The epitope of b12 is expressed on numerous recombinant Env-derived immunogens that have been tested pre-clinically and clinically for the elicitation of antibody responses similar to b12, but there is no indication that b12-like neutralizing antibodies have been elicited by such constructs. The reasons for the failure of recombinant Env immunogens to elicit b12-like neutralizing activities are not well understood. Understanding therefore how b12 was generated in the context of natural infection is a critical piece of information that is currently missing; information that would be very useful for the design of immunogens capable of eliciting b12-like antibodies.
Our results indicate that the germline version of b12 is not recognized by diverse Envs from clades A, B or C; Envs which are recognized by the mature b12 antibody. Our results therefore suggest that a potential reason why recombinant soluble Env immunogens have failed so far to elicit b12-like antibody responses may be the inability of these constructs to engage the germline BCR versions of such antibodies. In the absence of such an engagement, the process of B cell maturation that would lead to the eventual production of b12-like antibody is unlikely to be initiated.
The variable antibody regions of the mature and germline sequences of b12 differ by 21% for the heavy chain and 19% for the light chain and even more importantly, the four VH amino acids of mature b12 that make direct contact with Env, differ between the mature and germline sequences. When these four key amino acids are mutated (individually or in combination) into those present in the mature VH sequence (germline-to-mature mutations) the germline heavy chain acquires Env-binding abilities. However, the binding is dependent on the co-expression of a mature variable light chain and is Env target dependent. In fact, in certain cases, germline-to-mature conversion of all four key positions in the VH chain does not lead to recognition of certain Envs that are readily recognized by the fully matured antibody. These observations suggest that the broad Env-recognition ability of b12 necessitates a larger number of somatic mutations to take place in the VH region during affinity maturation. Also, germline-to-mature conversion of all four positions did not always lead to HIV-neutralization even when they allowed Env-binding. Therefore the broad neutralizing activity of b12 was acquired following the introduction of multiple mutations in the germline VH region. Some of these mutations may have allowed the maturing b12-like antibody to bind Env without displaying its broad neutralizing potential. Identifying the minimal number of germline-to-mature mutations that are necessary for b12-like antibodies to display cross-neutralizing activities will assist the design of immunogens and immunization protocols that will lead to the minimum maturation required of the germline b12 VH sequence to adopt its broadly neutralizing form.
The critical role of the heavy chain of b12 for Env-binding is highlighted by the fact that when the mature heavy chain is substituted by the germline heavy chain (while the light chain is kept in its mature form) only a single recombinant Env (out of 53 tested) was recognized. This Env, QH0692, is derived from a clade B virus isolated during the first 3 months of infection from a patient from Trinidad 
. If only the heavy chain, and not the light chain of b12 interacts with Env (as revealed by the crystal structure of b12 bound to the HxB2 gp120 core 
, then it is possible that the germline heavy chain of b12 binds the QH0692 Env. In this regard, the QH0692 Env is unique (for currently unknown reasons) among all the recombinant Envs tested here.
The identification of a recombinant Env that is recognized by the germline heavy chain of b12 supports the hypothesis that a very small number of Envs are capable of engaging the germline BCRs of b12-like antibodies. Also, this finding supports the utilization of QH0692-derived constructs as future immunogens to engage the fully germline version of the b12 BCR. Interestingly, the QH0692 Env was not recognized by the gHC/mLC of the other two anti-CD4-BS antibodies studied here. In fact the gHC/mLC versions of NIH45-46 and of 3BNC60 did not recognize any of the Envs tested. These observations suggest that a single Env might not engage the germline BCR of these three antibodies. From the point of view of immunogen design then, more than one Env immunogen may need to be engineered to engage all three germline BCRs tested here.
Reversion of the mature light chain to germline light chain abrogates the ability of the antibody to bind the majority of Envs. Thus, somatic mutations in the germline light chain are required to take place for the most optimal orientation of the heavy chain for Env-binding, or for stabilizing the interaction of the heavy chain with Env. The optimal engagement of b12-like antibodies with Env appears therefore to require multiple somatic mutations to take place in both light and heavy chains. Similar observations were made with two additional anti-CD4-BS antibodies, NIH45-46 and 3BNC60. Overall our results support the hypothesis that the pathway of affinity maturation which is required for the binding of anti-CD4-BS antibodies to diverse HIV Envs depends on somatic mutations introduced within both the heavy and light chains of these antibodies. Our data also suggests that depending on the Env target, the relative contribution of the heavy and light chains to Env recognition will vary.
Our results also indicate that the ability to bind diverse Envs is acquired gradually during affinity maturation, and is not inherent in the germline versions of the antibodies studied here. Importantly, the above described results show that some Envs can bind partial germline antibodies. Such Envs can serve as platforms to design Env proteins capable of engaging the germline versions of the antibodies examined here.
Although the above observations were made with the soluble versions of the mature, germline and chimeric antibodies, they are biologically relevant. The germline b12 BCR was unable to mediate intracellular calcium mobilization (an indication of B cell activation via BCR-antigen engagement). When the germline heavy chain was combined with the mature light chain, Env-mediated B cell activation was not observed. In contrast when the mature heavy chain was combined with the germline light chain, Env-specific B cell activation was observed, but at levels far reduced as compared to those recorded when the mature heavy and mature light chains were combined. Despite the fact that trimeric QH0692 gp140 bound to both gHC/mLC and mHC/gLC, it was not able to mediate calcium mobilization through these two BCRs. This contrasts with the above-discussed observation made with trimeric SF162 gp140, which bound the mHC/gLC chimera and mediated calcium mobilization through the corresponding BCR. At this stage we can only speculate that the inability of the trimetric QH0692 gp140 to activate the mHC/gLC BCR is related to inefficient BCR cross-linking on the surface of B cells.
Although we examined diverse Envs from three different clades, potentially a much larger screen of Envs is required to identify those rare constructs that engage the germline BCRs of the type of antibodies we examined here. For instance, it may be that the development of such antibodies during natural HIV-1 infection depends on the emergence of particular Env clones in certain HIV-infected subjects; clones with particular structural characteristics that allow them to engage the germline BCR form of bNAb to the CD4-BS. In the context of HIV-1-infection, an alternative possibility is that the maturation of bNAb precursors was initiated in response to non-HIV antigens, and that the resultant maturation antibody intermediates were capable of recognizing bNAb epitopes on Env, as proposed for anti-gp41 antibodies 
. We also note the possibility (not investigated here) that in the context of infection, the germline BCRs of these antibodies are initially engaged by membrane-anchored Env (for example, on the virion surface or on the surface of infected cells).
As mentioned above it is not possible to predict the exact amino acid composition of the VD and DJ joining regions in the true germline versions of the antibodies examined here. Therefore, the amino acids present in the mature antibodies in these two joining regions were left unchanged. One could argue therefore that these amino acids are not present in the ‘true’ germline BCRs of these antibodies, and that in the absence of other mutations in V and J regions, they play an inhibitory role in binding to certain recombinant Envs. Finally, we note that the results presented here were obtained with three NAbs whose epitopes are located within the same region of Env; the CD4-BS. It remains to be determined whether similar observations will be made with the germline versions of NAbs that bind to epitopes outside the CD4-BS. In this regard, it was reported that the germline versions of two MAbs, CH01 and CH04, which bind complex quaternary epitopes located outside the CD4-BS, can recognize certain recombinant Envs 
. Similarly, germline versions of the anti-gp41 MAb 2F5 recognize certain forms of recombinant gp41 
Overall our study provides new insights as to a potential reason why recombinant Env immunogens have failed to elicit anti-CD4-BS antibodies that display cross-neutralizing activities. Our results indicate that, as a first step in eliciting broadly neutralizing anti-CD4-BS antibodies by immunization, Env immunogens should be designed that engage the germline BCR versions of broadly neutralizing anti-CD4-BS antibodies.