The animal study was carried out in strict accordance with the UK Animals (Scientific Procedure) Act 1986, and the protocol was approved by the local Ethical and Welfare Committee of the University of Cambridge and the UK Home Office (Project license no. 80/2238).

HIV-1 gp140_SF162 has been previously characterized and was obtained as culture supernatant from Novartis Vaccines & Diagnostics, USA. The gp140 glycoprotein was purified using a published protocol [48]. Briefly, supernatant was loaded onto a Galanthus nivalis agarose affinity column equilibrated with buffer (20 mM Tris, 100 mM NaCl, pH 7.4), and bound gp140 was eluted with 500 mM methyl mannose pyranoside in equilibration buffer. The eluate was then loaded onto a DEAE column equilibrated with buffer (20 mM Tris, 100 mM NaCl, pH 8.0), such that contaminants were bound to the column while the gp140 flowed through. The flow-through was adjusted to pH 6.8 with 10 mM NaPO₄ before loading onto a ceramic hydroxyapatite (CHAP) column that has been previously equilibrated with buffer (10 mM Na₂HPO₄, 100 mM NaCl, pH 6.8). The flow through, which contained the purified glycoprotein, was collected and then concentrated to 1 µg/µl using an AMICON YM-30 (30 kDa cut-off) ultrafiltration disc (Millipore, UK) and stored at −80°C until use.

The gp140_SF162 (25 µg per animal) was formulated with different adjuvant(s) immediately prior to immunization. A detailed example on the dosages for each adjuvants and antigen is shown in Table S1. In order to keep the final volume identical in all groups, 25 µg of gp140_SF162 (1 µg/µl) was first diluted 1∶1 (v/v) with PBS, or 25 µl glycoprotein to 25 µl of PBS, before mixing 1∶1 (v/v) with Freund's complete or incomplete adjuvant (Sigma-Aldrich, UK), or MF59 (Novartis Vaccines & Diagnostics, USA). The mixture, with a final volume of 100 µl (per animal), was then incubated for 10 min before injection. For Carbopol-971P (Lubrizol, USA) and 974P (Particle Sciences, USA), both polymers were first prepared to a 2% (w/v) suspension with enough PBS, before mixing 1∶1 (v/v), or 25 µl to 25 µl, with gp140_SF162. The mixture was then brought to the same final volume of 100 µl with PBS and incubated for 30 min before injection [44], [45]. For the combination groups, gp140 was first mixed 1∶1 (v/v), or 25 µl to 25 µl, with 2% Carbopol-971P/974P and incubated for 30 min. The mixture was then added 1∶1 (v/v) with 50 µl of MF59 and incubated for another 10 min before injection. For the no adjuvant group, 75 µl of PBS was added to gp140_SF162 (25 µl) to make the final volume 100 µl for the injection. Due to the high viscosity of the adjuvants, substantial vortexing was required during the formulation process to ensure the antigen-adjuvant mixture was homogenized before use.

The conformation of the gp140, in the presence or absence of adjuvants, was assessed by reducing, non-reducing and Blue-Native (BN) polyacrylamide gel electrophoresis using standard protocols. Briefly, gp140_SF162 was formulated with or without adjuvant(s) as described above. To separate the glycoprotein from the adjuvant(s), the mixture was centrifuged at high speed for 30 min and the aqueous phase, which contained the glycoprotein, was extracted and quantified with a spectrophotometer. The extracted glycoprotein was then denatured at 90°C and reduced with beta-mercaptoethanol (βME) for reducing SDS-PAGE. The glycoproteins were not heat denatured or treated with βME for the non-reducing condition. Coomassie Blue dye was added to 5 µg of gp140 and then separated in the absence of sodium dodecyl sulphate for BN-PAGE.

The effect of adjuvant on antigenicity was also analyzed by antibody binding. Briefly, gp140_SF162 was first formulated with adjuvant(s) and then extracted before being coated (0.5 µg/well) to microwell plates for 2 h at 16°C. The plates were washed 3 times with 1× PBS and blocked with 5% BSA for 30 min, followed by another 3 washes with PBS. Monoclonal antibodies labeled with Europium (Perkin Elmer, UK) were added (100 ng/well) and incubated for 1 h at 16°C. The plates were washed 3 times with wash buffer (25 mM Tris-HCl pH 7.8 containing 1% Triton X-100). Enhancement solution (200 µl/well; Perkin Elmer) was added and the plates were developed for 15 min and the Relative Luminescence Unit (RLU) emitted were measured with a 1420 Victor Multilabel Counter (Perkin Elmer, UK).

Twelve groups of 6 New Zealand white female rabbits each received 3 intramuscular immunizations of gp140_SF162 at week 0, 4 and 12 (Figure S1). Except for animals in group 1, in which 75 µg/injection of gp140 was used, all other animals received 25 µg of gp140 for each immunization. Animals in both groups 1 and 2 received FCA for the first immunization and FIA for the next two. Animals in groups 3–5 received MF59, Carbopol-971 or, Carbopol-974P for all three injections, respectively. For groups 6 and 7, rabbits received either Carbopol-971P or Carbopol-974P for the first injection, followed by MF59 for the next two. The regimen was reversed for groups 8 and 9, in which animals were injected twice with MF59, followed by a final injection of either Carbopol-971P or Carbopol-974P. For group 10, animals received a mixture of Carbopol-971P and MF59 adjuvants for all three immunizations. Similarly, animals in group 11 received a mixture of Carbopol-974P and MF59 for all three immunizations. Finally, animals in group 12 received only gp140_SF162 for all three injections, without any adjuvant. Serum samples were collected 2 weeks before and after each immunization at week −2, 2, 6 and 14, and a terminal sample was collected at week 16. One of the animals in group 1 died before the end of the study of cause unrelated to adjuvant toxicity.

The endpoint titers of Env-specific antibodies in the sera were measured by standard ELISA. Briefly, 96 maxisorp microwell plates (Nunc, UK) were coated overnight with purified gp140_SF162 (50 ng/well) in 50 mM sodium carbonate-bicarbonate buffer (pH 9.6) at 4°C, and then blocked with PBS containing 0.05% Tween-20 and 5% non-fat milk for 2 h at 16°C. Serum samples were serially diluted (2-fold) and 50 µl were added to triplicate wells and incubated for 1 h at 16°C, and then washed 3 times with wash buffer (PBS containing 0.05% Tween-20). A rabbit IgG specific secondary antibody conjugated to HRP (Sigma-Aldrich, UK), diluted 1∶10,000, was added and incubated for 1 h at 16°C. After three washes with wash buffer, the plate was developed with a 1-Step 3,3,5,5-tetramethylbenzidine (Pierce Thermo Scientific, UK) for 20 min in the dark before being stopped with 2 M H₂SO₄. Absorbance values were detected at 450 nm using a Bio-Rad iMark Microplate Reader (Bio-Rad, UK).

The avidity indices of vaccinated sera were measured by its resistance to 8 M urea in binding to gp140_SF162 antigen [49]. Briefly, microwell plates were prepared and blocked as described above. Serum samples were diluted to give an OD_{450 nm} readout between 1.0 and 1.5 in endpoint ELISA and were added to two sets of triplicate wells to incubate for 1 h at 16°C. The wells were then washed three times with either PBS-Tween or 8 M urea in PBS-Tween, before incubating with an anti-rabbit secondary antibody. The plates were washed and developed with TMB as described above. The avidity index was calculated as the percentage of average urea treated OD_{450 nm}/average PBS-Tween OD_{450 nm}. Antisera with index values >50% were designated high avidity, 30–50% were designated intermediate avidity and <30% were designated low avidity.

Galanthus nivalis lectin (GNA) was coated at 100 ng/well onto microwell plates in PBS buffer at 4°C overnight. The microwell plates were blocked with 2% BSA for 2 h, before gp140_SF162 (100 ng/well) was added in triplicate and incubated for 2 h at 16°C. The microplate was washed 3 times with wash buffer (25 mM Tris-HCl pH 7.8 containing 1% Triton X-100). To determine the approximate epitopes that the polyclonal sera were raised to, antisera from vaccinated animals were diluted according to their endpoint titer to approximately 1×10⁵ IgG to compete with Eu³⁺-labeled mAbs (100 ng/well) for binding to antigens. Unbound antisera and mAbs were removed by 6 washes. Enhancement solution (Perkin Elmer, UK) was added and incubated for 20 min and the RLU_{620 nm} was measured in a Envision luminometer (Perkin Elmer, UK). Avidity of the antisera to the epitopes was determined by reduction in the Eu³⁺ luminescence intensity of the specific mAb that the serum was competing with. The 2F5 and 4E10 mAbs were purchased from Polymun (Vienna, Austria) and other mAbs were kindly provided upon request (see Acknowledgements).

The TZM-bl assay was carried out using a standard protocol with molecularly cloned pseudoviruses [50]. Briefly, pseudovirus was incubated with serially diluted rabbit antisera for 1 h at 37°C before being placed into wells of 96-well plates seeded with TZM-bl cells. After 48 h incubation at 37°C, the cells were lysed and luciferase signal in the lysate was developed with Britelite Plus substrate (1∶1 v/v; Perkin Elmer) and read in a luminometer.

Comparison of log transformed binding and avidity data between groups was tested using a one way ANOVA, with a correction for multiple comparisons between groups conducted using Tukey's Honestly Significant Differences. Comparison of log-transformed neutralization titers between groups was performed using a linear mixed effects model, fitted using a Bayesian Markov Chain Monte Carlo approach. Markov chains were run for 110,000 iterations, with a burnin of 10,000 steps, and the resulting chain was thinned to provide 1,000 samples of parameter values.

To examine if the adjuvants modified antigen conformation, gp140_SF162 was formulated in a single adjuvant or a combination of adjuvants described in the immunization protocol in Materials and Methods. The samples were extracted and analyzed for conformational changes and antigenicity by gel electrophoresis and antibody binding ELISA.

When gp140_SF162 was formulated in FCA or FIA, there was unavoidable and substantial spillage of the samples from the wells during loading, particularly for the former. To extract only the aqueous gp140 from the mineral oil, samples were centrifuged at high speed and the oil phase was removed. However, a smearing separation was still observed in both reduced and non-reduced, as well as BN-PAGE gels for these two samples, making interpretation difficult (data not shown). We observed no biophysical differences with any of the other adjuvants, including MF59, in either reducing or non-reducing gels, compared to the unadjuvanted gp140 (Figures 1a & 1b). Similarly, no differences were observed when samples were assayed by BN-PAGE (Figure 1c).

To further analyze whether gp140_SF162 underwent conformation changes that could affect its antigenicity in the presence of adjuvants, we assayed for binding of a panel of mAbs to the antigen by ELISA. The aqueous glycoprotein was separated from the oil phase from samples containing FCA, FIA or MF59. The amount of gp140_SF162 extracted was quantified and normalized before coating for ELISA. There was significant loss in mAb binding in the presence of FCA (data not shown). We suspected this was due to the high viscosity and hydrophobicity of the residual mineral oil, which either compromised gp140 binding to the plate, or heavily masked most of the epitopes and prevent mAb from accessing them. On the contrary, in the presence of FIA, which is substantially less viscous, we observed no change in antigenicity of the extracted gp140 as compared to the unadjuvanted sample (Figure 2). MF59, the other oil-in-water emulsion that also has lower viscosity than FCA, showed no significant changes in gp140 antigenicity as well.

In contrast to the oil-in-water emulsions, gp140_SF162 could not be separated from the Carbopol polymers since both the antigen and the adjuvants were in aqueous phase. Formulation with Carbopol-971P alone, or in a mixture with MF59, did not appear to interfere binding of b12 (CD4bs) or HGN194 (V3 loop), while modest reduction was observed with other mAbs targeting the same or other antigenic sites (Figure 2). Similar result was also demonstrated in a recent study, where mAbs b12 and CD4IgG2 showed indistinguishable level of binding to both native and Carbopol-971P formulated gp140, while binding of 17b (CD4i) and F425-B4e8 (V3 loop) was significantly compromised [45]. Since Carbopol is polyanionic, its interaction with the positively charged side chains of gp140 might induce partial masking and prohibit mAb binding to its epitope. Nevertheless, further investigation will be required to study the molecular interaction between the carbomers and the antigen for an accurate interpretation.

Twelve groups of 6 rabbits each were immunized 3 times with gp140_SF162 with different adjuvant regimens (Figure S1). Minor and transient swelling with reddening at the injection site were observed in rabbits that received FCA/FIA, which was resolved within days. No local or systemic reactogenicity was observed in rabbits that received other adjuvant formulations.

Serum antigen-specific IgG was measured by ELISA at the terminal bleed, 4 weeks after the 3^rd immunization (Figure 3). Rabbits immunized without adjuvant (group 12) were used as our low-end benchmark. These animals developed a relatively high mean titer (6.4×10⁵) after three immunizations over the course of 16 weeks, despite no adjuvant being used. FCA/FIA adjuvants were used as the high-end benchmark, and as expected, groups 1 and 2 had the highest binding IgG titers among all groups (mean = 6.2–6.5×10⁶, p<0.05). Although animals in group 1 received 3-fold more gp140_SF162 immunogen than those in group 2, there was no significant difference in the mean binding titers, suggesting that increasing the quantity of antigen does not offer additional benefits in the presence of a potent adjuvant. Among all other adjuvanted groups, only those that were immunized with the combination of Carbopol-971P and MF59 (group 10) elicited titers comparable to the FCA/FIA groups (mean = 5.8×10⁶, p = 1.00). When used individually, the titers elicited by MF59 or Carbopol-971P were 7- to 8-fold lower than the combination group or the FCA/FIA groups (p<0.001). Sequential immunization of Carbopol-971P followed by MF59 (group 6) or in reverse order (group 8) also resulted in titers similar to those of the single adjuvant protocols, indicating that concomitant use of both adjuvants was needed to enhance antibody titers. In addition, the high antibody titer induced by the combination group was very specific to the Carbopol-971P formulation, as the mean titer of those immunized with combination of Carbopol-974P and MF59 (group 11) was 6-fold lower in comparison (mean = 9.1×10⁵, p<0.005). Individually, Carbopol-971P appeared to be slightly more active and elicited higher binding titers of antibodies than Carbopol-974P (means = 7.24×10⁵ and 6.45×10⁵, respectively), but overall differences were not significant.

To estimate antibody affinity maturation, avidity of the antisera was measured by the 8 M urea displacement method as described previously [49]. Although all antisera tested have high relative avidity indices (RAI), defined as >50% residual binding, the mean RAI was significantly lower in those immunized without adjuvant (58.5%, p<0.05; Figure 4). The oil-in-water emulsion adjuvants, such as FCA/FIA and MF59, elicited antibodies with higher RAI, ranging from 71–82%, whether they were used alone or in combination with Carbopols. Comparing the two carbomers, antisera from rabbits immunized with Carbopol-971P had modestly higher avidity than those immunized with Carbopol-974P (mean = 74.0% and 66.7%, respectively; Figure 4), although the difference is not significant between the two groups. Again, as observed with IgG binding titer, increasing the immunogen dosage did not have any effect on the avidity, as the polyclonal antisera from groups 1 and 2 were found to have no significant differences in their RAI (mean = 76.5% and 76.9%, respectively; Figure 4).

To investigate if antibodies elicited by different adjuvants were qualitatively distinct in the epitopes that they targeted, the polyclonal antisera were tested for their ability to compete with a panel of well-characterized mAbs in binding to gp140_SF162. Two mAbs targeting each distinct antigenic site were used; these included b12 and VRC01 (CD4bs), 3019 and HGN194 (V3 loop), 17b and E51 (CD4i) and 2F5 and 4E10 (MPER). Antisera were normalized for endpoint IgG titers, such that similar amounts of polyclonal antibodies were used to compete each mAb. The comprehensive results are shown in Figure S2.

Across all groups, it appeared that most of the antibodies elicited were directed against the CD4bs and the CD4i regions of Env, as the antisera could out-compete up to 50% of the respective mAbs (Figure 5). Antisera from animals immunized with the combination of Carbopol-971P and MF59 (group 10) displayed the strongest activities to both CD4bs and CD4i, as they could potently out-compete an average of 80% of the four mAbs tested (shown as 20% residual mAb binding; Figure 6). To our surprise, binding activity to the V3 loop was only weakly detectable in most of the antisera, except those immunized with FCA/FIA (groups 1 and 2) or the combination of Carbopol-971P and MF59 (group 10), where the polyclonal antibodies modestly to potently out-competed both 3019 and HGN194 mAbs (Figure 6). Similarly, serum antibody binding to the MPER was either completely absent or only weakly detected. While animals immunized with FCA/FIA or combination of Carbopol-971P and MF59 (groups 1&2 and 10, respectively) competed with 2F5 and 4E10 binding to a modest degree, antisera from other groups, at best, only weakly out-competed 4E10 and showed limited, if any, competition to 2F5 (Figure 6).

To examine whether different adjuvant formulations had any effect on neutralization breadth and depth, rabbit antisera were assessed against a panel of tier 1 pseudoviruses using the standard TZM-bl assay (comprehensive results shown in Figure S3). Among animals immunized with either a single adjuvant or different adjuvants in sequential order, the neutralization titers against the autologous SF162.LS virus were the highest in animals immunized with FCA/FIA (groups 1 and 2) with mean group titers of 4,663 to 4,178, respectively (Figure 7). Although the neutralization titer was slightly higher in group 1 than group 2, the former of which received 3 times more antigen, the difference in titer was not significant. When formulated together, this mixture of Carbopol-971P and MF59 (group 10) was found to be highly immune stimulating, and elicited neutralizing titers higher than the FCA/FIA groups against SF162.LS (mean = 4,776; Figure 7). By contrast, the titers were at least 2-fold lower when MF59 (group 3) or Carbopol-971P (group 4) was used alone (mean = 885 and 2,175, respectively, p<0.05; Figure 7). The same effect was observed when the antisera were assayed with other pseudoviruses, in which the combination of Carbopol-971P and MF59 elicited the highest neutralization titers across all groups and against all pseudoviruses but BaL.26 (Figure 7). On the other hand, sequential immunization of Carbopol-971P followed by MF59 (group 6), or vice versa (group 8), did not have any additive or synergistic effect and the overall neutralization titers were either similar or lower than using either Carbopol-971P or MF59 alone (p>0.5; Figure 7). The increase in titers was specific for the 971P formulation of Carbopol, as combining Carbopol-974P and MF59 (group 11) resulted in significantly lower titers against all viruses tested (p<0.05). This observation might in part be due to the fact that Carbopol-974P (group 5) was less immunogenic than Carbopol-971P and elicited lower nAb titers on its own (Figure 7). Of note, even in the absence of adjuvant (group 17) the antigen was capable of eliciting low titer of nAbs against the readily neutralized tier 1A pseudoviruses, but the response was almost completely absent against the more neutralization resistant tier 1B viruses (Figure 7 and Figure S3).