Neoantigen-specific cytotoxic Tr1 CD4 T cells suppress cancer immunotherapy – Nature

Mice

WT male 129S6 mice (for experiments involving T3 and F244 cells) and female C57BL/6 mice (for experiments involving 1956 cells) were purchased from Taconic Farms and Charles River, respectively. Gzmb^−/− on the C57BL/6 background and Rag2^−/− mice (on the 129S6 and C57BL/6 backgrounds) were bred and housed in our specific pathogen-free facility. cDC2-deficient (Δ1 + 2 + 3)^44,45,46 and cDC1-deficient (IRF8Δ32)^44,45,46 mice on the C57BL/6 background were obtained from K. Murphy (Washington University in St. Louis). Lilrb4a (encoding gp49b)-knockout mice were obtained from M. Colonna (Washington University in St. Louis) and were generated by the Genome Engineering and iPSC Center (GEiC) and the Transgenic, Knockout, and Micro-Injection Core, Department of Pathology and Immunology, at Washington University in St. Louis. A single guide RNA (cctttagttgcagctcatccata) and Cas9 proteins were electroporated into zygotes before being transferred into pseudo-pregnant B6/J female mice. The resulting founders were screened using next-generation sequencing (NGS). A male founder with a 21-nucleotide deletion in exon 4 was bred with B6/J female mice obtained from the Jackson Laboratory (JAX stock #000664) for one generation, and heterozygous F1 mice were intercrossed to generate knockouts. The integrity of the highly homologous gene Lilrb4b was confirmed by NGS. The absence of LILRB4A protein expression in blood monocytes and neutrophils was confirmed by flow cytometry using the H1.1 antibody⁵⁷ clone (BioLegend). Lineage tracing eGFP mice on the C57BL/6 background were obtained from J. Kipnis (Washington University in St. Louis). Mice were used between 8 and 10 weeks of age. All in vivo experiments were performed in our specific pathogen-free facility using procedures approved by the AAALAC-accredited Animal Studies Committee of Washington University in St Louis and followed all relevant ethical regulations.

Tumour transplantation

T3, F244 and 1956 are MCA-induced sarcoma cells that were previously generated in our laboratory in male 129S6 (T3 and F244 cell lines) and female C57BL/6 mice (1956 cell line). Tumour cells were cultured in RPMI medium (Hyclone) supplemented with 10% FCS (Hyclone). All cell lines used were negative for mycoplasma and other infectious agents. For tumour inoculation, tumour cells were harvested by trypsinization, washed three times with PBS and resuspended in PBS at a density of 10⁷ cells per millilitre. Then, 150 μl was injected subcutaneously into the rear flanks of syngeneic recipient mice. Tumour growth in individual mice was determined using callipers and expressed as the average of two perpendicular diameters. Mice were euthanized when the tumour diameter reached 20 mm in any direction.

NeoAgs and vaccination protocol

T3 neoAgs were previously described^8,19,20 and are listed in Supplementary Table 1. The peptide of HPV (DKCLKFYSKISEYRHY CYSLYGTTL) was used as an irrelevant non-sarcomas antigen. All peptides were purchased from Peptide 2.0 or Genscript with a specified purity of over 95%. Endotoxin levels in all peptides tested were below 0.5 EU ml⁻¹ (Leinco). The indicated doses of peptide were mixed with 50 μg of polyinosinic–polycytidylic acid complexed with poly-l-lysine and carboxymethylcellulose (poly-ICLC; Oncovir) diluted in PBS. For therapeutic vaccination, tumour-bearing mice were injected intravenously or subcutaneously with the vaccines (200 μl) on days 6 and 17 post-tumour transplant. In experiments using the combination of vaccination and ICT, vaccines were administered intravenously on days 3 or 6 post-tumour transplantation. In the sorting experiments, 25 μg of anti-CD40 was mixed with the second dose of the vaccine to increase the yield of the antigen-specific CD4⁺ T cells.

In vivo antibody treatment

For in vivo treatment, LPS-free and pathogen-free anti-PD1 (rat IgG2a, clone RMP1-14), anti-CTLA4 (mouse IgG2b, clone 9D9) or anti-4-1BB (rat IgG2a, clone 3H3) monoclonal antibodies were purchased from Leinco Technologies. Tumour-bearing mice were vaccinated on day 3 post-tumour implantation and then injected intraperitoneally with 200 μg of each antibody on days 3, 6 and 9. To assess the role of IL-10 in HDVax-induced inhibition, monoclonal antibodies to IL-10 (rat IgG1, clone JES5-2A5) and IL-10R (rat IgG1, clone 1b1.3A) were purchased from Bioxcell and injected with HDVax. For anti-IL-10 blockade, mice were treated with 250 μg 1 day before tumour implantation, followed by treatment every 3 days for the duration of the experiment. For anti-IL-10R blockade, mice were treated with 1 mg per mouse every 7 days, starting 1 day before tumour injection. Neutralizing but not depleting hamster monoclonal antibody specific for LILRB4 (clone# 2F1.F9.E6) was provided by N. Sharma and J. Allison (University of Texas (MD Anderson Cancer Center)), and mice were treated with 250 μg on days 3, 6 and 9 post-tumour inoculation.

Tetramer staining

Tetramer staining for mLama4-specific CD8⁺ T cells and mItgb1-specific CD4⁺ T cells was performed as previously described^8,21. TILs or splenocytes (1–4 million) in 50 μl FACS buffer were incubated with PE/APC-labelled mLama4 peptide–H-2K^b tetramer at 37 °C for 20 min. Then, an antibody master mix consisting of CD90.2, CD4, CD11b and live/dead dye (NIR) in 50 μl FACS buffer was added to each well and incubated for 30 min at 4 °C. Tetramers were obtained from the Bursky Center Immunomonitoring Laboratory at Washington University School of Medicine in St. Louis.

Flow cytometry antibodies

Flow antibodies were: IFNγ (XMG1.2; 1:100; 505808), TNF (MP6-XT22; 1:100; 506323), CD200 (OX-90; 1:200; 123820), PD1 (29F.1A12; 1:200; 135228), CCL5 (2E9/CCL5; 1:500; 149106), GZMB (QA16A02; 1:50; 372216), TIM3 (RMT3-23; 1:200; 119723), CD25 (PC61; 1:100; 102036), CD154 (MR1; 1:100; 106506), CD152 (UC10-4B9; 1:100; 106318), CD4 (RM4-5; 1:500), IL-2 (Jes6-5H4; 1:50; 503826), LILRB4 (H1.1; 1:200; 144904), CD11b (M1/70; 1:800; 101226), XCR1 (ZET; 1:100; 148206), MHC-II (M5/114.15.2; 1:1000; 107641), CD11c (N418; 1:200; 117336), CD172a (P84; 1:500; 144008), Zombie NIR fixable viability dye (1:500; 423106), CD86 (GL-1; 1:200; 105042), CD80 (16-10A1; 1:200; 104712) and CD40 (FGK45; 1:100; 157506) all from BioLegend; CD90.2 (53-2.1; 1:800; 565257), CD45 (HI30; 1:800; 563791), CD39 (Y23-1185; 1:400; 567105), CD153 (RM153; 1:200; 740751), CD70 (FR70; 1:100; 740741) and CD8 (53.6.7; 1:200; 564920) from BD Biosciences; or FOXP3 (Fjk-16a; 1:50; 11-5773-82) and SEMA4a (5E3; 1:50; 46-9753-41) from eBioSciences. Foxp3/Transcription factor Staining kit (00-5523-00, Thermo Fisher) was used to stain FOXP3 and other intracellular proteins according to the manufacturing protocol. BD FACSDIVA software V9.1 on Symphony 3, LSRFortessaX20 or ARIA-II was used for collecting the flow cytometry data and for sorting, respectively. Data were analysed using FlowJo software v10.10.

Multiplex assay and antibody ELISA

TILs were harvested 7 days post-vaccination, stained with mItgb1–I-A^b tetramer prepared by the Immunomonitoring Laboratory core at Washington University in St. Louis using mItgb1 peptide covalently attached to the I-A^b β-chain, and total mItgb1-specific CD4⁺ T cells (Fig. 2c) or cells populating different clusters (Fig. 3c) were sorted by flow cytometry. Sorted cells (n = 200,000) were stimulated in a serum-free medium with 10⁶ irradiated splenocytes (isolated from naive mice) pulsed with 1 μg ml⁻¹ mItgb1 SLP. Following a 72-h incubation, secretion of multiple cytokines was measured using a flow-based customized ProcartaPlex cytokine panel (Luminex Technologies) following the manufacturer’s protocol.

To determine antibody responses to the mItgb1 peptide, ELISA plates were coated with 10 μg ml⁻¹ of either mItgb1 or irrelevant SLP at 4 °C overnight. Plates were washed and blocked with 4% goat serum for 2 h at room temperature, and then different dilutions of serum from LDVax or HDVax mice were added for 2–6 h. Plates were washed again and incubated with peroxidase-conjugated goat antibody to total mouse IgG (H + L) (115-035-003, Jackson ImmunoResearch) or different isotype-specific secondary reagents conjugated with horseradish peroxidase for 2 h at room temperature, followed by sequential addition of tetramethylbenzidine substrate (TMB). The reaction was stopped by acidification, and the optical density of each well was measured at 405 nm.

Adoptive transfer experiments

For adoptive transfer experiments into Rag2^–/– mice, WT mice were vaccinated with LDVax, and 9 days later, 5 million total T cells were transferred into Rag2^−/− mice that were injected with 2 × 10⁵ tumour cells 1 day earlier. To isolate mItgb1-specific CD4⁺ T cells from HDVax mice, TILs and spleens were harvested from HDVax-treated mice, and CD4⁺ T cells were enriched using the CD4⁺ T cell isolation kit from Miltenyi (130-117-043) stained with mItgb1 tetramer, and CD25-negative mItgb1-specific CD4⁺ T cells were purified by flow cytometry. Half a million mItgb1-specific CD4⁺ T cells were injected intratumourally (in 50 μl) into T3 tumour-bearing Rag2^−/− mice.

For sorting and adoptive transfer of individual clusters, CD4⁺ T cells were enriched by positive bead selection and stained with mItgb1–I-Ab⁺ tetramers, CD25, CD200, CD153 and CD39. T_reg cells were sorted based on the positive expression of CD25, and non-cluster 3/5 cells were defined by the negative expression of CD25 and positive expression of CD200 and/or CD153. Cluster 3 cells were sorted based on the negative expression of CD25, CD200 and CD153 and the positive expression of CD39. An equal number of cells (2 × 10⁵ cells) sorted from these clusters were resuspended in 50 μl PBS and were injected intratumourally. To sort for LILRB4-expressing and SEMA4-expressing cells, CD25, CD200 and CD39 were used to pregate on cluster 3 cells, and 1 × 10⁵ cells were injected into T3 tumour-bearing Rag2^−/− mice that also received T_reg cell-depleted total T cells from LDVax mice.

Intracellular cytokine and CD40L staining

Splenocytes (10⁵) harvested from naive mice were irradiated (30 Gy) and pulsed with 1 μg ml⁻¹ peptide, and TILs (1–2 × 10⁶) were subsequently added, and the cell suspension was incubated at 37 °C. GolgiPlug (BD Biosciences) was added 1 h later and incubated for another 4 h. Cells were stained for different surface markers, including the live/dead marker (NIR), then permeabilized using the intracellular permeabilization kit (BD Biosciences), followed by staining for IFNγ, TNF and CD40L (clone; SA047C3; 157010, BioLegend).

Ex vivo antigen presentation assay

Naive WT mice were intravenously vaccinated with HDVax consisting of 15 μg of mItgb1 SLP and 50 μg poly-ICLC (HDVax). At the specified time point, spleens and lymph nodes were harvested and digested in the presence of collagenase. APCs were FACS sorted into four different subpopulations: cDC1 (MHCII⁺CD11c⁺XCR1⁺), cDC2 (MHCII⁺CD11c⁺CD172a⁺) MHCII⁺CD11c⁻ and MHCII⁻CD11c⁻. Sorted APCs (5 × 10³) were incubated with 5 × 10⁴ mItgb1-specific hybridoma cells⁸ in IL-2 precoated ELISPOT plates (IMMUNOSPOT). Twenty-four hours later, plates were developed following the manufacturer’s protocol, and IL-2 spots were quantified using a CTL ImmunoSpot S6 machine.

In vitro dendritic cell killing assay

To isolate dendritic cells, spleens were harvested from naive WT mice and disrupted by collagenase digestion. CD11c⁺ cells were enriched using a positive selection kit (130-125-835, Miltney Biotec). Enriched CD11c⁺ cells were pulsed with mItgb1 SLP at 1 μg ml⁻¹ and incubated with 1 × 10⁴ tumour-specific CD4⁺ T cells at a 1:1 ratio for 12 h. Cells were washed twice with PBS, stained with MHC-II, CD11c, XCR1, CD172a and Zombie NIR dye to determine the frequency of NIR in cDC1s and cDC2s using flow cytometry.

CD8–IL-2

CD8–IL-2 was provided by Asher Biotherapeutics⁴¹. CD8–IL-2 was generated via fusion of an IL-2 mutein that does not bind to IL-2Rα and displays significantly reduced binding to IL-2Rβγ. This mutein was then coupled to a monovalent mouserized anti-mouse CD8 antibody. Monovalent IL-2 coupling was achieved using bispecific charge pair technology, and FcγR binding was abolished via mutating the FcγR binding region of an anti-CD8 antibody. CD8–IL-2 was expressed in HEK293 cells and purified via protein A affinity chromatography, followed by ion-exchange chromatography and then size-exclusion chromatography. Therapeutic doses (1 mg kg⁻¹) of CD8–IL-2 induced significant antitumour activity (approximately 80% response rate) against 8-day-established T3 tumours when they became insensitive to anti-PD1 therapy. For the current study, CD8–IL-2 was administered at a subtherapeutic dose of 0.3 mg kg⁻¹ diluted in PBS and injected intraperitoneally. WT IL-2 (202-IL/CF) was purchased from (R&D Systems). Each mouse was treated with 25,000 IU in PBS injected intraperitoneally daily for 5 days.

scRNA-seq analysis

UMAP clustering and separation of total and antigen-specific cells

T3 tumour-bearing mice were treated with HDVax, LDVax or PBS 6 days post-tumour transplantation. Seven days later, single-cell suspensions were prepared from TILs (pooled from seven mice for each group). Total CD4⁺ T cells were enriched using a CD4⁺ T cell-positive selection kit (Miltenyi). Enriched CD4⁺ T cells were split into two fractions. One fraction was labelled with TCRβ (TotalSeq-C0120 anti-mouse TCRβ chain antibody, 109259, BioLegend) and used as a source of total CD4⁺ T cells. The other fraction was labelled with CD90.2 (TotalSeq-C0075 anti-mouse CD90.2 antibody, 105353, BioLegend) and used to isolate mItgb1-specific cells based on mItgb1–I-A^b tetramer staining and flow sorting. The two fractions (total and antigen-specific CD4⁺ T cells) were mixed at a 1:1 ratio and submitted to the Genome Institute at Washington University to generate 10X libraries using 10X 5′ v2 single-cell RNA-seq. Alignment, barcode assignment and unique molecular identifier counting with Cell Ranger (v6.1.1) were used to prepare count matrices for the gene expression library using the mouse genome (GRCm38) as a reference⁵⁸. Barcodes in all samples representing low-quality cells were filtered out using the standard knee-inflection strategy available in Cell Ranger. For downstream analysis, the Seurat package (v4.0.4) was used; genes expressed in fewer than three cells were additionally filtered out from expression matrices, and cells that contained fewer than 200 expressed genes were removed. The mitochondrial gene fraction was calculated for all cells, and cells with a mitochondrial fraction more than the highest confidence interval for scaled mitochondrial percentage were filtered out. Each sample was normalized using the SCTransform function with mitochondrial content as a variable to regress out in a second non-regularized linear regression. For integration purposes, variable genes across the samples were identified by the SelectIntegrationFeatures function with a number of features equal to 2,000. Then, the object was prepared for integration (PrepSCTIntegration function), the anchors were found (FindIntegrationAnchors function) and the samples were integrated into the whole object (IntegrateData function)⁵⁹. Principal component analysis (PCA) was used for dimensionality reduction, and the first 20 principal components were used further to generate UMAP dimensionality reduction by the RunUMAP function. The clustering procedure was performed by FindNeighbors and FindClusters with a range of resolutions (from 0.2 to 1.0, with 0.2 as a step) and the first 20 principal components as input. Antibody-derived tags (ADT) data were normalized by a centred log-ratio transformation method, scaled and transformed to its own PCA space. Finally, cells were defined as mItgb1-specific cells and total cells (predominantly mItgb1-nonspecific cells) based on the scaled value of corresponding ADTs. Each object was iteratively cleaned for doublets as well as low-quality cells.

TCR data analyses

TCR data were aligned to the reference mouse genome GRCm38 and counted using Cell Ranger (v6.1.1) vdj workflow⁵⁸. The Seurat object was converted to the h5ad format, and cells that passed quality control for gene expression and TCR levels were integrated into the objects (mItgb1-specific cells and total cells). Downstream analyses, including clonotype expansion analysis and estimation of used VDJ gene pairings, were executed using the scirpy toolkit. Proportions of shared and unique clonotypes with CD4⁺FOXP3⁺ T_reg cells (cluster 5) were counted. TCR analyses were done using the Immunarch package.

Bulk RNA-seq

Three cohorts (n = 5) of T3 tumour-bearing mice were vaccinated with HDVax, and 7 days later CD4⁺ T cells were enriched from TILs using the Miltenyi mouse CD4⁺ T cells isolation kit, stained with the mItgb1–I-A^b tetramer and then sorted into three replicates of CD25⁺ (T_reg cells), CD25⁻CD39⁻CD200⁺ (non-T_reg cells/non-cluster 3) and CD25⁺CD200⁻CD39⁺ (cluster 3). Cells were lysed, and mRNA was extracted using Nucleospin RNA Plus (740990.50, TaKaRa). RNA was sequenced by the Washington University Genome Institute. In brief, total RNA integrity was determined using Agilent Bioanalyzer. Library preparation was performed with 10 ng of total RNA with a Bioanalyzer RNA integrity number (RIN) score greater than 8.0. double-stranded DNA (ds-cDNA) was prepared using the SMARTer Ultra Low RNA kit for Illumina sequencing (740990.50, TaKaRa) per the manufacturer’s protocol. cDNA was fragmented using a Covaris E220 sonicator using peak incident power of 18, duty factor of 20%, and cycles per burst of 50 for 120 s. cDNA was blunt-ended, had an A base added to the 3′ ends, and then had Illumina sequencing adapters ligated to the ends. Ligated fragments were then amplified for 12–14 cycles using primers incorporating unique dual index tags. Fragments were sequenced on an Illumina NovaSeq-6000 using paired-end reads extending 150 bases. For analysis, raw data were aligned and counted to the GRCm38.101 reference genome by the Dragen workflow of the GTAC@MGI service. For downstream analysis, 12,000 genes with the highest number of counts across the samples were selected. All samples were normalized by regularized log transformation (rlogTransformation function) and variance stabilizing transformation (vst function), which are accomplished in the DESeq2 (v1.30.1) package. PCA was run based on the vst output, and no outliers were identified. log₂ fold change values were shrunk for all comparisons (CD25 versus CD39, and CD200 versus CD39) by the lfcShrink function, and its output was used for volcano plot construction⁶⁰.

Gene signature analysis

To evaluate the significance of enrichment for highly expressed genes within cluster 3, we identified the top 200 differentially expressed genes from a bulk RNA-seq dataset (Supplementary Table 2). Subsequently, we conducted a rigorous analysis of differential gene expression using the limma package on normalized count data obtained from publicly available datasets. Specifically, we performed the following pairwise comparisons.

For the cytotoxic CD4⁺ T cells, data from GSE141540 were analysed, and a comparison of T_H cell-ctx versus T_H cell was performed. For the Tr1 gene signature, GSE158703 and GSE139990 were obtained, and a comparison of Tr1–IL-10^Pos versus T_H17–IL-10^Neg or Tr1 versus naive was performed, respectively.

For the T_reg and T_FH gene signature, GSE68242 or GSE140187 were obtained, and a comparison of FOXP3 versus other or lymph node WT T_FH versus lymph node naive was performed, respectively. For the chronically stimulated exhausted CD4⁺ T cells, data from GSE30431 were analysed, and a comparison of CD4 D30 chronic versus CD4 D30 acute was performed. We used GSEA to ascertain the significance of enrichment for the gene signature associated with cluster 3. This method evaluates whether cluster 3-predefined gene signatures are statistically overrepresented or underrepresented within the differentially expressed gene lists derived from the publicly available bulk RNA-seq datasets for each CD4⁺ T cell subtype.

To determine the frequency of Tr1-like cells in patients with cancer, we first subset the quality-controlled cells from each manuscript down to CD4⁺ T cells. After this, we used AUCell⁶¹ to determine which cells match the human Tr1 gene set. Following this, we determined the proportion of the Tr1 cells within the CD4⁺ T cell population of each patient. We tested for differences in Tr1 cell proportions between responders and non-responders using Student’s t-tests without assuming equal variances. There is a significant difference in the proportion of Tr1 cells between responders and non-responders from the Sade-Feldman et al. dataset (P = 0.002, t = 4.98, d.f. = 6.11). There was no significant difference in the Awad et al. dataset (P = 0.55).

GSEA for the IL-2 pathway

The FindMarkers function was used to find genes specific to both LDVax and HDVax samples. The average log₂ fold change values were used as an input in GSEA implemented in the fgsea R package (v1.16.0). Forty-three genes associated with the IL-2 signalling pathway were used for the gene set enrichment plot⁶². For violin plots, the summary expression of IL-2 signalling genes was calculated using the sctransform (SCT) assay (data slot), scaled and divided by the number of genes in the pathway.

Statistics

GraphPad Prism software (v10.2.2) was used to perform all statistical analyses. Significance was determined using one-way ANOVA with multiple comparisons corrected with Tukey’s method unless otherwise stated in the figure legends.

Reporting summary

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