Comparison of the immunogenicity of mRNA-encoded and protein HIV-1 Env–ferritin nanoparticle designs

ABSTRACT

Nucleoside-modified mRNA technology has revolutionized vaccine development with the success of mRNA COVID-19 vaccines. We used modified mRNA technology for the design of envelopes (Env) to induce HIV-1 broadly neutralizing antibodies (bnAbs). However, unlike SARS-CoV-2 neutralizing antibodies that are readily made, HIV-1 bnAb induction is disfavored by the immune system because of the rarity of bnAb B cell precursors and the cross-reactivity of bnAbs targeting certain Env epitopes with host molecules, thus requiring optimized immunogen design. The use of protein nanoparticles (NPs) has been reported to enhance B cell germinal center responses to HIV-1 Env. Here, we report our experience with the expression of Env–ferritin NPs compared with membrane-bound Env gp160 when encoded by modified mRNA. We found that well-folded Env–ferritin NPs were a minority of the protein expressed by an mRNA design and were immunogenic at 20 µg but minimally immunogenic in mice at 1 µg dose in vivo and were not expressed well in draining lymph nodes (LNs) following intramuscular immunization. In contrast, mRNA encoding gp160 was more immunogenic than mRNA encoding Env–NP at 1 µg dose and was expressed well in draining LN following intramuscular immunization. Thus, analysis of mRNA expression in vitro and immunogenicity at low doses in vivo are critical for the evaluation of mRNA designs for optimal immunogenicity of HIV-1 immunogens.

IMPORTANCE

An effective HIV-1 vaccine that induces protective antibody responses remains elusive. We have used mRNA technology for designs of HIV-1 immunogens in the forms of membrane-bound full-length envelope gp160 and envelope ferritin nanoparticle. Here, we demonstrated in a mouse model that the membrane-bound form induced a better response than envelope ferritin nanoparticle because of higher in vivo protein expression. The significance of our research is in highlighting the importance of analysis of mRNA design expression and low-dose immunogenicity studies for HIV-1 immunogens before moving to vaccine clinical trials.

INTRODUCTION

Broadly neutralizing antibodies (bnAbs) for HIV-1 target conserved regions on the envelope (Env) glycoprotein and neutralize heterologous HIV-1 isolates (1–3). Recent HIV Vaccine Trial Network (HVTN) trials HVTN 703 and HVTN 704—also known as the antibody mediated protection (AMP) trials—provided proof of concept that high titers of serum bnAbs can prevent the acquisition of sensitive HIV-1 strains (4, 5). Inducing bnAbs is a current major goal of HIV-1 vaccine development (1). However, despite decades of research, a safe and effective bnAb-inducing HIV-1 vaccine remains elusive, due to unusual features of bnAbs—including long heavy chain complementary domain region 3 (HCDR3) and high levels of somatic hypermutation (SHM)—that are required for neutralization breadth and potency (1, 6, 7). Extensive SHM in bnAbs means that B cells bearing B cell receptors (BCRs) of bnAb precursors must go through iterative cycles of antibody gene diversification and selection in germinal centers (8). Moreover, many of these mutations are in cold spots of activation-induced cytidine deaminase (AID) activity and rarely occur (8). To induce bnAbs with many improbable mutations through vaccination, a strategy called “mutation-guided B cell lineage immunogen design” has been proposed that involves the use of a priming immunogen to activate and expand naive bnAb B cells bearing the unmutated common ancestor (UCA) of a bnAb B cell lineage, followed by multiple boosting immunogens engineered to guide the bnAb UCA and intermediates to affinity-mature into bnAb B cells (1, 6, 8). HIV-1 immunogens need to be engineered such that the priming immunogen binds the UCA of a bnAb B cell lineage at high affinity (9) followed by sequential immunogens with an affinity gradient between immunization steps to create an “affinity pull” with immunogen fast on-rates such that a ceiling of affinity is not reached until the B cell lineage achieves full maturation (1, 8, 10–12).

Nucleoside-modified mRNA design for COVID-19 vaccines has been remarkably successful (13, 14) and required a double-proline stabilization mutation in the spike protein (S-2P) for enhanced protein stability and expression (15). We have recently demonstrated stabilizing mutations that optimize various forms of HIV-1 Env expression by modified mRNA (16, 17). In those studies, we used relatively high levels of Env-encoding mRNA in mice with 20 µg per intramuscular dose. Here, we have further evaluated modified mRNA immunogenicity in a wide dose range in rhesus macaques and bnAb UCA knock-in mice. We found that a modified mRNA in lipid nanoparticles (mRNA–LNP) encoding an Env SOSIP trimer ferritin nanoparticle (Env–NP) was not sufficiently immunogenic in rhesus macaques even at 200 µg dose. The evaluation of expression levels of this mRNA demonstrated both inabilities to form high levels of well-folded Env–NPs in vitro and limited expression in vivo. In contrast, mRNA–LNP encoding membrane-bound Env gp160 expressed higher levels of protein in vivo and demonstrated better immunogenicity at lower doses in mice. Thus, testing for adequate levels of in vivo expression of well-folded, stabilized immunogens is critical for optimal immunogenicity of modified mRNA encoding HIV-1 Env immunogens.

RESULTS

Protein CH848 10.17DTe Env–NP induced higher binding and neutralizing antibody responses than mRNA–LNP in rhesus macaques

The V3–glycan bnAb B cell lineage DH270 was isolated from a chronically infected person living with HIV-1 (PLWH) (18). The mature bnAb DH270.6 neutralized approximately 50% of 207 HIV-1 isolates. To induce DH270-like bnAb B cell lineages, we engineered a priming Env by removing two glycans on the V1 loop (CH848 10.17 N133D N138T, hereafter CH848 10.17DT) which allowed DH270 UCA to bind with high affinity (10). To prevent antibody responses targeting CH848 strain-specific glycan holes, an enhanced (e) version of CH848 10.17DT (CH848 10.17DTe) was engineered by introducing mutations D230N, H289N, and P291S to occupy these glycan holes on the CH848 Env (16). We have also recently described a CH848 Env with the V1 loop glycans restored (CH848 10.17) as the first boosting immunogen following priming by CH848 10.17DT for the stimulation of the DH270 bnAb B cell lineage (19). In DH270 UCA V_H and V_L heterozygous knock-in (V_H^+/−, V_L^+/−) mice (termed DH270 UCA KI mice), we have tested protein Env–NP of CH848 10.17DTe priming followed by CH848 10.17 boosting and demonstrated the induction of key DH270 bnAb B cell lineage improbable mutations (10, 19).

Thus, in this study, we aimed to evaluate the immunogenicity of this same Env–NP prime encoded by modified mRNA. We vaccinated rhesus macaques intramuscularly following a prime–boost regimen. One group of rhesus macaques (n = 4) received four immunizations of priming mRNA (CH848 10.17DTe Env–NP, 50 µg for each immunization) followed by three immunizations of boosting mRNA (CH848 10.17e Env–NP, 50 µg for each immunization). As controls, another group of rhesus macaques (n = 4) received the same immunogens in the form of purified recombinant proteins (100 µg each immunization) following the same immunization regimen, with 5 µg of the TLR7/8 agonist 3M-052 in an aqueous formulation admixed with 500 µg of alum in PBS as adjuvant (Fig. 1A) (20, 21).

Fig 1.

Protein Env–NP induced higher binding and neutralizing antibody responses than mRNA–LNP in rhesus macaques. (A) Protein and mRNA–LNP prime-boost regimens in rhesus macaques (n = 4). One group of rhesus macaques was primed intramuscularly with 50 µg per dose of CH848 10.17DTe Env–NP mRNA and were boosted with 50 µg per dose of CH848 10.17e Env–NP mRNA at indicated time points. The other group of rhesus macaques was immunized with 100 µg per dose of the same immunogens in the form of purified recombinant protein adjuvanted with 3M-052 and alum following the same regimen. (B) Quantification of serum-binding IgG levels to CH848 10.17 gp120, CH848 10.17DT SOSIP, CH848 10.17 SOSIP, H. pylori ferritin, and human ferritin by ELISA. Each thin line represents an individual rhesus macaque; thick lines represent means in each group. Mann-Whitney U test was performed on the Log-transformed area-under-curve of each group. (C) Quantification of serum autologous tier-2 neutralizing antibody titers 2 weeks after prime or boost by TZM-bl pseudovirus neutralization assay. Data shown are 50% inhibition dilutions (ID50s). Murine leukemia virus Senecavirus A (MLV-SVA) is a negative control virus. LLOD, lower limit of detection; GMT, geometric mean titer of ID50s. (D) Lack of V3-glycan bnAb epitope N332-dependent neutralization activities in mRNA–LNP- and recombinant protein-vaccinated rhesus macaques. Data are from one experiment. *P < 0.05; ns, not significant; Mann-Whitney U test.

We found protein Env–NP vaccinations induced higher binding antibody responses to CH848 10.17DT SOSIP (P = 0.0286, Mann-Whitney U test) and CH848 10.17 gp120 than mRNA encoding Env–NP (P = 0.0286). Antibody levels in both groups did not increase after boosting ×3 (Fig. 1B).

Two weeks after priming ×4, geometric mean titer (GMT) of 50% inhibition dilution (ID50) of neutralizing antibodies against autologous tier-2 (vaccine-matched, difficult-to-neutralize) virus CH848 10.17DT in protein-vaccinated rhesus macaques (GMT = 16893) was 27-fold higher than that in mRNA-vaccinated rhesus macaques (GMT = 615, P = 0.0286, Mann-Whitney U test) (Fig. 1C). GMTs of neutralizing antibodies against CH848 10.17DT decreased after boosting ×3 in both protein- (2.1-fold) and mRNA-vaccinated (22-fold) rhesus macaques; titers induced by protein were still higher than with mRNA (284-fold), with GMTs at 7,963 vs 28, respectively (P = 0.0286, Mann-Whitney U test) (Fig. 1C). Similar trends were observed for glycan-holes-filled virus CH848 10.17DT D230N H289N P291S (Fig. 1C). The decreased titers after boost could be explained by circulating antibody suppression due to repeated immunization (22). Epitope mapping of neutralizing antibody responses showed that most Env-specific antibodies targeted the glycan-deficient sites on the V1 loop in CH848 10.17DT, instead of the N332 site at the base of the V3 loop where V3-glycan bnAbs bind (P = ns, Mann-Whitney U test) (Fig. 1B through D).

We next asked if mRNA-encoded Env SOSIP trimer could be a better boosting immunogen than Env–NP. We primed rhesus macaques following the same regimen and boosted them with an mRNA–LNP encoding a CH848 10.17 SOSIP trimer without ferritin (Fig. 2A). Binding and neutralizing antibody responses followed the same pattern as seen in Fig. 1C of those in Env–NP-boosted rhesus macaques, showing reduced antibody titers after boosts (P = ns) (Fig. 2B and C). Similar to the observation in Fig. 1D, neutralization activities were not N332-dependent (Fig. 2D).

Fig 2.

CH848 10.17e SOSIP mRNA boost induced similar responses to CH848 10.17e Env–NP boost. (A) Protein and mRNA–LNP prime-boost regimens in rhesus macaques (n = 4 each group). One group of rhesus macaques was primed with 50 µg per dose of CH848 10.17DTe Env–NP mRNA–LNP at indicated time points; the other group of rhesus macaques was primed with 100 µg per dose of the same immunogen in the form of purified recombinant protein adjuvanted with 3M-052 and alum. Both groups of rhesus macaques were boosted with CH848 10.17e SOSIP trimer mRNA–LNP. (B) Serum-binding IgG levels to CH848 10.17 gp120, CH848 10.17DT SOSIP, CH848 10.17 SOSIP, H. pylori ferritin, and human ferritin measured in ELISA. Each thin line represents an individual rhesus macaque; thick lines represent the means in each group. (C) Serum autologous tier-2 neutralizing antibody titers 2 weeks after prime or boost measured in TZM-bl pseudovirus neutralization assay. Murine leukemia virus Senecavirus A (MLV-SVA) is negative control. Data shown are ID50s. LLOD, lower limit of detection; GMT, geometric mean titer. (D) Lack of V3-glycan N332-dependent neutralization activities in mRNA–LNP- and recombinant protein-vaccinated rhesus macaques. (E) ELISA for serum IgG or IgM for anti-dsRNA antibodies. No anti-dsRNA antibodies were induced by mRNA–LNP or by protein. (F) ELISA for serum IgG or IgM for anti-PEG antibodies. Low levels of anti-PEG antibodies were induced by mRNA–LNP but not by protein. *P < 0.05; area-under-curve (AUC) of each group were compared in Mann-Whitney U test. Data are from one experiment.

We asked if antibodies against components of mRNA–LNP, such as poly-ethylene glycol (PEG) or RNA, were induced in rhesus macaques that could have inhibited the response to boosting mRNA–LNP. To address this question, we quantified serum anti-PEG and anti-RNA IgG and IgM levels by ELISA. No anti-RNA IgG or IgM was detected in any rhesus macaques using the immunizing mRNA as the ELISA antigen (Fig. 2E). Low levels of anti-PEG IgG and IgM were detected in mRNA–LNP-vaccinated rhesus macaques, but not in protein-vaccinated control rhesus macaques (Fig. 2F). Thus, it is unlikely that decreased immunogenicity of mRNA–LNP was mediated by antibodies targeting PEG or RNA thus inactivating mRNA–LNP.

Modified mRNA that encoded the CH848 10.17DTe Env–NP showed a single peak at expected size in capillary gel electrophoresis (Fig. 3A) and contained minimal levels of double-stranded RNA (dsRNA) (Fig. 3B). Modified mRNA transient transfection in vitro in 293F cells yielded Env forms that showed antigenicity for binding to DH270 UCA and bnAbs 2G12, N6, and trimer-specific bnAbs PGT145 and V3–glycan bnAb PGT125 and showed low binding to V3-loop non-neutralizing antibody (nnAb) 19b and no binding to CCR5-binding-site nnAb 17b (Fig. 3C).

Fig 3.

Characterization of modified mRNA-encoded CH848 10.17DTe Env–NPs. (A) Capillary gel electrophoresis of CH848 10.17DTe Env–NP mRNA showed a single peak at expected size. (B) Minimal levels of dsRNAs were detected in CH848 10.17DTe Env–NP mRNA materials. Exposure time was 6 min. (C) Antigenicity of modified mRNA-encoded CH848 10.17DT Env–NPs that were expressed in 293 F cells and purified by bnAb PGT145. Data are from one experiment.

In summary, modified mRNA encoding Env–NP had the correct RNA size and contained minimal levels of dsRNA. However, mRNA-induced binding and neutralizing antibody responses in rhesus macaques were not optimal for further use as an immunogen.

CH848 10.17DTe Env–NP mRNA was immunogenic in DH270 UCA KI mice only at high immunogen doses

The results in rhesus macaques suggested that CH848 10.17DTe Env–NP mRNA did not express sufficient well-folded Env–NP protein in vivo. We previously showed that this mRNA is immunogenic in DH270 UCA KI mice at 20 µg dose (16). Thus, we studied the immunogenicity of this mRNA–LNP in DH270 UCA KI mice in a wide dose range. In DH270 UCA KI mice, 5%–10% of naive B cells express DH270 UCA BCR (10). We vaccinated DH270 UCA KI mice with 20, 1, 0.5, or 0.25 µg of CH848 10.17DTe Env–NP mRNA three times with 2 weeks interval and quantified serum binding and neutralizing antibody responses 1 week after the third dose (Fig. 4A). As we previously described (16), 20 µg of mRNA induced strong antigen-specific binding (Fig. 4B) and autologous tier-2 neutralizing antibody responses to CH848 10.17DT (GMT: 69,700) (Fig. 4C). In contrast, 1 µg of mRNA induced significantly lower binding (3.8-fold lower logAUC, P = 0.0079, Mann-Whitney U test) and neutralizing antibody responses (GMT: 1081, 65-fold lower, P = 0.0079, Mann-Whitney U test). Binding and neutralizing antibody responses in mice vaccinated with 0.5 or 0.25 µg of mRNA were low or undetectable (Fig. 4B and C). These data supported the notion that CH848 10.17DTe Env–NP mRNA did not express protein sufficiently well in vivo for optimal immunogenicity and, thus, was only optimally immunogenic at 20 µg in DH270 UCA KI mice.

Fig 4.

CH848 10.17DTe Env–NP mRNA–LNP was immunogenic at high dose only in DH270 UCA KI mice. (A) Vaccination regimen. DH270 UCA KI mice were vaccinated with 20, 1, 0.5, or 0.25 µg CH848 10.17DTe Env–NP mRNA–LNP intramuscularly every 2 weeks for three times. (B) Quantification of serum-binding IgG levels by ELISA 1 week after the third vaccination. (C) Quantification of serum autologous tier-2 neutralizing antibody titers 1 week after the third vaccination by TZM-bl pseudovirus neutralization assay. Data shown are ID50s. Horizontal bar: GMT of ID50s. * P < 0.05; ** P < 0.01; ns, not significant; Mann-Whitney U test.

To evaluate this hypothesis, we quantitated the expression of well-folded Env–NP by CH848 10.17DTe Env–NP mRNA in vitro using negative-stain electron microscopy (NSEM). Env–NP was purified from mRNA-transfected ExpiCHO cell supernatant using bnAb 2G12-conjugated agarose beads for NSEM analyses. 2D-classification of NSEM images demonstrated the existence of forms of well-folded Env–NP. However, well-folded Env–NPs consisted of only 0.7% (1,096 out of 157,663) of total particles observed (Fig. 5). We previously reported that 21% of total particles were well-folded Env–NP from this same mRNA design when expressed in 293F cell line and purified by trimer-specific bnAb PGT145 (16). ExpiCHO cells typically yield more proteins but have more host proteins in supernatants. Additionally, 2G12 purification is less Env-specific than PGT145, leading to more host protein contaminations in the purified sample, especially when Env expression is low. The data presented here and the data we published before all supported the notion that mRNA yielded well-folded NPs at very low levels and resulted in insufficient immunogenicity in rhesus macaques or in DH270 UCA KI mice for further studies.

Fig 5.

Quantification of CH848 10.17DTe Env–NPs expressed from mRNA in ExpiCHO cells. Modified mRNA-encoded CH848 10.17DTe Env–NPs were purified from transfected ExpiCHO cell supernantant using bnAb 2G12 affinity column. Well-folded Envs show the triangular forms attached to ferritin (central circle) in the 2D classification images above. NSEM analysis and quantification of mRNA-expressed Env–NPs showed that only 0.7% of observed particles were well-folded Env–NPs. Data are representative of two experiments.

Based on these observations, we asked if we could induce better humoral responses in rhesus macaques by vaccinating with higher doses of the same lot of CH848 10.17DTe Env–NP mRNA. We next vaccinated rhesus macaques with 50, 70, and 200 µg of CH848 10.17DTe Env–NP mRNA–LNP or 50 µg of purified recombinant protein Env–NP using LNP as adjuvant (Fig. 6A). This allowed us to directly compare modified mRNA-produced immunogens to purified recombinant protein immunogens using the same adjuvant (23). We found that antigen-specific-binding IgG levels to CH848 10.17DT SOSIP in rhesus macaques that received 200 µg mRNA were comparable to titers in rhesus macaques that received 50 µg mRNA and were lower than those in rhesus macaques vaccinated with Env–NP proteins (P = 0.0286, Mann-Whitney U test) (Fig. 6B). Binding to CH848 10.17 SOSIP was low, and serum binding was not affected by N332T mutation (Fig. 6B). Similar trends were observed for serum autologous tier-2 neutralizing antibody titers (Fig. 6D). Thus, increasing the dose of CH848 10.17DTe Env–NP mRNA vaccinations up to 200 µg (4-fold) did not improve immunogenicity in rhesus macaques.

Fig 6.

CH848 10.17DTe Env–NP mRNA–LNP lacked significant dose-dependent response in rhesus macaques. (A) Vaccination regimen. Rhesus macaques were vaccinated intramuscularly with 50, 70, or 200 µg per dose of mRNA–LNP encoding CH848 10.17DTe Env–NP or 100 µg per dose of the same immunogen as protein adjuvanted in LNP at indicated time points. (B) Quantification of serum-binding IgG levels by ELISA 2 weeks after the third vaccination. (C) Quantification of serum autologous tier-2 neutralizing antibody titers 2 weeks after the third vaccination by TZM-bl pseudovirus neutralization assay. Data shown are ID50s. Horizontal bar: GMT of ID50s. (D) Lack of V3-glycan N332-dependent neutralizing activities in vaccinated rhesus macaques. Data are from one experiment. *P < 0.05; Mann-Whitney U test.

mRNA-encoded CH848 10.17DT gp160 was more immunogenic than mRNA-encoded Env–NP

We previously described a CH848 10.17DT immunogen in the form of full-length, cell membrane-bound Env gp160 encoded by mRNA (16). We sought to further assess the immunogenicity of the CH848 10.17DT gp160 mRNA at low doses (e.g., 1 µg) in DH270 UCA KI mice and compare with Env–NP mRNA. We vaccinated DH270 UCA KI mice with 20 or 1 µg of mRNA–LNP encoding gp160 and compared serum binding and neutralizing antibody titers with those induced by Env–NP mRNA–LNP.

Both gp160 or Env–NP mRNA induced comparable serum-binding IgGs to CH848 10.17DT SOSIP at 20 µg doses (Mann-Whitney U test, P = ns). However, gp160 mRNA induced higher serum IgG at 1 µg dose compared with Env–NP mRNA (Mann-Whitney U test, P = 0.0043).

When serum-binding IgG levels to CH848 10.17 SOSIP were compared, gp160 mRNA induced higher binding responses at both 20 µg and 1 µg doses (Mann-Whitney U test, P = 0.0043, Fig. 7A). Additionally, 1 µg gp160 mRNA also induced higher autologous tier-2 neutralization titers against CH848 10.17DT than did Env–NP mRNA (P = 0.0043) (Fig. 7B).

Fig 7.

mRNA–LNP-encoded CH848 10.17DT gp160 was more immunogenic than Env–NP. (A) Comparison of CH848 10.17DT SOSIP- and 10.17 SOSIP-specific serum IgG levels induced by gp160 mRNA vs Env–NP mRNA. Sera from 1 week after the third vaccination were tested. (B) Comparison of autologous tier-2 neutralizing antibody titers against CH848 10.17DT pseudovirus induced by gp160 mRNA vs Env–NP mRNA. (C and D) NGS analysis of frequencies of key improbable mutations in DH270 antibody gene induced by 20 or 1 µg of Env–NP mRNA (C) or gp160 mRNA (D). Data shown are percentage of DH270 antibody genes with a specific mutation. Horizontal bar: median. Dotted line: expected mutation frequencies without antigen selection. (E) Standard curve for measuring Env expression by Western blot. Purified recombinant CH848 10.17DT Env–NP was ran in reduced SDS-page electrophoresis and blotted with anti-Env. Gray dotted line: 95% CI. (F) Quantification of Env–NP and gp160 from ExpiCHO cell mRNA transfection by Western blot. Values are interpolated from the standard curve in E. (G) Representative confocal images of mRNA–LNP-encoded gp160 and Env–NP expression in vivo. DH270 UCA KI mice were vaccinated intramuscularly with 20 µg of mRNA–LNP encoding gp160 or Env–NP; draining inguinal lymph nodes were collected 24 h after vaccination for IHC analysis. Blue, IgD; green, CD169; red, antigen stained by bnAb DH270.6. Images of whole LN sections are shown on the left; regions of interest (ROIs) are shown on the right. Scale bar in whole section image: 200 µm; in ROIs: 50 µm. (H) Percent area of lymph node sections covered by gp160 or Env–NP staining quantified by Fiji. Data shown are from 17 LN cryosections from three mice for each group. (I) Mean intensity of gp160 or Env–NP staining per µm² tissue quantified by Fiji. Data shown are from 17 LN cryosections from three mice for each group. *P < 0.05; **P < 0.01; ****P < 0.0001; ns, not significant; Mann-Whitney U test.

Quantification of frequencies of DH270 improbable mutations by NGS induced by either mRNA encoding Env–NP or by mRNA encoding gp160 showed that at 20 µg Env–NP mRNA induced lower light chain improbable mutation S27Y (P = 0.0476, Mann-Whitney U test) and trended lower frequencies of L48Y mutation (Fig. 7C), contrasting the results seen with gp160 mRNA where there was no difference between 20 and 1 µg for any of the four improbable mutations R98T, S27Y, or L48Y (Fig. 7D). Thus, in vivo-expressed mRNA-encoded gp160 was more efficient in selecting improbable mutations than mRNA-encoded Env–NP.

Multiple mechanisms could contribute to better immunogenicity of gp160 mRNA. One of the most obvious reasons was that gp160 mRNA expressed more antigen in vivo than Env–NP mRNA. Western blot with mRNA-transfected ExpiCHO cells showed higher protein levels of mRNA-expressed gp160 than Env in vitro (Fig. 7E and F). Additionally, we performed immunohistochemistry (IHC) to directly assess mRNA-expressed antigen in vivo in draining inguinal lymph nodes (iLNs). We vaccinated DH270 UCA KI mice intramuscularly with 20 µg mRNA–LNP encoding gp160 or Env–NP. iLNs were collected for IHC 24 h post-vaccination.

BnAb DH270.6 was used in IHC to stain for mRNA-expressed gp160 or Env–NP. Strong gp160 staining was detected in iLNs 24 h after vaccination. Most gp160s were expressed in the paracortex or medulla of iLN. Interestingly, a proportion of gp160s were colocalized with CD169⁺ subcapsular sinus macrophages. In contrast, mRNA-encoded Env–NP was near undetectable in iLNs at 24 h after one vaccination (Fig. 7G). Quantification of gp160 or Env–NP staining demonstrated that gp160 covered larger areas (Fig. 7H) and had higher mean intensity per µm² tissue (Fig. 7I) on iLN sections than Env–NP. These results demonstrated that gp160 expressed more antigen than did Env–NP mRNA in vivo mRNA 24 h after one vaccination.

DISCUSSION

In this paper, we have demonstrated the importance of in vitro and in vivo expression studies to determine the optimal modes of modified mRNA designs for HIV-1 envelope immunogens.

We previously demonstrated that modified mRNA vaccines encoding both Env–NP and gp160 were strongly immunogenic in DH270 UCA KI mice at a 20 µg dose and induced autologous tier-2 neutralizing antibodies and selected for key improbable mutations (16). In this study, we demonstrated that particular Env–ferritin NP design had low immunogenicity in rhesus macaques that was found only after mRNA–LNP production for a clinical trial. We found Env–NP mRNA was less immunogenic than its recombinant protein counterpart in rhesus macaques and demonstrated that this particular Env–NP mRNA design lacked sufficient correct folding and expression of well-folded Env–NP in vitro and in vivo for testing clinically. We vaccinated DH270 UCA KI mice with doses of mRNA encoding Env–NP, ranging from 20 to 0.25 µg and found it was less immunogenic at a 1 µg dose in mice as was a 20 µg of mRNA. Driven by the observation that the two COVID-19 mRNA vaccines use SARS-CoV-2 S protein in the full-length membrane-bound form, we compared the immunogenicity of mRNA-encoded gp160 with mRNA-encoded Env–NP and found that mRNA-encoded gp160 was more immunogenic at 1 µg dose. Finally, Western blot with mRNA-transfected cell lysates and supernantants showed that mRNA-encoded gp160, indeed, yielded more Env than mRNA encoding Env-ferritin NP (Fig. 7E and F). Direct in vivo analysis by immunohistochemistry also demonstrated minimal expression of Env–NP by mRNA as measured by bnAb DH270.6 that binds to well-folded V3-glycan bnAb epitope in draining LN 24 h after one immunization.

There are several key points based on the observations from this study. First, for HIV-1 vaccine development, mRNA expression of a stable, well-folded Env form is critical to minimize the expression of nnAb-inducing Env forms (1, 16). Second, the data presented in this paper is a cautionary tale of evolved appreciation of emphasizing the importance of taking the percent well-folded Env–NPs among total protein forms that are expressed as important. For in vivo studies, we found that mRNA should be sufficiently immunogenic at low doses such as 1 µg dose in mouse. Quantitation of the percent of correct Env forms by either NSEM for Env–NP or flow cytometry for gp160 in vitro and by immunohistochemistry is important for the evaluation of in vivo mRNA expression. We hypothesize that the inadequate expression of Env-ferritin NPs is due to insufficient Env stabilization. Work is ongoing to design new stabilization strategies for mRNA-encoded nanoparticles that have acceptable in vitro expression profiles. The general design of gp160 discussed in this study has now been successfully manufactured and will be tested in humans in a Phase I HIV-1 vaccine clinical trial.

In summary, mRNA technology can be a powerful vaccine platform for many vaccines. For HIV-1 vaccine development, ensuring optimal in vitro and in vivo expression of the desired protein form is essential for ensuring immunogenicity potency for the desired antibody specificities prior to consideration for immunogens being used for human clinical trials.

MATERIALS AND METHODS

Animals

All rhesus macaques were housed at Bioqual, Inc, Rockville, MD. All rhesus macaque studies were conducted in compliance with local, state, and federal regulations and were approved by the Bioqual Institutional Animal Care and Use Committee (IACUC) per a memorandum of understanding with the Duke University IACUC.

DH270 UCA V_H and V_L heterozygous knock-in (V_H^+/−, V_L^+/−) mouse (hereafter termed DH270 UCA KI mouse) was described previously (10). All mice were cared for in a facility accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International (AAALAC) at Duke University Medical Center. All study protocols and all veterinarian procedures were approved by the Duke University IACUC.

Vaccination

Rhesus macaques were primed and boosted intramuscularly in both quadriceps with 50 µg mRNA–LNP or recombinant protein adjuvanted with 5 µg of 3M-052 aqueous formulation admixed with 500 µg of alum in PBS. In the dose study, rhesus macaques were primed intramuscularly in both quadriceps with 50, 100, or 200 µg mRNA–LNP or 100 µg recombinant protein adjuvanted with LNP, which allowed a direct comparison of mRNA–LNP and recombinant protein with the same adjuvant (23).

For mRNA–LNP immunizations in mice, DH270 UCA KI mice that were 8–12 weeks old were used. Mice were randomly assigned into groups, and each group had at least one female mouse. Mice were vaccinated with mRNA–LNP at various doses intramuscularly. Vaccinations were 2 weeks apart from each other. Tail vein blood collection was performed 1 week after each vaccination. Necropsy was performed 1 week after the third or fourth vaccination and blood, spleen, and inguinal and axillary lymph nodes were collected.

Cell lines

Freestyle 293F cell line was purchased from Thermo Fisher (Thermo Fisher, Cat# R79007). Cells were cultured in Freestyle 293 Expression Medium (Thermo Fisher, Cat# 12338026) at 37°C with 8% CO₂; cell density was maintained between 0.3 × 10⁶ cells/mL to 3 × 10⁶ cells/mL. Cells were diluted to 1.25 × 10⁶ cells/mL and cultured for 2–3 h before transient transfection of mRNA. A new vial of cells was thawed before cell reached passage 30. ExpiCHO cell line was purchased from Thermo Fisher (Thermo Fisher, Cat# A29127). Cells were cultured in ExpiCHO Expression Medium (Thermo Fisher, Cat# A2910002) at 37°C with 8% CO₂; cell density was maintained between 1 × 10⁶ cells/mL 6 × 10⁶ cells/mL. A new vial of cells were thawed every 3 months. Both cell lines were routinely tested for mycoplasma at Duke Cell Culture Facility, and results were negative.

Production of nucleoside-modified mRNA

Processes for the production and analytical characterization of nucleoside-modified mRNAs have been detailed before (24). Empty LNP was provided by Acuitas Therapeutics and was used as described (25–27). LNP formulation contains ionizable cationic lipid, phosphatidylcholine, cholesterol, and PEG-lipid. Ionizable cationic lipid is proprietary to Acuitas Therapeutics. The cationic lipid and LNP composition are described in US patent US10,221,127.

Modified mRNA transfection

Transient transfection of modified mRNA was performed using TransIT-mRNA Transfection Kit (Mirus, Cat# MIR2250) following manufacturer’s instructions. 0.4 µg/mL cell of HIV-1 Env–NP mRNA and 0.1 µg/mL cell of furin mRNA were transfected. mRNAs were diluted in Freestyle 293 Expression Medium. Then, mRNA boost reagent and TransIT-mRNA transfection reagent were added and gently mixed with mRNA. mRNA transfection mixture was incubated at room temperature for 2–5 min and then added to cells. Transfected cells were cultured for 72 h before harvest.

Modified mRNA-encoded Env–NP purification

Modified mRNA-encoded Env–NP was purified from supernatant of transient transfection in 293F or ExpiCHO cells for ELISA and NSEM. Cells were harvested 3 days after mRNA transfection and were spun down at 3,000 × g for 30 min to remove cells and debris. Supernatant was collected and filtered through a 0.8-µm filter and concentrated about 30× with a Vivaspin 20 100 kDa MWCO sample concentrator (Cytiva, Cat# 28932363). mRNA-expressed Env forms were purified from concentrated supernatant using bnAb PGT145- or 2G12-conjugated agarose beads for NSEM and ELISA.

Western blot (WB)

ExpiCHO cells were transfected with modified mRNA. Three days post transfection cell culture supernatant and cell pellets were collected. Cell pellets were lysed with three sequential freeze thaw cycles of cells in RIPA buffer. Lysates were clarified by centrifugation and collection of the soluble protein in the supernatant. Cell lysates were concentrated 1.4x using a Vivaspin 20 with a 10kD MWCO (Cytiva, Cat#28932360). Cell culture supernatants were concentrated ~50x using a Vivaspin 20 with a 100kD MWCO (Cytiva, Cat# 28932363). Protein concentration was determined for the concentrated cell culture media and lysates using the Quick Start Bradford Protein Assay (Bio-Rad, Cat# 5000204). Five microliters of concentrated cell culture media or cell lysate were loaded onto a NuPAGE Novex 4-12% Bis-Tris Gel (Life Technologies Cat# NP0335BOX). For quantifying protein concentration, purified recombinant HIV-1 envelope ferritin nanoparticles matching those expressed by the mRNA were serially diluted and protein was loaded at concentration between 350 and 5 ng onto the gel to serve as a standard curve. The SDS-PAGE gel was run at 200 V for 52 min and transferred to PVDF membrane by iBlot Gel Transfer Device (Life Technologies Cat# IB1001)and iBlot transfer stack (Life Technologies Cat# IB3010-01). The membrane was blocked with 1x PBS with 1% Casein Blocker (Bio-rad Cat# 161-0783). The membrane was blotted with 1 μg/mL anti-Env antibody 16H3 for 1 h at room temperature. The primary was washed away and the blot was incubated at room temperature for 1 h with secondary antibody anti-mouse IgG (whole molecule) (SIGMA Aldrich Cat#A3565) at a ratio of 1:2,500. The blot was washed with 1x PBS + 0.1% Tween-20, incubated with Western Blue substrate (Promega Cat# S3841), and rinsed with milli-Q water. Blots were visualized with a Bio-Rad Gel Chemi-Doc, and bands were quantified using its densitometry software. Band intensities for the standard curve were plotted in GraphPad Prism v10. A non-linear regression curve was fit to the data and used to interpolate the Env concentration in the cell culture media or cell lysate. The calculated Env concentration was normalized to total protein loaded to give a final concentration as µg of Env per microgram of total protein.

ELISA

Quantification of rhesus macaque and mouse serum binding IgG

Antigens were diluted to 2 µg/mL in 0.1M Sodium bicarbonate buffer and were coated onto 384-well microplates (CoStar, Cat# 3700) overnight at 4°C. To retain HIV-1 Env SOSIP trimers in native-like conformation, bnAb PGT151 was coated on the microplates to capture HIV-1 Env SOSIP trimers. The next day, plates were washed with wash buffer (1× PBS + 0.1% Tween-20) once and were blocked with blocking solution (1× PBS with 40 g/L whey powder, 15% goat serum, 0.5% Tween-20, 0.05% Sodium Azide) for at least 1 h at room temperature. Serum samples were serial-diluted in blocking solution. After blocking, blocking solution was removed and serum samples were added and incubated for 1.5 h at room temperature. Then, microplates were washed 2× in wash buffer. For rhesus samples, Mouse Anti-Monkey IgG-HRP (SouthernBiotech, Cat# 4700-05) was added and incubated for 1 h at room temperature; for mouse samples, Goat Anti-Mouse IgG, Human ads-HRP (SouthernBiotech, Cat# 1030-05) was used. Next, microplates were washed 4× with wash buffer. Next, plates were developed with 20 µL/well SureBlue Reserve TMB 1-Component Microwell Peroxidase Substrate (Seracare, Cat# 5120-0083) for 15 min at room temperature and were stopped with 20 µL 1% HCl solution. Absorbance at 450 nm was measured by SpectraMax Plus 384 microplate/well reader (Molecular Devices). Log-transformed area-under-curve (logAUC) was calculated using Prism (Graphpad) and reported.

Quantification of serum IgG and IgM binding to immunogen RNA in vaccinated rhesus macaques

For RNA binding, 384-well plates were coated with 2 µg/mL poly-lysine overnight at 4°C, washed 1× with PBS and followed with the addition of CH848 10.17DTe Env–mRNA at 2 µg/mL in saline sodium citrate buffer or buffer alone for 1 h. Plates were washed 1×, blocked with assay diluent (1× PBS 2%BSA 0.05% Tween-20) for 1 h at room temperature. Plates were washed 2× and rhesus macaque serum samples in threefold serial dilution starting at 1:100 were then added and incubated for 1 h at room temperature. Goat Anti-Monkey IgG (gamma chain) Antibody HRP conjugated (Rockland, Cat# 617-103-012) or Goat Anti-Monkey IgM (mu chain) Antibody HRP conjugated (Rockland, Cat# 617-103-007) was added and incubated for 1 h. Plates were then washed 4× and developed by 20 µL/well SureBlue Reserve TMB 1-Component Microwell Peroxidase Substrate. Reactions were stopped after 10 min by 20 µL/well 1% HCl. Absorbance at 450 nm was measured by SpectraMax Plus 384 microplate reader (Molecular Devices). Absorbance values from the buffer-only plate were subtracted from the RNA-coated plates.

Quantification of serum IgG and IgM binding to PEG in vaccinated rhesus macaques

Methods for the quantification of anti-PEG IgG and IgM by ELISA were modified from a previously published method (28). 384-well plates were coated with 15 µL/well of 0.5 µg/mL PEG-BSA in 10 mM Tris-HCl (pH 8.0) 150 mM NaCl for 60 min at 37°C. Then, plates were washed with 50 µL wash buffer (0.25% n-Dodecyl-β-d-maltoside [DDM](Chem-Impex International, Cat# 21950) in 1× PBS) and were blocked with 40 µL/well 0.25% DDM in StartingBlock (TBS) Blocking Buffer (Thermo Fisher, Cat# 37542). Rhesus macaque serum samples were serial-diluted in 0.25% DDM in StartingBlock (TBS) Blocking Buffer. After blocking, blocking buffer was removed and 10 µL/well of serum samples were added to the plates and incubated for 60 min at room temperature. Then, plates were washed twice with 50 µL/well of wash buffer. Mouse Anti-Monkey IgG-HRP (SouthernBiotech, Cat# 4700-05) was diluted with 0.25% DDM in StartingBlock (TBS) Blocking Buffer and added to the plates and incubated for 60 min at room temperature. Plates were then washed 4× with 100 µL/well of wash buffer. Plates were then developed with 20 µL/well SureBlue Reserve TMB 1-Component Microwell Peroxidase Substrate (Seracare, Cat# 5120-0083) for 10 min and were stopped with 20 µL/well 1% HCl solution. Absorbance at 450 nm was measured by SpectraMax Plus 384 microplate reader (Molecular Devices). Log-transformed area-under-curve (logAUC) was calculated using Prism (Graphpad) and reported.

Quantification of antibody binding to mRNA-encoded Env–NP from transfected cell lines

Antibodies were diluted to 2 µg/mL in 0.1M sodium bicarbonate buffer and were coated onto 384-well microplates overnight at 4°C. The next day, plates were washed once with wash buffer and were blocked with blocking solution (1× PBS with 40 g/L whey powder, 15% goat serum, 0.5% Tween-20, 0.05% Sodium Azide) for at least 1 h at room temperature. Purified mRNA-encoded Env–NP were serial-diluted in blocking solution. After blocking, blocking solution was removed and samples were added and incubated for 1 h at room temperature. Then, microplates were washed 2× with wash buffer. Serum from a CH848 10.17DT Env vaccinated monkey was then added at 1:1,000 dilution factor and incubated for 1 h at room temperature followed by incubation with Mouse Anti-Monkey IgG-HRP (SouthernBiotech, Cat# 4700-05) for 1 h. Next, microplates were washed 4× with wash buffer. Next, plates were developed with 20 µL per well SureBlue Reserve TMB 1-Component Microwell Peroxidase Substrate (Seracare, Cat# 5120-0083) for 15 min at room temperature and were stopped with 20 µL per well 1% HCl solution. Absorbance at 450 nm were measured by SpectraMax Plus 384 microplate reader (Molecular Devices). Log-transformed area-under-curve (logAUC) was calculated using Prism (Graphpad) and reported.

Negative stain electron microscopy (NSEM)

Sample was diluted to 200 µg/mL with 5 g/dL glycerol in HBS (20 mM HEPES, 150 mM NaCl pH 7.4) buffer containing 8 mM glutaraldehyde. After 5 min incubation, glutaraldehyde was quenched by adding sufficient 1 M Tris stock, pH 7.4, to give 80 mM final Tris concentration and incubated for 5 min. A 5-µl drop of quenched sample was applied to a glow-discharged carbon-coated EM grid for 8-10 s, then blotted and stained with 2 g/dL uranyl formate for 1 min, blotted, and air-dried. Grids were examined on a Philips EM420 electron microscope operating at 120 kV and nominal magnification of 49,000x, and ~70 images were collected on a 76 Mpix CCD camera at a nominal calibration of 2.4 Å/pixel. Images were analyzed by 2D class averages using standard protocols with Relion 3.0.

Next-generation sequencing

Next-generation sequencing was performed with mouse spleen B cells to quantify the frequencies of DH270 bnAb improbable mutations as previously described (8, 16). Expected mutation frequencies of improbable mutations were calculated for each group of mice as median total mutation frequency for each group multiplied by mutation probability for each improbable mutations (29).

Immunohistochemistry

Mouse lymph nodes and spleens were collected and stored in cold RPMI medium with 10% FBS. Tissues were freshly embedded in TFM (General Data Healthcare, Cat# TFM-C) and were flash-frozen on mashed dry ice and were stored at −80°C. Five-micrometer cryosections were sliced using a CryoStar NX50 Cryostat (Thermo Fisher). Cryosections were dried at room temperature and were stored at −80°C. For IHC staining, tissue cryosections were thawed from −80°C, dried at room temperature, and then rehydrated in IHC wash buffer (1× PBS with 0.1% Tween-20) for 15 min. Next, tissue sections were fixed with 200 µL 4% paraformaldehyde (PFA) for 15 min at room temperature in dark; sections were then washed in IHC wash buffer for three times, 5 min for each wash. Next, 200 µL IHC blocking solution [1× PBS with 0.5% BSA, 10% goat serum, 5 µg/mL mouse Fc block (BD, Cat# 553142), 500 µg/mL rat IgG (Sigma, Cat# I8015-50MG)] were added and were incubated for 15 min at room temperature, followed by three washes in wash buffer. Antibodies were diluted in 1× PBS with 5% BSA and 0.1% Tween-20; antibodies used for IHC staining were Brilliant Violet 421 anti-mouse IgD (BioLegend, Cat# 405725, 1:200), Alexa Fluor 594 anti-mouse CD169 (Siglec-1) (BioLegend, Cat# 142416, 1:250), Alexa Fluor 647 DH270.6 (in-house, 5 µg/mL). DH270.6 was labeled using Alexa Fluor 647 Antibody Labeling Kit (Thermo Fisher, Cat# A20186) following manufacturer’s instructions. Two hundred microliters of antibody mix was added to each tissue sections and was incubated at room temperature protected from light for 3 h. After incubation, tissue sections were washed three times in wash buffer, mounted in Fluoromount-G (Thermo Fisher, Cat# 00-4958-02), and stored at 4°C protected from light. Imaging was performed on a Zeiss LSM 880 Confocal Laser Scanning Microscope within 24 h. Data were collected and processed using ZEN software (Zeiss).

Quantification of confocal microscopy images

Quantification of antigen staining signal was performed in Fiji (30). Raw images of antigen channel were imported into Fiji. Images were scaled to micrometers. A binary image was generated by setting threshold of 15–255. Next, a region of interest (ROI) was drawn along the border of the whole LN tissue section. Particle analysis was then performed. Particle size was set at 10–indefinite, circularity was set at 0.0–1.0. Whole tissue area, particle area, mean fluorescent intensity, and percent area covered by antigen were measured. Mean intensity per µm² was calculated by particle area * mean fluorescent intensity/total tissue area.

TZM-bl pseudovirus neutralization assay

For mouse serum samples, pre- and post-vaccinated sera were collected from DH270 UCA KI mice and heat inactivated for 15 min at 56°C and diluted 1:10 with Dulbecco’s modified Eagle’s medium (DMEM) (Thermo Fisher, Cat# 11995-073). The sera and DMEM were initially diluted 1:200 followed by a threefold serial dilution in a 96-well plate. Pseudovirus was added to the sera and incubated at 37°C for 1 h. Following the incubation period, the TZM-bl cells were mixed with DMEM and DEAE-Dextran (Sigma-Aldrich, Cat# D9885) and added to each well. The sera, pseudovirus, and TZM-bl cell mixture was incubated at 37°C for 72 h. After the incubation period, Bright-GloTM reagent (Promega Corporation, Cat# E2650) was placed in each well to lyse the cells. Then, the Bright-Glo reagent and media mixture was placed in a 96-well black plate and inserted into the VICTOR Nivo Luminometer (PerkinElmer, Cat# HH35940301). Raw Relative Luminescence units (RLU) for each well were recorded. Neutralization activity is quantified by a reduction in RLU in the presence of serum compared to wells without serum. Neutralizing antibody titers are shown as the reciprocal serum dilution that inhibits 50% of viral gene expression (ID50 values). The method for rhesus macaque serum samples was the same as for mouse, except that sera were initially diluted 1:20 followed by threefold serial dilution.

Statistical analyses

Mann-Whitney U test was performed at the alpha 0.05 level to compare single endpoint values or area-under-curve (AUC) for multiple endpoints between groups using SAS 9.4 (SAS Institute, Cary, NC). Significant results are indicated in figures and figure legends as follows: *P < 0.05; **P < 0.01; ****P < 0.0001 with no adjustment for multiple comparisons.

DATA AVAILABILITY

Data presented in this paper are available upon request to Barton Haynes at barton.haynes@duke.edu.