This study shows that large and complex synthetic DNA sequences can be successfully cloned in a single step into two poxvirus vectors: MVA and FPV and recombinant poxviruses could be grown to high titres without the recombinants reverting to their wild-type form. The vaccine candidates showed appropriate expression of recombinant proteins in infected/transfected cells and the b12 epitope (coincident with the CD4 binding site (CD4bs) of gp120) was shown to be held in common by the vaccine candidates. The CD4bs is an important target for NAb responses identified in HIV-1 infected individuals . In addition human cells infected/transfected with the vectors showed expression of authentic HIV-like VLPs. The HIV vaccine candidates were delivered by intramuscular injection of Chinese cynomolgus macaques in a prime-boost-boost vaccination protocol. The vaccines were tolerated without any adverse reactions. The vaccines elicited modest T cell responses in the immunised macaques but only macaque 1057 produced an HIV-specific antibody response which was highest after the third heterologous immunisation. However, the antibodies did not neutralise the panel of primary HIV isolates or the laboratory adapted, b12-sensitive isolate SF162 using the TZM-bl β-galactosidase assay. The TZM-bl neutralising antibody readout has been validated against protection from SHIV infection in passive transfer experiments . Our immunisation protocol was shorter than those generally used for subunit vaccines aimed at eliciting antibody responses but in keeping with those used for heterologous prime-boost aimed at eliciting T cell responses. Antibody responses capable of neutralising SHIV are generally apparent after the second subunit boost , but in natural HIV infection it can take some time to emerge [67–69]. We detected no evidence of NAb responses 5 weeks after the third heterologous immunisation.
The vaccine candidates directed VLP secretion from infected/transfected cells in vitro, however, we have not demonstrated VLP production following vaccination in vivo: a difficult subject to study without biopsying vaccination sites. The rMVA produced a prolific number of VLPs from infected HEK293 cells compared to the DNA and rFPV vaccines. Recombinant proteins in MVA were expressed from combination early/late promoters (both being effective in mammalian cells) whereas the recombinant proteins in FPV were expressed from early promoters alone (only early promoters being effective in FPV-infected mammalian cells). We have not proved that Env is incorporated in the membranes of the VLPs, although the appearance of Env spikes on TEM is highly suggestive. Others have also reported expression of the b12 epitope on Gag-Env pseudovirions  but not in the context of carriage by poxviruses. Expression and VLP formation from the plasmid constructs used in the DNA vaccine would probably have been enhanced if a single plasmid expressing both Env and Gag were used, but we were unable to obtain such materials. The Env expression plasmid employed is rev-independent. We used codon-optimised env consensus sequences for clades A and C which are known to be functional and CCR5-using. No consensus sequence for clade D env was available at the time, so we derived a codon-optimised version from the CCR5-using infectious molecular clone U88824. Functional consensus sequences were used where possible because these are believed to enhance NAb responses .
The reason for the failure to generate NAbs is not clear. It may be that the vectors employed simply do not generate good antibody responses despite the attempts to improve this with hC3d and CTB. The hC3d was incorporated towards the N-terminus of Env (essential if it is to be displayed on the external surface of a VLP) whereas the original work in rodents with hen egg lysozyme emphasised the importance of incorporation at the C terminus . Furthermore, most reports describe the use of murine C3d as molecular adjuvants [51, 72] but here we used hC3d because we reasoned it was more relevant for human vaccine development and our NHP model. In addition, we have not used triplet sequences of hC3d because highly repetitive sequences are rapidly deleted by poxviruses, and we predicted the trimeric structure of HIV Env would perform this function naturally anyway. The approach of using C3d as a molecular adjuvant in recombinant viral vectors has recently been shown to hamper antibody responses to certain antigens  and this study suggests that encoding C3d was counterproductive to the vector design. CTB was preferentially expressed in FPV not MVA, because MVA is known to block the effect of interleukin-1β  by production of a soluble receptor, and this would likely interfere with the adjuvant effect of CTB . Furthermore, the CTB was designed to be secreted from poxvirus-infected cells with no fusion with candidate HIV antigens. We have not proved that the CTB and hC3d expressed by the poxviruses are functional.
Since these experiments were conceived it has also become apparent that the native b12 epitope is a poor immunogen: it is located deep in the CD4bs, so the b12 MAb has an unusually extended variable loop in order to bind the epitope. Studies suggest that steric hindrance, e.g. by beta20-beta21 loop , would prevent good immune responses to this epitope in a similar fashion to steric hindrance of any coreceptor binding site epitopes . Furthermore, naturally glycosylated HIV-1 Env trimers are poor immunogens, so it is possible that further modifications to the Env amino acid sequence in order to better expose neutralising epitopes might be beneficial in addition to the cross-clade immunisation employed here.
Although we have focused on the b12 epitope it is quite possible that there were other cross-clade neutralising epitopes present in the vaccine candidates, whether on gp120 or gp41. For example, the highly conserved caveolin-binding motif (623WNNMTWMEW631) of gp41 is represented in the amino acid sequence of all the constructs [77, 78], although this does not appear to be immunogenic except when expressed in isolation. The TZM-bl β-galactosidase assay we employed would be expected to detect the effect of any antibody such as the gp120 MAb IgG1b12 that interfered with HIV-CD4 binding, HIV-coreceptor binding or fusion of HIV Env and target cell membrane. It is known that certain antibody subpopulations such as 2G12-like antibodies, may not be detected through the use of the TZM-bl assay  and that high levels of CCR5 expression can reduce sensitivity for antibodies such as 4E10 . Even so, this assay is the most standardised and widely applied assay for the measurement of neutralising antibodies [65, 80, 81] and alternative formats such as PBMC-based assays show great variability in sensitivity in inter-laboratory comparisons .
T cell responses were clearly seen on ELISpots to conserved Gag and Env peptides in the macaques at the end of the study (five weeks after the final rMVA immunisation). This finding is consistent with previous studies in cynomolgus macaques using DNA prime, MVA boost regimens [82, 83]. T cell responses in DNA prime, poxvirus boost regimens generally peak earlier than this at around 1 week post-immunisation , so it is possible that more vigorous T cell responses have been missed. It may also be the case that cross-clade T cell responses in macaques may not translate to humans, because the T cell epitopes are different and many are clade-specific.
Of note both antibody and T cell responses were best in the heaviest macaque 1057, the other two macaques were significantly smaller (see Additional File 4, Table S2). There was no obvious pathology at post mortem in any of the macaques. HIV and SIV vaccine candidates have not been extensively studied in Chinese cynomolgus macaques, and there is no data on MHC types, so future investigations may be better performed in the rhesus macaque model.
In conclusion, FPV and MVA are ideal replication-deficient viral vectors for HIV-1 vaccines due to their excellent safety profile for use in humans. This study shows that the DNA and poxvirus vectors used according to the immunisation protocol were poorly immunogenic in Chinese cynomolgus macaques. Furthermore, the antibodies elicited in the macaque did not neutralise primary or lab adapted isolates of HIV-1. Clearly it is very difficult to prove a negative result, and we cannot exclude the possibility that the viral vectors may elicit NAbs in combination with other vaccine candidates or in different model systems (e.g. Rhesus macaques), or with modifications to the vaccine vectors or adjuvants. The level of VLP production by the MVA recombinant was prolific, and this rMVA vaccine candidate may be worth revisiting with DNA and FPV vaccine candidates that are equally prolific producers of VLPs. We draw attention to the fact that we have published comprehensive sequence data and we will make our reagents available to bona fide vaccine researchers who wish to explore these issues.