Vector-directed immunity and boosting capacity of VRP-srRNA anti-tumor vaccines

Abstract

Self-replicating RNA vaccines delivered by viral replicon particles (VRPs) induce strong immunity, but repeated dosing on short intervals may be limited by vector responses. In patient samples and mice, VRP-srRNA vaccination generated VRP-neutralizing antibodies and T-cell responses to replicase. In mice, a third dose at two-week intervals minimally boosted transgene immunity, and prior VRP exposure reduced responses to a different-transgene VRP. Optimized schedules or heterologous prime–boost may sustain immunogenicity.

The success of mRNA vaccines has renewed interest in self-replicating/self-amplifying RNA (srRNA/saRNA) vaccines. These vectors typically encode both the target antigen and viral-derived replicases (four separate non-structural proteins, nsP1-4), enabling intracellular amplification and prolonged antigen expression. This allows protection comparable to mRNA vaccines but at lower doses¹. Delivery of srRNA vaccines can occur using lipid nanoparticles (LNPs) or as alphavirus replicon particles (viral replicon particles (VRPs)). Of these different formats, VRP-srRNAs have a longer track record and often generate stronger CD4⁺ and CD8⁺ T cell responses due to their natural targeting of antigen-presenting cells, particularly dendritic cells^2,3. In past studies, we have developed and clinically tested a Venezuelan Equine Encephalitis Virus (VEEV)-based VRP-srRNA vaccine, encoding the extracellular and transmembrane domains of HER2 (Fig. 1A, srRNA sequence in Supplementary Fig. S1 and GenBank)⁴. The srRNA is encapsulated in VRPs composed of VEEV structural proteins, allowing infection of host antigen-presenting cells without subsequent viral propagation (Fig. 1A). Previous studies have shown that VRP vaccination induces neutralizing antibodies against the viral particle structural proteins^4,5, but the kinetics of these responses remain undefined. Similarly, whether the replicase-derived non-structural proteins (nsPs) elicit immunity in VRP-based vectors has not been characterized. Moreover, despite the presence of neutralizing antibodies, VRP-based srRNA vaccines have been administered repeatedly in both preclinical and clinical studies^4,6–11, and homologous boosting has been reported^4,6,7. However, the optimal timing of boosting and its ability to further enhance antigen-specific immunity are unclear.

Fig. 1. Induction of VRP-neutralizing antibodies in patient sera following VRP-HER2 vaccination.

A Schematic of the VRP-HER2 structure and genome. B Timeline of VRP-HER2 administration and patient sample collection in clinical trial (NCT01526473). C Patient sera collected over time were analyzed in VRP neutralization assays to determine VRP-neutralizing antibody titers. Titers were determined by taking the inverse of the IC₅₀ value divided by 50, so that titers at week 0 are 1. An ordinary one-way ANOVA with Tukey’s multiple comparisons was performed. ns P > 0.05, *P < 0.05, ****P < 0.0001. A, B Schematics were created with BioRender.

To address these gaps, we assessed both serum antibody and T cell responses against vector-specific epitopes from HER2+ breast cancer patients treated with a self-replicating RNA vector derived from Venezuelan Equine Encephalitis (VEE) encapsulated in VRP that expresses HER2 from a past Phase I clinical trial⁴ (NCT01526473, Fig. 1A and Supplementary Fig. S1). In this study, patients received three doses of HER2-VRP (4 × 10⁸ IU) at weeks 0, 2, and 4 with biospecimens (serum and PBMCs) collected at screening, weeks 0, 2, 4, 6, and 8, with additional specimens from some patients at months 3, 6, and 12 (Fig. 1B). Using a microscopy-based VRP neutralization assay, we first assessed the magnitude and kinetics of neutralizing antibody responses (shown as reverse IC50 titers) for all study time points, up to 1-year post-vaccination. Notably, all patients had no pre-existing immunity or neutralizing antibodies against the viral vector prior to treatment. Analysis of serum samples revealed that neutralizing antibody titers were surprisingly low at 2 weeks after the priming vaccine, but homologous boosting significantly augmented responses at week 4, post-injection of two doses of vaccine, with 13 of 16 patients (83%) at this time point demonstrating increased neutralizing antibody titers (Figs. 1C and S2). At week 6 and week 8, all patients developed increased neutralizing antibody titers, although neutralizing antibody responses overall were not significantly different from those at week 4 (Fig. 1C). Neutralizing antibody titers then gradually declined but remained detectable compared to baseline through month 12, the last time point assessed (Figs. 1C and S2).

To next determine whether VRP-HER2 vaccination induced T cell responses to vector-encoded nsPs, PBMCs collected from patients pre- and post-vaccination were stimulated with VEE nsP overlapping peptide pools (Table S1) and analyzed by IFNγ ELISPOT. In our previous studies, CyToF analysis revealed that only certain patients had detectable target antigen HER2-specific CD8+ T cell responses, defined by perforin expression⁴. Our pooled ELISPOT data (reported as Z-scores as described in “Methods”) revealed significant nsP-specific responses, defined as having Z-scores > 5 at more than one time point for each peptide pool, in a minority of patients against nsP1–3 and even less frequently against nsP4 (Fig. 2A–D). ELISPOT responses against nsPs appeared as early as week 2 in certain patients, peaked at weeks 6–8, and generally declined by month 3. Individual patient data (Fig. S3) showed that seven of 18 patients developed nsP-specific responses, with only two patients responding to multiple nsP pools and five to a single pool. ELISPOT responses were confirmed by multi-cytokine flow cytometry of peptide-stimulated PBMCs from patient AVX_02_002, which demonstrated only modest nsP1 responses (Fig. S4). Flow cytometry further revealed that nsP1-specific T cells were poly-functional CD8+ (not CD4+), expressing both IFNγ and TNFα cytokines post-stimulation (Fig. S4). Of note, patients AVX_02_09 and AVX_02_10, who mounted responses to multiple nsP pools, were previously shown to lack perforin⁺ CD8⁺ T cell responses against HER2⁴. To examine whether vector-directed immunity interfered with antigen-specific responses, we assessed the correlation between the two using the post-vaccination percentage change in perforin⁺ CD8⁺ T cells responding to HER2 as a proxy. However, limited by this small Phase I sample size, our analysis revealed a non-significant inverse correlation between the summed nsP-specific Z-scores and the percentage of HER2-reactive perforin⁺ CD8⁺ T cells, as well as no significant correlation between VRP-neutralizing antibody titers and HER2-specific responses (Fig. S5). Collectively, these studies establish T cell responses against nsPs in patients, but demonstrate that they are highly heterogeneous and only occur in a minority of patients. Moreover, these responses declined post-vaccination and were skewed towards nsP1–3 but not nsP4. Whether these responses are enriched in patients who did not elicit significant HER2-specific T cell responses requires a larger sample size in future research to ascertain.

Fig. 2. Development of IFNγ response to VRP non-structural proteins in patient PBMCs post-vaccination.

Patient PBMC samples collected over time were stimulated with nsP1-4 peptide pools, and IFNγ secretion was detected by ELISPOT. Patient PBMC ELISPOT IFNγ responses to nsP1 (A), nsP2 (B), nsP3 (C), and nsP4 (D) peptide pools. Data are presented as Z-scores calculated by the difference from the control SIINFEKL peptide, divided by the standard deviation of the control. For (A–D), one-way ANOVA was performed. Differences between bars in each graph are not statistically significant (P > 0.05) because of high variability between patients.

While these Phase I clinical data demonstrate that VRP-srRNA vaccines elicit both T cell and humoral responses against the VRP vector components, all patients received two homologous boosting vaccines. Thus, it remained unclear whether repeated doses boost responses or if vector-elicited immunity is limiting further responses from VRP-srRNA vaccines after homologous VRP vaccination. To address these unknowns, we performed a preclinical mouse study with six vaccination regimens: no vaccine, one dose of VRP-CEA⁵ (vaccine encoding irrelevant antigen), two doses of VRP-CEA followed by one dose of VRP-HER2, one dose of VRP-HER2, two doses of VRP-HER2, or three doses of VRP-HER2 (Fig. 3A). To mirror the setup in our clinical study, all vaccines were given intramuscularly, 2 weeks apart and mice were staggered so that all vaccinated groups received the last dose of vaccine at the same time (Fig. 3A). We first performed VRP neutralization assays that revealed a single dose of VRP-CEA or VRP-HER2 produced higher titers in 80% of the vaccinated mice than unvaccinated controls (Fig. 3B), although the difference to baseline was not statistically significant, similar to responses in patients (Fig. 1C). In contrast, two doses of VRP-HER2 induced neutralizing antibodies in all mice by 4 weeks, with titers significantly higher than baseline and one-dose groups (neutralization curves of individual mice shown in Fig. S6). However, additional vaccination (three total doses, either 2× VRP-CEA + 1× VRP-HER2 or 3× VRP-HER2) did not further increase neutralization titers beyond those achieved with two doses of VRP-HER2. Given the level of neutralizing antibodies, we next determined if antibody responses to the encoded transgene, HER2, could still be boosted by homologous vaccination. Evaluation of anti-HER2 IgG titers by ELISA revealed that all groups receiving VRP-HER2 developed significantly higher anti-HER2 IgG titers, compared to the unvaccinated group or 1 dose VRP-CEA group (Fig. 3C). Like boosting of VRP neutralization titers (Fig. 3B), a second administration of VRP-HER2 boosted levels of HER2-specific antibodies, although a third dose did not significantly augment these responses, demonstrating a limitation on VRP-srRNA boosting capacity on a 2-week dosing schedule. Notably, heterologous vaccination (2× VRP-CEA + 1× VRP-HER2) generated significantly lower HER2 antibody titers than a single dose of VRP-HER2, suggesting that pre-existing immunity against VRP may dampen but did not eliminate responses to new antigens by reducing effective VRP-srRNA transduction (Fig. 3B, C).

Fig. 3. Antibody responses from homologous VRP-srRNA vaccinated mice.

A Schematic of the mouse vaccination experiment design created with BioRender. B Mouse sera from the vaccination experiment were tested in VRP neutralization assays to determine VRP-neutralizing antibody titers. Titers were determined the same way as in the human sera VRP neutralization experiment. C Mouse sera from the same experiment were serially diluted and tested in the HER2 protein-based ELISA. Anti-HER2 IgG titers were calculated by the reciprocal of the serum dilution at which serum dilution curves reached the 95% cutoff, as described in “Methods.” For (B, C), an ordinary One-way ANOVA followed by Tukey’s multiple comparisons test was performed. ns P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Given induction and boosting of humoral immune responses after homologous vaccination, we next investigated if VRP-srRNA vectors also elicited and boosted antigen-specific T cell immunity against the vector-encoded antigen and nsPs. To address this question, splenocytes from groups of mice in Fig. 3A were stimulated with CEA peptides, HER2-ECD peptides, and nsP1-4 peptide pools in IFNγ ELISPOT assays. These studies revealed significant T cell responses against nsP3 in all vaccinated groups, suggesting a strong immunodominance in BALB/c mice, in contrast to responses against nsP1, 2, and 4 epitopes (Fig. 4A–D). Notably, nsP3-specific T cell responses were relatively equivalent between mice given one or multiple doses of VRP-CEA and/or VRP-HER2 (Fig. 4C), suggesting a lack of T cell boosting against nsP epitopes and indicating generally muted responses to nsPs, similar to the modest responses we observed clinically (Fig. 2A–D). Mice receiving 1 dose of VRP-CEA or 2 doses of VRP-CEA followed by VRP-HER2 developed significant IFNγ responses to CEA compared to control, but responses were not boosted (Fig. 4E). In contrast, we found that T cell responses against HER2 were significantly boosted with an additional VRP-HER2 vaccination, but were not further elevated by a second boost (third vaccination) (Fig. 4F). These results indicate that, under the 2-week dosing regimen, homologous boosting with VRP-srRNA showed limited incremental benefit by week 4, when neutralizing antibody titers were high and nsP3-reactive cellular responses were detectable (Figs. 3B and 4C). Consistently, mice primed with 2 doses of VRP-CEA and then given a single VRP-HER2 dose at week 4 failed to stimulate HER2-specific T cell IFNγ responses, in contrast to naïve mice vaccinated with VRP-HER2 (Fig. 4F). The early prevalence and magnitude of VRP-neutralizing antibodies in mice suggest that vector-directed humoral immunity is a major barrier to effective boosting within this short interval, although our data do not establish causality or exclude a contribution from VRP structural protein- or replicase-directed cellular immunity. Further mechanistic studies assessing cellular responses to VRP structural proteins and antibody/T cell depletion or passive transfer experiments will be required to delineate the relative contributions of humoral versus cellular anti-vector immunity to impaired transgene-directed responses.

Fig. 4. T cell responses from homologous VRP-srRNA vaccinated mice.

Mouse splenocytes were stimulated with nsP1 (A), nsP2 (B), nsP3 (C), nsP4 (D), CEA (E), and HER2-ECD (F) peptide pools, and splenocyte IFNγ responses were measured by ELISPOT. Data from (A–F) were analyzed together in a two-way ANOVA followed by Tukey’s multiple comparisons. ns P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Our study uncovers a previously underappreciated dimension of VRP immunity, in that beyond inducing potent neutralizing antibodies, VRP vectors can also drive T cell responses against their non-structural replicase proteins. In patients, nsP1–3 emerged as immunodominant antigens, though individual patients showed distinct hierarchies of response. Notably, we found that patients exhibited a non-significant inverse correlation between HER2-specific and nsP-specific T cell responses. Further investigation in larger clinical trials is needed to evaluate the potential trade-off between vector-specific and transgene-specific immunity. Additionally, longitudinal analyses of patient samples revealed that both neutralizing antibody and replicase-specific T cell responses peaked after three doses of VRP-HER2, around weeks 6–8, and declined sharply by 3 months.

In BALB/c mice, cellular response to the viral replicase narrowed to a striking nsP3 bias, likely reflecting host MHCI restrictions or the unique biology of nsP3, which forms cytoplasmic aggregates and interacts broadly with host proteins¹². In the same model, repeated homologous dosing on our 2-week schedule yielded limited incremental boosting by the third dose, and prior exposure to two VRP doses blunted responses to a subsequent VRP encoding a different transgene. This regimen-dependent limitation parallels challenges faced by other viral vectors, such as AdHu5 and MVA^13,14. However, comparisons of our mouse data to human settings require caution. While mouse and human anti-vector humoral responses in our study were measured using the same assay and are therefore more readily comparable, anti-vector cellular responses were assessed using IFN-gamma ELISPOT assays performed from different biospecimen sources (murine splenocytes versus human PBMCs). These differ in composition and may influence response magnitude, although this magnitude was generally higher in the murine studies. Accordingly, we view the murine studies as providing qualitative, directional insight into how repeated VRP-based vaccination may shape both anti-target and anti-vector immunity, rather than as a quantitative proxy for human or non-human primate responses.

Our findings suggest a potential trade-off intrinsic to VRP-srRNA vaccines. Their unparalleled ability to prime innate and adaptive immunity may be offset by vector-directed responses that restrict iterative dosing within short 2-week intervals and potentially limit responses against transgenes in a minority of patients. Yet, the observation that vector immunity wanes by 3 months (Fig. 1C) suggests windows of opportunity remain, given that other viral vectors can retain boosting capacity at delayed time points¹⁵. To define the optimal timing for homologous boosting with VRP-srRNA vaccines and to determine the extent to which vector immunity limits boosting, additional studies evaluating longer inter-dose intervals (e.g., 4–12 weeks) are warranted. Given that our data highlights the prevalence and large magnitude of humoral responses to the VRP in human and mouse (Figs. 1C and 3B), alternative or heterologous srRNA delivery regimes^16,17 may be warranted to minimize the impact of neutralizing antibodies on transgene immunity. Notably, a recent study showed that LNP-srRNA elicited replicase (nsP)-directed humoral and T cell responses that impaired transgene-directed immunity but not protection against Influenza A in murine experimental settings¹⁸. This indicates that srRNA delivery via alternative strategies may not necessarily eliminate the potential impact of replicase-directed (nsP) immunity on srRNA vaccine immunogenicity. In sum, our data support the view that future development of VRP-srRNA vaccines should couple their strong immunogenicity with strategies to mitigate immune responses to both structural (vector) and non-structural (replicase) components.

Methods

Human clinical trial and patient samples

Human samples analyzed were collected as part of a phase I clinical trial performed under an FDA-approved Investigational New Drug Exemption and registered at ClinicalTrials.gov (NCT01526473). Participants provided written consent under a protocol approved by the Duke University Medical Center Institutional Review Board. The study was conducted in accordance with all recognized institutional and federal guidelines, including those put forth in the Declaration of Helsinki. Additional information about participant demographics and trial specifics can be found in ref. ². Samples analyzed in this report were exclusively from Cohort 2.

Mouse experiments

BALB/c (BALB/cAnNTac; model BALB) animals were purchased from Taconic Biosciences (Germantown, NY, USA) and bred at Duke University. Mice between 6 and 12 weeks of age were vaccinated in the left quadriceps while anesthetized with isoflurane (4 × 10⁶ particles in 0.05 mL per mouse). Upon sample collection, mice were euthanized by carbon dioxide at 30–70% displacement flow rate until 1 min after unconsciousness/death, followed by decapitation/heart puncture as a secondary means to ensure death. All animal studies were performed in accordance with Duke IACUC-approved protocol (A080-20-04 and A043-23-02) and housed by the Division of Laboratory Animal Resources.

nsP peptide pools

nsP peptide pools were made using overlapping 15-mers spanning the length of nsP1, nsP2, nsP3, or nsP4 from VEEV TC-83’s non-structural protein sequences with 4 amino acid overlaps (ordered from GenScript). Peptides were dissolved in DMSO, and pooled stocks were made at 100 µg/mL/peptide. For a complete list of peptide sequences, see Supplementary Table 1 in Supplementary Data.

ELISpot assay

Human IFNγ ELISpot assays (MabTech, Kit 3420-2H) and mouse IFNγ ELISpot assays (MabTech, Kit 3321-2H) were performed according to the manufacturer’s instructions. Human patient PBMCs (2.5 × 10⁵ cells/well) were incubated in RPMI1640 medium (Gibco) with 10% heat-inactivated fetal bovine serum for 24 h. During incubation, human PBMCs were stimulated with HER2-ECD peptide (1 µg/mL, JPT), NSP1 peptide pool, NSP2 peptide pool, NSP3 peptide pool, NSP4 peptide pool (1 µg/mL, GenScript), non-specific SIINFEKL peptide (1 µg/mL, JPT), or positive control PMA/Ionomycin (0.5 ng/ml for PMA and 1 µg/ml for Ionomycin, Sigma). For human ELISPOT graphs, Z-scores for each peptide stimulation were calculated as the difference from the control HIV-gag peptide mix, divided by the standard deviation of the control. Mouse splenocytes (5 × 10⁵ cells/well) were incubated in RPMI1640 medium (Gibco) with 10% heat-inactivated fetal bovine serum for 24 h and stimulated with CEA peptide (1 µg/mL, JPT) and positive control VEEV E2 peptide (1 µg/ml, JPT) in addition to the same peptides used in human ELISpot assays.

Flow cytometry on peptide-re-stimulated human PBMCs

Human PBMCs (2 × 10⁶ cells/well in 100 µL of RPMI1640 with 10% human serum (Sigma H3667)) were plated in 96-well round-bottom plates. All PBMCs received Brefeldin A (1:1000 Invitrogen 00450651), Monensin (1:1000 Invitrogen 00450551) 1:1000, anti-CD28 (10 µg/ml BD Pharmingen 340975) antibodies and anti-CD49d (10 µg/ml BD Pharmingen 340976). Depending on the treatment group, PBMCs were cultured in one of the following conditions: media control, OT-1 control peptide, HER2-ECD peptide, NSP1 peptide pool, NSP2 peptide pool, NSP3 peptide pool, NSP4 peptide pool, or Staphylococcal Enterotoxin B positive control (0.1 mg/mL, Toxin Technology #BT202). All peptides were given at 1 µg/mL. After 6 h of incubation at 37 °C, PBMCs were stained with live/dead dye and surface-targeting antibodies diluted in PBS/1%BSA in the dark. Surface antigen-stained PBMCs were washed with PBS/1%BSA, then fixed and permeabilized with FOXP3-fix/perm buffer (Tonbo TNB-0607-KIT) for 1 h at 4 °C, after which they were washed with 1x Permwash (Tonbo TNB-1213-L150) and stained with intracellular antibodies diluted in Permwash for 1 h at 4 °C. Finally, samples were washed with PBS and collected for acquisition by a Cytek 3L V/B/R Northern Light cytometer.

The flow panel used to stain human PBMCs can be found in Supplementary Table 2.

VRP neutralization assay

On day 1 of the assay, Vero cells (ATCC CCL-81) were plated (4000 cells/well) in 200 µL of DMEM medium (Gibco) with 10% heat-inactivated fetal bovine serum in the inner wells (B2-G11) of a 96-well flat-bottom plate, while the outer wells were filled with DPBS. Vero cells were maintained in a 37 °C, 5% CO₂ humidified incubator overnight. Simultaneously, on day 1, mouse or human sera were serially diluted to halve the final concentrations in HBSS + 1%BSA (Gibco & Sigma). Diluted serum (30 μL/well) was added to V-bottom plates pre-added with 30 μl/mL of VRP-CEA vaccine (800,000 IU/mL). Plates were sealed and incubated overnight at 4 °C. All sera final dilution started at 1:50. Human sera were serially diluted in a two-fold series, while mouse sera were serially diluted in a five-fold series. All serum concentrations were run in quadruplicate. On day 2, the mixture of VRP and serum (50 μL/well) from the previous day was added to Vero cells and cultured at 37 °C overnight. On day 3, Vero cells were fixed with 10%NBF (100 μL/well) for 15 min and washed once with 100 μL of DPBS + 1%BSA. Cells were then permeabilized, blocked, and stained with Alexa Fluor 594 anti-CEA Ab (ASL-32, BioLegend, 1.25 µg/mL) and DAPI (1 drop/mL NucBlue DAPI, Invitrogen). Plates were sealed and stored at 4 °C until imaged using an Agilent BioTek Cytation 7.

On the day of imaging, one 4× image was captured in two channels (DAPI & Texas Red 586) from the center of each well. Images were analyzed in Cell Profiler (version 4.2.8), identifying nuclei with IdentifyPrimaryObjects on DAPI. Nuclei were expanded using Texas Red. Intensity of secondary objects as measured on Texas Red, before objects were filtered based on MaxIntensity as CEA positive, using an empirically determined minimum cutoff. The percentage of CEA positively staining cells was calculated and exported on a per-image basis. Percent positivity was normalized to respective control sera on a per-concentration basis. Human sera were normalized to each patient’s own Week 0 sample, while mouse serum was normalized to the average of the control group. Neutralization curves and the IC₅₀ for each sample were calculated in Python using the neutcurve package¹⁹. Titers of VRP neutralization were calculated by the reciprocal of the IC₅₀ value divided by 50, so that titers at week 0 are 1.

ELISA

Immulon 4 HBX plates (Thermo Scientific, Waltham, MA, USA) were coated with 50 µL/well of 1 µg/mL HER2-ECD protein at 4 °C overnight. Plates were washed with DPBS + 0.05% Tween 20 before being blocked with DPBS + 1% BSA (Sigma-Aldrich, Inc., St. Louis, MO, USA) for 1 h at 37 °C. Serum samples were diluted in DPBS + 1% BSA and incubated on blocked plates for 2 h at 37 °C. Plates were washed with DPBS + 0.05% Tween 20, and anti-mouse IgG streptavidin-HRP-conjugated antibody (1:2000 in DBPS + 1% BSA; Cell Signaling Technology) was added for 1 h at 37 °C. Plates were washed with DPBS + 0.05% Tween 20, added with TMB substrate (100 µL/well, BioLegend), allowed to develop at 37 °C, stopped with 0.18 mol/L H₂SO₄, and read at 450 nm on an Agilent BioTek Cytation 7. 4-parameter curves were fit and cutoff values were determined using unvaccinated control serum to a 99% confidence level as described in this paper²⁰. Titers were calculated as the reciprocal of the serum dilution at which serum dilution curves reached the cutoff.

Statistical analyses

Data were plotted and analyzed with GraphPad Prism, Python nuetcurve or matplotlib packages. Detailed statistical analysis used in each graph can be found in the figure legends. P values of 0.05 or less were considered statistically significant. Not all statistically significant differences are shown in graphs (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001).

Author contributions

This work was conceived by Z.C.H., R.D.M., and X.M.; R.D.M., A.S., S.G., X.M., T.W., C.L., G.L., and J.W. conducted the experiments and data analysis. A.H., M.A.M., and H.K.L. conducted a human clinical trial. E.J.C. contributed critical instrumentation (Cytek flow cytometer). Funds from Z.C.H and H.K.L supported this study. The manuscript was drafted by X.M. and Z.C.H.; all authors edited the manuscript and approved its submission.

Data availability

All raw data and analyzed data have been deposited on the Duke Department of Surgery Protected Server and are available upon request. Privacy regulations prohibit public submission of patient data, but with patient consent, anonymized data are available from the corresponding author upon reasonable request. Plasmid sequences of self-replicating RNA vaccines were deposited in GenBank (accession numbers VRP-CEA: PX960553 and VRP-HER2-ECDTM PX960554). Schematics created in BioRender by X.M. can be accessed online (Fig. 1A at [https://BioRender.com/d7k4da3]; Fig. 1B at [https://BioRender.com/pwqgr3o]; Fig. 3A at [https://BioRender.com/fg6ghdh]).

Competing interests

Z.C.H. and H.K.L are co-founders, and both have equity stakes in Replicate Biosciences (San Diego, CA), a srRNA platform biotech that is developing breast cancer vaccines. Z.C.H is an Associate Editor of Cancer Research Communications (CRC) and has been a guest editor at BMC Cancer and Frontiers in Immunology. None of these journals was involved in the journal’s review of, or decisions related to, this manuscript.

Supplementary information

The online version contains supplementary material available at 10.1038/s41541-026-01423-x.