A Genetic Vaccine Encoding Shared Cancer Neoantigens to Treat Tumors with Microsatellite Instability

Cancer vaccination had limited success in the past likely because of (i) the use of tumor-associated antigens as targets; (ii) the presence of an immunosuppressive tumor microenvironment; and (iii) the use of suboptimal vaccination platforms for induction of T-cell immunity (1).

Today, vaccination is thought to be a viable therapeutic option in oncology if potent antigens arising from tumor-specific mutations can be selected and combined with drugs reversing tumor-mediated immune suppression (e.g., checkpoint inhibitors; ref. 2).

The vaccine delivery system plays a crucial role in the quality and strength of induced immune responses. A heterologous prime/boost vaccination platform based on Great Ape Adenoviruses (GAd) and Modified Vaccinia Ankara (MVA) vectors has already been tested in over 4,000 volunteers in several clinical trials with different antigen payloads (relevant to Ebola, Malaria, HCV, HIV, and RSV) and shown to generate potent, durable, and high-quality T-cell responses (3–8).

Microsatellite instability (MSI) tumors, characterized by a defective DNA mismatch repair system, have the highest mutational burden and, accordingly, show a good clinical response to checkpoint inhibitors (9, 10). In MSI tumors, the presence of indel mutations in microsatellites regions of coding genes results in the synthesis of shared frameshift peptides (FSP) that are expected to be very potent and safe neoantigens (11–15).

Therefore, MSI offers a unique opportunity for the development of an “off-the shelf” neoantigen-based cancer vaccine that can be combined with PD-1 checkpoint inhibition in patients with metastasis and may lead to improved clinical outcomes. Furthermore, given the extremely high risk of developing tumors for Lynch syndrome (LS) mutation carriers, a vaccine inducing a broad immune response could be used to prevent cancer occurrence/recurrence in this population (16, 17).

To design the Nous-209 vaccine, we analyzed a dataset from 320 patients annotated as MSI in The Cancer Genome Atlas (TCGA). For the validation, 20 additional samples were examined. Next generation sequencing (NGS) data (DNA exome and RNA) from 14 colorectal cancer MSI samples from the European Genome-Phenome archive (ID: EGAS00001000288) were downloaded with the permission of Genentech Inc. Four tumor Flash-Frozen biopsies of MSI colorectal cancer and matched healthy mucosa including 1 diagnosed with Lynch syndrome were obtained from iSpecimen Inc. Finally, 2 additional MSI colorectal cancer samples with LS were obtained from “Regina Elena” National Cancer Institute in Rome. DNA and RNA were extracted from each collected sample and sequenced at CeGaT GmbH. For all subjects, appropriate Institutional Review Board approval and written-informed consent were obtained according to the declaration of Helsinki.

Chromosomic coordinates of mononucleotide repeats (MNR) present in the human genome (Gchr37) with length ≥6 bp were determined by MISA software (18). Note that 7,524,435 MNRs were mapped on the human REFSEQ transcriptome (V73) to derive a subset of 30,678 MNRs located in the sequence of protein-coding transcripts. Protected Mutation Annotation Format files available from TCGA (release date 4.0—October 31, 2016) were analyzed for the presence of frameshift mutations (FSM) overlapping with the 30,678 MNRs. FSMs derived from a 1 nucleotide deletion and with a number of reads harboring the mutation significantly higher in the tumor sample compared with the matched normal sample (FDR-corrected Fisher test P value ≤ 0.05) were accepted. The resulting list of FSMs was then further refined by retaining only FSMs that fulfill the following criteria:

• Presence in at least 5% of tumor samples in one of the three analyzed tumor types;

• Allele frequency of the FSM in the matched normal samples ≤ 25%;

• Presence of the FSM in < 2% of alleles of samples from a collection of healthy tissues in the ExAC database (19);

• Expression (RSEM log₂ expression value downloaded from TCGA) of the FSM-carrying gene in all the three tumor types higher than the lower 20th percentile value of all expressed protein-coding genes determined considering all three tumor types.

Exomeseq and RNA sequencing (RNA-seq) reads were aligned on the Gchr37 human genome using HISAT2 2.0.4 (20). Multimapping reads were filtered out using Samtools 0.1.19 (21). Optical duplicates were marked using Picard's MarkDuplicates tool v1.14 (https://broadinstitute.github.io/picard/). DNA alignments were further optimized at regions around indels, and base scores were recalibrated after the optimization step using GATK software v3.4.46 (22). A “look-up” analysis was performed by identifying MNRs mutated in the exome DNA with a minimum threshold of 3 mutated reads and a minimum mutation allele frequency of 10%. The number of reads at each FSM locus was estimated by using the mpileup utility of samtools 0.1.19 (21). Somatic variant calling of small indels was performed utilizing mutect2 (23), Varscan2 (24), and SCALPEL (25) with default parameters. All FSMs detected by at least one variant caller were considered.

Gene expression was computed by counting the number of reads mapped on each gene annotated in the human REFSEQ transcriptome (V73) with Rsubreads package (26) and estimating transcript per million (TPM) values. Look-up analysis was performed on RNA-seq data to estimate the FSM loci covered by mutated reads.

For each patient, the HLA class-I and class-II haplotypes were determined from the Exomeseq of the healthy tissue, using the Optitype and PHLAT software (27, 28). MHC class I binding predictions were performed using the IEDB-recommended method included in IEDB 2.17 (29). Eight to 10 mer HLA binders with a predicted IC₅₀ ≤ 500 nmol/L were considered. MHC class II binding predictions were performed using the NetMHCIIpan method included in IEDB. HLA-II–predicted binders with a percentile score ≤10% were considered.

M9 cells are derivative of Hek293 cells that stably express the Tet repressor gene (TetR). Adherent M9 cells were grown in DMEM supplemented with 10% FBS, 1% Pen/Strep, 2 mmol/L l-glutamine, and 0.4 mg/mL Geneticin. Suspension M9 cells were grown in CD293 medium supplemented with 6 mmol/L l-glutamine and 0.4 mg/mL Geneticin. For all experiments, M9 cells were used between passage 2 and passage 12 after thawing. Primary chicken embryo fibroblasts (CEF) cells were prepared from SPF premium plus eggs (Charles River) and grown in VP-SFM medium supplemented with 1% Pen/Strep and 4 mmol/L l-glutamine. CEF cells were frozen after preparation with no additional passages and used for all the experiments between passage 1 and passage 3 after thawing. All cell culture supernatants were routinely tested with EZ-PCR Mycoplasma Detection Kit (Biological Industries) following the manufacturer's instruction, to exclude Mycoplasma contaminations.

All cell culture reagents were purchased by Gibco, Thermo Fisher Scientific. The EBV-B cell line was established in the laboratory of Pierre van der Bruggen by coculturing in IMDM (Life Technologies, Gibco) complemented with 10% FCS (Sigma-Aldrich), the peripheral blood mononuclear cell (PBMC) of the donor with the supernatant of cell line B95-8 that contains infectious particles of EBV.

Media were supplemented with 0.24 mmol/L l-asparagine, 0.55 mmol/L l-arginine, 1.5 mmol/L l-glutamine (AAG), 100 U/mL penicillin, and 100 μg/mL streptomycin. Human recombinant IL2 (Proleukin) was purchased from Novartis, IL6 and IL12 from Peprotech, IL7 from Miltenyi Biotec, and GM-CSF (Leukine Sargramostim) from Sanofi. Human recombinant IL4 was produced at de Duve Institute. For immunological analysis, a set of peptides covering the entire neoantigenic sequence encoded by Nous-209 were obtained from JPT Peptide Technologies and used as antigens for ex vivo immunological assays. FSPs shorter than 27 amino acids (aa) are represented by a single peptide, whereas for longer FSPs, multiple 15 mers, overlapping by 11aa, were synthesized. Nine mer peptides used for in vitro stimulation of human T cells were synthesized, purified, and characterized at the de Duve Institute.

The amino acid sequence of the four artificial FSP-A genes (FSPA1-A4) was converted in nucleotide sequence based on human codon usage. At the N-terminus of each artificial gene, a sequence was added, corresponding to the aa 1–29 of the human tissue plasminogen activator (TPA) protein (NP_000921.1). At the C-terminus, an HA tag was added. The resulting transgenes were synthesized by GeneArt (Thermo Fisher Scientifics) and subcloned via EcoR1-Not1 restriction enzymes (New England Biolabs) into a shuttle plasmid containing the CMV promoter with two Tet Operator repeats and a BGH polyA. The expression cassettes were then transferred into the E1 deletion locus of pGAd20 plasmid by homologous recombination in BJ5183 cells (Agilent). pGAd20 contains the genome of a gorilla adenovirus (serotype group C) deleted in E1 and E3 regions. The four GAd vectors were produced by transfection of adherent M9 cells with Lipofectamine 2000 CD (Invitrogen, Thermo Fisher Scientific) and amplification in suspension M9 cells. Vectors were purified from infected cells by Vivapure Adenopack 20 RT (Sartorius). After purification, the four vectors were mixed to generate the GAd-209-FSP vaccine.

The four artificial FSP-B genes (FSPB1–FSPB4) were generated as described for the FSP-A genes. TPA and HA sequences (N- and C-term respectively) were also added. The resulting transgenes were synthesized by GeneArt (Thermo Fisher Scientifics) and subcloned via BamH1-Asc1 restriction enzymes (New England Biolabs) into a shuttle plasmid under the control of the P7.5 promoter. In addition, the shuttle plasmid carries an eGFP expression cassette and sequences homologous to the deletion III locus of MVA, to allow insertion of both expression cassettes (eGFP and transgene) in this locus. Recombinant MVA vectors were obtained by homologous recombination in CEF cells as previously described (30). The final recombinant vectors (devoid of any fluorescent gene) were further isolated by a limiting dilution step in CEF cells to obtain a pure viral population (30). After amplification in CEF cells, MVA vectors were purified from infected cells by centrifugation on a Sucrose cushion. After purification, the four vectors were mixed to generate the MVA-209-FSP vaccine.

Six-week-old female CB6F1 mice were purchased from Envigo. All day-to-day care was performed by trained mouse house staff at Allevamenti Plaisant SRL. All experimental procedures were approved by the Italian Ministry of Health (Authorizations 213/2016 PR) and have been done in accordance with the applicable Italian laws (D.L.vo 26/14 and following amendments) at the Institutional Animal Care and ethic Committee of Allevamenti Plaisant SRL. Viral vectors were administered i.m. in the quadriceps by delivering a total volume of 100 μL (50 μL per side).

IFNγ ELISpot assays were performed on isolated murine splenocytes as previously described (31). In details, we used 16 pools of peptides (P1–P16) encompassing the entire vaccine sequence. P1–P16 are composed of multiple sets of peptides (around 60 peptides per pool) covering the encoded FSPs. FSPs shorter than 27aa are represented by a single peptide, whereas for longer FSPs, multiple 15 mers, overlapping by 11aa, were synthesized. Peptides were diluted in DMSO and mixed to obtain 4 peptide pools per vector (pool 1 to pool 4 for vector FSP1; pool 5 to 8 for vector FSP2; pool 9 to pool 12 for vector FSP3; pool 13 to pool 16 for vector FSP4). DMSO (Sigma-Aldrich) and concanavalin A (Sigma-Aldrich) were used, respectively, as negative and positive controls. ELISpot data were expressed as IFNγ spot forming cells (SFC) per million splenocytes. ELISpot responses were considered positive if all the following conditions occurred: (i) IFNγ production present in concanavalin A-stimulated wells, (ii) the number of spots seen in positive wells was 3 times the number detected in the mock control wells (DMSO), (iii) at least 30 specific spots/million splenocytes. Intracellular IFNγ staining was performed by overnight stimulation of splenocytes with peptide pools, as previously described (31).

Peripheral blood was obtained from hemochromatosis patient LB6019 as standard buffy coat preparation, which was laid down on a 15-mL Lymphoprep layer (Axis-Shield PoCAS) in 50-mL Falcon Tubes. Isolation of CD8 T cells was performed as previously described (31). Lymphocyte-depleted PBMCs were left to adhere for 1 hour at 37°C in a culture flask at a density of 2 × 10⁶ cells per cm² in IMDM/10% FCS. Non-adherent cells were discarded, and adherent cells were cultured in the presence of IL4 (200 U/mL) and GM-CSF (70 ng/mL) in complete IMDM medium. Cultures were fed on days 2 and 4 by removing half of the medium and adding fresh medium with cytokines. On day 6, monocyte-derived immature dendritic cells (DC) were incubated for 6 hours with 1 μg/mL of peptide SLMEQIPHL (Ly4_24), IL4 (200 U/mL) and GM-CSF (70 ng/mL), in the presence of 1 μg/mL of Ribomunyl (INAVA, Pierre Fabre Medicament Production) and 500 U/mL of IFNγ to induce their maturation, irradiated (50 Gray), and washed. Recombinant HLA-A*0201 molecules were folded in vitro with β2-microglobulin and peptide Ly4_24 or a control peptide (SLEPWIPYL). Peptides were purified by gel filtration, biotinylated, and mixed with streptravidin-PE (BD Pharmingen). For staining and sorting of CD8 T cells, cells were processed as previously described (32). Positive T cells were sorted by MACS technology and cultured in round-bottomed low-adherence microwells (Costar) in the presence of 200 μL of IMDM supplemented with AAG, 10% human serum, IL2 (50 U/mL), IL6 (1,000 U/mL), IL12 (12 ng/mL), and IL7 (10 ng/mL). They were stimulated on days 0 and 7 with irradiated peptide-pulsed mature autologous DC at a ratio of 1:1. After two rounds of stimulation, approximately 10⁵ cells from each microculture were stained with multimer and screened with a FACS Fortessa (BD Biosciences) using the DIVA software (BD Biosciences) to identify multimer⁺ cells. The remaining cells in each positive microculture were collected 2 days later, stained with anti-CD8 and fluorescent multimers, and passed through a flow cytometer to isolate positive cells. These cells were restimulated every 7 to 12 days with irradiated (100 Gray) T2 cells loaded with 1 μg/mL of Ly4_24 peptide in presence of IL2 (50 U/mL) and IL7 (10 ng/mL).

EBV-B cells (2 × 10⁶) from HLA*A2 patient CP50 were resuspended in a 15-mL tube in 250 μL of RPMI 1640 (Gibco) complemented with 2% human serum containing MVA-FSP-B2 at multiplicity of infection (MOI) of 5 and unrelated MVA encoding hcRed at MOI of 2 to control for infection efficiency. Cells were incubated for 1 hour at 37°C with 5% CO₂. Cells were washed, resuspended at 10⁶/mL in RPMI 2% HS, and plated overnight at 37°C. Infection rate was measured by flow cytometry, and cell mortality was analyzed with Trypan Blue. EBV-infected cells were then tested for their ability to stimulate IFNγ secretion by the T-cell clones. In total, 10,000 T cells were added in microwells containing 10,000 EBV-B cells either untouched, loaded with 1 μg/mL of Ly4_24 peptide, or infected with MVA-FSP-B2 encoding for Ly4_24 in a total volume of 200 μL of complete IMDM supplemented with 25 U/mL of IL2. After 20 hours, the IFNγ released in the supernatant was measured by ELISA using anti-IFNγ capture antibody (clone 350B10G6, Thermo Fisher) and biotin-conjugated anti-IFNγ monoclonal antibody (clone 67F12A8, Thermo Fisher).

FSMs detected in 320 MSI tumors of different histologies, 69 colorectal cancer, 85 gastric cancer, and 166 endometrial cancer, were subjected to a multistep filtering procedure to identify those shared across the three tumor types (33–35). Note that 1,494 FSMs present in the tumor and absent in matched healthy tissue were selected (Fig. 1A and Materials and Methods).

FSMs were mapped on the REFSEQ transcriptome to determine the amino acid sequences of the resulting FSPs, allowing for a single FSM generation of multiple FSPs due to alternatively spliced isoforms. To optimize for the presence of immunogenic epitopes, FSPs shorter than 4aa were removed and those shorter than 10aa were lengthened by adding up to 4 wild-type (wt) amino acids to the N-terminus to allow for generation of CD8 epitopes involving the junction between wt and FSP sequences. To minimize the risk of inducing autoimmunity, any FSP segment ≥ 8aa matching with wt human proteome was excluded. Finally, we obtained 1,087 FSPs suitable for a vaccine (Fig. 1B). We considered 1,087 FSPs too many for vaccine feasibility and developed an algorithm further reducing the number of FSPs to allow for their overall sequence to be encoded by 4 viral vectors. We determined that an antigenic sequence of 400aa contains on average 3 immunogenic CD8 epitopes (min 1, max 12; Supplementary Fig. S1) by reviewing published data on vaccination in healthy volunteers with antigens from HIV and HCV polyproteins encoded by viral vectors (4, 8). Therefore, we developed an algorithm to select FSPs, from the initial collection of 1,087, such that the maximum number of MSI tumor biopsies examined had in common with the putative vaccine a protein sequence of at least 400aa, intended as the sum of the lengths of FSPs encoded by the FSMs detected in the exome of each individual tumor. The final number of FSPs selected for the vaccine was 209 arising from 204 FSMs and encoding overall sequence of total 6,021aa corresponding to the maximum protein size that can be encoded by four viral vectors (Fig. 1A, Supplementary Table S1). The median length of selected FSPs is 17aa (Supplementary Fig. S2). More than 80% (166 out of 204) of the FSMs selected by the algorithm are shared across the 3 analyzed MSI tumor types (Fig. 1C) and have comparable mutation allele frequency in early- and late-stage tumors (Supplementary Fig. S3).

A median number of 50, 46, and 21 FSPs is shared between the 209 FSPs selected for the vaccine and each patient with colorectal cancer, gastric cancer and endometrial cancer, respectively (Table 1). Moreover, the vaccine has a sequence of at least 400aa in common with 98% of colorectal cancer, 95% of gastric cancer samples, and 70% of the endometrial cancer samples from TCGA database (Table 1).

To validate the FSPs' selection, an independent dataset of 20 samples of patients with MSI colorectal cancer, 3 of which were LS, was analyzed.

First, a “look-up” analysis (Materials and Methods) of tumors and matched healthy tissues was performed to confirm that Nous-209 FSMs were present in the tumor but absent as germline mutations. Detected FSMs encode a median number of 51 FSPs for a cumulative length ranging from 65 to 2,415aa (Fig. 2A and B), confirming data obtained by analyzing TCGA tumors. Across the 20 analyzed samples of patients with MSI, only 2 Nous-209 FSMs were detected in the DNAs from healthy mucosa of 2 patients (1 FSM/patient) with an allele frequency closed to the threshold of detection (10%; Supplementary Fig. S4A).

A median of 31 FSPs (62% of FSPs present in tumor DNA; min 1–max 57 FSPs/sample) was found to be expressed in tumor cells (red bars; Fig. 2C). Importantly, the 2 FSMs present in DNA of healthy tissues were not detected in the RNA (Fig. 2C). Moreover, each patient's tumor RNA had in common with the vaccine between 26 and 1,475aa (median 718aa), intended as the cumulative length of shared FSPs (Fig. 2D).

Exome sequence data were also analyzed by somatic variant calling, confirming almost all FSMs detected in tumor DNAs by the look-up analysis, with the only exception of sample 004, for which variant callers identified more FSMs than look-up analysis likely because FSMs detected in this patient have allele frequencies below the threshold used for look-up analysis (Supplementary Fig. S4B).

We estimated the potential of Nous-209 FSPs to encode CD8-binding peptides, by predicting FSP-derived 8–10-mer peptides that bind patient's HLA-I haplotypes, with an IC₅₀ ≤ 500 nmol/L (Fig. 3A). The Nous-209 vaccine encodes a median number of 335 HLA-I–predicted binder peptides per patient (Fig. 3A; orange bars), of which, a median of 27 (minimum 0 and maximum 100) is detected in both DNA and mRNA extracted from each patient's tumor biopsy (Fig. 3A; blue bars). Only in 1 patient (ID:12919), no CD8 binder peptides were predicted, likely because this patient harbors very rare HLA-I haplotypes for which a limited amount of experimental data are available in IEDB database, lowering the performance of the prediction algorithm. By looking at class-II predictions, we found a median of 751 predicted binder peptides, of which, 90 were encoded by FSMs expressed in each patient (Fig. 3B).

The 209 FSPs were assembled into 4 artificial genes by joining their sequences one after the other in a head-to-tail configuration. Each artificial gene was assembled in two layouts (A and B) containing the same set of FSPs, but in a different order. Genes in the layout A were used for the construction of 4 GAd vectors (GAd-FSP-A1 to A4), whereas genes in the layout B were encoded in 4 MVA vectors (MVA-FSP-B1 to B4). The scrambling of FSPs between the 2 layouts was designed to avoid GAd and MVA encoding the same junctional aminoacidic sequences between adjacent FSPs, which could boost immune responses against epitopes at the junctions. The 4 GAd and the 4 MVA vectors were then mixed to generate, respectively, the polyvalent GAd-209-FSP and MVA-209-FSP vaccines, which are collectively referred to as Nous-209 (Supplementary Fig. S5A and S5B).

To demonstrate the feasibility of our vaccination approach in generating T-cell response against multiple target antigens, the immunogenicity of Nous-209 was evaluated in CB6F1 mice using heterologous prime/boost regimen according to the scheme depicted in Fig. 4A. GAd-209-FSP and MVA-209-FSP were administered i.m. at the dosage of 4 × 10⁸ vp and 4 × 10⁷ ifu respectively. Mice were primed at week 0 with GAd-209-FSP. Two weeks later, a group of animals was sacrificed to evaluate GAd-primed immune responses, whereas a second group was boosted with a single dose of MVA-209-FSP. FSP-specific T-cell responses post prime (week 2) and post MVA boost (week 3) were measured by IFNγ ELISpot on splenocytes restimulated with peptides covering the entire neoantigenic sequence encoded by Nous-209 (Fig. 4B–C). Priming with GAd-209-FSP elicited a strong T-cell–mediated immunity. Responses were efficiently boosted by MVA-209-FSP, with approximately overall 10,000 SFCs/10⁶ splenocytes (Fig. 4B). Antigen-specific immune responses against each part of the neoantigen sequence covered by 16 peptide pools were detected in vaccinated animals (Fig. 4C and Materials and Methods). These results, although do not directly allow to derive information about immunogenicity in human, demonstrate the feasibility of employing Nous-209 vaccine for effective prime/boost immunization.

To exclude immunological interference due to vectors' coadministration, we selected FSP1 vector for proof of concept. Therefore, we measured the immune responses induced after vaccination with GAd-FSP-A1 as a single vector (1 × 10⁷ vp) or coadministered with the 3 other GAd vectors (each of them at the dose of 1 × 10⁷ vp). Animals receiving GAd-FSP-A1 were boosted with the corresponding MVA vector (MVA-FSP-B1) at 1 × 10⁷ pfu, whereas the second group of mice, receiving the mixture of 4 GAds, was boosted with the 4 MVAs. T-cell responses against the pools P1–P4 covering the FSPs in GAd-FSP-A1 and MVA-FSP-B1 were measured in both groups and found comparable, thus excluding immunological interference in case of concomitant, multivalent vaccination with the 4 viral vectors (Fig. 4D). Moreover, the quality of induced T-cell responses was investigated by flow cytometric intracellular staining on splenocytes harvested post MVA GAd-FSP-A1/MVA-FSP-A1 vaccine. Cells were stimulated with pools 1 and 3, which yielded the strongest IFNγ ELISpot responses against the FSPs encoded by the first vector. Both CD8 and CD4⁺ IFNγ⁺ FSP-specific T cells were induced in vaccinated mice (Fig. 4E).

To demonstrate that Nous-209–encoded epitopes can be processed, loaded in the MHC pocket of human cells and recognized by specific human T cells, we performed an in vitro experiment. Ninety-five FSPs present in Nous-209 and observed in the majority of patients with colorectal cancer (≥25% of patients) were analyzed for epitopes predicted to bind the HLA-A*02:01 molecule. Predicted epitopes were ranked by predicted IC₅₀ (Supplementary Table S2). Among the top scoring peptides, we selected the Ly4_Ly24 peptide (derived from FSP CKAP2) for proof-of-concept in vitro stimulation study, after confirming its binding to HLA-A*02:01 by T2 binding assay. We synthesized an HLA multimer folded with the Ly4_Ly24 peptide to enrich for FSP-specific CD8 T cells from an HLA-A*02:01 subject (36). The sorted T cells were expanded by two rounds of in vitro antigenic stimulation with autologous DCs pulsed with the Ly4_Ly24 peptide. The microcultures were screened with a fluorescent HLA-A2/Ly4_Ly24-multimer, and 3 out of 301 were found positive. Clonal populations were isolated from each of the 3 microcultures by flow cytometry, and amplified in vitro by several rounds of antigenic stimulation. Using flow cytometry, we isolated 3 T-cell clones able to bind specifically the fluorescent HLA-A2/Ly4_Ly24 multimer, but not a multimer with an unrelated peptide (Fig. 5A and C; Supplementary Fig. S6). Moreover, we showed that those clones selectively secrete IFNγ when incubated with HLA-matched EBV-B cells, either loaded with the neoantigen peptide or infected with the MVA-FSP-B2 encoding the Ly4_Ly24 peptide (Fig. 5B).

The knowledge that passive transfer of neoantigen-specific T cells resulted in complete remissions of patients with advanced malignancies makes neoantigens a highly desirable targets for cancer vaccination (37–40).

MSI solid tumors are characterized by a high number of indel mutations in coding genes that can give rise to FSPs that are entirely novel, nonself peptides, and, as such, potentially the most immunogenic and safe type of neoantigens (14, 15).

Interestingly, a positive correlation between the total number of FSPs detected in MSI colorectal cancers and the CD8 TIL density was observed (13), highlighting an important role for FSP-specific CD8 T-cell response in controlling MSI tumors and providing a rationale for developing cancer vaccines targeting FSPs to strengthen and broaden the spontaneous immune response. Moreover, this type of mutations was shown to be shared to a variable extent across patients with MSI (41, 42), offering a unique opportunity for an “off-the-shelf” vaccination approach. Indeed, some shared FSPs derived from frequently mutated genes have been tested as peptide vaccine in patients with MSI colorectal cancer (NCT01885702 and NCT01461148), and vaccination was shown to be safe and immunogenic (43, 44).

Our work describes the widespread occurrence of shared FSPs in MSI tumors and translates those findings into the generation of Nous-209, the first genetic vaccine for MSI tumors. We selected 209 tumor-specific FSPs using TCGA dataset by taking into account several parameters to maximize the probability of inducing a tumor-specific immune response intrapatient and across different patients' HLA haplotypes. Nous-209 was conceived to achieve a very broad coverage by targeting many different FSPs to face highly heterogeneous and rapidly evolving tumors.

Off-the-shelf cancer vaccines represent a very attractive approach as they overcome the complexity of a personalized cancer vaccine, including potential delays between biopsy, vaccine manufacturing, and release tests, as well as the high costs associated with vaccine production. To date, the current “off-the-shelf” neoantigen-based cancer vaccines target single or few frequently shared driver mutations (45). This implies: (i) the need for patient preselection based on the presence of the targeted mutations and (ii) the high likelihood of tumor escape if the mutation(s) in common with the vaccine is lost under the immune pressure. Our vaccine does not require MSI patient screening for specific mutations by NGS and has a high probability of inducing effective antitumor activity by targeting a wide number of tumor neoantigens.

To validate the selection of FSPs included in vaccine, we used an independent cohort of samples of patients with MSI colorectal cancer. The presence of vaccine-encoded mutations was confirmed in this cohort with the same frequencies observed in the TCGA dataset. Importantly, none of the FSMs was expressed in the normal adjacent mucosa of patients with MSI, highlighting the anticipated safe profile of the vaccine.

The value of Nous-209 is further strengthened by the use of a potent genetic vaccine platform based on GAd and MVA viral vectors. This platform has been validated in the field of infectious diseases for its potent induction of T cells in humans (4, 8). Here, we demonstrated a powerful immunogenicity of the Nous-209 vaccine in mice in which we could demonstrate the prime/boost effect with a balanced induction of both CD4 and CD8 T cells and the lack of interference among the four coadministered vectors. Although we could not evaluate the efficacy of the Nous-209 vaccination in an MSI model, we recently provided a proof of concept of our vaccination platform encoding neoantigens in murine cancer models. Vaccination as standalone treatment was demonstrated to be 100% effective in a prophylactic setting, and to synergize with anti–PD-1 in advanced therapeutic settings of tumor-bearing mice, with a 3-fold increase of the cure rate compared with anti–PD-1 monotherapy (31).

MSI tumors showed indeed a good response to anti–PD-1 treatment, and the expectation is to increase the response rate by expanding T cells against relevant tumor neoantigens (9, 10).

The presence of a human T-cell repertoire against FSPs has been demonstrated both in healthy volunteers and patients with cancer (13). Here, we provide further evidence by reporting an additional reactivity. Although this finding is limited to only one FSP in Nous-209, we provided a demonstration that human antigen-presenting cells can process and present an epitope derived from a Nous-209 vaccine–encoded FSP, resulting in activation of human T-cell clones specific for that reactivity.

The breadth of the immune response inducible by Nous-209 vaccine was investigated by predicting the number of potential MHC binders present in the vaccine according to the patient's HLA class I and II haplotypes. Nous-209 targets a median of 27 class-I and 90 class-II–predicted MHC binders in FSPs expressed by each patient's tumor and median number of 335 class-I and 678 class-II additional epitopes in FSPs potentially arising over the course of disease progression/relapse.

Mutations in the β-2-microglobulin (B2M) gene, frequently observed in these patients, could represent an obstacle to vaccine effectiveness (46). Nevertheless, recently published evidence indicates that 85% of B2M-mutated MSI colorectal cancer tumors do benefit from checkpoint inhibitor therapy (47). Based on these results, we expect that only a limited number of patients will not be able to respond to the vaccine because of impaired HLA-I presentation in the tumor.

We previously showed that benign polyps acquire a hypermutated phenotype and start to accumulate FSPs derived from indel mutations as they become more advanced (48). Nous-209 could therefore represent a prophylactic approach to attack these types of lesions and prevent their transformation into malignant tumors in LS carriers, for which preventive options are today limited to aspirin that was shown to slightly reduce cancer incidence in the CAPP2-randomized trial (49). This study opens new avenues for treatment and prevention of MSI tumors. Nous-209 vaccine is currently under clinical evaluation in patients with metastatic colorectal cancer, gastric, and gastro-esophageal cancer in combination with Pembrolizumab (NCT04041310).

G. Leoni reports being the inventor and patent applications related to the manuscript: “A Universal Vaccine Based on Shared Tumor Neoantigens for Prevention and Treatment of Micro Satellite Instable (MSI) Cancers,” WO2019/012082. M. Yadav is an employee of Genentech Inc. M. Gordon-Alonso is currently an employee of F-star Therapeutics outside the submitted work. C. Vanhaver reports other funding from Nouscom (the work on specific T cells was performed in the frame of a research contract with Nouscom) during the conduct of the study. M. Panigada reports personal fees from Nouscom during the conduct of the study. E. Soprana reports personal fees from Nouscom during the conduct of the study. A. Siccardi reports personal fees from Nouscom during the conduct of the study. S. Colloca reports being the inventor and one patent application related to the manuscript “Non Human Great Apes Adenovirus Nucleic Acid- and Amino Acid-Sequences, Vectors Containing Same, and Uses Thereof,” WO 2019/008111. P. van der Bruggen reports other funding from Nouscom (the work on specific T cells was performed in the frame of a research contract with Nouscom) during the conduct of the study. A. Nicosia reports being the inventor and two patent applications related to the manuscripts (1) “Non Human Great Apes Adenovirus Nucleic Acid- and Amino Acid-Sequences, Vectors Containing Same, And Uses Thereof,” WO2019/008111 and (2) “A Universal Vaccine Based on Shared Tumor Neoantigens for Prevention and Treatment of Micro Satellite Instable (MSI) Cancers,” WO2019/012082. A. Lahm reports being the inventor and two patent applications related to the manuscript (1) “Non Human Great Apes Adenovirus Nucleic Acid- and Amino Acid-Sequences, Vectors Containing Same, and Uses Thereof,” WO2019/008111 and (2) “A Universal Vaccine Based on Shared Tumor Neoantigens for Prevention and Treatment of Micro Satellite Instable (MSI) Cancers,” WO2019/012082. E. Scarselli reports being the inventor and one patent application related to the manuscript “A Universal Vaccine Based on Shared Tumor Neoantigens for Prevention and Treatment of Micro Satellite Instable (MSI) Cancers,” WO2019/012082. No potential conflicts of interest were disclosed by the other authors.

G. Leoni: Conceptualization, data curation, software, formal analysis, validation, investigation, visualization, methodology, writing-original draft, writing-review and editing. A.M. D'Alise: Conceptualization, data curation, formal analysis, investigation, methodology, writing-original draft, writing-review and editing. G. Cotugno: Conceptualization, resources, investigation, writing-original draft. F. Langone: Resources, formal analysis, investigation. I. Garzia: Resources, formal analysis, investigation. M. De Lucia: Resources, formal analysis, investigation. I. Fichera: Resources, formal analysis, investigation. R. Vitale: Resources, formal analysis, investigation. V. Bignone: Resources, formal analysis, investigation. F.G. Tucci: Software, formal analysis, investigation. F. Mori: Resources, investigation. A. Leuzzi: Resources, investigation. E. Di Matteo: Resources, formal analysis, investigation. F. Troise: Resources, formal analysis, investigation. A. Abbate: Resources, formal analysis, investigation. R. Merone: Resources, formal analysis, investigation. V. Ruzza: Resources, formal analysis, investigation. M.G. Diodoro: Resources. M. Yadav: Resources, writing-review and editing. M. Gordon-Alonso: Resources, investigation. C. Vanhaver: Resources, investigation. M. Panigada: Resources, investigation. E. Soprana: Resources, investigation. A. Siccardi: Resources, formal analysis, investigation, writing-review and editing. A. Folgori: Formal analysis, writing-review and editing. S. Colloca: Formal analysis, writing-review and editing. P. van der Bruggen: Conceptualization, formal analysis, investigation, writing-review and editing. A. Nicosia: Conceptualization, supervision, funding acquisition, visualization, methodology, project administration, writing-review and editing. A. Lahm: Conceptualization, data curation, software, formal analysis, validation, investigation, methodology, writing-original draft, writing-review and editing. M.T. Catanese: Conceptualization, formal analysis, investigation, writing-original draft, project administration, writing-review and editing. E. Scarselli: Conceptualization, supervision, funding acquisition, validation, investigation, visualization, methodology, writing-original draft, project administration, writing-review and editing.

We are grateful to Professor Riccardo Cortese, Nouscom founder and former CEO. Most of this work would not have been possible without his ideas and guidance. We thank Marina Udier, CEO of Nouscom, for critical reading of the article and helpful discussion.

We acknowledge the animal facility of Plaisant in Castel Romano (Rome) for the maintenance and care of the mice used in this study.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.