Background

Cytotoxic CD8+ T cells can specifically recognize and eliminate cancer cells when detecting tumor-derived antigens presented via major histocompatibility complex (MHC) molecules.1 Mutated neoantigens (NeoAg) are tumor-specific antigens derived from cancer cell-specific genetic alterations and can be effective targets for T cell-mediated tumor cell killing.2–4 The quality or immunogenicity of individual NeoAg is defined by multiple parameters, most importantly MHC-binding characteristics, and the likelihood of recognition by T cells.5 6 Recent findings however indicate that NeoAg architectures portray an additional key determinant for the immunogenicity of NeoAg, thus impacting the strength of the antitumor T-cell response.7–9 NeoAg architectures are defined by the overall abundance as well as the clonality of specific NeoAg9 and are shaped during tumor evolution, which frequently leads to a coexistence of phenotypically distinct tumor subclones.10–12 This intratumoral heterogeneity (ITH) is inherent to most solid cancers10–12 and can lead to the expression of clonal as well as subclonal NeoAg, which are respectively shared between all, or limited to a fraction of tumor cells. ITH and subclonal NeoAg expression have been shown to effectively blunt antitumor T-cell responses.13 14 Moreover, clinical data have established NeoAg clonality as a key predictor of response to immune-checkpoint blockade immunotherapy (ICB) in patients with cancer.14–16 While this clinical correlation is well established,16 it is poorly understood how ITH impairs antitumor immunity on a mechanistic level.

NeoAg architectures could impact antitumor T-cell responses both through the abundance as well as the clonality of tumor NeoAg. Concurrent T-cell responses targeting distinct antigens are known to influence one another,7 8 but the determinants dictating their interplay during tumor evolution are incompletely understood. The establishment of antigen hierarchies between CD8+ T-cell responses has been described in murine cancer models,7 mirroring the interplay between T-cell responses observed in viral disease.17 In these hierarchies, a dominant NeoAg-specific immune response is enhanced at the expense of subdominant immune responses.7 Enhanced NeoAg-specific responses on the other hand can be observed in the context of CD4+ T-cell response-mediated help based on tumor expression of MHC-II-restricted NeoAg18–21 and, as described more recently, in the context of concurrent CD8+ T-cell responses to MHC-I-restricted NeoAg lacking observable immunodominance.8 How NeoAg clonality and the interplay between T-cell responses act together to impact antitumor immunity in complex, subclonal NeoAg architectures is unknown. This is a critical issue, as human cancers, with few exceptions, are subclonal diseases, typically consisting of one to three distinct tumor subclones.10 Patients with tumors exhibiting high ITH currently derive little or no benefit from ICB.9 16 A better understanding of how ITH blunts antitumor immune responses could enable the development of novel treatment approaches for patients. A deepened understanding is further pivotal for the design of multivalent cancer vaccines. Defining ideal targets and antigen combinations could allow therapeutically leveraging of (neo-)antigen architectures to maximize antitumor immunity.

In this study, we therefore investigated how NeoAg architectures impact antitumor T-cell responses. We used a reductionist preclinical mouse model with defined NeoAg expression to decipher the interplay between concurrent NeoAg-specific CD8+ T-cell responses. This further allowed the modeling of complex, clinically relevant NeoAg architectures to study the impact of this interplay in tumors with heterogenous NeoAg expression. We found that antigen presentation by cross-presenting dendritic cells (cDC1) in the lymph node mirrors tumor NeoAg architectures as they orchestrate the interplay between concurrent T-cell responses. Mediated by cDC1, NeoAg expression patterns in the tumor thus define the NeoAg architecture-dependent immunogenicity (NADi) of individual NeoAg and tumor subclones. In tumors with heterogenous NeoAg expression, suppressed NADi drove immune evasion and mediated resistance to ICB. Therapeutic RNA-based vaccination targeting suppressed NeoAg responses synergized with ICB to overcome impaired immunogenicity of tumors with heterogenous NeoAg expression and enabled tumor control. Combining ICB with clonal NeoAg-targeting vaccination might thus represent a tailored treatment approach for patients with tumors exhibiting high ITH.

Materials and methodsMice

C57BL/6 wildtype mice (B6, strain #000664) and B6-CD45.1+ mice (strain #002014) were purchased from Jackson Laboratories. Rag2-knockout mice (Rag2−/−, strain #008449) were purchased from Jackson Laboratories and bred in-house. Mice were housed under specific pathogen-free conditions at the Koch Institute animal facility. Mice were gender-matched and age-matched for experiments (6–12 weeks old at the time of experimentation). All animal procedures were approved by the Committee on Animal Care at MIT.

Tumor cell lines, cell culture and tumor injections

Parental cancer cell lines KP1233 and RMA-S were gifts from Tyler Jacks (Koch Institute for Integrative Cancer Research at MIT). KP6S was subcloned from KP1233. Tumor cell lines were cultured at 37°C and 5% CO2 in complete media (Dulbecco's Modified Eagle Medium (DMEM) (Gibco) supplemented with 10% heat-inactivated Fetal Bovine Serum (FBS) (Atlanta Biologicals), 1% penicillin/streptomycin (Gibco), and 20 mM 4-(2-Hydroxyethyl)piperazine-1-ethanesulfonic acid (HEPES) (Gibco)). For tumor injections, tumor cells were harvested by trypsinization (Gibco), washed three times with phosphate-buffered saline (PBS) (Gibco), resuspended in PBS and 1×106 tumor cells were injected subcutaneously on the flank. For tumor outgrowth experiments, the subcutaneous tumor area (length×width) was measured every 2–3 days using digital calipers.

Generation of NeoAg expression vectors and NeoAg-expressing cell lines

To generate NeoAg expression vectors, a fluorescent protein (mCherry, ZsGreen or Cerulean) was first cloned into the backbone vectors pLV-EF1α-IRES-puro and pLV-EF1α-IRES-Blast (Addgene plasmids #85132 and #85133). These were then linearized by enzymatic digestion with HF-SpeI and HF-EcoRI restriction enzymes (NEB) before oligonucleotides (Genewiz) containing the NeoAg-genetic barcode sequences were cloned into the linearized vectors. Fluorophores were expressed upstream of the NeoAg, separated by a spacer sequence. NeoAg sequences encoded for the mutant peptide flanked by two amino acid overhangs of the wildtype sequence on each side. When expressing multiple NeoAg, these were separated by spacer sequences. The length of the NeoAg-encoding sequence within the construct was kept constant between different constructs (online supplemental figure S1A). The resulting constructs were amplified and sequenced for accuracy. Constructs were subsequently used to generate third generation lentiviruses using Lenti-X 293 T cells. A functional lentivirus titer was obtained through serial dilution of the virus and infection of KP6S cells followed by selection with puromycin (Gibco) or blasticidin (Gibco) (7 and 10 days, respectively). Cells were then fixed (10% formalin) and stained with crystal violet stain before colonies were counted to determine the functional virus titer. To generate NeoAg-expressing cell lines, the parental line KP6S was transduced with titered lentiviruses at an multiplicity of infection (MOI) of 0.1, diluted in complete media supplemented with 4 µg/mL protamine sulfate (Sigma-Aldrich). The virus-containing media was replaced with complete media after 24 hours, selection was started 48 hours after transduction using selection media (containing either 2 µg/mL puromycin or 10 µg/mL blasticidin). Flow cytometry-based analysis of fluorescent protein expression was used to confirm transduction and to determine construct expression levels (online supplemental figure S1B).

Neoantigens and peptides

Naturally presented, endogenous NeoAg were identified in the literature.4 13 20 22 23 NeoAg sequences and sources are listed in online supplemental table S1. MHC-binding prediction was performed using NetMHCpan-4.1.24 NeoAg peptides were purchased from GenScript (Adpgk, Spb2, Aatf, Cpne1, Lama4, Alg8) and Peptide2 (Intb1) at >95% purity and reconstituted in 100% Dimethylsulfoxide (DMSO) at a stock concentration of 10 µg/µL.

Tissue processing for flow cytometry, cell sorting and ELISpot

Tumors, tumor-draining lymph node (tdLN) and spleens were resected and stored in Roswell Park Memorial Institute Medium (RPMI) (Gibco) or tumor digestion buffer (RPMI supplemented with 250 µg/mL Liberase (Roche) and 2 mg/mL DNAse (Roche)) on ice before further processing. LN were either directly mashed through a 70 mm filter into RPMI (Gibco) for T-cell analysis, or processed using a method adapted from Ruhland et al25 as described previously.26 Spleens were mashed through 70 µm filters to obtain a single-cell solution and red blood cells were lysed with ACK lysis buffer (Gibco) for 5 min on ice. Splenocytes were washed twice with RPMI before further processing. Tumors were weighed and minced using razor blades before enzymatic digestion at 37°C for 30 min. Digested tumors were mashed through 70 µm filters to obtain a single-cell solution. For ≥day 10 tumor analyses, lymphocytes were isolated using Ficoll (Sigma-Aldrich). Cells were washed twice with RPMI before further processing.

Flow cytometry staining and analysis

For flow cytometry staining, cells were first resuspended in Fluorescence-activated cell sorting (FACS) buffer (PBS (Gibco) with 1% FBS (Atlanta Biologicals) and 2 mM EDTA (Invitrogen)) containing Fixable Viability Dye eFluor 780 to distinguish live and dead cells and αCD16/CD32 (clone 93, BioLegend) to prevent non-specific antibody binding, and incubated for 20 min at 4°C. Cells were then washed with FACS buffer and stained for surface proteins using fluorophore-conjugated antibodies resuspended in FACS buffer. Following surface staining, cells were washed twice with FACS buffer and analyzed directly or fixed for downstream intracellular staining and/or analysis the next day. Cell fixation was performed using the Foxp3 Transcription Factor Fixation/Permeabilization Buffer (eBioscience) following manufacturer’s instructions. Following fixation, cells were washed twice with FACS buffer and stained for intracellular markers overnight at 4°C. Finally, cells were washed twice with FACS buffer prior to flow cytometry analysis. Precision count beads (BioLegend) were used to determine the total numbers of cells present in a sample. Sample acquisition was performed on BD flow cytometers (FACSCanto II, Symphony A3, Fortessa) and analyzed using FlowJo V.10 (TreeStar). For cell sorting, the surface staining was performed as described above under sterile conditions and cells were sorted on an FACSAria III sorter (BD). For CD8+ T-cell analyses, cells were pre-gated on live, singlets, CD45+ (or CD45.1+ when congenically labeled), CD3e+, CD4–, CD8+ surface markers. For DC analyses, cells were pre-gated on live, singlets, CD45+, CD19–, CD3e–, NK1.1–, Ly6C+, MHC-II+, F4/80–, CD11c+ and mCherry+/ZsGreen+ when applicable.

Tetramers and tetramer-staining

Biotinylated MHC monomers were generated at the NIH Tetramer Core Facility and tetramerized using fluorophore-linked streptavidin (premium-grade PE-Streptavidin or APC-Streptavidin (Invitrogen)). Cells were pre-incubated with 50 nM Dasatinib (Sigma-Aldrich) for 25 min at 37°C to enhance tetramer binding27 before staining at 4°C for 30 min together with surface marker staining.

Adoptive bulk CD8+ and tetramer-sorted T-cell transfers

Splenocytes were isolated from spleens of naïve or tumor-bearing mice on day 7 after tumor injection and CD8+ T cells were enriched using magnetic cell separation (CD8a+T Cell Isolation Kit (Miltenyi Biotec)). For bulk CD8+ T-cell transfers, 5×106 CD8+ T cell-enriched cells per donor group were transferred into tumor-bearing recipient mice (bearing single NeoAg-expressing tumors on opposite flanks; day 4 after tumor injection). The total number of transferred CD8+ T cell-enriched cells was kept constant between recipient groups (table 1).

Table 1

Adoptively transferred CD8+ T cells

For the transfer of Tetramer-sorted cells, CD8+ T cell-enriched cells were stained, and tetramer-positive cells were sorted on an FACSAria III cell sorter (BD). 5,000 tetramer-positive cells per mouse were then transferred into tumor-bearing (day 4 after tumor injection) recipient mice.

Interferon-γ ELISpot assay

Enzyme-linked immunosorbent spot (ELISpot) plates (EMD Millipore) were coated overnight at 4°C with anti-interferon (IFN)-γ capture antibody (BD Biosciences). Plates were washed and blocked with complete media for 2 hours at room temperature. Splenocytes were plated in complete media at 0.5×106 or 1×106 cells/well with either 50 ng/µL (10 µg/well) NeoAg peptide, negative control (complete media) or positive control (complete media supplemented with 100 ng/mL PMA (Sigma-Aldrich) and 1 mg/mL ionomycin (Sigma-Aldrich)). Plates were incubated overnight (16–18 hours) at 37°C and 5% CO2 and developed using a mouse IFN-γ ELISpot kit (BD Biosciences), following manufacturer’s instructions. After drying (overnight at room temperature), spot counts were determined using an ImmunoSpot (CTL) ELISpot reader.

In vivo cytotoxicity assay

In vivo cytotoxicity was determined as described previously.20 In brief, donor splenocytes were isolated from naïve B6 and B6-CD45.1+ mice and stained with either carboxyfluorescein succinimidyl ester (CFSE) (0.5 µM or 5 µM for CFSElow and CFSEhi) or cell trace violet (CTV) (0.5 µM or 5 µM for CTVlow and CTVhi). Cells were washed with PBS and pulsed with 10 µg/mL peptide for 2 hours in lymphocyte media (complete media supplemented with non-essential amino acids (Thermo Fisher), 1 mM sodium pyruvate (Thermo Fisher) and 55 µM ß-mercaptoethanol (Gibco). Unpulsed splenocytes served as a control population. Cells were washed three times with PBS, mixed in 50/50 ratios and 5×106 cells were retro-orbitally transferred into day 10 tumor-bearing recipient mice. After 20 hours, spleens were harvested and the ratio of peptide-pulsed to unpulsed cells was determined by flow cytometry. The ratio was normalized to the ratio determined in naïve control mice. Per cent specific lysis was calculated as (1 − (naïve control ratio/experimental ratio)) × 100.

Peptide:MHC binding assays

Affinity of peptide:MHC (pMHC) binding was determined using TAP-deficient RMA-S cells. In brief, RMA-S cells were first incubated at 26°C for 16 hours to increase the surface expression of empty MHC molecules.28 Cells were then incubated with peptide at 26°C for 2 hours before degrading empty MHC molecules at 37°C for 1 hour. The relative quantity of peptide-stabilized MHC molecules was assessed by staining with anti-H2-Db and/or anti-H2-Kb antibodies for 20 min at 4°C followed by fixation and flow cytometry-based analysis.

DNA extraction and quantitative PCR

Genomic DNA was isolated using the GenElute Mammalian Genomic DNA Miniprep Kit (Sigma-Aldrich) following manufacturer’s instructions. Isolated DNA was quantified by NanoDrop and diluted to 1 ng/µL for quantitative PCR (qPCR). Standard curves were generated using genomic DNA extracted from respective tumor cell lines. qPCR was performed using an SYBR Green PCR Master Mix (Applied Biosystems) and run on a StepOne Real-Time PCR System (Applied Biosystems). Genetic barcode copy numbers were interpolated using a standard curve for each barcode. Subclonal fractions were normalized using an aliquot of the injected tumor cell solution as a reference.

In vivo antibody treatments

CD8+ T-cell depletion: CD8+ T cells were depleted in vivo by administering 200 µg anti-CD8-antibody (Bio X Cell) intraperitoneally (i.p.) 2 days prior to tumor injection followed by 100 µg twice weekly during ongoing experiments. H2 allele-specific blockade: Allele-specific H2-Kb in vivo blockade was performed by administering 400 µg anti-H2-Kb-antibody (Bio X Cell) i.p. 1 day prior to tumor injection followed by 200 µg on day 2 and 4 after tumor injection. Immune checkpoint blockade therapy (ICB): Mice were injected i.p. with 100 µg of anti-cytotoxic T-lymphocyte associated protein 4 (Bio X Cell) and 100 µg anti-programmed death-ligand 1 PD-L1 (Bio X Cell) in PBS on days 7, 10, 13 and 16 after tumor injection.

Replicon RNA synthesis

Venezuelan equine encephalitis virus (VEE) replicon plasmid DNA was prepared based on mutant constructs and cloned after the subgenomic promoter as described previously.29–31 VEE DNA was linearized via endonuclease digestion and purified with PureLink PCR Purification columns (Thermo Fisher) following manufacturer’s instructions. To synthesize RNA, 20 µL in vitro transcription (IVT) reactions were performed using the HiScribe T7 High Yield RNA Synthesis Kit (NEB) and 1–2 µg of linear DNA template (scaled as needed). The IVT product was purified using PureLink RNA Mini columns (Thermo Fisher) following manufacturer’s instructions. RNA was then capped and methylated using the ScriptCap Cap 1 Capping System (CellScript) following manufacturer’s instructions, after which RNA was purified a final time using PureLink RNA Mini columns. The quality of the resulting replicons was assessed using UV-Vis spectrophotometry and gel electrophoresis.

Lipid nanoparticle formulation and replicon RNA vaccines

Self-replicating (replicon) RNA vaccines were designed and synthesized as described previously.32 Lipid nanoparticles (LNPs) were composed of N1,N3,N5-tris(3-(didodecylamino)propyl)benzene-1,3,5-tricarboxamide (TT3), (6Z,9Z,28Z,31Z)-Heptatriaconta-6,9,28,31-tetraen-19-yl 4-(dimethylamino) butanoate (DLin-MC3-DMA; MedChemExpress), 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine (DOPE; Avanti Polar Lipids), Cholesterol (Avanti Polar Lipids), and 1,2-dimyristoyl-rac-glycero-3-methoxypolyethylene glycol-2000 (DMG-PEG2k; Avanti Polar Lipids), mixed at a molar ratio of 10:25:20:40:5 in ethanol. Replicon-RNA was diluted in 10 mM citrate buffer (pH 3.0) and formulated in LNPs at amine-to-phosphate (N/P) ratio of 2:1 through microfluidic nanoprecipitation (NanoAssemblr Ignite instrument, Precision NanoSystems) at a volume ratio of 1:2 (organic:aqueous) and flow rate of 12 mL/min. Replicon-RNA containing LNPs were dialyzed against PBS prior to intramuscular injection into the gastrocnemius muscle (1 µg RNA/dose).

Bone marrow-derived dendritic cell generation and culture

Bone marrow-derived dendritic cells (BMDCs) were generated as described previously.33 In brief, bone marrow was isolated from femurs and tibias of naïve mice, red blood cells were lysed using ACK lysis buffer, and remaining cells were washed and cultured in BMDC media (RPMI supplemented with 10% FBS, 1 mM HEPES, 55 µM β-mercaptoethanol, non-essential amino acids, 2 ng/mL murine GM-CSF (BioLegend), and 100 ng/mL human Flt-3L-Ig (Bio X Cell). Media was changed every 2 days and cells were split on day 4 before harvest on day 7. BMDCs were frozen in 10% DMSO (Sigma-Aldrich) in FBS (Atlanta Biologicals) and stored in liquid nitrogen. BMDCs were phenotyped using flow cytometry.

ß2-mikroglobulin (B2M)-knockout cell line generation

B2M−/− cell lines were generated as described previously.33 In brief, the KP6S parental line was transiently transfected with three pooled CRISPR guides targeting exon 2 of murine B2M cloned into the px459-Cas9-puro vector (Addgene #62988) followed by selection with 2.5 µg/mL puromycin for 48 hours. B2M-KO was confirmed by flow cytometry-based analysis of surface H-2Db/Kb expression. NeoAg-expression in B2M−/− cell lines was engineered as described above.

Tumor cell irradiation and co-culture with BMDCs

To generate tumor debris, tumor cells were trypsinized, counted and irradiated with 20 Gy (gray) on ice before being cultured for 96 hours. Non-adherent tumor debris was then harvested, counted and added to BMDCs (thawed 24 hours earlier). BMDCs were co-cultured with tumor cell debris for 18 hours, collected, washed with PBS and flash-frozen in liquid nitrogen before processing for mass spectrometry.

Heavy isotope-labeled peptide synthesis

Heavy amino acid labeled Alg8-peptides were generated at the Biopolymers and Proteomics core facility at MIT. Synthesized peptides were cleaved using a standard cleavage cocktail and purified to >95% using high-performance liquid chromatography (HPLC). Molecular weight was confirmed using a MALDI mass spectrometer (Bruker microflex). Heavy isotope-labeled amino acids used for synthesis were purchased from Cambridge Isotope Laboratories (table 2).

Generation of recombinant heavy isotope-labeled peptide MHCs

Heavy amino acid-labeled Alg8 peptides and positive control peptides (provided by manufacturer) were loaded on recombinant, empty mouse H2-Kb monomers (easYmers, Immunaware) according to manufacturer’s protocol. The concentration of stable complexes post loading was quantified using a protocol adapted from Flex-T HLA class I ELISA assay (BioLegend).

pMHC isolation

Cell pellets containing 2.2–5×106 co-cultured BMDCs were resuspended in 1 mL MHC lysis buffer (20 mM Tris-HCl pH 8.0, 150 mM NaCl, 0.2 mM PMSF (Sigma), 1% CHAPS (Sigma), and 1× Halt Protease/Phosphatase Inhibitor Cocktail (Thermo Scientific)), followed by brief sonication at 4°C (3×10 s microtip sonicator pulses at 30% amplitude) to disrupt cell membranes. Lysates were cleared by centrifugation at 16,000 ×G for 15 min at 4°C. pMHCs were isolated from lysates by immunoprecipitation (IP) and size exclusion filtration, as previously described.34 In brief, for each sample 0.1 mg of anti-mouse H2-Kb antibody (clone Y-3 clone, InVivoMAb, catalog #BE0172) was conjugated to 20 µL FastFlow Protein A Sepharose bead slurry (Cytiva). Beads were washed 1× with IP wash buffer (20 mM Tris-HCl pH 8.0, 150 mM NaCl), followed by the addition of lysate and known amounts of isotopologue heavy isotope-labeled peptide MHCs (see Table 2 below), and incubated rotating overnight at 4°C to immobilize pMHCs on beads. Beads were washed with 1× TBS and 2× water, and pMHCs were eluted using 10% acetic acid for 20 min at RT. Peptides were isolated from antibody and MHC molecules using a 10 K molecule weight cut-off filter (PALL Life Science), lyophilized, and stored at −80°C until analysis.

Mass spectrometry data acquisition

Samples were analyzed using an Orbitrap Exploris 480 mass spectrometer (Thermo Scientific) coupled with an UltiMate 3000 RSLC Nano LC system (Dionex), Nanospray Flex ion source (Thermo Scientific), and column oven heater (Sonation). pMHC samples were resuspended in 5 µL of 3% acetonitrile, 0.1% formic acid, and loaded onto a 10–12 cm analytical capillary chromatography column with an integrated electrospray tip (~1 µm orifice), prepared and packed in house (50 µm ID and 1.9 µM C18 beads, ReproSil-Pur) through WPS-3000 autosampler (Dionex).

Survey analyses: Peptides were eluted using a gradient with 6–25% buffer B (70% acetonitrile, 0.1% formic acid) for 53 min, 25–45% for 12 min, 45–97% for 3 min, and 97–3% for 1 min. Standard mass spectrometry parameters were as follows: spray voltage, 2.5 kV; no sheath or auxiliary gas flow; heated capillary temperature, 280°C. The Exploris was operated in data-dependent acquisition (DDA) mode with an inclusion list of the 4H trigger peptide. Full scan mass spectra (300–1,500 m/z, 60,000 resolution) were detected in the orbitrap analyzer after accumulation of 3×106 ions (normalized AGC target of 300%) or 50 ms. For every full scan, up to 20 ions were subsequently isolated if the m/z was within ±5 ppm of the targeted 4H trigger peptide and reached a minimum intensity threshold of 1×105. Ions were collected with a maximum injection time of 250 ms, normalized AGC target=1000%, and fragmented by higher energy collisional dissociation (HCD) with a collision energy (CE): 30%. A library of acquired spectra was generated using Skyline software for SureQuant-IsoMHC targeted analyses.

SureQuant-IsoMHC targeted analyses: The custom SureQuant acquisition template available in Thermo Orbitrap Exploris Series 2.0 was used for this method. All acquisition parameters for heavy labeled Alg8 isotopologues are located within a distinct 4-node branch stemming from a full scan node. In the full scan, the trigger 4H peptide m/z and intensity thresholds are defined in the “Targeted Mass” filter node as 1% of the intensity from the DDA survey run. Next, parameters for the low resolution, trigger peptide MS2 scan are defined, followed by the “Targeted Mass Trigger” filter node, which defines the six product ions used for pseudo-spectral matching. To connect each set of product ions within the targeted mass trigger node to a given precursor mass, a group ID feature was used to define the precursor m/z associated with each group of product ions. Finally, the isolation offset (m/z) corresponding to each of the four MS2 scans of the endogenous and 1–3 hours peptides was defined in the scan parameters within each node. Standard mass spectrometry (MS) parameters for SureQuant acquisition were as follows: spray voltage: 2.5 kV, no sheath or auxiliary gas flow, heated capillary temperature: 280°C. Full-scan mass spectra were collected with a scan range: 380–1,200 m/z, AGC target value: 300% (3e6), maximum IT: 50 ms, resolution: 120,000. 4H Alg8 peptide matching the m/z (within 10 ppm) and exceeding the intensity threshold defined on the inclusion list were isolated (isolation window 1 m/z) and fragmented (nCE (normalized collision energy): 30%) by HCD with a scan range: 150–1,200 m/z, maximum injection time: automatically determined from the resolution, AGC (Automatic gain control) target value: 1,000% (1e6), resolution: 15,000. A product ion trigger filter next performs pseudo-spectral matching, only triggering an MS2 event of the endogenous target peptide and heavy standard peptides at the defined mass offset if n≥3 product ions are detected from the defined list. If triggered, the subsequent light, 1H, 2H, and 3H peptides MS2 scans are initiated at the defined mass offsets. Scan parameters have the same CE, scan range, and AGC target as the heavy trigger peptide, but with a higher maximum injection time and resolution (3H&2H: max IT: 250 ms, resolution, 120,000; 1H&light: max IT: 1 s, resolution 480,000). Triggered MS2 scans are performed in the following order: 3H, 2H, 1H, light. Ions for pseudo-spectral matching ar listed in table 3.

SureQuant-IsoMHC data analysis

Skyline software was used to quantify the abundance of the standard and endogenous peptide. For each sample, the abundance of each Alg8 peptide (1H, 2H, 3H and light) was approximated using an average of the maximum intensity of the top three product ions across the elution chromatogram. The average intensity of the three heavy Alg8 standards was regressed against the amount added in IP to generate a standard curve. The absolute amount of endogenous light Alg8 peptide was calculated using the embedded standard curve in each sample. Copies of Alg8 peptide per cell were calculated using the absolute amount and input cell number in the IP.

Human data analysis

Patient with pan-cancer cohort data sets: Data from a total of 922 patients undergoing ICB treatment for the following cancer entities from publicly available studies were gathered and processed in a harmonized manner: Lung cancer,35–37 colorectal cancer,38 melanoma,39–44 gastric cancer,45 urothelial cancer,46 renal cancer,47 48 multiple tumor entities.49 50

Preprocessing: Patient FASTQ files were obtained from respective publications. The Sarek V.3.1.1 pipeline51 52 was used for harmonized alignment and variant calling to hg38. pVACseq via pVACtools V.3.1.3 and vatools V.5.0.1 was used to perform in silico binding affinity predictions of patient NeoAg. HLA-HD V.1.6.1 was used to call patient HLA types. Pyclone V.0.13.1 was used to perform mutation clonality clustering.

Analysis: NeoAg from pVACseq were classified as strong or weak binders according to a NetMHCpan-4.1 percentile rank of <0.5 and <2, respectively. Mutations and their associated NeoAg were classified as clonal if they were found in the pyclone cluster with the greatest number of mutations for each patient. Immunodominance was measured using the greatest number of NeoAg that were strongly presented by any of a patient’s HLA alleles.

Statistical analysis

Statistical analyses were performed using GraphPad Prism V.10. Data are shown as mean±SEM, unless otherwise indicated. Comparisons between groups were performed using parametric (Student’s t-test, analysis of variance (ANOVA)) or non-parametric tests (Mann-Whitney U test, Kruskal-Wallis test) after testing for normality of data distribution (Shapiro-Wilk test). Tumor outgrowth curves were compared using two-way ANOVA. Correction for multiple comparisons was performed where applicable. P values<0.05 were considered statistically significant (*p<0.05; **p<0.01; ***p<0.001; ****p<0.0001; ns=not significant).

ResultsPeptide:MHC binding characteristics alone fail to predict immunogenicity of co-expressed NeoAg

To study the impact of NeoAg architectures on antitumor immune responses, we developed a preclinical mouse model with defined NeoAg architectures. First, to reduce intra-cell line heterogeneity, a subclone of the KrasG12D Trp53−/− (KP) lung adenocarcinoma cell line53 was generated (KP6S; parental cell line). This cell line was subsequently engineered to express either single or multiple NeoAg linked to a fluorescent protein followed by a genetic barcode (figure 1A, online supplemental figure S1A). Individual or mixes of NeoAg-expressing KP6S derivative cell lines were used to model tumors with clonal and subclonal NeoAg expression, respectively (figure 1A). All engineered NeoAg were previously described, endogenous, naturally MHC-presented NeoAg13 20 22 23 (figure 1B) and expressed at similar levels (online supplemental figure S1B). Predicted24 and observed pMHC binding affinities correlated with in vivo immunogenicity of individually expressed NeoAg, as assessed by IFN-γ ELISpot assays and tumor outgrowth experiments (figure 1B–D, (online supplemental figure S1C,D). We categorized NeoAg as strong (highly immunogenic, NeoAg S1-4) and weak (poorly immunogenic, NeoAg W1-2), when their expression respectively did or did not improve tumor control compared with the parental cell line (figure 1C,D, online supplemental figure S1E). CD8+ T-cell depletion resulted in a loss of NeoAg expression-induced tumor control (online supplemental figure S1E), indicating that tumor control was predominantly mediated by NeoAg-specific CD8+ T cells.

Figure 1

pMHC binding characteristics alone fail to predict immunogenicity of co-expressed NeoAg. (A) Transplantable KP lung adenocarcinoma mouse model, genetically engineered to express single or multiple NeoAg and injected subcutaneously as tumors with clonal or subclonal NeoAg expression. (B) Highly and poorly immunogenic (strong and weak) NeoAg, peptide sequences and pMHC binding predictions using NetMHCpan-4.1. #Cpne1 is predicted to also bind H2-Kb at a lower affinity (IC50=170.9 nM) (C–D) Tumor outgrowth of (C) strong and (D) weak NeoAg-expressing tumors compared with parental cell line tumors. (E–F) IFN-γ ELISpot counts on day 10 after tumor injection for (E) weak and (F) strong NeoAg expressed alone or clonally co-expressed with a (E) strong or (F) weak NeoAg. (G–H) IFN-γ ELISpot counts on day 10 after tumor injection for (G) weak and (H) strong NeoAg expressed alone or subclonally co-expressed with a (G) strong or (H) weak NeoAg. (I–J) IFN-γ ELISpot counts on day 10 after tumor injection for strong NeoAg expressed alone or clonally co-expressed with a second strong NeoAg restricted to (I) the same or (J) a different MHC allele. (Data for single NeoAg S1 (Adpgk) tumors in (I) and (J) partially from overlapping experiments.) (K) IFN-γ ELISpot counts on day 10 after tumor injection for strong NeoAg expressed alone, or subclonally co-expressed with a second strong NeoAg restricted to the same MHC allele. (Data for single NeoAg S1 (Adpgk) and S2 (Spb2) tumors partially from overlapping experiments with (H)). (C–D) Representative display of ≥2 independent experiments (n≥3 per group). Data are represented as mean±SEM. (E–K) Pooled data from ≥2 independent experiments (n≥3 per group). Results are expressed as mean±SEM. (C–D) Two-way analysis of variance. (E–K) Two-tailed Student’s t-test. ns, not significant; *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001. See also online supplemental figure S1. ELISpot, Enzyme-linked immunosorbent spot; IFN, interferon; MHC, major histocompatibility complex; NeoAg, neoantigens; pMHC, peptide:MHC.

Previous work by our group using a reductionist model to compare the clonal or subclonal expression of two NeoAg (S1 (Adpgk) and W1 (Aatf)) demonstrated that clonal NeoAg expression resulted in enhanced antitumor T-cell immunity.8 To decipher the dynamic interplay between concurrent T-cell responses and its mechanistic underpinnings systematically, we first used this reductionist system and co-expressed pairs of NeoAg either clonally or subclonally (figure 1E–K). These paired NeoAg models uncovered two consistent patterns of interplay between concurrent NeoAg-specific CD8+ T-cell responses, indicating that the antigenic context contributes to define NeoAg immunogenicity. First, a mutually beneficial interplay was observed between responses against co-expressed weak and strong NeoAg and depended on clonal expression (NeoAg synergy, figure 1E–H). And second, an establishment of immunodominance hierarchies with dominant and subdominant responses was observed between co-expressed strong NeoAg. This observation depended on clonal NeoAg expression as well as restriction to the same MHC allele (NeoAg competition, figure 1I-K).

Clonal expression of weak and strong NeoAg enhances CD8+ T-cell responses against weak NeoAg

Stronger T-cell responses against weak NeoAg were consistently observed when these were clonally expressed with a strong NeoAg. This NeoAg synergy pattern was observed across multiple independent combinations of weak and strong NeoAg (figure 1E). In the same setting, T-cell responses against the strong NeoAg benefited to a lesser extent (figure 1F). Co-expression of two weak NeoAg on the other hand did not induce synergistic effects, suggesting that one potent NeoAg response was required to initiate the observed effects. In contrast, we observed a diminished W1 response in clonal W1-W2 tumors, suggesting competition between the two “weak” T-cell responses (online supplemental figure S1F) in the absence of a strong NeoAg. Notably, NeoAg synergy depended on clonal NeoAg expression and was absent in subclonal expression of the same NeoAg pairs (figure 1G,H). Moreover, the lower NeoAg density in tumors with subclonal NeoAg expression (clonal fractions of 50%) was associated with overall weaker responses against all NeoAg (figure 1G–H, online supplemental figure S1G).

Clonal expression of strong NeoAg induces immunodominance in an MHC allele-dependent manner

Contrasting the synergistic effects mutually benefiting weak and strong NeoAg responses when expressed clonally, we observed an establishment of immunodominance hierarchies between concurrent T-cell responses against two strong, clonal NeoAg. Subdominance in these hierarchies weakened respective immune responses, whereas the dominant response remained unaffected. Competition between T-cell responses was thereby dependent on clonal expression and restriction to the same MHC allele (figure 1I–K). In line with previous reports,7 54 immunodominance was established by the NeoAg with the most favorable pMHC binding characteristics (online supplemental figure S1D).

The consistent patterns of interplay observed between concurrent T-cell responses and their dependence on NeoAg clonality indicated that NeoAg expression patterns in the tumor directly impact the strength of respective T-cell responses.

Synergy between CD8+ T-cell responses enhances T-cell expansion and functionality

Next, we used the two NeoAg model systems to study NeoAg synergy in detail. To this end, we profiled CD8+ T-cell responses against the strong NeoAg S1 (Adpgk) and the weak NeoAg W2 (Cpne1), when these were expressed either alone (strong NeoAg and weak NeoAg tumors, respectively) or in clonal, synergistic combination (NeoAg synergy tumors). Although expression of the weak NeoAg alone did not improve tumor control in outgrowth studies, NeoAg synergy tumors expressing both NeoAg were better controlled than strong NeoAg tumors (figure 2A). To resolve which NeoAg-specific response was mediating this effect, we performed adoptive CD8+ T-cell transfer (ACT) experiments in Rag2−/− mice bearing single NeoAg-expressing tumors on opposite flanks. ACT from donors injected with NeoAg synergy tumors slowed the outgrowth of weak NeoAg tumors in Rag2−/− mice, whereas ACT from donors injected with weak NeoAg tumors did not (figure 2B, left panel). This suggested that the establishment of a productive T-cell response against the weak NeoAg depended on NeoAg synergy. In contrast, ACT was only minimally more efficacious in controlling strong NeoAg tumors in Rag2−/− mice when donors were injected with NeoAg synergy rather than strong NeoAg tumors (figure 2B, right panel). Augmented tumor control in the context of ACT from NeoAg synergy tumor-bearing donors was thus notable particularly for the immune response against the weak NeoAg (figure 2B).

Figure 2

Synergy between CD8+ T-cell responses enhances T-cell expansion and functionality. (A) Tumor outgrowth of the parental cell line, single NeoAg and NeoAg synergy tumors. (B) Tumor outgrowth in Rag2−/− mice following ACT on day 4 from naïve donors or from donors bearing single NeoAg or NeoAg synergy tumors. (C) Flow cytometry-based assessment of NeoAg-specific CD8+ T-cell expansion in the tdLN of mice injected with single NeoAg or NeoAg synergy tumors. (D) Flow cytometry-based assessment of (left panel) NeoAg-specific and (right panel) overall CD8+ T-cell tumor infiltration in mice injected with single NeoAg or NeoAg synergy tumors. (E) In vivo killing capacity of mice bearing single NeoAg or NeoAg synergy tumors on day 10 after tumor injection. (F) Flow cytometry-based analysis of GzmB expression of S1 (Adpgk)-specific and W2 (Cpne1) -specific CD8+ T cells in the tdLN in mice injected with single NeoAg or NeoAg synergy tumors. (G) Flow cytometry-based assessment of tumor debris uptake of cDC1 on day 7 after injection of tumors with clonal and subclonal NeoAg expression. (H–I) Flow cytometry-based assessment of expression of (H) co-stimulatory and (I) inhibitory markers of cDC1 engulfing debris from the parental line, single NeoAg or NeoAg synergy tumors on day 4 after tumor injection. (A–B) Representative display of ≥2 independent experiments. Data are represented as mean±SEM. (C–I) Pooled data from ≥2 independent experiments (n≥3 per group). Results are expressed as mean±SEM. (A–B) Two-way ANOVA. (C,F,G) Two-tailed Student’s t-test. (D left panel), (E) Mann-Whitney U test. (D right panel), (H–I) One-way ANOVA. ns, not significant; *p<0.05; **p<0.01; ***p<0.001; ****p<0.0001. See also online supplemental figure S2 and S3. ANOVA, analysis of variance; cDC1, cross-presenting dendritic cells; GzmB, granzyme B; MFI, Median fluorescence intensity; NeoAg, neoantigens; PD-L1, programmed death-ligand 1; tdLN, tumor-draining lymph node.

Analysis of CD8+ T-cell expansion kinetics in the tdLN revealed that NeoAg synergy induced a greater expansion of both NeoAg-specific T-cell responses, particularly at early time points (figure 2C), with similar kinetics observed in the spleen (online supplemental figure S2A). In line with the greater T-cell expansion in the tdLN, NeoAg synergy induced a greater infiltration of NeoAg-specific and overall CD8+ T cells into the tumor (figure 2D). Functionally, in vivo cytotoxicity assays showed a significant increase in killing capacity for both T-cell responses when mice were bearing NeoAg synergy tumors (figure 2E, online supplemental figure S2B) compared with weak NeoAg and strong NeoAg tumors. Phenotypically, NeoAg-specific T cells showed higher expression of effector molecule granzyme B (GzmB) in NeoAg synergy tumors compared with respective single NeoAg-expressing tumor controls (figure 2F). Generally, the response to the weak NeoAg showed a stronger relative benefit from NeoAg synergy than the response to the strong NeoAg (online supplemental figure S2C).

As NeoAg synergy resembled CD4+ T cell-mediated help,55 we next compared the effects of NeoAg synergy to canonical CD4+ T-cell help using the naturally presented, MHC-II-restricted NeoAg integrin ß1N710Y (Intb1).20 IFN-γ ELISpot assays showed that CD4+ T-cell help similarly induced a greater expansion of the T-cell response against the weak NeoAg (online supplemental figure S2D). Notably, CD4+ T cell-mediated help depended on clonal NeoAg expression in the tumor and was absent in the case of subclonal NeoAg expression (online supplemental figure S2E), mirroring the observations made for synergistic, MHC-I-restricted NeoAg (figure 1E–H, figure 2A–F).

Synergistic effects are induced by highly stimulatory cDC1 in the tdLN mirroring NeoAg expression patterns of the tumor

Because NeoAg synergy was observed at early time points (figure 2C) and in circulation (figure 2E), we hypothesized that it could be induced during T-cell priming in the tdLN. We therefore analyzed NeoAg presentation patterns of cDC1 in the tdLN in tumors with clonal and subclonal NeoAg expression (online supplemental figure S2F). To this end, we generated cell lines in which each NeoAg was fused to a separate fluorophore, enabling the use of acquired fluorescence as a surrogate for the engulfment of NeoAg-containing tumor debris, and ultimately NeoAg cross-presentation. As previously reported by our group,8 cDC1s positive for both fluorescent proteins, a surrogate for simultaneous engulfment and processing of both NeoAg, were predominantly observed in tumors with clonal NeoAg expression and almost absent in the case of subclonal NeoAg expression (figure 2G). This suggested that cDC1 engulf tumor debris from a limited number of tumor cells before migrating to the tdLN, thus retaining the NeoAg expression pattern observed in the tumor. This further suggested that concurrent priming of different NeoAg-specific T-cell populations by the same cDC1 is limited to clonal NeoAg, potentially explaining observed synergistic effects. Phenotypic analysis of tumor-debris engulfing cDC1 revealed that simultaneous processing and presentation of synergistic NeoAg, but not of single NeoAg was associated with a highly stimulatory phenotype. When compared with cDC1 engulfing debris from the parental tumor cell line, higher surface expression of co-stimulatory molecules CD40, CD80 and CD86 and lower expression of immunoregulatory molecule PD-L1 was observed (figure 2H,I, online supplemental figure S3A). Of note, we previously demonstrated that this observation similarly applied to cDC1 co-presenting synergistic antigens derived from distinct subclones in genetically heterogenous tumors.8 In line with the earlier expansion observed for T-cell responses in NeoAg synergy, these differences in cDC1 phenotype were observed particularly at early time points (day 4 after tumor injection, figure 2H,I) and were less pronounced at later time points (day 7 after tumor injection, online supplemental figure S3B,C). Global C