PD-1 genetic fate mapping uncovers immune cell diversity mediating the efficacy of combined PD-1 blockade and chemotherapy

ABSTRACT

Combined cytotoxic chemotherapy and immune checkpoint inhibition (ICI) improves outcomes in PD-L1–low lung cancer, but transient and broad PD-1 expression across immune cells complicates the understanding of the underlying mechanisms. We generated Pdcd1-CreERT2 fate-mapping mice to trace PD-1–expressing cells via tdTomato during PD-1 blockade. PD-1–fate-mapped lymphocytes downregulated PD-1 in the spleen but largely retained it in tumors, except for NK cells, which lost PD-1 and regained function. Single-cell transcriptional profiling was performed on immune cells in PD-L1–low Lewis lung carcinoma (LLC) treated with cyclophosphamide (CTX) and/or anti–PD-1 antibodies. Anti–PD-1 monotherapy showed limited efficacy, whereas CTX plus anti–PD-1 markedly improved tumor control. Single-cell analysis identified 15 transcriptionally distinct immune clusters with treatment-dependent abundances. Combination therapy expanded cytotoxic CD8 T cells and a dysfunctional Treg cluster, enhancing CTL activity, including PD-1–fate-mapped CD8 T cells expressing Tpex1 markers. Single-cell TCR analysis revealed that clonotypes selectively expanded by combination therapy, mediating potent cytotoxicity against LLC tumors. PD-1 blockade synergizes with cytotoxic chemotherapy to diversify and expand PD-1 lineage–traced CTL clonotypes, driving robust antitumor immunity. Thus, our fate-mapping system is a valuable tool to search for immune cells responsive to ICI therapy.

Introduction

Lung cancer is a leading cause of cancer-related mortality worldwide.¹ Before the advent of targeted therapies for driver mutations and immune checkpoint inhibitors (ICIs) that block programmed death-1 (PD-1), programmed death-ligand 1 (PD-L1), or cytotoxic T-lymphocyte–associated antigen 4 (CTLA-4), cytotoxic chemotherapy was the primary treatment for patients with advanced lung cancer. The introduction of ICIs has dramatically reshaped lung cancer treatment strategy.² While ICI monotherapy provides clinical benefit primarily to a subset of patients with PD-L1–high tumors,^1,3 the majority of patients encountered in clinical practice exhibit low or indeterminate tumor PD-L1 expression, necessitating the use of combination strategies.^4-10

Preclinical murine studies suggest that the enhanced anti-tumor efficacy of combining ICIs with cytotoxic chemotherapy is associated with increased infiltration of CD8 T cells, Th1 cells, M1 macrophages, and dendritic cells (DCs) within the tumor microenvironment.^11-13 The beneficial roles of both CD4 and CD8 T cells in mediating therapeutic efficacy have been highlighted by depletion experiments in murine models of triple-negative breast cancer and renal cell carcinoma.¹⁴ However, the specific clonal diversification of immune cells induced by combination therapy remains unclear. Thus, the precise mechanisms through which ICIs and cytotoxic chemotherapy cooperate to orchestrate the tumor immune network and potentiate anti-tumor immunity are yet to be fully elucidated.

One major challenge in elucidating these mechanisms is the diversity of PD-1–expressing cells. Recent single-cell analysis revealed that PD-1 is expressed on various immune cells, including T cells, B cells, innate lymphoid cells, myeloid-derived suppressor cells (MDSCs), and DCs, where it plays a critical role in maintaining immune tolerance.^15-22 Additionally, the transient nature of PD-1 expression on immune cells makes it complicated to decipher the overall immune responses induced by ICIs. PD-1 is transiently expressed on a subset of thymocytes during T cell development and is downregulated upon their migration to peripheral tissues.²⁰ Therefore, comprehensive analysis of PD-1⁺ cells which are possibly modulated by ICIs have not been conducted.

Lewis lung carcinoma (LLC) is a murine lung cancer with mesenchymal features harboring KRas/NRas mutations, resembling human lung adenocarcinoma.²³ Due to their low expression of MHC class I molecules, LLC cells can evade CTL–mediated killing.²⁴ Similar to many human lung cancers that exhibit minimal PD-L1 expression, LLC cells also express low levels of PD-L1,²⁴ which contributes to their poor responsiveness to PD-1/PD-L1 blockade therapy.^25,26 Therefore, LLC represents a clinically relevant preclinical model for lung cancer in which PD-1 blockade alone has limited efficacy and effective antitumor responses require combination strategies with chemotherapy.

In this study, we generated Pdcd1 fate-mapping (Pdcd1^fm) mice, in which tdTomato serves as a reporter to trace PD-1–expressing cells during ICI therapy in combination with cytotoxic chemotherapy. Although cisplatin is commonly used with ICIs for lung cancer, cisplatin monotherapy is largely ineffective in the LLC model.²⁷ Therefore, we selected cyclophosphamide (CTX) as a cytotoxic backbone to evaluate the combinatorial efficacy with PD-1 blockade, given its potent cytotoxic activity against LLC cells, although it is no longer commonly used in current clinical practice for lung cancer.²⁸ Using the LLC model of combined CTX and ICI therapy, we analyzed tdTomato⁺ immune cells within the tumor by single-cell RNA-seq (scRNA-seq), single-cell TCR-seq (scTCR-seq), and flow cytometry. Combination therapy enhanced efficacy due to high CTL activity compared with CTX or ICI alone and was associated with enrichment of CD8 T cells which expressed progenitor exhausted CD8 T cell (Tpex) markers SLAMF6 and CD69 among tdTomato⁺ cells, whereas CTX primarily expanded tdTomato⁺ myeloid populations such as MDSCs and DCs. Finally, we identified CTL clonotypes uniquely induced by the combination therapy, suggesting that the clonal diversification of tumor-reactive CTLs underlies the enhanced anti-tumor efficacy of the combined regimen.

Materials and methods

Mice

All the mice were bred and maintained in a specific pathogen–free facility at the Akita University Graduate School of Medicine, Akita, Japan. All animal procedures were approved by the Ethics Review Board of Akita University (a-1-0441) and were conducted in accordance with the guidelines of the American Veterinary Medical Association. The mice were euthanized either in a CO₂ chamber with controlled flow rates. Pdcd1^Cre-ERT2 mice were generated from JM8.N4ES cells (MMRRC), whose genomes were manipulated using CRISPR/Cas9 technology by replacing the endogenous stop codon of the Pdcd1 gene with a P2A self-cleaving peptide sequence followed by Cre-ERT2, which was inserted at the 3′ end of the coding region in our animal facility. For fate mapping of PD-1-positive cells, these mice were then crossed with Rosa26-loxP-STOP-loxP-tdTomato mice (#007914, the Jackson Laboratory). The resulting Pdcd1-CreERT2/Rosa26-lsl-tdTomato mice are hereafter designated as Pdcd1^fm mice. This study adhered to the ARRIVE guidelines.

Tumor implant model

Lewis lung carcinoma (LLC, ATCC) cells were cultured in DMEM supplemented with 10% fetal bovine serum (FBS) and penicillin–streptomycin. 6 × 10⁴ LLC cells in 100 μL of PBS was subcutaneously injected into the left flank of female Pdcd1^fm mice aged 7–13 weeks. The tumor size was measured three times weekly, and the tumor volume was calculated as (length × width × height)/2. To induce tdTomato expression, the mice received an intraperitoneal injection of 2 mg tamoxifen (Tront Research Chemicals, cat# T006000) dissolved in 100 μL corn oil on the first day of drug treatment. Tamoxifen was subsequently provided orally by mixing 20 mg of tamoxifen with 50 g of powdered CE-2 chow. On day 12 after implantation, the mice were randomized into treatment groups with comparable tumor volumes. The mice were treated intraperitoneally with CTX (100 mg/kg) and/or an anti–PD-1 antibody clone (RMP1-14, Bio X Cell, 400 μg) on days 0 and 7; rat IgG2a (clone 2A3, Bio X Cell) was used as an isotype control.

Single-cell analysis

The tumors were harvested on day 11 after treatment, excised, and mechanically dissociated. Tumor tissue (0.5 g) was digested for 45 min at 37 °C with agitation in RPMI containing 2% FBS, collagenase IV (C5138; Sigma–Aldrich), and DNase (043-26773; FUJIFILM). The cell suspensions were filtered through a 70-μm strainer, and red blood cells were removed using ACK lysis buffer. The immune cells were then enriched by two rounds of 40% Percoll gradient centrifugation. tdTomato⁺/viability-dye⁻ cells (7500 cells per sample, four samples) were sorted and multiplexed using TotalSeq-A anti-mouse Hashtags A0301–A0304. BD Rhapsody whole-transcriptome and paired TRA/TRB libraries were prepared according to the manufacturer's protocols, quantified, and sequenced on an Illumina NovaSeq 6000 after pool balancing. The reads were processed using the BD Rhapsody pipeline to generate a gene-by-cell matrix (GRCm39, release-107). Hashtag reads were used for sample demultiplexing and multiplet exclusion. TCR reads were processed with STARsolo and MiXCR to assemble TRA/TRB contigs and identify CDR3 sequences. For each cell, the TRA and TRB contigs with the highest UMI support were retained, and clonotypes were defined by the TRB CDR3 amino-acid sequence and V/J usage. Clonotype information was integrated with the gene-expression matrix for downstream single-cell analyses.

Downstream computational analysis

Analyses were performed in R (details in the Supplementary Code). After quality control and hashtag demultiplexing, the data were analyzed using Seurat v5 with SCTransform normalization (v2, default settings). PCA-based nearest-neighbor graphs were used for clustering, and UMAP was used for visualization. Marker genes were identified using Wilcoxon rank-sum tests with FDR correction. T-cell states were annotated by mapping productive TCR-bearing cells to the mouse ProjecTILs reference.²⁹ Clonotype information was integrated by exact barcode matching and summarized by cluster and treatment group.

Flow cytometry

The following antibodies were used for flow cytometry: APC/Cy7 anti-mouse CD45 (30-F11), BV421 anti-mouse CD3 (17A2), AF700 or APC anti-mouse CD4 (GK1.5), APC or PE/Cy7 anti-mouse CD8α (53-6.7), AF700 anti-mouse MHC II (M5/114.15.2), APC or Pacific Blue anti-mouse CD25 (PC61), BV421 anti-mouse CD19 (6D5), APC/Cy7 or FITC anti-mouse CD44 (IM7), FITC anti-mouse PD-1 (29F.1A12), BV421 anti-mouse TCRβ (H57-597), APC anti-mouse CD27 (LG.3A10), APC anti-mouse Ly108 (330-AJ), PE/Cy7 anti-mouse CD69 (H1.2F3), Pacific Blue or PE/Cy7 anti-mouse CD11b (M1/70), APC anti-mouse Ly6G (1A8), APC/Cy7 anti-mouse Ly6C (HK1.4), PE/Cy7 anti-mouse CD11c (N418), and biotin-conjugated anti-mouse TCRγδ (GL3) and NK1.1 (PK136), followed by streptavidin-BV421 or −APC/Cy7. All antibodies and streptavidin reagents were purchased from BioLegend. These cells were stained with a Zombie Aqua Fixable Viability Kit (BioLegend) to exclude dead cells. Data were acquired on a FACSAria (BD Biosciences) and analyzed using FlowJo software.

CTL assay

LLC cells (1 × 10⁴ cells) were plated on 96-well round-bottom plates and co-cultured with 1 × 10⁴ tdTomato-positive CD8 T cells sorted from tumors using a FACSMelody (BD Biosciences) for 48 h. GolgiPlug was added to the media for the last 4 h, and interferon-γ production by T cells was analyzed using flow cytometry. For intracellular IFN-γ staining, the cells were fixed and permeabilized using Cytofix/Cytoperm (BD Biosciences) and stained with a FITC- or APC-conjugated anti–IFN-γ antibody (clone B27, Biolegend). TCR clones were retrovirally transfected into murine CD8 T cells using murine TCR OT-I-2A.pMIG II (Addgene, cat. #52111), as described previously.³⁰ Briefly, Vα sequences (Supplementary Table S1) with overhangs of 5′-CGCCGGAATTCAGATCTACC-3′ and 5′-ATCCAGAACCCAGAACCTGCTG-3′ and Vβ sequences with overhangs of 5′-GGAAGAAAACCCCGGTCCCATG-3′ and 5′-GATCTGAGAAATGTGACTCCACCCAAG-3′ were obtained from IDT. The TCR alpha constant gene was PCR-amplified from the backbone vector using primers 5′-GATCTGAGAAATGTGACTCCACCCAAG-3′ and 5′-GCGGAATTGATCCCGCTCGAG-3′. The Vα sequences were then assembled with the TCR alpha constant gene into the EcoRI/XhoI site using NEBuilder HiFi DNA Assembly Master Mix (NEB). Prior to retroviral transfection, MACS-sorted CD8 T cells (Miltenyi Biotec, cat. #130-104-075) were stimulated with plate-bound anti-CD3/anti-CD28 antibodies for 48 h. After spin inoculation at 37 °C for 2 h, the transfected cells were cultured on an LLC cell layer in RPMI supplemented with 10% FBS and penicillin–streptomycin for an additional 48 h. GFP⁺ CD8 T cells were then analyzed for intracellular IFN-γ.

Statistical analysis

The sample size per group was determined based on prior studies under similar conditions that showed significant differences, with comparable numbers of animals. No animals or data points were excluded from the analysis. No potential confounders were expected to influence the results. No randomization or blinding was performed. The data were analyzed using GraphPad Prism software. Unpaired two-tailed Student's t-test with or without Welch's correction was used for comparisons between two groups. For comparisons involving more than two groups, one-way ANOVA followed by Tukey's multiple comparisons test was performed. p < 0.05 were considered statistically significant.

Results

Pdcd1 fate-mapping mice reveal transient PD-1 expression by lymphocytes

To track PD-1–expressing cells via tdTomato expression, we generated Pdcd1^Cre-ERT2 mice by inserting a P2A sequence followed by Cre-ERT2 at the 3′ end of the Pdcd1 coding sequence, replacing the stop codon (Figure 1A). These mice were then crossed with Rosa26-loxp-stop-loxp-tdTomato (Rosa26-tdTomato) mice to obtain Pdcd1^fm mice. To validate tdTomato expression in PD-1–expressing cells, we orally administered tamoxifen to Pdcd1^fm mice for either 2 or 7 d and examined tdTomato expression in splenocytes. No leaky tdTomato expression was observed in Pdcd1^fm mice without tamoxifen treatment (Figure 1B). On day 2, tdTomato expression correlated with PD-1 expression in splenocytes. On day 7, the tdTomato⁺ cells exhibited lower PD-1 expression than at day 2, indicating that the majority of the tdTomato⁺ cells had lost PD-1 expression in the spleen (Figure 1B–D). These tdTomato⁺ cells included CD4 T cells (71.13%), B cells (9.3%), CD8 T cells (8.0%), γδ T cells (7.7%), and NK cells (0.7%), all of which transiently expressed and subsequently downregulated PD-1 (Figure 1C–E). In contrast, tdTomato expression was scarcely detected in splenic myeloid cells. The proportion of tdTomato⁺ cells was relatively higher among γδ T cells and CD4 T cells compared to other lymphocyte subsets (Figure 1F). Continuous tamoxifen administration in Pdcd1^fm mice for up to four weeks led to a progressive increase in the number of tdTomato⁺ cells across all lymphocyte subsets, indicating ongoing replenishment and prolonged persistence of tdTomato⁺ lymphocytes in the spleen. Thus, Pdcd1^fm mice are a useful tool for tracking cells that transiently express PD-1.

Figure 1.

tdTomato expression marks cells that transiently express PD-1. (A) Schematic figure depicting the genome structure of Pdcd1 fate-mapping (Pdcd1^fm) mice. (B and C) Expression of PD-1 and tdTomato in bulk splenocytes (B) and in the indicated splenic subsets (C) isolated from Pdcd1^fm mice following oral administration of tamoxifen for 2 or 7 d. (D) Quantification of PD-1 downregulation in the tdTomato⁺ cells as measured in C. (E) A pie chart depicting the proportions of splenic subsets in the Pdcd1^fm mice following the oral administration of tamoxifen for 7 d. (F) Frequency of tdTomato⁺ cells in the indicated splenic subsets following the oral administration of tamoxifen for the indicated periods. (G) Frequency of tdTomato⁺ cells in thymocytes isolated from Pdcd1^fm mice administered tamoxifen for 7 d. **p < 0.01 and ***p < 0.01, and ****p < 0.001 as determined by an unpaired two-tailed t-test with/without Welch's correction. The data represent at least two independent experiments (mean ± SEM of three to four mice in D–G).

Since T cells comprised the majority of tdTomato⁺ splenocytes, we next investigated the timing of tdTomato induction during T cell development. To this end, we examined thymocytes from Pdcd1^fm mice treated with tamoxifen for seven days. Thymocyte development proceeds through double-negative (DN), double-positive, and single-positive stages, which are defined by CD4 and CD8 expression. tdTomato expression peaked at the DN4 stage, coinciding with thymocyte proliferation following TCRβ rearrangement, and gradually declined as the cells matured toward the single-positive stage (Figure 1G). Thus, a subset of T cells transiently expresses PD-1 and tdTomato during thymic development and subsequently downregulates PD-1 expression in the periphery.

The tumor microenvironment sustains PD-1 in immune cells except in NK cells

To assess whether tumors induce systemic or local alterations in tdTomato⁺ cells, we subcutaneously inoculated Pdcd1^fm mice with LLC cells, administered tamoxifen for 10 d, and analyzed tdTomato expression in splenocytes and tumor-infiltrating leukocytes (TILs). Overall, tumor burden did not markedly affect the proportions of tdTomato⁺ cell subsets in the spleen (Figure 2A,B; Supplementary Figure S1). In contrast, the frequencies of CD8 T cells, regulatory T (Treg) cells, and DCs among total tdTomato⁺ cells were higher in the tumor than in the spleen (Figure 2A). Notably, tdTomato⁺ B cells were scarcely detected in the tumor.

Figure 2.

Tumor-infiltrated tdTomato⁺ cells except NK cells maintain PD-1 expression. (A) Proportions of cell subsets among total tdTomato⁺ cells in the spleen of Pdcd1^fm mice with or without LLC tumors and within LLC tumors. (B) Frequency of tdTomato⁺ cells among the indicated cell subsets in the spleen of Pdcd1^fm mice with or without LLC tumors and within LLC tumors. (C and D) Expression of PD-1 and tdTomato (C) and quantification of PD-1⁺ cells among all tdTomato⁺ cells (D) in the indicated cell subsets in the spleen of Pdcd1^fm mice with or without LLC tumors and within LLC tumors. (E and F) Expression (E) and quantification (F) of IFN-γ by tdTomato⁺ PD-1^– NK cells and tdTomato⁺ PD-1⁺ NK cells. *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.001 as determined by unpaired two-tailed t-test with/without Welch's correction or one-way ANOVA followed by Tukey's multiple comparisons test. The data represent at least two independent experiments (mean ± SEM of four mice in A, B, D, and F).

We next quantified the frequencies of tdTomato⁺ cells within individual immune cell subsets in the tumor. tdTomato expression was markedly elevated in CD4 T cells, CD8 T cells, NK cells, DCs, gMDSCs, and mMDSCs in the tumor compared with those in the spleen (Figure 2B,C). Although these myeloid populations specifically expressed tdTomato within the tumor, they each represented only a minor fraction of their respective cell subsets (Figure 2B). In contrast, tdTomato expression was downregulated in γδ T cells in the tumor relative to those in the spleen.

To examine whether the tumor burden influences PD-1 expression in tdTomato⁺ cells, we analyzed PD-1 expression in tdTomato⁺ cells from the spleen and tumor. Regardless of tumor presence, tdTomato⁺ lymphocytes—including CD4 T cells, CD8 T cells, B cells, γδ T cells, and NK cells—generally downregulated PD-1 expression in the spleen (Figure 2C,D). Compared with their splenic counterparts, tdTomato⁺ lymphocytes exhibited higher PD-1 expression within the tumor microenvironment. Among these populations, NK cells were unique in undergoing partial PD-1 downregulation; however, despite this reduction, tumor-infiltrating NK cells still expressed higher levels of PD-1 than splenic NK cells (Figure 2C,D).

PD-1 expression within the tumor microenvironment impairs antitumor immune responses. We therefore assessed how PD-1 downregulation influences NK cell function. Notably, tdTomato⁺ PD-1^– NK cells secreted higher levels of IFN-γ than tdTomato⁺ PD-1⁺ NK cells in response to IL-12 and IL-18 (Figure 2E,F), suggesting that PD-1 downregulation may contribute to the sustained activity of NK cells in LLC tumors. These data indicate that, among tdTomato⁺ immune populations, NK cells uniquely retain functional capacity by partially escaping tumor microenvironment-mediated suppression.

Single-cell analysis reveals PD-1⁺ cells affected by cyclophosphamide and/or ICIs

To establish a mouse model for combination therapy with cytotoxic chemotherapy and ICIs, we treated Pdcd1^fm mice bearing subcutaneous LLC tumors with CTX and/or anti-PD-1 antibodies. The dosage and schedule of CTX and anti-PD-1 antibodies were optimized to observe the additive effects of PD-1 blockade in the presence of CTX, as shown in Figure 3A. This dose of anti-PD-1 antibodies alone had a limited effect on tumor growth, probably because LLC cells express low levels of PD-L1.²⁶ Meanwhile, CTX alone significantly inhibited tumor growth. Tamoxifen was administered to induce tdTomato expression in PD-1⁺ cells during treatment, and tdTomato⁺ TILs were analyzed using scRNA-seq/TCR-seq (Figure 3B–F and Supplementary Tables S2 and S3).

Figure 3.

Clonal variation of tdTomato⁺ immune cells in the tumor microenvironment following cyclophosphamide (CTX), anti–PD-1 antibody (ICI), or their combination. (A) Growth of LLC tumors subcutaneously implanted on the backs of Pdcd1^fm mice treated with CTX, ICI, or their combination, along with oral tamoxifen. (B) Uniform manifold approximation and projection (UMAP) visualization of single-cell RNA sequencing of tumor-infiltrated immune cells from tumor-inoculated Pdcd1^fm mice under different treatment conditions. (C) Proportional representation of immune cell clusters among total tdTomato⁺ cells in each treatment group. (D) Proportional representation of treatment groups within individual immune cell clusters. (E) UMAP plots showing the expression patterns of selected genes. (F) Gene expression heatmap of each immune cell clusters. *p< 0.05, **p < 0.01, and ***p < 0.001, as determined by one-way ANOVA followed by Tukey’s multiple comparisons test. The data represent at least two independent experiments (mean ± SEM of five mice in A).

Approximately 5000 cells per experimental group passed quality control and were subjected to uniform manifold approximation and projection (UMAP) for dimensionality reduction. Clustering analysis of a total of 21,078 tdTomato⁺ cells revealed 15 distinct clusters, including four CD4 T cell clusters, a CD8 T cell cluster, a γδ T cell cluster, an NK cell cluster, and three myeloid cell clusters (Figure 3B–D). The cluster-defining genes and selected functional markers are presented in Figure 3E and F, and Supplementary Table S2. Clusters 0 and 4 were identified as regulatory T cell (Treg) subsets, based on high expression of Foxp3, Il2ra, and Areg. Additionally, cluster 0 expressed elevated levels of Tnfrsf18, Tox, and Tbx21, indicating features of Th1-like Treg cells within the tumor microenvironment.³¹ In contrast, cluster 4 represented dysfunctional Treg cells characterized by low expression of Foxp3 and Il2ra, ³¹ and high expression of predicted genes such as Gm3500, Gm3373, Gm8281, and Gm26630. Although these Gm genes are annotated as protein-coding in the NCBI database, they lack identifiable functional domains and human orthologs, suggesting that they may represent long non-coding RNAs (lncRNAs). Cluster 2 displayed a transcriptional profile consistent with that of naïve T cells, marked by high expression of Tcf7, S1pr1, Klf2, and Cxcr4, and low expression of effector-like Th1-associated genes such as Tbx21, Ifng, and Il18r1. Cluster 7 was defined by co-expression of activation markers (Tnfsf11, Tnfsf4), exhaustion markers (Pdcd1, Lag3, Il1r2), and T follicular helper (Tfh)-related genes (Rora, Tnfsf8). The low expression of effector genes (Ifng, Tnf, Il2) in this cluster supports the classification of these cells as exhausted CD4 T cells, which are known to adopt Tfh-like transcriptional programs in chronic virus infections.^32-34 Lineage-specific immune subsets were delineated based on canonical marker expression: Cluster 1 (Cd8b1, CD8 T cells); Cluster 9 (Trgc1, γδ T cells); Cluster 3 (Ncr1, NK cells); Cluster 12 (S100a8, monocytic myeloid-derived suppressor cells, mMDSCs); and Cluster 11 (Mafb, granulocytic MDSCs, gMDSCs). Cluster 14 was annotated as dendritic cells (DCs), based on expression of Ccl22, which is predominantly produced by DCs in the tumor microenvironment.³⁵ In addition, several clusters reflected distinct cellular states. Cluster 6 showed a strong type I interferon-stimulated gene (ISG) signature, with high expression of Mx1, Ifit1, and Ifit3. Proliferative cells were captured in cluster 5, which exhibited elevated expression of Mki67, Cdc13, and Cdca8. Consistent with proliferation, cluster 13 showed increased expression of histone genes. Cluster 10 was defined by the upregulation of genes associated with ribosome modification and biosynthesis, including Fbl, Bkc1, and Nhp2, indicating a transcriptional program linked to cellular proliferation. The sequence data for cluster 8 met basic quality control criteria; however, owing to their relatively lower quality compared with other clusters, they could not be reliably assigned to a specific cell type (Supplementary Figure S2).

To identify cell populations associated with the therapeutic effects of combined CTX and ICI treatment, we compared the proportions of each treatment group within individual clusters. The combined therapy group exhibited the highest representation in cluster 1 CD8 T cells and cluster 4 dysfunctional Treg cells (Figure 3C,D and Supplementary Table S3). Furthermore, CTX treatment was associated with an increased proportion of tdTomato⁺ myeloid cell clusters, including mMDSCs, gMDSCs, and DCs. This shift likely reflects the recruitment of phagocytic myeloid cells to the tumor in response to CTX-induced immunogenic cell death.

ICI therapy increases the CD8 to CD4 T cell ratio within the tdTomato⁺ cell population

To minimize clustering driven by the cell state and to characterize T cells in greater detail, we extracted transcriptomic data from cells with TCR information and analyzed them using ProjecTILs (Figure 4A–D and Supplementary Table S4).²⁹ This analysis revealed that tumor-infiltrating T cells were predominantly Treg cells and Th1 cells, while CD8 T cells were classified into five distinct populations (Figure 4A–C). Across the treatment groups, the CD8/CD4 T cell ratio was higher in the ICI monotherapy group compared with control, and in the CTX + ICI group compared with CTX alone group (Figure 4B). The CTX + ICI group was enriched mostly in early activated CD8 T cell and terminally-exhausted CD8 T cell clusters (Figure 4C). Despite this, clustering analysis showed only subtle differences between the CTX alone and combination therapy groups (Figure 4B,C), suggesting that alterations in the TCR repertoire may underlie the additive effect of ICI when combined with CTX.

Figure 4.

Transcriptomic profiling of tdTomato⁺ T cells in the tumor microenvironment following cyclophosphamide (CTX), anti–PD-1 antibody (ICI), or their combination. (A) UMAP visualization of single-cell RNA sequencing of tumor-infiltrating T cells from tumor inoculated Pdcd1^fm mice under different treatment conditions (left) and a bar graph of the relative frequency of each immune cell subset within the total tdTomato⁺ T cells (right). (B) Proportional representation of T cell subsets among total tdTomato⁺ T cells in each treatment group. (C) Proportional representation of treatment groups within individual immune cell subsets. (D) Dot plots showing the expression levels of specific marker genes in each T cell subset. The size of the dots represents the proportion of cells expressing the marker genes, and the spectrum of color represents the expression levels of the genes.

We next focused on identifying proliferative and ISG⁺ clusters, as shown in Figure 3B. The expression of markers such as Mki67, Mx1, H3C10, and Fbl was evenly distributed across T cell populations (Figure 4D), indicating that all T cell subsets in these specific cell states were classified into clusters 5, 6, 10, and 13 in Figure 3B.

Combination therapy of CTX and ICI induces tdTomato⁺ Tpex1-like CD8 T cells in the tumor

To validate treatment-related shifts in tdTomato⁺ TIL populations suggested by scRNA-seq analysis, we performed flow cytometric profiling of tumors following CTX and/or ICI treatment (Figure 5A‒D). ICI treatment significantly increased the proportion of tdTomato⁺ CD8 T cells among the total number of tdTomato⁺ cells, independent of CTX administration (Figure 5A). To characterize these tdTomato⁺ CD8 T cells in more detail, we next focused on cell surface markers for the Tpex subsets. Tpex cells represent a population of exhausted CD8 T cells that retain proliferative potential and respond to PD-1 blockade therapy. They can be classified into SLAMF6⁺CD69⁺ Tpex1 and SLAMF6⁺CD69⁻ Tpex2 subsets, with Tpex1 being more quiescent and self-renewing than Tpex2.²² While a high frequency of Tpex2-like cells was observed in the ICI group, the CTX + ICI group preferentially increased the proportion of Tpex1-like cells, likely reflecting the cytotoxic effect on the more proliferative Tpex2-like subset (Figure 5B–D). Consequently, tumors treated with the combination of CTX and ICI exhibited the highest frequency of Tpex1-like CD8 T cells.

Figure 5.

Combination therapy with CTX and ICI increased the frequency of Tpex1-like tdTomato⁺ CD8 T cells in the tumor microenvironments. (A–D) Flow cytometric analysis of tdTomato⁺ immune cells in the tumor microenvironment following cyclophosphamide (CTX), anti–PD-1 antibody (ICI), or their combination. (A) Frequency of CD8 T cells among total tdTomato⁺ T cells in each treatment group. (B) Expression of SLAMF6 and CD69 in the tdTomato⁺ CD8 T cells. (C) Frequency of SLAMF6⁺ CD69⁺ CD8 T cells as determined in B. (D) Frequency of the indicated cells among total tdTomato⁺ immune cells in each treatment group. *p < 0.05, **p < 0.01, and ***p < 0.001, as determined by one-way ANOVA followed by Tukey's multiple comparisons test. The data represent at least two independent experiments (mean ± SEM of four to five mice in A–D).

Among myeloid populations, both CTX monotherapy and combination therapy significantly expanded tdTomato⁺ mMDSCs and DCs, which was consistent with single-cell analyses (Figures 3C,D and 5D). DCs are typically categorized into CD11b⁻ CD8α⁺ cDC1, which possess cross-priming capacity for CD8 T cells, and CD11b⁺ CD8α⁻ cDC2, which primarily present antigens to CD4 T cells.³⁶ Based on CD11b and CD8α expression, the tdTomato⁺ DCs were identified as cDC2 (Supplementary Figure S1). In contrast, no significant differences were observed in gMDSCs across treatment groups, despite the single-cell data indicating the highest frequency of the CTX group in the gMDSC cluster (Figures 3C,D and 5D). The frequencies of tdTomato⁺ γδ T cells, CD4⁺ CD25⁺ Treg cells, and NK cells remained largely unchanged following CTX + ICI treatment (Figure 5D). Collectively, these results indicate that the enhanced antitumor efficacy of CTX + ICI therapy is associated with the selective expansion of tdTomato⁺ Tpex1-like CD8 T cells within the tumor microenvironment.

Because tdTomato expression was confined to a small subset of cells, we next examined overall changes in tumor-infiltrating immune cell populations irrespective of tdTomato expression across treatment groups (Supplementary Figure S3). Lymphocyte subsets, including CD8 T cells, Treg cells, γδ T cells, and NK cells, were proportionally reduced following CTX alone or CTX + ICI therapy. In contrast, the frequency of mMDSCs increased under both CTX alone and CTX + ICI conditions. Notably, the proportion of CD8 T cells was markedly reduced in the CTX + ICI group compared with the ICI group, suggesting that the therapeutic outcome may be determined by the specific TCR clonotypes present rather than the overall abundance of CD8 T cells in the tumor microenvironment.

Combined CTX and ICI therapy promotes clonal diversification of CTLs targeting the tumor

The principal effector cells mediating direct tumor cell killing are NK cells and cytotoxic T lymphocytes (CTLs). Because LLC cells exhibit low MHC class I expression, which can trigger NK cell activation through a missing-self mechanism, we evaluated the contribution of NK cells to the additive antitumor effect of ICIs combined with CTX. However, NK cell depletion did not result in significant changes in tumor volume compared with control treatment, indicating that NK cells are dispensable for the efficacy of the combined therapy (Supplementary Figure S4). We then hypothesized that the combination of CTX and ICI therapy would most effectively induce CTLs-targeting LLC tumors. To test this, CD8 T cells were isolated from tumors and cultured on an LLC tumor cell layer for three days, and IFN-γ production was assessed. CD8 T cells derived from tumors treated with the combination therapy secreted the highest levels of IFN-γ among all treatment groups (Figure 6A,B), indicating that the combination therapy efficiently enhanced CTL activation.

Figure 6.

Combination therapy with CTX and ICI specifically induced CTL TCR clonotypes. (A and B) Expression of IFN-γ in tumor-derived CD8 T cells cultured with LLC cells across the indicated treatment groups (A), and quantification of IFN-γ⁺ cells as determined in A (B). (C and D) Heatmap of the top 20 tumor-enriched TCR clonotypes determined by single-cell TCR sequencing across treatment groups (C) and pie charts depicting the distribution of T cell subsets in each indicated TCR clone (D). (E and F) Expression of IFN-γ in CD8 T cells retrovirally transfected with the indicated TCR clonotypes and cultured with LLC cells (E) and quantification of IFN-γ⁺ cells as determined in E (F). *p < 0.05, **p < 0.01, and ***p < 0.001, as determined by one-way ANOVA followed by Tukey's multiple comparisons test (mean ± SEM of three mice in B and technical triplicates in F).

To identify TCR clonotypes enriched in tumors following combination therapy, TCR sequences in TILs from four treatment groups—control, CTX, ICI, and CTX + ICI combination—were examined by single-cell TCR repertoire analysis (Figure 6C,D and Supplementary Table S1). TCR clonotypes were ranked according to their overall frequency across all samples, and eight of the most abundant TCR clonotypes (Top 1, 5, 8, 10, 11, 14, 15, and 20) were preferentially detected in the tumors from the combined therapy group. Notably, clones Top 5, 8, 10, 14, and 20 were predominantly found in Treg cells, while clones Top1, 11, and 15 were frequently observed in CD8 cells such as Tpex, Tex, and effector memory T cells. To further evaluate whether CD8 T cells bearing TCR clonotypes Top1, 11, and 15 could specifically recognize LLC tumor cells, we retrovirally transduced CD8 T cells with these TCRs and co-cultured them with LLC tumor cells. CD8 T cells expressing Top1 and Top15 produced significantly higher levels of IFN-γ, demonstrating that these clonotypes mediate tumor-specific CTL responses (Figure 6E,F). Collectively, our data suggest that CTX and ICI combination therapy selectively induces tumor-reactive CTL clones, which can be efficiently identified using our Pdcd1 fate-mapping system.

Discussion

We established a comprehensive map of PD-1–expressing cells that are modulated by ICI therapy with or without cytotoxic chemotherapy using Pdcd1^fm mice. In this system, PD-1–expressing cells are efficiently labeled with tdTomato, enabling the identification of TCR clonotypes of CTLs targeting LLC cells. Notably, these CTL clones were specifically induced by the combination of CTX and ICI, resulting in enhanced CTL activity against LLC tumors. Thus, the clonal diversification of CTLs elicited by the combination therapy likely contributes to its superior efficacy compared with ICI or chemotherapy alone in LLC tumor.

A limitation of this study is that we employed CTX to model the efficacy of combination therapy with ICI and a cytotoxic agent in lung cancer using LLC cells. CTX had been used for the treatment of lung cancer until the 1970s; however, since the 1980s, platinum-based agents have largely replaced CTX owing to their superior efficacy and more favorable toxicity profiles.²⁸ In our experimental setting, we were unable to consistently detect an additive effect of ICI in combination with cisplatin in LLC-bearing mice, which led us to adopt CTX as the cytotoxic backbone. Although CTX and cisplatin exert distinct effects on tumor and immune cells, both agents are known to induce immunogenic cell death, which may represent a key mechanistic basis for the efficacy observed with combination therapy.^28,37

Recently, a neoantigen in LLC cells, termed mRiok1, was identified through peptide screening; however, ICIs combined with mRiok1 peptide vaccination failed to induce tumor regression.³⁸ Consequently, TCR clonotypes associated with effective anti-tumor immune responses against LLC have remained undefined. In this study, we identified two distinct CTL TCR clonotypes, the Top1 and Top15 clonotypes, that were specifically induced by the combination of CTX and ICI. We acknowledge that the antigen specificity and cognate neoantigens of these clonotypes were not formally tested. The use of LLC-OVA cells, rather than parental LLC cells, may be advantageous for directly assessing tumor antigen–specific T cell responses elicited by combination therapy. Owing to this limitation, it remains unclear whether the emergence of these clonotypes is indispensable for the therapeutic efficacy of combined treatment. Nevertheless, CTX induces immunogenic cell death in tumor cells, leading to the release of neoantigens and DAMPs.^37,39 These signals, in combination with the enhanced antigen presentation to Tpex1 cells promoted by ICIs, may collectively contribute to the expansion of the observed CTL clonotypes against LLC cells.

Several studies have reported that cytotoxic chemotherapy increases the frequency of DCs in tumors and lymph nodes.^13,40 However, these studies did not specify which DC subsets are expanded by treatment. In contrast, our findings demonstrate that tdTomato⁺ DCs exhibit a cDC2 phenotype (CD11b⁺CD8α⁻) with high Ccl22 mRNA expression and are proportionally increased following treatment. CCL22 is known to recruit Treg cells via CCR4 into the tumor microenvironment, thereby suppressing antitumor immune responses.³⁵ However, cDC2 cells can promote antitumor immunity in the absence of Treg cells. Indeed, depletion of Treg cells has been shown to drive the migration of cDC2 from tumors to lymph nodes, where they induce tumor-specific Th1 responses.⁴¹ Because we and another group have shown that Treg cells within tumors are depleted by CTX treatment,⁴² the pronounced reduction of Treg cells by CTX in the tumor microenvironment may unleash tdTomato⁺ cDC2 cells from Treg-mediated suppression, thereby enabling them to promote CD4 T cell-dependent tumor rejection.

We identified that dysfunctional Treg cells in the tumor were characterized by coordinated upregulation of multiple, largely unannotated lncRNAs. Although intriguing, this pattern was not observed in datasets from other studies containing Treg cells, and its physiological significance remains unclear. PD-1 fate mapping may have revealed an activation state in which this specific chromosomal region is engaged in dysfunctional Treg cells. Importantly, combination therapy with CTX and ICI did not increase the overall abundance of Treg cells within the tumor microenvironment, and lncRNA⁺ dysfunctional Treg cells did not constitute a dominant Treg population. Thus, any potential increase in lncRNA⁺ dysfunctional Treg cells is unlikely to play a major role in mediating anti-tumor immune responses following CTX and ICI treatment.

Activated NK cells express PD-1, which suppresses NK cell cytotoxicity through interaction with PD-L1/PD-L2 on tumor cells.^43,44 However, other reports suggest that PD-1 is not a major checkpoint molecule for NK cells in tumors due to its limited expression.^16,45 Although around 10% of NK cells were positive for tdTomato in LLC tumor, more than 70% of them down-regulate PD-1 to overcome inhibitory interaction with PD-L1-expressing tumor cells. The down-regulation of PD-1 may in part explained by the internalization of PD-1 with PD-L1 because PD-1-expressing NK cells retrieves PD-L1 from tumor cells, allowing them to kill the tumor by PD-1⁺ NK cells. Thus, PD-1 blockade therapy does not effectively reactivate NK cells in tumor as previously reported.¹⁶

Our system for tracing PD-1–expressing cells provides a valuable tool for comprehensively analyzing anti-tumor immunity elicited by ICI therapy, particularly in settings where neoantigens and tumor-specific CTL clonotypes remain undefined. Using this approach, we were able to identify dominant TCR clonotypes responsive to the combination of CTX and ICIs. In this study, we focused primarily on acute responses to combination therapy; however, the Pdcd1^fm mouse model also offers unique opportunities to investigate the long-term effects of ICI therapy, which are clinically relevant in patients who discontinue treatment owing to adverse events. Furthermore, this system may be applied to study immune-related adverse events in preclinical models. Collectively, our findings highlight the Pdcd1^fm mouse system as a powerful platform for generating clinically relevant insights in tumor immunology.

Supplementary Material

Supplementary material

Supplementary Figure 3.tif

Supplementary material

Supplementary Table S1.xlsx

Supplementary material

Supplementary Table S2.xlsx

Supplementary material

Supplementary Figure 2.tif

Supplementary material

Figure_legends_for_supplements.docx

Supplementary material

Supplementary Table S4.xlsx

Supplementary material

Supplementary Table S3.xlsx

Supplementary material

Supplementary Figure 1.tif

Supplementary material

Supplementary Figure 4.tif

Supplementary material

Author_Checklist_E10_only_TE

Funding Statement

This study was supported in part by the Grants-in-Aid for Scientific Research Program of the Japanese Society for the Promotion of Science (grant numbers 24K02256 to T.E and 222K08746 and 25K11901 to K.I.), the Takeda Science Foundation (to T.E), the Yasuda Memorial Medical Foundation (to T.E.), Key Research Laboratory at Akita University (to T.E.), and Multilayered Stress Diseases (JPMXP1323015483) at Science Tokyo (to T.E.).

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Data availability statement

Sequencing data are available under the ArrayExpress (EMBL-EBI) accession number E-MTAB-15883.The data generated in this study are available upon request from the corresponding author.

Supplemental material

Supplemental data for this article can be accessed at https://doi.org/10.1080/2162402X.2026.2639723.