Mutant KRAS peptide vaccine with dual checkpoint blockade in metastatic colorectal cancer: a phase I trial Authors: Hejia Henry Wang, Amanda L. Huff, S. Daniel Haldar, Maureen Berg, Christopher Thoburn, Thatcher R. Heumann, Robert A. Anders, Benjamin Barrett, Katherine M. Bever, Michael J. Pishvaian, Valerie Lee, Dung T. Le, Eric S. Christenson, Marina Baretti, Mark Yarchoan, Daniel Laheru, Amy Thomas, Jennifer N. Durham, Julie M. Nauroth, Jiayun Lu, Hao Wang, Elizabeth M. Jaffee, Nilofer S. Azad, Neeha Zaidi Published: 2026 ABSTRACT Immune checkpoint inhibitors (ICIs) have limited activity in mismatch repair proficient or microsatellite stable (MMRp/MSS) colorectal cancer (CRC). KRAS mutations, present in approximately 40% of these cancers, can generate neoantigens that are targets for therapeutic vaccines. In this single-arm, phase I study (NCT04117087), we evaluated mKRAS-VAX, a pooled mutant KRAS (mKRAS) peptide vaccine targeting six KRAS mutations with nivolumab and ipilimumab in 13 patients with pretreated metastatic MMRp/MSS CRC. Both primary endpoints of safety and immunogenicity (within 17 weeks post-vaccination) were met. Secondary endpoints included treatment efficacy defined by RECIST v1.1 criteria. All adverse events attributed to mKRAS-VAX were grade 1 or 2, and the addition of mKRAS-VAX did not increase the frequency of severe immune-related adverse events beyond the expected profile of dual ICIs alone. mKRAS-VAX elicited an increase in tumor-specific mKRAS-reactive T-cells in 8/12 biomarker-evaluable patients (75%) by direct ex vivo IFNγ ELISpot and in 12 patients (100%) following in vitro expansion. Our findings support further development of mKRAS vaccines with ICIs for advanced MMRp/MSS CRC. Introduction Colorectal cancer (CRC) is the second leading cause of cancer deaths in the United States with a 5-year survival of approximately 13% in patients with metastatic disease. Immune checkpoint inhibitors (ICIs) have transformed the treatment of advanced mismatch repair deficient or microsatellite instability-high CRC, achieving durable responses and significant improvements in overall survival (OS)1–4. However, ICIs have limited efficacy in the ~95% of metastatic colorectal cancer (mCRC) that are mismatch repair proficient or microsatellite stable (MMRp/MSS), where chemotherapy and/or targeted therapy remain the treatment backbone3. Specifically, ICI monotherapy with anti-PD-(L)1 antibodies has a 0% objective response rate (ORR), while dual checkpoint blockade against both PD-(L)1 and CTLA-4 has shown minimal activity4. ORRs range from 1% with durvalumab (anti-PD-L1) and tremelimumab (anti-CTLA-4) in unselected populations to 10% with nivolumab (anti-PD-1) and ipilimumab (anti-CTLA-4) in the rare subset of MMRp/MSS mCRC with high tumor mutational burden (TMB)5–7. A phase I study of balstilimab (anti-PD-1) combined with botensilimab, a second generation anti-CTLA-4 antibody, reported the highest efficacy to date in chemorefractory MMRp/MSS mCRC, with an ORR of 17% in 101 patients, most of whom did not have liver metastases, with a randomized study ongoing in that setting8. Importantly, no objective responses have been observed with PD-1 and CTLA-4 blockade alone in patients with active liver metastases, a disease state increasingly recognized as being refractory to immunotherapy and on that represents >60% of patients with mCRC8–12. Taken together, these findings underscore the therapeutic need for additional agents that enhance the efficacy of ICIs in MMRp/MSS mCRC. Resistance to ICIs in MMRp/MSS CRC is driven by both a highly immunosuppressive tumor microenvironment and a low TMB, the combination of which limits the activation, infiltration, and effector function of tumor-reactive T-cells13–15. Therapeutic cancer vaccines offer a promising approach to expand T-cell clones reactive to the relatively few neoantigens expressed in MMRp/MSS CRC and have been shown to synergize with ICIs across a range of tumor types15–27. Oncogenic KRAS mutations, which occur in over 40% of CRC, have emerged as prime neoantigen targets because they arise early in tumorigenesis, reducing the likelihood of subclonal immune escape3,28,29. Adoptive T-cell therapies targeting KRAS G12D with HLA-C*08:02-restricted T-cells have led to tumor regression in mCRC in case reports, but its use has been hindered by logistical complexity and its limitation to specific mKRAS and HLA allele combinations30,31. In contrast, a long peptide vaccine targeting mKRAS can be off-the-shelf and expand mKRAS-specific T-cells in a HLA-agnostic manner. Therapeutic vaccines targeting mKRAS have been tested clinically, but only in limited contexts. A lymph node-targeted peptide vaccine against KRAS G12D and G12R showed evidence of immunogenicity in the high-risk adjuvant setting, primarily in patients with pancreatic ductal adenocarcinoma32. Conversely, a fixed vaccine encoding 20 public neoantigens—including four KRAS mutations—tested in combination with nivolumab and subcutaneous ipilimumab in patients with advanced solid tumors showed no objective responses and elicited low mKRAS-specific immune responses, thought to be due to immunodominance of TP53 neoantigens27. Thus, it remains unknown whether mKRAS-targeted vaccines can (1) reliably induce mKRAS-specific T-cells that traffic into tumors in the metastatic setting, and (2) lead to tumor regression when combined with ICIs in cancers, such as MMRp/MSS CRC, that are not typically responsive to immunotherapy. We developed a pooled synthetic long peptide vaccine, mKRAS-VAX, targeting six KRAS mutations (G12V, G12A, G12R, G12C, G12D, and G13D) that collectively occur in ~80% of CRC expressing mKRAS29. We conducted a phase I clinical trial (NCT04117087) evaluating mKRAS-VAX formulated with the adjuvant poly-ICLC (Hiltonol®) in combination with nivolumab and ipilimumab (“mKRAS-VAX combination”) in patients with heavily pretreated MMRp/MSS mCRC (Fig. 1A). We have previously reported on the adjuvant pancreatic adenocarcinoma cohort of this trial33. The primary endpoints were safety and immunogenicity of the mKRAS-VAX combination. Secondary and exploratory endpoints included antitumor activity and T-cell profiling in serial peripheral blood and paired tumor biopsy samples.Fig. 1Trial design and clinical outcomes.A Schema of trial design and administration of mKRAS-VAX combined with nivolumab and ipilimumab for heavily pretreated metastatic MMRp/MSS colorectal cancer. B Consort diagram. C Waterfall plot of the best change from baseline in the sum of target lesions per RECIST v1.1 (n = 12 evaluable patients). One patient withdrew shortly after receiving the first dose; a RECIST read was not available. * denotes a patient with progressive disease due to mixed response on initial on-treatment scan who remained on-treatment past initial progression. D Swimmer’s plot of clinical outcomes in efficacy-evaluable patients. Source data provided in Source Data File. Trial design and clinical outcomes. A Schema of trial design and administration of mKRAS-VAX combined with nivolumab and ipilimumab for heavily pretreated metastatic MMRp/MSS colorectal cancer. B Consort diagram. C Waterfall plot of the best change from baseline in the sum of target lesions per RECIST v1.1 (n = 12 evaluable patients). One patient withdrew shortly after receiving the first dose; a RECIST read was not available. * denotes a patient with progressive disease due to mixed response on initial on-treatment scan who remained on-treatment past initial progression. D Swimmer’s plot of clinical outcomes in efficacy-evaluable patients. Source data provided in Source Data File. Here we report that, in patients with heavily pretreated MMRp/MSS mCRC, including in those with liver metastases, the mKRAS-VAX combination is well tolerated and elicits peripheral mKRAS-reactive T-cells that traffic into tumors and, importantly, are associated with tumor regression. Results Patient characteristics, treatment, and safety assessments Between June 2021 and October 2024, we enrolled 14 patients with MMRp/MSS mCRC who had received prior fluoropyrimidine, irinotecan, and oxaliplatin chemotherapy in the metastatic setting. Baseline demographics and clinical characteristics of the efficacy-evaluable population are shown in Table 1. Tumors were confirmed to be MMRp/MSS and harbor a KRAS mutation included in mKRAS-VAX through molecular testing performed in a CLIA-certified laboratory (Table 1 and Supplementary Table 1). As expected in MMRp/MSS mCRC, tumors had a low TMB, with a median of 3.68 Mut/Mb (range 0.9–12.3), and only one patient had a TMB > 10 Mut/Mb. Patients had bulky disease with a median sum of target lesion diameters of 5.95 cm (range 1.9–17.3), and 8 of 13 patients (62%) had active liver metastases. The safety- and efficacy-evaluable population included the 13 patients who received at least one dose of study drug. The biomarker-evaluable population consisted of 12 patients, excluding one patient who withdrew prior to on-treatment biomarker sample collection as per protocol (Fig. 1B). This patient entered the study with significant cancer-associated pain due to disease progression, with enlarging hepatic metastases, lymphadenopathy, and abdominal wall implant. After 12 days on study, the patient withdrew treatment consent because of poor quality of life partly related to the burden of extensive travel required for trial visits. Imaging performed on day 6 showed modest lymph node enlargement accompanied by surrounding inflammatory changes, while other metastatic sites remained stable. These findings were interpreted as treatment-related inflammatory changes rather than disease progression.Table 1Patient demographics and clinical characteristicsCharacteristicOverall, n = 13Age, median (range), years49 (21, 66)Male, n (%)9 (69)Race, n (%) White10 (77) Black2 (15) Other1 (8)Baseline ECOG PS, n (%) 08 (62) 15 (38)Primary location, n (%) Colon, right2 (15) Colon, left3 (23) Rectum8 (62)Stage at diagnosis, n (%) I–II2 (15) III4 (31) IV7 (54)Prior radiation, n (%)7 (54)Prior lines of therapy, median (range)2 (1, 3)Previous systemic therapy, n (%) Fluoropyrimidine13 (100) Oxaliplatin13 (100) Irinotecan13 (100) Bevacizumab13 (100) Trifluridine-Tipiracil2 (15) Other2 (15)Liver metastases, n (%)8 (62)Lung metastases, n (%)11 (85)TMBa, median (range)3.68 (0.9, 12.3)KRAS mutation, n (%) G12V5 (38) G12D4 (31) G13D2 (15) G12C1 (8) G12R1 (8)Baseline sum of target lesionsb, c, median (range), cm5.95 (1.9, 17.3)a TMB not available for 1 patient.b Sum of longest diameter for all target lesions as per RECIST 1.1c Not available for 1 patient who withdrew consent prior to any RECIST 1.1 reads.ECOG PS Eastern Cooperative Oncology Group performance status, TMB tumor mutational burden. Table 1Patient demographics and clinical characteristicsCharacteristicOverall, n = 13Age, median (range), years49 (21, 66)Male, n (%)9 (69)Race, n (%) White10 (77) Black2 (15) Other1 (8)Baseline ECOG PS, n (%) 08 (62) 15 (38)Primary location, n (%) Colon, right2 (15) Colon, left3 (23) Rectum8 (62)Stage at diagnosis, n (%) I–II2 (15) III4 (31) IV7 (54)Prior radiation, n (%)7 (54)Prior lines of therapy, median (range)2 (1, 3)Previous systemic therapy, n (%) Fluoropyrimidine13 (100) Oxaliplatin13 (100) Irinotecan13 (100) Bevacizumab13 (100) Trifluridine-Tipiracil2 (15) Other2 (15)Liver metastases, n (%)8 (62)Lung metastases, n (%)11 (85)TMBa, median (range)3.68 (0.9, 12.3)KRAS mutation, n (%) G12V5 (38) G12D4 (31) G13D2 (15) G12C1 (8) G12R1 (8)Baseline sum of target lesionsb, c, median (range), cm5.95 (1.9, 17.3)a TMB not available for 1 patient.b Sum of longest diameter for all target lesions as per RECIST 1.1c Not available for 1 patient who withdrew consent prior to any RECIST 1.1 reads.ECOG PS Eastern Cooperative Oncology Group performance status, TMB tumor mutational burden. Patient demographics and clinical characteristics a TMB not available for 1 patient. b Sum of longest diameter for all target lesions as per RECIST 1.1 c Not available for 1 patient who withdrew consent prior to any RECIST 1.1 reads. ECOG PS Eastern Cooperative Oncology Group performance status, TMB tumor mutational burden. The combination of mKRAS-VAX with nivolumab and ipilimumab had an acceptable safety profile, meeting the primary safety endpoint. Treatment-related adverse events (TRAEs) as of data cut-off (February 27, 2025) are summarized in Table 2. All 13 patients in the safety-evaluable population had at least one TRAE. All vaccine site-related or mKRAS-VAX-related adverse events were grade 1 or 2, with the most common being injection site pain (92%), muscle cramps (38%), and fatigue (31%). Nine patients (69%) had an immune-related adverse event (irAE), and four patients (31%) had a grade 3 or 4 irAE—consisting of grade 3 adrenal insufficiency in two patients, grade 3 arthritis in one patient, and grade 4 hyperglycemia in one patient. Three patients (23%) discontinued ipilimumab and nivolumab due to irAEs.Table 2Treatment-related adverse events (TRAEs)Adverse event termOverall (n = 13)Alla, n (%)> Grade 3a, n (%)Treatment-related AEs13 (100)4 (31)mKRAS-VAX related AEs5 (38)0 Fatigue4 (31) Headache2 (15) Fever2 (15) Chills1 (8) Diarrhea1 (8) Flu-like symptoms1 (8) Myalgia1 (8) Nausea1 (8) Rash maculopapular1 (8)Vaccine site related AEs13 (100)0 Pain12 (92) Muscle cramps5 (38) Erythema3 (23) Bruising2 (15) Induration1 (8) Pruritus1 (8)Immune-related AEs9 (69)4 (31) Hyperthyroidism4 (31)0 Fatigue5 (38)0 Adrenal insufficiency3 (23)2 (15) Hypothyroidism3 (23)0 Rash maculo-papular3 (23)0 Anorexia1 (8)0 Arthralgia1 (8)1 (8) Colitis1 (8)0 Confusion1 (8)0 Diarrhea1 (8)0 Fever1 (8)0 Hyperglycemia1 (8)1 (8) Hypotension1 (8)0 Malaise1 (8)0 Nausea1 (8)0 Pruritus1 (8)0 TSH increased1 (8)0 Vomiting1 (8)0 Weight gain1 (8)0aNumber of patients that experienced reported adverse event. If a patient reported multiple adverse events of the same preferred term, the highest grade event was counted.TSH thyroid stimulating hormone. Table 2Treatment-related adverse events (TRAEs)Adverse event termOverall (n = 13)Alla, n (%)> Grade 3a, n (%)Treatment-related AEs13 (100)4 (31)mKRAS-VAX related AEs5 (38)0 Fatigue4 (31) Headache2 (15) Fever2 (15) Chills1 (8) Diarrhea1 (8) Flu-like symptoms1 (8) Myalgia1 (8) Nausea1 (8) Rash maculopapular1 (8)Vaccine site related AEs13 (100)0 Pain12 (92) Muscle cramps5 (38) Erythema3 (23) Bruising2 (15) Induration1 (8) Pruritus1 (8)Immune-related AEs9 (69)4 (31) Hyperthyroidism4 (31)0 Fatigue5 (38)0 Adrenal insufficiency3 (23)2 (15) Hypothyroidism3 (23)0 Rash maculo-papular3 (23)0 Anorexia1 (8)0 Arthralgia1 (8)1 (8) Colitis1 (8)0 Confusion1 (8)0 Diarrhea1 (8)0 Fever1 (8)0 Hyperglycemia1 (8)1 (8) Hypotension1 (8)0 Malaise1 (8)0 Nausea1 (8)0 Pruritus1 (8)0 TSH increased1 (8)0 Vomiting1 (8)0 Weight gain1 (8)0aNumber of patients that experienced reported adverse event. If a patient reported multiple adverse events of the same preferred term, the highest grade event was counted.TSH thyroid stimulating hormone. Treatment-related adverse events (TRAEs) aNumber of patients that experienced reported adverse event. If a patient reported multiple adverse events of the same preferred term, the highest grade event was counted. TSH thyroid stimulating hormone. mKRAS-VAX combination has antitumor activity in previously treated mCRC Secondary endpoints included the clinical efficacy of the mKRAS-VAX combination, which were determined by ORR and progression-free survival (PFS), as assessed by RECIST v1.1, and OS. Two patients (15%) achieved a partial response (PR) and the disease control rate (DCR) was 38% (5 of 13 patients) (Fig. 1C, D). The median PFS was 2.04 months, and the median OS was 24.85 months with a median follow-up of 17.52 months (Supplementary Fig. 1). After discontinuing from the trial, all patients received subsequent systemic anticancer treatments, with a median of 2 subsequent lines (range 1–5), the most common of which were trifluridine-tipiracil +/− bevacizumab (8 of 13) or another clinical trial (8 of 13) (Supplementary Table 2). Among the eight patients with active liver metastases, one (13%) achieved a PR, and two patients had stable disease, yielding a DCR of 38% (3 of 8 patients). Patient 21 with progressive disease (PD) had a mixed response on the first on-treatment scan (shrinkage of target lesions alongside emergence of new non-target lesions) and elected to remain on study beyond progression as per protocol. This patient subsequently had partial regression of the new lesions. The five patients that had any tumor regression while on study were designated as “patients with tumor response” for correlative studies. mKRAS-VAX combination elicits mKRAS T cell responses To determine whether the mKRAS-VAX combination elicited functional mKRAS-reactive T-cells, the immunogenicity primary endpoint, we performed ex vivo interferon-γ (IFNγ) ELISpot assays using all six mKRAS 21-mer SLPs included in mKRAS-VAX (Supplementary Fig. 2a). Within 17 weeks post-vaccination, all biomarker-evaluable patients showed an increase in their T-cell response against their tumor-specific KRAS mutation, with a median 8.9-fold change (range 1.8–25.1) from baseline, whereas no responses were detected against an irrelevant 24-mer control peptide (Fig. 2A). Of these, eight responses (75%) exceeded the threshold for a true increase beyond baseline variability (“Methods”). All twelve patients showed an increase in their average T-cell response across all six mKRAS antigens, with a median 9.2-fold change (range 4.1–29.1) from baseline (Supplementary Fig. 2b). Similar to the PDAC cohort of this trial, KRAS G12V and G12R appeared to be the most immunogenic, regardless of the patient’s tumor-specific KRAS mutation33.Fig. 2mKRAS-VAX induces polyfunctional T-cell responses against tumor-specific KRAS mutations.A Left: Ex vivo IFNγ ELISpot quantifying the peripheral T-cell response against each patient’s tumor-specific KRAS mutation at baseline and at the maximal on-treatment timepoint within 17 weeks following vaccination. Right: Representative raw IFNγ ELISpot images of pre- and post-vaccination PBMCs from two patients stimulated with either a negative control or a tumor-specific mKRAS peptide. 2 × 105 PBMCs plated per well. B Post-IVS IFNγ ELISpot quantifying the peripheral T-cell response against each patient’s tumor-specific KRAS mutation following in vitro mKRAS peptide expansion at baseline and on-treatment. C Representative flow cytometry showing CD4+ and CD8+ cytokine responses in expanded peptide restimulated T-cells. D Proportion of cytokine-positive (IFNγ+, TNFα+, or IL-2+) CD4+ or CD8+ T-cells measured by intracellular cytokine staining on expanded peptide restimulated T-cells. Cytokine responses are considered significant and displayed if they were > 50% above the control peptide background. E Detailed summary of cytokine staining in cytokine-positive T-cells per patient. F Memory phenotype distribution of cytokine-positive T-cells per patient. Central memory (CD45RA−, CCR7+), effector memory (CD45RA−, CCR7−), terminal effector memory (TEMRA: CD45RA+, CCR7−), naïve (CD45RA+, CCR7+). Data in (A–D) represent the means of 3 technical replicates and are background subtracted using control peptide stimulations. * in (A, B) indicates that the data exceed the threshold for a true change beyond expected baseline variability (“Methods”). Source data provided in Source Data File. mKRAS-VAX induces polyfunctional T-cell responses against tumor-specific KRAS mutations. A Left: Ex vivo IFNγ ELISpot quantifying the peripheral T-cell response against each patient’s tumor-specific KRAS mutation at baseline and at the maximal on-treatment timepoint within 17 weeks following vaccination. Right: Representative raw IFNγ ELISpot images of pre- and post-vaccination PBMCs from two patients stimulated with either a negative control or a tumor-specific mKRAS peptide. 2 × 105 PBMCs plated per well. B Post-IVS IFNγ ELISpot quantifying the peripheral T-cell response against each patient’s tumor-specific KRAS mutation following in vitro mKRAS peptide expansion at baseline and on-treatment. C Representative flow cytometry showing CD4+ and CD8+ cytokine responses in expanded peptide restimulated T-cells. D Proportion of cytokine-positive (IFNγ+, TNFα+, or IL-2+) CD4+ or CD8+ T-cells measured by intracellular cytokine staining on expanded peptide restimulated T-cells. Cytokine responses are considered significant and displayed if they were > 50% above the control peptide background. E Detailed summary of cytokine staining in cytokine-positive T-cells per patient. F Memory phenotype distribution of cytokine-positive T-cells per patient. Central memory (CD45RA−, CCR7+), effector memory (CD45RA−, CCR7−), terminal effector memory (TEMRA: CD45RA+, CCR7−), naïve (CD45RA+, CCR7+). Data in (A–D) represent the means of 3 technical replicates and are background subtracted using control peptide stimulations. * in (A, B) indicates that the data exceed the threshold for a true change beyond expected baseline variability (“Methods”). Source data provided in Source Data File. To quantify low-frequency mKRAS-reactive T-cell responses, we performed in vitro stimulation (IVS) to expand tumor-specific mKRAS-reactive T-cells and then repeated IFNγ ELISpot assays after peptide restimulation in the four patients who lacked a significant ex vivo response. Following T-cell expansion, all four patients had an increase in their tumor-specific mKRAS T-cell response on-treatment (Fig. 2B, “Methods”). We did not observe a correlation between the magnitude of peripheral tumor-specific mKRAS-T-cell responses and clinical outcomes. As part of an exploratory correlative analysis, we profiled the phenotype of vaccine-induced tumor-specific mKRAS-reactive T-cells by performing intracellular cytokine staining (ICS) for IFNγ, tumor necrosis factor alpha (TNFα), and interleukin-2 (IL-2) on expanded SLP-restimulated T-cells. All twelve patients had CD4+ T-cell responses and two (17%) had both CD4+ and CD8+ T-cell responses to their tumor-specific KRAS mutation (Fig. 2C, D and Supplementary Fig. 3). The majority of CD4+ T-cell responses were polyfunctional, with 53.9% producing two or more effector cytokines; 41.2% of CD8+ T-cell responses were polyfunctional (Fig. 2E). Most responses originated from antigen-experienced effector memory (CCR7−CD45RA−) or central memory (CCR7+CD45RA−) T-cell subsets (Fig. 2F). Because both the ex vivo and IVS assays used 21-mer mKRAS long peptides, which may be processed and presented inefficiently by PBMCs and thereby reduce assay sensitivity, we also evaluated overlapping minimal 9-, 10-, or 15-mer overlapping peptide pools (Supplementary Table 3). Expansion and restimulation with these minimal peptide pools did not identify additional CD4+ or CD8+ T-cell responses beyond those detected with the 21-mer mKRAS long peptides (Supplementary Fig. 4). The induction of primarily CD4+ T-cell responses against mKRAS is consistent with findings from prior long peptide neoantigen vaccine studies17,19,24. mKRAS-VAX combination induces T cells that traffic into metastatic tumors To assess if vaccine-induced T-cells trafficked into tumors, we profiled the peripheral and intratumoral T-cell receptor (TCR) repertoires in the five patients (three patients with tumor response and two without tumor response) who had adequate paired core tumor biopsies collected on study. To identify putative vaccine-induced mKRAS-reactive T-cell clonotypes, we stimulated on-treatment peripheral blood mononuclear cells (PBMCs) in vitro with tumor-specific mKRAS peptides, expanded them over 14 days, confirmed the outgrowth of tumor-specific mKRAS-reactive T-cells, followed by bulk T-cell receptor β-chain (TCRβ) sequencing (TCR-seq, Supplementary Fig. 5). Putative vaccine-induced mKRAS-reactive T-cell clonotypes were defined as clonotypes that emerged in on-treatment peripheral blood and showed preferential expansion under tumor-specific mKRAS peptide stimulation relative to irrelevant control peptide stimulation. Specifically, nominated clonotypes met all of the following criteria: (1) absence in baseline PBMCs, (2) > fivefold enrichment after mKRAS peptide expansion relative to a control peptide expansion, (3) post-mKRAS-expansion frequency >0.025%, and (4) significant clonotype-level enrichment in the mKRAS expansion relative to control (one-sided Fisher’s exact test with false discovery rate-adjusted p < 0.01) (Fig. 3A). Across patients, we identified a median of 98 newly emergent tumor-specific mKRAS-reactive T-cell clonotypes in the peripheral blood (range 66–216) following mKRAS-VAX administration (Fig. 3B and Supplementary Data 1). In patient 8, we performed in vitro expansions using each of the six mKRAS peptides included in mKRAS-VAX and found that 34% of their tumor-specific G13D T-cell clonotypes were cross-reactive with other KRAS mutations (Supplementary Fig. 6). Additionally, we identified three public vaccine-induced mKRAS T-cell clonotypes—TCRβ sequences shared across multiple individuals (Supplementary Table 4). Although single-cell TCR-seq followed by TCR reconstruction is needed to definitively confirm antigen-specificity, one public clonotype (CDR3β CASRQGNTGELFF), identified as cross-reactive to G13D and G12R in our in vitro expansions, has the same CDR3β sequence as a previously validated G13D-reactive tumor-infiltrating lymphocyte (TIL) from a patient with colon cancer34.Fig. 3Peripheral mKRAS-VAX-induced mKRAS-reactive and treatment-enriched T-cell clonotypes traffic into metastatic tumors.A Representative T-cell clonotype scatter plot for Patient 20 following in vitro expansion with G12V peptide. Each dot represents a unique T-cell clonotype and G12V clonotypes are colored orange. B, C Bar graphs showing the number of mKRAS-VAX-induced tumor-specific mKRAS-reactive T-cell clonotypes (B) and treatment-enriched T-cell clonotypes (C) identified in the peripheral blood for indicated patients. D Number of unique intratumoral mKRAS or treatment-enriched T-cell clonotypes identified in the baseline biopsy only, the on-treatment biopsy only, or both biopsies for each patient. N indicates the total number of unique mKRAS-reactive or treatment-enriched clonotypes identified within the tumor of that patient. E Proportion of unique intratumoral mKRAS T-cell clonotypes classified as pre-existing or de novo for each patient. De novo clonotypes were not detected in baseline peripheral blood or baseline tumor. Source data provided in Source Data File. Peripheral mKRAS-VAX-induced mKRAS-reactive and treatment-enriched T-cell clonotypes traffic into metastatic tumors. A Representative T-cell clonotype scatter plot for Patient 20 following in vitro expansion with G12V peptide. Each dot represents a unique T-cell clonotype and G12V clonotypes are colored orange. B, C Bar graphs showing the number of mKRAS-VAX-induced tumor-specific mKRAS-reactive T-cell clonotypes (B) and treatment-enriched T-cell clonotypes (C) identified in the peripheral blood for indicated patients. D Number of unique intratumoral mKRAS or treatment-enriched T-cell clonotypes identified in the baseline biopsy only, the on-treatment biopsy only, or both biopsies for each patient. N indicates the total number of unique mKRAS-reactive or treatment-enriched clonotypes identified within the tumor of that patient. E Proportion of unique intratumoral mKRAS T-cell clonotypes classified as pre-existing or de novo for each patient. De novo clonotypes were not detected in baseline peripheral blood or baseline tumor. Source data provided in Source Data File. We next looked at TCR clonotypes that were enriched in the blood on-treatment relative to baseline to capture other T-cells that may be correlated with clinical responses, such as T-cells that recognize tumor-specific mKRAS, but have limited proliferative capacity in our in vitro expansions, or those that recognize non-mKRAS tumor neoantigens. We identified a median of 49 treatment-enriched T-cell clonotypes per patient (range 1–71), defined as high-frequency clones that newly emerge in unexpanded peripheral blood on-treatment (Fig. 3C and Supplementary Data 1; see “Methods”). Among these high-frequency clonotypes, several treatment-enriched T-cell clonotypes (6 of 201) overlapped with mKRAS T-cell clonotypes identified by in vitro expansion. Notably, we cannot exclude that these treatment-enriched T-cell clonotypes could be T-cells that recognized KRAS mutations not found in the patient’s tumor or pathogen targets. To assess intratumoral T-cell dynamics and determine if tumor-specific mKRAS-reactive T-cell clonotypes trafficked to the tumor bed, we performed bulk TCR-seq on tumor biopsies. We observed an increase in intratumoral T-cell clonality following treatment across all evaluable paired samples, although this change did not reach statistical significance due to limited power from the small sample size (Supplementary Fig. 7a). This clonal expansion is consistent with prior analyses of TILs after neoadjuvant ICI35,36. Immunohistochemistry (IHC) revealed an increase in CD8+ T-cell densities in all samples following treatment, whereas changes in CD4+ T-cell and CD20+ immune cell density were more variable (Supplementary Fig. 7b–d). Due to the limited sample size, we were underpowered to statistically compare patients with tumor response with those without tumor response. However, tumors that regressed showed greater increases in T-cell clonality, CD4+ T-cell density, and CD8+ T-cell density. All tumors retained MHC class I expression and largely remained MHC class II negative following treatment (Supplementary Fig. 7e). By tracking T-cell clonotypes via their unique CDR3β sequences, we found that tumor-specific mKRAS-reactive T-cell clonotypes and treatment-enriched T-cell clonotypes identified in peripheral blood were detectable in tumors following treatment in all five patients (Fig. 3D). Although mKRAS-reactive T-cell clonotypes were defined to be newly emergent in the peripheral blood on-treatment, a subpopulation was detected in the tumor at baseline. This suggests that the mKRAS-VAX combination not only primed de novo mKRAS-reactive T-cells that infiltrated tumors, but also expanded pre-existing intratumoral mKRAS-reactive T-cells to levels detectable in the peripheral blood. Interestingly, tumors from patients with tumor response primarily harbored de novo intratumoral mKRAS-reactive T-cell clonotypes (89% de novo), whereas tumors from patients without tumor response were enriched for pre-existing clonotypes (62% pre-existing; Fig. 3E). Finally, we assessed whether the peripheral blood frequency of tracked clonotypes predicted their abundance in on-treatment tumors (Supplementary Fig. 8). Following vaccination, neither mKRAS-reactive nor treatment-enriched clonotypes showed a clear relationship between their frequency in peripheral blood and their frequency in tumors. Intratumoral clonal expansion of mKRAS-VAX-induced T cells in patients with tumor response Given that regressing tumors exhibited greater increases in intratumoral T-cell clonality compared to progressing tumors (Supplementary Fig. 7a), we next investigated whether this was accompanied by clonal expansion of infiltrating vaccine-induced mKRAS-reactive or treatment-enriched T-cell clonotypes. Patient 20, who had a KRAS G12V tumor with a bulky axillary mass (6.8 cm target lesion) and liver metastases, achieved a PR (Fig. 4A). A baseline core biopsy of the axillary mass revealed pancytokeratin+ tumor cells and T-cells that were excluded from the tumor (Fig. 4B). Pre-existing G12V-reactive and treatment-enriched clonotypes were detectable in the tumor at baseline, but at low aggregate frequencies, accounting for 0.29% and 0.24% of intratumoral T-cell repertoire, respectively (Fig. 4C, D). An on-treatment biopsy of the shrinking axillary mass showed a regression bed with dense T-cell infiltration, but no residual pancytokeratin+ tumor cells (Fig. 4B). G12V-reactive and treatment-enriched clonotypes increased in aggregate frequency, now accounting for 9.31 and 3.67% of the on-treatment intratumoral T-cell repertoire, respectively (Fig. 4C, D). While pre-existing G12V-reactive and treatment-enriched T-cell clonotypes exhibited modest expansion, de novo G12V-reactive T-cell clonotypes underwent substantial clonal expansion with a single clonotype accounting for up to 8.79% of the on-treatment intratumoral T-cell repertoire.Fig. 4Intratumoral expansion of de novo mKRAS-reactive T-cell clones in a patient with an objective response (Patient 20).A Top: Plot of percentage change from baseline in sum of target lesion diameters, as per RECIST v1.1, with indicators showing timing of baseline and on-treatment biopsies. Bottom: Serial CT scans displaying representative target lesions (red outlines) and additional non-target lesions (red arrows). B Representative regions of interest of IHC staining of axillary metastasis biopsies at baseline and on-treatment. Top: pancytokeratin (brown) staining outlines tumor cells. Bottom: dual-color staining for CD4+ (red) and CD8+ (brown) T-cells. Scale bar 250 µm. C Scatter plot of intratumoral T-cell clonotype frequencies at baseline versus on-treatment. G12V-reactive and treatment-enriched T-cell clonotypes are highlighted in orange and blue, respectively. D Stacked bar plot showing cumulative intratumoral frequencies of pre-existing and de novo G12V or treatment-enriched T-cell clonotypes at baseline and on-treatment. Source data provided in Source Data File. Intratumoral expansion of de novo mKRAS-reactive T-cell clones in a patient with an objective response (Patient 20). A Top: Plot of percentage change from baseline in sum of target lesion diameters, as per RECIST v1.1, with indicators showing timing of baseline and on-treatment biopsies. Bottom: Serial CT scans displaying representative target lesions (red outlines) and additional non-target lesions (red arrows). B Representative regions of interest of IHC staining of axillary metastasis biopsies at baseline and on-treatment. Top: pancytokeratin (brown) staining outlines tumor cells. Bottom: dual-color staining for CD4+ (red) and CD8+ (brown) T-cells. Scale bar 250 µm. C Scatter plot of intratumoral T-cell clonotype frequencies at baseline versus on-treatment. G12V-reactive and treatment-enriched T-cell clonotypes are highlighted in orange and blue, respectively. D Stacked bar plot showing cumulative intratumoral frequencies of pre-existing and de novo G12V or treatment-enriched T-cell clonotypes at baseline and on-treatment. Source data provided in Source Data File. Patient 21, who had a KRAS G12D tumor with predominantly pulmonary metastases and no liver involvement, exhibited a mixed response on the initial on-treatment scan with regression of pulmonary lesions, but emergence of new lymphadenopathy (Supplementary Fig. 9a). Given overall clinical benefit, the patient elected to stay on study per protocol beyond initial RECIST v1.1 PD and subsequently underwent a research biopsy of an enlarging lymph node. Over time, the lymphadenopathy stabilized and partially regressed, suggesting pseudoprogression. An on-treatment biopsy of a regressing lung metastasis revealed a highly inflamed tumor with significant treatment effect, whereas the transiently enlarging lymph node displayed sparse tumor-infiltrating T-cells (Supplementary Fig. 9b). Bulk TCR-seq revealed robust expansion of treatment-enriched T-cell clonotypes in the regressing lung lesion, accounting for 9.22% of the on-treatment intratumoral T-cell repertoire (Supplementary Fig. 9c, d). In the lymph-node metastasis, treatment-enriched T-cell clonotypes represented 3.11% of the intratumoral repertoire, although no baseline biopsy of that site was available for comparison (Supplementary Fig. 9d). In contrast to the two aforementioned patients with tumor response, on-treatment tumors from patients with PD were characterized by moderate aggregate frequencies of pre-existing mKRAS-reactive T-cell clonotypes and rare treatment-enriched T-cell clonotypes (Supplementary Fig. 10a, b). These findings suggest that the mKRAS-VAX combination is associated with clonal expansion of de novo tumor-infiltrating T-cells in regressing tumors; while pre-existing or non-expanded T-cell clones predominate in growing tumors. Discussion Most prior neoantigen vaccine studies in the metastatic setting have focused on tumors intrinsically responsive to immunotherapy, such as melanoma18,20,22,25. In this phase 1 study, we tested a mKRAS vaccine in combination with ICIs in patients with heavily pretreated MMRp/MSS mCRC, a tumor type historically non-responsive to ICIs. mKRAS-VAX was well tolerated when combined with nivolumab and ipilimumab. The primary toxicities attributable to mKRAS-VAX alone were limited to low-grade injection site reactions and mild systemic symptoms, and the overall frequency and distribution of severe adverse events were not different from what is typically observed with nivolumab and ipilimumab alone1,37,38. The mKRAS-VAX combination demonstrated encouraging clinical activity, though efficacy outcomes should be interpreted with requisite caution given the small single-arm patient cohort. Although cross-trial comparisons are inherently limited, the ORR of 15% observed in our study compares favorably with prior immunotherapy studies in MMRp/MSS mCRC, where responses have generally been rare, including an ORR of 1% with durvalumab (anti–PD-L1) plus tremelimumab (anti–CTLA-4)6. In particular, the tumor regression and PR observed in a patient with liver metastases is notable, as this disease setting has historically been refractory to ICIs, including an ORR of 0% among 25 patients with active liver metastases treated with the second-generation balstilimab (anti–PD-1) plus botensilimab (anti–CTLA-4) combination8,10. We also noted biopsy-documented pseudoprogression in this study, which is a common clinical dilemma for patients treated on novel immunotherapy trials. Although KRAS mutations are associated with inferior survival in mCRC, our study had a compelling median OS of 24.85 months, which exceeds the OS in the trials that led to the approval of trifluridine-tipiracil plus bevacizumab, fruquintinib, and regorafenib in a similar disease setting39–42. The observed survival in the present study, despite a median PFS of only 2.04 months, may reflect selection bias of trial-eligible patients or the impact of post-progression systemic therapies, but it also raises the intriguing hypothesis that the mKRAS-VAX combination may exert durable effects on disease trajectory, which would require confirmation in a randomized trial. mKRAS-VAX elicited tumor-specific mKRAS-reactive T-cell responses in 75% of patients after direct ex vivo peptide stimulation of on-treatment PBMCs and polyfunctional responses in all patients following in vitro expansion. Notably, mKRAS-VAX induced up to hundreds of tumor-specific mKRAS-reactive T-cell clonotypes in peripheral blood, which were capable of homing into metastatic tumors. By priming T-cells outside the immunosuppressive tumor microenvironment, mKRAS-VAX may generate less dysfunctional tumor-reactive T-cells than those primed within tumors, potentially contributing to its efficacy even in patients with liver metastases15. Consistent with prior neoantigen vaccine studies in the metastatic setting, we did not observe a correlation between the magnitude of peripheral mKRAS T-cell responses and clinical outcomes20–22,25–27. In our study, peripheral frequencies of expanded mKRAS-reactive and treatment-enriched T-cell clonotypes were not predictive of intratumoral T-cell infiltration. Given that <2% of the body’s T-cells circulate in the peripheral blood, the lack of correlation is not unexpected43. Intratumoral analyses revealed that regressing tumors primarily contained de novo vaccine-induced mKRAS T-cell clonotypes, which in one patient with an objective response, underwent significant intratumoral clonal expansion. These findings suggest that intratumoral clonal expansion of vaccine-induced mKRAS-reactive T-cells may be a therapeutic mechanism for the mKRAS-VAX combination. They also highlight the importance of profiling the intratumoral T-cell repertoire, rather than relying on the peripheral blood T-cell repertoire, as a biomarker in future cancer vaccine studies in the metastatic setting. Deeper understanding of vaccine-induced intratumoral T-cell repertoire and immune dynamics may ultimately enable the identification of peripheral immune markers that more faithfully reflect clinically relevant antitumor immunity. mKRAS-VAX predominantly induced tumor-specific mKRAS-reactive CD4+ T-cell responses in peripheral blood, consistent with prior studies of neoantigen long peptide vaccines17,19,24. There is increasing recognition of the importance of CD4+ T-cells in effective antitumor immunity, either indirectly by enhancing anti-tumor CD8+ T-cell or myeloid cell activity, or, in some contexts, directly via effector cytokine secretion or cytotoxicity44–46. Recent findings from a trial of neoantigen-reactive TILs in gastrointestinal cancers illustrate the importance of the CD4+ T-cell response: infusion products enriched for neoantigen-reactive CD4+ T-cells correlated with clinical benefit47. Because MHC class II was not detected on tumor cells in on-treatment biopsies, vaccine-induced mKRAS CD4+ T-cells likely mediate antitumor effects without directly interacting with mKRAS peptides presented by tumor cells. One possibility is that these mKRAS-reactive CD4+ T cells help recruit tumor-reactive CD8+ T cells that recognize non-mKRAS tumor neoantigens, which would be consistent with the greater increase in intratumoral CD8+ T-cell density observed after treatment. Future studies are needed to further characterize the intratumoral function of vaccine-induced mKRAS CD4+ T-cells and determine whether they mediate tumor killing through epitope spreading. Finally, given the increased infiltration of both CD4+ and CD8+ T-cells in regressing tumors, vaccine-induced mKRAS-reactive CD8+ T-cells may predominantly reside within tissue compartments, limiting our ability to detect mKRAS-reactive CD8+ T-cell responses in the peripheral blood. In contrast to personalized cancer vaccines targeting private neoantigens, mKRAS-VAX is an off-the-shelf, fixed vaccine directed against public KRAS neoantigens, which offers distinct advantages. First, personalized vaccines require tumor biopsies for neoantigen prediction followed by individualized manufacturing, which can require at least 4 weeks with mRNA platforms under best turn-around times and up to several months with peptide-based vaccines24,26. Given that MMRp/MSS mCRC can have a high disease tempo, manufacturing delays may preclude patients from receiving timely treatment or necessitate bridging therapies. Second, intratumoral neoantigen heterogeneity is a negative predictive factor for ICI response across multiple tumor types48–50. While clonal neoantigens are prioritized during personalized vaccine design, sampling bias is inherent to metastatic tumor biopsies and subclonal immune editing of vaccine-targeted neoantigens is a known resistance mechanism25,28,51. In contrast, the oncogenic KRAS mutations targeted by mKRAS-VAX are among the most clonally conserved driver mutations in MMRp/MSS CRC28,29,52. However, future studies are needed to determine whether subclonal variation in KRAS mutation status constitutes a common resistance mechanism against mKRAS-targeted cancer vaccines. Other approaches to target mKRAS in MMRp/MSS mCRC include adoptive T-cell therapies and KRAS inhibitors. Although case reports show evidence of clinical efficacy for mKRAS-targeted cell therapies, their implementation is constrained by the need for bespoke development for each mKRAS peptide—HLA allele pair, which is currently applicable to only a quarter of patients with mKRAS tumors30,31. Inhibitors of KRAS G12C are in clinical use, but the G12C mutation is found in only ~2% of patients with MMRp/MSS CRC and responses are often short due to rapid development of resistance29,53–55. Second-generation mKRAS and pan-RAS inhibitor are in clinical trials, but resistance is also anticipated56,57. Recent preclinical studies have shown that KRAS inhibition enhances tumor antigen presentation and promotes T-cell infiltration, highlighting the need to evaluate whether mKRAS vaccines can synergize with KRAS inhibitors58,59. We acknowledge limitations to this study. First, the trial was designed to evaluate safety and immunogenicity, and not efficacy. Its single-arm design and small size preclude any definitive conclusions regarding whether mKRAS-VAX causally contributed to clinical outcomes. While MMRp/MSS CRC are considered intrinsically resistant to nivolumab and ipilimumab, we cannot exclude that tumors from patients with tumor response may harbor rare genetic alterations or a tumor immune microenvironment not captured by routine clinical molecular testing that confers increased sensitivity to ICIs alone. Accordingly, our observed responses are hypothesis-generating and future prospective randomized trials are needed to determine whether the addition of mKRAS-VAX confers clinical benefit beyond that of dual checkpoint blockade against PD-1 and CTLA-4. Given the toxicity of dual checkpoint blockade, randomized trials are needed to confirm a favorable benefit–risk profile for the mKRAS-VAX combination as well. Second, while all patients had at least one fresh biopsy on trial, only 5 of 12 patients had adequate paired tumor biopsies, limiting the generalizability of our intratumoral T-cell repertoire analysis. Larger datasets from future mKRAS-VAX trials are needed to validate these findings and better define correlates of response. Third, while bulk TCR-seq can nominate putative mKRAS-reactive clonotypes, definitive validation of mKRAS recognition requires single-cell TCR-seq to recover paired α/β chains, followed by TCR reconstitution and functional testing. A companion report on the pancreatic adenocarcinoma cohort of this trial performed this orthogonal validation to support our bulk TCR-seq analysis33. However, for the CRC cohort, future studies are needed to definitively validate mKRAS-specificity of tumor-infiltrating T-cell clonotypes and assess whether intratumoral mKRAS-specific TCRs from patients with tumor response exhibit distinctive features33. Finally, our study was insufficient to determine whether immunodominance by vaccine-included non-tumor-specific mKRAS epitopes influences intratumoral T-cell dynamics or clinical benefit. In conclusion, mKRAS-VAX was well tolerated when combined with nivolumab and ipilimumab in patients with heavily pretreated MMRp/MSS mCRC. Clinical benefit was observed with confirmed responses in select patients, including those with disease states historically unresponsive to dual checkpoint blockade alone, and encouraging OS in the total population. The mKRAS-VAX combination elicited mKRAS T-cells capable of infiltrating metastatic tumors and undergoing intratumoral clonal expansion. These findings support further clinical development of mKRAS vaccines and underscore their potential to immunologically target oncogenic KRAS mutations. The durability of neoantigen vaccine-induced clinical responses may be enhanced in the first-line maintenance setting, where tumor burden is lower following induction chemotherapy. Thus, we are conducting an ongoing phase Ib clinical trial (NCT06411691) evaluating mKRAS-VAX in combination with balstilimab and botensilimab in the first-line maintenance setting for patients with metastatic MMRp/MSS colorectal and pancreatic cancers. Methods Study overview This study was conducted in accordance with the Declaration of Helsinki and good clinical practice guidelines. The study protocol and all amendments were approved by the institutional review board at Johns Hopkins University. This study was an open-label, single-arm, first-in-human phase I trial conducted at the Johns Hopkins Sidney Kimmel Comprehensive Cancer Center, evaluating a pooled synthetic long peptide mutant KRAS vaccine (mKRAS-VAX) in combination with ipilimumab and nivolumab in patients with either resected pancreatic ductal adenocarcinoma or heavily pretreated MMRp/MSS mCRC. This manuscript reports data from the mCRC cohort. All participants provided written informed consent. The study was preregistered at ClinicalTrials.gov (NCT04117087) on October 3rd 2019 and the full trial protocol is provided in the Supplementary Information. Patients Eligible patients had histologically confirmed mCRC and prior exposure to fluoropyrimidine, irinotecan, and oxaliplatin chemotherapy in the metastatic setting. Molecular testing performed in a CLIA-certified laboratory was required to confirm that the tumor was MMRp/MSS and had a KRAS mutation included in mKRAS-VAX. Patients had to have at least one tumor lesion that was amenable to repeated biopsies with acceptable clinical risk. Key exclusion criteria included prior treatment with ICIs and the presence of brain metastases, peritoneal disease, or bulky liver metastases (defined as a single hepatic lesion > 5 cm or > 50% liver involvement). Both male and female participants were included in this study. Sex was recorded from the electronic medical record based on participant self-report. Analyses were not stratified by sex because the study was not powered to detect sex-specific differences in outcomes. Full eligibility criteria are provided in the study protocol in the Supplementary Information. Patients were enrolled between June 2021 and October 2024. Study endpoints The co-primary endpoints were to evaluate the safety of mKRAS-VAX in combination with ipilimumab and nivolumab and its immunogenicity within 17 weeks following initial vaccination. Secondary endpoints were immunogenicity any time after vaccination, ORR, progression-free survival (PFS), and OS. Exploratory endpoints included immunologic profiling of serial peripheral blood and tumor biopsy samples. Treatment and clinical assessment Treatment consisted of a prime and boost phase. During the prime phase, patients received mKRAS-VAX weeks 1, 2, 3, 4, administered alongside ipilimumab 1 mg/kg on weeks 1, 7, and nivolumab 3 mg/kg on weeks 1, 4, 7, and 10. In the boost phase, beginning at week 13, patients received mKRAS-VAX every 8 weeks and nivolumab 480 mg every 4 weeks, for up to one year of total treatment. Safety was assessed during each administration of mKRAS-VAX, ipilimumab, and nivolumab, at the end-of-treatment visit, and 100 days from the last dose of nivolumab or ipilimumab (or within 7 days prior to initiation of a subsequent antineoplastic therapy, whichever occurred first). Adverse events were graded according to NCI Common Terminology Criteria for Adverse Events (CTCAE) version 5.0. Tumors were assessed for radiographic response every 8 weeks. Treatment was discontinued upon withdrawal of consent, unacceptable toxicity, or radiographic disease progression. Given that a minority of patients treated with immunotherapy may derive benefit from continued treatment despite initial evidence of PD, patients were permitted to continue past initial RECIST v1.1-defined PD if the patient had no signs of clinically significant disease progression, was tolerating study treatment, and was judged to derive clinical benefit by the investigator. Such patients were required to discontinue treatment upon confirmed progression on any subsequent imaging. Study endpoints The primary safety endpoint was the frequency and severity of TRAEs, graded according to CTCAE version 5.0. The safety-evaluable population included all patients who received at least one dose of any study drug. The primary immunogenicity endpoint was defined as the maximal fold-change in IFNγ-producing mKRAS-reactive T-cells within 17 weeks following vaccination, relative to baseline, as measured by an ex vivo IFNγ ELISpot assay. Secondary efficacy endpoints included ORR, PFS, and OS. ORR was defined as the proportion of patients achieving a complete response (CR) or PR per RECIST v1.1 criteria. PFS was defined as the time from the first vaccination to the date of disease progression or death, whichever occurred first. OS was defined as the time from the first vaccination until death from any cause, or was censored at the last known date the subject was last known to be alive. The efficacy-evaluable population included all patients who received at least one dose of study drug. PFS and OS were evaluated using the Kaplan–Meier method. For correlative analyses, per protocol, the trial aimed to enroll 12 biomarker-evaluable patients. This sample size was based on the estimate that the mKRAS-VAX combination will result in at least a 2.5-fold-change in IFNγ-producing mKRAS-reactive T-cells following vaccination based on prior studies60. Assuming that the coefficient of variation for IFNγ-producing mKRAS-reactive T-cells was 100% and the correlation of pre- and post-treatment measurements is weak at 0.2, a sample size of 12 biomarker-evaluable patients would provide an 81% power to detect a 2.5-fold change from baseline based on a two-sided paired-sample t-test at a significant level of 0.05. mKRAS-VAX formulation and administration mKRAS-VAX consists of a pool of six 21-mer synthetic long peptides corresponding to common KRAS mutations found in PDAC and CRC. Clinical grade lyophilized peptides were synthesized by JPT technologies at > 90% purity and had the following sequences: G12V—YKLVVVGAVGVGKSALTIQLI G12A—YKLVVVGAAGVGKSALTIQLI G12R—YKLVVVGARGVGKSALTIQLI G12C—YKLVVVGACGVGKSALTIQLI G12D—YKLVVVGADGVGKSALTIQLI G13D—YKLVVVGAGDVGKSALTIQLI Peptides were reconstituted in DMSO, aliquoted, and stored at −80 °C until vaccination. On the day of administration, all six peptides (0.3 mg each, 1.8 mg total) were admixed with 0.5 mg poly-ICLC (Hiltonol, Oncovir) in saline by the Investigational Drug Service Pharmacy at Johns Hopkins Hospital. The mKRAS-VAX and poly-ICLC doses were selected to align with dosing used in prior neoantigen peptide vaccine trials17,19. The vaccine was administered as up to five subcutaneous injections in the upper thighs, arms, and/or back. A topical lidocaine-based anesthetic cream was applied to each injection site at least one hour prior to vaccination. Patient samples Patients underwent core needle biopsies of tumors at baseline (prior to first treatment dose) and at the same site approximately 7 weeks after starting treatment. Optional tumor biopsies at the time of radiographic progression were permitted at the discretion of the principal investigator. Biopsies were either formalin-fixed and paraffin-embedded (FFPE) or flash frozen. Peripheral blood was collected for research purposes at baseline, every 3 weeks during the prime phase, every 8 weeks during the boost phase, and at the end-of-treatment visit. Ex vivo and post-in vitro stimulation (IVS) IFNγ ELISpot assays The frequency of IFNγ-producing mKRAS-reactive T-cells was assessed by the Johns Hopkins Immune Monitoring Core under a validated ELISpot test method for ex vivo (fresh) PBMC using precoated stripwells (Mabtech, 3420-4HST-2). PBMCs were cultured in RPMI-1640 (Gibco, 11875-093) supplemented with L-glutamine (Mediatech, 25-005-Cl), 1× Pen/Strep (Quality Biological, 120-095-721), 50 µM β-mercaptoethanol (Sigma), and 10% fetal calf serum (Hyclone). Briefly, 2 × 105 fresh PBMCs were seeded per well in technical triplicates and stimulated with individual mKRAS peptides at 2 µg/mL for 18 h. Controls included unstimulated PBMCs and PBMCs stimulated either a viral peptide pool (CTL, CTL-CEF-002) or an irrelevant control peptide (JPT, RKREIFDRYGEEVKEFLAKAKEDF), both at 2 µg/mL. For post-IVS assays, mKRAS-reactive T-cells were expanded for 14 days in the presence of peptides and cytokines prior to assessment via IFNγ ELISpot assays. PBMCs were cultured in R10 media—RPMI 1640 (Gibco, 11875-085) supplemented with 5% human serum (GeminiBio 100-512), 1× Pen/Strep (Gibco 15140-122), and 12.5 mM HEPES (Gibco 1563-080). Cryopreserved PBMCs from baseline and the on-treatment timepoint corresponding to the maximal ex vivo IFNγ ELISpot response were thawed and seeded at 2 × 106 cells per well. PBMCs were stimulated with the patient’s tumor-specific mKRAS peptide at 10 µg/mL and maintained in R10 media containing 50IU/mL IL-2 (PeproTech 200-02), 25 ng/mL IL-7 (PeproTech 200-07), and 25 ng/mL IL-15 (PeproTech 200-15). Cytokine-containing R10 media was replaced every 2–3 days with fresh R10 media plus 100 IU/mL IL-2, 25 ng/mL IL-7, and 25 ng/mL IL-15. On day 15, PBMCs were transferred to cytokine-free R10 media. On day 16, post-IVS IFNγ ELISpot assays were performed using precoated IFNγ plates (Mabtech, 3420-2HPW-2) as per manufacturer’s instructions. Briefly, 2 × 105 expanded PBMCs were seeded per well in technical triplicates and restimulated with the patient’s tumor-specific mKRAS peptide or control peptide, both at 10 µg/mL, for 24 h. Plates were read using an iSpot Spectrum Analyzer (AID) and results are reported as spot forming units (SFU) per million cells based on the means of technical triplicates. Background responses from irrelevant control peptide restimulations were subtracted from results. To determine whether an individual patient’s on-treatment change in IFNγ SFU responses represents a true increase from baseline, we first calculated the repeatability coefficient (RC)61. The RC was defined as:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${RC}=1.96\sqrt{2}\cdot {within}\,{subject}\,{standard}\,{deviation}\left({wSD}\right)=2.77\cdot {wSD}$$\end{document}RC=1.962⋅withinsubjectstandarddeviationwSD=2.77⋅wSDSince repeated baseline measurements within individual patients were not available, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${wSD}$$\end{document}wSD was derived from the baseline SFU measurements across all 12 patients. For each patient, a 95% confidence interval (CI) for the observed change was then calculated as:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$({Maximal}\,{Absolute}\,{Change}-{RC},{Maximal}\,{Absolute}\,{Change}+{RC})$$\end{document}(MaximalAbsoluteChange−RC,MaximalAbsoluteChange+RC)If the 95% CI did not include 0, the patient’s change was considered a true increase beyond baseline variability rather than due to chance. Intracellular cytokine staining (ICS) Prior to intracellular cytokine staining (ICS), mKRAS-reactive T cells were expanded for 9 days in assays with SLPs only or for 14 days in assays where both SLPs and minimal peptide pools were used. 9-, 10-, and 15-mer minimal peptides were synthesized as lyophilized peptides (Peptide 2.0) at ≥ 90% purity, reconstituted as overlapping peptide pools in DMSO, aliquoted, and stored at −80 °C until use. Peptide pool constituents are listed in Supplementary Table 3. Cryopreserved PBMCs from the on-treatment timepoint with the maximal ex vivo IFNγ ELISpot responses were thawed and expanded as above with the following modifications. For 9 day expansions, PBMCs were cultured in ImmunoCult-XF T cell expansion media (“XF media”, Stemcell 10981). XF media was refreshed with 100 IU/mL IL-2, 10 ng/mL IL-7, and 25 ng/mL IL-15 every 2–3 days. On day 9, PBMCs were transferred to cytokine-free XF media. On day 10, expanded PBMCs were replated at 1.5–2 × 106 cells per well in technical triplicates. They were restimulated with the patient’s tumor-specific mKRAS peptide or control peptide, both at 10 µg/mL, in the presence of costimulatory antibodies anti-CD28 (1 µg/mL, BD 347690) and anti-CD49d (1 µg/mL, BD 347690) as well as degranulation marker anti-CD107a (PE/Dazzle, 1:40, BioLegend clone H4A3, 328645). For 14 day expansions, PBMCs were thawed and stimulated with the patient’s tumor-specific mKRAS SLP at 2 µg/mL or minimal overlapping peptide pools at 2 μg/mL/peptide in R10 media containing 50 IU/mL IL-2, 25 ng/mL IL-7, and 25 ng/mL IL-15. Cytokine-containing R10 media was replaced every 2–3 days with fresh R10 media plus 100 IU/mL IL-2, 25 ng/mL IL-7, and 25 ng/mL IL-15. On day 14, expanded PBMCs were transferred to cytokine-free R10 media. On day 16, PBMCs were restimulated with 2 µg/mL tumor-specific mKRAS SLP or 2 μg/mL/peptide peptide pools in the presence of anti-CD107a. For both 9 day and 14 day expansions, after 1 h of peptide restimulation, 1× eBioscience protein transport inhibitor (ThermoFisher, 00-4980-03) was added, and cells were incubated for an additional 5 h. For flow cytometry, cells were collected, washed with PBS, and stained with Zombie NIR fixable viability stain (Biolegend, 423105) for 10 min at room temperature. Samples were then washed in flow buffer (PBS + 2% fetal bovine serum) and stained with the surface markers anti-CD3 (FITC, 1:400, Biolegend clone SK7, 344804), anti-CD8 (BV421, 1:250, Biolegend clone RPA-T8, 301036), anti-CD4 (BV510, 1:250, BD Biosciences clone SK3, 562971), anti-CCR7 (BV605, 1:100, BD clone 2-L1-A, 566755), anti-CD14 (APC, 1:250, Biolegend clone M5E2, 301807), anti-CD16 (APC, 1:250, BD Biosciences clone 3G8, 561248), anti-CD19 (APC, 1:250, Biolegend clone SJ25C1, 363006), and anti-CD45RA (AF700, 1:100, BD Biosciences clone HI100, 560673) in flow buffer at 37 °C for 20 min. Cells were then washed in flow buffer, permeabilized with BD Perm/Fix kit (BD Biosciences, 554714), and stained for intracellular cytokines using anti-IFNγ (PE, 1:100, Biolegend clone 4S.B3, 502508), anti-TNFα (BV650, 1:100, Biolegend clone Mab11, 502938), and anti-IL2 (BV785, 1:100, Biolegend clone MQ1-17H12, 500348) in Perm/Wash buffer at room temperature for 30 min. Cells were then washed in Perm/Wash buffer, resuspended in flow buffer, and run on a BD Celesta flow cytometer at the Johns Hopkins Flow Cytometry Core. Data (means of technical triplicates) were analyzed in Flowjo version 10.10.0. Upon restimulation with mKRAS peptides, CD4 + CD107a+ or CD8 + CD107a+ cytokine responses were considered positive if they were ≥ 50% above the control peptide background. Background responses were subtracted from mKRAS responses to determine the frequency of mKRAS-reactive cytokine responses. Responses that did not meet the threshold were not plotted. The gating strategy is shown in Supplementary Fig. 3. mKRAS-reactive T-cell expansion and TCR sequencing Cryopreserved PBMCs from the on-treatment timepoint with the maximal ex vivo IFNγ ELISpot response prior to on-treatment tumor biopsies were thawed, stimulated with the patient’s tumor-specific mKRAS peptide or a control peptide at 10 µg/mL. PBMCs were expanded for 14 days following the post-IVS IFNγ ELISpot expansion protocol. For patient 8, PBMCs were stimulated with each of the six mKRAS peptides or the control peptide at 2 µg/mL and cultured in R10 media with 100 IU/mL IL-2 for 7 days without media refresh. After expansion, cells were harvested for genomic DNA (gDNA) isolation using the DNeasy Blood and Tissue kit (Qiagen, 69504). gDNA was also isolated from unexpanded baseline and on-treatment PBMCs (same timepoint as expansions). TCR CDR3β regions were sequenced via the ImmunoSeq assay (Adaptive Biotechnologies) as previously described in ref. 62. Tumor TCR sequencing and immunohistochemistry Serial 5 µm sections were cut from FFPE blocks onto glass slides, and several slides from each block were stained with haematoxylin and eosin (H&E). A board-certified pathologist reviewed H&E slides to identify patients with paired biopsies containing adequate viable tumor (or regression bed if on-treatment biopsy). Tumor tissue was obtained via manual microdissection of slides and gDNA was extracted using the QIAamp DNA FFPE Tissue Kit (Qiagen #56404). TCR CDR3β regions were sequenced via the ImmunoSeq assay (Adaptive Biotechnologies). IHC was performed by the Tumor Microenvironment Core at the Johns Hopkins School of Medicine. Single-color IHC included staining for HLA-A/B/C (clone EMR8-5, Abcam ab70328-1001), HLA-DR/DP/DQ (clone CR3/43, Santa Cruz sc-53302), pancytokeratin (clone AE1/AE3, Dako M351529-2), and CD20 (clone L26, Leica CD20-L26-L-CE). Dual-color IHC panel included staining for CD4 (clone EP204, Sigma 104R-24) plus CD8 (clone 4B11, Leica CD8-4B11-L-CE). Slides were scanned with a Hamamatsu NanoZoomer and cell densities were quantified using HALO software (Indica Labs). Increases in cell densities in on-treatment versus baseline tumor biopsies were determined using a two-tailed Wilcoxon signed-rank test. TCR repertoire analysis and T-cell clonotype tracking T-cell clonotypes were defined by unique TCRβ sequences (CDR3 amino acid sequence combined with TRBV and TRBJ alleles). In each bulk TCRβ sequencing sample, the total number of sequenced T-cells, nn, was defined as the sum of all productive clonotype reads. The frequency of each clonotype, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${p}_{i}$$\end{document}pi, was defined as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${clonotype\; reads}/n$$\end{document}clonotypereads/n. For log-log scaled scatter plots, clonotypes absent in a condition were assigned a pseudo-frequency of 1/3n1/3n and their absence was indicated with a dashed line. An outlier clonotype of unknown antigen specificity in Patient 29 paired tumor biopsies, representing 47% of baseline and 86% of on-treatment tumor TCR reads, was excluded from the intratumoral TCR repertoire analysis to avoid skewing comparisons. Treatment-enriched clonotypes were defined as newly observed clonotypes in on-treatment peripheral blood that were significantly enriched versus baseline peripheral blood. Significant enrichment was defined as \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{{adjusted}} < 0.01$$\end{document}Padjusted<0.01 via a one-sided Fisher’s exact test with false discovery rate controlled via the Benjamini–Yekutieli procedure. Putative mKRAS-VAX-induced mKRAS-reactive clonotypes were defined as newly emergent clonotypes in on-treatment peripheral blood that were preferentially expanded after in vitro expansion with mKRAS peptide compared with irrelevant control peptide. For each clonotype, enrichment was assessed by comparing its read count in the mKRAS peptide expansion library with its read count in the matched control peptide expansion library using a one-sided Fisher’s exact test. False discovery rate was controlled across clonotypes within each patient using the Benjamini–Yekutieli procedure. Clonotypes were annotated as putative mKRAS-reactive clonotypes if they met all of the following criteria: (1) \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${P}_{{adjusted}} < 0.01$$\end{document}Padjusted<0.01 for enrichment in the mKRAS expansion relative to control, (2) at least fivefold enrichment in the mKRAS expansion, (3) frequency ≥ 0.025% in the mKRAS expansion, and (4) absence from baseline peripheral-blood. Clonotypes meeting these criteria for multiple mKRAS peptides were labeled as cross-reactive. mKRAS-reactive clonotypes sharing identical TCRβ sequences with previously published mKRAS-specific TCRs, mKRAS-reactive clonotypes from other patients, or a clonotype found in tumor biopsies from other patients were classified as public. Clonality was calculated as 1 - Shannon entropy:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${Clonality}=1-\frac{-{\sum }_{i=1}^{n}{p}_{i}{\mathrm{ln}}({p}_{i})}{{\mathrm{ln}}(n)}$$\end{document}Clonality=1−−∑i=1npiln(pi)ln(n) Statistics and reproducibility Due to the limited availability of primary patient samples, experiments were typically performed once with technical replicates. In rare instances, experiments were repeated if control conditions failed and sufficient patient samples were available. Specific statistical methods are detailed in the corresponding subsections in “Methods”. All data visualization, unless otherwise stated, were graphed using Graphpad Prism version 10.4.2. T-cell clonotype scatter plots were graphed using R (version 4.5.1). Kaplan–Meier curves were graphed using R (version 4.3.0). Figures were prepared using Affinity Designer 2.6.3 (Serif). Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Reporting Summary Transparent Peer Review file Supplementary Data 1 Supplementary Information Reporting Summary Transparent Peer Review file Supplementary Data 1 Source data Source Data Source Data