Antibody-mediated targeting of SOSIP HIV-1 Env to skin Langerhans cells potentiates humoral responses

Summary

Background

Despite decades of research and multiple phase 2b/3 trials, an effective HIV vaccine remains out of reach. A key obstacle is the need for innovative delivery strategies that can induce potent and durable protective immune responses. Targeting antigen-presenting cells in the skin, specifically Langerhans cells (LCs), offers a promising approach to enhance vaccine efficacy by promoting early and efficient activation of the humoral immune system.

Methods

We developed an Antibody-Mediated targeting Vaccine (AMV) platform that delivers HIV-1 Env trimeric antigens directly to LCs via an anti-Langerin scFv domain. Two constructs were tested: LC³.Env³, carrying a trimeric gp140z Env fusion protein, and LC³.SOSIP, carrying the well-characterised BG505 SOSIP.664 trimer. Immunogenicity of these constructs was evaluated in mice and rabbits without adjuvant, assessing the germinal centre (GC)/Tfh reaction in draining lymph nodes (dLN), Env-IgG production and HIV-1 neutralisation.

Findings

We show that, as compared to an anti-Langerin mAb fused with Env monochains, LC-targeted delivery of Env³ through LC³.Env³ significantly enhanced Tfh and GC B cells (B_GC) formation (p < 0.0001), with an expansion of Env-specific B_GC cells in mice (p < 0.05). Env IgG responses was enhanced with LC³.Env³ as compared to non-targeted trimeric Env (p < 0.05). Results were extended using LC³.SOSIP constructs which maintained the structural integrity of trimeric Env, but also markedly promoted as compared to Env³ and LC³Env³ the presentation of the native-like, prefusion-closed structure that displays all major quaternary-dependent broadly neutralising antibodies (bNAbs) regions. LC³.SOSIP elicited in rabbits robust, Env-specific antibody responses with cross-clade reactivity, including neutralising antibodies (NeutAbs), without the use of adjuvants. Finally, as compared to non-targeted SOSIP, LC³.SOSIP elicited stronger Env IgG responses (p < 0.01), even at low antigen doses, and neutralising activity against to Tier-1A but also autologous BG505 Tier-2 viruses (2 rabbits out of 8).

Interpretation

These results validate LC-targeting as an efficient, adjuvant-free, strategy for inducing potent humoral responses against HIV-1. The AMV platform enables precise delivery of structurally intact Env trimers to skin-resident LCs facilitating early B cell activation and maturation. This approach offers distinct advantages over conventional adjuvanted sub-unit protein vaccines supporting its potential for clinical translation.

Funding

This work was supported by Inserm, the Vaccine Research Institute (VRI), the French National Research Agency (ANR-10-LABX-77, ANR-10-INBS-05-02, ANR-10-EURE-0003), and the European Union’s Horizon 2020 (grant 681032).

Research in context.

Evidence before this study

We conducted a systematic search of major biomedical literature sources, including PubMed, to identify relevant preclinical and clinical studies on HIV vaccination strategies, dendritic cell (DC) targeting, and germinal centre (GC) responses. The search covered the period from database inception to August 1, 2025, without date restrictions, to ensure inclusion of both seminal and recent work in the field, including preprint papers. Search queries combined controlled vocabulary and free-text terms, including “HIV vaccine”, “Env trimer” OR “SOSIP”, “dendritic cells” OR “antigen-presenting cells” OR “Langerhans cells”, “germinal centre” OR “T follicular helper cells” OR “Tfh”, “antibody affinity maturation” OR “neutralising antibodies”, using appropriate Boolean operators. This search identified a broad range of alternative vaccination strategies aimed at overcoming the inherent challenges of inducing robust and durable humoral immunity against HIV. Despite their diversity, these approaches converge on a shared goal: enhancing antigen-specific and epitope-focused GC responses to promote affinity maturation and the generation of long-lived antibody responses. Collectively, these findings underscore the need for continued innovation, particularly through improved immunogen design, such as native-like SOSIP trimers, optimisation of antigen delivery, and refinement of immunisation regimens.

Added value of this study

Targeting vaccine antigens to endogenous antigen-presenting cells (APCs), particularly dendritic cells (DCs), has emerged as a promising strategy to enhance immune priming. By directly engaging DCs, this approach combines antigen delivery with coordinated T and B cell activation, thereby improving the efficiency and quality of vaccine-induced responses. Preclinical studies have demonstrated that DC-targeted vaccines can elicit strong and durable humoral and cellular immunity, and early phase I clinical trials in healthy volunteers, including those evaluating the CD40.HIVRI.Env vaccine, have shown a favourable safety profile. Within this framework, increasing evidence indicates that epidermal Langerhans cells (LC) represent a specialised APC subset with a critical role in promoting T follicular helper (Tfh) cell differentiation and GC-dependent B cell responses. However, whether LC targeting could efficiently deliver structurally complex, native-like Env trimers and improve the magnitude and quality of neutralising antibody responses remained unresolved.

This study demonstrates that antibody-mediated targeting of native-like trimeric HIV-1 Env immunogens to epidermal LCs markedly enhances humoral immune responses. Using trivalent anti-Langerin scFv constructs fused to stabilised Env trimers (LC³.Env³), including BG505 SOSIP antigens (LC³.SOSIP), we show efficient in vivo targeting of LCs, enhanced antigen transport to draining lymph nodes, and robust induction of Tfh cells and GC B cells. Compared with non-targeted or earlier LC-targeting mAb, LC³.Env³ and LC³.SOSIP improved the kinetics, magnitude, and quality of GC responses and elicited higher titres of Env-specific IgG with neutralising activity against Tier-1A and Tier-2 HIV-1 strains. These responses were achieved at low antigen doses and without adjuvant. Structural and antigenic analyses confirmed that LC³.SOSIP proteins preserve a native-like, closed prefusion Env trimer conformation and expose key broadly neutralising antibody (NeutAb) epitopes.

Implications of all the available evidence

Together with prior studies, these findings establish epidermal LCs as a key target for vaccine strategies aiming to induce strong and durable humoral responses. By shaping GC dynamics and promoting Tfh differentiation, LC targeting supports affinity maturation and the generation of high-quality NeutAbs against complex pathogens such as HIV.

Beyond identifying LCs as relevant targets, this work introduces the LC³ platform as a technological advance in antigen delivery. The trivalent anti-Langerin scFv-based construct enables multivalent, high-avidity targeting to LCs while preserving conformational epitopes critical for broadly neutralising antibody (bNAb) recognition. This is particularly important for HIV immunogens like native-like Env trimers (e.g., SOSIP), whose prefusion structure must remain intact. LC³ thus bridges antigen design and delivery, allowing structurally fragile proteins to be efficiently targeted without compromising their integrity. More broadly, LC³ provides a modular and adaptable framework for antigen delivery to skin-resident APCs. Notably, the Langerin-targeting platform induces robust neutralising antibody responses at low antigen doses without requiring adjuvant, reinforcing its potential for safe, cost-effective vaccine development.

Introduction

A protective vaccine against HIV is still lacking despite more than 40 years of intense research. The HIV RV-144 clinical trial, testing a combination of a recombinant canarypox vector and HIV-1 Envelop gp120 protein, demonstrated some promising, albeit modest, protection against HIV acquisition.¹ A correlate of protection linked to the production of anti-Env binding antibodies targeting the V1V2 region from HIV envelope was identified. However, these responses diminished over time, highlighting that the vaccine failed to induce enduring V1V2 antibodies. Since then, these findings were not replicated in various subsequent phase 2b/3 HIV prophylactic studies, emphasising the urgency to adopt a fresh perspective and devise innovative strategies. In response, several alternative vaccination strategies have been explored to overcome the intrinsic difficulty of eliciting robust and persistent humoral immunity against HIV.² These strategies include germline-targeting approaches,³ which are currently under evaluation in experimental Phase I clinical trials (e.g., IAVI C101, assessing the BG505 SOSIP.GT1.1 gp140 immunogen), as well as approaches that have already demonstrated successful priming of naïve B-cell precursors in Phase I trials (e.g., IAVI G001⁴ and G003⁵). Additional avenues include the assessment of novel immunogens (e.g., HVTN 133⁶) or adjuvants designed to enhance germinal centre (GC) responses (e.g., HVTN 137⁷), fractional dosing and slow-release antigen delivery strategies that extend antigen availability within lymphoid tissues,^8,9 and recent technological advances in antigen and adjuvant formulation.¹⁰ Collectively, these approaches converge on a common goal: increasing antigen-specific and epitope-focused GC activity to support affinity maturation and durable antibody responses. It is therefore imperative to explore the development of innovative vaccines by designing new immunogens, altering delivery methods, and reassessing immunisation regimens to create effective vaccination approaches.

Targeting antigens to endogenous antigen presenting cells (APC), including dendritic cells (DC), represents another promising strategy for immune system reprogramming.¹¹ The innovation lies in several crucial aspects. This approach combines antigen delivery and T/B cell activation by directly targeting DC, paving the way for a more efficient immune response. Extensive preclinical data confirm the capacity to elicit strong and durable B and T cell responses, including CD8⁺ T cells, neutralising serum antibodies, ADCC-mediated responses, and specific antiviral control in various models.12, 13, 14, 15 By reducing the required antigen dosage to sub-milligram amounts in human vaccination, production costs in clinical development can be substantially lowered. Experience in manufacturing several GMP clinical batches, including an HIV DC-targeting vaccine through the CD40 receptor (CD40.HIVRI.Env), further supports the feasibility of this approach. Finally, the phase I clinical trials in healthy volunteers demonstrated a favourable safety profile of the CD40.HIVRI.Env vaccine.¹⁶

Activation of the cutaneous immune system promotes APC migration from the skin to the dLN, facilitating antigen transport and immune priming. Novel APC vaccination strategies aim to target specialised dermal or epidermal DC subsets for more efficient vaccine responses.17, 18, 19 Recent studies highlight the critical role of epidermal Langerhans cells (LC) in priming T follicular helper (Tfh) and germinal centre (GC) B cell responses.20, 21, 22 LCs, particularly when delivering foreign antigens through the Langerin receptor, play a dose-dependent role in supporting GC-Tfh formation.²³ In a non-human primate model, targeting the HIV-1 Gag antigen to the Langerin receptor induced Tfh and B cell responses.^24,25 Moreover, targeting human LCs with anti-Langerin mAbs fused with the HIV-1 gp140z (96ZM651 clade C) envelope (LC.Env) triggered autologous naïve CD4⁺ T cell differentiation into Tfh cells and specific IgG production by memory B cells from people living with HIV. The LC.Env prototype vaccine demonstrated significant Env-specific IgG induction without adjuvant in mice; a GC/Tfh reaction in the dLN; and sustained humoral responses independently of dermal cDC1.²⁶ The intradermal route of immunisation yielded the highest Env-specific IgG titres and strongest Tfh responses, underscoring the pivotal role of LC in shaping effective humoral responses.²⁷

In rational vaccine design, especially for challenging NeutAb targets such as HIV-1 gp120 envelope protein of HIV, the strategic delivery of antigen to LCs is therefore highly compelling. The development of native-like soluble forms of the HIV-1 Env trimer ectodomain (e.g., Env SOSIP trimers) has emerged as a promising vaccine platform for stimulating germline precursors of bNAbs-producing B cells.28, 29, 30 Therefore, we surmised that incorporating Env trimers into an LC-targeting delivery system may profoundly reshape B cell composition within the GC, enhancing Tfh support, B_GC cell dynamics and circulating IgG titres. Accordingly, we developed an Antibody Mediated targeting Vaccine (AMV) strategy to enhance HIV Env delivery to LC by fusing a trimeric Env immunogen to three single-chain variable fragments (scFv) derived from the anti-Langerin 4C7 mAb.²⁶ This approach was initially established using the gp140z protype trimer and subsequently validated with closed, stabilised SOSIP antigens. This strategy significantly enhanced anti-Env humoral responses in vivo, improving the kinetics, magnitude, and the quality of Tfh/B_GC cell responses, as well as the levels and the neutralisation capacity of circulating IgG antibodies.

Methods

Experimental models and ethics

C57BL/6J mice (RRID: IMSR_JAX:000664, male and female, 8 weeks old, n = 105) were manually and sequentially allocated to experimental groups as they became available. All samples from different experimental groups were processed simultaneously during technical procedures (tissue collection, staining, flow cytometry acquisition) to avoid batch effects. Formal computer-based randomisation was not employed. Blinding during animal injections and in vivo manipulations was not implemented due to logistical constraints inherent to coordinating live procedures in preclinical mouse studies. Both male and female C57BL/6J mice were used in all experimental groups to avoid sex bias. The study was not designed or powered to detect sex-specific effects, and sex-disaggregated statistical analyses were therefore not performed. Accordingly, data from males and females were pooled for all analyses. All procedures involving animals were conducted in accordance with the ethical standards and guidelines for the care and use of laboratory animals. The study protocol was reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) of Anses/ENVA/UPEC (CEE-016), under approval number [20-043 #25329]. The study conducted at the Mondor Institute of Biomedical Research facilities was authorised by the Ministry of Research, Innovation and Education under registration number 25329-2020051119073072 v4. All efforts were made to minimise animal suffering and to reduce the number of animals used in the experiments. Female New Zealand White rabbits (10–12 weeks old, n = 39) were housed and immunised at Charles River Biologics under approved protocols (APAFIS#39244-2022110910396896 v3). Only female animals were used to reduce biological variability and ensure consistency across cohorts; sex was therefore not included as a variable in the study design or analysis. Independent animal cohorts were used for each experimental endpoint. Schematic representations of animal experiments were created with BioRender and added in each corresponding figure.

Cord blood samples were obtained from the CRB-BSC (Biological Sample Collection, Saint-Louis Hospital, Unit Cellular Therapy, Paris, France), following informed consent from donors and in accordance with authorisation from the French Ministry of Health for the transfer of biological samples (CODECOH No. AC-2024-6640).

Inclusion/exclusion criteria and sample size determination

Pre-specified inclusion criteria consisted of healthy C57BL/6J mice meeting standard housing and experimental conditions. Predefined exclusion criteria included technical failure of vaccination or sample collection, or failure to meet assay quality control standards. No animals, samples, or data points met these exclusion criteria; therefore, no exclusions were made. Sample size was determined prospectively based on the primary endpoint of Env-specific IgG levels measured by Luminex. The calculation was informed by prior data,²⁷ assuming mean values of 184 MFI (SD = 129) in vaccinated mice and 28 MFI (SD = 8) in controls. Using a two-tailed Student’s t-test, a significance level of 5%, and 80% power, the required sample size was estimated at 11 mice per group. Full details of the calculation are provided in the detailed methods.

Immunogens

Targeted vaccine constructs consisted of HIV-1 Env antigens genetically fused to an anti-Langerin mAb to enable specific targeting of LCs.²⁶ Both monomeric and trimeric Env formats were used, including gp140z (96ZM651 clade C) and stabilised BG505 SOSIP.664-derived constructs.^28,29 The anti-Langerin antibody was based on the human IgG4 clone 4C7 (GenBank: JX002669.1 and JX002668.1), which cross-reacts with human and rabbit Langerin and recognises the ectodomain of mouse Langerin. Env antigens were genetically fused to the C-terminus of the IgG4 heavy chain via a FlexV1 linker or in trimeric constructs, to N-terminus of single-chain variable fragments (scFv) via a FlexV1 linker. Expression constructs were generated by cloning synthetic genes into a UCOE® expression vector using standard restriction enzyme-based strategies, followed by transformation, endotoxin-free plasmid amplification, and sequence verification. Trimeric constructs incorporated either a T4 fibritin trimerisation domain or SOSIP.664 stabilisation and included purification/labelling tags (6xHis and AviTag) where appropriate. Additional BG505 SOSIP.664 variants (including in-house designs) were used for both immunisation and serological assays. Non-targeting Env constructs and PBS were used as controls. All recombinant proteins were produced in mammalian expression systems (Expi-CHO-S, RRID:CVCL_5J31; or Expi293 cells, RRID:CVCL_D615), purified by affinity and size-exclusion chromatography (SEC) (ÄKTA Pure, Cytiva; RRID: SCR_023461). For selected constructs, co-expression with furin was performed to ensure proper Env cleavage. Proteins were quality-controlled prior to use by SDS-PAGE/Native PAGE, SEC profiling, endotoxin quantification (<0.5 ng/mg) as previously described.²⁷

Vaccine characterisation

ELISA

Binding of vaccine constructs to human, mouse, or rabbit Langerin ectodomains was assessed by ELISA. Plates were coated with recombinant Langerin (2 μg/mL in carbonate buffer, pH 9.6, overnight at 4 °C), incubated with serially diluted vaccine constructs or antibodies (1 h at RT after blocking with Superblock, Thermo Fisher Scientific), and binding was detected using HRP-conjugated secondary antibodies (e.g., goat anti-mouse IgG-HRP, Jackson ImmunoResearch, RRID: AB_10015289). Absorbance was measured at 450 nm (reference 570 nm) using a Tristar2 plate reader (Berthold Technologies).

In vitro LC binding

Binding of vaccine constructs to human Langerhans-like cells derived from CD34⁺ progenitors was evaluated by flow cytometry using fluorescently labelled or biotinylated vaccine proteins, as previously described.²⁶ CD34-LC phenotyping was performed by flow cytometry using the fixable Aqua Live/Dead marker and the following anti-human antibodies: CD45 (RRID: AB_1645735), CD11c (RRID: AB_11153662), HLA-DR (RRID: AB_1727526), CD1c (RRID: AB_1953227), CD1a (RRID: AB_2561931), and CD207 (RRID: AB_10900993). CD34-LCs were incubated with serial dilutions of vaccine constructs (30 min, 4 °C in PBS/0.5% BSA), and binding was revealed using anti-biotin-PE (RRID: AB_2661377) for trimeric biotinylated constructs or anti-human IgG Fc-AF647 (RRID: AB_2563329) for anti-Langerin mAbs. Data were acquired on a BD LSR II cytometer (RRID: SCR_002159) and analysed using FlowJo v10 (TreeStar).

Surface plasmon resonance (SPR) binding assay

Binding kinetics between Langerin and Env-based vaccine constructs were measured by SPR using Biacore instruments (Biacore 1S+ and Biacore 3000, RRID: SCR_018044), as previously described.³¹ Ligands were captured via anti-His, anti-Env (10–1074), or anti-human IgG Fc antibodies on CM5 sensor chips using amine coupling. Multiple-cycle kinetics experiments were performed using analyte concentrations up to 160 nM, with defined association (∼180 s) and dissociation (∼400–600 s) phases in HBS-EP + buffer. Association and dissociation rates were determined using a 1:1 Langmuir binding model, yielding kon, koff, and KD values, after double-referencing and fitting using Biacore evaluation software.

bNAb binding integrity assay

The structural integrity of Env antigens was assessed by measuring the binding of a panel of bNAbs to vaccine-coupled Luminex® beads, using nonlinear regression to determine binding profiles, as previously described.³² Env proteins (5 μg) were covalently coupled to MagPlex microspheres using EDC/S-NHS chemistry (Bio-Plex Amine Coupling Kit, Bio-Rad), followed by blocking and storage at 4 °C. Binding was measured as MFI across concentration series and analysed using GraphPad Prism (v10.5.0).

Negative staining electron microscopy (EM)

Protein samples were visualised by negative-stain EM. Samples (4 μL at 0.09 mg/mL) were applied to glow-discharged carbon-coated 400-mesh copper grids (20–45 s, 20–25 mA), stained with 1–2% (w/v) Sodium Silico Tungstate for 20–45 s, and air-dried. Imaging was performed on Tecnai T12 or F20 microscopes (FEI) operating at 120–200 kV, with pixel sizes of 2.1–2.8 Å and electron doses of ∼80 e⁻/Ų. Particles were subjected to 2D classification and low-resolution reconstruction using cryoSPARC v3,³³ and maps were visualised in ChimeraX.³⁴

Animal immunisation

Mice were immunised intradermally (i.d.) in the ear with targeted or non-targeted Env vaccines without adjuvant, using prime-boost regimens. Briefly, mice received 10 μL per ear at day 0 and day 21, corresponding to 250 ng or 5 μg of Env formulated as LC³.Env³, LC.Env, or non-targeted Env³, with PBS-injected animals as controls. Rabbits were immunised subcutaneously (s.c.) with Env-based vaccines in the absence or presence of a TLR4 agonist containing liposomal adjuvant (MPLA in a liposomal formulation, referred to as “MPLA”). Rabbits received 20 μg or 4 μg Env-equivalent doses depending on the regimen, administered at days 0, 28, and 42 or at days 0, 42, and 84, with study termination 2 weeks after the last boost, MPLA (30 μg; Polymun Scientific) was used where indicated. Vaccine doses were normalised to Env content. Blood samples were collected longitudinally, and lymphoid tissues were harvested at defined time points for immunological analyses. In mice, sera were collected weekly until day 28 (and up to day 450 in long-term studies), and spleens and auricular dLNs were collected at sacrifice for downstream analyses. For biodistribution studies, fluorescently labelled vaccines (AF647, Thermo Fisher Scientific) equivalent to 5 μg Env were injected i.d., and dLNs were analysed 24 h post-injection.

Cell phenotyping

Single-cell suspensions from spleens and dLN were analysed by flow cytometry. Panels were designed to characterise B_GC cells, Tfh cells, effector and memory T cells, and myeloid cell subsets. Dead cells were excluded using fixable viability dyes, and analyses were performed on live CD45⁺ singlet populations. Spleens were mechanically dissociated and subjected to red blood cell lysis (ACK buffer), whereas dLNs were enzymatically digested using collagenase D (Sigma–Aldrich) and DNase I (Qiagen). Approximately 5 × 10⁶ cells per sample were stained with antibody panels including (i) B_GC cells: CD45 (RRID: AB_2737976), CD3 (17A2), CD45R/B220 (RRID: AB_2738470), CD138 (281-2), IgG (Poly4060), IgD (11-26c.2a), Fas (RRID: AB_2737690), GL7 (RRID: AB_2800677); (ii) T cells: CD45, CD3 (17A2), CD4 (RM4-5, RRID: AB_2728707), CD8α (RRID: AB_1272235), CD62L (RRID: AB_1645210), CD45R/B220 (RRID: AB_2738470), CD44 (RRID: AB_2650923), CD69 (H1.2F3), CD185(CXCR5) (L138D7), CD279(PD-1) (29F.1A12), and FOXP3 (RRID: AB_11218868); and (iii) myeloid populations: CD45, CD11b-BV711 (RRID: AB_2716860), Ly6C (HK1.4), CD11c (RRID: AB_493568), MHC-II (I-1/I-E) (M5/114.15.2), CD8α (RRID: AB_1272235), CD103 (2E7), XCR1 (RRID: AB_2783118), CD207 (929F3.01). Data were acquired on a LSR II flow cytometer and analysed after gating on live CD45⁺ singlets.

Env-specific B cells

Env-specific B cells were identified using biotinylated, deglycosylated trimeric Env probes followed by fluorescent anti-biotin antibodies, in combination with germinal centre markers (as in²⁷). Briefly, lymph node cells were incubated with 3 μg of biotinylated deglycosylated Env³ per 10⁶ cells, followed by staining with 2 anti-biotin antibodies (RRID: AB_2661379; 1D4-C5) and B_GC cell markers prior to fixation.

Luminex® Env-specific IgG assay

Env-specific IgG levels in mouse and rabbit sera were quantified using a bead-based Luminex® assay. Recombinant Env proteins were covalently coupled to magnetic beads and incubated with diluted serum samples. Bound IgG was detected using PE-conjugated secondary antibodies, and results were expressed as median fluorescent intensity (MFI). Env proteins (50 μg) were coupled to MagPlex beads (Bio-Rad) using EDC/S-NHS chemistry, and sera (mouse and rabbit sera were diluted 1:100 and 1:200, respectively) were incubated for 1 h at 750 rpm. Detection was performed using anti-mouse IgG-PE (RRID: AB_11044909) or anti-rabbit IgG-PE (AB_2795937), and acquisition was carried out on a Bio-Plex 200 system (RRID: SCR_018026).

IFN-γ ELISpot

Env-specific T cell responses were quantified by IFN-γ ELISpot using splenocytes or lymph node cells stimulated with overlapping Env peptide pools. Cells (2.5 × 10⁵ per well) were stimulated for 16 h with 1 μg/mL peptide pools, and IFN-γ secretion was detected using anti-IFN-γ antibody (RRID: AB_2280104) followed by streptavidin-ALP and BCIP/NBT substrate. Spot-forming units were enumerated after enzymatic detection using an ELISpot reader (AID Diagnostika).

Neutralisation assays

Rabbit sera were evaluated for HIV-1 neutralising activity using TZM-bl assays against a panel of pseudoviruses. Neutralisation titres were defined as the reciprocal serum dilution achieving 50% inhibition of viral infection (ID₅₀). TZM-bl cells (RRID: CVCL_B478) were infected with pseudoviruses generated in 293 T cells using Env expression plasmids (MW965.26, SF162.NW, 96ZM651, BG505) and pSG3Δenv backbone. Neutralisation was quantified based on reduction in relative luminescence units (RLU), and ID₅₀ values were calculated.

Histology and immunofluorescence (IF)

Draining lymph nodes were cryosectioned (6 μm; cryostat, RRID: SCR_018453) and analysed by haematoxylin-eosin staining (HES) or IF to assess lymphoid architecture and GC formation. For IF, sections were fixed in cold acetone, blocked, and stained with antibodies including CD20 (RRID: AB_3678438) revealed with anti-rabbit IgG (RRID: AB_2722519), CD3 (RRID: AB_2563427), CD21/35 (RRID: AB_2632698), IgD (RRID: AB_10598660) and Peanut Agglutinin (PNA)-FITC (Eurobio Scientific, France). Whole slide images of HES were acquired by a scanner (RRID: SCR_027284) at magnification 20×. IF Images were acquired on a widefield fluorescent microscope (RRID: SCR_024706) with dry EC.Plan-Neofluar 20× or 40× objectives, and analysed using ImageJ. Confocal imaging was performed on a Nikon Ax.R system at the ICM Quant Core Facility (RRID: SCR_026393).

Quantification and statistical analysis

Normality of the data was assessed using the Anderson-Darling, D’Agostino and Pearson, and Shapiro–Wilk tests. Depending on distribution, statistical analyses were conducted using one-way ANOVA with Tukey’s post hoc test for normally distributed data or the Kruskal–Wallis non-parametric test with Dunn’s post hoc test for non-normally distributed data. For two-group comparisons, non-parametric Mann–Whitney tests were applied when normality was not observed. For multiple comparisons in mixed-effects analyses, Šidák’s or Dunnett’s corrections were applied as appropriate. A significance threshold of p < 0.05 was used (∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001; ns, non-significant). Mean and standard error of the mean (SEM) are plotted depending on the scale in all figures. Bars represent the medians, while whiskers indicate the lowest and highest values in the data set. Results are presented as mean ± SEM or median with range (minimum–maximum), as appropriate to the data distribution. All over, the statistical analysis application (Graph-Pad Prism Software, version 11) was utilised.

Detailed methodologies for quality controls of vaccines, and in vitro and in vivo experiments can be found in the Detailed Methods. All unique reagents generated in this study are listed in the Supplementary Resources Table.

Role of funders

The funding sources were not involved in the study design, data acquisition, data analysis, data interpretation, or writing of the manuscript.

Results

Development and design of a trimeric LC³.Env³ vaccine targeting the Langerin receptor

We produced the 4C7 mAb targeting human, mouse and rabbit Langerin receptors^22,26 (Supplementary Fig. S1A). We previously reported that immunisation with the anti-Langerin 4C7 mAb fused to HIV-1 envelope gp140z from clade C (96ZM651) (LC.Env), without adjuvant, induced Env-specific serum IgG responses in mice and demonstrated a prominent role of skin LC in the Tfh/B_GC cell responses.²⁶ In order to improve the immunogenicity of the vaccine, we designed a new anti-LC HIV construct where the two monomeric Env chains fused at the C terminal region of the two Heavy Chains (HC) of the 4C7 mAb were replaced by three Env gp140 arranged as a trimer using the T4 fibritin trimerisation domain.³⁵ To overcome the rigid structure of the dimeric Heavy chain (HC) constant domain of the 4C7 mAb, we coupled the Env trimer with three scFv of the 4C7 variable domain through a flexible linker (Fig. 1A). The resulting vaccine, designed thereafter LC³.Env³, was produced in stably transfected CHO cells, as well as the non-targeting Env trimer (Env³) used as a control. Quality assessments of the vaccine batches using SDS-PAGE revealed the glycosylated Env monochains under reduced conditions, while the trimeric form of the LC³.Env³ vaccine and non-targeting Env³ were observed without reduction (Fig. 1B). SEC further confirmed the production of a single trimerised product, with no detectable monomeric chains, aggregates, or degraded products (Fig. 1C and Supplementary Fig. S8A). Production of the LC.Env vaccine and the non-targeting IgG4.Env was also carried out in CHO cells, as previously documented.²⁶ LC³.Env³ and LC.Env bound to the human Langerin receptor with equivalent affinity as shown by ELISA (1.42 ± 0.11 nM versus 0.59 ± 0.01 nM, Fig. 1D), and by FACS using human in vitro derived CD34-LC²⁶ (Fig. 1E and Supplementary Fig. S2), indicating that the conversion of the variable domain to three scFv fragments did not affect targeting to the Langerin receptor. In contrast the binding of non-targeting hIgG4.Env and Env³ constructs serving as controls was barely detectable. Surface plasmon resonance (SPR) was employed to compare the binding affinities of LC.Env, LC³.Env³, and Env³ to human, mouse, and rabbit Langerin receptors (Fig. 1F). Vaccine constructs were captured using immobilised anti-Env 10–1074 mAb, and Langerin receptors at the indicated concentrations were used as analytes (Supplementary Fig. S1B and C). Robust and specific binding of both LC.Env and LC³.Env³ to the Langerin from all three species was observed, whereas Env³ did not show specific interactions, even at high concentration (160 nM). The differential binding between Env³ and LC³.Env³ was further validated using alternative assay configurations, either by capturing vaccine constructs with an anti-His mAb or by capturing Langerin with an anti-Fc mAb and using the vaccines as analytes (Supplementary Fig. S1D and E). Kinetic analyses were subsequently performed by injecting increasing concentrations of Langerin receptors over 10-1074-immobilised LC.Env and LC³.Env³, yielding characteristic sensorgrams (Fig. 1G). Binding affinities of LC.Env were comparable across species (KD = 0.7 ± 0.15 nM), as were those of LC³.Env³ (KD = 3.5 ± 1.5 nM) (Fig. 1H, left panel). For both vaccine constructs, increased association and dissociation rates were observed with rabbit Langerin (Fig. 1H, middle and right panels). Overall, SPR analyses confirmed high-affinity interactions between the scFv domains of LC³.Env³ and all Langerin receptors in the nanomolar range, while these affinities were lower (0.7 log₁₀) than those observed for the LC.Env original mAb.

Fig. 1.

Production and quality control of the LC³.Env³ vaccine. (A) Schematic overview of the cloning strategy used to generate the trimeric LC³.Env³ construct. The diagram illustrates the fusion of a fibritin-stabilised HIV-1 Env gp140z trimer (Env³) to the single-chain variable fragment (scFv) domain of the anti-Langerin antibody LC.Env. Additional tags and linkers (FlexV1) are indicated. (B) The purified LC³.Env³ and Env³ vaccine were subjected to SDS-PAGE and stained with Coomassie blue for quality control. Arrows highlight trimeric forms (estimated at 531 versus 456 kDa for LC³.Env³ and Env³, respectively) versus the monomeric (177 versus 152 kDa) chains of the vaccine. (C) Size-exclusion chromatography (SEC) diagram of LC³.Env³, depicting elution volume. Blue line corresponds to the elution of the vaccine, and position of standard molecular weights are mentioned. (D) ELISA binding assay to the human Langerin receptor of LC³.Env³ (red plain circles) and LC.Env (blue plain circles) compared to non-targeting Env trimer (Env³, orange open circles, dotted line), using the Env-binding 16H3 mAb for detection.³⁶ Data are presented as mean ± SEM from duplicate measurements across three independent experiments (total n = 4 for LC³.Env³ and Env³; n = 3 for LC.Env). (E) Binding of the LC³.Env³ versus LC.Env vaccine was assessed in a dose-dependent manner by FACS on human LC generated in vitro from CD34⁺ HSC (CD34-LC), compared to Env³ and non-targeting hIgG4.Env (blue open circles). Data are presented as mean ± SEM from duplicate measurements across three independent experiments. (F) Surface plasmon resonance (SPR) analysis of LC³.Env³ (red), LC.Env (blue), and Env³ (orange) captured via the 10–1074 mAb and assessed using 80 nM human (h, top), mouse (m, middle), or rabbit (r, bottom) Langerin as analytes. b: binding phase of analytes; d: dissociation phase. (G) Multiple-cycle kinetics (MCK) were performed as described in panel F, using the indicated analyte concentration ranges. (H) Binding affinity parameters (KD, kon, and koff) were derived from the corresponding sensorgrams. All measurements were performed in duplicate (n = 3 for LC³.Env³; n = 3 for LC.Env with mouse Langerin; n = 5 for LC.Env with other receptors). Dotted lines indicate medians for comparison: panel KD shows the median of LC.Env versus LC³.Env³; the Kon/Koff panel shows medians for human and mouse Langerin versus rabbit.

LC³.Env³ immunisation induced specific changes in Tfh and B cell populations in the draining lymph nodes

To assess the immunogenicity of LC³.Env³, C57BL/6J mice were intradermally injected without adjuvant twice, three weeks apart, with either LC.Env (10 μg/injection) or LC³.Env³ (7.2 μg to respect equimolar concentration of Env antigen, i.e., 5 μg) (Fig. 2A). We first looked at specific changes in cell populations in the dLN. We showed that frequency of CXCR5⁺ PD-1⁺ Tfh cells within the CD4⁺ T cell population was significantly higher in LC³.Env³ (mean of 0.9 ± 0.08%) vaccine group than unvaccinated (0.3 ± 0.06%) or LC.Env-vaccinated mice (0.34 ± 0.04%) (p < 0.0001) (Fig. 2B and C and Supplementary Fig. S3A). The Tfh cell response was accompanied by significant changes in the phenotype of B cells (Supplementary Fig. S3B). In LC³.Env³ vaccinated mice, total B cells (CD45⁺ B220⁺; 34 ± 1.4%), antibody-secreting cells (ASC; IgD⁻, CD138⁺, 0.12 ± 0.01%) and memory B cells (IgD⁻ CD138⁻, 6.4 ± 0.3%) were significantly increased as compared to unvaccinated mice (p < 0.01 for ASC; p < 0.0001 for total and memory B cells) and LC.Env-vaccinated group (p < 0.01 for memory and total B cell comparisons) (Fig. 2D–F). LC³.Env³ immunisation induced a significantly higher expansion of Fas ⁺ GL-7⁺ B_GC cells (57 ± 4.2% of memory B cells) than the LC.Env and control groups (p < 0.0001) (Fig. 2G and H).

Fig. 2.

LC³.Env³ elicits robust GC/Tfh responses in dLN. (A) C57BL/6J mice were intradermally immunised with 5 μg of the HIV-1 envelope antigen, equivalent to 7.2 μg of LC³.Env³, and 10 μg of the LC.Env mAb vaccine, without adjuvant. A prime-boost schedule was implemented, with a boost at day 21, followed by sacrifice at day 28. Created with BioRender.com (CC BY 4.0; https://BioRender.com/b2lmy7a). (B) Tfh cell phenotyping with representative dot plots of Tfh cells (PD-1⁺ CXCR5⁺) in each mouse group. (C) Percentage of Tfh cells induced post-boost in each group (PBS, n = 11; LC.Env, n = 12; LC³.Env³, n = 15). (D–F) Percentages of total B cells, antibody-secreting cells (ASCs), and memory B cells (member) were analysed by flow cytometry (PBS, n = 11; LC.Env, n = 13 or 14; LC³.Env³, n = 14 or15). (G and H) Expansion of B_GC cells (GL-7⁺/FAS⁺) post-boost in each vaccinated group (PBS, n = 9; LC.Env, n = 11; LC³.Env³, n = 13). Statistical analysis was conducted using one-way ANOVA with Tukey’s post hoc test: ∗∗p < 0.01; ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001; ns, non-significant. Bars represent the medians, while whiskers indicate the lowest and highest values in the data set of individual mice.

LC³.Env³ improved responses to Langerin targeting vaccine

We confirmed that these cellular changes in dLNs were associated with potent immunogenicity of the LC³.Env³ construct (Fig. 3A). We found that one-week post boost, LC³.Env³ vaccine induced significant Env-specific IFN-γ secreting T cells responses (mean of 61 ± 9 spot forming units (SFU) per million of cells, p < 0.0001 as compared to non-stimulated cells) (Fig. 3B and Supplementary Fig. S4A). Significant but lower IFN-γ T cell responses were also observed in the spleens of LC³.Env³-vaccinated animals (average of 46 ± 9 SFU per million of cells, p < 0.01) (Supplementary Fig. S4B). Expansion of Env-specific B_GC cells was confirmed using biotinylated Env trimers²⁶ (Fig. 3C). Env-specific B_GC cells were significant in the LC³.Env³ group only (0.011% of total B cells, p < 0.01) (Fig. 3D). Humoral responses against the trimeric Env antigen (Env³) were measured by Luminex®. Env-specific IgG were rapidly detectable after LC³.Env³ prime (day 14, average MFI of 891 ± 162), and increased after the boost, heightened at day 28 (average MFI of 3088 ± 300) (Fig. 3E). In LC.Env-vaccinated animals, responses were elevated compared to the PBS group at day 28. The levels still remained significantly lower than those observed in LC³.Env³-vaccinated animals, with average AUCs of 1119 ± 601 for LC.Env versus 27,625 ± 3393 for LC³.Env³ (p < 0.0001) (Fig. 3F). In summary, the LC³.Env³ immunisation induced specific changes in dLN cell populations with higher frequency of Env-specific B cells and magnitude of Env-specific IgG than LC.Env.

Fig. 3.

LC³.Env³ improves LC targeting. (A) On the day of euthanasia, auricular dLN were collected, and cells were harvested for INF-γ T cell ELISpot analysis. Created with BioRender.com (CC BY 4.0; https://BioRender.com/25qqg7e). (B) Responses were quantified as the number of spots per million cells. Black open circles represent unstimulated cells, while open red circles depict stimulation with Env peptide pools. Šidák’s multiple comparison test were applied: ∗∗∗∗p < 0.0001; ns, non-significant (PBS, n = 4; LC.Env, n = 5; LC³.Env³, n = 6). (C) Representative dot plots of Env-specific B_GC cells identified by biotinylated Env³ detected with two anti-biotin PE/anti-biotin APC secondary antibodies, and (D) the percentage of Env^+/+ B_GC cells in each mouse group. Statistical analysis was conducted using one-way ANOVA with Tukey’s post hoc test: ∗p < 0.05; ∗∗p < 0.01; ns, non-significant. Data represent the pooled results of three independent experiments (PBS, n = 4; LC.Env, n = 11; LC³.Env³, n = 10). (E) Evaluation of Env-specific IgG in the blood sera of mice vaccinated with LC.Env and LC³.Env³ using Luminex® at days 14 and 21 post-prime, and day 28 (one week post-boost). Data are presented as the median fluorescent intensity (MFI), with the dotted line indicating the boost at day 21. Red circles represent the LC³.Env³ group, blue circles represent the LC.Env group, and black circles represent non-immunised mice (PBS). Means (±SEM) are indicated, with statistical analysis comparing each vaccinated and naïve group of mice at each time-point, using two-way ANOVA with Tukey’s post hoc test: ∗∗∗p < 0.001; ∗∗∗∗p < 0.0001; non-significant not depicted. (F) Overall intensity of the responses calculated by the area under the curve (AUC), using Kruskal–Wallis non-parametric test with Dunn’s correction: ∗∗∗∗p < 0.0001; ns, non-significant. Bars represent the medians, while whiskers indicate the lowest and highest values in the data set of individual mice. In panels E and F, data represent the pooled results of three independent experiments (PBS, n = 11; LC.Env, n = 14; LC³.Env³, n = 15).

LC³.Env³ promoted efficient trimeric Env uptake by Langerhans cells, delivery into the dLNs and early GC formation

We then analysed GC structure in auricular dLNs of LC³.Env³ vaccinated mice by immunofluorescence (IF) and markers of B cell follicles (CD20), naïve B cells (IgD), follicular DC (FDC, CD35) and PNA staining of GC (Fig. 4A). As compared to the unimmunised group (PBS), haematoxylin/eosin staining (HES) revealed an early development of B cell follicles at day 8 (Fig. 4B), reaching a peak at day 14, as evidenced by PNA staining (green) of GC surrounded by naïve B cells (purple) (Fig. 4C, zooms a to c), and the FDCs confined to the central area of primary follicles, in the LC³.Env³ group (Fig. 4C, zooms d). Noteworthily, at day 28, we could observe FDC polarised arrangement that defines the germinal centre’s light zone³⁷ (Fig. 4C, zoom e). Next, we analysed the potency of LC³Env³ to elicit T and B cell Env-specific immune responses as compared to the non-targeted trimerisation form of the antigen (Env³). First, we tracked the uptake of LC³.Env³ versus Env³ that were labelled with AF-647 fluorochrome and injected intradermally (Fig. 5A and Supplementary Fig. S5). Draining lymph nodes collected after 24 h showed an increase in the absolute numbers of both resident (res) and migratory (mig) conventional dendritic cells (cDCs), similarly, in both vaccinated conditions as compared to the unimmunised condition (PBS) (mean increase by 2.16-fold for resDCs and 2.37 for migDCs) (Fig. 5B). Still, the number of vaccine-positive cells was increased in LC³.Env³ vaccinated animals as compared to Env³, with significant increase for LC (mean increase by 11.5 fold, p < 0.05) suggesting that LCs play a crucial role in taking up LC³.Env³ to the dLN (Fig. 5C). Second, we compared the immunisation effect of reduced doses of LC³.Env³ and the equimolar equivalent amount of non-targeting Env³ immunogens, both without adjuvant (Fig. 5D). Significant increases in circulating Env IgG were observed with LC³.Env³, at the lowest administered dose (250 ng of Env antigens) (p < 0.05, AUC D0–D28; Fig. 5E). Difference in terms of circulating Env IgG was maintained over 200 days (Fig. 5F). Although limited in number, even after 450 days post-immunisation, some mice still exhibited Env IgG titres (Supplementary Fig. S6A) and showed B cell follicles with typical structure of GC (PNA staining and/or CD35⁺ FDCs) significantly detected within dLN (p < 0.05, Supplementary Fig. S6B and C). Taken together, these results suggest that targeting trimeric Env antigens to Langerin alters vaccine uptake and delivery to the dLNs enhancing humoral responses in terms of onset, magnitude, and durability.

Fig. 4.

Targeting the HIV Env trimer to Langerhans cells induces GC organisation within the dLN. (A) Experimental design. Created with BioRender.com (CC BY 4.0; https://BioRender.com/01uj3em). (B) Haematoxylin and eosin staining (HES) was performed on days 8, 14, and 28 in LC³.Env³-vaccinated mice (equivalent to 5ug of Env) and compared to non-vaccinated controls (D28 post-immunisation). At the periphery of the dLN, round GC structures (Top, dotted circles and arrows) were observed in areas corresponding to B-cell zones (bottom, CD20⁺ cells in red), in contrast to the T-cell zones (yellow), as visualised by widefield microscopy for large image acquisition (mosaic, Zeiss AxioImager M2, 40 × objective). (C) GC structure was further confirmed by additional immunofluorescence (20× objective). Cryosections (6 μm) were stained with anti-CD20, anti-IgD Ab, and PNA to identify B cells, naïve B cells, and GC, respectively. Images are representative of at least 3 mice per group from 2 independent experiments. A series of zoomed-in fluorescence images (panels a–e; 20× snap images from the scan) are shown in the top corners. Light (LZ) and dark zones (DZ) within the dLN were inferred from the density of CD35⁺ follicular dendritic cells (panel e). Scale bars: 100 μm (whole section) and 50 μm (zoomed images).

Fig. 5.

Targeting the HIV Env trimer to Langerhans cells enhances humoral antiviral responses. (A) The Env³ HIV-1 trimer and LC³.Env³ were labelled with the AF647 fluorochrome and administered intradermally in mice. The fluorochrome solution from the labelling kit was injected as a control (PBS group). Created with BioRender.com (CC BY 4.0; https://BioRender.com/a7p9kr5). (B) After 24 h, dLNs were collected for FACS analysis, evaluating absolute numbers of resident versus migratory DC populations expressing or not expressing the Langerin receptor, and (C) absolute numbers of AF-647⁺ cells (considered as vaccine +) among the DC populations. The data presented are the mean percentages (±SEM) of three different independent experiments (PBS, n = 5; Env³, n = 5; LC³.Env³, n = 7). Statistical analysis was conducted using Šidák’s multiple comparison test, comparing within each vaccinated group the frequency of vaccine + DC population (right panel; ∗p < 0.05 for LC; non-significant not depicted). (D) Comparison of Env IgG titres induced by LC³.Env³ and the non-targeting Env³ HIV-1 trimer at low dose immunisation (250 ng of Env). Created with BioRender.com (CC BY 4.0; https://BioRender.com/a6ti83i). (E) Data are represented as AUC (D0–D28) from 3 independent experiments. Red circles, LC³.Env³; orange circles, Env³. Bars represent the medians, while whiskers indicate the lowest and highest values in the data set of individual mice. Groups were compared using a Mann–Whitney non-parametric t-test: ∗p < 0.05 (LC³.Env³, n = 7; Env³, n = 5) (F). Env-specific IgG titres were measured along 200 days in each group (n = 3 per group).

Design and characterisation of LC³.SOSIP constructs

Leveraging on results reported above, we further designed new LC-targeting immunogens by incorporating the clinically tested native-like BG505 SOSIP.664 structure; an Env immunogen adapted to be more stable and less prone to conformational changes, which can improve ability to elicit HIV-1 NeutAbs³⁸ (constructs illustrated in Supplementary Fig. S7A). Two LC³.SOSIP constructs were produced in CHO cells: LC³.SOSIP(s), which contains a glycine-serine (GS) flexible linker between the HIV-1 envelope gp41 and gp120 domains, and LC³.SOSIP(w), an in-house variant with two amino acid substitutions and a mutated furin cleavage site (Supplementary Table S1). The trimeric architecture and characteristic propeller-like structures of the Env trimers were further confirmed by SEC, native PAGE and electron microscopy (Supplementary Figs. S7 and S8).

LC³ stabilisation enhances bNAbs recognition of HIV Env immunogens

The impact of LC³ stabilisation on antigen integrity was evaluated by testing using Luminex the binding of a panel of bNAbs targeting distinct regions of HIV-1 Env to LC³-modified immunogens and their corresponding controls (Fig. 6). Env³ and LC³.Env³ showed binding by the CD4 binding site (CD4bs) VRC07 and 3BNC117 mAb, as well as the V3 glycan-directed 10–1074 mAb (Fig. 6A). In contrast, binding of trimer-specific apex mAb PGT145 and PG9, as well as the gp120/gp41 interface antibody PGT151, remained minimal for both constructs, consistent with limited formation of native-like trimer epitopes in this context. In contrast, LC³.SOSIP(s) and classical SOSIP(s) trimers exhibited robust and dose-dependent binding to all tested mAbs (Fig. 6B). LC³ incorporation did not impair recognition by any bNAb class, indicating that LC³ stabilisation preserved native trimer architecture in well-ordered SOSIP constructs. Compared to SOSIP(w), LC³.SOSIP(w) showed substantially high-affinity binding to quaternary epitope-dependent antibodies PGT145 and PG9 targeting the trimer apex, and PGT151 targeting the gp120/gp41 interface (Fig. 6C). Taken together, these data show that LC³ stabilisation improves Env antigenicity, did not alter epitope exposure of Env³ maintaining native antigenicity in well-formed SOSIP trimers, and improved the presentation of native-like, quaternary bNAb epitope, in weakly stabilised SOSIP constructs (SOSIP (w)).

Fig. 6.

Comparative profiling of bNAb binding to HIV Env vaccines and protein constructs. Purified Env³, LC³.Env³, SOSIP(s), LC³.SOSIP(s), SOSIP(w) and LC³.SOSIP(w) proteins were covalently coupled to MagPlex beads and used for comparative binding studies with a panel of anti-HIV Env bNAbs. In this Luminex assay, binding concentration response curves were generated for bNAbs binding the CD4bs (VRC07 and 3BNC117), V3 site (10–1074), trimer apex (PGT145), V1/V2 region of apex (PG9) and gp120/gp41 interface (PGT151) regions, with relative binding compared for: (A) Env³ (orange circles) and LC³.Env³ (red circles), (B) SOSIP(s) (orange circles), LC³.SOSIP(s) (purple circles), and (C) SOSIP(w) (orange squares) and LC³.SOSIP(w) (purple squares). Duplicates were performed for each bNAb concentration and data shown is representative of 3 independent experiments. Nonlinear regression binding curves were generated using Prism GraphPad Software.

LC³.Env³ and LC³.SOSIP immunisation of rabbits elicited Tier-1A and Tier-2 HIV-1 neutralising antibodies

The 4C7 mAb constitutive of the anti-Langerin AMV platform cross-reacts with the rabbit Langerin receptor (see Fig. 1 and Supplementary Fig. S1). Thus, we evaluated NeutAb responses elicited by LC³ vaccine platform in New Zealand White rabbits (Fig. 7A). As shown in Fig. 7B (top), immunisation with LC³.Env³, LC³.SOSIP(s), and LC³.SOSIP(w) without adjuvant led to a significant increase in Env-specific IgG levels following the second boost (p < 0.05), reaching comparable antibody titres (mean Log = 3.97 ± 0.11). These elevated responses were maintained for at least 40 days after the final boost. Notably, Env-specific IgG induced by LC³.SOSIP(s) (clade A) demonstrated cross-reactivity with the trimeric Env³ protein (gp140z, clade C) (Fig. 7B, bottom). In contrast, IgG responses elicited by LC³.Env³ (antigen from 96ZM651 clade C) exhibited a lower binding against heterologous clade A BG505 Env antigens (p < 0.05). Of note, antibody responses against Env³ and Env were similar in LC³.Env³ vaccinated rabbits, suggesting that the immune responses were not primarily directed against the trimerisation domain.

Fig. 7.

LC³.SOSIP induces Env-specific IgG and Tier-1A and -2 NeutAb in Rabbits. (A) New Zealand White rabbits were subcutaneously immunised with 20 μg of Env antigen without adjuvant, receiving either LC³.Env³ (red circles), LC³.SOSIP(s) (purple circles, dotted line), or LC³.SOSIP(w) (purple squares). NaCl injection served as the control (Mock, black crosses). The immunisation was performed at three time points (indicated by dotted lines). Created with BioRender.com (CC BY 4.0; https://BioRender.com/hwplg0r). (B) (top) Env-specific IgG levels in sera were measured by Luminex®, using the corresponding Env antigen for each vaccine (Env³ and BG505 SOSIP.664 in animals vaccinated with LC³.Env³ and LC³.SOSIP, respectively). MFI values ± SEM are shown (n = 12 and n = 5 for naïve versus vaccinees, respectively). Statistical analysis was performed using Dunnett’s multiple comparison tests, comparing each vaccinated group to the PBS control group at each serum collection time point (∗p < 0.05 for LC³.SOSIP(s) at D51 and D71, for LC³.SOSIP(w) at D51, for LC³.Env³ at D71, ∗∗p < 0.01 for LC³.SOSIP(w) at D71; non-significant not depicted) (bottom) Env-specific IgG titres in rabbits immunised with either of the vaccine group (see panel A) were measured against HIV-1 Env antigens: trimeric gp140z (Env³, red circles), monomeric gp140z Env (Env, blue circles), and the BG505 SOSIP.664 glycoprotein (purple squares). Titres are expressed as Log₁₀(AUC) from day 0 to study termination. Statistical analysis was performed using two-way ANOVA followed by Sidak’s multiple comparisons test: ∗p < 0.05; ns, non-significant. Bars represent the medians, while whiskers indicate the lowest and highest values in the data set of individual mice (n = 5 in each group) (C) Neutralisation assays were conducted to evaluate the sera’s ability to neutralise Tier-1A HIV-1 strains: MW965.26 (top) and SF162.NW (middle), and autologous Tier-2 BG505 (bottom). Titres are expressed as Log IRD₅₀ values for each vaccinated animal (LC³.Env³, n = 4; LC³.SOSIP(s/w), n = 5), plotted together with mean values (±SEM) of the mock group (black dotted line) (n = 4). Rabbits (r) showing significant neutralising titres as compared to the naïve animals (see Supplementary Fig S9) are indicated. The grey zone (titres below 20) indicates low or non-neutralising activity. Data shown are representative of two independent experiments. (D) Env-specific IgG titres were compared in rabbits immunised with a low dose (4 μg) of BG505 SOSIP.664 either targeted to the Langerin receptor (LC³.SOSIP(w), purple squares) or non-targeted (orange squares), with or without MPLA adjuvant. A mock group receiving MPLA was used as negative control (black crosses). Created with BioRender.com (CC BY 4.0; https://BioRender.com/9nzvq8l). (E) Mean MFI values ± SEM are reported for animals immunised without adjuvant (top, open symbols) or co-administered with MPLA (bottom, plain symbols). Statistical analysis was performed using two-way ANOVA with Tukey’s post hoc test: ∗∗p < 0.01; non-significant not depicted. (F) As in (C), NeutAb responses were assessed against the MW965.26 and BG505 strain in rabbits immunised with a low dose of either LC³.SOSIP(w) or BG505 SOSIP.664(w), with or without adjuvant. In panels E and F, n = 4 in each group.

Results showed that sera from all immunised animals with non-adjuvanted immunogens, LC³.Env³, LC³.SOSIP(s) or LC³.SOSIP(w) (n = 5/group), neutralised the clade C Tier-1A MW965.26 HIV-1 strain, with similar inhibitory reciprocal dilution (IRD₅₀) values (maximum titres ranging from 102 for LC³.SOSIP(s) to 285 for LC³.SOSIP(w), Fig. 7C top panels). Neutralisation of clade B SF162.NW Tier-1A strain appeared earlier with LC³.SOSIP(s) or LC³.SOSIP(w) as compared to LC³.Env³ (Fig. 7C middle). No neutralisation was detected against the Tier-2 HIV-1 strain 96ZM651 (not shown). However, we found that LC³.SOSIP(w) elicited NeutAbs against the autologous clade A BG505 Tier-2 strain in 2 out of 5 vaccinated rabbits (Fig. 7C lower panel and Supplementary Fig. S9), while no neutralisation against BG505 was induced by the prototype LC³.Env³ vaccine. These findings confirmed the ability of the LC³.SOSIP(w) platform to induce NeutAb, particularly when SOSIP-like antigens are targeted to Langerin receptors.

To better characterise the potency of the anti-Langerin targeting, we compared the immunogenicity of low doses of LC³.SOSIP(w) and non-targeted SOSIP(w) (equivalent to 4 μg of Env) given with or without adjuvant (MPLA) (Fig. 7D, n = 4/group). LC³.SOSIP(w) elicited significantly higher Env-specific binding IgG responses than control SOSIP(w) (p < 0.01, Fig. 7E). MPLA adjuvant did not enhance significantly IgG magnitude in LC³.SOSIP(w) animals. Analysis of NeutAbs confirmed the superiority of targeted SOSIP(w) as compared to non-targeted SOSIP (Fig. 7F) with higher neutralising titres (13-fold) against Tier-1A MW965.26 strain, in LC³.SOSIP(w) vaccinated animals (Fig. 7F). In addition, neutralisation against BG505 was confirmed in the LC³.SOSIP(w) group but not in the SOSIP(w) group at lower dose (2 out 8 vaccinated animals). Taken together, these results showed a potential advantage to improve the delivery of BG505 SOSIP.664 constructs to Langerin, even without adjuvant, for inducing HIV-specific NeutAbs.

Discussion

We report here that targeting a trimeric HIV-1 Env antigen to Langerhans cells (LCs) represents a highly effective vaccination strategy. The LC³.Env³ and LC³.SOSIP construct efficiently targets LCs in vivo, promoting antigen transport to B-cell follicles inducing strong GC and Tfh responses. This results in the generation of anti-Env IgG with neutralising activity against both Tier-1A and Tier-2 HIV-1 strains.

These findings build on previous studies, including from our group, demonstrating that antigen delivery through the Langerin receptor elicits potent immune responses.^20,22, 23, 24, 25, 26^,39,40 Targeting LCs has been shown to be superior to targeting dermal cDC1s, another Langerin-expressing population in mice, in terms of migration to dLNs and induction of B_GC cells responses in a dose-dependent manner.²² The trivalent LC³ design, composed of three anti-Langerin scFv, enables the fusion of structurally complex immunogens such as gp140z Env trimers while maintaining high-affinity Langerin binding at the nanomolar range. The enhanced immunogenicity of LC³.Env³ may rely on improved LC targeting and the multimeric presentation of Env. In addition, the absence of a human IgG Fc domain may limit background immune activation and non-specific Fc receptor-mediated uptake and degradation of Env, although this was not directly assessed.

Previous work has shown that Langerin-targeted dermal cDC1s can induce humoral responses primarily through extrafollicular pathways, whereas LCs preferentially promote GC-dependent responses.²² Consistent with this, LC³.Env³ induced significantly higher frequencies of Tfh cells, B_GC cells, and Env-specific B_GC cells than LC.Env in vivo. Tracking in vivo fluorescent LC³ versus non-targeted Env³ vaccines further revealed that distinct DC subsets capture the vaccine depending on whether the Env trimer is Langerin-targeted, supporting the conclusion that LC-mediated delivery of Env shapes the magnitude and durability of GC responses and antibody production, particularly at low doses of antigen. Direct antigen delivery to LCs may therefore represent an efficient strategy for prophylactic vaccination, as LCs can contribute to GC formation through antigen presentation and interactions with Tfh cells.^41,42 Within GCs, follicular dendritic cells (FDCs) in the light zone retain opsonised antigen and sustain B cell selection and affinity maturation.^37,43,44 It is likely that FDCs emerging after LC³.Env³ immunisation support the maturation of B_GC cells into memory B cells and antibody-secreting cells, ultimately driving Env-specific humoral immunity. Notably, these responses were achieved without adjuvant, an important advantage for clinical translation. While the mechanisms by which steady-state LCs promote humoral immunity remain incompletely defined, recent evidence suggests that some inflammatory cytokines may be dispensable in this process.⁴⁵

The LC³ platform further enables delivery of Env trimers that better preserve native quaternary structure. LC³.Env³ incorporates a T4 fibritin-stabilised Env trimer, a design initially developed decades ago³⁵ and later adapted by Nussenzweig’s lab for improved detection of Env-specific memory B cells.⁴⁶ Consistent with prior reports demonstrating cross-clade neutralisation induced by T4-stabilised gp140 trimers,⁴⁷ LC³.Env³ elicited NeutAbs in rabbits, with anti-Env IgG detectable after the second boost and increasing after the third. Comparable antibody titres were observed using Env monomers or trimers, confirming Env-specificity. While the T4 fibritin was not immunodominant in our experimental setting with LC³.Env³ in the absence of adjuvant, repeated immunisations employing heterologous trimerisation domains or nanoparticle scaffolds have been reported to induce strong immune responses that may sometimes overshadow anti-Env IgG responses.^48,49

Thus, to further optimise antigen quality, we evaluated foldon-free SOSIP trimers. LC³.SOSIP constructs were produced in CHO cells, purified by SEC, and shown by EM to form homogeneous trimeric assemblies. Binding to a diverse panel of bNAbs covering the principal epitope specificities confirmed that LC³.SOSIP proteins present a native-like, closed prefusion Env trimer, exposing all major neutralising epitopes. In particular, strong recognition by apex-specific (PGT145, PG9) antibodies, demonstrated that the quaternary structure of Env in both the LC³.SOSIP(s) and the LC³.SOSIP(w), which contrasted with the more limited epitope exposure observed for Env³ and LC³.Env³. Furthermore, the gp120/gp41 interface epitope targeted by PGT151 is effectively presented by LC³.SOSIP proteins, with LC³.SOSIP(w) showing improved recognition and tighter binding than all other protein constructs, including SOSIP(w) comparator proteins These data aligned with the in vivo immunogenicity results and highlighted the high structural quality of LC³.SOSIP antigens. LC³.SOSIP induced Env-specific IgG titres comparable to those elicited by LC³.Env³, demonstrating that native-like trimers can be efficiently delivered to LCs and that the trimerisation via foldon is not required for robust humoral responses. The antibodies generated were cross-reactive with clade C Env and capable of neutralising both Tier-1A viruses and the Tier-2 autologous BG505, representing an improvement over the prototype LC³.Env³. Furthermore, LC³.SOSIP incorporates the T332N mutation, which restores a conserved glycan and favours the elicitation of antibodies targeting the V3-glycan supersite.²⁹ However, the immunodominance of responses directed toward the S241 glycan hole⁵⁰ should be addressed in a next-generation design.⁵¹ Although neutralisation remained modest, probably due to the low antigen dose used, these data indicate that LC targeting can enhance the quality of SOSIP-induced responses and support the delivery of rationally designed immunogens, including those incorporating rare mutations or germline-targeting features.⁵²

In vivo tracking confirmed efficient migration of skin DCs to dLNs and prominent uptake by LCs, although contributions from other DC subsets cannot be excluded. Nevertheless, the key role of LCs in driving humoral immunity through Langerin targeting is supported by results obtained in XCR1^DTA mice lacking Langerin ⁺ dermal DCs.²⁶ Further research is required to clarify how this vaccine is handled in vivo and to determine whether the trimeric Env structure influences B-cell priming. It will also be important to assess whether the trimeric scFv domains provide shielding at the base of the Env trimer and affect epitope presentation. Given that rabbit antibodies often possess longer CDR loops than those of mice, which may facilitate penetration of the HIV Env glycan shield and support the induction of neutralising responses, rabbits were used to assess NeutAbs activity. While cross-reactivity of the targeting antibody with rabbit Langerin was confirmed by SPR, LC subsets in rabbit skin were not directly characterised, and future studies should address to assess their immunological functions upon targeted delivery.

To conclude, this study established LC targeting as a versatile and promising platform for HIV vaccine development. By enhancing antigen delivery, GC responses, and antibody quality without adjuvant, this strategy may facilitate the induction of broader neutralising immunity and contribute to next-generation HIV vaccines.

Contributors

Conceptualisation: SC, YL, VG, GP, MC; Vaccine production and qualification AH, AK, ME, ES, GF, SZ, GZ, CF, TB, SC; Mouse Experimentation: AH, MS, JL, FP, GH, JM, TA, VG, SC; Rabbit experimentation: AH, BP, DG, CM, WW, Funding acquisition: YL, SC. Writing – original draft: AH, SC, YL, VG. Access and verification of underlying data: AH, SC, YL, VG. All authors participated in discussions of experimental results, read and approved the final version of the manuscript.

Data sharing statement

All raw datasets and original images of SDS-PAGE and native PAGE gels are freely accessible via the Mendeley Data repository (https://doi.org/10.17632/gsxj3sdr3r.2; version 2; February 16, 2026). No custom code or scripts were employed beyond standard use of GraphPad Prism. Additional information can be obtained from the corresponding authors upon reasonable request. The sequences used in this study were deposited in GenBank as follows: anti-Langerin clone 4C7 variable regions (heavy chain: JX002669.1; light chain: JX002668.1), HIV-1 Env Clade C 96ZM651 gp140 (AY181197.1), FlexV1 linker (AJD85777.1), T4 fibritin trimerisation domain (ADJ39878.1), and furin (NP_001276752.1). BG505 SOSIP.664 was derived from the BG505 Env sequence (GenBank accession numbers: DQ208458, ABA61516.1), as originally described by Sanders et al.²⁹ Pseudovirus production utilised HIV-1 envelope expression plasmids corresponding to MW965.26 (GenBank: U08455.1), SF162.NW (GenBank: EU123924), 96ZM651 (GenBank: AF286224), and BG505 (GenBank: DQ208458; protein: ABA61516.1), co-transfected with the env-deficient HIV-1 backbone plasmid pSG3Δenv (GenBank: L02317). All unique reagents generated in this study are listed in the Supplementary Information and are available from the corresponding authors upon completion of a Materials Transfer Agreement (MTA). Schematic representations of experimental designs and SPR binding assays were created using BioRender (Cardinaud S., 2026). The figures are distributed under a Creative Commons Attribution 4.0 International Licence (CC BY 4.0), with the licence details (https://creativecommons.org/licenses/by/4.0/) indicated within each figure.

Declaration of interests

Authors declare no financial or commercial conflicts of interest, except GZ and SZ are named inventors on patent filings for the use of antigens directed to LCs as vaccines (WO 2011/032161 A3; Vaccines directed to Langerhans cells; August 25, 2011), filed by Baylor Institute for Immunology Research (Dallas, TX, USA) and licenced to Inserm Transfert (Paris, France). AH, MS, VG, YL and SC are named inventors of (i) Langerhans cells targeting HIV-1 vaccines, filed by Inserm and University Paris-Est Créteil (France) licenced to Inserm Transfert (International Procedure PCT/IB2024/000163; March 24, 2024), and (ii) Vaccine delivery of HIV-1 Env trimers to Langerhans cells. filed by Inserm; AP-HP Assistance Publique - Hôpitaux de Paris; and Université Paris-Est Créteil Val de Marne (France) (European Procedure EPA #24 315 512.4; Nov. 6, 2024).