Mother-to-child transmission of HIV-1: the ‚all mucosal‚ hypothesis as a predominant mechanism of transmission

In contrast with HIV-2 which is rarely transmitted from mother to child [1], HIV-1 is transmitted to foetus/infants by 13-48% of infected mothers [2]. HIV-1 can be transmitted from mother to child at three different times: antenatally, peri partum or postnatally [3]. At least 65% of infections in a non-breastfed paediatric population are acquired around the day of birth, presumably during labour and delivery [4-6] even in the context of antiretroviral treatment of pregnant women [7]. By contrast, recent reports considering the temporal acquisition of a positive PCR by HIV-infected infants who had been breastfed, suggest that breastfeeding may contribute to between one- and two-thirds of infections [8,9]. If the timing of mother-to-child transmission is not fully understood, the mechanisms of transmission, such as transplacental, blood-borne, iatrogenic, ascending infections and mucosal exposure, are even less clear [10]. Is there a single mechanism, or are there multiple mechanisms of acquisition of HIV-1 in foetus/infants? Studies have attempted to address directly or marginally this question by sequencing various HIV genes in mother-child pairs and comparing HIV-1 diversity in the two related hosts [11-17]. With a notable exception [12], these studies have suggested that in most cases, child HIV-1 variants are phylogenetically clustered around a single maternal variant, and therefore, that a single or closely similar mechanism of transmission may be involved in most cases. However, these studies suffered many methodological difficulties in terms of sample size, standardization of laboratory techniques (such as the choice of genes to be sequenced) and interpretation of results. As shown in Table 1, mucosal exposure to HIV-1 can be considered at each period of transmission: by this it is proposed to unify route and cellular mechanism of contamination to explain both the timing of transmission and response to prophylaxis by antiretroviral agents. It is hypothesized that mother-to-child transmission occurs predominantly through contact between HIV-1 virions or infected cells and foetal/infant mucosal surfaces, and by using alternative membrane receptors such as galactosyl ceramide and 38 sulfogalactosyl ceramide (GalCer/GalS), C3 receptor or Fc receptor. The hypothesis is based on three indirect but converging arguments: the mode of action of antiretrovirals; the lack of impact of chemokine receptors polymorphism on mother-to-child transmission of HIV-1; and the ontogeny of inflammatory and epithelial cells.Table 1: Potential mucosal exposure to HIV-1 and timing of transmission of HIV-1 from mother to foetus/child.HIV-1 phenotype and mode of action of antiretrovirals A body of evidence suggests that syncytium-inducing (SI) strains of HIV-1, which encompass T-cell tropic, CCR5 and/or CXRC4 receptor-using variants, are more frequently involved in mother-to-child transmission than in any other mode of HIV-1 transmission [18,19]. In a group of 30 HIV-1-infected mothers and 12 HIV-1-infected babies, Scarlatti et al. showed that mothers who harbour virus able to replicate in cell lines (rapid/high virus, corresponding mainly to SI viruses) have a significantly higher risk of infecting their children than mothers with slow/low [corresponding to non-SI (NSI)] virus [18]. However, numerous studies have demonstrated that most vertical transmission occurs with NSI, macrophage-tropic strains [20,21]. Antiretroviral beneficial effects seem to be highly dependent on viral phenotypes [22]. A marked differential efficacy of zidovudine has been observed depending on the HIV-1 phenotype [22] and conversion from a NSI phenotype to a SI phenotype is not prevented by zidovudine [23]. In patients carrying both SI and NSI variants, changes in viral load as a result of zidovudine therapy are explained by the loss of coexisting NSI virus [24]. As the only one proven effective agent, zidovudine is the most widely used drug in prophylaxis of mother-to-child transmission of HIV-1. As in the ACTG 076-ANRS 024 trial, which showed that zidovudine is highly effective in reducing the risk of vertical transmission, its efficacy was only partly explained by a decrease in viral load [25]. This, together with the observation that CD4 counts and viral load measurements are not predictive of zidovudine efficacy in patients carrying an HIV-1 SI virus [26], strongly suggest that, at least in women treated with zidovudine, SI strains are involved in a substantial proportion of vertical transmission events. However, most of the data available to date suggest that the mother-to-child transmission of HIV occurring despite maternal zidovudine treatment is predominantly with NSI strains. Impact of chemokine receptors polymorphism on mother-to-child transmission To enter susceptible cells, HIV-1 uses not only the CD4 receptor but also different kinds of chemokine receptors as co-receptors [27,28]. Schematically, NSI viruses are implicated in the vast majority of HIV-1 sexual transmission events and are predominantly macrophage-tropic viruses which preferentially use CCR5 as a co-receptor for viral entry; on the other hand SI viruses demonstrate a marked T-cell tropism and use both CCR5 and CXCR4 co-receptors. In sexual transmission of HIV, the virus encounters stratified epithelia of vagina, exocervix and anal margin which are densely infiltrated by dendritic cells bearing the CCR5 but not the CXCR4 receptor [29]. Without any evidence of epithelial cell infection, HIV enters dendritic cells via CD4 and CCR5 [29] and is transported to lymph nodes [30,31]. It has been shown that some highly exposed but HIV-1 uninfected individuals are carrying a homozygous 32 base-pair deletion of CCR5 which leads to a truncated protein that is not expressed on the cell membrane [32]. Heterozygosity for this D32 allele is associated with a slower course of HIV disease, with moderate or no reduction of homosexual transmission of HIV-1 but with some reduction of heterosexual transmission [33]. Indeed, D32 heterozygous cells express detectable levels of CCR5 co-receptors on the cell membrane and are infectable by HIV but with a much lower efficiency than wild-type cells [34]. Recent studies on a very large number of mother-child pairs (overall more than 1800 have been reported) suggest that D32 heterozygosity in the child is associated with a slower course of paediatric HIV disease but has no effect on mother-to-child transmission of HIV-1 [33,35-40]. Table 2 summarizes the results of published reports on CCR5 chemokine receptor alleles in children and the risk of mother-to-child transmission of HIV-1 as well as a preliminary pooled analysis. Taken together, these data suggest that heterozygosity for the D32 allele of CCR5 alone is not protective against mother-to-child transmission as it appears to be against heterosexual transmission. The lack of association between D32 heterozygosity and the risk of mother-to-child transmission of HIV-1 suggests that the mechanism of entry of HIV-1 into the front-line cell in the foetus/infant occurs via a receptor other than CD4/CCR5. Indeed, several alternative receptors for HIV cell attachment and entry/transport in mucosal epithelial cells have been demonstrated or suggested such as GalCer/GalS [41,42], C3 receptor [43] and Fc receptor (for opsonized HIV) [44].Table 2: CCR5 chemokine receptor alleles in children and the risk of mother-to-child transmission of HIV-1: summary of published reports.Human intestine has been infected successfully by HIV-1 in vitro [45] most probably through a GalCer/ GalS-dependent process. GalCer/GalS are present on the cell surface of both absorptive enterocytes and M cells. M cells are highly specialized epithelial cells representing about 2% of follicle-associated epithelial cells (human Peyer‚s patches). Conversion of human enterocytes into an M cells is dependent on mucosal associated lymphoid tissue (MALT), and in particular, lymphocytes present in the lamina propria [46]. M cells have very particular functions of trans-cellular transport of antigens (called transcytosis) from the lumen, eased by the absence of a brush border and glycocalyx, and priming of the MALT consisting of B and CD4 T cells, macrophages and dendritic cells that protrude into a large intraepithelial pocket [47]. Recently, HIV-1 virions have been shown to be rapidly transported by transcytosis across rabbit M cells in vitro [48]. Moreover, transcytosis of HIV-1 delivered by polarized infected cells has been observed across human absorptive enterocytes [49], a process that can be blocked by HIV-1 specific secretory IgA [50]. Recently, antigen- transporting M cells, or functionally similar cells [51] situated in the lymphoepithelium have been shown to similarly provide access of HIV-1 to MALT nasopharyngeal tonsils and adenoids [52]. M cells and/or enterocytes are, therefore, plausible candidates for a portal of entry for HIV-1 transmitted from mother to child, through a CD4/CCR5-independent mechanism. The front-line cell encountered by HIV-1 is in this case an epithelial cell, by contrast with CD4/CCR5-bearing dendritic cells generally involved in sexual and blood-borne transmission of HIV-1. Ontogeny of intestinal inflammatory and epithelial cells In the normal foetal ileum, CD3 T cells are present in the lamina propria as small aggregates beneath the epithelial layer by the 11-19th week of gestation [53]. The CD4:CD8 cell ratios of these lymphocytes are higher in foetal than in postnatal ileum [53] and there are approximately three to five intra epithelial lymphocytes per 100 epithelial cells [54]. At 11 weeks gestation, embryonic Peyer‚s patches are identifiable as small aggregates of HLA-DR cells and by 19 weeks gestation B-cell follicles are constituted and T cells occupy the inter-follicular zones [55]. Therefore, it is only by mid-gestation that follicle-associated epithelial and MALT are fully mature in the foetal gut. If this hypothesis is true, this could explain why HIV-1 acquisition is uncommon in early gestation [56], which is compatible with the observed timing of mother-to-child transmission of HIV-1 [57]. However, when it occurs, it may well be through mucosal contact of the foetus with contaminated amniotic fluid (ascending infection) rather than through transplacental transmission. However, as amniocentesis is contraindicated in most HIV-infected pregnant women, there is only a limited number of small-scale studies on the presence of HIV in human amniotic fluid. Conclusions As HIV-1 is present either as free virions or as HIV-1 infected cells in amniotic fluid [58], maternal cervico-vaginal secretions [59-61], and in colostrum and breast milk [62,63], the foetus, neonate and breastfed infant/child is in close mucosal contact - oral, tonsilar, bronchial, intestinal etc. - with the virus for a long period of time. Based on converging arguments, it is hypothesised that HIV-1 involved in mother-to-child transmission is more frequently of the SI phenotype and has CCR5-independent usage than viruses involved in other transmission routes. It first encounters susceptible cells by receptors other that CD4/ CCR5, and it uses a mucosal portal of entry in most cases to secondarily infect submucosal lymphocytes, macrophages or dendritic cells. This hypothesis is compatible with the observed higher risk of contamination of the first-born of twins of HIV-1 infected mothers, as their mucosal surfaces are exposed to a larger amount of contaminated body fluids [64]. It may also explain why Caesarean section has a synergistic effect over zidovudine in preventing mother-to-child transmission of HIV-1, presumably by avoiding mucosal exposure of the new-born with genital secretions of its mother during vaginal delivery [65]. This hypothesis does not rule out uncommon transmission by maternal microtransfusion or infection of trophoblastic cells. Absence of transmission is also consistent with this model, as transmission can conceivably be modulated by host response such as the relative maturity and efficacy of the foetus, new-born and infant local immune responses. Indeed, secretrory IgA and IgM are extremely potent antibodies able to block transcytosis of HIV-1 [50], a mechanism that could be taken into account in preventing mother-to-child transmission by active or passive immunoprophylaxis. If this hypothesis is correct, it may have profound implications for designing and implementing interventions aimed at the reduction of mother-to-child transmission of HIV-1. On the one hand, interventions that are delivered predominantly during the intra partum phase (e.g. short regimens of antiretroviral agents) or primarily addressed to prevent mucosal infections (e.g. vitamin A supplementation) should be encouraged as they are likely to be effective in early transmission through their predominant and direct effect on HIV in the maternal genital tract and on early breastfeeding transmission. Other interventions that limit mucosal contacts or that reduce viral burden in genital secretions, such as obstetrical procedures, use of prophylactic antimicrobial agents or local microbicides during pregnancy, labour and delivery may also be of some benefit. On the other hand, in prolonged breastfed children born to HIV-1 infected mothers treated with antiretrovirals, it has to be feared that those who are born without the virus as a result of the efficacious treatment of their mothers, but who remain with a particular mucosal susceptibility toward HIV-1, will acquire HIV-1 postnatally. Breastfeeding may well ultimately erase the net beneficial effect of maternal antiretroviral treatment observed on mother-to-child transmission of HIV-1 in non-breastfed children [66]. This hypothesis of a predominant mucosal transmission of HIV-1 from mother to child will have to be scientifically validated but can immediately be translated into designing and evaluating combinations of interventions, especially in developing countries, with the aim of providing the most appropriate and efficacious treatment at the lowest possible cost. Acknowledgements I thank N. Meda (Centre Muraz), F. Dabis (University of Bordeaux), T. Mertens (WHO), I. de Vicenzi (UNAIDS), M. Carael (UNAIDS) and L. Mandelbrot (Hôpital Cochin-Port Royal, Paris) for their comments and criticisms on this manuscript.