Individualized, heterologous chimpanzee adenovirus and self-amplifying mRNA neoantigen vaccine for advanced metastatic solid tumors: phase 1 trial interim results

Abstract

Checkpoint inhibitor (CPI) therapies provide limited benefit to patients with tumors of low immune reactivity. T cell-inducing vaccines hold promise to exert long-lasting disease control in combination with CPI therapy. Safety, tolerability and recommended phase 2 dose (RP2D) of an individualized, heterologous chimpanzee adenovirus (ChAd68) and self-amplifying mRNA (samRNA)-based neoantigen vaccine in combination with nivolumab and ipilimumab were assessed as primary endpoints in an ongoing phase 1/2 study in patients with advanced metastatic solid tumors (NCT03639714). The individualized vaccine regimen was safe and well tolerated, with no dose-limiting toxicities. Treatment-related adverse events (TRAEs) >10% included pyrexia, fatigue, musculoskeletal and injection site pain and diarrhea. Serious TRAEs included one count each of pyrexia, duodenitis, increased transaminases and hyperthyroidism. The RP2D was 10¹² viral particles (VP) ChAd68 and 30 µg samRNA. Secondary endpoints included immunogenicity, feasibility of manufacturing and overall survival (OS). Vaccine manufacturing was feasible, with vaccination inducing long-lasting neoantigen-specific CD8 T cell responses. Several patients with microsatellite-stable colorectal cancer (MSS-CRC) had improved OS. Exploratory biomarker analyses showed decreased circulating tumor DNA (ctDNA) in patients with prolonged OS. Although small study size limits statistical and translational analyses, the increased OS observed in MSS-CRC warrants further exploration in larger randomized studies.

Main

Cancer immunotherapies have shown promise in harnessing the immune system to target and destroy cancers, leading to clinical benefit enriched in patients with a high mutational burden^1,2,3,4,5. Multiple studies indicate that cytotoxic CD8 T cells targeting tumor neoantigens are critical to tumor control and clearance in response to immunotherapies targeting CTLA-4 or PD-1^6,7,8,9,10. Clinical responses to CPI therapy rely mostly on reinvigorating preexisting tumor-specific T cell responses¹¹, and active vaccination to expand preexisting and prime de novo tumor-specific T cells is anticipated to overcome this limitation.

The limited success of cancer vaccines in the past can be attributed to a number of factors, including selection of poorly immunogenic self-antigens¹², insufficiently immunogenic vaccine platforms and immunosuppressive milieus in patients with advanced cancers⁴. Accordingly, peptide-based neoantigen vaccine platforms have to date failed to consistently induce robust neoantigen-specific CD8 T cell responses in the majority of patients^13,14,15. Although more immunogenic, a homologous prime boost messenger RNA (mRNA)-based vaccination approach elicited predominantly CD4 T cell responses^16,17,18. Cumulatively, previous findings suggest that a successful cancer vaccine should (1) target tumor-specific neoantigens, (2) use highly immunogenic vaccine platform(s), (3) expand and prime T cells, (4) be combined with CPI therapy¹⁹ and (5) generate long-term memory responses to ensure continuous tumor control for durable clinical benefit.

Viral vector-based vaccine platforms, such as recombinant adenovirus, are able to prime robust T cell responses^{20,21,22,23,24}. Although high seroprevalence of anti-adenoviral antibodies in human populations can lead to diminished immune responses to antigens delivered via adenoviral-based vaccines^20,25, preexisting anti-vector immunity can be circumvented by using non-human adenoviral species, such as chimpanzee adenovirus (ChAd). RNA virus-based vaccines have been tested in humans in a range of cancer and infectious disease indications^26,27, and samRNA vectors are of particular interest due to their ability to express high and durable antigen levels²⁸.

In this study, the quality of antigen-specific T cell responses induced by a heterologous vaccine regimen consisting of a ChAd vector (ChAd68) and a fully synthetic Venezuelan equine encephalitis virus-based samRNA vector was assessed in non-human primates (NHPs) and patients with advanced cancers in combination with CPIs. The potency and longevity of the immune response against simian immunodeficiency virus (SIV) model antigens²⁹ and the impact of subcutaneous (s.c.) anti-CTLA-4 administration were assessed in NHPs. Immunogenicity against individualized (patient-specific) predicted cancer neoantigens³⁰ and in-depth analysis of T cell responses elicited by vaccination are being assessed in an ongoing first-in-human phase 1/2 clinical trial (GRANITE, NCT03639714) in patients with metastatic MSS-CRC, non-small cell lung cancer (NSCLC) or gastroesophageal adenocarcinoma (GEA). Primary endpoints of the phase 1 part of the study were safety and tolerability and RP2D. Secondary and exploratory endpoints include immunogenicity, OS, feasibility of manufacturing patient-specific vaccines and changes in ctDNA. Induction of neoantigen-specific CD8 T cell responses was observed in all patients, including patients with MSS-CRC whose tumors lacked immune reactivity before vaccination. Decreases in ctDNA and stabilization of disease and/or reduction in tumor burden in these patients suggest early signs of potential therapeutic benefit in this CPI-refractory patient population^31,32,33.

Results

Vaccination induces durable CD8⁺ T cell responses in NHPs

To assess the potency and durability of the immune responses induced by this heterologous vaccine regimen preclinically in NHPs, a model antigen cassette encoding six well-defined NHP-specific MAMU-A*01-restricted SIV antigens²⁹ was inserted into the ChAd68 and samRNA vaccine platforms. MAMU-A*01-positive rhesus macaques were immunized intramuscularly (i.m.) with ChAd68 prime (n = 11), followed by i.m. samRNA boosts at 4, 12 and 20 weeks, with (n = 6) or without (n = 5) concomitant administration of s.c. anti-CTLA-4 antibody ipilimumab (Fig. 1a and Supplementary Table 1). ChAd68 induced immune responses against all six antigens (mean 923 and 313 spot-forming units (SFU)/10⁶ cells at week 4 ± s.c. ipilimumab, respectively; Fig. 1b). Primed T cell responses were boosted following administration of samRNA, with T cell responses still detectable 32 weeks after prime in both groups. Concomitant administration of s.c. ipilimumab resulted in maintenance of an overall higher T cell response (mean 843 and 277 SFU/10⁶ cells ± s.c. ipilimumab, respectively; Fig. 1b). ChAd68 readministration at week 32 boosted T cell responses ~15-fold and ~12-fold ± concomitant ipilimumab, respectively (Fig. 1c). Immunization resulted in the generation and maintenance of antigen-specific effector memory T cells (Extended Data Fig. 1a), which have been demonstrated to be predictive of providing therapeutic benefit to patients with cancer³⁴.

Fig. 1: Heterologous vaccination with ChAd68 and samRNA induces broad, durable CD8⁺ T cell responses in NHPs that are detectable long-term and can be boosted ≥2 years after prime.

a, Schematic of study treatment schedule. b, Antigen-specific T cell response (IFN-γ SFU per 10⁶ cells) at each measured timepoint for duration of study for each group (mean ± standard error of the mean for each antigen stacked), n = 6 per group up to week 20 (n = 5 for group without ipilimumab at weeks 24–107; Supplementary Table 1). c, Antigen-specific T cell response at weeks 32 and 33 (before and after ChAd68 boost) for each group (n = 5 without ipilimumab, n = 6 with ipilimumab), average response per animal (SFU per 10⁶ cells) for all antigens stacked. d, Antigen-specific T cell response at weeks 106 and 107 (before and after samRNA boost) for each group (n = 5 without ipilimumab, n = 6 with ipilimumab), average response per animal (SFU per 10⁶ cells) for all antigens stacked.

Source data

In both therapeutic and prophylactic vaccine settings, the generation and maintenance of long-lived memory responses is highly desirable to maintain disease control or provide protection from future infections. To investigate whether antigen-specific memory T cells were detectable in NHPs up to 2 years after the ChAd68 prime vaccination, six animals that had not received additional treatment and remained available from the initial study were assessed. Tetramer analysis of peripheral blood mononuclear cells (PBMCs) 80 weeks after ChAd68 prime identified the presence of antigen-specific central memory T cells (Extended Data Fig. 1b). Accordingly, interferon-γ (IFN-γ) ELISpot at week 106 revealed positive responses to four or more of the SIV antigens in all six animals, and samRNA administration at week 106 resulted in an ~8-fold and ~17-fold increase in antigen-specific T cell response detectable at week 107 ± s.c. ipilimumab, respectively (Fig. 1d and Extended Data Fig. 1c). Finally, the ability of the boosted antigen-specific CD8 T cells to kill target cells presenting SIV epitopes was determined. CD8 T cells from all six animals showed cytotoxic activity against their cognate targets, with CD8 T cells from four of six animals achieving killing of >70%, even at the lower effector-to-target ratio of 10:1 (Extended Data Fig. 1d).

Individualized vaccine regimen is safe and well tolerated

After establishing the potency and longevity of T cell responses induced by this vaccine in NHPs, the vaccine was assessed as an individualized, heterologous regimen expressing patient-specific neoantigens in an ongoing phase 1/2 clinical study in patients with advanced cancers (Extended Data Fig. 2a; NCT03639714). The primary objectives for the phase 1 portion of this study were assessment of safety, tolerability and RP2D. A single priming dose of ChAd68 was administered i.m. followed by multiple i.m. samRNA boosts with escalating doses of samRNA (Fig. 2a). Eligible patients included those with metastatic MSS-CRC, NSCLC or GEA who were beginning treatment with standard-of-care chemotherapy during vaccine manufacturing and had tissue available for neoantigen prediction, normal organ function and no contraindications for treatment with immunotherapy (Methods). As of data cutoff, enrollment in phase 1 had been completed, and the study is ongoing. For each patient meeting a threshold for cumulative predicted neoantigen score³⁰ (Extended Data Fig. 2a), the 20 highest ranking mutations were selected for inclusion in a codon-optimized vaccine expression cassette. The same expression cassette was inserted into both patient-specific vaccine vectors for manufacturing. Patient-specific vaccines were successfully manufactured for 15 patients enrolled for vaccine manufacturing, confirming manufacturing feasibility (a secondary endpoint) of this individualized vaccination approach (one patient subsequently elected not to receive study treatment). Fourteen patients received vaccination and monthly intravenous (i.v.) anti-PD-1 antibody nivolumab administration (480 mg every 4 weeks). Patients at dose level 3 (DL3) and DL4 also received s.c. ipilimumab (30 mg) with each vaccination (Fig. 2a and Table 1). Four patients (G1, G8, G9 and G10) received continued standard-of-care chemotherapy with study treatment, and eight patients received a single ChAd68 boost ≥24 weeks after the prime vaccination (Extended Data Table 1). Planned dosing intervals and timepoints with blood draws for immunogenicity are shown in Extended Data Fig. 2b. The median age of the 14 phase 1 patients was 59 years; tumor types included metastatic MSS-CRC (n = 7), GEA (n = 6) or NSCLC (n = 1); and patients had received one or two prior lines of therapy (Table 1 and Extended Data Table 1). The individualized vaccine regimen was safe and well tolerated, with no dose-limiting toxicities. The primary safety endpoint of this vaccine regimen was reflective of a viral-based vaccine regimen with TRAEs consistent with a vaccine-induced immune response (Fig. 2b). TRAEs >10% included pyrexia (86%), fatigue (43%), musculoskeletal and injection site pain (both 36%) and diarrhea (29%). Serious TRAEs were largely consistent with the known safety profile of CPI therapy and included one count each of pyrexia, duodenitis, increased transaminases and hyperthyroidism (Fig. 2b). Based on the totality of the safety and immunogenicity data (secondary endpoint; see next section) from the phase 1 part of the study, the primary endpoint of RP2D was determined to be 10¹² VP ChAd68 and 30 µg samRNA.

Fig. 2: First-in-human data show the vaccine regimen is safe and induces CD8⁺ T cell responses against predicted patient-specific cancer neoantigens.

a, Phase 1 study schematic outlining dose escalation. b, Safety data: TRAEs >10% and all treatment-related serious adverse events (AEs) are shown (n = 14 patients). Graph shows counts of grade 1/2 (blue bars) and 3/4 (orange bars) events. c, Patient PBMC ex vivo IFN-γ ELISpot responses to patient-specific peptide pools (Supplementary Table 2) are shown for baseline (left panel), maximum response following ChAd prime and samRNA boosts (middle panel) and maximum response post ChAd boost (right panel). Graphs show mean SFU per 10⁶ PBMCs ± standard deviation (s.d.) for triplicate ELISpot wells. Assay limit of detection (LOD) and upper limit of quantitation (ULOQ) are indicated by dotted lines. d, Flow plot showing relative CD8 and CD4 T cell frequencies in CD4- or CD8-depleted (depl) PBMC populations. Corresponding ELISpot data (mean SFU of technical replicates ± s.d.) are shown for whole PBMCs (black bars), CD8^enr PBMCs (navy blue bars) and CD4^enr PBMCs (orange bars) stimulated with dimethylsulfoxide (DMSO), patient-specific CD8 or 15mer pools for patients G2 and G8.

Source data

Table 1 Baseline and disease characteristics by DL

Vaccine induces neoantigen-specific CD8⁺ T cells

The key secondary endpoint of immunogenicity was assessed for neoantigen-specific T cells. To preferentially stimulate CD8 T cells via major histocompatibility complex (MHC) class I presentation in immune assays^30,35,36, individual patient minimal epitope (8–11mer) peptide pools were designed to include the 40 highest ranked predicted neoepitopes (CD8 pool; Supplementary Table 2). Because this exploratory minimal epitope pool design does not include all potential neoantigens per mutation and can disproportionately favor some mutations with high prediction scores, additional patient-specific peptide pools consisting of overlapping 15mers spanning each vaccine cassette were also used to interrogate T cell responses more broadly (15mer pool; Supplementary Table 2). Posttreatment PBMC samples were available for 13 of 14 patients (all except patient G5). Stimulation of PBMCs with patient-specific peptide pools resulted in IFN-γ responses detectable by ex vivo ELISpot following heterologous prime boost vaccination in 100% of patients (13/13), with a ChAd68 boost further increasing peripheral T cell levels in 67% of patients (4/6) with evaluable post-boost samples (Fig. 2c and Extended Data Fig. 3a). Only 15% of patients (2/13) had responses above LOD to their cognate neoantigen peptide pool at baseline, which were increased following vaccine administration (Fig. 2c). Similarly to observations in NHPs, vaccine-induced T cell responses were long-lived, and T cell levels of 700–5,000 SFU/10⁶ cells were maintained for >52 weeks in three patients with available longitudinal samples (Extended Data Fig. 3a,b). Notably, readministration of ChAd68 resulted in elevated T cell levels detectable 12–24 weeks after ChAd68 boost (patients G1, G3, G8 and G11; Fig. 2b and Extended Data Fig. 3a). Dose escalation of samRNA suggested that more consistent boosts in T cell responses were observed in patients receiving 30-µg and 100-µg doses of samRNA (Extended Data Fig. 3c). Prolonged samRNA dosing intervals of 8–16 weeks occurring due to logistical constraints during coronavirus disease 2019 (COVID-19) restrictions revealed interesting insights into the kinetics of samRNA-boosted T cell responses (Extended Data Fig. 3a,b). In patient G1 (GEA, concomitant chemotherapy), T cells expanded 1.5-fold during an 8-week interval between samRNA boosts and remained consistently >700 SFU/10⁶ cells during a 16-week period following samRNA administration at week 52. A sevenfold increase in CD8 T cell responses between samRNA doses at weeks 24 and 36 was observed in patient G3 (NSCLC). Patient G2 (GEA) T cell responses showed ~1.5-fold contractions and expansions of the T cell response during a 9-week samRNA dosing interval between weeks 4 and 13. These data suggested that lower doses and longer boost intervals may be beneficial in driving T cell responses with this new samRNA platform. This led to the selection of 30 µg samRNA as the RP2D, a primary endpoint (see previous section) and longer boost intervals (8 weeks) being assessed in the ongoing phase 2 part of this study.

CD8⁺ T cells recognize neoantigens in multiple HLA alleles

Responses to cognate CD8 pools and 15mer pools were analyzed for CD8 versus CD4 T cell bias in five patients with available samples (G1, G2, G8, G11 and G14). As expected based on peptide length, IFN-γ responses following CD8 pool stimulation were driven by CD8 T cells, whereas responses to 15mer pool stimulation were either CD8-driven or mixed CD4/CD8 T cell responses (Fig. 2d, Extended Data Fig. 4a). Active vaccination against tumor neoantigens should lead to expansion of preexisting and priming of de novo tumor-specific T cell responses. To assess the breadth of responses with limited available PBMC samples, 4 minipools covering the 20 patient-specific vaccine neoantigens (Supplementary Table 2) were used to partially deconvolute T cell responses in 12 of 13 patients. Positive responses to four different minipools indicate responses to at least four different mutations. Antigen-specific expansion of T cells in short in vitro stimulation (IVS) cultures^30,37 confirmed expansion of preexisting and priming of de novo T cell responses after treatment (Extended Data Fig. 4b). All 12 patients tested had detectable T cell responses to multiple mutations at posttreatment timepoints (Fig. 3a). Voluntary leukapheresis and pooled posttreatment timepoints enabled full deconvolution of T cell responses against all 40 minimal epitopes included in the CD8 pool for patients G1, G2 and G8, revealing positive responses to ≥50% of vaccine mutations in multiple HLA class I alleles for all three patients (Fig. 3b,c and Extended Data Fig. 5). These data indicate that partial deconvolution via minipools (often necessitated due to limitations in sample availability) likely underestimates the true breadth of neoantigen-specific responses induced by this heterologous vaccine. Peptide-HLA (pHLA) tetramers were generated to assess the breadth and memory phenotype of antigen-specific T cells. pHLA binding was detected in the context of multiple HLA class I alleles for patients G1, G2 and G8. Tetramer-positive CD8 T cells were characterized by 1.3- to 1.5-fold higher frequencies of effector memory and up to 20-fold lower frequencies of naive T cells when compared to total CD8 T cells (Fig. 3d and Extended Data Fig. 6a). ChAd68 boost in patient G1 resulted in a 10-fold increase in tetramer positive cells, with post-boost tetramer-positive CD8 T cells consisting of ≥93% effector memory cells in the context of both class I alleles tested (Fig. 3e and Extended Data Fig. 6b).

Fig. 3: Vaccine induces polyclonal effector memory CD8 T cell responses across multiple neoantigens and HLAs.

a, Patient PBMCs from available posttreatment timepoints (G1 W12, G2 W5, G3 W8, G4 W6, G7 W4, G8 W12, G9 W8, G10 W16, G11 W16, G12 W12, G13 W12, G14 W13 and G15 W12) were stimulated overnight in ex vivo or post-IVS IFN-γ ELISpot with patient-specific minipools or DMSO. Graphs show percentage of total response of mean IFN-γ responses to minipool 1 (turquoise), minipool 2 (light blue), minipool 3 (medium blue) and minipool 4 (navy blue) for each patient. b, Number of positive minimal epitopes detected (positive by ex vivo or post-IVS ELISpot) for each mutated gene. Undetectable responses against epitopes tested are indicated by ‘x’. c, Pie charts showing patient-specific predicted HLA class I presentation of positive neoantigen responses shown in panel b. d, Density plots showing recognition of pHLA (A*30:04, A*02:01, B*08:01) and corresponding T cell memory phenotypes for patient G2. Pie charts showing frequencies of naive (CD45RA⁺CCR7⁺; purple), central memory (CM; CD45RA⁻CCR7⁺; blue), effector memory (EM; CD45RA⁻CCR7⁻; green) and T effector memory RA⁺ (TEMRA; CD45RA⁺CCR7⁻; orange) cells in total CD8⁺ population (left-most chart) and CD8⁺tetramer⁺ populations from corresponding dot plots. e, Density plots and corresponding pie charts showing recognition of pHLA (B*44:05) and corresponding T cell memory phenotypes before and after ChAd68 boost for patient G1.

Source data

CD8⁺ T cells are polyfunctional and kill target cells

In addition to driving broad T cell responses, a potent vaccine regimen should induce cytotoxic T cells that are able to kill target cells presenting cognate epitopes in the context of MHC class I. Secretion of cytotoxic markers assessed by Meso Scale Discovery (MSD) and triplex FluoroSpot was confirmed in eight patients (Extended Data Fig. 6c,d). Notably, readministration of ChAd68 resulted in increased granzyme B secretion following CD8 pool stimulation in four of six patients with evaluable post-ChAd68 boost timepoints (Extended Data Fig. 6e). To assess whether this cytotoxic signature was driven by CD8 T cells, cytokine secretion was analyzed via intracellular cytokine staining (ICS). Stimulation of ex vivo PBMCs with CD8 pool resulted in detectable frequencies of IFN-γ⁺, CD107α⁺, tumor necrosis factor α (TNF-α)⁺ and interleukin-2 (IL-2)⁺ CD8 T cells in patients G2 and G8, which were increased ~2-fold to ~700-fold following IVS expansion (Fig. 4a and Extended Data Fig. 7a). Boolean gating confirmed the presence of quadruple-, triple-, and double-positive polyfunctional CD8 T cells for both patients (Fig. 4b, Extended Data Fig. 7b). Polyfunctional CD4 T cell responses following cognate 15mer pool stimulation were also detectable in patients G2 and G8 following IVS culture (Extended Data Fig. 7c). The ability of antigen-specific CD8 T cells to kill target cells presenting neoantigen was determined. Co-culture of expanded neoantigen-specific patient T cells with single HLA allele-expressing target cells pulsed with cognate neoantigen peptides resulted in a 2.5-fold to 6.4-fold decrease in live target cells over time compared to controls, indicating antigen-specific T cell-driven killing of target cells presenting cognate pHLA for both patients G2 and G8 (Fig. 4c and Extended Data Fig. 7d).

Fig. 4: Neoantigen-reactive CD8 T cells are polyfunctional, kill targets presenting cognate peptides in multiple HLA alleles and are increased in on-treatment tumor samples.

a, Representative density plots for patient G2 ICS for CD8⁺ cells producing IFN-γ, CD107α, TNF-α and IL-2 in ex vivo or IVS-expanded PBMCs stimulated overnight with DMSO or patient-specific CD8 pool. Cells were gated on PBMCs (forward scatter (FSC)-A versus side scatter (SSC)-A), singlets (FSC-A versus FSC-H), live cells (FSC-A versus Live/Dead), CD8 or CD4 T cells (CD8 versus CD4) and cytokine⁺ cells. b, Polyfunctionality was assessed via Boolean gating of CD8⁺cytokine⁺ populations (background subtracted). Pie charts show polyfunctionality of ex vivo PBMCs and post-IVS PBMCs. Pie arcs depict individual cytokines in red (CD107α), lime green (IFN-γ), turquoise (TNF-α) and navy blue (IL-2). Pie sections depict frequencies of cells expressing multiple cytokines as outlined in pie category legend. Main functional pie slices are colored red (quadruple positive for CD107α, IFN-γ, TNF-α and IL-2), orange (triple positive for CD107α, IFN-γ and TNF-α) and yellow (double positive for CD107α and IFN-γ). c, Target cell killing by IVS-expanded PBMCs from patients G2 (B*08:01, B*57:01, A*30:04). Graphs show relative viability based on NucRed count and normalized to effector-to-target ratio of 10:1 DMSO control over time. Data are shown as mean ± s.d. from duplicate wells for targets alone stimulated with DMSO (gray circles) or cognate peptide (black squares) and for 10:1 effector/target co-cultures stimulated with DMSO (light blue circles) or cognate peptide (navy blue squares). d, DNA sequencing of TCR-β chains in tumor samples in baseline and on-treatment samples from patients G1 and G3. TCR-β chains expanded in PBMCs, in both PBMCs and tumors, and contracted are shown.

Source data

T cells are increased in tumors following vaccination

To exert anti-tumor effects and clinical benefit in patients, vaccine-induced T cells need to infiltrate tumors and lyse cells presenting their cognate pHLA antigen through recognition via the T cell receptor (TCR). Paired TCR sequencing and gene expression profiling via single-cell RNA sequencing was performed on baseline and on-treatment PBMC timepoints from patients with available corresponding tumor biopsies (G1 and G3; Extended Data Fig. 8a). Neoantigen-reactive CD8 T cells were increased 1.6-fold to 6.4-fold in on-treatment samples compared to baseline (Extended Data Fig. 8b). Analysis of transcriptional activation signatures confirmed an enrichment for CD8 T cells in on-treatment samples (Extended Data Fig. 8c). TCR analysis revealed a total of 12 and 30 distinct TCR clonotypes (α-β pairs) detected in PBMCs following vaccination for patients G1 and G3, respectively (Extended Data Fig. 8b). Analysis of IFNG, GZMB, TNFA and IL2 transcripts in T cell clonotypes revealed 12- to 98-fold increases in frequencies of T cells expressing two or more cytokine transcripts in on-treatment samples compared to baseline for both patients (Extended Data Fig. 8d,e and Supplementary Table 3). Baseline and on-treatment tumor biopsy samples from patients G1 and G3 were analyzed for the presence of TCR-β CDR3 sequences via bulk DNA sequencing. Analysis of TCR-β CDR3 sequences obtained independently via single-cell RNA sequencing in PBMCs versus DNA sequencing in tumor samples indicated overlap of putative neoantigen-specific T cells in both the blood and tumor. For patient G1, 341 TCR-β CDR3 clonotypes were identified across baseline and on-treatment tumor samples with an enrichment of nine clones in the on-treatment biopsy. Two of these nine clonotypes were also found in PBMCs sampled at the time of the on-treatment biopsy (Fig. 4d). In patient G3, a total of 294 TCR-β CDR3 clonotypes were identified via DNA sequencing in both tumor samples, with 40 clonotypes enriched in the on-treatment sample. Five of the 40 TCR-β CDR3 clonotypes enriched in the on-treatment biopsy overlapped with the activated TCR clonotypes in PBMCs at the time of the on-treatment biopsies (Fig. 4d). Representation of clonotypes from neoantigen-specific CD8 T cells in tumor tissue that were also detected in PBMCs provides proof-of-concept that vaccine-induced neoantigen-specific T cells in the periphery are able to infiltrate tumors.

Prolonged survival and decreases in ctDNA in MSS-CRC

Multiple measures of activity were assessed, including OS and changes in tumor size via imaging (objective response rate and progression-free survival per RECIST v1.1, all secondary endpoints), as well as transcriptional profiling and changes in ctDNA (both exploratory endpoints). Standard genomic correlates of responsive tumors to CPIs (namely, tumor mutational burden, CD274 (PD-L1) and IFNG RNA sequencing signatures) were analyzed in patient pretreatment biopsies and compared to relevant cancer types from The Cancer Genome Atlas (TCGA). Pretreatment patient genomic data demonstrate that patients with MSS-CRC and GEA in this study had low tumor mutational burden and PD-L1 expression and that these tumors had low immune reactivity (Fig. 5a,b and Supplementary Table 4). Such cold tumors are known to be refractory to CPI therapy. Objective response rate and progression-free survival per RECIST v1.1 show a patient with GEA with a complete response and several patients across tumor types with stable disease (SD) (Extended Data Table 1). Given the observation of SD in several patients with MSS-CRC, we focused on analyzing these seven patients with MSS-CRC for the secondary endpoint of OS and an exploratory endpoint of changes in ctDNA from baseline (Extended Data Table 1 and Supplementary Table 4). For patients with MSS-CRC, 3 of 7 (42.9%) remained alive (median OS of 8.7 months; Extended Data Table 1), with an OS rate at 12 months of 42.9% reflecting multiple patients experiencing durable benefit.

Fig. 5: Early signals of clinical efficacy in patients with MSS-CRC in advanced disease setting.

a, Scatter plot showing tumor sample RNA sequencing (RNA-seq) for patients before study treatment. PD-L1 normalized RNA-sequencing expected counts (y axis) versus tumor mutational burden (x axis) for patients and relevant TCGA cancer types. b, Patient RNA-sequencing IFN-γ signature average z-scores (x axis) calculated across patients and relevant TCGA cancer types (y axis). TCGA samples: MSS-CRC (n = 286), MSI-CRC (n = 62), gastric (n = 406), lung adenocarcinoma (Lung Ad.)_ (n = 498), lung squamous (Lung Sq.) (n = 463) and melanoma (n = 436). GRANITE patients: MSS-CRC (n = 7), GEA (n = 6) and NSCLC (n = 1). c, Percent change in ctDNA following initiation of study treatment relative to baseline levels shown as best response and colored by RECIST-based response and time on treatment. Patients in blue were all on study treatment >24 weeks with two patients treated beyond RECIST-based progressive disease (PD) that was not confirmed by 24 weeks and continued on study treatment. d, Probability of survival in MSS-CRC patients with (n = 4) or without (n = 3) reduction in ctDNA since ChAd68 prime. Data cutoff 31 January 2022. e, Radiological images showing tumor lesions in the lung and liver of patient G8 before initiating study treatment (baseline) and following study treatment (16 and 48 weeks). Fluorodeoxyglucose-positron emission tomography imaging of tumor lesions in the liver when patient was diagnosed in February 2018 and 7 months after starting study treatment. Arrows indicate relevant target lesions.

Source data

Emerging evidence suggests changes in ctDNA levels on-treatment relative to baseline are a useful biomarker to stratify patients as molecular responders or nonresponders and differentiate patients with a good prognosis, especially those with SD per RECIST v1.1 criteria³⁸. Of these patients, 100% (3/3) with treatment durations <6 months had increases in ctDNA concentration, whereas 75% of patients (3/4) with treatment durations ≥6 months demonstrated a molecular response (<50% from baseline; Fig. 5c and Supplementary Table 4). Two patients had complete ctDNA clearance while on treatment (G8 and G14; Fig. 5c). Most notably, these decreases in ctDNA were more frequent in patients with prolonged OS (median OS not reached versus 7.8 months; Fig. 5d). Patient G8, a 51-year-old patient with MSS-CRC who received chemotherapy for 21 months, had progressed on first-line chemotherapy before study treatment initiation but continued concomitant standard of care chemotherapy during study treatment (Extended Data Table 1) and showed clear evidence of tumor regression following vaccine administration via both ctDNA and radiologic assessments. Computed tomography scans showed size reduction in multiple lung lesions within 16 weeks and persistent reduction in multiple lung and one liver lesion through 48 weeks (Fig. 5e). Fluorodeoxyglucose-positron emission tomography imaging of the stable liver lesions showed a lack of hypermetabolic cells following study treatment administration, suggesting anti-tumor activity of this vaccine regimen not evidenced by computed tomography imaging (Fig. 5e). In totality, the safety, immunogenicity, OS and ctDNA signals observed in this trial warrant further investigation of this vaccine platform in patient populations that are refractory to CPI monotherapy^31,32,33.

Discussion

The quality of antigen-specific T cell responses induced by a heterologous vaccine regimen, consisting of ChAd68 and samRNA vectors in combination with CPI, was assessed in NHPs and patients with advanced cancers. The potency and longevity of T cell responses to model antigens following vaccination were studied in NHPs, whereas safety, immunogenicity and feasibility of an individualized patient-specific vaccine were assessed in patients with late-stage cancer. This vaccine regimen was well tolerated and induced neoantigen-specific CD8 T cell responses in all patients. Vaccine-driven T cell responses were (1) polyclonal across multiple mutations and HLA alleles, (2) polyfunctional and (3) able to kill target cells presenting cognate peptide. Of note, more consistent boosts in T cell responses were observed in patients receiving lower doses of samRNA. This could be due to less innate immune activation via mRNA-sensing pattern recognition receptors that might interfere with samRNA amplification and antigen expression. Increased frequencies of vaccine-induced TCR clonotypes were found in on-treatment tumor samples, providing proof-of-concept for tumor infiltration of vaccine-induced T cells. Patients with molecular responses following vaccination showed stabilization and/or reduction of target lesions and prolonged OS, indicating potential early signs of clinical activity in this CPI-refractory patient population. A randomized phase 2/3 study has been initiated to demonstrate the efficacy of this vaccine regimen in patients with MSS-CRC in the first-line metastatic setting (NCT05141721).

A vaccine that induces neoantigen-specific CD8 T cells provides a foundation for combination with CPIs, making their combined administration highly desirable. Although the effects of s.c. ipilimumab were difficult to discern in this small cohort of patients, administration of ipilimumab in addition to a vaccine in NHPs resulted in increased breadth and magnitude of antigen-specific T cell responses in all animals. This may be due to a lower threshold for T cell activation against weaker epitopes with ipilimumab compared to vaccine alone. The lower threshold of T cell activation provided by the addition of ipilimumab might allow induction of an overall broader immune response to vaccine-encoded neoantigens and a more robust immune attack of the tumor. Driving broad T cell responses against multiple tumor antigens across multiple HLA alleles, as observed in this study, may limit the impact of common HLA-related immune escape mechanisms^39,40,41. As a patient’s disease progresses, increased heterogeneity can lead to immune evasion. Treating patients earlier in their disease, such as in the adjuvant setting or early in the treatment of metastatic disease, has the potential benefit of completely eradicating tumor cells by vaccine-induced T cells before resistance mutations arise. Also, given the time lag between vaccine administration and T cell expansion, patients with lower initial disease burden and more time to generate a T cell response before disease progression are more likely to receive long-term benefit and durable disease control from our vaccination approach. Increasing evidence suggests that a clear survival benefit (even without a radiologic response) may be a hallmark of immunotherapies⁴². Follow-up of this ongoing phase 1/2 study is ongoing, and one phase 2/3 and one phase 2 study are assessing potential clinical activity of this vaccine regimen in patients with MSS-CRC in the metastatic and adjuvant treatment settings, respectively. In addition to benefiting patients with cancer, this versatile and potent vaccine can be applied to other indications such as prophylactic vaccination against SARS-CoV-2 (ref. ⁴³) (NCT05148962, NCT04776317). In sum, this versatile heterologous vaccine regimen generates strong, persistent and functional immune responses that have broad applicability to a wide range of disease settings and support continued exploration in the clinical setting to bring better treatment options to patients with cancer and people at risk of infection.

Methods

Ethics statement

The research presented in this paper complies with all the relevant ethical regulations. Animal studies were performed according to institutional animal care and use (IACUC)-approved protocol at the California National Primate Research Center at the University of California, Davis. The clinical study and all related analyses were carried out in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines and was approved by the appropriate institutional review board (IRB) or ethics committees at each participating site (University of Chicago Comprehensive Cancer Center, Chicago, IL; Columbia University Medical Center, New York, NY; The Ohio State University Medical Center, Columbus, OH; Virginia Cancer Specialists, Fairfax, VA; Memorial Sloan Kettering Cancer Center, New York, NY; Peter MacCallum Cancer Centre, Melbourne, VIC, Australia; Mayo Clinic, Phoenix, AZ; Mayo Clinic, Rochester, MN; The University of Texas MD Anderson Cancer Center, Houston, TX). All patients provided written, informed consent. Further details can be found at https://clinicaltrials.gov/ct2/show/NCT03639714.

NHP study

Animal selection and dosing

Study was performed according to IACUC-approved protocol at the California National Primate Research Center at the University of California, Davis. Animals did not receive any prior treatment with immune-modulatory antibodies or vaccination against SIV and had no prior exposure to SIV. Animal groups and dosing are outlined in Supplementary Table 1. Briefly, both ChAd68 and samRNA were administered as bilateral i.m. injections into the quadriceps muscle. All ChAd68 doses were at 10¹² VP with 2 ml total injection volume (1 ml per leg). All samRNA doses were at 300 μg with 1 ml total injection volume (500 μl per leg). Anti-CTLA-4 (ipilimumab) was administered as either an i.v. injection of 5 mg kg⁻¹ or a flat dose of 50 mg per animal injected subcutaneously bilaterally near each vaccine site draining lymph node. Initial weights, doses of ipilimumab and treatment groups are shown below. Animals 37576 and 42300 were excluded from the study after week 20 with no further immune monitoring or immunizations due to recurrent issues with anemia and low-yield blood draws.

SIV model antigen cassettes

A model antigen cassette containing six previously published SIV viral epitopes²⁹ (SIV Gag CM9 (CTPYDINGQM), SIV Env TL9 (TVPWPNASL), SIV Env CL9 (CAPPGYALL), SIV Tat TL8 (TTPESANL), SIV Gag LW9 (LSPRTLNAW) and SIV Pol SV9 (SGPKTNIIV)) was designed. Each minimal epitope is embedded in a 25 amino acid sequence with 8 amino acids flanking each minimal epitope sequence. The cassette also contains class II helper epitopes. The cassette was synthesized by GenScript and designed to allow cloning into plasmids containing ChAd and samRNA vectors. The ChAd vector is an E1, E3 deleted virus, and the cassette expression was flanked by a cytomegalovirus promoter and an SV40 polyadenylation signal. The ChAd plasmid was digested with PacI and then transfected into HEK293 cells to generate virus. The virus for the NHP studies was made by CsCl gradient centrifugation and for the good manufacturing practice studies was purified by anion exchange chromatography. The same cassette was cloned into BstB1–PacI sites of a pBAC-VEE alphavirus-based plasmid that was used to make the samRNA RNA by in vitro transcription (TriLink) and then formulated into lipid nanoparticles (Arbutus).

Test articles (NHP study)

Anti-CTLA-4 monoclonal antibody (mAb) (ipilimumab) was provided by Bristol Myers Squibb. ChAd68-MAG was generated by Gritstone Oncology (non-GLP) and prepared at 5 × 10¹¹ VP ml⁻¹ in Adenovirus Reference Material buffer (10 mM Tris, pH 8.0, 25 mM NaCl and 2.5% glycerol buffer). samRNA was generated by TriLink Biotechnologies and formulated in lipid nanoparticles by Arbutus.

Isolation of PBMCs and serum (NHP study)

For immune monitoring, 10–20 ml blood was collected into vacutainer tubes containing heparin and maintained at room temperature until isolation. PBMCs were isolated by density centrifugation using lymphocyte separation medium and Leucosep separator tubes. PBMCs were counted using the Attune NxT (Thermo Fisher Scientific) and resuspended at a concentration of 4 × 10⁶ cells ml⁻¹ in complete RPMI.

IFN-γ ELISpot assay (NHP study)

IFN-γ enzyme-linked immunospot (ELISpot) assays were performed using precoated 96-well plates (MAbtech, Monkey IFNγ ELISpot PLUS, ALP) following the manufacturer’s protocol. Then, 10⁵ PBMCs per well were plated in triplicate with 10 µg ml⁻¹ peptide stimuli (GenScript) and incubated overnight in complete RPMI (RPMI + 10% FBS). A human hepatitis B virus S-antigen peptide not contained in the cassette (WLSLLVPFV, GenScript) was used as a negative control for each sample. Plates were washed with PBS and incubated with anti-monkey IFN-γ mAb biotin (MAbtech) for 2 h, followed by an additional wash and incubation with Streptavidin-ALP (MAbtech) for 1 h. After final wash, plates were incubated for ten minutes with BCIP/NBT (MAbtech) to develop the immunospots and dried overnight at 37°C. Spots were imaged and enumerated using AID reader (Autoimmun Diagnostika). For data processing and analysis, samples with replicate well variability (variability = variance/(median + 1)) greater than 10 and median greater than 10 were excluded⁴⁴. Spot values were adjusted based on the well saturation according to the formula⁴⁵:

$${{{\mathrm{AdjustedSpots}}}} = {{{\mathrm{RawSpots}}}} + 2 \ast \left( {{{{\mathrm{RawSpots}}}} \ast {{{\mathrm{saturation}}}}/\left( {100 - {{{\mathrm{saturation}}}}} \right)} \right.$$

Each sample was background corrected by subtracting the average value of the negative control peptide wells. Data are presented as SFU per 10⁶ PBMCs. Wells with well saturation values greater than 50% were labeled as too numerous to count, and their value was set to the maximum measured SFU for that experiment.

ICS assay (NHP study)

Freshly isolated PBMCs were distributed at 10⁶ per well into V-bottom 96-well plates. Cells were pelleted and resuspended in 100 μl complete RPMI containing 1 μg ml⁻¹ anti-CD28 mAb (CD28.2, BD Biosciences), 1 μg ml⁻¹ anti-CD49d mAb (9F10, BD Biosciences) and 10 μg ml⁻¹ pooled peptide antigens or human hepatitis B virus S-antigen peptide (WLSLLVPFV) as a negative control stimulant (GenScript). After 1 h of stimulation, Brefeldin A (Biolegend) was added to a final concentration of 5 μg ml⁻¹ and incubated for an additional 16 h. Following stimulation, cells were washed with PBS and stained with fixable viability dye (eBioscience). Extracellular staining was performed in FACS buffer (PBS + 2% FBS + 2 mM EDTA) with anti-CD3, anti-CD4, anti-CTLA-4, anti-PD1 and anti-CD8 antibodies. Cells were washed, fixed and permeabilized with Transcription Factor fixation/permeabilization kit (eBioscience), and then intracellular staining was performed in permeabilization buffer. Samples were stained for IFN-γ, TNF-α, granzyme B and CD69. Samples were collected on the Attune NxT (Thermo Fisher Scientific). Gating strategy was as follows (Supplementary Fig. 1): lymphocytes (SSC-A versus FSC-A), single cells (SSC-H versus SSC-A), viable cells (SSC-A versus LD-506), CD4⁺ or CD8⁺ (CD4-phycoerythrin (PE) versus CD8-e450), cytokine⁺ (TNF-α-BV421 versus IFN-γ-APC-R700).

Tetramer staining assay (NHP study)

Following overnight rest at 4°C, isolated PBMCs were distributed into 96-well V-bottom plates at a density of 10⁶ cells per 100 µl and stained with a viability dye (Live/Dead e506, eBioscience) in PBS for 20 min at 4°C. TCR recycling was inhibited by treatment with 50 nM Dasatinib for 1 h at 4°C. Cells were stained for 1 h at 4°C in FACS buffer with Mamu-A*01 tetramers for each of the peptide antigens with a unique combination of two fluorophores per antigen (produced in-house). Cells were co-stained for CD3 (A700, SP34.2, BD Biosciences), CD8 (e450, SK1, BD Biosciences), CD45RA (FITC, 5H9, BD Biosciences) and CCR7 (SB780, 3D12, eBioscience). Samples were collected on the Attune NxT (Thermo Fisher Scientific). The gating strategy was as follows (Supplementary Fig. 1): lymphocytes (SSC-A versus FSC-A), single cells (SSC-H versus SSC-A), viable cells (SSC-A versus LD-506), CD3⁺ cells (SSC-A versus CD3-A700), CD8⁺ (SSC-A versus CD8-e450), tetramer double positive (tetramer-UV405 versus tetramer-UV605; tetramer-UV405 versus tetramer-PE; tetramer-PE versus tetramer-UV675; tetramer-PE-Cy7 versus tetramer-UV405) and memory populations (CD45RA-FITC versus CCR7-SB780).

Cytotoxicity (T cell killing) assay (NHP study)

Freshly isolated PBMCs for each sample were divided into two portions. T-cell enrichment was performed on one portion using the Miltenyi NHP T-cell isolation kit (130-092-143), following manufacturer’s protocol and using MS columns (Miltenyi) to deplete non-T cells. The second portion (referred to as PBMCs) was stained with carboxyfluorescein succinimidyl ester (CFSE) at 1 mM (eBioscience). The CFSE-labeled PBMCs were then loaded with 50 mg ml⁻¹ SIV peptides or a negative control peptide (WLSLLVPFV) by incubation for 1 h at 37˚C in RPMI + 1% FBS. Isolated T cells and peptide-loaded target PBMCs were then counted and plated in a 20:1 ratio (T cells to target cells) in a V-bottom plate. Then, 400,000 T cells and 20,000 target cells were plated per well. Target cells only were also plated as a control. Cells were incubated for 6 h at 37˚C in RPMI + 10% FBS. Following incubation, cells were stained with 7-AAD (Thermo Fisher Scientific) and immediately analyzed by flow cytometry using a Cytoflex LX (Beckman). CFSE⁺ target cells were gated and 7-AAD⁺ cells measured (as a percentage of CFSE⁺ cells). Remnant T cells were also analyzed by flow cytometry to confirm T cell isolation. Samples were collected on the Attune NxT (Thermo Fisher Scientific). The gating strategy was as follows (Supplementary Fig. 1): lymphocytes (SSC-A versus FSC-A), single cells (SSC-H versus SSC-A), 7-AAD⁺CFSE⁺ (7-AAD versus CFSE).

Anti-ChAd68 neutralizing antibody assay (NHP and clinical study)

Dilutions of heat-inactivated human and NHP sera were evaluated for the ability to neutralize a ChAd virus expressing the reporter protein beta-galactosidase (ChAd-LacZ). The reciprocal of the dilution that gives 50% or greater neutralization is used to assign a neutralization titer. In brief, pre- and post-vaccination serum samples were heat inactivated at 56˚C for 30 min. Sera samples were then diluted twofold in serum-free media (DMEM) starting at 1:20 dilution down to 1:1,280. If needed, additional dilutions were made to get a titer within the range of the assay. The diluted samples were then mixed with an equal volume of media containing diluted virus sufficient to infect 293 A cells at an approximate multiplicity of infection (m.o.i.) of 1. The diluted sera and virus were incubated for 1 h at 37˚C before adding onto 293 A cells plated at 2.5 ×10⁴ cells/well on a 96-well tissue culture plate 1 day earlier. The cells and neutralizing mix were incubated for 1 h at 37˚C and then an equal volume of media with 20% FBS was added. The cells and virus were then incubated for an additional 18–20 h before lysis and measurement of B-galactosidase activity. Media was aspirated and the cells lysed in 100 µl lysis buffer from Thermo Fisher Scientific (Galacto-Star System). The cells were then freeze thawed twice (alternating −80˚C and 37˚C), mixed, and then 50 µl was mixed with an equal volume of Galacto-Star substrate (Thermo Fisher Scientific) for B-galactosidase measurement. Wells with media alone were used to determine expression in the absence of sera (100% B-galactosidase activity) and wells with no virus were used to determine the background of the assay. The percent inhibition was calculated for each dilution and plotted to generate a curve of percent neutralization versus dilution. A titer was assigned either as a limiting dilution titer, which is the reciprocal dilution that gives 50% or greater inhibition, or as interpolated from the dilution curve.

Phase 1 clinical trial

Study design

The study protocol is available in the Supplementary Information. The main objectives of this phase 1/2 study (NCT03639714) were to evaluate the safety and tolerability, immunogenicity and early clinical activity to determine and evaluate the RP2D of i.m. administration of GRT-C901 (ChAd68) and GRT-R902 (self-amplifying RNA formulated in lipid nanoparticles; samRNA), an individualized neoantigen cancer vaccine, in combination with i.v. nivolumab and s.c. ipilimumab, in patients with metastatic NSCLC, MSS-CRC, GEA and metastatic urothelial cancer. The phase 1 portion of the study was designed to find an RP2D using an adaptive dose-escalation method and modified toxicity probability interval-2 (mTPI-2) among the planned doses; DL1 through DL4 are outlined in Fig. 2a, with the goal of determining the highest vaccine dosage with a dose-limiting toxicity (DLT) rate below 30%.⁴⁶ All patients received 10¹² VP ChAd68 prime and monthly i.v. anti-PD-1 antibody nivolumab administrations (480 mg every 4 weeks). Patients at DL1 and DL2 received 30 µg and 100 µg samRNA boosts, respectively. Patients at DL3 and DL4 received 100 µg and 300 µg samRNA, respectively, in addition to s.c. ipilimumab (30 mg) with each vaccination. ChAd68 boost was administered to eight patients with varying intervals at a dose of 10¹² VP (Extended Data Table 1). The sample size was not driven by a statistical power due to the nature of an adaptive dose-escalation method. Therefore, the sample size of the phase 1 portion of the study ranged from 11 to maximum of 24 patients which is dependent on occurrence of DLTs. This study was carried out in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines and was approved by the appropriate IRB or ethics committees at each participating site (University of Chicago Comprehensive Cancer Center, Chicago, IL; Columbia University Medical Center, New York, NY,; The Ohio State University Medical Center, Columbus, OH; and Virginia Cancer Specialists, Fairfax, VA). All patients provided written, informed consent. Further details can be found at https://clinicaltrials.gov/ct2/show/NCT03639714. Per protocol, patients had the potential to receive study treatment concurrent with ongoing standard-of-care chemotherapy.

The first patient enrolled on 24 September 2018 and the last patient enrolled on 25 November 2019 for the phase 1 portion of the study. As of the data cutoff date of 31 January 2022, a total of 29 patients (i.e., 14 patients in phase 1 and 13 patients in phase 2) have been treated, and the study is still ongoing to evaluate the efficacy of the RP2D using overall response rate, duration of response, clinical benefit rate, deepening of response, progression of survival and OS. Of the 14 phase 1 patients reported on in this study, patients G8, G9 and G10 received continued standard-of-care chemotherapy with study treatment. G8 received fluorouracil (5-FU) and bevacizumab (initiated 21 months before study treatment), and G9 and G10 received fluoropyrimidine/irinotecan (FOLFIRI) (initiated 12 and 8 months before study treatment). Four clinical sites enrolled and treated patients for the phase 1 part of the study: University of Chicago Comprehensive Cancer Center, Columbia University Medical Center, The Ohio State University Medical Center and Virginia Cancer Specialists. There is no data safety monitoring board for this study. A study committee consisting of the investigators and study sponsor meets regularly to assess the safety profile of the study treatment and make recommendations for dose escalation. Patient data were collected via electronic data capture and analyzed as described in the Statistics and reproducibility section.

Study population

Inclusion criteria

Inclusion criteria for vaccine production stage

Patients must meet the following inclusion criteria to be eligible for vaccine production stage:

• 1. Provide a signed and dated Informed Consent Form (ICF) document before initiation of study-specific procedures.

• 2. Patients with the indicated advanced or metastatic solid tumor as follows:

  • a. NSCLC who are planned for or have received no more than one cycle of systemic treatment with cytotoxic, platinum-based chemotherapy (patients receiving anti-PD-(L)1 monotherapy are eligible)

  • b. GEA who are planned for or have received no more than one cycle of systemic treatment with cytotoxic, platinum-based chemotherapy

  • c. metastatic urothelial cancer who are planned for or have received no more than one cycle of systemic treatment with cytotoxic, platinum-based chemotherapy

  • d. CRC-MSS who are receiving first-line systemic therapy or who are planned for or have received no more than one cycle of second-line systemic therapy including a fluoropyrimidine and oxaliplatin or irinotecan

• 3. Availability of formalin-fixed, paraffin-embedded (FFPE) tumor specimen for neoantigen assessment.

• 4. Measurable disease according to RECIST v1.1.

• 5. Lesions amenable to biopsy.

• 6. ≥12 years of age (patients 12–17 years of age must weigh >40 kg).

• 7. Life expectancy of > 6 months per the investigator.

• 8. ECOG performance status of 0 or 1 (or equivalent for patients 12–17 years of age based on Karnofsky or Lansky performance scale).

• 9. Patient has adequate organ function using values obtained within 4 weeks before starting routine therapy as defined below. The sponsor may consider patients eligible despite having a laboratory value that does not meet inclusion criteria provided the Investigator reports the patient is otherwise in good health and the history and kinetics of the change in the laboratory value do not raise a substantial safety concern over including the patient in the study or compromise the integrity of the study data. The sponsor and investigator must agree to include any patients with a laboratory value below the defined criteria below:

  • a. Peripheral white blood cell (WBC) ≥3,000/mm³

  • b. Absolute lymphocyte count ≥800/mm³

  • c. Absolute neutrophils count (ANC) ≥ 1,500/mm³

  • d. Platelets ≥100,000/mm³

  • e. Hemoglobin ≥ 9 g dl⁻¹

  • f. Albumin ≥ 3 g dl⁻¹

  • g. Calculated creatinine clearance >50 ml min⁻¹ using Cockcroft–Gault equation

  • h. Alanine aminotransferase (ALT) and aspartate aminotransferase (AST) ≤3× upper limit of normal (ULN)

  • i. Total bilirubin ≤1.5× ULN or direct bilirubin ≤1× ULN (patients with Gilbert’s disease may be included if their total bilirubin is ≤3.0 mg dl⁻¹)

  • j. International normalized ratio or prothrombin time (PT) or partial thromboplastin time (PTT) ≤1.5× ULN

• 10. For NSCLC only, a history of smoking, defined as more than five packs, or 100 lifetime cigarettes.

• 11. For women of childbearing potential, willing to undergo pregnancy testing and agreeing to use at least one highly effective contraceptive method during the study treatment period and for 5 months after last investigational study treatment.

Inclusion criteria for study treatment stage

Patients must sign a second ICF and meet the following inclusion criteria to be eligible for the study treatment stage:

• 12. If patient did not participate in the vaccine production part of this study, patient has been enrolled in the screening protocol GO-003.

• 13. Provide a signed and dated ICF document before initiation of any treatment-related study-specific procedures.

• 14. Patient has received routine therapy as follows:

  • i. Patient with NSCLC, metastatic urothelial cancer or GEA who has received systemic treatment with cytotoxic, platinum-based chemotherapy, has measurable disease according to RECIST v1.1 and:

    • a. has chemotherapy and has an assessment of SD, partial response or complete response: will either receive study treatment in addition to continued routine therapy, or will receive study treatment after stopping routine therapy, per Investigator discretion; or

    • b. has experienced disease progression: will receive study treatment as next line of therapy (i.e., without chemotherapy); or

    • c. is intolerant to chemotherapy: will receive study treatment as next line of therapy (i.e. without chemotherapy)

  • ii. Patient with CRC-MSS who has received second-line systemic therapy including a fluoropyrimidine and oxaliplatin or irinotecan, has measurable disease according to RECIST v1.1 and:

    • a. has assessment of SD, partial response or complete response: will receive study treatment in addition to continued routine therapy per investigator discretion; or

    • b. has experienced disease progression: will receive study treatment as next line of therapy; or

    • c. is intolerant to chemotherapy: will receive study treatment as next line of therapy

• 15. Patient agrees to undergo research biopsies before study treatment and approximately 16 weeks after first vaccine administration (if patient has lesions amenable to biopsy).

• 16. ECOG performance status of 0 to 2 (or equivalent for patients 12–17 years of age based on Karnofsky or Lansky performance scale).

• 17. Patient has adequate organ function as defined below. The sponsor may consider patients eligible despite having a laboratory value that does not meet inclusion criteria provided the investigator reports the patient is otherwise in good health and the history and kinetics of the change in the laboratory value do not raise a substantial safety concern over including the patient in the study or compromise the integrity of the study data. The sponsor and investigator must agree to include any patients with a laboratory value below the defined criteria below.

  • a. Peripheral WBC ≥2,000/mm³

  • b. Absolute lymphocyte count ≥500/ mm³

  • c. ANC ≥1,000/mm³

  • d. Platelets ≥75,000/mm³

  • e. Hemoglobin ≥ 9 g dl⁻¹

  • f. Calculated creatinine clearance >40 ml min⁻¹ using Cockcroft–Gault equation

  • g. ALT and AST ≤3× ULN unless liver metastases are present in which case patients are included if ALT and AST ≤5 × ULN

  • h. Total bilirubin ≤1.5× ULN or direct bilirubin ≤1× ULN (patients with Gilbert’s disease may be included if their total bilirubin is ≤ 3.0 mg dl⁻¹).

  • i. International normalized ratio and PT and PTT ≤1.5× ULN, unless patient is receiving anti-coagulant therapy, in which case patients are eligible if PT and PTT is within therapeutic range of intended use of anti-coagulants.

Exclusion criteria

Exclusion criteria for the vaccine production stage

Patients meeting any of the following criteria are not eligible for enrollment in the vaccine production stage:

• 1. Tumors with genetic characteristics as follows:

  • a. For NSCLC, patients with a known genetic driver alteration in EGFR, ALK, ROS1, RET or TRK

  • b. For CRC and GEA, patients with known microsatellite instability-high disease based on institutional standard

  • c. For CRC, patients with a known BRAF V600E mutation or patients with peritoneal carcinomatosis, and for GEA, patients with peritoneal carcinomatosis as their only evidence of disease

• 2. Patient has a known tumor mutation burden more than one nonsynonymous mutations/megabase.

• 3. Known exposure to ChAd or any history of anaphylaxis in reaction to a vaccination or allergy or hypersensitivity to study drug components.

• 4. Bleeding disorder (e.g., factor deficiency or coagulopathy) or history of substantial bruising or bleeding following i.m. injections or blood draws.

• 5. Patient has received prior therapy consisting of anti-CTLA-4, anti-PD-1, anti-PD-L1 or any other antibody or drug specifically targeting T cell co-stimulation or checkpoint pathways.

  Patients with NSCLC and GEA who have been, or are currently being, treated with anti-PD-(L)1 monotherapy are eligible.

• 6. Immunosuppression that is expected to be present at the time of study treatment after vaccine production and coming from:

  • a. Concurrent, recent (≤4 weeks) or anticipated treatment with systemic corticosteroids (>10 mg daily prednisone equivalent) or other immunosuppressive medications such as OKT3, ATG/ALG, methotrexate, tacrolimus, cyclosporine, azathioprine or rapamycin. Inhaled or topical steroids, physiologic corticosteroid replacement therapy for adrenal or pituitary insufficiency, steroid pretreatment for chemotherapy (e.g., pemetrexed), antihistamines, nonsteroidal anti-inflammatory drugs and aspirin are permitted in the absence of active autoimmune disease

  • b. Conditions such as common variable hypogammaglobulinemia or radiation exposure such as large field radiotherapy

• 7. History of allogeneic tissue/solid organ transplant.

• 8. Patients who have had a history of life-threatening TRAEs with prior immunotherapy or have not recovered from prior cancer therapy-induced AEs (i.e., any AE that remains ≥ grade 2 or that has not returned to baseline with the exception of peripheral neuropathy, alopecia and hypothyroidism that is controlled with hormone replacement therapy).

• 9. Active, known or suspected autoimmune disease.

  • a. Patients with type I diabetes mellitus, hypothyroidism, adrenal or pituitary insufficiency that is controlled with hormone replacement therapy, skin disorders (such as vitiligo, psoriasis or alopecia) not requiring systemic treatment, or conditions not expected to recur in the absence of an external trigger are permitted. Replacement therapy (e.g., thyroxine, insulin or physiologic corticosteroid replacement therapy for adrenal or pituitary insufficiency) is not considered a form of systemic treatment.

• 10. History of other cancer within 2 years with the exception of basal or squamous cell carcinoma of the skin, superficial bladder cancer or carcinoma in situ of the breast, cervix, prostate or other neoplasm that has undergone potentially curative therapy with no evidence of disease recurrence following agreement with sponsor.

• 11. Any severe concurrent non-cancer disease (including active systemic infection and/or uncontrolled hypertension) that, in the judgment of the investigator, would make the patient inappropriate for the current study.

• 12. Active tuberculosis or recent (<2 week) clinically substantial infection, or evidence of active hepatitis B or hepatitis C.

• 13. Known history of positive test for HIV or known acquired immunodeficiency syndrome.

• 14. History of pneumonitis requiring systemic steroids for treatment (with the exception of prior resolved in-field radiation pneumonitis).

• 15. Patient with known central nervous system (CNS) metastases and/or carcinomatous meningitis.

• 16. Myocardial infarction within 6 months of study initiation, active cardiac ischemia, myocarditis or New York Heart Association grade III or IV heart failure.

• 17. Pregnant, planning to become pregnant or nursing.

• 18. Medical, psychiatric, cognitive or other conditions that compromise the patient’s ability to understand the patient information, give informed consent, comply with the study protocol or complete the study.

Exclusion criteria for the study treatment stage

Patients meeting any of the of the following exclusion criteria will not be eligible for the study treatment stage:

• 19. Immunosuppression from:

  • concurrent, recent (≤ 4 weeks) or anticipated treatment with systemic corticosteroids (>10 mg daily prednisone equivalent) or other immunosuppressive medications such as OKT3, ATG/ALG, methotrexate, tacrolimus, cyclosporine, azathioprine or rapamycin. Inhaled or topical steroids, physiologic corticosteroid replacement therapy for adrenal or pituitary insufficiency, steroid pretreatment for chemotherapy (e.g., pemetrexed), antihistamines, nonsteroidal anti-inflammatory drugs and aspirin are permitted in the absence of active autoimmune disease

  • conditions such as common variable hypogammaglobulinemia or exposures such as large field radiotherapy

• 20. Patients who have had history of life-threatening TRAEs with prior immunotherapy or who have not recovered from prior cancer therapy-induced AEs (i.e., any AE that remains ≥ grade 3, or that has not returned to baseline with the exception of alopecia, neuropathy and hypothyroidism if controlled by hormone replacement).

  Only applicable to those AEs not covered by other inclusion and exclusion criteria (e.g., ANC or ALT values).

• 21. Active, known or suspected autoimmune disease.

  Patients with type I diabetes mellitus, hypothyroidism if requiring hormone replacement, skin disorders (such as vitiligo, psoriasis or alopecia) not requiring systemic treatment, or conditions not expected to recur in the absence of an external trigger are permitted. Replacement therapy (e.g., thyroxine, insulin or physiologic corticosteroid replacement therapy for adrenal or pituitary insufficiency) is not considered a form of systemic treatment.

• 22. Received radiation therapy within 2 weeks before first dose of study treatment.

• 23. Any severe concurrent non-cancer disease (including active infection and/or uncontrolled hypertension) that, in the judgment of the investigator, would make the patient inappropriate for the current study (study treatment may be delayed until patient has recovered from disease (such as an infection) or has stabilized).

• 24. Active tuberculosis or recent (<2 week) clinically substantial infection, or evidence of active hepatitis B or hepatitis C.

• 25. History of pneumonitis requiring systemic steroids for treatment (with the exception of prior resolved in-field radiation pneumonitis).

• 26. Patient with uncontrolled CNS metastases or brain stem metastasis, patient receiving steroids for CNS disease and/or carcinomatous meningitis; patients with irradiated CNS metastases should be discussed on a case-by-case basis with the sponsor before enrollment.

• 27. Myocardial infarction, active cardiac ischemia, myocarditis or New York Heart Association grade III or IV heart failure since inclusion in the study.

• 28. Pregnant, planning to become pregnant or nursing.

• 29. Treatment with botanical preparations (e.g. herbal supplements or traditional Chinese medicines) intended for general health support or to treat the disease under study within 2 weeks before treatment that may interfere with study treatment per investigator discretion.

• 30. Medical, psychiatric, cognitive or other conditions that compromise the patient’s ability to understand the patient information, give informed consent, comply with the study protocol or complete the study.

Individual patient-specific vaccine cassettes

Patient tissue next-generation sequencing, analysis and neoantigen prediction using the EDGE model were performed as previously described³⁰. Twenty neoantigens were selected for each patient, as summarized in Supplementary Table 2. The 20 neoantigen epitopes were designed with flanking sequences so that each are 25 amino acids in length and arranged as a cassette to minimize generation of new junction epitopes. A class II helper domain consisting of universal class II epitopes PADRE and tetanus toxoid were included at the C terminus of the cassette. The cassettes were codon optimized for maximum protein expression in humans and for ease of synthesis. The cassettes were synthesized from overlapping oligonucleotides (Integrated DNA Technologies (IDT)), that cover both DNA strands of the cassette sequence plus sequences in the plasmid backbones and assembled into either ChAd or samRNA plasmid backbones by Gibson assembly (Codex DNA). The assembled plasmids were then transformed into bacteria and screened by Sanger sequencing for correct clones which served as the templates for good manufacturing practice ChAd68 production or for samRNA transcription. ChAd68 plasmids were linearized with PacI and then transfected with TransIt Lenti (Mirus Bio) into 293F (Thermo Fisher Scientific) cells to initiate virus production. The virus was amplified through successive rounds of infection before the final production run at scale. The virus was purified at 48–72 h after infection and purified by anion exchange chromatography (Satorius) and formulated in 5 mM Tris, pH 7.9, 1 mM MgCl₂ and 75 mM NaCl. VP titers were determined by absorbance at 260 nm after SDS disruption and infectious unit titers were determined by immunostaining with an anti-Ad antibody (Abcam). samRNA plasmid was used as a template for RNA transcription using T7 polymerase, and the RNA was capped using a vaccine-capping enzyme and mRNA CAP 2′-O-methyl transferase (NEB). The RNA concentration was measured by Ribogreen quantitation and then formulated into lipid nanoparticles (Genevant).

Patient PBMC isolation and storage

Patient demographics and dosing schedules are shown in Table 1 and Extended Data Table 1. Due to COVID-19 restrictions, several blood draws were missed. Ad hoc dosing and blood draw timepoints were added with IRB approval for several patients. Whole blood for immunogenicity testing was collected before administration of the first dose (ChAd68 prime) to assess baseline levels and at 14 and 28 days (2 and 4 weeks) after ChAd68 prime. Subsequent planned sample collection post samRNA boosts include days 7, 14 and 28 after dose 2 (samRNA boost 1) and at days 7 and/or 28 after each additional samRNA boost. Two patients (G2 and G8) consented to an optional leukapheresis procedure (week 19 for G2, week 25 for G8). PBMCs from whole blood or leukopak were isolated at local PBMC processing sites according to standardized protocols. Briefly, cells were isolated using density gradient centrifugation on Ficoll Paque Plus (GE Healthcare), washed with D‐PBS (Corning), counted and cryopreserved in CryoStor CS10 (STEMCELL Technologies) at 5 × 10⁶ cells ml⁻¹ for whole blood and 2 ×10⁷ cells ml⁻¹ for leukopak. Cryopreserved cells were stored in liquid N₂, shipped in cryoports and transferred to storage in liquid N₂ upon arrival. Cryopreserved cells were thawed and washed twice in OpTmizer T Cell Expansion Basal Medium (Gibco) with Benzonase (EMD Millipore) and once without Benzonase. Cell counts and viability were assessed using the Guava ViaCount reagents and module on the Guava EasyCyte HT cytometer (EMD Millipore). Cells were rested overnight before use in functional assays.

Peptides (clinical study)

Custom-made, recombinant, lyophilized peptides specific for each patient were produced by GenScript and reconstituted at 5 mg ml⁻¹ per peptide in sterile DMSO (VWR International), aliquoted and stored at −80˚C (Supplementary Table 2). Control peptides to assess responses to infectious disease antigens from cytomegalovirus, Epstein–Barr virus and influenza (CEF peptide pool) were purchased from JPT Peptide Technologies.

IVS cultures

Neoantigen-reactive T cells from patient samples were expanded in the presence of cognate peptides and low-dose IL-2 as described previously³⁰. Briefly, thawed PBMCs were rested overnight and stimulated in the presence of minimal epitope peptide pools (10 µg ml⁻¹ per peptide, 40 patient-specific peptides per pool) or control peptides (CEF) in ImmunoCult-XF T Cell Expansion Medium (IC media; STEMCELL Technologies) with 10 IU ml⁻¹ recombinant human IL-2 (R&D Systems) for 14 days in 48- or 24-well tissue culture plates. Cells were seeded at 1-2 ×10⁶ cells/well and fed every 2–3 days by replacing two thirds of the culture media with recombinant human IL-2.

IFNγ ELISpot assay (clinical study)

Detection of IFN-γ-producing T cells was performed by ex vivo ELISpot assay⁴⁷. Briefly, cells were harvested, counted and resuspended in media at 4 × 10⁶ cells ml⁻¹ (ex vivo PBMCs) or 2 × 10⁶ cells ml⁻¹ (IVS-expanded cells) and cultured in the presence of DMSO (VWR International), phytohemagglutinin-L (Sigma-Aldrich), CEF peptide pool or cognate peptides in ELISpot Multiscreen plates (EMD Millipore) coated with anti-human IFN-γ capture antibody (Mabtech). Following 18–24 h incubation in a 5% CO₂, 37˚C, humidified incubator, supernatants were collected, cells were removed from the plate and membrane-bound IFNγ was detected using anti-human IFN-γ detection antibody (Mabtech), Vectastain Avidin peroxidase complex (Vector Labs) and AEC Substrate (BD Biosciences). ELISpot plates were allowed to dry, stored protected from light and sent to Zellnet Consulting for standardized evaluation⁴⁵. Data are presented as SFU per million cells.

ICS (clinical study)

PBMCs or IVS-expanded PBMCs were stimulated with either cognate minimal (8–11mer) or long overlapping (15mer) peptides, DMSO (vehicle control; VWR) or PMA/ionomycin cell stimulation cocktail (positive control; Affymetrix) in the presence of anti-human CD28/CD49d antibody (BD Biosciences), anti-human CD107α-BV421 (BioLegend), BD GolgiStop (BD Biosciences) and Brefeldin A (BioLegend) over a period of 18 h. Following overnight incubation, cells were stained with live/dead ZombieRed (BioLegend) and surface marker antibodies (CD8-PerCP-Cy5.5, CD3-BV605, CCR7-PE-Cy7 from BioLegend; CD4-APC-eF780, CD45RA-SuperBright702 from eBioscience; PD-1-BUV737 from BD Biosciences) before fixation and permeabilization with FIX & PERM Cell Permeabilization Kit (Thermo Fisher Scientific). Following permeabilization, cells were stained for intracellular cytokines using anti-human IFN-γ-APC, TNF-α-BV785, and IL-2-PE antibodies (BioLegend) before data acquisition on a BD LSRFortessa flow cytometer (BD Biosciences).

HLA class I fluorophore-conjugated tetramer generation and staining (clinical study)

HLA class I tetramers were generated from Flex-T biotinylated HLA monomers (BioLegend) or MBL International with target peptides loaded onto the monomers by UV exchange according to the manufacturer’s instructions. Successful loading of target peptides by UV exchange was assessed by ELISA (BioLegend, according to manufacturer’s instructions). Alternatively, biotinylated monomers were generated in-house and prefolded with target peptides. pHLA monomers were assembled into tetramers using fluorophore-conjugated Streptavidin (PE from BioLegend; APC from eBioscience). Using generated fluorophore-labeled peptide-MHC tetramers, PBMCs were tetramer stained to help identify antigen-specific T cells. Patient PBMC samples were thawed and resuspended in a 96-well V-bottom plate with FACS buffer (PBS, 10% FBS [v/v]). Before staining, samples were incubated with 50 nM Dasatinib (Sigma-Aldrich) in a 5% CO₂, 37˚C incubator for 30 min. Subsequently, HLA class I fluorophore-labeled tetramers loaded with target peptides of interest were added directly to each corresponding PBMC sample. In instances where multiple tetramers needed to be added to the same sample, a cocktail was made with equal amounts of each tetramer, then mixed and added at the same time. Samples were incubated at room temperature for 1 h in the dark. Following incubation, samples were washed in FACS buffer and stained with an antibody master mix containing purified anti-human HLA-A,B,C antibody, anti-human CD8-PerCP-Cy5.5, anti-human CCR7-PE-Cy7, anti-human CD3-BV421 (all BioLegend), anti-human CD45RA-SB702, anti-human CD4-APC-eF780 (both eBioscience) and Live/Dead Green (Thermo Fisher Scientific). Sample were mixed and incubated at 4˚C in the dark protected from light for 20 min. Samples were washed once with FACS buffer and then resuspended in FACS buffer prior to acquisition on a BD LSRFortessa flow cytometer (BD Biosciences).

Flow cytometry analyses (clinical study)

Samples acquired on flow cytometers were analyzed using FlowJo software (FlowJo). Gating strategies for human samples are as follows (Supplementary Figs 2 and 3). For CD4 versus CD8 T cell assessments: lymphocytes (SSC-A versus FSC-A), single cells (FSC-H versus FSC-A), viable cells (FSC-A versus LD-ZombieRed), CD3⁺ cells (FSC-A versus CD3-BV605), CD4⁺ and CD8⁺ (CD4-APCeF780 versus CD8-PerCP-Cy5.5). For tetramer studies: lymphocytes (SSC-A versus FSC-A), single cells (FSC-H versus FSC-A), viable cells (FSC-A versus LD-AF488), CD3⁺ cells (FSC-A versus CD3-BV421), CD8⁺Tet⁺ (CD8-PerCP-Cy5.5 versus tetramer-PE or tetramer-APC), memory phenotypes (CD45RA-SB702 versus CCR7-PE-Cy7). For ICS studies: lymphocytes (SSC-A versus FSC-A), single cells (FSC-H versus FSC-A), viable cells (FSC-A versus LD-ZombieRed), CD3⁺ cells (FSC-A versus CD3-BV605), CD4⁺ and CD8⁺ (CD4-APCeF780 versus CD8-PerCP-Cy5.5), CD8⁺ or CD4⁺ cytokine⁺ cells (CD8-PerCP-Cy5.5 versus CD107a-BV421, IFN-γ-APC, TNF-ɑ-BV786, IL-2-PE; CD4-APCeF780 versus CD107a-BV421, IFN-γ-APC, TNF-ɑ-BV786, IL-2-PE). Polyfunctionality analyses were performed using Boolean gating (FlowJo) and graphed using SPICE software v6 (ref. ⁴⁸). Results are represented as the percentage of positive cell populations (frequency of parent). Data are shown as background subtracted where indicated.

MSD U-plex assay

Detection of secreted IL-2, TNF-α and granzyme B (GRZB) in ex vivo ELISpot supernatants was performed using a U-plex assay MSD U-PLEX Biomarker assay. Assays were performed according to the manufacturer’s instructions. Data were acquired on the MESO SECTOR S 600 instrument and DISCOVERY WORKBENCH assay analysis software. Analyte concentrations (pg ml⁻¹) were calculated using serial dilutions of known standards for each cytokine. For graphical data representation, values below the minimum range of the standard curve were represented as zero.

Triplex FluoroSpot assay

A multiplexed FluoroSpot assay was performed according to manufacturer’s instructions (MABtech) on a subset of patient samples for the detection of IFN-γ, IL-2 and granzyme B-producing T cells. Briefly, patient PBMCs were thawed and rested overnight at 2 × 10⁶ cells ml⁻¹. The following day, cells were harvested, counted and resuspended at 4 × 10⁶ cells ml⁻¹ before plating. Precoated FluoroSpot plates were washed with PBS and blocked with culture media before patient-specific peptide pools and cells were added (2 × 10⁵ PBMCs per well). After incubating in a 5% CO₂, 37˚C humidified incubator for 18–24 h, secreted IFN-γ, IL-2 and granzyme B were detected using anti-human IFN-γ-BAM/anti-BAM-490, biotinylated anti-human granzyme B/Streptavidin-550 and anti-human-IL-2-WASP/anti-WASP-650 antibody pairs followed by fluorescence enhancer. Plates were imaged and enumerated on an AID iSpot reader (Autoimmun Diagnostika). Data are presented as SFU per million cells.

Single-cell TCR sequencing and digital gene expression

Single-cell library preparation and sequencing was completed according to 10x Genomics v1 5’ Chromium Single Cell V(D)J protocol CG000186 Rev A. The only deviation in library preparation was in V(D)J library step 5.4 with five cycles of PCR. All sequencing was completed on the Illumina NovaSeq at a loading concentration of 450 pM with 5% PhiX. The 10x Genomics Cell Ranger version 3.1 was used for V(D)J and Digital Gene Expression library analysis and was current at time of research.

Adaptive TCR sequencing of RNA later-preserved tumor samples

Tumor DNA was extracted via Qiagen AllPrep DNA/RNA kit and quantified with both Qubit high sensitivity dsDNA assay (Thermo Fisher Scientific) and Agilent Genomic DNA tape on the 4200 TapeStation system. Between 300 and 1,000 ng extracted DNA was sent to Adaptive Biotechnologies for T cell receptor beta sequencing, and results were returned to us via the immunoSEQ Analyzer 3.0.

TCR sequencing analysis

Significant tumor TCR expansion was determined via a binomial test with Benjamini–Hochberg correction and P value <0.05, analyzing TCRs with five cells or more in either timepoint. IVS TCR expansion was analyzed by filtering 10× V(D)J results for each timepoint for TCRs (1) with a full alpha/beta pair, (2) found above an abundance threshold and (3) not found in the CEF IVS. The proportion of the resulting TCRs was then tracked from baseline to on treatment, and the polyfunctional activation of these expanded TCR populations was also analyzed by searching in the transcriptome for non-zero expression of IFN-γ, granzyme B, TNF and IL-2. The expanded TCRs were cross-referenced with those expanded in the tumor biopsies, and overlap was statistically significant by Fisher test (P < 0.001).

cfDNA collection and isolation

Whole blood was collected in two 10-ml cell-free DNA (cfDNA) BCT (Streck) starting at the time of the prime, and subsequent draws were collected at dosing visits. Due to COVID-19 restrictions, several blood draws were missed. Whole blood underwent a double-spin protocol to first separate plasma from WBCs and red blood cells before a second spin to remove any remaining cellular debris. The separated plasma was frozen and shipped to Gritstone Bio and stored at −80°C until extraction. cfDNA was extracted from the entire plasma volume of a single draw using the Apostle MiniMax cfDNA Isolation kit (ApostleBio) and quantified using the Qubit 1x dsDNA High Sensitivity Assay (Thermo Fisher Scientific).

ctDNA sequencing and analysis

Custom hybrid capture panels were designed for each patient enrolled in the study. Each patient panel was designed to capture the EDGE-predicted neoantigens and all predicted coding (nonsynonymous) mutations present in the original screening biopsy specimen. Target mutations for each patient were curated and sent for customized design to IDT as xGen Lockdown Probes. Each capture pool contained up to nine patients per pool. Patient-matched genomic DNA was fragmented before library preparation using the NEB FS module (NEB). Shotgun libraries for cfDNA (up to 30 ng) and the fragmented, patient-matched genomic DNA (20–30 ng) were prepared using the KAPA HyperPrep (KAPA Biosystems) kit using a customized pool of duplex adapters containing unique molecular identifiers for duplex sequencing (IDT). Shotgun libraries were captured with patient-specific probes overnight using the IDT xGen Hybridization and Wash kit. Enriched libraries were sequenced on an Illumina NovaSeq to a minimum mean raw depth of 80,000×. Briefly, unique molecular identifiers were clipped from the raw sequencing reads before alignment to hg38 using BWA-MEM. Using fgbio, aligned reads were grouped by position and duplex identity. Consensus reads were created using a duplex of 3× (three supporting reads from each strand) and realigned to hg38. Variant calling was performed using FreeBayes and filtered for the patient-specific variants. Variant allele frequency was converted to mutated haploid genomic equivalents per milliliter plasma (hGE ml⁻¹) using the extracted plasma volume and total cfDNA yield from the extraction. Percent change in ctDNA was calculated as the change of the median mutated hGE ml⁻¹ from the baseline sample.

Patient tumor biopsy whole-exome and transcriptome analysis and TCGA data comparison

Pretreatment GRANITE patient FFPE biopsy-derived DNA and RNA and matched whole blood-derived DNA were whole-exome sequenced (IDT xGen V1) and analyzed as described previously³⁸. Relevant tumor types for genomic comparison from TCGA Pan-Cancer Atlas dataset included gastric (n = 405 stomach adenocarcinoma), colon adenocarcinoma (n = 332, MSS-CRC (281) and MSI-CRC (51)), lung adenocarcinoma (n = 498), lung squamous cell carcinoma (n = 463, LUSC) and melanoma (n = 436, skin cutaneous melanoma). For RNA expression analyses, all GRANITE and TCGA sample RNA-seq by Expectation-Maximization expected counts were normalized utilizing DESeq2 (ref. ⁴⁹). Normalized counts were used to determine patient PD-L1 RNA expression and the IFN-γ RNA expression signatures, which were calculated as the average of 28 IFN-γ-associated gene specific z-scores across all GRANITE and TCGA samples (n = 2,494)⁵⁰. Tumor mutational burden was calculated for each GRANITE patient as the total of all somatic variants called >4% variant allele frequency divided by the exome bait panel size (39 Mb). For TCGA samples, total mutations called were divided by exome bait panel size used (38 Mb)⁵¹ (Supplementary Table 4).

Transduction, culture and live sorting of single HLA-expressing A375 NucLight Red cell lines

A375 cells were obtained from ATCC (CRL-1619). Endogenous class I HLAs were knocked out using CRISPR-Cas9 ribonucleoprotein complexes. HiFi S.p. Cas9 Nuclease V3 (1081060), Cas9 Electroporation Enhancer (1075915) and custom guide RNAs were obtained from IDT. Ribonucleoproteins were electroporated into A375 cells using the Neon transfection system from Thermo Fisher Scientific with a voltage of 1,400 and two pulses for 20 ms. CRISPR-treated A375 cells were stained using a 1:100 anti-human HLA-A,B,C antibody from Biolegend (311405), and HLA-null cells were sorted using the Cell Sorter SH800S (SONY Biotechnology) (Supplementary Fig. 4). Single-allele A375 cell lines were generated with HLA sequences obtained from the coding sequence of the human genome (HLA-B*08:01:01:01, HLA-B*57:01:01:01, HLA-A*30:04:01:01, HLA-A*03:01:01:01, HLA-C*04:01:01:01). HLAs were cloned into the pCDH-EF1α-MCS1 lentiviral expression backbone vector from System Biosciences (CD502A-1). Lentivirus was generated for each HLA and titer determined using the Lenti-X qRT-PCR titration kit from Takara Bio (631235). Single HLAs were transduced into A375 HLA-null cells at an m.o.i. of 100 with 8 ug ml⁻¹ polybrene (EMD Millipore, TR-1003). Cells were spinoculated at 800g for 45 min at room temperature at a density of 8 × 10⁵ cells per ml. Fresh media was added and cells were incubated overnight. HLA-positive cells were sorted after at least 48 h using the method described above. Fluorescent single-allele A375 cells were generated with NucLight Red (IncuCyte, 4476) at an m.o.i. of 3 with 8 μg ml⁻¹ polybrene (EMD Millipore, TR-1003). Cells were spinoculated at 800g for 45 min at room temperature at a density of 8e5 cells ml⁻¹. Fresh media was added and cells were incubated overnight. Red fluorescent and HLA-positive cells were sorted after at least 48 h using the method described above.

IncuCyte killing assays

Single HLA-expressing A375 NucLight Red cell lines were seeded in 96- or 48-well plates at concentrations of 2.5 × 10⁴ or 3.5 ×10⁴ cells per well in DMEM with 10% heat-inactivated FBS. The plates were placed in the IncuCyte S3 (Essen Biosciences) and 24 h after the seeding, the effector cells were plated at a concentration of 2.5 × 10⁵ or 3.5 × 10⁵ cells per well in a 96- or 48-well plate, respectively, for an effector to target ratio of 10:1. Individual peptides (GenScript; Supplementary Table 2) were added to the treated wells for a final concentration of 10 µg ml⁻¹, and DMSO was used for the control wells. The plates were imaged with the IncuCyte for up to 72 h, after which the data were analyzed using the IncuCyte S3 2018 analysis software. Viability of A375 was assessed by red cell count, and relative viability was calculated relative to DMSO co-culture control wells.

Statistics and reproducibility

For the clinical study, the sample size is not driven by a statistical power due to the nature of an adaptive dose-escalation method. Therefore, the sample size of the phase 1 portion of the study ranges from 11 to maximum of 24 patients, which is dependent on occurrence of DLTs. No data were excluded from the analyses, and experiments were not randomized. The investigators were not blinded to allocation during experiments and outcome assessment. For the NHP study, the number of animals per group is n = 3/sex, which is the minimum number of animals needed to do simple statistics. The total number of animals used in this study is considered the minimum required to properly characterize the effect of the test articles (i.e., vaccine and anti-CTLA-4) and compare their route of administration (i.e., s.c. versus i.v.) and has been designed such that it does not require an unnecessary number of animals to accomplish its objectives. For both the NHP and clinical study, descriptive statistics are described throughout. Where applicable, statistical analyses applied for specific methods are outlined in the relevant methods sections.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.