In this study, we demonstrate that the elicitation of co-receptor site directed antibodies by the YU2 gp140-F trimers requires the presence of primate CD4. We show that the relatively homogenous, soluble, stable YU2 trimers bind to human CD4 with high affinity in a solution-based assay that, by design, should be independent of oligomeric influences on functional affinity by avidity-dependent interactions. Analysis of the interaction between the prototypic co-receptor antibody, 17b, and the trimers demonstrated that high-affinity and high-avidity binding is detectable even in the absence of CD4. Binding of the trimers to primate cell-surface CD4, but not to cell-surface rabbit CD4 was also shown. WT rabbits and monkeys inoculated with the CD4-binding YU2 Env trimers formulated in the same adjuvant system elicited an overall similar pattern of HIV-1 in vitro
neutralization against the viruses tested. However, cross neutralization of HIV-2 in presence of sCD4, an assay that is diagnostic for the detection of CD4i antibodies, was observed initially in sera derived from monkeys inoculated with the YU2 trimers but not in WT rabbits. Taken together, these data strongly suggest that Env-CD4 complexes generated in vivo
upon inoculation are the source for elicitation of the CD4-induced antibodies following vaccination. This observation was confirmed by the inoculation of Env trimers into rabbits transgenic for human CD4 and the detection of CD4i antibodies in the sera of these animals, in contrast to WT rabbits, revealing conclusively the mechanism of their generation (see schematic ). Consistent with this observation, high levels of co-receptor-directed, HIV-2 (+CD4) cross neutralizing antibodies were detected in 20 of 20 human serum samples from the VaxGen phase III clinical trial using monomeric gp120 
, suggesting that during natural infection shed, soluble gp120 can elicit these antibodies 
Schematic representations of the pre-CD4 and post-CD4 conformations of Env.
The induction of co-receptor site directed antibodies in non-human primates and humans is consistent with previous reports that detected the presence of CD4i, co-receptor directed antibodies in gp140-immunized humans (GMS, unpublished observations), as well as in naturally infected humans 
and following SHIV challenge of naïve non-human primates 
. However, the mechanistic basis for the elicitation of CD4i antibodies was not previously addressed in a direct manner. In this study, we present a controlled experiment, which demonstrates that the elicitation of the co-receptor binding site antibodies by Env alone requires the presence of, and likely direct interaction with, primate CD4. This requirement has not previously been defined, despite numerous Env trimer immunogenicity experiments performed to date in both monkeys and non-primate species 
. This is in part, because the HIV-2-based neutralization assay diagnostic for CD4i antibodies was a relatively recent development and is more definitive for neutralizing capacity directed at the co-receptor binding site then are binding assays employed by us and others previously 
. Elicitation of CD4i antibodies in primates by Env trimers also nicely illustrates another potential mechanism of immune escape by HIV-1. Not only does Env binding to CD4 obscure a conserved surface that, if it was highly immunogenic, might elicit antibodies capable neutralizing a broad array of isolates (essentially antibodies mimicking the soluble primary receptor), but the binding event induces a second conserved and apparently immunogenic region: the co-receptor binding site. The CD4i antibodies directed against this region are not generally able to neutralize primary isolates in vitro 
. This is likely due to a commonly elicited selection pressure that renders the co-receptor binding inaccessible to most antibodies of this type 
. The inability of the CD4i antibodies to control HIV-1 infection is supported by data from a recent study where elicitation of CD4i antibodies can be detected prior to the detection of autologous virus neutralization capacity in sera derived from HIV-1 infected patients 
, as well as the data here, which indicates that they were elicited in the phase III Vaxgen clinical trial where no protection was observed. However, a recent study in non-human primates suggests that the presence of CD4i antibodies (as determined in vitro
by the same HIV-2 detection assay as used here) is associated with more rapid viral clearance following SHIV162P3 challenge 
, illuminating that this is an area worthy of further investigation.
In the present study, the most likely in vivo
source for presentation of the CD4i region to the humoral immune system is by direct interaction of the trimers with cell-surface CD4 displayed on CD4-expressing T cells or other CD4-positive cells of the hematopoetic lineage. It is also possible that low levels of CD4 are shed from CD4+
cells into interstitial spaces and soluble complexes are formed. Detection of low levels of sCD4 has been previously reported in humans 
, although in the assay used here we could not detect soluble CD4 in the sera of animals examined. It is also possible that trimer binding to cell-surface CD4 induces shedding of complexes, but we could find no such evidence for soluble Env-CD4 complexes in the sera.
The implications of inducing the co-receptor binding site has been discussed extensively at the level of entry, but less so at the level of antibody elicitation. That the CD4i antibodies are not elicited by trimers in the absence of CD4, even though the gp140-F molecules are well recognized by 17b, and that CD4 induction of the epitope in the trimer context is not a requirement for 17b binding, may seem to be a bit of a paradox. However, we interpret these data to indicate that the conformational fixation imparted by CD4 binding to gp120 is a critical requirement for the naïve, germline B cell repertoire to efficiently recognize the site as opposed to the affinity-matured 17b antibody, which can likely induce the fit of its epitope in the absence of CD4 (see Schematic ). This is more likely to be an affinity limitation of the non-affinity matured BCR repertoire, although it is possible that it is somehow related to binding limitations of Ig molecules on the surface of B cells and epitope accessibility issues.
Another possible interpretation of the data is that elements of the pre-CD4 co-receptor binding site are immunogenic, but do not elicit antibodies that cross react efficiently with the fully formed site induced by CD4. However, the 17b blocking assay, using a form of the gp120 core with the potential to be recognized by any antibodies to the co-receptor site indicated that the pre-CD4 state of the trimers did not elicit many antibodies directed toward this region compared to those elicited in the presence of primate CD4 (). Also, we cannot rule out that array of gp140 or gp120 on the surface of CD4+ cells might also enhance the elicitation of CD4i antibodies in a manner independent of conformational fixation. However, the study by DeVico et al 
clearly demonstrates that covalent gp120-CD4 complexes incapable of binding cell-surface CD4 efficiently elicit CD4i antibodies, so CD4-dependent cell-surface array cannot be the only explanation for their elicitation in the presences of primate CD4.
The implications of the data presented here are also an important consideration for vaccine candidates designed to elicit neutralizing antibodies against the conserved gp120 CD4 binding site. The Env CD4bs likely remains fully accessible in animals without human or primate CD4, however the elicitation of the CD4i antibodies in animals with primate CD4 indicates that this is likely not the case in species harboring CD4 molecules with a high affinity to Env. These results suggest that a fraction of the population of a CD4-binding-competent immunogen will interact with primate CD4 and thereby occlude the CD4 binding region on this protein subset. It is possible that the subtle differences detected in the neutralization profile between WT rabbits and monkeys occur as a result of such an interaction, partially altering the spectrum of antibodies that are elicited. However, the fractional component of the inoculate which binds to CD4 as yet remains to be determined, and may not be absolute as the overall HIV-1 neutralization profile elicited by the trimers used in this study was similar between the rabbits and the non-human primates.
It was also previously shown that HIV Env-CD4 interaction resulted in altered CD4+
T cell function in vitro 
and it was suggested that elimination of Env interaction with CD4 in the context of vaccination might be beneficial to better elicit functional T cell help and more potent neutralizing antibody responses. From that study and the data presented here it will be interesting to assess if Env variants that do not bind CD4, but still retain the ability to bind CD4 binding site antibodies might make better immunogens than do unmodified YU2 gp140-F proteins. Alternatively, redirecting the immunogen more efficiently to B cell and antigen presenting cells might also overcome any potential detrimental effects of Env-based immunogens interacting with primate CD4. Follow up immunogen trimer design, characterization and immunogenicity studies are warranted to clarify these issues further in the near future.