HLA and AB0 Polymorphisms May Influence SARS-CoV-2 Infection and COVID-19 Severity

Abstract

Background. 

SARS-CoV-2 infection is heterogeneous in clinical presentation and disease evolution. To investigate whether immune response to the virus can be influenced by genetic factors, we compared HLA and AB0 frequencies in organ transplant recipients and waitlisted patients according to presence or absence of SARS-CoV-2 infection.

Methods. 

A retrospective analysis was performed on an Italian cohort composed by transplanted and waitlisted patients in a January 2002 to March 2020 time frame. Data from this cohort were merged with the Italian registry of COVID⁺ subjects, evaluating infection status of transplanted and waitlisted patients. A total of 56 304 cases were studied with the aim of comparing HLA and AB0 frequencies according to the presence (n = 265, COVID⁺) or absence (n = 56 039, COVID^–) of SARS-CoV-2 infection.

Results. 

The cumulative incidence rate of COVID-19 was 0.112% in the Italian population and 0.462% in waitlisted/transplanted patients (OR = 4.2; 95% CI, 3.7-4.7; P < 0.0001). HLA-DRB1*08 was more frequent in COVID⁺ (9.7% and 5.2%: OR = 1.9, 95% CI, 1.2-3.1; P = 0.003; P_c = 0.036). In COVID⁺ patients, HLA-DRB1*08 was correlated to mortality (6.9% in living versus 17.5% in deceased: OR = 2.9, 95% CI, 1.15-7.21; P = 0.023). Peptide binding prediction analyses showed that these DRB1*08 alleles were unable to bind any of the viral peptides with high affinity. Finally, blood group A was more frequent in COVID⁺ (45.5%) than COVID^– patients (39.0%; OR = 1.3; 95% CI, 1.02-1.66; P = 0.03).

Conclusions. 

Although preliminary, these results suggest that HLA antigens may influence SARS-CoV-2 infection and clinical evolution of COVID-19 and confirm that blood group A individuals are at greater risk of infection, providing clues on the spread of the disease and indications about infection prognosis and vaccination strategies.

INTRODUCTION

Although infectious disease threats remain one of the major global challenges, the associated mortality decreased in recent years, with a general increase in life expectancy in many parts of the world. The 2019 novel coronavirus (2019-nCoV or SARS-CoV-2) has recently been added to the list of problematic emerging pathogens of the 21st century.¹ Due to the increasing number of cases in China and other countries, the WHO has declared the SARS-CoV-2 outbreak a global health emergency of international concern on January 30, 2020, and classified it as a pandemic on March 11, 2020.² As of April 26, Italy is one of the European countries with the highest numbers of infected people with over 195 000 cases and >26 000 deaths, occurring mainly in the northern regions, as Lombardy.³

Clinically, the infection ranges from very mild flu-like symptoms to acute respiratory distress syndrome associated with intensive care unit admission and a high mortality.^4,5

Individual variability in building an immune response to the virus is among the hypotheses that can justify this phenotypic heterogeneity. It is known that the genetic background controlling immune responses may affect the ability of the virus to infect and of the host to defend. Based on experience with other infectious agents, such as HCV or HIV and the previous SARS-CoV, it is plausible that polymorphisms in genes coding for innate and adaptive immunity proteins or those coding for transmembrane receptors mediating virus entry in target cells might be responsible for a different host response to infection and hence disease severity, depending on HLA-peptides binding affinity.

Among key proteins of the adaptive immune response, HLA plays perhaps the most important role.^6,7 HLA genes code for highly polymorphic surface molecules that bind peptides derived from endogenous or exogenous proteins, including viral ones, restricting T-cell recognition, and activation. In silico, SARS-CoV-2 peptides have different affinities for HLA alleles,^8,9 raising the question of whether different HLA alleles influence infection and clinical manifestations, as demonstrated for the previous SARS-CoV, where HLA-B*46 was significantly more frequent in severely affected patients than in unaffected persons.¹⁰

Our study aims to assess whether some genetic factors influence the risk of infection. The Italian Health system offers a unique opportunity for this purpose. In fact, the “Istituto Superiore di Sanità” (ISS—Higher Institute of Health, a technical center of the Italian Ministry of Health) collects the registries of both COVID⁺ subjects (COVID Registry) and of all organ transplanted recipients and waitlisted candidates in Italy (Transplant Registry—TR). By matching data of these 2 registries, we identified the subset of patients who had received—or were awaiting—a transplant who were also positive for COVID.

Since AB0 blood groups and HLA-A, B, and DR patients’ typing are included in the TR, we compared the frequencies of these polymorphisms according to the presence (COVID⁺) or absence (COVID^–) of SARS-CoV-2 infection, using the known AB0 and HLA frequencies of the Italian population as control.

MATERIALS AND METHODS

Patients

As of March 22, 2020, 67 822 SARS-CoV-2 positive (COVID⁺) subjects were registered in the database of the Italian Ministry of Health (COVID Registry), with a cumulative incidence of 0.112%. For each patient, clinical history of the infection and outcome were available.

At the same date, the TR database contained 59 941 organ transplants, consecutively performed in Italy since 2002: 55 982 from deceased and 3959 from living donors. Transplanted patients who died during transplant follow-up before COVID exposure (n = 10 349), as well as recipients who underwent multiple transplants (n = 1944), were excluded from this study, resulting in an enrolled cohort of 47 648 transplanted patients (median follow-up from transplant date: 7.9 y; IQR, 3.5–12.8). In addition, on March 22, 2020, 10 139 patients were awaiting organ transplantation. Patients with double registrations (n = 1483) were excluded, resulting in 8656 waitlisted patients (median follow-up from inscription in waiting list: 2.0 y; IQR, 0.8–4.2). Thus, the total eligible cohort for analysis was made of 56 304 patients (Figure 1).

FIGURE 1.: Flow chart of data included in the study.

AB0, HLA-A, HLA-B, and DRB1 Typing

AB0 typing was performed by Blood Bank centers supporting organ transplantation activities. Frequencies of AB0 reported in the different Italian regions were used as control.¹¹

HLA-A, B, and DRB1 typing was performed by the Italian HLA typing laboratories supporting transplantation activities using mainly DNA-based techniques. All labs are accredited by the European Federation for Immunogenetics (https://efi-web.org/) and follow the guidelines issued by CNT and the reference scientific society (https://aibt.it/it/). HLA typing was available for 219 COVID⁺ and 40685 COVID^–. The HLA frequencies used as controls were those previously reported for the Italian population.¹²

Prediction of Possible HLA-A, B, and HLA-DRB1 Ligands of the SARS-CoV-2 Proteins

In silico prediction of the affinity between HLA antigens and peptides generated from spike glycoproteins [6VSB_C-1288aa] of SARS-CoV-2 (which also contains receptor binding domain—RBD [6M17_F-223aa]), utilized the Immune Epitope Database (IEBD—https://www.iedb.org/).¹³ Based on the availability of predictors and previously observed predictive performance, IEDB calculates for each predicted peptide a median percentile rank (PR) by comparing the score of the peptide against the scores of 5 million random 15 mers selected from the SWISSPROT database. A numerically low PR indicates a high affinity. The IEDB currently recommends making selections based on a PR of ≤1% for each HLA allele-peptide combination to cover most of the immune responses.

Statistical Analyses

AB0 and HLA-A, B, DRB1 phenotypic frequencies—assessed by direct counting—and all other categorical variables were compared between COVID⁺ and COVID^– by Chi-square test and Bonferroni post hoc correction for multiple comparisons, when appropriate. They are presented as numbers and percentage (%) based on the total number of patients included in the study. Two-tailed t test or Mann-Withney U test were performed as appropriate, testing normality distribution by Kolmogorov-Smirnov test. Continuous variables, as occurrence, were categorized by ROC curve analysis. Multivariable analysis was performed by adjusted logistic regression model, including sex, age, and phenotype as covariates. A P value <0.05 was considered statistically significant. Analyses were performed using IBM Corp. Released 2020. IBM SPSS Statistics for Windows, Version 26.0. Armonk, NY: IBM Corp.

Italian COVID Registry

The epidemiological surveillance of SARS-CoV-2 was entrusted to the ISS on February 27, 2020, by order of the Italian Ministry of Health of February 27, 2020 (http://www.trovanorme.salute.gov.it/norme/dettaglioAtto?id=73469).

Italian Transplant Registry

The TR is entrusted to the Transplant Information System, which is an infrastructure for the management of data related to the activity of the National Transplant Network, established and regulated by Italian Laws (no. 91/99 and Decree of the Ministry of Health n. 130 of August 20, 2019).

Ethics Statement, Privacy, and Informed Consent

Notification to the Ministry of Health of COVID-19⁺ patients is mandatory by law. All patients included in TR consented to registration and use of their data, which were treated and analyzed in compliance with the European General Data Protection Regulation (EU GDPR), anonymizing sensitive data and processing aggregate data. ISS Data Protection Officer authorized this study.

RESULTS

Combining COVID and Transplant registries, we identified 265 COVID⁺ subjects: 192 transplanted recipients and 73 waitlisted patients, with a cumulative incidence of 0.471, markedly higher than that of the general population (0.112; OR = 4.2; 95% CI, 3.7-4.7; P < 0.0001). These patients underwent molecular testing for COVID either because they were at high risk, that is, they had been in close contact with an infected person, or because they were symptomatic. Patients defined as COVID^– were the untested, asymptomatic counterpart. Transplant recipients and waitlisted patient’s characteristics are shown in Table 1. As expected from Italian epidemiological data, males were more affected than females (67% and 33%, respectively; OR = 1.3, 95% CI, 1.02-1.75; P = 0.039). In addition, COVID⁺ patients were older than those without infection (59.8 ± 11.9 and 56.4 ± 15.3 y, respectively; P < 0.001) and had a higher probability of being resident in Lombardy than in other Italian regions (60.7% and 17.3%, respectively; OR, 7.4; 95% CI, 5.7-9.4; P < 0.001).

TABLE 1. - Characteristics of patients enrolled in the study compared for COVID-19 status

		Transplanted recipients (n = 47 648) and waitlist patients (n = 8656)
		Total	COVID+	COVID–	OR [95% CI]	P	P c *
n (%)		56 304	265 (0.47)	56 039 (99.53)
Gender, a n (%)	Male	37 565 (66.7)	193 (72.8)	37 372 (66.7)	1.33	0.039
	Female	18 728 (33.3)	72 (27.2)	18 656 (33.3)	[1.02–1.75]
Age, mean (SD), y		56.4 (15.3)	59.8 (11.9)	56.4 (15.3)		<0.001 b
PRA ≥80%, c n (%)		1942	13 (6.6)	1929 (6.8)		0.6
Blood group A, d n (%)		21 942	120 (45.5)	21 822 (39.0)	1.30 [1.02-1.66]	0.03
Lombardy residence, n (%)		9838	161 (60.7)	9677 (17 3)	7.4 [5.7-9.4]	<0.001
HLA-DRB1*08, e n (%)		2138	21 (9.7)	2117 (5.2)	1.9 [1.2-3.1]	0.003	0.036
HLA-DRB1*07, e n (%)		8778	33 (15.2)	8744 (21.6)	0.6 [0.5-0.9]	0.023	0.276
HLA-B*51, f n (%)		7202	55 (25.1)	7147 (17.7)	1.6 [1.1-2.1]	0.004	0.092
Organ: n, (%)	Kidney	34 040 (60.5)	186 (0.55g)	33 854		<0.001	0.007
	Liver	15 189 (27.0)	41 (0.27 g )	15 148
	Heart	3990 (7.1)	19 (0.48 g )	3971
	Lung	1464 (2.6)	10 (0.68 g )	1454
	Pancreas	469 (0.8)	3 (0.64 g )	466
	Combined	1116 (2.0)	6 (0.54 g )	1110
	Bowel	36 (0.1)	0	36

*P value with Bonferroni correction (P_c value). For HLA comparisons, the P value was corrected for the number of alleles at each locus. For organ-specific transplants, the P value was corrected for the type of transplant.

^a11 gender missing data in COVID^–.

^b2-sided t test. Normality was assessed by Kolmogorov-Smirnov test.

^cPanel reactivity antibodies: 27 711 missing data (98 in COVID+, 27 613 in COVID^–).

^d111 blood group missing data (1 in COVID⁺, 110 in COVID^–).

^e15 513 HLA-DRB1 typing missing data (48 in COVID⁺, 15 465 in COVID^–).

^f15 676 HLA-B typing missing data (46 in COVID⁺, 15 630 in COVID^–).

^gCumulative incidence COVID⁺.

Stratification of COVID⁺ patients on the basis of transplanted (or to be transplanted) organ, highlighted differences in the cumulative incidence of infection, with the highest value scored by lung (0.68%), followed by pancreas (0.64%), kidney (0.55%), combined transplants (0.54%), heart (0.48%), and liver (0.27%) (Table 1).

HLA-A, B, and DRB1 frequencies were analyzed in 40 904 individuals (32 294 transplant recipients and 8610 waitlisted patients), divided according to SARS-CoV-2 infection (Table 1). HLA-DRB1*08 was associated with a higher risk of COVID (OR = 1.9; 95% CI, 1.2-3.1; P = 0.003; P_c = 0.036). HLA frequencies in COVID^– were comparable to those of the Italian control population (HLA-DRB1*08: 6.0%).

Since selective pressure usually benefits from heterozygosity, we determined whether the frequency of HLA homozygous individuals differed between COVID⁺ and COVID^– patients, failing to highlight significant differences (data not shown).

Blood group A was found to be more frequent in COVID⁺ (45.5%) than in COVID^– (39.0%; OR = 1.3, 95% CI, 1.02-1.66; P = 0.03). In the latter subset, the frequency was comparable to that of the Italian population (39.0%).

The presence of HLA-DRB1*08 was the biological factor conferring the greatest risk of infection, as determined by logistic regression analysis adjusted for all the variables with statistical significance in the univariate analysis (Table 2). Lombardy residence and age were confirmed independent factors that significantly influence the risk of infection, while gender and blood group A were not statistically significant. Even considering only the subset of patients residing in Lombardy, HLA-DRB1*08 retained a significantly higher risk of infection in multivariate analysis (OR, 1.9; 95% CI, 1.1-3.3; P = 0.026) (not shown).

TABLE 2. - Adjusted logistic regression model for COVID-19 infection

	OR	95% CI, OR		P
		Lower	Upper
HLA-DRB1*08	1.814	1.151	2.860	0.010
Gender, male	1.331	0.996	1.784	0.055
Blood group A	1.077	0.820	1.414	0.593
Age ≥60 y a	1.475	1.127	1.931	0.005
Lombardy residence	7.229	5.499	9.504	<0.001

^aCutoff resulting by ROC curve analysis (Figure S1a, SDC, https://links.lww.com/TP/C47).

Finally, we compared age and gender as well as HLA and AB0 frequencies in COVID⁺ patients, comparing those who died to those still alive. As expected, those with a more advanced age had a higher risk of death (P < 0.001; Table 3), while gender did not influence disease outcome. The frequency of blood group A did not differ between those who died (44.4%) and those still alive (45.6%). Among HLA alleles, HLA-DRB1*08 was significantly associated with a higher risk of death (OR = 2.9; 95% CI, 1.15-7.21; P = 0.023, Table 3). Adjusted logistic regression of COVID mortality confirmed a significant effect of HLA-DRB1*08 (OR = 8.6; 95% CI, 1.7-43.9; P = 0.01), older age (≥63 y: OR = 2.5; 95% CI, 1.2-5.3; P = 0.018) and residence in Lombardy (OR = 2.4; 95% CI, 1.08-5.2; P = 0.031; Table S1, SDC, https://links.lww.com/TP/C47).

TABLE 3. - Main characteristic of COVID⁺ patients according to mortality

		COVID+ patients (n = 265)
		Dead	Alive	OR [95% CI]	P
n (%)		72 (27.2)	193 (72.8)
Gender, n (%)	Male	54 (75.0)	139 (72.0)	1.17 [0.63-2.16]	0.628
	Female	18 (25.0)	54 (28.0)
Age, mean (SD), y		65.4 (9.9)	57.7 (12.0)		<0.001 a
PRA ≥80%, b n (%)		4 (9.1)	9 (7.3)	1.27 [0.37-4.34]	0.71
Blood group A, n (%)		32 (44.4)	88 (45.6)	0.95 [0.55-1.65]	0.86
HLA-DRB1*08, n (%) c		10 (17.5)	11 (6.9)	2.9 [1.15-7.21]	0.023
Lombardy residence		50 (69.4)	111 (57.5)	1.7 [0.94-2.99]	0.077

^a2-sided t test. Normality was assessed by Kolmogorov-Smirnov test.

^bPanel reactivity antibodies: 98 missing data (28 in COVID⁺ dead, 70 in COVID⁺ alive).

^c48 HLA-DRB1 typing missing data (15 in COVID⁺ dead, 33 in COVID⁺ alive).

We then useed bioinformatics tool IEDB to predict possible HLA-A, B, and DRB1 ligands of spike peptides. The purpose of this simulation was to identify which HLA alleles can bind with high affinity a broad range of viral peptides, being potentially “good presenters,” as opposed to alleles that have more limited binding capabilities. The simulation was done with the entire spike protein, of which RBD is a part. Since the protein is made of 1288 amino acids, there are 2679 potential peptides for class I alleles and 2202 for class II alleles. Only HLA-A, B, and DRB1 alleles that had a binding affinity with 1 or more of these peptides with a PR ≤1 were considered, highlighting 539 combinations out of 4699 observed, involving 16 HLA-A, 11 HLA-B, and 8 HLA-DRB1 alleles. Importantly, DRB1*08 alleles were unable to bind any of the viral peptides (Tables S2 and S3, SDC, https://links.lww.com/TP/C47).

Since HLA typing was performed at low resolution level, we attempted to convert the alleles presenting a high affinity for the viral peptides into the corresponding HLA antigens. For each of these alleles, the corresponding HLA antigen was defined (Table S2, SDC, https://links.lww.com/TP/C47). Some HLA antigens, at least in the Italian population, have full correspondence with the indicated allele (such as HLA-A*23:01). For a few HLA antigens considered, such as B35, DR4, DR11, only in minority of cases the corresponding HLA alleles are coincident. Due to the poor correspondence between alleles and HLA-DRB antigens, we decided not to consider them in further analyses. We therefore reevaluated the HLA-A and B phenotypes of the enrolled patients, to measure how many of them did not present any of these “in silico” protective antigens, and how many up to a maximum of 4 of these antigens for the A and B loci. This distribution, however, did not present significant differences between COVID⁺ and COVID^– (data not shown).

DISCUSSION

Despite an unprecedented global effort of the scientific community to improve knowledge about SARS-CoV-2 and COVID-19, many aspects of this infection remain unknown. One open point concerns the identification of the factors responsible for the spread of the disease and for its different clinical manifestations.³ It is well documented that male sex, comorbidities, and age are unfavorable prognostic factors.⁵ Individual heterogeneity in responding to a viral infection, linked both to environmental and genetic factors, is one the hypotheses that can be postulated to justify the different outcomes. Among nongenetic conditions limiting immune responses, immunosuppression due to different clinical conditions plays a major role. Even if in this context, solid organ transplant recipients who undergo permanent immunosuppressive therapy should be at higher risk of infection, but literature is so far limited and with conflicting results.^14-18 The Italian Health system offers a unique opportunity to investigate these aspects, because it centralizes collection of epidemiological data of both COVID⁺ patients and patients waiting for organ transplantation or already transplanted in national Institutions. By combining these data, we identified the subgroup of patients who received, or were awaiting, a transplant and were simultaneously COVID⁺. The first result is that the infection risk for transplanted and waitlisted patients is markedly higher compared with the control population, in line with higher risk of a patient population requiring frequent hospital care and immunosuppression.

Beside environmental conditions including residence in Lombardy region, which was the worst hit Italian area, individual genetic variability can result in a more susceptible or more resistant phenotype to viral infection and disease progression.¹⁹ The AB0 blood group and the HLA-A, B, and DRB1 typing data are also included in the TR, allowing comparison of the frequencies of these polymorphisms in COVID⁺ and COVID^– individuals.

HLA genes code for highly polymorphic (>26 000 variants described to date) surface molecules that bind peptides derived from different sources, including viruses. The affinity of SARS-CoV-2 peptides for HLA molecules varies between polymorphic HLA alleles, potentially influencing antigen presentation and the strength of the immune response.^8,20 While studies on HLA and SARS-CoV-2 associations have never been published for European populations, the possible association between HLA polymorphisms and susceptibility to the previous SARS-CoV was amply addressed.^9,21-23 Importantly, many of the identified SARS-CoV associated HLA antigens were rare in the European or African populations while frequent in Asians, where SARS-CoV spread.

Several in silico studies showed that the various HLA alleles bind to peptides derived from SARS-CoV-2 nucleocapsid with different affinity.^7,24 Successful presentation of peptides depends on their effective binding to HLA molecules via hydrogen bonds and salt bridge interactions, allowing for high affinity with a broader specificity. We studied the potential interaction of the putative viral epitopes recognized by T cells with their corresponding representative HLA class I and class II allele most frequently observed in the Italian population. The surface spike glycoprotein (S), which is critical for virus entry through engaging the host receptor and mediating virus-host membrane fusion, is the major putative antigen of coronaviruses. Recent studies focused on neutralizing antibodies, also taking advantage from experience with SARS-CoV infection and identified binding sites for neutralizing antibodies.^25,26 One of them, CR3022, targets a highly conserved epitope, distal from RBD that enables crossreactive binding between SARS-CoV-2 and SARS-CoV, which have an aminoacid identity of ~70%.^1,26 For these reasons, we predicted the binding affinity between peptides derived from S protein of SARS-CoV-2, which include RBD, and HLA class I and class II molecules, identifying a set of HLA-A, B, and DRB1 molecules with high binding affinity for S peptides (Tables S2 and S3, SDC, https://links.lww.com/TP/C47). While they represent a limited portion of HLA antigens, they are relatively frequent in the Italian population. Unfortunately, however, we were unable to document increased frequencies of these alleles in COVID^– compared to COVID⁺ patients, likely because HLA typing was carried out at low resolution (1 “digit”). The same HLA antigen can in fact be split in several variants, which have a different degree of affinity for viral peptides.

With this limitation, we investigated a possible association between HLA polymorphisms and susceptibility or evolution of SARS-CoV-2 infection in Italian COVID⁺ patients and found that HLA-DRB1*08 represents a risk factor both for COVID infection and related death. In the Italian population, ~6% of individuals carry HLA-DRB1*08, 72% of whom are DRB1*08:01, 8% DRB1*08:02, 9% DRB1*08:03, 9% DRB1*08:04, 1% DRB1*08:10, and all other subtypes with frequencies close to 0.¹² In our prediction analyses, none of these DRB1*08 alleles were able to bind any of the viral peptides with high affinity, raising the hypothesis that HLA-DRB1*08-expressing individuals are less able to recognize SARS-CoV-2, and therefore more susceptible to infection and to severe manifestations of COVID-19. By looking at the distribution of HLA-DRB1*08 frequency in Italy, we noticed that this allele is more frequent in the north, where COVID presented the highest cumulative incidence, compared to both the south and the islands (Figure 2A). In the North of Italy, a lower cumulative incidence of COVID-19 was recorded in the 2 delta Po river province of Rovigo and Ferrara where frequency of HLA-DRB1*08 was intriguingly lower (Figure 2B). Even considering patients resident in Lombardy region, the presence of HLA-DRB1*08 represented an independent risk factor of infection and even more of death in univariate e multivariate analyses. A unique feature of HLA-DRB1*08 is that the HLA haplotype that contains it, does not usually present other functioning DRB genes.²⁷ In fact, DR8-positive subjects have a lower set of HLA-DRB molecules, with an expected reduced repertoire of presented peptides. We were unable to confirm the association between COVID-19 and HLA-B*44, recently reported and identified based on “an ecological approach” in the Italian population.²⁸ This result may be explained on the basis of the analytical approach used, based on an epidemiological analysis in the Correale et al’s paper instead of on laboratory data in the present study.

FIGURE 2.: HLA-DRB1*08 allele frequencies and cumulative COVID-19 incidence in Italy. A, HLA-DRB1*08 allele frequencies in Italian control population composed by 340 000 bone marrow donors. Colors define in each province the allelic frequencies (expressed as %), as indicated in the figure. Modified from Rendine et al³⁵ Lombardy region is highlighted in green, while the provinces of Ferrara and Rovigo are in blue. Gray areas refer to provinces where no data is available. B, Cumulative incidence of COVID-19 patients in different Italian provinces as of May 10, 2020 (from: https://lab.gedidigital.it/gedi-visual/2020/coronavirus-i-contagi-in-italia/). The color code shows the ratio between healthy individuals over COVID⁺ patients, as indicated in the vertical bar in the figure. Lombardy region is highlighted in green, while the provinces of Ferrara and Rovigo are in blue.

Another finding of this study concerns the risk conferred by the blood group A. We have shown that in COVID⁺ patients, the frequency of blood group A is significantly higher than the negative counterpart, in line with a recent study.^29,30 These findings are also consistent with similar risk patterns of AB0 blood groups for SARS-CoV infection found in previous works.³¹ A likely explanation is that anti-A natural agglutinins specifically inhibits adhesion of SARS-CoV S protein to angiotensin-converting enzyme 2-expressing cells.³² Given the similarity in nucleic acid sequence and angiotensin-converting enzyme 2-receptor binding between SARS-CoV and SARS-CoV-2,³³ a similar explanation could hold true also for the latter virus. Blood group A, however, does not change the risk of mortality once infected with the virus, confirming a recent study²⁹ and previous work for SARS-CoV infection.³¹

In conclusion, this study demonstrates for the first time that HLA polymorphisms influence the rate and outcome of SARS-CoV-2 infection in Europeans. It also confirms that blood group A is associated to an increased risk of infection. Together, these findings can explain at least in part, the different worldwide spread of the virus.³⁴ They also make it possible to identify individuals at greater risk of infection, or of inferior outcome. Lastly, through the identification of the HLA antigens that bind with greater affinity the peptides derived from the virus, these results can be helpful in setting up population-specific vaccination strategies.

Despite promising results, this study also has few limitations. First, we examined a retrospective cohort of patients. Second, we studied patients at higher risk of COVID and at higher risk of fatal complications. Third, the identification of positive individuals—at least in Italy—depended on testing availability, which was not homogenous at the national level. Fourth, HLA typing was limited (only A, B, and DR loci) and mostly at low resolution. Future work is needed to confirm these results in other patient populations and to determine whether other HLA loci, such HLA-C, HLA-DQ, and HLA-DP, play a role in influencing SARS-CoV-2 infection and COVID-19 evolution.

ACKNOWLEDGMENTS

This study was carried out thanks to the support and cooperation of the Italian National Transplant Network. Specifically, we want to acknowledge the Regional Coordinators: (*) Daniela Maccarone, Angelo Saracino, Pellegrino Mancini, Antonio Corcione, Gabriela Sangiorgi, Roberto Peressutti, Nicola Torlone, Andrea Gianelli Castiglione, Giuseppe Piccolo, Francesca De Pace, Federico Genzano Besso, Loreto Gesualdo, Lorenzo D’Antonio, Giorgio Battaglia, Adriano Peris, Lucia Pilati, Peter Zanon, Atanassios Dovas, Giuseppe Feltrin. We wish to thank staff at Transplant Centres, the hospitals that took care of waitlisted and transplanted patients, as well as those that made organ donation possible, even during this difficult COVID-19 pandemic period. A special thanks to blood transfusion centers and HLA tissue typing laboratories. Finally, yet importantly, our deepest gratitude is for the donor families, whose generosity makes, each day, transplantation accessible for many patients. We would like also to thank the Task Force ISS for the Italian COVID-19 epidemiological surveillance. The authors thank Paola Di Ciaccio for the article linguistic review and Fabrizio Fop for statistical supports. The authors are indebted with Giuseppe Remuzzi for his critical revision of the article.