Characterization and Immunogenicity of a Novel Mosaic M HIV-1 gp140 Trimer

Production and expression of mosaic HIV-1 Env proteins.

The mosaic M Env gene sequences have been described previously (12, 15, 16). The mosaic gp140s were engineered to contain point mutations to eliminate cleavage and fusion activity (11, 12). To maximize expression in human cell lines, human codon-optimized mosaic M gp140s were synthesized by GeneArt (Life Technologies) with a C-terminal T4 bacteriophage fibritin foldon trimerization domain. A polyhistidine motif was included to facilitate protein purification in one version of the protein. Genes were cloned into the SalI-BamHI restriction sites of a pCMV eukaryotic expression vector, inserts were verified by diagnostic restriction digests, DNA was sequenced, and expression testing was performed using 10 μg of DNA with Lipofectamine (Life Technologies) in 293T cells. Stable cell lines for natural C clade isolate C97ZA.012 (NatC) (17) and MosM gp140 Env trimers were generated by Codex Biosolutions.

For protein production, the stable cell lines were grown in Dulbecco's modified Eagle medium (DMEM) (supplemented with 10% fetal bovine serum [FBS], penicillin-streptomycin and puromycin) to confluence and then were changed to Freestyle 293 expression medium (Invitrogen) supplemented with the same antibiotics. Cell supernatants were harvested at 96 to 108 h after medium change, and His-tagged gp140 proteins were purified by Ni-nitrilotriacetic acid (NTA) (Qiagen) and size exclusion chromatography as previously described (17, 18). The synthetic gene for full-length MosM gp120 was generated from the MosM gp140 construct. The synthetic gene for full-length MosM gp160 used in the TZM.bl assay was synthesized by GeneArt (Life Technologies) and cloned into a pcDNA3.1/V5-His-TOPO vector (Invitrogen).

The MosM gp140 without a polyhistidine tag was inserted into the pcDNA2004(Neo) vector (GeneArt), which is optimized for expression in PER.C6 cells. Stable MosM gp140-expressing PER.C6 suspension cells were used to produce MosM gp140 trimer without a polyhistidine tag either in batch or fed-batch mode. The MosM gp140 was Galanthus nivalis lectin (Vector Laboratories)-purified from the supernatant followed by gel filtration chromatography.

Biophysical characterization of MosM gp140.

The biophysical properties of MosM gp140 without a polyhistidine tag produced in PER.C6 cells were determined by size exclusion chromatography-multiangle light scattering (SEC-MALS), SEC–quasi-elastic light scattering (SEC-QELS), dynamic light scattering (DLS), far-UV circular dichroism (CD), SDS-PAGE, and differential scanning calorimetry (DSC) as detailed below.

SEC-MALS/SEC-QELS.

Size exclusion chromatography (SEC) was performed using an analytical column (TSKgel G3000SWxl; Tosoh Bioscience) equilibrated with 150 mM sodium phosphate and 50 mM sodium chloride at pH 7.0. Typically, 125 μg of protein was injected and separated at a flow rate of 1 ml/min. For molar mass determination, in-line UV (Agilent 1260 Infinity MWD; Agilent Technologies), refractive index (Optilab T-rEX; Wyatt Technology), and eight-angle static light-scattering (Dawn HELEOS; Wyatt Technology) detectors were used. Replacement of one of the static light-scattering detectors by a dynamic light-scattering detector (DynaPro NanoStar; Wyatt Technology) enabled determination of hydrodynamic radii simultaneously. The stability of the MosM gp140 protein was studied by SEC-multiangle light scattering (SEC-MALS) upon incubation for 30 min at 50, 60, 70, 80, or 90°C. Astra Software, including the protein conjugate analysis function, was used for data analysis.

For determination of hydrodynamic radii in batch mode, a dynamic light-scattering detector was used (DynaPro Plate Reader; Wyatt Technology). Light-scattering intensity was detected at 25°C at a concentration of 1 mg/ml during 20 acquisitions of 5 s each. Hydrodynamic radii were determined using the regularization function in the Dynamics software.

Far-UV CD.

Secondary structure analysis was performed using a circular dichroism (CD) spectrometer (model 420; AVIV Biomedical, Inc.) equipped with a recirculating chiller (Thermocube 200/300/400; Solid State Cooling Systems). Far-UV CD spectra were recorded with rectangular 1-mm-path-length quartz cuvettes at concentrations of 3 to 4 μM, using a 0.5-nm (determination of intrinsic properties) or 1-nm (stability study) step width and an averaging time of 4 s. Far-UV CD was recorded after incubation at 40, 50, 60, 70, 80, and 90°C for 10 min. After correction of the signal for baseline drift and contribution of buffer components, the molar residual ellipticity (MRE; in degrees cm² dmol⁻¹) was calculated based on the following equation: (0.1 × θ_λ)/[d × M × (number of amino acids], where θ_λ is the observed ellipticity in millidegrees, d is the path length of the cuvette in cm, and M is the molar concentration.

Sodium dodecyl sulfate-polyacrylamide gel electrophoresis SDS-PAGE.

Samples were prepared under nonreducing conditions by addition of LDS sample buffer (NP0007; Life Technologies) and incubation for 10 min at 70°C or 30 min at 98°C. For reducing conditions, LDS sample buffer and a dithiothreitol (DTT)-based reducing reagent (NP0004; Life Technologies) were added, and samples were incubated for 10 min at 98°C. Samples (10 μg/lane) and HiMark Prestained Protein Standard (LC5699, Life Technologies) were applied on the gel (EA03752BOX; Life Technologies). The gel was stained with Coomassie blue.

Differential scanning calorimetry (DSC).

Measurements were performed using a differential scanning calorimeter (MicroCal VP-Capillary DSC System; GE Healthcare). Samples were buffer exchanged to phosphate-buffered saline and diluted to 3 to 4 μM. Scans were recorded from 25°C to 110°C at a scan rate of 60°C/h. Data were processed using MicroCal VP-Capillary DSC Control Software, version 2.0.

Recombinant adenovirus serotype 26 vector.

Replication-incompetent, E1/E3-deleted recombinant adenovirus serotype 26 (rAd26) vector expressing MosM gp140 was prepared as previously described (12).

Western blot immunodetection.

Supernatants (20 μl) obtained 48 h after transient transfection of 293T cells with pCMV-MosM.3.1 (MosM), -MosM.3.2, or -MosM.3.3 gp140 expression constructs were separately mixed with reducing sample buffer (Pierce), heated for 5 min at 100°C, and run on a precast 4 to 15% SDS-PAGE gel (Bio-Rad). Protein was transferred to a polyvinylidene difluoride (PVDF) membrane using an iBlot dry blotting system (Invitrogen), and membrane blocking was performed overnight at 4°C in Dulbecco's phosphate-buffered saline–0.2% (vol/vol) Tween 20 (Sigma)–5% (wt/vol) nonfat milk powder (PBS-T). Following overnight blocking, the PVDF membrane was incubated for 1 h with PBS-T containing a 1:2,000 dilution of monoclonal antibody penta-His-horseradish peroxidase (HRP) (Qiagen), washed five times with PBS-T, and developed using an Amersham ECL Plus Western blotting detection system (GE Healthcare). For Western blot immunodetection using 2F5 and 4E10 monoclonal antibodies, clade A (92UG037.8) gp140 (17, 18) and MosM gp140 proteins were processed as described above and detected with an anti-human IgG (Jackson ImmunoResearch).

SPR.

Surface plasmon resonance (SPR) binding assays were conducted on a Biacore 3000 (GE Healthcare) at 25°C utilizing HBS-EP (10 mM HEPES [pH 7.4], 150 mM NaCl, 3 mM EDTA, 0.005% P20 surfactant) running buffer (GE Healthcare). Immobilization of soluble two-domain CD4 (19) (1,500 resonance units [RU]) or protein A (ThermoScientific) to CM5 chips was performed according to the manufacturer's (GE Healthcare) recommendations. Immobilized IgGs were captured at 300 to 750 RU. Binding experiments were conducted at a flow rate of 50 μl/min with a 2-min association phase and 5-min dissociation phase. Regeneration was conducted with a single injection of 35 mM NaOH and 1.3 M NaCl at 100 μl/min, followed by a 3-min equilibration phase in HBS-EP buffer. Injections over blank surfaces were subtracted from the binding data for analyses. Binding kinetics were determined using BIAevaluation software (GE Healthcare) and the Langmuir 1:1 binding model with the exception of PG16, which was determined using a bivalent analyte model. All samples were run in duplicate and yielded similar kinetic results. Soluble two-domain CD4 and PG9 and PG16 Fabs were produced as described previously (17, 19). 17b hybridoma was provided by James Robinson (Tulane University, New Orleans, LA) and purified as previously described (17). VRC01 was obtained through the NIH AIDS Reagent Program. VRC01 was provided by John Mascola (VRC, NIH, Bethesda, MD) (20). 3BNC117 was provided by Michel Nussenzweig (Rockefeller University, New York, NY). PGT121, PGT126, and PGT145 were provided by Dennis Burton (The Scripps Research Institute, La Jolla, CA). 2F5, 4E10, PG9, and PG16 were obtained commercially (Polymun Scientific). GCN gp41-inter (21) was used as a positive control in 4E10 and 2F5 SPR analyses.

Animals and immunizations.

Outbred female Hartley guinea pigs (Elm Hill) (n = 10/group) were housed at the Animal Research Facility of Beth Israel Deaconess Medical Center under approved Institutional Animal Care and Use Committee (IACUC) protocols. Guinea pigs were immunized intramuscularly (i.m.) with either MosM or NatC gp140 Env trimer or both (100 μg/animal) at weeks 0, 4, and 8 in 500-μl injection volumes divided between the right and left quadriceps. The adjuvant was either 15% (vol/vol) oil-in-water Emulsigen (MVP Laboratories)-PBS and 50 μg of immunostimulatory dinucleotide type B oCpG DNA (5′-TCGTCGTTGTCGTTTTGTCGTT-3′) (Midland Reagent Company) (17) or the immune-stimulating complex (ISCOM)-based Matrix M (Novavax). The groups of animals receiving the mixture of NatC and MosM (NatC+MosM) gp140s were given a dose of 50 μg of each trimer, for a total of 100 μg per animal. In heterologous prime-boost regimens, a rAd26 vector expressing MosM gp140 (1 × 10¹⁰ virus particles [vp]/animal) was administered i.m. at week 0 in 500 μl of PBS-sucrose, and animals were boosted at weeks 8, 12, and 16 with either MosM, NatC, or bivalent NatC+MosM gp140 Env protein trimers (100 μg/animal) in Adju-Phos alum (Brenntag, Denmark) adjuvant. Serum samples were obtained from the vena cava of anesthetized animals at 4 weeks after each immunization.

ELISAs.

Serum binding antibody titers against MosM or NatC gp140 Env trimer were determined by endpoint enzyme-linked immunosorbent assays (ELISAs) as previously described (17, 18). Endpoint titers were defined as the highest reciprocal serum dilution that yielded an absorbance of >2-fold over background values. For the detection of membrane-proximal external region (MPER) epitopes in the MosM trimer, ELISA plates were coated with either 2F5 or 4E10 IgG, antigen was added, and epitopes were detected utilizing anti-His tag-HRP monoclonal antibody (MAb; Abcam). 2F5 and 4E10 were obtained commercially (Polymun Scientific), and GCN gp41-inter (21) was used as a positive control in 4E10 and 2F5 ELISAs.

TZM.bl neutralization assay.

Neutralizing antibody responses against tier 1 HIV-1 Env pseudoviruses were measured using luciferase-based virus neutralization assays with TZM.bl cells as previously described (17, 18, 22). These assays measure the reduction in luciferase reporter gene expression levels in TZM.bl cells following a single round of virus infection. The 50% inhibitory concentration (IC₅₀) was calculated as the serum dilution that resulted in a 50% reduction in relative luminescence units compared with the virus control wells after the subtraction of cell control relative luminescence units. The panel of 11 tier 1 viruses analyzed included easy-to-neutralize tier 1A viruses (SF162.LS and MW965.26) and an extended panel of tier 1B viruses (DJ263.8, Bal.26, TV1.21, MS208, Q23.17, SS1196.1, 6535.3, ZM109.F, and ZM197M) (22) (23). Murine leukemia virus (MuLV) negative controls were included in all assays.

A3R5 neutralization assay.

Additional nAb responses were evaluated using an A3R5 assay as previously described (17). Briefly, serial dilutions of serum samples were performed in 10% RPMI growth medium (100 μl per well) in 96-well flat-bottomed plates. Infectious molecular clone (IMC) HIV-1 expressing Renilla luciferase (24) was added to each well in a volume of 50 μl, and plates were incubated for 1 h at 37°C. A3R5 cells were then added (9 × 10⁴ cells per well in a volume of 100 μl) in 10% RPMI growth medium containing DEAE-dextran (11 μg/ml). Assay controls included replicate wells of A3R5 cells alone (cell control) and A3R5 cells with virus (virus control). After samples were incubated for 4 days at 37°C, 90 μl of medium was removed from each assay well, and 75 μl of cell suspension was transferred to a 96-well white, solid plate. Diluted ViviRen Renilla luciferase substrate (Promega) was added to each well (30 μl), and after 4 min the plates were read on a Victor 3 luminometer. The A3R5 cell line was provided by R. McLinden and J. Kim (U.S. Military HIV Research Program, Rockville, MD). Tier 2 clade B IMC Renilla luciferase-expressing viruses included SC22.3C2.LucR, SUMA.LucR, and REJO.LucR. Tier 2 clade C IMC Renilla luciferase-expressing viruses included Du422.1.LucR.T2A.ecto, Ce2010_F5.LucR.T2A.ecto, and Ce1086_B2.LucR.T2A.ecto and were provided by C. Ochsenbauer (University of Alabama at Birmingham, Birmingham, AL). Viral stocks were prepared in 293T/17 cells as previously described (24).

Statistical analyses.

For both the TZM.bl and A3R5 data described above, the neutralization scores used for plotting and for statistical considerations are provided as the IC₅₀ titer of the postvaccination serum with the prevaccination background subtracted as previously described (25). All reactivity that was below the level of assay detection (<20) was assigned the value 20 for plotting and statistical purposes. Titers within a 3-fold range are considered “concordant”; thus, we considered titer differences of the geometric mean group response that were >0.5 log as indicative of robust differences. Statistical analyses were performed using the statistical package R (http://www.r-project.org/). Nonparametric Wilcoxon tests were used to compare distributions of values between vaccine groups, Fisher's exact test was used to compare the relative proportions of responses that were above the level of detection in different vaccine groups, and the R package lme4 (26) was used to model the relative impact of different variables on neutralization sensitivity levels.

Expression and stability of the mosaic M gp140 trimer.

We have previously described the mosaic M Env gene sequences (12, 16). MosM.3.1 Env (whose nomenclature has been simplified here to MosM Env) is more closely related to clade B natural strains, whereas MosM.3.2 Env is most closely related to clade C natural strains, and MosM.3.3 is not particularly associated with any clade but captured complementary forms of epitopes that are common throughout the M group and not already represented in MosM and MosM3.2 (Fig. 1). The mosaic forms, in combination, optimize potential coverage of linear epitopes in the population, but MosM combined with the natural isolate C97ZA.012 (17, 18, 27) (whose nomenclature has been simplified here to NatC) enhanced potential linear epitope coverage to levels approached using two mosaic trimers (Fig. 1).

Eukaryotic pCMV DNA vectors expressing HIV-1 MosM, MosM.3.2, and MosM.3.3 gp140 were transfected into 293T cells, and protein expression and stability were assessed after 48 h. Western blot analysis revealed a band of the expected size only for MosM gp140 (Fig. 2A), suggesting that MosM gp140 was more stable than MosM.3.2 gp140 and MosM.3.3 gp140. Size exclusion chromatography (SEC) using material generated from a stable 293T cell line expressing the MosM gp140 demonstrated a monodispersed peak, confirming the homogeneity of the trimer (Fig. 2B). To evaluate the preliminary stability of the MosM trimer, 100 μg of protein underwent a freeze-thaw cycle or was stored for 7 days at 4°C and reevaluated by SEC and showed no signs of aggregation, dissociation, or degradation (Fig. 2C). Given the reduced stability of MosM.3.2 and the good theoretical coverage of MosM combined with NatC, we focused on the combination of MosM and NatC for the rest of this study.

Biophysical characterization of the MosM gp140 trimer.

The mass and size of MosM gp140 produced in stable PER.C6 cell clones were assessed under native conditions by size exclusion chromatography (SEC) coupled with online multiangle light-scattering (MALS) and dynamic light-scattering (DLS) detectors. The molar mass of the glycoprotein was determined by SEC-MALS to be 454 ± 36 kDa, which indicated that the MosM gp140 was trimeric (Fig. 3A). By applying protein conjugate analysis, the estimated molecular mass of the protein component was 258 ± 20 kDa, which corresponds within experimental error to the theoretical molecular mass of an unglycosylated protein trimer. The amount of glycosylation was estimated to be 198 ± 18 kDa, which showed that the level of glycosylation in MosM trimer is significantly higher than previously reported for the BG505 SOSIP.664 trimer (28). The tight packing of the MosM gp140 was further confirmed by both batch and online DLS measurements, which showed that the estimated hydrodynamic radius of the protein was 8.2 to 8.8 nm, similar to the reported 8.1-nm hydrodynamic radius of the BG505 SOSIP.664 trimer (28), accounting for the increased glycosylation (Fig. 3B). The secondary structure of the MosM trimer was assessed by far-UV circular dichroism (CD) (Fig. 3C) and Fourier transform infrared (FTIR) spectroscopy (data not shown). As expected, both far-UV CD and FTIR spectroscopy confirmed the presence of alpha-helical and beta-sheet secondary structure elements. The mean residue molar ellipticity was assessed from the far-UV CD spectrum as approximately −11,000 degrees cm² dmol⁻¹.

The thermal stability of the MosM trimer was assessed by differential scanning calorimetry (DSC), far-UV CD, and SEC-MALS. No unfolding transitions were detected by DSC upon applying a temperature ramp up to 110°C (Fig. 4A). Furthermore, no significant changes in secondary structure were detected upon exposure of the protein to temperatures up to 90°C, as shown by the unaltered far-UV CD spectra (Fig. 4B). The robustness of the protein was also confirmed by FTIR spectroscopy (data not shown). The SEC-MALS experimental data indicated that the trimer molecule underwent a partial dimerization to a hexamer at temperatures above 70°C, but no further aggregation or dissociation (Fig. 4C). In particular, based on the quantitative recoveries from the SEC column, no higher-molecular-weight entities were formed at elevated temperatures. Furthermore, the hydrodynamic radius of the trimer did not change upon its exposure to elevated temperatures, thus confirming its remarkable stability.

The high stability of the MosM trimer was also shown by the difficulty unfolding the protein even in SDS-PAGE loading buffer at elevated temperatures. The MosM trimer could still be observed on a nonreduced gel after the protein was incubated with SDS-containing buffer at 70°C (Fig. 4D). The trimer band disappeared while some amounts of dimer were still detected when the sample was incubated for 30 min at 98°C in SDS loading buffer on a nonreduced gel, indicating that the trimer units were not held together via disulfide bridges (Fig. 4D).

Antigenicity of the mosaic M gp140 trimer.

Surface plasmon resonance (SPR) studies were next performed to determine whether the MosM trimer could bind CD4 and broadly neutralizing monoclonal antibodies. The MosM trimer bound CD4 at a high affinity, demonstrating that the CD4 binding site (CD4bs) is present and accessible (Fig. 5A). We next evaluated the ability of the MosM trimer to bind the monoclonal antibody (MAb) 17b in the presence and absence of bound CD4. 17b recognizes a CD4-induced (CD4i) epitope exposed by CD4 binding and the formation of the bridging sheet and coreceptor binding site in gp120 (29, 30). The MosM trimer showed 17b binding in the absence of CD4, but there was a clear increase following CD4 binding, as expected (Fig. 5B). We next observed that the broadly neutralizing CD4bs-specific MAbs VRC01 and 3BNC117 (20, 31) bound the MosM gp140 with high affinity (Fig. 5C and D). Moreover, the broadly neutralizing N332 glycan-dependent MAbs PGT121 and PGT126 also bound the trimer with high affinity (Fig. 6A and B). PG9 and PG16 have been reported to bind preferentially to Env trimers and target V2 glycans (32, 33). PG9 bound the MosM gp140 trimer with high affinity (5.8 × 10⁻⁸ M) and the corresponding MosM gp120 monomer with a 22-fold lower affinity (1.3 × 10⁻⁶ M) and substantially lower magnitude (Fig. 6C). The PG9 Fab bound to the MosM gp140 trimer with a similar affinity (5.5 × 10⁻⁸ M) as the complete PG9 IgG (data not shown), confirming the high-affinity PG9 binding. Similarly, PG16 bound the MosM gp140 trimer with a higher affinity than the corresponding gp120 monomer (Fig. 6D) although the off-rate was higher for PG16 than PG9.

We also assessed 2F5 and 4E10 binding to membrane-proximal external region (MPER) epitopes. 2F5 and 4E10 bind to linear epitopes (34, 35). The 2F5 epitope was present in the linear MosM gp140 sequence, as confirmed by sequence alignment (Fig. 7A) and Western blot (Fig. 7B) analyses. However, by SPR and ELISA, the intact MosM gp140 trimer was unable to bind 2F5, suggesting that the MPER epitope is not accessible in the trimer structure, presumably as a result of being buried (35) (Fig. 7C and D), similar to our previous findings with the NatC trimer (17). Alternatively, it is possible that the absence of a lipid membrane (36, 37) in the MosM gp140 trimer may have reduced 2F5 binding.

Taken together, the biochemical, biophysical, and antigenicity data suggest the robustness of the MosM gp140 trimer. This is based on the expected molecular mass and hydrodynamic radius, its remarkable stability, and its capacity to bind all CD4 binding site-, V3 glycan-, and V2 glycan-specific broadly neutralizing MAbs that we have tested, except for 2F5 and 4E10 as expected.

Functionality of mosaic M gp140 trimer.

We next evaluated whether the MosM Env protein was functional. Full-length MosM gp160 was used to generate pseudovirions to assess infectivity in TZM.bl cells (17, 18, 22, 40). We observed that, over a broad titration range, MosM gp160 Env readily infected TZM.bl target cells overexpressing CD4 and coreceptors CCR5/CXCR4 (Fig. 8). These data show that the synthetic MosM Env has the functional ability to infect TZM.bl target cells.

Immunogenicity of the mosaic M gp140 trimer.

Guinea pigs (n = 10/group) were immunized three times at monthly intervals with 100 μg of MosM gp140 trimer, 100 μg of NatC gp140 trimer (17, 18), or a mixture of 50 μg of both trimers. Half of the animals were immunized with ISCOM-based Matrix M adjuvant, and half were immunized with CpG/Emulsigen adjuvants (17, 41). High-titer, binding antibodies by ELISA were elicited by all the vaccination regimens with comparable kinetics (Fig. 9). These responses were detectable after a single immunization and increased after the second and third immunizations. Peak binding antibody titers ranged from 5 to 7 logs. Binding antibody titers elicited by the NatC and MosM trimers were higher against their respective homologous antigens (P < 0.05), whereas the bivalent MosM and NatC mixture induced comparable responses to both antigens (Fig. 9). ELISA titers were similar using trimers with or without the His tag as coating antigens (data not shown).

We next assessed serum nAb responses elicited in these animals after the third vaccination using both TZM.bl and A3R5 nAb assays (17, 18, 22, 40). For the TZM.bl assays, we included a multiclade panel of tier 1 pseudoviruses comprising easy-to-neutralize tier 1A viruses (SF162.LS and MW965.26) and an extended panel of intermediate tier 1B viruses (DJ263.8, Bal.26, TV1.21, MS208, Q23.17, SS1196.1, 6535.3, ZM109.F, and ZM197M) (22, 23). Background prevaccination titers were negative or low in all animals (<20) (data not shown). We observed a trend for slightly increased responses with the Matrix M adjuvant compared with CpG/Emulsigen adjuvant (P = 0.05 and 0.06 for the TZM.bl and A3R5 assays, respectively) (Fig. 10).

The nAb responses elicited by the bivalent NatC+MosM trimer mixture were comparable to the better of the two immunogens in the cocktail for each virus tested (Fig. 11A). Thus, there was no apparent loss of potency due to dilution or competition in the NatC+MosM mixture, and the overall response to the NatC+MosM cocktail was thus superior to either immunogen alone. A statistical breakdown of the data, comparing response level distributions by a nonparametric Wilcoxon rank sum test, is provided in Table 1 and summarized in Fig. 11B. Specifically, using the TZM.bl assay, for 2 (ZM197M and 6535) of the 11 viruses, the responses were indistinguishable between the three vaccine groups although responses to both of these viruses were borderline. For 6 of the 11 viruses, the responses to the individual MosM and NatC monovalent immunogens were significantly different, and the NatC+MosM combined vaccine was comparable to the higher of the two monovalent vaccines in each case. A similar trend was observed for SF162.LS. The NatC+MosM combination yielded a more potent response than both monovalent vaccines for one virus (MS208), whereas the NatC gp140 slightly outperformed the bivalent NatC+MosM vaccine for one virus (ZM109F). Overall, these data suggest that the MosM and NatC gp140 Env trimers were immunologically complementary and could be effectively combined as a mixture that was superior to either trimer alone in terms of nAb coverage. Less clear differences were observed for tier 2 viruses in A3R5 neutralization assays (Fig. 12).

To evaluate the complexity of the interactions between variables that might impact neutralization sensitivity, each assay was analyzed separately using an inverse Gaussian generalized model. Animals and viruses were included in the model as random effects. For fixed effects, we started testing the larger possible model, which included vaccine, clade, adjuvant, and all their possible interactions, and then scaled down the model in a stepwise manner, eliminating one variable at a time until we minimized the Akaike information coefficient (AIC). We initially fit the whole data set with 11 viruses and both adjuvants (Table 2). The best model had a significant interaction between vaccine and clade (P < 2.2e−16). This statistical analysis confirms that the MosM trimer, which is more clade B like (Fig. 1A), elicits stronger responses to the clade B viruses, whereas the NatC trimer elicits stronger responses to clade C as well as clade A viruses. A significant effect from the adjuvant (P = 0.0074) was also observed, indicating the higher potency of the Matrix M adjuvant.

Finally, we assessed nAb responses following priming with a replication-incompetent rAd26 vector expressing the MosM gp140 protein, followed by three protein boosts with the NatC, MosM, or bivalent NatC+MosM gp140 trimers (Fig. 13; Table 3). This experiment suggests that the mixture of the MosM and NatC trimers was superior to either trimer alone also in the context of a prime-boost vaccine regimen.