2. The ABO Blood Group System at a Glance
The ABO blood group system was discovered more than a century ago and its understanding allowed the development of blood transfusion. However, the corresponding antigens are expressed on a large number of cell types in addition to erythrocytes [
42]. They are of the carbohydrate type, constituting terminal motifs of either N-linked or O-linked chains of glycoproteins as well as of glycolipids. Their synthesis proceeds by addition of monosaccharide units to precursor glycan chains through specific glycosyltransferases. It first requires the synthesis of the histo-blood group H precursor antigen, which is catalyzed by alpha1,2fucosyltransferases that add a fucose in α1,2 linkage to a terminal β-galactose of the subjacent glycan chain. The FUT1 enzymes are responsible for this activity in erythroblasts, megakaryocytes, vascular endothelial cells, and several other cell types, while it is the FUT2 enzyme that catalyzes synthesis of the H antigen in most epithelial cells such as the upper airways, the digestive tract, and the lower genito-urinary tracts. Once the H antigen is produced, addition in α1,3 linkage of an N-acetylgalactosamine or of a galactose to the same subjacent galactose unit by the A or B blood group enzymes generates the A and B antigens, respectively (
Figure 1). The A and B enzymes are coded by distinct alleles of the
ABO gene, whereas the
O alleles correspond to null alleles unable to generate any active enzyme. These three major types of alleles generate the four major phenotypes O, A, B, and AB [
43]. Both the
FUT1 and
FUT2 genes also present null alleles that lead to a lack of precursor H antigen synthesis in the corresponding cell types and therefore to a lack of A and B blood group antigens expression in these cells [
44].
FUT1 null alleles are responsible for a rare red cell phenotype called “Bombay”. Given its rare occurrence, it will not be discussed any further. By contrast, null alleles of the
FUT2 gene are common and their frequency varies across populations. These alleles are responsible for the so-called “nonsecretor” phenotype which by contrast with the “secretor” phenotype is characterized by a lack of A, B, and H antigens in many secretions such as saliva and in epithelia. In the Western world, the secretors represent around 80% of the population and nonsecretors, the remaining 20% [
44].
In addition to its antigens, the ABO system is characterized by the presence of antibodies against the A and B antigens. Thus, blood group O individuals possess anti-A and anti-B antibodies, blood group A individuals possess anti-B antibodies, and blood group B individuals have anti-A antibodies. Only blood group AB individuals are devoid of both anti-A and anti-B antibodies. This system of antigens and their cognate antibodies defines the basic rules of transfusion where blood group O constitutes a universal donor, whereas blood group AB represents a universal receiver [
45]. The origin of the natural anti-ABO antibodies is still debated. Nonetheless, it seems that most of these antibodies appear during the first year of life under stimulation of microorganisms either pathogenic or from the microbiota that carry similar antigens [
46,
47]. Their amounts are highly variable between individuals and some data suggest that they may decrease with improved hygiene conditions [
48,
49].
The
ABO gene and the
FUT2 gene, which controls expression of ABH antigens in epithelia, are among the few human genes clearly under frequency-dependent balanced selection, suggesting important roles in interactions with environmental factors [
50,
51,
52,
53,
54,
55]. Histo-blood group antigens, including ABO blood groups, have previously been implicated in the genetic susceptibility to several infectious diseases, including viral diseases. This has been particularly well documented for human noroviruses and rotaviruses that together are responsible for the majority of gastroenteritis cases worldwide. These non-enveloped RNA viruses attach to the carbohydrate antigens expressed in the gastrointestinal mucosa. They have evolved so that distinct strains recognize preferentially different carbohydrate motifs, resulting in a strain-dependent susceptibility in accordance with the person’s blood type [
56]. Rabbit Hemorrhagic Disease Virus (RHDV) is a highly pathogenic rabbit calicivirus related to noroviruses also attaches to blood group antigens expressed in the rabbit respiratory and gut epithelia. The rabbit-RHDV pair allowed the documentation of a natural example of host–pathogen co-evolution involving the recognition of A, B, and H blood groups antigens by the virus [
57]. These co-evolving host–pathogen pairs explain, at least partly, the maintenance of the ABO polymorphism. Nevertheless, using mathematical modeling, it has been argued that its maintenance would additionally require the role of the anti-ABO antibodies to be taken into account [
58]. It has additionally been argued that viral-mediated selection may explain why South American native populations are exclusively of blood group O [
59].
5. Consequences of between-Populations Differences in ABO Blood Types Frequencies
Frequencies of the A, B, and O alleles and of the respective ABO phenotypes are quite variable across human populations [
50]. Overall, on the Eurasian continent, there exists a gradient of blood group B whose frequency increases as one moves from west to east. This increase takes place at the expense of both blood groups A and O that are less frequent in East Asia than in Western Europe, the Middle East presenting intermediate frequencies [
94]. The African continent is characterized by a rather large and complex diversity of ABO blood groups frequencies between geographic locations and ethnic groups [
33]. The American continent is characterized by an extremely high frequency of the O blood type among Amerindian populations, most notable in its Central and Southern parts [
94]. Considering these large differences between geographical areas and human populations, variation in the impact of ABO phenotypes on COVID-19 is to be expected. Documenting an effect is more difficult in ethnically mixed populations such as those of Western Europe or of the Americas. As already mentioned, special care should therefore be given to alleviate biases introduced by the population admixture as the risk of COVID-19 is clearly associated with socioeconomical conditions, themselves associated with the ethnic background.
For all hypotheses raised to explain the links between ABO phenotypes and COVID-19, variations in relative frequencies represent a basic statistical issue. Obviously, it is more difficult to observe significant associations for less well represented phenotypes and this may contribute to explain some of the divergences observed between studies. In addition, and quite distinctly, within the framework of the anti-ABO hypothesis, the variation in ABO phenotypes frequencies affects the outcome in a more complex manner. Thus, anti-B antibodies from blood group A individuals may be protective from viral particles emitted by blood group B individuals. Reciprocally, anti-A antibodies from blood group B individuals can only be protective in the case of an ABO incompatible transmission event originating from a blood group A person [
60]. The relative proportion of blood group A and B in the population will therefore be critical. This may explain why blood group A appears as a risk factor more often than blood group B. Indeed, many studies were performed on population showing a higher representation of blood group A than of blood group B. Likewise, O blood group individuals if protected from transmission from non-O individuals, cannot be protected from transmission by other O blood group individuals. Consequently, the frequency of the O blood type in a population will be critical. At the individual level, the maximal protection of O blood group individuals will take place when they only represent a small fraction of the population. Inversely, when present in a large proportion their protection by anti-ABO antibodies will vanish and completely disappear if all individuals of the population were of the O blood type. With this in mind, we looked for a correlation between the reported odd ratios of O versus non-O blood types to the frequencies of blood group O in the studied population listed in Table I. To do this, we used studies that reported the O blood group odd ratios in comparison with all other groups and we also used the frequency of blood group O in the studied populations as reported by the authors or deduced from previously published data.
Figure 2A shows the results of the analysis, indicating that, as predicted by the hypothesis involving anti-ABO antibodies, O blood group individuals may benefit from a more important protection from SARS-CoV-2 infection in locations where blood group O frequencies are lower.
As the above observation suggested that the variable ABO frequencies across populations may differentially impact the outcome of the pandemic between countries, we looked for a potential correlation between ABO frequencies and SARS-CoV-2 attack rates worldwide. As already mentioned above, the relationship between an ABO blood type and susceptibility to COVID-19 is not straightforward, therefore individually relating each ABO blood type to the virus attack rate may be misleading. At the population level, the optimal protection that anti-ABO antibodies may provide should be attained when A, B, and O phenotypes are more or less evenly distributed because protection should occur only in ABO incompatible transmission events. In extreme situations, where one or two of these three major phenotypes largely dominate, incompatible events will be less frequently encountered. Thus, rather than looking for correlations between each ABO phenotype and SARS-CoV-2 attack rates, we looked for correlations between ABO coefficients of variation in different countries or geographical locations and SARS-CoV-2 attack rates. Protection conferred by anti-ABO antibodies should be optimal in populations presenting low coefficients of variation and inversely it is expected to be far less efficient or even not efficient at all in populations where they are elevated. ABO coefficients of variation were calculated using published frequencies of ABO blood types in different regions or countries. However, obtaining reliable SARS-CoV-2 attack rates that can be compared between countries is rather complex. Rates of detection obtained following RT-PCR testing are particularly unreliable since testing strategies have been very different between countries and within countries during the unfolding of the pandemic. Using data such as the number of COVID-19-related deaths/10
6 inhabitants may appear easier to compare. Yet, these may be greatly affected by variable reporting strategies between countries and most importantly by the age structure of the population as the risk of death increases exponentially with age [
95]. Older populations will therefore appear disproportionately affected, regardless of other parameters. More reliable data were recently provided by O’Driscoll et al. [
95] using seroprevalence analyses from 45 countries or regions. Using these published median values of proportions of SARS-CoV-2 seropositive individuals as of September 2020, we found 41 countries with reliable published national ABO frequencies. As depicted on
Figure 2B, a strong positive relationship can be detected between infection rates and the ABO coefficients of variation. It is noteworthy that countries of Central and South America where blood group O is highly prevalent show the highest SARS-CoV-2 attack rates, as predicted by the hypothesis on anti-ABO antibodies protection limited to ABO incompatible transmission events. Looking at a more homogeneous group of countries, both in terms of ABO coefficient of variation and of socioeconomic status, European countries also show a clear relationship between the distribution of ABO blood groups and infection attack rates (r
2 = 0.29,
p = 0.012).
Infection attack rates are very dependent on demographic and geographic variables, the national policies of protection that have been set up by governments, as well as by the socioeconomic conditions and inequalities [
96,
97,
98]. To determine whether the observed associations between ABO coefficient of variation and SARS-CoV-2 attack rates actually revealed underlying covariation with historical between-population socioeconomic inequalities, we looked for a potential relationship between the ABO coefficient of variation and the inequality-adjusted human development index (IHDI). This index has been created by the United Nation Development Program (UNDP) in order to best define the degree of human development by taking inequalities into account. As shown on
Figure 2C, we controlled that there was no relationship between ABO coefficient of variation and the IHDI. As unfavorable socioeconomic and demographic conditions have been discussed as important drivers of the epidemic [
99], we also tested whether the IDHI was associated with the attack rates for the 41 countries that could be included in our analysis. This was indeed the case (
Figure 2D), but the correlation was not as strong as that observed between attack rates and the ABO coefficients of variation. More refined indexes might be able to better capture the effect of inequalities and poverty on the impact of COVID-19, but that is beyond the scope of this article. Regardless, these observations suggest that the distribution of ABO blood groups strongly impacts development of the COVID-19 pandemic when looking at the level of populations. They suggest that the protective effect of anti-ABO antibodies is particularly high in Asian countries, important in Europe and North America, but unfortunately less important in countries of Central and South Americas where the frequency of blood group O is disproportionately high in comparison with those of blood groups A and B. This highly ABO-biased distribution is due to the large American Indian populations in these countries as American Indians are almost exclusively of blood group O [
100]. Interestingly, in the USA, the impact of COVID-19 is extremely high among non-Hispanic American Indians and Alaska Natives with attack rates 3.5 times higher than in the white population [
101]. It is also high among Hispanics (3 times higher compared to whites) [
21], followed by African-American (2.3 times higher compared to white) in parallel with the frequencies of blood group O in these subgroups [
21,
102]. Similarly, a study from the UK aimed at relating ABO phenotypes to the risk of COVID-19 in pregnant mothers, reported the results of separate analyses of White and BAME (Black, Asian, Minority Ethnic) women, the latter being overrepresented among the COVID-19 positive cases [
35]. The ABO coefficients of variation were widely different between the two subgroups, 60% and 28% for the White and BAME women groups, respectively. In accordance with the anti-ABO hypothesis, in the White COVID-19+ group, O blood group was modestly decreased in comparison to the White COVID-19− group (37% vs. 43%), but in the BAME cohort, the effect was much more pronounced (25% vs. 40%). It is thus tempting to hypothesize that the disproportionately high COVID-19 impact on the disadvantaged subgroups as well as in some countries might be explained by a synergistic effect between ABO blood group distribution and socioeconomic factors.
Although correlative and potentially influenced by other demographic covariates that cannot be ruled out, the data presented in
Figure 2 suggest that the protective effect of anti-ABO antibodies could greatly contribute to the large COVID-19 disparities between countries and population subgroups. It seems that non-pharmaceutical interventions, such as population confinement and social distancing, are efficient when the ABO coefficients of variation are low, as seen in some Asian countries that remarkably succeeded in controlling the epidemic. However, when they are high, the non-pharmaceutical interventions fail as illustrated by the examples of Latin American countries. When intermediate, as in Europe, they work to some extent, but are still affected by the West–East gradient of ABO frequencies, Eastern European countries tending to fare better, as of September 2020. All this seems to indicate that when the frequency of blood group O is below 40% and blood groups A and B are well balanced, the situation remains manageable, but when blood group O frequency increases above 40% and that the frequencies of blood groups A and B are unbalanced, management of the epidemic in absence of a vaccine turns out to be more difficult.
6. Conclusions
Collectively, the data presented above indicate that ABO blood groups influence the risk of SARS-CoV-2 infection and several studies additionally allow to ascertain an effect of the ABO phenotypes on disease severity. In both situations, blood group O appears protective in comparison with non-O types. As discussed above, the protective effect on infection could be mediated either by natural anti-A and anti-B antibodies or by a lower efficiency of furin cleavage in blood group O individuals. Both could lead to either a complete protection or to lowering the initial viral load, which may have important consequences by facilitating viral clearance by the immune system and preventing the cytokine storm and ensuing ARDS. The observation of a link between ABO coefficient of variation in different geographical areas and the odds ratios of blood group O relative to other blood types as well as the COVID-19 attack rates suggest that anti-ABO antibodies play a prominent role in protection against infection but that their impact is heavily influenced by the relative frequencies of ABO phenotypes in the population.
The previously well-documented influence of ABO phenotypes on hemostasis and vascular function, likely contributes to the lower severity of the disease observed among already severely affected patients of blood group O in comparison with non-O blood group patients. Other mechanisms could also contribute, including effects of the ABO polymorphism on inflammation and immune functions as well as on lipid metabolism.
At the individual level, the increased risk of severe COVID-19 symptoms associated with non-O blood groups is generally modest. Therefore, it is not clear if that information can be useful at the clinical level. Nonetheless, in selected groups of patients, such as patients with underlying cardiovascular diseases it may be of importance, as illustrated by the study of Sardu et al. on hypertensive patients which detected a 3.7 times increased risk of death in non-O blood group patients [
30]. Special attention could be important for such high-risk patients who may need early anticoagulant therapies in order to reduce cardiac injury, as suggested by the authors.
Until recently, the only evidence that natural anti-ABO antibodies played any significant biological role, besides their importance for blood transfusion and organ transplantation, came from artificial
in vitro observations and indirectly from the association between ABO phenotypes and SARS [
74]. The new data accumulated on COVID-19 point to a role of these antibodies in the control of outbreaks at the population level. Anti-ABO antibodies could also help controlling infection by other coronaviruses as these viruses are highly glycosylated and replicate in epithelial cells of the upper respiratory tract that strongly express ABH antigens in accordance with the
ABO and
FUT2 genes polymorphisms. There is evidence that anti-A and B titers are decreasing in relation with improved living conditions [
49]. This phenomenon could contribute to facilitate virus transmission in contemporary societies with high standards of living. Moreover, COVID-19 patients have lower levels of anti-ABO antibodies than control individuals, suggesting that these antibodies are protective only when present in sufficient amounts [
72]. Interestingly, levels of natural anti-glycan IgM, including anti-ABO, decrease with aging, which may contribute to the increased risk of infection in the elderly [
103]. Acting to increase these antibodies titers at the population level would thus be desirable in order to take full profit of this natural antiviral defense mechanism. Although a detailed discussion of the means by which anti-ABO antibodies could be raised at the population level is beyond the scope of this article, a possibility would be to use selected probiotic bacterial strains that express the A and/or B antigens as it is largely under stimulation of bacteria from the microbiota that these antibodies are naturally raised. A transfusion accident was reported where a platelet donor had an extremely high anti-B titer following ingestion of a probiotic preparation containing several species of harmless bacteria [
104]. Similarly, pediatric patients developed anti-B in association with probiotic use [
105], suggesting that this could be a broadly applicable strategy as a complement to the SARS-CoV-2 vaccine strategy. Populations where ABO coefficients of variation are low or intermediate could slow down the virus transmission to a large extent, making it easier to control the epidemic. Even populations that present high ABO coefficients of variation could benefit from the anti-ABO protective effect, albeit to a lesser degree. That would increase the efficacy of both non-pharmaceutical interventions and of vaccines if used as a complementary tool in the fight against COVID-19.