Asn64Lys, p.Gln102Pro and p.His378Arg) were also observed,[72] suggesting that these variants could contribute to weakening or enhancing the PD–RBD affinity, respectively. The dose of a viral inoculum[124] is an important factor for infection or severity of infection; a weakened/enhanced affinity could influence the dose required for infection in individuals carrying variants, and these observations encourage the investigation of the consequences of host variants in SARS-CoV-2 pathogenesis.

These variants, however, are rare (gnomAD[30] allele frequencies, 10–3 to 10–5) and are unlikely to occur in ACE2 at the same time. The occurrence frequency of these variants also varies between n class="Species">human subpopulations. For example, in gnomAD,[30] the p.Thr27Ala variant is seen only in the Latino population (allele frequency, 7.3 × 10–5), p.Glu37Lys is observed in the Finnish European (3.2 × 10–4) and African (1.0 × 10–4) populations, and p.Ser19Pro is observed only in the African (3.3 × 10–3) population. With a population of ∼1.3 billion in Africa (and assuming 50% men), the p.Ser19Pro variant, with a frequency of 3.3 × 10–3, is likely to be present in a significant number of people. In women, these variants are likely to occur predominantly as heterozygotes (ACE2 escapes XCI). In heterozygotes, oligomeric protein complexation can be affected (e.g., the sickle-cell trait of heterozygotes[20] or oligomeric haptoglobin of heterozygotes and homozygotes[125]). The recent structural observation of ACE2 chaperoning B0AT1 (also known as SLC6A19) suggests that ACE2 likely dimerizes within the membrane. The positions of the PD–RBD interface variants are distal from the homodimer interface, suggesting that the variants are also less likely to affect homodimerization. The ACE dimers in heterozygous women could also be affected by PD–RBD interface variants. The ACE2–B0AT1 interface is also important in this respect, especially in the kidney and the intestine, where B0AT1 is predominantly expressed. Proteolytic cleavage of ACE2 by TMPRSS2 for the endosomal release of SARS-CoV-2 is an important step, and B0AT1 could possibly compete with TMPRSS2 for ACE2 access (Figure D). ACE2–B0AT1 interfacial variants could therefore also indirectly influence pathogenesis. The ACE2 transmembrane helix is the primary site of B0AT1 interaction[57] (Figure B), and four ACE2 missense variants (p.Pro734Leu, pIle741Val, p.Val745Ile, and p.Ile753Met) can be mapped onto the ACE2–B0AT1 interface (Figure C), with predicted ΔΔG values in a similar range as above. Among the four variants, only p.Pro734Leu has a PolyPhen score of 1.0 (i.e., likely deleterious), and this variant is seen in the South Asian population with an allele frequency 1.1 × 10–4 (also observed in 1 out of ∼5000 South Asian men in gnomAD). That is, despite being predicted to be deleterious, men possessing the variant are known to have survived. It is, however, unclear whether these variants could interfere with B0AT1 interactions to influence TMPRSS2 access. Missense variants (p.Asn638Ser and p.Arg710His) at the ACE2 homodimer interface (Figure C) could also influence viral entry (i.e., homodimer interface disruption could enhance TMPRSS2 accessibility). Asn638 and Arg710 are engaged in ACE2 intersubunit hydrogen bonding and salt bridge formation, respectively. In particular, p.Arg710His has a PolyPhen score of 1.0 (i.e., deleterious). p.Arg710His is observed in non-Finnish Europeans (also observed in 1 out of ∼40 000 Europeans) and the East Asian population, with allele frequencies of 6.1 × 10–5 and 1.5 × 10–4, respectively. The predicted ΔΔG of these two naturally occurring mutants is 1.95 and 0.24 kcal/mol, respectively, and they would be useful in revealing mechanistic details of the endosomal release of the trimeric S protein docked onto the dimeric ACE2 protein, similar in flavor to a previous in vitro mechanistic study[126] of HIV-1 entry disruption by CCR5 nonsense variant C100X (a rare allele, 6.6 × 10–4). Similarly, one of the ACE2 disulfide (Cys133–Cys141) disrupting variants (p.Cys141Tyr, 5.8 × 10–6) would be useful. In addition to the ACE2 missense variants discussed here in the context of ACE2 structure and complex, a discussion of ACE2 nonmissense variants would also be very important. gnomAD reports ACE2 nonmissense variants (Figure A, B), although only a few due to high pLI score. There, respectively, are two, one, and two ACE2 rare nonsense (stop-gained, 5.9 × 10–6 and 6.4 × 10–6), frameshift (5.4 × 10–6), and inframe deletion (1.1 × 10–5 and 5.6 × 10–6) variants reported in gnomAD (Figure B). All nonmissense variants map to PD, suggesting that PD truncation and structural perturbations from the nonmissense variants are likely to negatively impact PD–RBD interactions and viral entry in carrier individuals.

Auxiliary Protein Variants

In addition to ACE2 (which directly imical">nteracts with n class="Species">SARS-CoV-2 proteins), other human proteins could also influence ACE2–S-protein interactions (Figure A,D). For example, interactions of B0AT1 with ACE2 could possibly compete with TMPRSS2 and ADAM17 to influence ACE2 proteolysis (Figure D). Proteins (e.g., B0AT1) with the potential to exert indirect influence on host–viral protein interactions (e.g., ACE2–S-protein) are referred to as auxiliary proteins here for convenience. The list of auxiliary proteins (e.g., B0AT1, TMPRSS2, ADAM17, etc.) that likely influence viral entry is provided in Figure A, highlighting different categories of variants associated with these auxiliary proteins. In this regard, B0AT1 missense variants at the ACE2–B0AT1 interface would be important in influencing the proteolytic outcome of ACE2. Six out of the 14 B0AT1 missense variants at the ACE2–B0AT1 interface have 30 Å2 or more of their surface area in contact with ACE2, with the predicted ΔΔG of four of these being quite high (1.20 to 2.02 kcal/mol). Of these four, p.Arg214Gly is engaged in intermolecular hydrogen bonding and is carried by non-Finnish Europeans (allele frequency, 6.1 × 10–5) and African (3.5 × 10–5) populations. Unlike ACE2, pLI for B0AT1 is 0.0 (and pNull = 1.0, see later), and several B0AT1 stop-gained and other nonmissense variants are observed in gnomAD (Figure A, C). Truncations/structural perturbations of B0AT1 due to nonmissense variants are likely to favor ACE2 proteolysis.

Analyzing expression quantitative trait loci (eQTL) polymorphic variants for examining ACE2 expressiomical">n imical">n the genotype-tissue expressiomical">n (GTEx) database, Chen et al.[10] imical">n a recent report showed that at the populatiomical">n level, n class="Gene">ACE2 expression is negatively correlated with COVID-19 severity.[10] To substantiate the counterintuitive observation, follow-up in silico expression analyses by Brest et al.[11] of variants of not only ACE2 but also the transmembrane proteases TMPRSS2 and ADAM17 suggest that protease expressions are likely to also contribute to disease severity.[11] The logical support for the argument[11] is that ACE2 is shed from membranes upon proteolysis[127] (Figure D). Proteolysis by ADAM17 results in ACE2 shedding into extracellular space, as ADAM17 operates on ACE2 and not the S protein.[11,127] TMPRSS2, on the contrary, functions on ACE2 as well as the S protein, leading to endosomal release (i.e., virus uptake) upon membrane fusion.[11] The authors suggest that for low viral loads, ACE2 shedding could serve as a barrier for infection, whereas for high viral loads, proteolysis would lead to infection[11] (Figure D). In short, these proteases and their variants are likely to play an important role in the clinical outcome. The variants probed by human population level GTEx expression analyses are genome variants (i.e., not exome variants necessarily) typically occurring within ±10 kb of the gene of interest with minor allele frequencies between 0.05 and 0.5. The focus here is on rare (allele frequencies, 10–3 to 10–5) exome variants, as these variants have the potential (see later) to influence clinical outcomes of an individual (with the prospect of understanding mechanisms from protein analysis), complementing the population level study. The list of gnomAD rare nonmissense variants for TMPRSS2 and ADAM17 shows several variants (Figure A,C), especially for TMPRSS2 (pLI = 0.0), that could influence viral entry.

Human Protein Variants with Potential to Influence Immune Response

Post-viral-entry protein machinery engaged in the host immune response is discussed next, as a number of recent studies[3−8] point to the contributions of innate immunity toward disease severity. For example, an important recent study from the Casanova laboratory[5] reports that ∼10% of patients with life-threatenimical">ng n class="Disease">COVID-19 pneumonia have neutralizing autoantibodies (auto-Abs) against interferons,[5] whereas the anti-interferon auto-Abs are absent in asymptomatic patients.[5] Interferons (cytokine subgroup) are key signaling proteins of innate immunity, especially in viral infections. It had previously been shown that phenocopies of auto-Abs against cytokines are similar to clinical phenotypes of germline variants of cytokines and cytokine-receptors.[5] In other words, the absence of functional innate immunity signaling proteins due to either germline mutations or auto-Abs against them could compromise the innate immune response, leading to a severe infectious. Here variants of proteins associated with innate immunity signaling pathway for the production of interferons and other cytokines (e.g., interleukins) are discussed in regard to COVID-19 disease severity.

Toll-like Receptor 7 (TLR7) Variants

van der Made et al.[3] reported an important case study highlighting the role of monogenic rare genetic variants in severe COVID-19 imical">n four young n class="Species">men (i.e., two pairs of brothers) from unrelated families with no prior history of major chronic diseases.[3] Rapid whole genome sequencing of the four (including one deceased) paired brother patients showed putative loss-of-function mutations in TLR7.[3] TLR7 functions as a pattern recognition receptor (e.g., recognizing ssRNA) and plays an important role in the viral (e.g., RNA viruses) immune response. TLR7 is encoded from the X chromosome. The two TLR7 variants found in the brothers are frameshift p.Gln710Argfs*18 (in one family) and missense p.Val795Phe (in the other) variants that map to the leucine-rich repeat (LRR) structural region of TLR7 (Figure A,B). These variants are not observed in gnomAD, suggesting that the two variants are rare (i.e., occur with very low allele frequency to be observed in ∼140 000 individuals in gnomAD). The functional assay of primary immune cells of the patients (i.e., carrying the TLR7 variants) by probing with TLR7 agonist showed the defective regulation of type-I IFN-related genes in comparison with normal cells.[3] This suggested that the functional (and likely structural) consequences of the rare variants compromise TLR7 (and downstream innate immunity signaling protein) functions, further underscoring the importance of the study of rare variants. The p.Gln710Argfs*18 frameshift variant is expected to alter the spacing of nonpolar residue patterns at distinctive intervals within the 22–29 residue repeating LRR motif, and p.Gln710Argfs*18 is thus expected to perturb the LRR fold. Given the functional consequence of the LRR frameshift (p.Gln710Argfs*18) variant in the above patient brothers, the identification of TLR7 nonmissense variants is of great importance. pLI for TLR7 is 0.98; therefore, only a few nonmissense (two stop-gained, three frameshift, and three inframe deletion) TLR7 variants are observed in gnomAD (Figure A), and all eight TLR7 nonmissense variants map to LRRs (Figure A). These nonmissense variants are likely to similarly compromise TLR7 function, and men carrying these nonmissense variants could be at risk for severe COVID-19.

Figure 3

TLR7 variants: (A) TLR7 nonmissense variants (as described in Figure B) are listed (top) and mapped onto the protein (bottom) showing the leucine rich region (LRR), the transmembrane (TM) and Toll/IL-1 receptor (TIR) regions. The mutants shown with a yellow background are reported in the van der Made et al.[3] study of brothers with severe symptoms. (B) Rhesus macaque TLR7 dimer structure with bound ssRNA (top) showing human TLR7 missense variants at interfacial positions and glycosylation site (bottom). (C) Select missense variants within TLR7 LRR (right) are mapped onto the 22–29 residue LRR structural motifs (middle) with the p.Val795Phe variant shown within the structural motif (left). TLR7 LRR structural motifs (1–27) were superposed using MUSTANG[128] structural alignment.

TLR7 variants: (A) n class="Gene">TLR7 nonmissense variants (as described in Figure B) are listed (top) and mapped onto the protein (bottom) showing the leucine rich region (LRR), the transmembrane (TM) and Toll/IL-1 receptor (TIR) regions. The mutants shown with a yellow background are reported in the van der Made et al.[3] study of brothers with severe symptoms. (B) Rhesus macaque TLR7 dimer structure with bound ssRNA (top) showing human TLR7 missense variants at interfacial positions and glycosylation site (bottom). (C) Select missense variants within TLR7 LRR (right) are mapped onto the 22–29 residue LRR structural motifs (middle) with the p.Val795Phe variant shown within the structural motif (left). TLR7 LRR structural motifs (1–27) were superposed using MUSTANG[128] structural alignment.

The structure of human n class="Gene">TLR7 is unavailable, but estimates of structural perturbations due to missense variants of human TLR7 can be made by utilizing the structure of Rhesus macaque TLR7[129] (98% sequence identity). On the basis of the Rhesus macaque TLR7 structure, gnomAD missense variants that are expected to be similar to p.Val795Phe (in the patient brothers) in their behavior are also important for discussion. The substituted valine in p.Val795Phe is at an inward facing position within the LRR structural motif (Figure C), and the estimated energetic contribution of Val795 (FoldX[130] alascan) is ∼2 kcal/mol. Several missense variants that substitute nonpolar residues (i.e., Leu/Ile) at inward facing positions within the LRRs (similar to that of Val795 and with estimated energetic contributions of 2–4 kcal/mol) are also present in gnomAD (Figure C), suggesting that the observed rare variants would be important for determining the risk of COVID-19 severity. The structure of Rhesus macaque TLR7[129] in complex with ssRNA in the functionally competent homodimer state (Figure B) also shows the presence of interfacial missense variants such as p.Arg473Lys (TLR7–ssRNA interface), p.Arg186Gln, and p.Arg553Trp (TLR7–TLR7 interface), along with the variant p.Asn523His at a glycosylation site. All of these variants, through structural or functional perturbation of TLR7, have the potential to contribute to the malfunctioning of TLR7-mediated signaling. For X-linked TLR7, men are expected to be affected more by TLR7 variants. The impact of TLR7 variants on females is less clear, as some studies report TLR7 to escape XCI,[131] whereas other studies have not included TLR7 among the list of genes escaping XCI.[118] However, the autosomal dominant behavior of TLR3 variants (see later) suggests that females could also be affected by TLR7 variants, likely due to the dimerization during RNA recognition.

Variants of Other Innate Immunity Proteins

Another important recent study[4] from the Casanova laboratory shows that rare loss-of-function variants in proteins associated with TLR3 and the n class="Gene">interferon regulatory factor 7 (IRF7)-dependent type-I interferon (IFN) signaling pathway are enriched in patients with life-threatening COVID-19 pneumonia relative to asymptomatic infections.[4] In the study, 13 genetic loci whose variants had been known to be associated with immunity to the influenza virus were probed experimentally to identify autosomal-dominant/-recessive loss-of-function variants in these loci in ∼4% of patients with life-threatening COVID-19 pneumonia.[4] Information on the loss-of-function variants of these 13 genes present in the human population would therefore be valuable for predicting susceptibility to the severe disease. In addition, because auto-Abs against interferons (i.e., resulting in the depletion of interferons) also results in severe COVID-19 disease,[5] information about loss-of-function variants of interferon genes present in the human population will also be valuable. The loss-of-function variants of the 13 and 14 interferon (IFNA1–8, IFNA10, IFNA14, IFNA16, IFNA17, IFNA21, IFNB1, and IFNW1) genes are listed in Figure A. It is important to note that the pLI score of many of these genes (e.g., TLR3, IRF7, IFNAR1, etc.) is ∼0.0, suggesting that loss-of-function variants of these genes are tolerated, especially TLR3, IRF7, and IFNAR1, which have 15, 10, and 17 truncation variants, respectively, in gnomAD along with having several other nonmissense variants (Figure A). It is interesting to note that for some of the loss-of-function variants (e.g., in TLR3, TBK1, IRF3, etc.) identified in the study[4] (or those known from previous studies), a single copy of the variant (i.e., heterozygous) could result in severe disease.[4] That is, with respect to disease severity, some of the variant alleles are dominant. In other words, a heterozygous individual can otherwise be healthy (i.e., in absence of the infection, unlikely to be affected by the loss-of-function) but could develop severe disease when challenged by the infection. For the autosomal dominant nature (i.e., with respect to disease severity) of several variants of the genes identified in the study,[4] other variants of these genes observed in gnomAD also have the potential to result in malfunction of the corresponding proteins. For example, like the TLR3 dominant p.Ser339 fs (frameshift) and p.Trp769* (stop-gained) loss-of-function variants observed in patients with severe pneumonia,[4] the list of other TLR3 stop-gained (e.g., p.Trp11*, p.Ser88*, p.Arg394*, etc.), frameshift (e.g., p.Leu243MetfsTer16, p.Leu255GlufsTer2, p.Lys345ArgfsTer10, etc.), and deletion/insertion (e.g., p.Asn285del, p.Asp348del, p.Leu482del, etc.) variants, IRF7 stop-gained (e.g., p.Trp30*, p.Trp214*, p.Trp391*, etc.) and frameshift (e.g., p.Asp207AlafsTer206, p.Leu185SerfsTer26, p.Glu95ArgfsTer3, etc.) variants, and others observed in gnomAD are of great interest for understanding the potential for risks of severe symptoms. In gnomAD, rare variants are typically observed in heterozygotes, and the rare variants of innate immunity proteins have the potential to function as autosomal dominant with respect to disease severity.

Figure 4

Variants of innate immunity proteins: (A) List of genes with their chromosomal location, UniProt accession number of the corresponding protein, pLI and pNull probabilities, and the number of different gnomAD variants. For the interferon genes, only the average values of the 14 genes are presented. (B) Mouse TLR3 dimer structure with bound dsRNA (left), showing p.Pro554Ser located between two sets of interfacial LRRs (middle) and the pair of disulfide disrupting variants (right). (C) TLR3 TIR domain modeled structure (from ModPipe[132]) showing p.Met870Val (middle) located within the nonpolar residues cluster specific for Group 3 TIR[133] (right). (D) Select missense variants within TLR3 LRR structural motifs (left) and the TLR3–dsRNA interfacial variant (right). (E) TBK1 homodimer structure showing interfacial missense variants (left). The mutants shown with a yellow background in panels B, C, and D are autosomal dominant variants reported in some patients with severe symptoms[4] or in herpes simplex virus-1 encephalitis (HSE).[28]

Variants of innate immunity proteins: (A) List of genes with their chromosomal location, UniProt accession number of the corresponding protein, pLI and pn class="Chemical">Null probabilities, and the number of different gnomAD variants. For the interferon genes, only the average values of the 14 genes are presented. (B) Mouse TLR3 dimer structure with bound dsRNA (left), showing p.Pro554Ser located between two sets of interfacial LRRs (middle) and the pair of disulfide disrupting variants (right). (C) TLR3 TIR domain modeled structure (from ModPipe[132]) showing p.Met870Val (middle) located within the nonpolar residues cluster specific for Group 3 TIR[133] (right). (D) Select missense variants within TLR3 LRR structural motifs (left) and the TLR3–dsRNA interfacial variant (right). (E) TBK1 homodimer structure showing interfacial missense variants (left). The mutants shown with a yellow background in panels B, C, and D are autosomal dominant variants reported in some patients with severe symptoms[4] or in herpes simplex virus-1 encephalitis (HSE).[28]

Whereas there are several variants of proteins (Figure A) with the potential to interfere with the type I interferon (n class="Gene">IFN) pathway (and therefore demand attention), only a few variants (e.g., missense) of TLR3 and TBK1 (Figure B–E) are discussed here in the context of protein structure for convenience. The autosomal dominant TLR3 missense variants p.Pro554Ser and p.Met870Val probed in the study[4] above are, respectively, within the LRR and TIR (Toll/IL-1 receptor) regions of the molecule. In the mouse TLR3 LRR structure (∼80% sequence identity with human), p.Pro554Ser is placed in the loop connecting two sets of LRRs engaged in the dsRNA and the dimer interface, respectively (Figure B). With an estimated energetic contribution of ∼2 kcal/mol (FoldX[130]), Pro55 likely plays an important role in the assembly for the TLR3 signaling complex. The known p.Leu360Pro dominant loss-of-function TLR3 missense variant, impacting herpes simplex virus-1 encephalitis (HSE),[28] resides in an inward facing position within the LRR motif. Similar to p.Leu360Pro, several missense variants that substitute nonpolar residues (i.e., Leu/Ile) at inward facing positions within the LRRs (with estimated energetic contributions of 2–4 kcal/mol) are also observed in gnomAD (Figure D, left). In addition, TLR3–dsRNA interface variant (p.Arg489Ser) (Figure D, right) and TLR3 disulfide disrupting variants (p.Cys28Gly and p.Cys651Ser) are also observed (Figure B, right). TLR3 TIR belongs to Group 3 TIR.[133] Group 3 TIR possesses a distinct cluster of bulky nonpolar residues (Figure C), and the modeled human TLR3 TIR structure suggests that p.Met870Val is likely to interfere with the nonpolar residue cluster (Figure C). The TBK1 homodimer structure indicates that gnomAD missense variants p.Arg444Gln and p.Arg547Lys in each protomer are likely to perturb several interfacial hydrogen-bonded interactions and that these variants have the potential to perturb TBK1 function (Figure E).

In addition to variants of the innate immunity pathway proteins, the genetic variability across the human n class="Gene">leukocyte antigen (HLA) A, B, and C genes could also contribute to SARS-CoV-2 susceptibility and disease severity. In this regard, in silico(134−136) assessment of the binding of SARS-CoV-2 peptides onto major histocompatibility complex (MHC) class I binding pocket across HLA-A, -B, and -C genotypes is also useful for the analysis of susceptibility.[134−136] Because the presentation of high-affinity peptide by MHC class I requires the peptide-loading complex[137,138] protein machinery, information about the variants of genes (TAPBP, PDIA3, CALR, TAP1, and TAP2, referred to here as MHCI-assembly genes) of the peptide-loading complex will also be useful in this regard. The variants of MHCI-assembly genes are therefore listed in Figure A (bottom), and it is interesting to note that pLI probabilities for TAP1, TAP2, and TAPBP are 0.0.

Figure 5

SARS-CoV-2 RPI and PPI protein: (A) List of genes (as described in Figures A and 4A) of RPI (top), PPI (middle), and MHCI-assembly proteins (bottom). For convenience, only a select few among the 13 RPI and 24 PPI with high pNull are shown. (B) pLI (left) and pNull (right) probability distributions of SARS-CoV-2 RPI (top) and PPI (middle) proteins compared with reference human proteins (bottom). The dotted line (pLI or pNull = 0.9) represents the cutoff. For the RPI proteins, only those showing at least two-fold enrichment in both postinfection time points (24 and 48 h)[94] were considered.

SARS-CoV-2 RPI and PPI proteimical">n: (A) List of genes (as described imical">n Figures A and 4A) of RPI (top), PPI (middle), and MHCI-assembly proteins (bottom). For convenience, omical">nly a select few amomical">ng the 13 RPI and 24 PPI with high pNull are shown. (B) pLI (left) and pNull (right) probability distributions of SARS-CoV-2 RPI (top) and PPI (middle) proteins compared with reference human proteins (bottom). The dotted line (pLI or pNull = 0.9) represents the cutoff. For the RPI proteins, only those showing at least two-fold enrichment in both postinfection time points (24 and 48 h)[94] were considered.

Human Protein Variants with Potential to Influence Virus–Host-Protein–Protein and RNA–Protein Interactions

The focus of the discussion above on variants was on a small set of genes that are engaged in a specific pathway (e.g., viral entry and innate immunity). Variants of proteins of other pathways could also contribute to susceptibility. In general, for a system-level view of variants in the context of host–virus interactions, proteins identified through proteomic-scale approaches provide a useful reference. In this regard, a discussion on variants associated with human proteins recently identified by Gordon et al.[92] and Flynn et al.[94] that interact with SARS-CoV-2 proteins and RNA, respectively, would be valuable. Gordon et al.[92] and Flynn et al.,[94] respectively, report 330 and 309 human proteins engaged in high-confidence SARS-CoV-2–human protein–protein interactions[92] (referred to as SARS-CoV-2 PPI proteins) and RNA–protein interactions[94] (referred to as SARS-CoV-2 RPI proteins). For convenience, the discussion on variants of SARS-CoV-2 PPI and RPI proteins is in the context of tolerance to loss-of-function variants. Genes can be categorized as (a) tolerant to loss-of-function variants as homozygous (pNull probability high), (b) tolerant to loss-of-function variants as heterozygous (pRec probability high), and (c) intolerant to loss-of-function variants as heterozygous (pLI probability high).[117] As discussed above, TLR3, IRF7, and IFNAR1 all have low pLI (i.e., 0.0, Figure A). IRF7 and IFNAR1 have high pNull (0.99 and 0.84, respectively), and TLR3 additionally has small pNull (i.e., 0.01, Figure A). That is, IRF7 and IFNAR1 are expected to be tolerant to loss-of-function variants as homozygous, whereas TLR3 is expected to be tolerant to loss-of-function as heterozygous. It is interesting to note that autosomal recessive variants (i.e., homozygous) for disease severity were observed for IRF7 (e.g., p.Pro364 fs/p.Pro364 fs)[4] and IFNAR1 (e.g., p.Trp73Cys/Trp73Cys),[4] which is consistent with their high pNull probability for an individual to carry both loss-of-function alleles. Autosomal dominant (i.e., heterozygous) variants were observed for TLR3 (e.g., p.Trp769*/wildtype),[4] which is consistent with the low pNull probability for TLR3. In fact, autosomal dominant variants identified in the above study[4] were predominantly in genes with low pNull probability (e.g., TLR3, UNC93B1, TBK1, IRF3, etc., Figure A). In short, the set of pLI, pNull, and pRec probabilities provides important information for expecting a variant to be observed as homozygous/heterozygous. Genes with low pLI (equal to 0) and high pNull (>0.8) probability would have a higher chance of having homozygous variants in a population. For example, FUT2 and CCR5 (pLI = 0 and 0; pNull = 1.0 and 0.98) genes have homozygous truncation variants present in the human population. Because these proteins directly participate in the entry of their respective virus, homozygous truncation variants of FUT2 and CCR5 provide protective advantage. Therefore, for SARS-CoV-2 PPI and RPI proteins that directly interact with viral machinery (and likely participate in the essential steps of the viral life cycle), a discussion on the behavior of pLI and pNull probabilities of these genes is very useful (Figure ). There are 24 and 13 genes of SARS-CoV-2 PPI and RPI proteins, respectively, with pLI = 0 and pNull >0.8 (Figure A), and individuals carrying homozygous truncation variants of these genes are likely to be found in the human population. For example, homozygous individuals carrying nonsense or frameshift variants of these genes (e.g., CNTRL, NINL, ANO6, CHD1L, ANXA5, NLRX1, etc.) are observed in gnomAD. The absence of functional forms of these proteins in homozygous individuals could provide a protective advantage. In general, with respect to the disease susceptibility, homozygous variants of genes with high pNull probability are expected to (a) provide protection if the protein is essential for the viral life cycle and (b) result in severe symptoms if the protein is a vital player in innate immunity. The distribution of the pLI score (Figure B) shows that ∼24 and ∼49% of human SARS-CoV-2 PPI and RPI proteins, respectively, have pLI > 0.9, whereas only ∼16% of human proteins, in comparison, have pLI > 0.9 (Figure B). This suggests that the virus tends to interact with machinery (especially for viral RNA maintenance) that is less prone to tolerating truncation variants than that expected from the reference background. This could be an adaptation to minimize host variability for viral survival. The distribution of pNull probability suggests that ∼10% of human genes are tolerant to loss-of-function genes as homozygous (i.e., pNull ≥0.9), whereas that for SARS-CoV-2 PPI and RPI protein genes is ∼5 and ∼3%, respectively, which is lower than the reference background (Figure B) and is consistent with an adaptation to minimize host variability. Among the SARS-CoV-2 PPI and RPI protein genes, 19 are X-linked (e.g., POLA1, PDZD11, DDX3X, MID1IP1, etc.), with none having high pNull probability, but three X-linked genes (PDZD11, HS6ST2, and MID1IP1) have low pLI, and their nonsense variants may provide protection in men. Finally, other variants (e.g., 5′ UTR, synonymous, splice-acceptor/donor, etc.) affecting tissue-level protein expression or stability-compromising missense variants that indirectly affect protein expression levels of the above-discussed pathway specific proteins could also be valuable for assessing the role of genotype in pathogenesis and susceptibility.

Conclusions

In recognition of the important role that variants could play in disease susceptibility, the human proteimical">n variants imical">nvolved imical">n n class="Species">SARS-CoV-2–host interactions have recently been reported, analyzed, and discussed.[33,34] The emphasis here has been to provide a comprehensive view of the roles that human protein variants could play not only in these interactions but also in all major aspects of viral biology, such as viral entry, innate immunity, and host–virus interactions (protein–protein and protein–RNA). Because rare variants have been observed to contribute to disease severity in otherwise healthy individuals,[4] a careful look at those involving pathway-specific proteins in a large reference database is needed, given that they typically occur in heterozygotes. The findings extracted from the gnomAD database suggest that heterozygous carriers of several rare variants are present in the human population and that disease risks may be associated with these individuals. In addition, the observed influence of rare homozygous recessive variants on disease severity[4] encouraged an assessment of tolerance for loss-of-function (involving pLI, pNull, and pRec probabilities[117]) of the genes discussed above. The degree of this tolerance helps estimate the likelihood of observing rare variant carriers (both homozygous and heterozygous) in a population. Although the discussion was centered around gnomAD, the rare variants not observed in this database (e.g., the variants of TLR7 and TLR3 discussed above) could play equally important roles. Because the profiling of an individual’s variants can now be performed with relative ease (e.g., using quantitative PCR[139] or MassARRAY[140−142]), this is especially true for specific loci, which would avoid whole-genome sequencing and thereby facilitate the assessment of COVID-19 disease risks. Such profiling would be valuable for optimizing national responses to this health crisis while we await an effective therapeutic or vaccine. In addition, knowledge about the protein structural consequences of these variants (Figures , 3, and 4) enables probing of the mechanistic basis of COVID-19 susceptibility. The significance of the discussion is thus to advocate: (a) rapid sequencing of specific loci to facilitate the assessment of COVID-19 disease risk and (b) experimental probing of the structural consequences of known variants to unravel the mechanistic basis of susceptibility.