Xinqidi,Your biological research partner
Xinqidi Biotech Co.,Ltd,Wuhan,China 2008-2021
R&D 13th year

Lipid nanoparticle encapsulated nucleoside-modified mRNA vaccines elicit polyfunctional HIV-1 antibo

Issuing time:2021-04-12 10:47


The development of an effective AIDS vaccine remains a challenge. Nucleoside-modified mRNAs formulated in lipid nanoparticles (mRNA-LNP) have proved to be a potent mode of immunization against infectious diseases in preclinical studies, and are being tested for SARS-CoV-2 in humans. A critical question is how mRNA-LNP vaccine immunogenicity compares to that of traditional adjuvanted protein vaccines in primates. Here, we show that mRNA-LNP immunization compared to protein immunization elicits either the same or superior magnitude and breadth of HIV-1 Env-specific polyfunctional antibodies. Immunization with mRNA-LNP encoding Zika premembrane and envelope or HIV-1 Env gp160 induces durable neutralizing antibodies for at least 41 weeks. Doses of mRNA-LNP as low as 5 μg are immunogenic in macaques. Thus, mRNA-LNP can be used to rapidly generate single or multi-component vaccines, such as sequential vaccines needed to protect against HIV-1 infection. Such vaccines would be as or more immunogenic than adjuvanted recombinant protein vaccines in primates.


Messenger ribonucleic acid (mRNA)-based vaccines have been shown to elicit protective immunity against Zika virus (ZIKV) infection after a single immunization of rhesus macaques1, and have been designed for many other pathogens including Ebola, influenza, Hepatitis C, Cytomegalovirus, and respiratory syncytial virus2,3,4,5,6,7.

The recent development of mRNA vaccines has overcome initial roadblocks to their use. mRNA vaccines were initially hindered by innate immune sensing of mRNA7,8,9. However, advances in modifying mRNA nucleosides have improved mRNA translation, while eliminating recognition by innate pattern recognition receptors7,8,10. Specifically, the incorporation of pseudouridine and 1-methyl-pseudouridine prevents recognition of mRNA by Toll-like receptor 7 and 8 and other nucleic acid sensing pattern recognition receptors7. Modification of mRNA in combination with HPLC and FPLC purification methods and codon optimization methods have increased the efficiency of mRNA production and the efficacy of mRNA vaccination8,11,12. For mRNA to be effective at transducing cells in vivo, it must be protected from RNAses. One method for protecting the mRNA from degradation has been its encapsulation in lipid nanoparticles (LNP)7,8,13. Importantly, small interfering RNAs (siRNAs) in LNP have been approved by the FDA for the treatment of a genetic form of amyloidosis14. Thus, advances in encapsulation and preventing mRNA immune sensing have made mRNA vaccines a feasible vaccine platform.

mRNA vaccines are also attractive as a vaccine platform, because they can accept the genes of various pathogen antigens, and they can be rapidly manufactured at scale. These two aspects made mRNA vaccines a prime candidate for responding to the SARS-CoV-2 outbreak15,16. High levels of SARS-CoV-2 neutralizing antibodies, defined as higher than geometric mean values of convalescent serum from COVID-19 patients, were observed for both nucleoside-modified mRNA-LNP and Matrix-M1 adjuvanted subunit protein in human trials16,17,18. Given the ability of mRNA vaccines to elicit neutralizing antibodies8,16,17, this platform could improve the elicititation of neutralizing antibodies against HIV-1. However, little data exists on the comparative immunogenicity of mRNA-LNP and proteins in non-human primates. Thus, a key question for HIV-1 vaccine development is whether the immunogenicity of mRNA vaccines encoding HIV-1 immunogens, a poorly immunogenic protein that requires extensive post-translational modifications, is comparable to that of proteins that can be purified after in vitro production.

HIV-1 vaccines aiming to elicit protective antibody responses will likely need to elicit polyfunctional non-neutralizing effector antibodies (nnAbs) and/or broadly neutralizing antibodies (bnAbs)19,20. NnAbs are easy to induce and bind to the surface of virus-infected CD4+ T cells where they can mediate antibody-dependent cellular cytotoxicity (ADCC). However, their efficacy in protecting against HIV transmission or control of the disease in the setting of high transmission rates was called into question with the HVTN 702 ALVAC-C, gp120-C trial failing to show efficacy21,22. The only HIV-1 clinical trial to date to demonstrate any efficacy, RV144, induced nnAbs by immunizing with gp120 proteins derived from HIV-1 isolates A244 and MN23. Protection in that trial correlated with binding IgG to the second variable region on Env and ADCC, but the protective antibody response was not durable24.

In contrast to nnAbs, bnAbs are difficult to induce for many reasons including shielding of bnAb Env epitopes by glycans, the metastability of Env conformation, conformational masking of neutralizing epitopes25,26,27, host immune controls preventing bnAb development28,29, and the requirement for multiple highly improbable somatic mutations that are needed for the acquisition of antibody neutralization breadth30,31,32. mRNA immunization elicits antigen-specific follicular T helper cells, which are key for affinity maturation of antibodies in germinal centers33,34, and has been shown to correlate with bnAb development during natural infection35,36. Regardless of whether it is nnAbs or bnAbs targeted, the poor durability of HIV-1 Env antibody responses37, and the necessity of designing a vaccine that has multiple components for bnAb induction19,20 necessitates additional vaccine platforms that can maintain protective antibody levels.

Here, we compare the antibody responses in rhesus macaques induced by either HIV-1 Env-encoded as nucleoside-modified mRNA-LNP or the same vaccine candidate administered as an adjuvanted HIV-1 Env recombinant protein. In each comparison, mRNA-LNP administration induced equal or better immune responses as proteins. Moreover, mRNA-LNP vaccination encoding soluble HIV-1 Env trimers or ZIKV prM-E elicited durable neutralizing antibodies that were stable for ~1 year.


Both protein and mRNA vaccination elicited high titers of serum HIV-1 binding antibodies

We compared the immunogenicity of HIV-1 Env A244 gp120 lacking the first 11 N-terminal amino acids of gp120 (Δ11 gp120, as in the RV144 trial)38 administered as either nucleoside-modified mRNA-LNP or adjuvanted recombinant protein in rhesus macaques. We immunized two groups of five macaques with recombinant protein formulated in different adjuvants. One group of five macaques received A244 Δ11 gp120 formulated with the aluminum hydroxide adjuvant Rehydragel (Fig. 1a). This group mimicked the A244 Δ11 gp120 protein aluminum hydroxide (Rehydragel) formulation used in the RV144 ALVAC/B/E gp120 trial23. Adjuvants vary in their immunostimulatory strength; thus, in the second group of five macaques, we assessed the immunogenicity of the A244 Δ11 gp120 Env in combination with a liposomal adjuvant formulation composed of monophosphoryl lipid A (a TLR-4 agonist) and QS21 (ALFQ)39. In addition, two groups of macaques were immunized with nucleoside-modified mRNA-LNP encoding different forms of A244 Δ11 gp120. One group of macaques received A244 Δ11 gp120, whereas the other group received mRNAs encoding A244 Δ11 gp120 with a D368R mutation that disrupted the CD4 binding site (Fig. 1a). The purpose of the CD4 binding site knockout mutation was to determine whether Env binding to CD4+ T cells in vivo altered the immunogenicity of HIV-1 Env gp120. Macaques were immunized with either protein or mRNA-LNP at weeks 0 and 6 and antibody responses were followed for 18 total weeks (Fig. 1a).

Fig. 1: Immunization of rhesus macaques with HIV-1 Env recombinant protein adjuvanted with ALFQ or HIV-1 Env-encoding mRNA-LNP elicits comparable titers of gp120-specific serum IgG.

a Rhesus macaque vaccine regimen and biospecimen collection schedule. The different immunogens are listed on the right side. b Serum IgG binding titers for each group are shown as mean log area-under-the-curve (AUC) against A244 gp120, A244 gp120 mutants, and B.63521 gp120. A244 gp120 was mutated at the ADCC site of immune pressure observed in the RV144 trial (A244 K169V). In addition, the V3-glycan (N295A/N301A/N332A), V2-glycan (N156K/N160K), CD4 binding site (delta371I/P363N) neutralizing epitopes were mutated on A244 gp120. The group mean and standard error are shown (n = 5 per group). Arrows indicate immunization timepoints.

Binding nnAbs to Env immunogens have been associated with protection in macaques from SHIV infection40,41. Thus, we first determined serum binding IgG elicited by each of the mRNA/LNP or protein groups. Regardless of the site of immunization, gp120-specific IgG titers in the macaque serum were detectable after a single immunization and continued to rise after the second immunization (Fig. 1b). mRNA-LNP elicited nearly identical serum IgG antibody titers as compared to macaques immunized with A244 Δ11 gp120 formulated with ALFQ (Fig. 1b). Comparison of mRNA-LNP that encoded wildtype gp120 versus CD4 mutated gp120 showed identical binding IgG elicitation (Fig. 1b). Macaques that received A244 Δ11 gp120 formulated with Rehydragel generated the lowest group mean serum IgG binding titers to A244 Δ11 gp120. For these ELISA assays, the differences between the Rehydragel group and the other groups were greatest after one immunization, but the small numbers of monkeys per group precluded statistical analysis. Nonetheless, ALFQ adjuvant-induced higher levels of A244 Δ11 gp120 antibody compared to gp120 in Rehydragel (Fig. 1b). The same group ranking was also observed for binding IgG to a HIV-1 Env gp120 B.63521 from a non-vaccine matched HIV-1 isolate (Fig. 1b). The binding of serum antibodies to A244 gp120 was not reduced by the introduction of mutations in the CD4 binding, glycan-V3, and glycan-V2 sites (Fig. 1b). Thus, mRNA-LNP gp120 vaccination produced comparable IgG titers in plasma as recombinant gp120 protein in the potent adjuvant ALFQ.

IgG responses to the second variable (V2) region of HIV-1 Env were a correlate of reduced infection risk in the RV144 trial24. Similarly, K169 in the V2 region was a site of immune pressure during the RV144 trial42. To assess whether nucleoside-modified mRNA-LNP vaccination elicited comparable V2 region IgG antibodies in plasma, we tested serum IgG binding to various minimal antigens that recapitulated the V2 region of HIV-1 Env. These antigens included V1V2 proteins, a V2 peptide, and V2 scaffolded on gp70. Binding to each antigen showed the same pattern as gp120 binding. Recombinant protein adjuvanted with AFLQ and mRNA-LNP immunization elicited nearly identical plasma IgG responses (Fig. 2). Recombinant gp120 adjuvanted with Rehydragel elicited the weakest responses for all the antigens tested, with group difference being largest after a single immunization (Fig. 2). V1V2-specific binding IgG titers were comparable whether the V1V2 antigen matched the vaccine strain or if it was from an unmatched virus 1086C (Fig. 2a). The first variable region (V1) was not necessary for V2 binding, as plasma IgG bound to a short V2 only peptide, K178 (Fig. 2b). The antigen for which binding IgG correlated with reduced infection risk in the RV144 trial was B.CaseA V1V2 scaffolded on gp7024,43. Each group of macaques generated binding IgG antibodies to gp70 B.CaseA V1V2 protein (Fig. 2c). The antibodies were specific for HIV-1 V1V2, as no antibodies were detected against the control gp70 scaffold presenting murine leukemia virus V1V2 (Fig. 2c). However, mutation of K169 had little effect on B.CaseA V1V2 binding IgG titers. Similarly, mutation of the K169 in the V2 region of A244 gp120 had little effect on binding IgG titers (Fig. 1b). Thus, V1V2 antibodies were not solely dependent on the K169 at the site of immune pressure observed in the RV144 trial.

Fig. 2: Adjuvanted recombinant protein and mRNA-LNP immunization of rhesus macaques elicits gp120 V2-specific plasma IgG responses.

Plasma IgG binding titers recombinant V1V2 proteins, V2 peptide and gp70 scaffolded V1V2 proteins. The group mean and standard error are shown (n = 5 per group). Arrows indicate immunization timepoints.

To determine the breadth of reactivity and epitope specificity of the first and second variable (V1V2) region antibodies induced by each vaccine, we performed linear V1V2 peptide arrays. Vaccination induced only low levels of V1-binding antibodies, but high titers of V2 antibodies across the different vaccines (Fig. 3a–d and Supplementary Fig. 1). V2 antibody binding magnitudes were similar for mRNA-LNP and protein formulated with ALFQ (Fig. 3). Each vaccine-elicited antibodies capable of binding V2 peptides from clade AE and C (Fig. 3 and Supplementary Fig. 1). Therefore, mRNA-LNP and adjuvanted protein vaccination elicited V2 antibody responses that were similar in magnitude, breadth, and epitope specificity.

Fig. 3: Adjuvanted recombinant gp120 and mRNA-LNP-encoded gp120 immunizations elicit plasma IgG with similar V2 specificities, but different C2 and C3 specificities.

Plasma IgG binding to HIV-1 Env peptides spanning (ad) the V1V2 region of gp120 or (ef) the entire gp120 subunit. Arrows indicate differences in C2 (red) and C3 (blue) binding antibodies. HIV-1 Env regions encompassed by the peptides are listed above the curves. Each panel shows the response for the immunization group listed in the title. Graph lines are colored based on HIV-1 isolate. The group medians are shown by solid lines. Dashed lines indicate individual macaques in (ad).

Next, we compared plasma antibody specificities induced by protein or mRNA-LNP immunization at HIV-1 Env gp120 sites outside of the V1V2 site. mRNA-LNP induced similar A244 gp120 linear epitope antibodies compared to protein formulated in adjuvant with two exceptions. First, recombinant protein, but not mRNA-LNP, elicited plasma IgG to the C-terminal portion of the second constant (C2) region (Fig. 3e–h and Supplementary Fig. 2). Second, mRNA-LNP immunization elicited plasma IgG against the third constant region but recombinant protein immunization did not (Fig. 3e–h and Supplementary Fig. 2).

To further compare polyclonal plasma IgG specificities, we assessed the ability of post-vaccination plasma to block the binding of gp120 monoclonal antibodies (mAbs) to HIV-1 Env. Plasma from either adjuvanted protein- or mRNA-LNP-immunized macaques was added to HIV-1 Env protein, followed by the addition of biotinylated mAbs to determine plasma antibody blocking of monoclonal antibody binding to Env. We examined blocking of nnABs CH58, which targets the V2 site of immune pressure identified in RV14444, and A32 which defines an immunodominant ADCC gp120 site45 that synergizes with V2 antibodies to mediate ADCC46 (Fig. 4a). Next, we determined whether plasma could block the HIV-1 entry receptor CD4 from binding to Env (Fig. 4b), or block V2-glycan bnAbs PG9 and CH01 and N332 glycan-dependent bnAbs 2G12 and PGT125 Env binding (Fig. 4c, d). Either immunization with adjuvanted recombinant gp120 protein or mRNA-LNP elicited plasma antibodies that blocked the binding of CH58, A32, CD4, CH01, 2G12, and PGT125 (Fig. 4). Comparison of mRNA-LNP-induced antibodies versus gp120 protein-induced plasma antibodies showed that mRNA-LNP immunization elicited the same magnitude of blocking after 2 immunizations as protein adjuvanted with ALFQ (Fig. 4). Blocking activity was lowest for animals immunized with protein adjuvanted with Rehydragel (Fig. 4). The blocking of CH01 binding to A244 most likely represented CH58-like antibody binding and not V2-glycan bnAb binding, since CH01 binding can be blocked by CH58-like linear V2 antibodies (Fig. 4).

Fig. 4: ALFQ-adjuvanted recombinant gp120 and mRNA-LNP-encoded gp120 immunizations in macaques elicit similar magnitudes of plasma antibodies capable of blocking ADCC-mediating antibodies, HIV-1 bnAbs, and CD4 binding to Env.

ae Plasma antibody blocking of (a) ADCC-mediating antibodies (CH58 and A32), b N332 glycan bnAbs (2G12 and PGT125), c, d V2-glycan CH01, PG9), and e soluble CD4. Antibody and Env names are listed in the graph title. Note, plasma IgG blocking is absent when non-vaccine matched Envs 9021 or B.6240 are used as the Env antigen in (d). The group mean and standard error are shown (n = 5 per group). Arrows indicate immunization timepoints.

Comparable neutralizing and non-neutralizing antibody functions induced by adjuvanted Env gp120 protein versus mRNA-LNP immunization

Several IgG Fc receptor (R)-mediated immune responses have been associated with decreased transmission risk either in the RV144 trial or in Env immunizations studies in macaques followed by low dose mucosal SHIV challenges41,47. These immune responses include binding to HIV-1 infected CD4+ T cells, ADCC, and antibody-dependent cellular phagocytosis (ADCP). To assess the potential for week 8 plasma IgG to mediate effector functions against infected cells, we first examined plasma IgG binding to HIV-1.CM235-infected T cells. Plasma IgG from all four groups of macaques was able to bind to HIV-infected cells (Fig. 5a, b). When binding levels of IgG was quantified as mean fluorescence intensity of bound IgG or the percentage of cells positive for HIV-1 protein p24 and plasma IgG, mRNA-LNP vaccination and recombinant protein adjuvanted with ALFQ were not different (Fig. 5a, b). In agreement with the overall lower IgG titers, Rehydragel-adjuvanted protein gave the weakest cell-binding responses (Fig. 5a, b).

Fig. 5: Adjuvanted recombinant protein and mRNA-LNP immunizations elicit antibodies that mediate comparable magnitudes of infected cell binding and effector functions.

a, b Plasma IgG binding to CM235-infected cells measured as mean fluorescence intensity (MFI) or percentage of infected (p24+) cells with detectable plasma IgG binding (Ab+). Each symbol represents an individual macaque. c Antibody-dependent cellular cytotoxicity (ADCC) of CM235-infected cells. dAntibody-dependent cellular phagocytosis (ADCP) of A244 ∆11 gp120-coated fluorescent beads. Symbols indicate scores for individual macaques immunized with adjuvanted recombinant A244 ∆11 gp120 protein (green and blue) or A244 ∆11 gp120 mRNA-LNP (red and orange). HIV-1 bnAb 2G12 and influenza bnAb CH65 were used as positive and negative controls respectively. Box and whisker plots show minimum values, maximum values, median, and interquartile ranges.

ADCC activity was measured in a flow cytometry-based granzyme B assay. There was a clear adjuvant effect between the two protein-immunized group. ADCC titers were approximately one order of magnitude higher when ALFQ was used as the adjuvant as compared to Rehydragel. Similarly, mRNA-LNP immunization (mean ± standard deviation = 22,916 ± 22,102) induced activity lower than ALFQ-formulated recombinant protein, but threefold higher than protein adjuvanted with Rehydragel (mean ± standard deviation = 7324 ± 7024) (Fig. 5c). Elimination of CD4 binding to A244 Δ11 gp120 had no effect on ADCC activity (Fig. 5c). Lastly, we compared ADCP activity of week 8 plasma antibodies from macaques administered adjuvanted recombinant protein or mRNA-LNP. Median plasma ADCP activity against A244 gp120-coated beads was similar across all groups with A244 Δ11 gp120 in Rehydragel eliciting a wider range of responses (Fig. 5d). In agreement with previous studies, nucleoside-modified mRNA-LNP vaccination elicited antibody effector functions that have been shown previously to correlate with reduced infection risk24,48.

While the goal of A244 Δ11 gp120 immunization was to induce non-neutralizing effector functions like those seen in the RV144 trial, we compared elicitation of neutralizing antibodies by each of the vaccines. We selected for testing the tier 1 HIV-1 strain CRF_01AE 92TH023, against which A244 has consistently induced neutralizing antibodies24,44. We found that there was no significant difference among the different vaccination regimens for induction of HIV-1 92TH023 neutralizing antibodies (Supplementary Fig. 3). These studies showed that for immunization with gp120 monomers aiming to elicit potentially protective non-neutralizing V2 antibodies, nucleoside-modified mRNA-LNP vaccination was superior to recombinant protein formulated with aluminum hydroxide and comparable to recombinant protein formulated with a more complex TLR-4/QS21/liposomal adjuvant.

Immunizations of macaques with sequential CH505 Env gp160 mRNA-LNP, SOSIP gp140 mRNA-LNP, or adjuvanted recombinant SOSIP gp140 proteins

We previously isolated CD4 binding site bnAbs from the African individual CH50549,50. From CH505, we cloned a series of different CH505 Envs for testing as immunogens to induce similar types of bnAbs with vaccination49,50. To this end, we studied a mRNA-LNP sequential immunization strategy where each individual envelope in the series is delivered one-at-a-time in a specific order. The HIV-1 envelope immunogens were designed as transmembrane HIV-1 gp160s (50 μg/immunization) or soluble, stabilized gp140 SOSIP trimers (50 μg/immunization; Fig. 6a). The SOSIP gp140s were stabilized by a chimeric gp41 and the addition of E64K and A316W mutations51. The immunogenicity of these two sets of mRNA-LNP was compared to that of 100 μg of the same set of CH505 Envs as soluble gp140 SOSIP proteins formulated in the TLR-4 adjuvant, GLA-SE (Fig. 6a). Four rhesus macaques were immunized every four weeks with either set of immunogens (Fig. 6a), and binding antibody and neutralizing antibodies were measured.

Fig. 6: Comparison of HIV-1 Env trimer immunogenicity after mRNA-LNP and recombinant protein immunization.

a Macaque immunization regimen and biospecimen collection. bd. Plasma IgG binding titers as log AUC to CH505 TF b gp120, c V2, and d V3. e Difference in binding titers to wildtype CH505 TF gp120 and CH505 TF gp120 with the CD4 binding site knocked out with a ∆371Ile mutation. Values above zero indicate higher binding to wildtype gp120 than the CD4 binding site mutant. f Plasma IgG binding titers as log AUC to HIV-1.MN gp41. Lines represent group mean values (n = 4 per group). gAntibody binding to HIV-1 gp160 expressed on the surface of Freestyle 293-F cells after mRNA transfection. Antibody binding is shown as mean fluorescence intensity measured by flow cytometry. Dark-colored and bright-colored bars indicate non-neutralizing and broadly neutralizing antibodies respectively. Antibody epitopes examined were: 7B2, gp41; 17B, co-receptor binding site; 19B, V3; PGT145, CH01 and PG9, V1V2-glycan; PGT125, V3-glycan; and CH65, anti-influenza heamaglutttin antibody. Independent replicates are indicated by black circles. Mean and standard error are shown for the 2–4 independent replicates.

All three vaccines were immunogenic in macaques. With regard to induction of binding titers of gp120 antibodies, mRNA-LNP encoding sequential CH505 gp160s induced higher gp120 titers than did either of the mRNA-LNP encoding soluble gp140 SOSIP trimers or soluble gp140 SOSIP trimers proteins (Fig. 6b). These binding titers rose dramatically after two immunizations with gp160 mRNA-LNP, but rose gradually over the course of five immunizations in macaques immunized with SOSIP gp140s (Fig. 6b). Plasma IgG binding titers to gp120 were equivalent between the mRNA-LNP and the adjuvanted soluble gp140 SOSIP trimer protein-immunized animals (Fig. 6b). Soluble gp120 exposes non-neutralizing epitopes, thus we examined whether the high titers of gp120 antibodies in macaques immunized with gp160 mRNA-LNP were due to antibodies targeting these sites. Indeed, the gp160 mRNA-LNP induced very high titers of antibodies to linear CH505 TF V2 peptides, whereas titers in the gp140 SOSIP trimer groups showed no or a slight increase from baseline (Fig. 6c). Linear V3 epitopes were also highly immunogenic in gp160 mRNA-LNP-immunized macaques, and to a lesser extent was immunogenic in soluble gp140 SOSIP trimer protein-immunized macaques (Fig. 6d). Interestingly, V3 antibodies were not elicited by gp140 SOSIP trimer mRNA-LNP, suggesting a difference in V3 exposure when the trimer was expressed in vivo or potential effects of the adjuvant. To examine the induction of gp120 antibodies that may target conserved neutralizing epitopes, we assessed the amount of gp120 antibody binding that was dependent on the CD4 binding site. We mutated the CD4 binding site with a deletion of the isoleucine at position 371 and determined the decrease in plasma IgG binding to CH505 TF gp120. Both groups of macaques that received SOSIP trimers as either protein gp140s or mRNA-LNP were superior to the mRNA-LNP encoding the gp160s (Fig. 6e). Immunization with gp140 SOSIP trimers elicited differential binding between the mutant and wildtype gp120 that rapidly receded after each immunization, but was maintained for 16 weeks once immunizations were stopped (Fig. 6e). Similar to linear V2 and V3 peptide antibodies, gp160 mRNA-LNP induced high levels of gp41 antibodies whereas gp140 SOSIP trimers delivered as mRNA-LNP or adjuvanted recombinant protein had levels close to baseline (Fig. 6f). Thus, gp160 mRNA-LNP immunization elicited higher titers of undesired antibodies targeting non-neutralizing gp120 and gp41 epitopes and lower titers of desired CD4 binding site antibodies than SOSIP trimer-encoding mRNA-LNP or adjuvanted recombinant protein. However, none of the regimens induced significant tier 2 or heterologous neutralizing antibodies due to the need for high-affinity germline B cell targeting and designs of sequential Env regimens that select for bnAb improbable mutations (Supplementary Tables 13)20,31,32,52,53.

The high levels of undesired antibodies elicited by gp160 mRNA-LNP immunization could be due to poorly folded HIV-1 envelope expressed on the surface of cells. We hypothesized that stabilization of HIV-1 envelope with amino acid substitutions that preserve Env trimers in their native conformation could reduce HIV-1 envelope binding to non-neutralizing antibodies. We introduced stabilizing amino acid changes H66A and A582T into envelope54. Also, we generated envelopes with G458Y and cysteines at positions 113 and 429 or 113 and 43155. The cysteines have been shown to form disulfide bonds that keep the HIV-1 envelope in the closed conformation55 and G458Y stabilizes the envelope fifth variable loop53. Among these variants of the M5 gp160 envelope, adding H66A and A582T reduced binding by non-neutralizing antibodies against V3 and gp41, while leaving trimer-specific (PGT145) or timer-preferring (PG9 and CH01) antibody binding unchanged (Fig. 6g). Thus, mRNA immunization with stabilized HIV-1 envelope gp160 is one potential approach to improve the elicitation of neutralizing antibodies, while not engaging the B cell receptor of B cells that produce antibodies that cannot bind native, fusion-competent HIV-1 envelope.

Nucleoside-modified mRNA-LNP immunization induces durable antibody responses against Zika prM-E and HIV-1 Env

Inducing durable Env antibody responses is a key goal of HIV vaccine development. Yet in the RV144 trial, protective antibodies fell dramatically over the first 42 weeks after vaccination24. Thus, if immunization with mRNA-LNP induced durable antibody responses it would benefit their use as a vaccine platform for many different infectious diseases. We assessed the durability of antibody responses using neutralization assays of tier 1 CH505 w4.3 virus, since antibodies capable of neutralization of the tier 2 CH505 transmitted founder (TF) virus were not elicited (Supplementary Tables 13). CH505 w4.3 is an early virus isolate that is identical to the tier 2 CH505 TF virus with the exception of a W680G mutation that makes it highly sensitive to HIV antibody neutralization49,56,57. Longitudinal comparative analyses of neutralization of the CH505 w4.3 virus showed that the gp160-encoding mRNA-LNP elicited higher titers of neutralizing antibodies than the gp140 SOSIP trimer-encoding mRNA-LNP or the gp140 SOSIP trimer proteins in GLA/SE58 (Fig. 7a). Notably, neutralization of the CH505 w4.3 virus was still detectable in all three groups 12 weeks after the last immunization (Fig. 7a). There were also sporadic low levels of neutralization of CH505 or 426 C viruses that were modified to be highly sensitive to CD4 binding site antibodies (Supplementary Tables 13).

Fig. 7: HIV-1 Env mRNA-LNP immunization, like Zika prM-E immunization, elicits durable HIV-1 serum neutralizing antibodies.

a Comparison of macaque serum neutralization of HIV-1 infection of TZM-bl cells. Neutralization titers are shown as the reciprocal serum dilution that inhibits 50% of virus replication (ID50). Lines represent the group geometric mean for macaques immunized with gp160 mRNA (purple), SOSIP gp140 mRNA (maroon), or SOSIP gp140 protein (light orange). b Vaccine-elicited serum neutralization titers as ID50 against Zika (forest green) and HIV-1.CH505 w4.3 (purple) in macaques immunized at different times with both Zika prM-E and a series of HIV-1 CH505 gp160 mRNA. ZIKV neutralization was measured by the PRNT assay. c Plasma IgG binding titers (log area-under-the-curve) to HIV-1 CH505 TF gp120 (purple) and Zika prM-E (forest green) for the macaques shown in (b). The line represents the group mean binding titer.

While these results indicated that durable neutralizing antibodies were elicited by each vaccine regimen, the downward trend of the neutralization titers in these macaques and the A244-immunized macaques raised the question of how long would neutralizing antibodies persist at detectable levels. A subset of the macaques used in this HIV-1 study was previously administered 50 μg of mRNA-LNP encoding ZIKV pre-membrane and envelope (prM-E). These macaques generated protective anti-Zika neutralizing antibody responses1. Using these macaques that were administered both HIV and Zika mRNA-LNP vaccines, we determined serum neutralizing antibodies titers over 52 weeks (Fig. 7b). We found that neutralizing antibodies against ZIKV were maintained until the last timepoint of follow-up 52 weeks after immunization. Neutralizing antibodies to HIV-1 persisted for the duration of the 41 weeks of follow-up. During these 41 weeks, HIV-1 titers initially fell ~10-fold after the last immunization but plateaued at ~1:100 titer (Fig. 7b). In contrast, Zika neutralizing antibody levels were maintained at the same level for the 52 weeks after being detected at week 2 (Fig. 7b). Similar patterns were observed for plasma binding IgG titers to CH505 gp120 and Zika prM-E (Fig. 7c). Taken together, nucleoside-modified mRNA-LNP immunizations induced durable serum neutralizing antibodies for both HIV-1 and ZIKV.

Neutralizing antibody titers are dependent on mRNA-LNP dose

In our previous ZIKV vaccine studies in nonhuman primates, we administered 50, 200, or 600 μg of mRNA-LNP1. Given the potent elicitation of neutralizing antibodies by mRNA administered at each of these doses, we sought to determine whether the mRNA dose could be further decreased. We immunized four macaques intramuscularly with either 50, 20, or 5 μg of mRNA-LNP encoding Zika prM-E (Fig. 8a) and compared titers of binding IgG and neutralizing antibodies. Fifty and twenty microgram doses of mRNA-LNP elicited similar titers of binding IgG and neutralizing antibodies (Fig. 8b, c). Titers of ZIKV binding IgG and neutralizing antibodies were substantially decreased when the mRNA-LNP dose was lowered to 5 μg, but was detectable with the administration of this single administration of a small amount of mRNA-LNP in macaques. The route of immunization was not critical as administering 50 μg of mRNA-LNP either intramuscularly or intradermally elicited comparable Zika envelope binding IgG and Zika neutralizing antibodies (Fig. 8b, c). In summary, the mRNA-LNP dose could be lowered from 50 to 20 μg in macaques without detrimental effects to antibody responses. Thus, more immunizations can be performed with each preparation of mRNA-LNP by reducing the dose by 60 percent.

Fig. 8: Neutralizing antibody titers are dependent on mRNA dose, but are similar for intramuscular and intradermal immunization routes.

a Immunization regimen for groups of 4 rhesus macaques immunized with different amounts of mRNA-LNP via intramuscular or intradermal routes. b Plasma binding IgG titers specific for ZIKV envelope protein were determined by ELISA before vaccination (week 0) or 6 weeks post-vaccination. Binding IgG titers are shown as log area-under-the-curve. Each symbol represents an individual macaque and the bar shows the group mean and standard error. c Serum neutralization titers against ZIKV determined by the reporter virus particle assay are shown as the 50% effective concentration (EC50). Symbols are the same as in (b) and bars represent the group geometric mean.


Here, we demonstrate in 28 rhesus macaques that immunization with nucleoside-modified mRNA-LNP (n = 16) was equal to or superior to the immunogenicity of adjuvanted Env protein. There are multiple factors for why nucleoside-modified mRNA immunization has potent immunogenicity for viral proteins. First, mRNA is delivered by LNP to dendritic cells and likely other immune cells that are able to activate naive T cells to respond to the vaccine immunogen59,60,61. Second, nucleoside-modified mRNA-LNP immunization generates robust antigen-specific germinal center T and B cell responses2. Within the germinal center, follicular T helper cells provide help to B cells undergoing affinity maturation34. The affinity maturation process is critical for the development of high-affinity antibodies after vaccination, and in particular for HIV-1 is required for bnAb development62. Animal models vaccinated with nucleoside-modified mRNA-LNP encoding HIV-1 Env or influenza virus hemagglutinin develop high levels of T follicular helper cells33. Given that HIV-1 neutralizing antibody breadth is correlated with the frequency of circulating T follicular helper cells35,36, nucleoside-modified mRNA-LNP immunization warrants further study as an HIV-1 vaccine platform. Nucleoside-modified mRNAs have many potential advantages over proteins including speed of production and cost-savings. Moreover, mRNA-LNP increases the feasibility of making under current good manufacturing practices the multicomponent vaccines that are needed to induce both protective HIV-1 bnAbs and nnAbs.

In RV144, aluminum hydroxide (Rehydragel) was used and poor durability of protective immune responses was observed24,63. ALFQ is a potent adjuvant that contains anionic liposomes, QS21, and TLR-4 agonist, MPL39, both potent adjuvants. Thus, comparison of mRNA-LNP with Rehydragel gives a direct indication of how it might fare versus the RV144 regimen, and comparison with ALFQ provides an indication of how it might fare in comparison with a stronger adjuvant. The adjuvant GLA (synthetic MPL) in SE (stable emulsion)58 has been used in HVTN 115 with multiple gp120s and was immunogenic, although in a head-to-head comparison of Rehydragel and GLA-SE in primates, they were similar in promoting immunogenicity except for V2 responses where Rehydragel was superior. While ALFQ and GLA stable emulsion contain lipids, they differ from the LNPs. The LNPs include ionizable cationic lipids that give the LNP a neutral surface charge, whereas most commercial liposomes lack ionizable cationic lipids. The neutral charge helps reduce non-specific protein binding in vivo. Unlike GLA or ALFQ, the LNP does not activate TLR-4, but instead activates follicular helper T cells33. In general, with the doses and adjuvants used in this study of proteins versus mRNA-LNP, nucleoside-modified mRNA-LNP immunization was superior or equal to protein in Rehydragel or GLA-SE and equal to protein in ALFQ. Future studies using different vaccination intervals, adjuvants, and protein doses could corroborate this conclusion.

There are currently two ongoing efficacy trials that test the protective capacity of non-neutralizing HIV-1 antibodies with viral vector prime, Env protein boosts to determine if the 42-month efficacy of ALVAC/gp120 can be replicated. In addition, the just-completed HVTN 702 trial (NCT02968849) uses an ALVAC with a clade C Env insert as prime and ALVAC-C + a bivalent C gp120 Env boost failed to confer any protection from HIV-1 infection in South Africa in an area of high HIV-1 infection rates. The ongoing HVTN 705 trial (NCT03060629) tests the efficacy of an adenovirus (Ad) 26 vector containing mosaic HIV-1 genes as a prime and the same Ad26 + a clade C gp140 Env as a boost, and the HVTN 706 trial (NCT03964415) tests the efficacy of Ad26 mosaic HIV as a prime and Ad26 + a bivalent clade C gp140 Env + a mosaic gp140 Env as boost64. The correlate of decreased transmission risk in macaque studies of the A26, Ad26 + gp140 boost SHIV challenge studies did not include neutralization breadth of antibodies, but instead included clade C Env ELISA binding antibodies and interferon-gamma Env ELISPOT values64. Thus, these latter two vaccines will test whether non-neutralizing antibody effector functions can mediate significant protection in humans.

The ease and cost-effectiveness of mRNA-LNP production compared to the production of multiple recombinant proteins by good manufacturing practice (GMP) techniques make mRNA-LNP particularly attractive. Furthermore, the scalability of mRNA-LNP for thousands to millions of doses again make the cost and ease of production of mRNA-LNP attractive. These manufacturing considerations are key for improving on RV144, and show great promise for COVID-19 vaccines currently under development17. However, the cold temperature required for the distribution and storage of mRNA vaccines remains a challenge for this vaccine platform65.

While mRNA vaccination is promising, there are still improvements that can be made for HIV mRNA-LNP vaccination. Specifically, for bnAb induction by mRNAs, the near-native Env trimers will need to be stabilized such that they will be able to be produced by the transfected cell in a well-folded state. This challenge is not unique to mRNA as DNA vaccination has also faced similar roadblocks66. This aspect of genetic vaccination differs from recombinant proteins that can be purified prior to immunization and recognition by the immune system. However, there are a plethora of available mutations for stabilizing HIV-1 envelope, and each envelope has different intrinsic stability that can be further augmented. We investigated only two potential sets and found H66A and A587T reduced non-neutralizing antibody binding to the envelope in vitro. The reduction in non-neutralizing antibody binding suggests a higher percentage of native-like envelope expressed by the mRNA. Future vaccine studies will need to determine whether reduced non-neutralizing antibody binding in vitro translates to improved neutralizing antibody elicitation in vivo. Despite the best envelope stabilization designs, it is possible that not all of the envelope will be well-folded, native-like trimers. In vivo studies by us and others are aiming to determine what the percentage of well-folded envelope needs to be in order to elicit optimal neutralizing antibody responses. Both of these important questions are areas of intense investigation and are critical for mRNA becoming a widely accepted HIV-1 vaccine platform.

In summary, comparison of nucleoside-modified mRNA-LNP with recombinant Env proteins in three adjuvants demonstrated the utility of mRNA-LNP as a mode of inducing nnAbs that are predicted to be protective against retrovirus challenge. We suggest that moving to nucleoside-modified mRNA-LNP for the next generation of clinical trials for multivalent Env immunization will be advantageous and speed the development of a globally available protective HIV-1 vaccine.

Article classification: Biological abstract
Share to:
Add:Room A11-329, 1st Floor, No.1, SBI Venture Street, Optics Valley, East Lake
New Technology Development Zone, Wuhan, China.
Certificate NO.:U18Q28010569R0S