Discussion
Probably due to mimicry of A or B antigens by flora or infectious bacteria, individuals acquire anti-A or -B blood group antibodies to the antigen that they do not synthesize. These so-called natural anti-histo-blood group antibodies have long been suspected to play a role in anti-viral immunity since viruses may carry ABH structures as terminal carbohydrate motifs of their envelope glycoproteins or possibly as inserted glycolipids (Greenwell 1997). In line with this concept, a monoclonal anti-A was shown to neutralize HIV produced by lymphocytes from blood group A donors only (Arendrup et al. 1991). More recently, anti-A or -B from human serum were shown to sensitize HIV to complement-mediated inactivation (Neil et al. 2005). Likewise, measles virus produced by cells expressing either A or B histo-blood group epitopes was neutralized by natural anti-HBGAs in a complement-dependent manner (Preece et al. 2002). Though these in vitro data suggest that natural anti-HBGAs may provide protection against some viruses, they have not been substantiated by epidemiological observations so far. If natural anti-A or -B serum antibodies provide protection, it is expected that during an outbreak, blood group O individuals should experience a lower risk of infection than non-blood group O individuals. This has not been observed for either HIV or measles virus at present. Yet, it is precisely what was observed in the case of a hospital outbreak of SARS in Hong Kong, where O blood group individuals appeared at a much lower risk of being infected by SARS-CoV than subjects of other blood types (Cheng et al. 2005). Interestingly, SARS-CoV infects cells that express ABH antigens according to the individual's ABO phenotype. Indeed, SARS-CoV infection has been documented in pneumocytes, enterocytes of the small intestine as well as in cells of the kidney distal tubular epithelium, all cell types known to be able to synthesize ABH antigens (Chen and Subbarao 2007; Gu and Korteweg 2007). Since the glycosylation of viral glycoproteins necessitates the glycosylation machinery of the infected cell, viral particles synthesized in cells that may express that histo-blood group antigens are expected to be tagged with these antigenic motifs, and therefore natural antibodies directed against these carbohydrate tags may have a protective role (Seymour et al. 2004).
In order to determine if anti-HBGAs could block SARS-CoV entry into target cells, we used an experimental model of cell adhesion that has been developed with the aim of screening molecules that block the virus entry without using infectious particles. The model allowed us to show that either a monoclonal anti-A antibody or natural plasma anti-A specifically inhibited the SARS-CoV S protein/ACE2-dependent adhesion. This is in accordance with the hypothesis of the protective role of natural anti-HBGAs and strongly suggests that the low risk of infection of blood group O individuals during the Hong Kong hospital outbreak was due to the presence of these antibodies prior to the outbreak.
Although the RBM of the SARS-CoV S protein does not involve any glycan chain, clusters of glycosylation sites are located in its vicinity (Li et al. 2005; Han et al. 2007). Large molecules such as lectins or antibodies binding to these glycans are thus expected to interfere with the S protein/ACE2 interaction. This has already been observed with mannose-specific lectins which have shown anti-viral activity against SARS-CoV by blocking virus attachment to its receptor (Keyaerts et al. 2007). Our data indicate that natural anticarbohydrate antibodies could have a similar effect. In addition to the blocking of virus attachment to its receptor, natural antibodies could block entry or opsonize viral particles leading to complement-mediated neutralization (Neil et al. 2005). Moreover, it was recently shown that natural antibodies can contribute to help the generation of cytotoxic T cells against the pathogen (Stäger et al. 2003; Dürrbach et al. 2007). These additional mechanisms of protection may also have participated to the protection of blood group O individual during the SARS outbreak.
Since the epidemiology of SARS has been well documented (Anderson et al. 2004; He and Chinese 2004), it was possible to develop a model of the virus transmission that takes into account the effect of the ABO polymorphism. Critical transmission parameters were deduced from those of the global SARS outbreak and from those of the Hong Kong hospital outbreak where the ABO effect was observed. The model considers that no prophylatic measures were taken; although this is clearly unrealistic, it allows us to fully appreciate the impact of the blood group polymorphism itself. Of note, the expression of ABH antigens in epithelial cells where SARS-CoV replicates is also controlled by polymorphisms of the FUT2 gene. Thus, individuals with two FUT2 null alleles, the so-called nonsecretors, are unable to synthesize H antigen and hence A or B antigens in these cells (Marionneau et al. 2001). For simplicity, the model did not consider such individuals since with regard to the virus transmission, they would behave as O blood group donors. Including them in the analysis is therefore similar to slightly increase the pool of O individuals. Since anti-A and anti-B titer interindividual variability is quite high, and since we observed that plasmas from O blood group individuals with low anti-A titers were not inhibitory in the cell adhesion assay (not shown), we considered transmission parameters allowing for a moderate ABO effect only. In this case, virus transmission in incompatible ABO situations remains possible but with a lower probability of occurrence than in compatible situations. That is certainly more realistic than the case of the strong ABO effect where ABO incompatibility completely impairs virus transmission. The latter case was analyzed in order to evaluate the maximal potential of the ABO polymorphism. The model indicated that both in the presence of a moderate or a strong group effect, virus transmission was decreased, supporting the hypothesis that natural anti-A and -B antibodies can contribute to protect against selected viral diseases at the population level. Less intuitively, the model shows that the main effect of the natural anti-histo-blood group antibodies is to delay and slow down the epidemic. In the case of a full protection (strong effect), the delay between the occurrence of the first cases and the full development of the outbreak can be very large. It is linked to the frequency of O individuals in the population, but remains very significant even in a population with an unusually low blood group O frequency such as the Aïnous of Japan. This delay, already clearly visible when taking into account a moderate ABO effect only, might have had an adaptative value, in past epidemics of other viruses with transmission characteristics similar to those of SARS, since it allows for modifications of behavior limiting the spread of epidemics. This could have contributed to the maintenance of the ABO polymorphism throughout human evolution and history. Mean natural anti-A or -B titers tend to decrease over the years in developed countries, possibly due to improved hygiene (Dr. A. Blancher, personal communication). It is thus possible that in the past their protective effect was higher than in most contemporary populations. It could thus be of interest to raise the levels of anti-A or -B in all populations so as to slow and limit the spread of some emergent pathogens. This could prove a valuable prevention strategy against SARS but also against other coronaviruses which are responsible for a significant proportion of common colds and can contribute to more severe respiratory tract infections (van der Hoek 2005).