Personalized neoantigen vaccine and pembrolizumab in advanced hepatocellular carcinoma: a phase 1/2 trial

Hepatocellular carcinoma (HCC) is the most common primary liver cancer and a leading cause of cancer mortality, with a 5-year survival rate under 10%. Advanced HCC is relatively immune resistant, often with low T cell infiltration and modest tumor mutational burden (TMB). Immune checkpoint inhibitors (ICIs) targeting PD-1 yield response rates of roughly 12–18% as monotherapy in HCC. Tumor-specific mutation-associated neoantigens (MANAs) can be leveraged to elicit anti-tumor immunity; patients with preexisting neoantigen-specific immunity frequently respond better to ICIs. Building on next-generation sequencing advances, personalized therapeutic cancer vaccines (PTCVs) can target an individual’s MANAs to induce or broaden anti-tumor T cell responses. The central hypothesis of this trial is that adding a DNA-based PTCV encoding up to 40 neoantigens (with a plasmid IL-12 adjuvant) to pembrolizumab can prime neoantigen-specific CD4+/CD8+ T cells, drive their expansion and tumor infiltration, and thereby improve clinical responses in advanced HCC beyond what is seen with PD-1 blockade alone.

Preclinical studies have shown that vaccines targeting MANAs can elicit tumor-specific T cell responses and impede tumor growth. Early clinical trials of various PTCV platforms (DNA, RNA, viral vectors) demonstrated induction of neoantigen-specific immunity in patients across tumor types, and a recent randomized phase 2b study (KEYNOTE-942) in resected melanoma showed improved outcomes when an mRNA neoantigen vaccine was combined with pembrolizumab versus pembrolizumab alone. In HCC specifically, prior registrational trials of PD-1/PD-L1 inhibitors as mono- or combination therapy reported ORRs of ~12–18% in second-line settings. Biomarkers like TMB, PD-L1, and T cell–inflamed gene expression profiles have variably correlated with response to ICIs across cancers, but robust predictors in HCC remain limited. These observations provide rationale for combining PD-1 blockade with PTCVs to generate de novo antitumor T cells in immune-cold tumors such as HCC.

Design: Single-arm, open-label, multicenter phase 1/2 trial (GT-30) in 36 patients with advanced HCC previously treated with a multityrosine kinase inhibitor (sorafenib or lenvatinib). Primary endpoints: safety/tolerability (CTCAE v5.0) and immunogenicity (IFNγ ELISpot in PBMCs). Secondary endpoints: objective response rate (ORR; RECIST v1.1 by investigator), progression-free survival (PFS), overall survival (OS), and feasibility. Exploratory endpoints: tumor/immune biomarkers, TCR repertoire dynamics, single-cell phenotyping, ctDNA kinetics. Data cutoff: August 18, 2023. Eligibility: Adults (≥18 years) with confirmed HCC, BCLC stage B or C, Child-Pugh A, ECOG 0–1, measurable disease (RECIST 1.1), life expectancy >6 months. Key exclusion: prior systemic therapy other than sorafenib/lenvatinib; active autoimmune disease. Interventions: Personalized DNA vaccine GNOS-PV02 (1 mg) encoding up to 40 neoantigens plus plasmid IL-12 adjuvant (pIL12, 0.34 mg) administered intradermally into both deltoids with in vivo electroporation (CELLECTRA 2000). Dosing: every 3 weeks (Q3w) for four doses, then every 9 weeks (Q9w) until year 2, then Q12w thereafter. Pembrolizumab 200 mg IV Q3w for up to 2 years or until progression, unacceptable toxicity, withdrawal, or study end. Vaccine design/manufacture: Tumor-normal exome and tumor RNAseq (Personalis ACE) identified somatic nonsynonymous variants (SNVs, indels, fusions) with DNA allelic fraction >0.05 and RNA FPKM >1. Predicted HLA-I epitopes (NetMHCpan 4.0) were ranked by binding affinity; up to 40 top candidates were included. Constructs encode strings of ~33-mer epitopes (including variant and flanking sequences) separated by synthetic furin cleavage sites, center-embedding CD8 epitopes; codon/RNA optimized and cloned into pGX0001. IL-12 DNA plasmid expressed p35 and p40 subunits from a dual-promoter vector. Plasmids were GMP manufactured and sequence verified. Immunogenicity assays: Ex vivo IFNγ ELISpot in PBMCs without cytokine stimulation; positivity defined as ≥2 SD above background, ≥2-fold above background, and ≥5 SFU per 10^6 cells. Intracellular cytokine staining after in vitro stimulation with patient-specific peptide pools assessed activation (CD69, CD107a), proliferation (Ki67), and cytokines (IFNγ, TNF, IL-2) in CD4+/CD8+ T cells; cytolytic potential via GZMA and PRF1. TCRβ CDR3 immunoSEQ profiling (Adaptive) in paired pre/post PBMCs and tumor biopsies assessed clonal expansion, clonality, richness, and trafficking to tumor. Single-cell RNAseq/TCRseq (10x Genomics) on week 12 PBMCs characterized T cell phenotypes and mapped expanded clones. ctDNA: Personalized targeted panels based on tumor mutations evaluated cfDNA at baseline, week 3, week 9, and thereafter; molecular response defined as >50% reduction from baseline. Imaging-based responses per RECIST v1.1 with first assessment at week 9. Statistical plan powered to detect ORR improvement over historical pembrolizumab ORR (16.9%) using one-sided exact binomial test (α=0.10).

Population and treatment exposure: 36 patients enrolled (median age 66.5 years; 69.4% male). All Child-Pugh A; ECOG 0/1: 69.4%/30.6%. BCLC B/C: 50%/50%. Prior 1L TKI: lenvatinib 91.6%, sorafenib 5.6%, both 2.8%. Median number of vaccinations 5 (range 1–18); median treatment duration 6.1 months. Median follow-up 21.5 months at cutoff. Safety: Treatment well tolerated; no dose-limiting toxicities or treatment-related grade ≥3 AEs. TRAEs were mainly grade 1/2 and occurred in 75.0% (27/36). Injection-site reactions were most common (41.6%). Immune-related AEs requiring systemic steroids occurred in 3 patients (8.3%): grade 2 nephritis, pneumonitis, and hepatitis. One patient (2.8%) discontinued pembrolizumab due to AE; none discontinued PTCV. Efficacy: mITT ORR (RECIST 1.1; confirmed + unconfirmed) 30.6% (11/36), including 3 CRs (8.3%) and 8 PRs (22.2%). Disease control rate 55.6% (20/36). Among 34 evaluable patients (≥1 on-treatment scan), two unevaluable due to unrelated SAEs were counted as nonresponders in mITT. Median time to response 9.3 weeks (range 8–46). Median PFS 4.2 months; median OS 19.9 months; median duration of response not reached. ORR significantly exceeded historical pembrolizumab control (16.9%) with one-sided P=0.031 (one-sided 90% CI, 20.4–100%). Clinical response (CR/PR vs SD/PD) associated with longer PFS/OS. ctDNA: In 13 patients with baseline and on-treatment cfDNA, percent ctDNA change at week 9 differed significantly between disease control (CR/PR/SD) and PD (P=0.006). A ctDNA decrease at week 9 associated with longer OS (P=0.01). In some cases, ctDNA suggested deeper molecular responses than imaging. Biomarkers: All patients had low TMB (median 2 mut/Mb); TMB did not differ between responders and nonresponders. Pretreatment PD-L1 (CD274), CD8A, KDR, and a T cell–inflamed GEP did not associate with response. Post hoc analysis found a positive correlation between the number of neoantigens encoded in the PTCV and clinical response (P=0.025); ≥30-neoantigen vaccines had ORR 41.2% vs 23.5% for <30. Immunogenicity and TCR dynamics: ELISpot in 22 patients showed increased cumulative neoantigen-specific T cell responses post-vaccination (P<0.0001); 19/22 (86.4%) had higher numbers of reactive epitopes post-treatment. Vaccine-encoded epitopes elicited responses to a median of 64% of encoded epitopes post-treatment (vs 11.8% pretreatment). Trends toward higher IFNγ response magnitude associating with longer OS and with CR/PR. Intracellular staining in selected responders showed increased activation/proliferation and cytolytic potential in both CD4+ and CD8+ T cells upon neoantigen stimulation. Bulk TCRβ sequencing in 14 patients revealed significant clonal expansion in all patients in blood and tumor; increased clonality in tumor post-treatment (P=0.035) without change in richness. Many newly expanded peripheral clones were detected in posttreatment tumors, indicating trafficking. Single-cell analyses showed expanded clones enriched in CD8+ effector memory (TEM) and proliferating compartments with cytotoxic phenotypes (GZMB, NKG7), less frequently preexhausted (GZMK). Engineering of high-frequency postvaccination TCRs confirmed specificity to vaccine-encoded neoantigens. Case study of escape: In a patient with initial PR and new adrenal lesion, strong ELISpot-reactive vaccine neoantigens were absent in the adrenal metastasis, consistent with immune editing and clonal escape; liver lesion showed robust CD8+ infiltration post-PTCV while adrenal lesion was less inflamed.

The combination of a personalized DNA neoantigen vaccine plus pembrolizumab demonstrated feasibility, acceptable safety, immunogenicity, and encouraging clinical activity in advanced HCC, a disease with limited responsiveness to PD-1 monotherapy. The observed ORR of 30.6% exceeds historical anti–PD-1 monotherapy benchmarks (12–18%) in similar settings, although causality cannot be definitively attributed due to the single-arm design. Immunologic analyses support the hypothesized mechanism: vaccination primed new neoantigen-specific T cell responses, broadened and expanded the TCR repertoire, and promoted trafficking of expanded clones into tumors, predominantly with cytotoxic CD8+ TEM phenotypes. Unlike reliance on preexisting immunity with PD-1 blockade, PTCV appears to generate de novo antitumor responses, potentially explaining responses in both inflamed and noninflamed tumors and the lack of predictive value for standard pretreatment biomarkers (PD-L1, TMB, T cell–inflamed GEP). The correlation between the number of encoded neoantigens and clinical response, and between immune response magnitude and survival trends, suggests antigen breadth and vaccine immunogenicity as key drivers of benefit. Evidence of immune editing and neoantigen loss in a case of acquired resistance highlights tumor heterogeneity as a challenge and potential role for iterative vaccine updates guided by longitudinal multi-lesion genomics and ctDNA monitoring.

A personalized multi-neoantigen DNA vaccine (GNOS-PV02) plus pembrolizumab is feasible, safe, and clinically active in advanced HCC, inducing robust neoantigen-specific CD4+ and CD8+ T cell responses, expansion and tumor infiltration of vaccine-enriched TCR clones, and an ORR that exceeds historical PD-1 monotherapy benchmarks. The data support a mechanism whereby PTCV primes de novo antitumor immunity synergizing with PD-1 blockade. Future work should include prospective randomized trials to confirm efficacy over PD-1 monotherapy, optimization of antigen selection and vaccine breadth, strategies to overcome tumor heterogeneity and immune editing (e.g., adaptive neoantigen panel updates), and integration with evolving first-line ICI combinations while maintaining a favorable safety profile.

Key limitations include the small sample size (n=36) and single-arm, open-label design without central radiology review, limiting causal inference and generalizability. Differences in study population versus historical controls may influence response comparisons. The rapidly evolving HCC treatment landscape (new first-line ICI combinations with improved survival) may constrain applicability. Biomarker analyses were exploratory and limited by available samples; some affiliations between immune metrics and outcomes showed trends rather than definitive associations. Additionally, mechanisms of resistance such as neoantigen loss and inter-lesional heterogeneity, observed in a case study, present challenges for durable control.