Therapeutic vaccines for HPV-associated neoplasia: a narrative review and reappraisal of ongoing clinical development pipelines

Introduction and historical context

Persistent infection with oncogenic (high-risk) human papillomaviruses (HPV) is a well-known cause of most cervical cancers and contributes substantially to vulvar, vaginal, anal, penile and oropharyngeal malignancies.1 2 Prophylactic L1 virus-like-particle vaccines are highly effective at preventing new infections, but they do not eradicate established infection or treat HPV-transformed cells.1 3 That biological gap created the rationale for therapeutic vaccination: unlike many non-viral tumours, HPV-associated neoplasia constitutively expresses foreign viral antigens that can, in principle, be recognised by T cells.3 The therapeutic objective is therefore not neutralising antibody production but durable cellular immunity capable of eliminating infected or transformed cells. The distinction is especially important in gynaecology, where a successful therapeutic vaccine could occupy several niches—from persistent high-risk HPV infection and cervical intraepithelial neoplasia (CIN) to recurrent or metastatic cervical cancer—each with a different balance of tumour burden, immune suppression, acceptable toxicity and clinical endpoints.2–4

When we first reviewed therapeutic DNA vaccines in late 2021, the field contained multiple promising clinical leads but also faced recurrent barriers: a narrow focus on HPV E6/E7 antigen selection, limited potency of naked DNA, dependence on electroporation, incomplete understanding of biomarkers of immune protection and the likelihood that advanced cancers would require combination immunotherapy.3 Five years have gone by, several of those predictions can be reappraised rather than simply repeated. Therapeutic vaccination still falls short of becoming standard care, but the evidentiary landscape has changed. Randomised and single-arm studies in cervical precancer now provide clinically interpretable lesion-regression endpoints; mature phase II studies in recurrent or advanced cervical cancer report objective responses and survival following vaccine–checkpoint combinations; and HPV-positive head and neck squamous cell carcinoma (HNSCC) has become an informative parallel disease model in which both encouraging and negative findings test assumptions about immune priming.5–10

Preclinical and clinical developments have also broadened technologically. DNA vaccines remain the most clinically mature nucleic-acid platform, but recent preclinical work has drawn attention to lipid nanoparticle (LNP)-delivered DNA, conventional messenger RNA (mRNA), self-amplifying RNA and circular RNA. These technologies permit rapid redesign, multivalent antigen expression, incorporation of immune modulators and delivery without a permanent viral vector. In parallel, antigen engineering has moved beyond simple expression of wild-type E6 and E7 toward detoxified sequences, fusion antigens, intracellular trafficking signals, antigen-presenting-cell (APC) targeting and—in selected preclinical programmes—extension to early proteins such as E2 (figure 1). The overall progression has been made through a series of linked bottlenecks rather than competing vaccine ‘generations’, as improvements in antigen, vector, delivery and immune context are interdependent; optimising only one component is unlikely to overcome the biology of established HPV-associated neoplasia.

The clinical context matters equally. Most studies focus on precancer patients withCIN2/3, which have low disease burden, an intact lymphatic environment, measurable virologic endpoints and time for vaccine-induced immunity to develop. By contrast, advanced and metastatic cancers are immunologically and therapeutically distinct in terms of exhausted T cells, immunosuppressive myeloid populations, heterogeneous antigen expression and multiple mechanisms of immune escape. Thus, the same vaccine can appear effective in a premalignant setting yet insufficient as monotherapy in metastatic stage. This helps explain why current development is bifurcating: non-surgical therapeutic vaccination is being tested as an alternative or adjunct to excision for high-grade lesions, whereas cancer programmes increasingly use vaccination as an immune primer that needs to be combined with immune checkpoint blockade (anti-PD-1/PD-L1) or other interventions that alter tumour microenvironment (TME).2 11 12

This narrative review is organised around the progression in technological and translational bottlenecks: antigen design, vector engineering, delivery technologies, immune modulation and clinical translation (figure 1). The emphasis is deliberately placed on nucleic-acid vaccines because they offer the clearest and most versatile path to iterative antigen engineering and scalable combination design. Peptide, bacterial and viral-vector vaccines are also discussed wherever their clinical results answer questions that nucleic-acid programmes must also confront. One central notion is clear: therapeutic HPV vaccination has moved beyond proof of immunogenicity, but not yet beyond the need for careful disease-stage selection, biomarker-guided combination therapy and randomised evidence based on adequate sample sizes.

Antigen design: from dependence on HPV E6/E7 to safer and broader immunogens

Antigen selection is the first constraint on every therapeutic HPV vaccine. To date, HPV E6 and E7 antigens have been compelling targets because they are constitutively expressed in HPV-driven malignant cells, and they are functionally required for maintenance of the transformed phenotype. Their viral origin also makes them more immunogenic than most self-derived cancer-associated antigens. These advantages explain why nearly all clinically advanced programmes—including VGX-3100, GX-188E, VB10.16, Vvax001, ISA101/ISA101b, BNT113 and HB-200—retain E6 and/or E7 as core antigens (figure 1 and table 1). The persistence of this design is not simply historical inertia: clinical responses and HPV-specific T-cell expansion demonstrate that E6/E7 can generate clinically relevant HPV-specific cellular immunity.5 8 9

Sole reliance on E6/E7 antigens inevitably compresses a biologically diverse infection into a small antigenic target, with a rather small epitope repertoire.3 13 A design of tandem E6/E7 repeats in the same vaccine construct might help with the processing of certain epitopes, at least in preclinical models.14 Nonetheless, HPV-transformed lesions vary in viral integration, transcriptional programmes, human leucocyte antigen (HLA) genotype, antigen processing, and immunoediting. A broader antigen repertoire could increase the number of targetable epitopes and reduce dependence on a small set of immunodominant responses. To this end, E1 and E2 are particularly attractive in earlier HPV-associated disease because they are expressed during viral replication and may provide additional targets before complete malignant transformation. However, viral integration frequently disrupts E1/E2 expression in invasive cancers, making their value stage-dependent. Thus, broadening antigen content should not be treated as universally superior; rather, it should be matched to the biology of persistent infection, precancer, or established malignancy.15 16

Recent mRNA studies provide direct experimental support for this broader design space17: mHTV-03E2, an LNP-formulated mRNA vaccine encoding E2, E6, and E7 from both HPV16 and HPV18, was capable of inducing strong antigen-specific cellular responses, intratumoural CD8⁺ T-cell infiltration, tumour regression, prolonged survival, and immune memory in preclinical models (table 2). The importance of this work is not that E2 has already been clinically validated—it has not—but that mRNA permits multiple antigens and trafficking elements to be assembled in a single manufacturable construct. Such modularity changes the practical cost of testing antigen breadth. A multiantigen hypothesis that would require several recombinant proteins or vectors in the past can now be evaluated through sequence redesign.

Safety engineering is equally important because E6 and E7 are themselves oncogenic proteins. Therapeutic constructs should preserve immunogenic epitopes while minimising transforming activity. Strategies include sequence rearrangement, point mutation, fusion of E6/E7 fragments and disruption of functional domains involved in p53 or retinoblastoma-protein pathways. Recent mRNA programmes use modified E6/E7 sequences rather than relying on native oncoprotein expression, and APC-targeted designs further couple detoxified antigens to trafficking or receptor-binding domains.17 18 This principle is particularly relevant for RNA platforms, where high transient expression is desirable immunologically but makes careful protein engineering essential.

Antigen architecture can also determine where and how the antigen is presented. Fusion to heat-shock proteins, secretion signals, Flt3 ligand or APC-targeting domains can enhance uptake and cross-presentation. VB10.16 provides a clinical example: its DNA-encoded Vaccibody construct is designed to direct HPV antigen to APCs rather than relying solely on intracellular antigen expression. Preclinical mRNA constructs have similarly incorporated targeting domains to improve dendritic-cell engagement. These designs blur the boundary between ‘antigen’ and ‘adjuvant’; the encoded molecule itself can determine antigen localisation, APC interaction and the quality of the resulting T-cell response (table 1).

The post-2021 evidence therefore supports a nuanced reappraisal, with HPV E6/E7 remaining as the most clinically validated targets. The main advance is instead the ability to engineer them more safely, combine them with additional HPV antigens and embed trafficking or APC-targeting functions in the same nucleic-acid product. The next decisive experiments should compare antigen breadth directly rather than infer benefit from immunogenicity alone. Trials should determine whether E2/E1-containing constructs improve lesion regression or cancer control, in which disease stages those antigens remain expressed and whether broader antigen coverage reduces immune escape without diluting dominant E6/E7 responses. Until such data emerge, antigen breadth is a promising design variable—not a settled clinical solution.

Vector engineering: expression, safety and manufacturability

For nucleic-acid vaccines, the vector is not just a passive container. Expression level, duration, innate sensing, plasmid backbone composition, RNA stability and subcellular localisation all shape immunogenicity. The first generation of therapeutic HPV DNA vaccines established the practical virtues of plasmid DNA—chemical stability, relatively simple manufacturing, repeat dosing and capacity for multigene constructs—but also exposed a central weakness: efficient expression in vivo is difficult without an enabling delivery technology.19 Consequently, vector engineering and delivery have evolved together.

DNA optimisation has historically included mammalian codon optimisation, strong eukaryotic promoters such as the cytomegalovirus immediate-early promoter, optimised Kozak sequences, removal of unnecessary bacterial sequences and fusion of antigens to immunologic targeting domains. Contemporary plasmid design increasingly aims to reduce backbone size and bacterial elements while maintaining high-copy manufacturing and robust mammalian expression. Nanoplasmid and minimal-backbone concepts are attractive because smaller constructs may improve transfection and reduce nonessential sequences, although direct HPV-specific clinical comparisons remain sparse. This is an area where the technological rationale is stronger than the disease-specific evidence base, and the review should distinguish those two levels of support.

The same logic now applies to RNA vaccines. Optimisation includes 5′ cap structure, untranslated regions, poly(A) length, codon and sequence engineering, modified nucleosides and purification strategies that limit unwanted double-stranded RNA. Circular RNA and self-amplifying RNA extend the design space by altering expression kinetics rather than merely increasing dose, as illustrated by an LNP-encapsulated circular RNA encoding fused HPV16 E7/E6, which generated functional antigen-specific cytotoxic T cells, remodelled the TME, produced tumour regression and sensitised non-inflamed tumours to PD-L1 blockade in mice.20 Such findings are again encouraging, but they are at a preclinical phase and should not be interpreted as evidence that circular RNA is clinically superior to linear mRNA for downstream translation.

Vector engineering should therefore be judged by translational outputs rather than expression alone: sufficient antigen production, appropriate innate stimulation, minimal oncogenic potential, reproducible manufacturing and compatibility with repeated administration. The field has gained a richer engineering toolkit, but HPV-specific head-to-head evidence is still lacking. This bottleneck is being solved incrementally and will likely become most visible when optimised vectors are paired with modern carriers and tested under real-world conditions.

Delivery technologies: from electroporation to LNPs

Delivery has been the defining practical bottleneck for therapeutic DNA vaccination. Naked plasmid DNA is stable and easy to manufacture but enters cells inefficiently after conventional intramuscular injection. Electroporation addressed this limitation by transiently increasing membrane permeability and local gene expression, and it enabled the clinical development of VGX-3100 and other DNA candidates. Its immunologic effectiveness is not in doubt: recent preclinical comparisons again show that intramuscular DNA followed by electroporation produces substantially greater antigen expression and HPV-specific cellular immunity than needle injection alone or needle-free jet injection.14 The problem is translational friction—device requirements, operator training, procedure time, local discomfort and the challenge of deploying repeated treatment at scale.

The clinical history of VGX-3100 illustrates both the achievement and the limitation of electroporation. The platform progressed to phase III cervical-high-grade squamous intraepithelial lesion (HSIL) trials, demonstrating that device-assisted DNA vaccination can reach late-stage clinical development. Yet the US programme did not obtain a straightforward registration path, and further US development was halted after additional controlled evidence was requested in a biomarker-defined population.21 In China, the related pivotal HPV-303CHN programme reportedly met its primary composite endpoint in 2026, pending detailed numerical results and peer-reviewed publication.22 These divergent trajectories should not be attributed to delivery alone, but they reinforce the need for simpler administration if therapeutic vaccination is ultimately intended for broad use in precancer.

LNPs now offer the most consequential alternative. Their success in licensed RNA products and vaccines has established scalable manufacturing principles and a clinically familiar carrier class. In HPV models, LNPs are being used not only for mRNA but also for plasmid DNA. Li et al directly compared naked DNA, electroporation and several ionizable LNP formulations for HPV16/18 E6/E7 plasmids. Naked intramuscular DNA generated little response; electroporation improved cellular immunity; and LNP-encapsulated DNA—particularly an SM-102 formulation—produced stronger HPV-specific cellular responses in mice.14 This conceptually important observation suggests that DNA need not remain tied to an electroporation device. However, as a preclinical immunogenicity study, this new line of evidence requires further validation in terms of tumour efficacy, biodistribution, plasmid persistence, toxicology and manufacturability of DNA-LNP formulations.

For mRNA, LNP delivery is already intrinsic to the platform. Recent HPV16/18 E6/E7 mRNA-LNP vaccines have induced strong CD8-dependent antitumour immunity and memory in mouse models, with route of administration affecting efficacy.23 More elaborate constructs add E2, APC-targeting domains or immune modulators without changing the basic manufacturing logic. Circular RNA-LNP further extends the platform by seeking prolonged antigen expression and improved stability.20 The practical attraction is clear: an injectable formulation can potentially replace an electroporation procedure while allowing rapid sequence updates and multivalent constructs.

Yet LNPs create their own optimisation problem. Ionizable lipid chemistry influences tissue distribution, cellular uptake, endosomal escape, reactogenicity and transfected APC populations of interest. Formulations optimised for prophylactic intramuscular mRNA vaccination may not be optimal for repeated therapeutic dosing or tumour-directed (intratumoural) immunity. Systemic delivery can also shift biodistribution toward the liver, whereas intramuscular or intratumoural administration may favour different immune compartments. The mHTV-02 study, for example, reported therapeutic benefit after intramuscular or intratumoural administration but little benefit with intravenous delivery, emphasising that ‘LNP delivery’ is not a single biological intervention.23

The delivery roadmap should therefore be viewed as convergence rather than replacement. Electroporation remains a validated benchmark for DNA potency and has generated the deepest clinical experience. Polymeric and other nonviral carriers may still have roles where local or mucosal delivery is desired. LNPs, however, provide the strongest current route toward a device-free nucleic-acid platform that can accommodate DNA, linear mRNA, self-amplifying RNA or circular RNA. A critical next step is clinical—not another demonstration that LNPs can produce T-cell responses in mice, but evidence that a scalable formulation improves lesion regression, cancer response, patient acceptability or dosing feasibility relative to established delivery approaches.

Immune modulation: making vaccine-induced T cells function in TME

The most important conceptual shift since 2021 is the recognition that generating HPV-specific T cells is necessary but often insufficient for therapeutic efficacy. Established cancers can fail at several downstream steps: APCs may be poorly activated, vaccine-expanded T cells may not reach tumour nests, and infiltrating lymphocytes may be restrained by (1) checkpoint signalling, (2) suppressive myeloid cells, (3) regulatory T cells or (4) stromal barriers. Immune modulation should therefore be viewed as a problem of matching the intervention to the underlying immune defect rather than simply increasing the magnitude of a peripheral T-cell response.

Two mechanistically distinct classes of immune modulation are worth noting. The first is encoded or vaccine-localised immunostimulation. Nucleic-acid constructs can incorporate cytokines, chemokines, heat-shock proteins, Flt3 ligand, CD40-related signals or APC-targeting domains together with the HPV immunogen. These elements are intended to act mainly during antigen expression and presentation, enhancing dendritic-cell recruitment or activation, cross-presentation and T-cell priming without requiring sustained systemic exposure (figure 1). Their appeal is modularity and temporal linkage to vaccination, while their limitations include construct complexity, uncertain dose relationships between antigen and immune modulator and the possibility that stronger innate stimulation may increase reactogenicity without improving tumour control. The relevant benchmark is therefore a demonstrable pharmacodynamic gain in priming or trafficking, not the number of functions encoded in the vaccine construct.

The second class comprises systemic or tumour-microenvironment-modifying therapies administered separately from the vaccine. PD-1/PD-L1 blockade, TGF-β pathway inhibition, IL-12-based immunocytokines, epigenetic modulators and related agents act beyond the injection site and can alter inhibitory signalling, stromal restraint or myeloid suppression over a broad spatial and temporal range. Their role is complementary to vaccination: vaccine supplies or expands HPV-specific T cells, whereas the partner therapy attempts to remove key barriers to immune function. Clinical examples include checkpoint combinations with GX-188E, VB10.16 and ISA101/ISA101b, as well as PDS0101- and PRGN-2009-based regimens that incorporate IL-12 or PD-L1/TGF-β-directed modulation.5 8–10 24 25 Trial-level efficacy data are discussed in the Clinical Translation section and summarised in tables 1 and 2; mechanistically, these studies show why immune priming and tumour conditioning should not be treated as interchangeable interventions.

A useful framework is to distinguish TME by the barrier that is most likely to limit vaccine efficacy. In an inflamed or T-cell-infiltrated tumour, HPV-specific lymphocytes are already present, but adaptive inhibitory pathways such as PD-1/PD-L1 may restrain T-cell function; vaccination may broaden or amplify relevant clones, while checkpoint blockade can release an existing brake. The immune-correlate findings from ISA101 plus nivolumab, in which activated T-cell and interferon-related signatures were associated with benefit, are consistent with this model.5 In an immune-excluded tumour, T cells may accumulate at the invasive margin or in stromal compartments without effectively entering the tumour nests. In that setting, simply generating additional circulating T cells is unlikely to be sufficient; intratumoural delivery strategies or combination partners that improve trafficking, stromal access or local inflammatory signalling are critical. A third pattern is a strongly immunosuppressive, myeloid-rich TME, in which myeloid-derived suppressor cells, suppressive macrophages, TGF-β-related signalling or other inhibitory networks can blunt both priming and effector function. Such tumours provide a strong rationale for therapeutic combinations that directly modify myeloid or TGF-β-dominant suppression rather than empiric addition of checkpoint blockade alone.11 12 16 24–26 These categories overlap and can evolve during treatment or disease course, but they offer a more useful trial-design framework than a single ‘hot-versus-cold’ label of TME.

The immune modulation framework also links vaccine platforms and delivery choices to TME context. A platform that produces strong local innate activation and APC engagement may be advantageous when priming is weak, whereas a highly immunogenic vaccine cannot by itself overcome physical exclusion or existing systemic suppression. Intramuscular electroporation and LNP delivery may differ not only in convenience but also in the cells transfected, the inflammatory signals generated, and the anatomical compartments in which immune priming occurs. Intratumoural delivery, wherever feasible, may add local inflammatory effects, whereas systemic delivery introduces different biodistribution constraints. Thus, platform, route, and combination partner should be co-designed around a testable immune deficit rather than optimised independently.

Biomarker selection follows directly from this mechanistic view. PD-L1 alone is unlikely to capture the full range of resistance states. Future studies may need to integrate viral genotype, HLA type, circulating and intratumoural HPV-specific T-cell clonotypes, baseline immune infiltration and spatial distribution, interferon signatures, PD-L1, circulating tumour HPV DNA, disease burden and—wherever feasible—myeloid and stromal features. Longitudinal sampling is especially informative because therapeutic vaccination provides a defined antigenic perturbation: paired blood and tumour analyses can distinguish inadequate priming from failed trafficking, antigen loss, T-cell dysfunction or persistent local suppression. Such pharmacodynamic classification could even make negative trial results informative by identifying the failing step in any given vaccine-to-tumour pathway.

Disease stage may be the most powerful modifier of immune and TME architecture. CIN and minimal residual disease generally impose less tumour burden and may require less extensive TME remodelling than bulky recurrent or metastatic cancer. In earlier disease, effective priming and tissue access may be sufficient for lesion regression; in advanced disease, combination therapy should increasingly be selected according to the dominant resistance mechanism (figure 1). The future of immune modulation is therefore not 'more adjuvants’ or universal checkpoint combinations, but mechanistically matched interventions that connect vaccine design, route of delivery, TME phenotype and pharmacodynamic biomarkers to an appropriate clinical setting.

Clinical translation: what has changed recently?

Recent progress in clinical translation, including trials in patients with cervical neoplasia and HPV-positive HNSCC (table 1), is now sufficiently mature to offer broad optimism with a more discriminating framework. The central question is no longer whether therapeutic HPV vaccination can induce immune responses; numerous platforms can achieve this goal. The question is where those responses produce clinically meaningful benefit and whether that benefit provides substantial added value to justify the complexity of vaccination relative to surgery, checkpoint blockade or other available standards of care in the context of broader HPV-associated cancer programmes (table 2).

For cervical precancer, the therapeutic goal is unusually tangible: eradicate or regress a high-grade lesion and clear the causative high-risk HPV while avoiding excisional morbidity. A 2023 systematic review and meta-analysis reported higher pooled CIN2/3 regression and high-risk HPV clearance with therapeutic vaccination than with controls, but the underlying trials were heterogeneous and the search ended in January 2022.4 More recent studies provide distinct proof-of-concept signals. In a randomised placebo-controlled trial of the oral Lacticaseibacillus paracasei vaccine IGMKK16E7 in HPV16-positive CIN2/3, high-dose vaccination produced complete histopathologic regression to normal in 31.7% vs 12.5% with placebo, with stronger E7-specific cellular responses associated with regression.6 Although this is not a nucleic-acid vaccine, it demonstrates that a therapeutic immune intervention can alter the natural history of high-grade cervical disease in a randomised setting.

Vvax001 provides a second clinically informative signal. In 18 patients with HPV16-positive CIN3, the replication-defective Semliki Forest virus vaccine produced histopathologic complete response in 9 of 18 and HPV16 clearance in 10 of 16 evaluable patients, with no serious adverse events and no reported recurrences at a median follow-up of 20 months.7 The small uncontrolled design precludes definitive efficacy conclusions, but the magnitude of lesion regression supports continued study of vaccination as a non-surgical strategy. For gynaecologic practice, this is potentially transformative: excision is effective, but repeated procedures can affect cervical anatomy and future obstetric outcomes. A therapeutic vaccine would need to match this clinical context with high regression rates, reliable exclusion of occult invasion, acceptable time to response and rigorous surveillance rather than merely demonstrate immunogenicity.

VGX-3100 remains the most instructive DNA-vaccine case because it reached phase III development for HPV16/18-positive cervical HSIL. Its U.S. trajectory shows that statistically positive biological activity does not guarantee a straightforward regulatory pathway. After the REVEAL programme, further US development was stopped in 2023 when additional controlled evidence was required in a biomarker-defined population.21 In contrast, the China HPV-303CHN pivotal programme reportedly met its primary composite endpoint in May 2026, with a regulatory submission planned. Until full numerical data and peer-reviewed results are available, the outcome seen in China should be presented as a pipeline development rather than established efficacy. The geographic divergence nevertheless underscores an important translational reality: therapeutic vaccines will be judged within different screening systems, treatment alternatives, regulatory expectations and unmet needs.

In advanced cervical cancer, the field has crossed a different threshold. In the final phase II report, GX-188E plus pembrolizumab produced a confirmed objective response rate (ORR) of 35.0% among 60 efficacy-evaluable patients, including complete responses, a median response duration of 12.3 months and median overall survival of 23.8 months; responses occurred in both PD-L1-positive and PD-L1-negative tumours.8 VB10.16 plus atezolizumab independently produced an ORR of 19.1% in 47 response-evaluable patients, rising to 29.2% in PD-L1-positive disease, with HPV16-specific T-cell responses increasing in 22 of 36 evaluable patients.9 These studies substantially narrow the clinical-efficacy gap identified in 2021 and support DNA vaccination as an immune primer in combination therapy. However, both were single-arm studies, so neither establishes the incremental contribution of vaccination over checkpoint blockade alone. The next generation of cervical-cancer trials should be designed to measure that added benefit, not merely whether the combination is active.

The HNSCC experience makes this requirement unavoidable. The updated ISA101 plus nivolumab study reported an ORR of 33% in incurable HPV16-positive cancers, durable responses in a subset and associations between benefit and baseline activated T-cell/interferon-related immune signatures.5 The later randomised phase II study of peltopepimut-S (ISA101b) plus cemiplimab vs cemiplimab alone in recurrent/metastatic HPV16-positive oropharyngeal cancer did not improve the overall-population outcome: response rates were similar and overall survival was not improved, although a signal in the PD-L1 combined positive score (CPS) CPS≥20 subgroup contrasted with unfavourable outcomes in CPS<20 disease.10 This negative randomised result is as informative as the earlier single-arm signal. It argues against assuming that any HPV-specific vaccine will synergise with any PD-1 inhibitor and instead supports biomarker- and immune-context-guided combinations. HNSCC therefore remains useful as a translational test bed for principles that directly inform cervical-cancer development and other gynaecology-anchored goals.

The next clinical transition is towards mRNA. BNT113, encoding HPV16 E6/E7, is being evaluated with pembrolizumab in the randomised AHEAD-MERIT programme for recurrent/metastatic HPV16-positive HNSCC.26 Other clinically advanced platforms include HB-200, an arenavirus-vector programme evaluated alone and in combination with pembrolizumab in HPV16-positive recurrent/metastatic cancers.27 As of the evidence set used here, mature peer-reviewed efficacy data are not yet available. Preclinical HPV mRNA-LNP studies, however, now show robust CD8-dependent tumour control, memory, multivalent HPV16/18 antigen expression, and compatibility with checkpoint blockade.17 18 Circular RNA extends this trajectory experimentally.20 These advances justify a focused RNA/LNP section, but not a claim that mRNA or circRNA has clinically surpassed DNA, which already has human efficacy signals; RNA currently has greater engineering flexibility and a strong preclinical imperative.

Overall, a rational development strategy may require the separation of three clinical niches. First, in CIN2/3 or persistent high-risk HPV, the goal is durable virologic and histologic clearance with minimal toxicity and simple outpatient delivery. Randomised trials need to compare vaccination with observation or accepted management pathways, stratify by genotype and lesion characteristics, and include reproductive outcomes where relevant. Second, in minimal residual disease after definitive therapy, vaccination could exploit low tumour burden to eradicate microscopic HPV-positive clones; circulating tumour HPV DNA may provide a particularly useful enrichment and monitoring biomarker. Third, in recurrent/metastatic cancer, vaccination should be embedded in biomarker-driven combination regimens that include an appropriate control arm and serial immune sampling to provide correlates of protection.

Manufacturing and implementation must also enter the clinical equation earlier. A therapeutic vaccine for cervical precancer could ultimately be used in settings where screening coverage and surgical access vary widely. Device-free formulations, thermostability, dose number, cold-chain requirements, and the ability to cover multiple high-risk genotypes will affect public-health value. DNA retains advantages in stability and manufacturing simplicity, whereas RNA/LNP offers rapid redesign and potent expression but imposes formulation and storage considerations. The optimal product may therefore differ between an oncology centre treating metastatic disease and a population-level programme managing screen-detected precancer.2

Finally, clinical endpoints should reflect mechanisms of action. In precancer, lesion regression and type-specific HPV clearance are directly relevant, but durability is essential because transient histologic improvement is insufficient. In cancer, ORR is useful but should be complemented by duration of response, progression-free and overall survival, and patient-reported outcomes. Immunological endpoints should be prespecified as mechanistic biomarkers rather than treated as surrogate efficacy. The field has accumulated many examples of strong T-cell responses without regulatory success. Translation will depend on demonstrating that a defined immune change mediates a clinically meaningful outcome in the intended patient population.

Conclusions and future directions

New evidence as summarised above now supports a more optimistic but also more disciplined view of therapeutic HPV vaccination. Several limitations identified in late 2021 now have clearer paths towards mitigation, although none can be considered fully resolved. Antigen engineering is safer and more modular; nucleic-acid vectors can encode multiple antigens and targeting functions; LNPs offer a credible path beyond electroporation; and vaccine-induced T cells can be pharmacologically supported with checkpoint blockade and other immune modulators. Encouragingly, clinical studies now document lesion regression in cervical precancer and objective responses with survival follow-up in advanced cervical cancer. These are substantive advances, not merely new platforms.

At the same time, the central uncertainties have become clearer. E6/E7 remain the clinical backbone, while the benefit of adding E1, E2 or other antigens remains largely preclinical. LNP delivery is compelling but has not yet shown superior clinical outcomes in HPV-related diseases. Vaccine–checkpoint combinations can be active, but the randomised HNSCC experience demonstrates that biological plausibility and single-arm response rates are not enough. Biomarkers, disease stage, and control-arm selection are now as important as vaccine potency. No therapeutic HPV vaccine is yet standard care, and regional regulatory divergence should not be mistaken for resolution of the efficacy question.

The most promising roadmap is therefore an integrated one (figure 1): engineer non-oncogenic, sufficiently broad antigens; optimise compact and manufacturable nucleic-acid vectors; use delivery systems that achieve APC access without excessive procedural burden; pair vaccination with immune modulation only when supported by tumour biology; and test the resulting product in the disease stage where an induced T-cell response has the greatest chance of changing outcome. Nucleic-acid vaccines—particularly RNA/LNP systems—are well suited to this iterative strategy because sequence, antigen breadth, targeting, and immune-modulatory features can be changed rapidly. Their promise, however, should be measured against clinical endpoints rather than platform enthusiasm.

A licensed therapeutic HPV vaccine is still elusive and may still require several years of further development, consistent with our cautious timeline proposed in 2021. What has changed is the path towards that goal. The field is no longer waiting for a single breakthrough vaccine platform. It is learning how to assemble antigen design, delivery, immune context, biomarkers, and clinical setting into a coherent therapeutic system. If that integration process is successful, therapeutic vaccination could reduce HPV-associated morbidity and mortality across a continuum from high-grade precancer to recurrent cancer while complementing, rather than replacing, prophylactic vaccination and screening.