Plasminogen Activator Inhibitor-1 Mediates Tolerance to Anti-PD-1 Immunotherapy in Non-Small Cell Lung Cancer.

Introduction
Lung cancer is the leading cause of cancer-related deaths worldwide (1), and its incidence is still expected to increase in the future (2). Lung cancer is histopathologically classified into small cell lung cancer and non–small cell lung cancer (NSCLC), with NSCLC accounting for 85% of cases. For approximately half of patients with advanced NSCLC without targetable driver mutations who cannot undergo curative surgery or radiotherapy, the standard first-line treatment is cytotoxic chemotherapy combined with immunotherapy using immune checkpoint inhibitors (ICI) that target PD-L1 or PD-1. However, only a small proportion of patients are cured, and more than half experience disease progression within 1 year from treatment initiation (3). Disease progression is partly attributable to NSCLC cells acquiring resistance to ICI treatment. The reported mechanisms include tumor antigen depletion, T-cell dysfunction, increased numbers of immunosuppressive cells, and alteration in the expression of PD-L1 within tumor cells (4).
Moreover, cancer cells undergo epithelial–mesenchymal transition (EMT), acquiring resistance to lysis by lymphocytes. EMT-induced cells express high levels of TGF-β, which promotes an immunosuppressive tumor-immune microenvironment (TIME; 5–7). In addition, drug-tolerant persister cancer cells (DTP) that survive within tumors during treatment and exhibit reversible resistance have recently been reported as a cause of subsequent tumor regrowth (8, 9). Eliminating DTPs may lead to a cure for advanced NSCLC. Increasing evidence indicates that DTP cells can reprogram diverse cellular processes, including epigenetic regulation, transcription, translation, and metabolism, to sustain survival under therapeutic pressure (10). In addition, recent studies have implicated involvement of insulin-like growth factor-1 receptor signaling, histone demethylases, and EMT of cancer cells (11–13).
Plasminogen activator inhibitor-1 (PAI-1) is a 47-kDa glycoprotein produced by various cells, including vascular endothelial cells, hepatocytes, adipocytes, macrophages, and fibroblasts, and functions as an inhibitor of the fibrinolytic system. PAI-1 has also been reported to be involved in systemic inflammation, atherosclerosis, and fibrosis in the lungs and kidneys (14–16). Additionally, high PAI-1 expression in tumor tissues has been identified as a poor prognostic factor not only in NSCLC but also in other cancer types, and its involvement in tumor progression has been described (17–19). We have previously reported that in lung cancer cells, PAI-1 contributes to acquired resistance to chemotherapy via EMT, which is one of the resistance mechanisms to anti–PD-1 (aPD-1) antibody treatment (20). We have also reported that PAI-1 is associated with the emergence of DTPs through EMT in osimertinib-treated EGFR-mutated NSCLC cells (21). Based on these findings, PAI-1 may be a significant influencing factor in the tolerance to aPD-1 antibody treatment.
TM5614, a PAI-1 inhibitor, is a small-molecule drug selected among more than 1,400 novel derivatives explored by in silico drug discovery based on the crystal structure of human PAI-1. A phase I trial in healthy individuals and phase II trials in chronic myeloid leukemia and coronavirus disease 2019 (COVID-19) have shown that oral administration of TM5614 is well tolerated and safe (22, 23). A recent phase II study has also shown the efficacy and safety of a combination treatment with nivolumab and TM5614 for patients with melanoma who experience disease progression during nivolumab treatment (24).
Based on these findings, this study aimed to investigate whether PAI-1 was involved in the emergence of DTPs in aPD-1 antibody–treated NSCLC cells using TM5614 and a mouse lung cancer model. Additionally, we examined the possibility of using PAI-1 as a therapeutic target to overcome this tolerance and whether combination treatment with aPD-1 antibody and TM5614 could be a new therapeutic strategy for NSCLC.

Materials and Methods

Cell lines and culture conditions
The murine lung adenocarcinoma cell line Lewis lung carcinoma (LLC; RRID: CVCL_4358, Anti-Cancer Japan, obtained in 2020), the murine lung squamous cell carcinoma cell line KLN205 (RRID: CVCL_3533, Cell Resource Center for Biomedical Research Institute of Development, Aging and Cancer, Tohoku University, obtained in 2021), the human lung fibroblast cell line MRC-5 (RRID: SCR_023187, Japanese Collection of Research Bioresources Cell Bank, obtained in 2002), and the murine macrophage cell line Raw264.7 (RRID: CVCL_0493, ATCC, obtained in 2005) were used in this study. These cells were cultured in RPMI-1640 medium supplemented with 7.5% FBS, penicillin, and streptomycin (100 U/mL and 100 μg/mL, respectively) in a humidified incubator with 5% CO2 at 37°C. All experiments were performed within 6 months after thawing and within 10 passages. Vendor certificates confirmed that all cell lines used in this study (LLC, KLN205, MRC-5, and RAW264.7) were free of Mycoplasma contamination at the time of receipt. No in-house Mycoplasma testing was subsequently performed. Because these cell lines were obtained directly from certified cell banks that perform characterization and were used within 6 months after resuscitation, reauthentication was not required under American Association for Cancer Research policy; vendor authentication data were considered sufficient. The biological sex of these cell lines was not specified.

Reagents
TM5614 (Renascience) was dissolved with carmellose sodium (Maruishi) in in vivo experiments and with DMSO (Sigma) in in vitro experiments. TM5614 was administered at a concentration of 62.7 μmol/L in in vitro experiments according to a previous study (25). Detailed information on TM5614 is provided in the Supplementary Materials and Methods and Supplementary References. SK-216 was chemically synthesized and supplied by Shizuoka Coffein, Co., Ltd. The IC50 value for SK-216 cells was determined to be 44 μmol/L, as reported in the international patent WO2004/010996 (World Intellectual Property Organization; https://patentscope2.wipo.int/search/en/WO2004010996). SK-216 corresponds to example number 125 described in this patent. A IgG2a isotype control antibody (#BE0089; Bio X Cell, RRID: AB_1107769) and an aPD-1 antibody (#BE0146; Bio X Cell, RRID: AB_10949053) were used.

qRT-PCR
Total cellular RNA was isolated using RNeasy Mini Kit (Qiagen) according to the manufacturer’s instructions. The isolated total RNA was reverse-transcribed into cDNA using a high-capacity RNA-to-cDNA kit (Thermo Fisher Scientific). qRT-PCR was performed using CFX96 (Bio-Rad, RRID: SCR_018064). mRNA expression levels were evaluated and normalized to β-actin levels as an endogenous reference. All PCR primers were obtained using TaqMan Gene Expression Assays (Thermo Fisher Scientific). The following primers were used: PAI-1 (Mm00435858_m1), E-cadherin (Mm01247357_m1), N-cadherin (Mm01162497_m1), fibronectin-1 (Mm01256744_m1), α-SMA (Mm00725412_s1), vimentin (Mm01333430_m1), TGF-β (Mm01178820_m1), TGF-β (Hs00171257_m1), and β-actin (Mm02619580_g1). All reagents were from Thermo Fisher Scientific (RRID: SCR_008452).

Microarray tests
Total RNA was extracted using RNeasy Mini Kit (Qiagen) according to the manufacturer’s instructions and quantified on an Agilent 2100 bioanalyzer (RRID: SCR_018043). The global expression of mRNAs in the cells was examined using a Clariom S Array (Thermo Fisher Scientific). The raw mRNA expression profile data were analyzed using iDEP version 0.96. Parametric gene set enrichment analysis was used to identify genes differentially expressed between groups. The Kyoto Encyclopedia of Genes and Genomes pathway database and diagrams were used to visualize and compare pathway expression. Genes with adjusted P values < 0.05 after FDR correction and a fold change of >2 were considered differentially expressed.

Single-cell RNA sequencing
LLC tumor-bearing mice were treated with an IgG2a isotype control antibody or an aPD-1 antibody for 7 days prior to tumor harvest. Single-cell RNA sequencing (sc-RNA seq) libraries were constructed using GEM-X Flex Gene Expression Mouse (10× Genomics) according to the manufacturer’s instructions. Libraries were sequenced on NovaSeq X plus at a read length of 28 × 90 to yield a minimum of 10,000 reads per cell. The resulting raw data were processed using Cell Ranger version 9.0.1 (10× Genomics). The data were visualized and analyzed by Loupe Browser 4.2.0 (10× Genomics). Cell type annotation was performed based on the expression of canonical marker genes extracted from each cluster using Loupe Browser. Putative cell identities were assigned by comparing marker gene profiles with reference datasets from CellMarker 2.0.

IHC staining
IHC staining was performed as previously described (20). The following primary antibodies were used: PAI-1 (#sc-8979; Santa Cruz, RRID: AB_2186884), CD206 (#24595; Cell Signaling Technology, RRID: AB_2892682), CD8α (#98941; Cell Signaling Technology, RRID: AB_2756376), PD-L1 (#64988; Cell Signaling Technology, RRID: AB_2799672), CD31 (ab182981; Abcam, RRID: AB_2920881), α-SMA (ab5694; Abcam, RRID: AB_2223021), and TGF-β (ab215715; Abcam, RRID: AB_2893156). The stained specimens were checked, and micrographs were obtained under the same photographic conditions at three representative locations where the aggregation of cancer cells could be clearly seen and where 3,3′-diaminobenzidine (DAB) coloration was considered the strongest. Micrographs were captured using a BZ-9000 microscope (Keyence, RRID: SCR_015486). The open-source plugin IHC Profiler, compatible with ImageJ (RRID: SCR_003070), was used to separate DAB- and hematoxylin-stained regions. The percentage of DAB-stained regions was then calculated and compared.

Quantification of PAI-1 protein levels
PAI-1 protein levels in the culture medium were quantified using ELISA. LLC cells were seeded in six-well plates at 5 × 104 cells/well. After 24 hours, TM5614 (62.7 μmol/L) was added in three wells (TM5614 group), and the same volume of DMSO was added to the remaining three wells (control group). The total PAI-1 secreted into serum-free culture medium for 48 hours was measured using an ELISA kit (RayBiotech) according to the manufacturer’s instructions.

Proliferation assay
PAI-1 was inhibited using TM5614. LLC or KLN205 cells were seeded in six-well plates at 5 × 104 cells/well. After 24 hours, TM5614 (62.7 μmol/L) was added in three wells (TM5614 group), and the same volume of DMSO was added to the remaining three wells (control group). The number of viable cells in each well was counted every 24 hours using trypan blue.

SB-431542 and TGF-β treatment
Stock solutions of TGF-β receptor inhibitor SB-431542 (Selleck Chemicals) were prepared in DMSO at 10 mmol/L, prepared in advance, and stored at −20°C until use. Recombinant Mouse TGF-β1 Protein (FUJIFILM Wako) was reconstituted in 1 mL PBS at a concentration of 5 μg/mL and stored at −80°C. The stock solution was diluted 1:1,000 to achieve a working concentration of 5 ng/mL immediately before use. On day 0, LLC cells were seeded at 2 × 105 cells per well in six-well plates and divided into the control group and the SB-431542 group (n = 3/group). On day 1, the culture medium was replaced with serum-free medium. On day 2, SB-431542 was added at a final concentration of 5 μmol/L, followed 30 minutes later by the addition of Recombinant Mouse TGF-β1 Protein (5 ng/mL). On day 4, cells were harvested, and the expression levels of EMT-related genes in each group were assessed using qRT-PCR.

Animals and subcutaneous tumor model
All animal experiments were approved by the Committee on Animal Research of Hiroshima University (approval no. A22-189) and conducted according to the Guide for the Care and Use of Laboratory Animals, eighth edition, 2010 (NIH).
Female C57BL/6 (RRID: IMSR_JAX:000664) and B6D2F1/Crl (RRID: MGI:5649818) mice (6–8 weeks old; Charles River Laboratories, RRID: SCR_003792) were housed in pathogen-free rooms in a controlled environment under a 12-hour light–dark cycle and allowed free access to laboratory chow and water. C57BL/6 mice were used for the experiments with LLC cells, whereas B6D2F1/Crl mice were used for those with KLN205 cells.
The mice were anesthetized using mixed anesthetic agents, including medetomidine (0.3 mg/kg body weight; Kyoritsu Seiyaku), midazolam (4 mg/kg body weight; Sandoz K.K.), and butorphanol (5 mg/kg body weight; Meiji Seika Pharma). The indicated cells (1 × 106) were suspended in 50 μL RPMI-1640 medium mixed with 100 μL Matrigel basement membrane matrix (Corning) and subcutaneously injected into the left flank of mice. Treatment was started when the tumor size reached approximately 200 mm3. Carmellose (0.5%) was used as vehicle control. The average body weights of the mice in each group were not significantly different at the beginning of the experiments. TM5614 (10 mg/kg) was prepared using 0.5% carmellose. Vehicle control or TM5614 were administered orally by gavage every day. IgG2a isotype control antibody or an aPD-1 antibody was intraperitoneally injected every 3 days at a dose of 250 μg. For SK-216 experiments, the mice were given drinking water with or without SK- 216 (500 ppm). Administration was performed in an open-label setting. Subcutaneous tumor size was measured using a caliper every 3 days, and tumor volume was calculated as follows: volume = 1/2 × [(shortest diameter)2 × longest diameter].

Clinical samples
Written informed consent for the use of surgical specimens in research was obtained from all patients. This study was approved by the Ethics Committee of Hiroshima University Hospital (approval no. E2024-0060) and conducted using the opt-out method in accordance with the principles of the Declaration of Helsinki.
The medical records of patients with NSCLC treated at Hiroshima University Hospital were retrospectively reviewed. Three patients with stage IIIA NSCLC who underwent surgery after preoperative aPD-1 antibody monotherapy in 2018 were consecutively evaluated. Among them, one patient was excluded because the preoperative treatment eliminated the cancer cells, with no cancer cells detected in the surgical specimens. As a control group, two patients with clinical stage IIIA NSCLC who underwent surgery without preoperative treatment in the same period were evaluated. Thus, four patients were included. The patient characteristics are summarized in Supplementary Table S1.

Statistical analysis
All results are expressed as means ± SEMs. The Mann–Whitney U test was used for two-group comparisons, whereas one-way ANOVA with Tukey test was used for comparisons among three or more groups. Statistical significance was set at P < 0.05. All statistical analyses were performed using JMP Pro 17 (RRID: SCR_014242).

Results

PAI-1 expression level in tolerant persister NSCLC cells at the early phase of aPD-1 antibody treatment
To investigate the level of PAI-1 expression in aPD-1 antibody-tolerant persistent cells (aPD-1-TP) in tumors at the early phase of aPD-1 antibody treatment, we established a subcutaneous tumor model using LLC and KLN205 cells and administered aPD-1 antibody or IgG2a isotype control antibody. The experimental design is shown in Supplementary Fig. S1A. Tumor growth was relatively suppressed until day 7 following treatment initiation, after which the tumors began to grow more rapidly. Therefore, cancer cells that survived within the tumors 7 days after aPD-1 antibody treatment were defined as aPD-1-TPs. To comprehensively compare gene expression in aPD-1-TPs and control cells, microarray analyses were performed on aPD-1 antibody–treated and untreated LLC and KLN205 tumor models. Parametric gene set enrichment analysis revealed upregulation of the TGF-β signaling pathway in the aPD-1–treated group in both the LLC and KLN205 tumor models (Fig. 1A and B). Given that TGF-β has been reported to induce PAI-1 expression and promote EMT in cancer cells (26), we next focused on the expression of PAI-1 and EMT-associated genes.
Microarray analyses of both models demonstrated that the aPD-1 group exhibited upregulated expression of PAI-1 and mesenchymal markers and downregulated expression of epithelial marker genes (Fig. 1C and D). qRT-PCR also showed that PAI-1 mRNA expression levels in aPD-1 antibody–treated LLC tumors were significantly higher than those in control tumors (Fig. 1E). IHC staining of tumor sections demonstrated that the PAI-1–positive area in the aPD-1 antibody–treated LLC tumors was significantly larger than that in the control tumors (Fig. 1F and G). These results show that PAI-1 protein expression levels in aPD-1-TPs are significantly higher than those in control cells. Additionally, in the subcutaneous KLN205 cell tumor model, PAI-1 mRNA expression levels were significantly higher in tumors treated with aPD-1 antibody than in control tumors (Fig. 1H). IHC staining of the tumor sections also showed that the PAI-1 protein expression level was significantly higher in aPD-1-TPs than in control cells (Fig. 1I and J). These results indicate that aPD-1-TPs express significantly higher levels of PAI-1 than control cells.
Subsequently, to investigate whether tumor size was associated with PAI-1 expression in cancer cells, PAI-1 levels in subcutaneous tumors were compared between tumors collected on day 7 after treatment initiation (small tumor group) and those collected at a later time point after tumor growth (large tumor group; Supplementary Fig. S1B). qRT-PCR showed that PAI-1 mRNA expression levels in LLC tumors were significantly higher in the large tumor group than in the small tumor group (Supplementary Fig. S1C). IHC staining of tumor sections showed a significantly larger proportion of PAI-1–positive area in the large tumor group than in the small tumor group (Supplementary Fig. S1D and S1E). Similar results were observed in the experiments using KLN205 cells (Supplementary Fig. S1F–S1H). These findings suggest that PAI-1 expression in cancer cells tends to increase with increasing tumor size during aPD-1 antibody treatment.

High PAI-1 expression in human NSCLC specimens after aPD-1 antibody treatment
To investigate whether high levels of PAI-1 expression also occur in human NSCLC, differences in PAI-1 expression were compared between surgically resected NSCLC specimens with and without aPD-1 antibody treatment. The patient characteristics are summarized in Supplementary Table S1. Although the small sample size and insufficient adjustment for potential confounding factors limited the analysis, IHC staining showed that PAI-1 expression was significantly higher in NSCLC specimens treated with an aPD-1 antibody than in controls (Supplementary Fig. S2A and S2B).

Effect of TM5614 on the antitumor efficacy of aPD-1 treatment
Administration of an aPD-1 antibody temporarily suppressed tumor growth in the subcutaneous tumors from LLC cells, but tumor regrowth was observed around day 30 (Fig. 2A). We then examined whether PAI-1 inhibition, highly expressed in aPD-1-TPs, could suppress subsequent tumor regrowth. The therapeutic potential of TM5614, a small-molecule PAI-1 inhibitor known to promote the degradation of PAI-1 protein (27), was investigated. TM5614 had a IC50 value of <6.95 μmol/L in a previous tPA-dependent hydrolysis assay (28, 29). The results showed significantly lower PAI-1 protein levels in the supernatant in the TM5614 group than in the control group (Supplementary Fig. S3A). With respect to the in vivo antitumor efficacy of TM5614 in combination with aPD-1 therapy, no tumor regrowth was observed in the aPD-1 antibody + TM5614 group (Fig. 2A and B). Similar results were observed in the subcutaneous tumors from KLN205 cells (Fig. 2C and D).
Given that the effect of TM5614 is exerted through PAI-1 degradation, it was next examined whether TM5614 treatment reduced PAI-1 expression within tumors. IHC analysis revealed that PAI-1 expression in tumor tissues was decreased in the TM5614-treated group (Fig. 2E and F). Subsequently, the antiproliferative effect of TM5614 on lung cancer cell lines in vitro was investigated. TM5614 significantly inhibited tumor cell proliferation in both the LLC and KLN205 models (Supplementary Fig. S3B and S3C). Collectively, these findings indicate that TM5614 reduces PAI-1 levels within tumors and exerts a suppressive effect on cancer cell proliferation. Additionally, PAI-1 is involved in the survival of aPD-1-TPs and subsequent tumor regrowth during aPD-1 therapy.
To rule out the possibility that TM5614 exerts off-target effects on molecules other than PAI-1, additional experiments were performed using SK-216, another PAI-1 inhibitor. Briefly, SK-216 has an IC50 value of 44 μmol/L as reported in international patent WO04/010996. In our previous subcutaneous lung cancer mouse model, SK-216 treatment also resulted in tumor regression comparable with that in PAI-1 knockout mice (20). The current study established subcutaneous tumor models using LLC and KLN205 cells, and the mice were randomized into four groups: control, aPD-1 monotherapy, SK-216 monotherapy, and aPD-1 plus SK-216 combination therapy. The results were similar to those obtained in experiments using TM5614. Particularly, compared with aPD-1 monotherapy, combined aPD-1 and SK-216 treatment was associated with significantly better antitumor efficacy in both LLC and KLN205 models (Supplementary Fig. S4). These findings suggest that PAI-1 contributes to the survival of aPD-1-TPs. Importantly, TM5614 inhibition of PAI-1 significantly enhances the efficacy of aPD-1 monotherapy, and TM5614 activity is specific to PAI-1.

PAI-1 involvement in EMT of aPD-1-TPs during aPD-1 therapy
The microarray results showed higher expression of mesenchymal genes and lower expression of epithelial genes in the aPD-1 antibody group than in the control group (Fig. 1C and D). These findings indicate that aPD-1-TPs persisted through the induction of EMT. qRT-PCR also showed that compared with the control group, the aPD-1 antibody group had significantly lower expression of epithelial genes and significantly higher expression of mesenchymal genes. In contrast, the aPD-1 antibody + TM5614 group showed significantly lower mesenchymal gene expression than the aPD-1 antibody group (Fig. 3A–E). Similar results were observed for the tumors from KLN205 cells (Fig. 3F–J). These findings support that PAI-1 is involved in EMT of aPD-1-TPs during aPD-1 therapy and that TM5614 suppresses EMT in aPD-1 antibody–treated NSCLC cells.
To investigate whether TM5614 inhibited EMT at a later time, the expression of EMT-related genes was compared between the aPD-1 antibody and aPD-1 antibody + TM5614 groups at day 36 (Supplementary Fig. S1B). The expression of mesenchymal genes in the aPD-1 antibody + TM5614 group was significantly lower than that in the aPD-1 antibody group (Supplementary Fig. S5A–S5E). Similar results were observed in the mouse lung cancer model using KLN205 cells (Supplementary Fig. S5F–S5J). These findings suggest that TM5614 suppresses EMT of cancer cells at a later time point.

PAI-1 involvement in TGF-β–induced EMT of tolerant persister NSCLC cells during aPD-1 antibody treatment
The results showed that PAI-1 was involved in EMT of aPD-1-TPs. Signaling pathways, such as TGF-β or the integrin-induced pathway, promote EMT in cancer cells. Further analysis of the microarray data revealed that the key transcription factors associated with TGF-β–induced EMT, Twist, Zeb, and Snail, were highly expressed in the aPD-1 group (Fig. 3K and L). In contrast, the expression of genes composing the integrin-induced pathway was not consistently increased (Supplementary Fig. S6A and S6B). Consistently, IHC staining showed that TGF-β expression was significantly increased in the aPD-1 antibody–treated tumors in both the LLC (Supplementary Fig. S6C and S6D) and KLN205 models (Supplementary Fig. S6E and S6F). To further clarify the role of TGF-β in PAI-1–mediated EMT, additional in vitro experiments were conducted using LLC cells and a TGF-β inhibitor. SB-431542, whose inhibitory effect on TGF-β signaling in cancer cells had been experimentally validated in previous reports (30), was used as a TGF-β inhibitor. The expression of epithelial genes was higher whereas that of mesenchymal genes was lower in LLC cells with SB-431542 treatment than in control LLC cells without SB-431542 treatment (Supplementary Fig. S7). Collectively, these findings suggest that TGF-β plays an important role in PAI-1–mediated EMT.
IHC staining showed positive TGF-β expression in cancer and stromal cells. TGF-β is secreted by tumor-associated macrophages (TAM) and cancer-associated fibroblasts (CAF), in addition to cancer cells (31, 32). Consistently, the current study found that TAMs and CAFs existed in the tumor stroma, indicating TGF-β expression in tumors (Supplementary Fig. S8A–S8D). PAI-1 promotes TGF-β expression from CAFs and TAMs (16, 20). Thus, we investigated the relationship between PAI-1 and TGF-β expression in fibroblasts and macrophages. The results showed that pharmacologic inhibition of PAI-1 by TM5614 significantly reduced TGF-β mRNA expression in both Raw264.7 macrophages and MRC-5 fibroblasts (Supplementary Fig. S8E and S8F).

scRNA-seq Identification of PAI-1–associated clusters among tolerant persister NSCLC cells after aPD-1 treatment
scRNA-seq was performed to investigate whether specific clusters of DTPs that survived following aPD-1 treatment were influenced by PAI-1 and underwent EMT. The distribution of all cell populations identified by scRNA-seq is shown in Fig. 4A. Tumor cells were further divided into six distinct clusters (Fig. 4B). Among these, cluster 2 was the most abundant in the aPD-1 treatment group (Fig. 4C). Compared with the control group, the aPD-1 group showed a higher proportion of cluster 2 cells, along with high expression of PAI-1 (Fig. 4D), TGF-β, and EMT markers (Fig. 4E). These findings suggest that cluster 2 comprises aPD-1-TPs and are influenced by PAI-1. Additionally, single-cell transcriptomic analysis showed that cluster 2 had elevated expression of genes involved in mitochondrial function (mt-Nd1 and mt-Nd2), DNA repair (Neil3 and Pif1), and EMT (Ankle1, Kif15, Atg9b, and Cep128; Fig. 4F). This gene signature is consistent with an EMT-like, oxidative phosphorylation–DTP phenotype.
In contrast, cluster 1 was enriched for genes regulating cell proliferation and nucleotide metabolism (e.g., Rrm2, Mthfd2, Tuba1c, and Shmt2), reflecting a highly proliferative, treatment-responsive cell population. Cluster 3 shared partial overlap with cluster 2 but was dominated by mitotic regulators (e.g., Espl1, Zwilch, Cdc7, and Spag5), suggesting active chromosome segregation and cell division. This potentially indicated a transitional state between proliferation and tolerance. Cluster 4 exhibited high expression of stress and inflammation-related genes (e.g., Trp53inp1, Cxcl1, and Gpnmb), indicating adaptation to microenvironmental stress and possible immune modulation. Cluster 5 was defined by the activation of embryonic and germline-associated genes (e.g., Spin2g, Duxf3, and Zscan4d), resembling diapause-like persister cells in a quiescent, therapy-resistant state. Finally, cluster 6 expressed metabolic stress–response genes (e.g., Txn1, Glrx5, and Ybx1), indicative of a metabolically adapted, low-proliferative state aligned with the integrated stress response (Fig. 4F).

PAI-1 induces survival of aPD-1-TP NSCLC cells by creating an immunosuppressive TIME
aPD-1-TPs expressed PAI-1 and TGF-β and underwent EMT. TGF-β was previously reported to form an immunosuppressive TIME (31). To evaluate the TIME status in the tumors treated with an aPD-1 antibody, the current study examined the number of tumor-infiltrating lymphocytes (TIL) and TAMs and PD-L1 expression in cancer cells. IHC staining revealed that in the tumors from LLC cells, the number of CD206-positive TAMs and PD-L1–expressing cancer cells was significantly higher in the aPD-1 antibody group than in the control group. Meanwhile, the number of CD8-positive TILs was significantly lower in the aPD-1 antibody group than in the control group. In contrast, the number of TAMs and PD-L1 expression in cancer cells was lower whereas the number of CD8-positive TILs was higher in the aPD-1 antibody + TM5614 group than in the aPD-1 antibody group (Fig. 5A–D). Similar results were obtained in the mouse lung cancer model using KLN205 cells (Fig. 5E–H). These findings indicate that PAI-1 is involved in the survival of tolerant persister cancer cells in tumors treated with an aPD-1 antibody by increasing the number of TAMs and PD-L1 expression while decreasing the number of TILs.
Additionally, TM5614 administration overcame tolerance to the aPD-1 antibody by improving the TIME. Subsequently, the TIME at a later time point during aPD-1 treatment was compared between the aPD-1 antibody and aPD-1 antibody + TM5614 groups. Compared with the aPD-1 antibody group, the aPD-1 antibody + TM5614 group showed a significantly lower number of TAMs and PD-L1 expression and higher number of TILs (Supplementary Fig. S9A–S9D). Similar results were obtained in the tumors from KLN205 cells (Supplementary Fig. S9E–S9H). These findings suggest that TM5614 improves the TIME at a later time point during aPD-1 treatment.

PAI-1 is involved in the survival of aPD-1-TPs via angiogenesis
Considering that PAI-1 was involved in tumor angiogenesis (18), we investigated whether this mechanism was responsible for PAI-1–induced tolerance to aPD-1 therapy by examining the degree of angiogenesis in the tumors at the tolerant phase. IHC staining revealed that the CD31-positive area was significantly larger in the tumors from LLC cells tolerant to aPD-1 antibody than in controls. Conversely, the CD31-positive area in the aPD-1 + TM5614 group was significantly smaller than that in the aPD-1 antibody group (Fig. 6A and B). Similar results were noted in the model using KLN205 cells (Fig. 6C and D). These findings suggest that PAI-1 is involved in the survival of tolerant persister cancer cells in tumors treated with an aPD-1 antibody via angiogenesis. Subsequent comparison of the CD31-positive area at a later time point during aPD-1 treatment showed that it was smaller in the aPD-1 antibody + TM5614 group than in the aPD-1 antibody group (Supplementary Fig. S10A and S10B). Similar results were observed in the mouse lung cancer model using KLN205 cells (Supplementary Fig. S10C and S10D). These findings suggest that TM5614 suppresses angiogenesis at a later time point during aPD-1 treatment.

Discussion
The current study showed that aPD-1-TPs expressed high levels of PAI-1 and that PAI-1 was involved in the survival of aPD-1-TPs via EMT. Furthermore, scRNA-seq revealed that a subset of aPD-1-TPs was characterized by PAI-1–driven EMT, which may contribute to their survival. Moreover, PAI-1 induced tolerance by shifting the TIME through two steps. First, it promoted immunosuppression (e.g., enhancing TAM and decreasing TIL counts) and increased PD-L1 expression in cancer cells. Second, it promoted tumor angiogenesis (Supplementary Fig. S11A). We subsequently showed that combination treatment with aPD-1 antibody and TM5614 suppressed EMT in NSCLC cells and improved TIME status, resulting in significantly higher antitumor efficacy than aPD-1 treatment alone (Supplementary Fig. S11B). Collectively, these results support that the combination of an aPD-1 antibody and TM5614 could be a novel treatment strategy for NSCLC. To our best knowledge, this is the first study to show that tolerant persister NSCLC cells treated with an aPD-1 antibody express high levels of PAI-1, resulting in treatment tolerance.
Most previous studies on drug resistance have focused on the mechanisms of acquired resistance that arises after long-term treatment. Considering evidence of DTPs that survive initial anticancer treatment and exhibit reversible resistance (9–11), eliminating DTPs may lead to a cure for advanced NSCLC by suppressing subsequently acquired resistance. We have previously reported that EGFR-mutated NSCLC cells treated with the EGFR tyrosine kinase inhibitor osimertinib expressed high levels of PAI-1 and developed acquired tolerance to osimertinib (21). Therefore, PAI-1 expressed by aPD-1-TPs can be an important target for eliminating aPD-1-TPs. In the current study, aPD-1-TPs in tumors treated with aPD-1 antibody had EMT characteristics associated with PAI-1. Cancer cells that undergo EMT become resistant to T-cell lysis, leading to resistance to immunotherapy (7). A previous study also reported that urothelial cancers with high expression of EMT-related genes showed a poorer response to aPD-1 antibody than did those with low expression of these genes (33). Thus, EMT is considered an important resistance mechanism to aPD-1 treatment.
The current study demonstrated that PAI-1 was involved in TGF-β–induced EMT of aPD-1-TPs during aPD-1 therapy. PAI-1 and TGF-β expression was higher in aPD-1–tolerant tumors than in control tumors. Previous studies have also shown a positive correlation between PAI-1 and TGF-β expression in NSCLC specimens (34). The current study also found that PAI-1 induced TGF-β expression in fibroblasts and macrophages, consistent with previous findings (16, 20). Collectively, these results support that PAI-1 induces TGF-β expression, resulting in EMT in NSCLC tumors treated with an aPD-1 antibody. We have also previously reported that PAI-1 is involved in the induction of EMT in lung cancer cells in vitro, providing direct evidence of the association between PAI-1 and EMT (20, 21). One proposed mechanism underlying this process involves PAI-1 binding to PIAS3 (a protein inhibitor of activated STAT3), a known suppressor of STAT3 activity. This interaction leads to STAT3 activation, subsequently inducing the expression of EMT-related genes (35).
This study showed that PAI-1 contributed to the survival of aPD-1-TPs by shifting the TIME to an immunosuppressive one by increasing the numbers of TAMs and decreasing that of TILs and enhancing PD-L1 expression in cancer cells. TGF-β contributes to the formation of an immunosuppressive TIME. PAI-1 also contributes to the shift to an immunosuppressive TIME by increasing the tumor infiltration of TAMs. A previous report has shown that various tumor tissues with high PAI-1 expression are characterized by increased TAM counts (36). Additionally, PAI-1 increases the number of TAMs by increasing the migration ability of macrophages (37, 38) and promoting polarization to M2 macrophages via the C-terminal uPA interaction domain (36, 39). Subsequently, CXCL10 and CCL22 from TAMs inhibit lymphocyte invasion into the tumor (40).
A meta-analysis on NSCLC reported that high TIL content was associated with the antitumor efficacy of aPD-1 antibodies (41, 42). Consequently, a low number of TILs induced by TAMs would be expected to reduce the efficacy of such antibodies. Additionally, TAMs promote PD-L1 expression in cancer cells, enhancing T-cell apoptosis (43–46). These findings indicate that TAMs contribute to the formation of an immunosuppressive TIME and that TAM infiltration may be closely associated with resistance to aPD-1 antibody treatment (47). The current study found that TM5614 could promote PAI-1 degradation. Therefore, the reduction in TAMs in the TM5614-treated group was likely due to decreased PAI-1 levels. This supports that TM5614 enhances the antitumor efficacy of aPD-1 antibodies by suppressing the intratumor migration of TAMs.
The current study found that aPD-1-TPs expressed high levels of PAI-1 and induced tumor angiogenesis. In addition, PAI-1–induced tumor angiogenesis was associated with the survival of aPD-1-TPs. We have previously reported that PAI-1 played an essential role in tumor angiogenesis (18, 19), an important mechanism of chemotherapy resistance. Tumor blood vessels are often twisted and do not have the typical hierarchical organization observed in normal vessels. Additionally, adhesion between endothelial cells is reduced, and pericyte coverage is decreased. These changes can increase vascular permeability and cause plasma leaking, thereby increasing interstitial pressure. This results in obstructed blood vessels and irregular and impaired blood flow, which in turn decreases drug delivery and immune cell infiltration into the tumor (48). Therefore, tumor angiogenesis is a resistance mechanism to aPD-1 therapy, and TM5614 administration, which limits tumor angiogenesis, would be a reasonable treatment option to inhibit the survival of aPD-1-TPs.
In the present study, TM5614 acted as a specific inhibitor of PAI-1. Its effects are presumed to be primarily mediated through the inhibition of cancer cell–derived PAI-1. However, our previous study using PAI-1 knockout mice showed that host cell–derived PAI-1 also played a critical role in tumor angiogenesis (18). Therefore, it is reasonable to consider that, in addition to targeting cancer cell–derived PAI-1, TM5614 may also inhibit host cell–derived PAI-1, resulting in the suppression of tumor angiogenesis, as observed in this study. Furthermore, tumors tolerant to aPD-1 therapy showed high PAI-1 expression, along with increased TAM infiltration. TAMs promote tumor angiogenesis by secreting proangiogenic factors, such as TGF-β and VEGF (49, 50). These findings suggest that PAI-1 may contribute to tumor angiogenesis.
This study demonstrates that PAI-1 is involved in the survival of aPD-1-TPs and that combining TM5614 with an aPD-1 antibody can enhance the antitumor effect against NSCLC. Many basic and clinical studies are currently being conducted to develop treatment strategies, including combination therapies with an anti–CTLA-4 antibody, to overcome the resistance mechanisms to aPD-1 antibodies. However, these strategies face several challenges, such as increased immune-related adverse events and high medical costs. TM5614 has been confirmed to be safe in a phase I trial in healthy individuals and in phase II trials in patients with COVID-19 and melanoma (23, 24). Furthermore, TM5614 is a small-molecule compound that can be administered orally and has a low production cost, thus having significant healthcare economic advantages. Based on this background and our research findings, we started a phase II investigator-initiated clinical trial in September 2023 aimed to evaluate the efficacy and safety of combination therapy with nivolumab and TM5614 as third-line or later treatment for advanced NSCLC.
In conclusion, PAI-1 is involved in the survival of aPD-1-TPs via EMT and by modifying the TIME status. Thus, PAI-1 is a potential therapeutic target to overcome tolerance to aPD-1 antibodies and improve the prognosis of patients with advanced NSCLC.

Supplementary Material
Supplementary Table S1Supplementary Table S1 shows the clinical characteristics of the four examined patients.

Supplementary MethodsSupplementary Materials and Methods shows Chemical Structure and Representative Synthesis of TM5614

Supplementary ReferencesSupplementary References list the references describing the chemical structure and representative synthesis of TM5614.

Supplementary Figure S1Supplementary Figure S1 shows the experimental scheme for aPD-1 antibody-tolerant cells and the induction of PAI-1 expression during tumor regrowth.

Supplementary Figure S2Supplementary Figure S2 shows tumor expression level of PAI-1 in surgically resected NSCLC samples.

Supplementary Figure S3Supplementary Figure S3 shows PAl-1 protein levels in the culture medium of control and TM5614 groups, and Effect of TM5614 on the lung cancer cells.

Supplementary Figure S4Supplementary Figure S4 shows that SK-216 significantly increases the antitumor efficacy of aPD-1.

Supplementary Figure S5Supplementary Figure S5 shows tumor expression of mesenchymal and epithelial genes.

Supplementary Figure S6Supplementary Figure S6 shows the mechanism of PAI-1 involvement in tolerance to aPD-1 antibody treatment, and TGF-β expression in Control and aPD-1 antibody group.

Supplementary Figure S7Supplementary Figure S7 shows LLC cells expression of mesenchymal and epithelial genes in Control and SB-431542 group.

Supplementary Figure S8Supplementary Figure S8 shows relationship of TGF-β secretion with TAMs and CAFs.

Supplementary Figure S9Supplementary Figure S9 shows tumor microenvironment of tumors composed of LLC and KLN205 cells.

Supplementary Figure S10Supplementary Figure S10 shows degree of angiogenesis in the tumors treated with aPD-1 antibody or aPD-1 antibody with TM5614.

Supplementary Figure S11Supplementary Figure S11 shows schematic illustration of our study.