TIMELESS upregulates PD-L1 expression and exerts an immunosuppressive role in breast cancer

The cancer genome atlas (TCGA) datasets were obtained by the R package “TCGAbiolinks” and divided into two parts using the median expression of TIM as a cutoff. One thousand ninety-seven samples of breast cancer were collected. Counting data is converted to transcripts per million (TPM) and normalized log₂ (TPM + 1) while keeping clinical information intact. The data on survival time and clinical characteristics were also obtained. The RNA sequencing data was the same in the article as our lab previously published. The NCBI (National Center for Biotechnology Information)-GEO datasets GSE161529, and GSE130472 and their figures were obtained from TISCH and TISMO datasets. Data of GSE46141 was obtained from NCBI.

Both human and mouse breast cancer cell lines were obtained from Renji Hospital, Shanghai Jiaotong University, School of Medicine, including MCF-7, T-47D, MDA-MB-231, BT-20, and 4T1. All the cells except MDA-MB-231 and T-47D were maintained in DMEM medium (Gibco, USA), supplemented with 10% Fetal Bovine Serum (Gibco, USA) and 1% Penicillin–Streptomycin (Gibco, USA). MDA-MB-231 was maintained in Leibovitz's 15 medium (Gibco, USA) while T-47D was maintained in RPMI 1640 (Gibco, USA), both supplemented with 10% Fetal Bovine Serum (Gibco, USA) and 1% Penicillin–Streptomycin (Gibco, USA). Cells were cultured at 37 ℃ with 5% CO₂.

Western blotting: TIMELESS (Abcam, ab109512, 1/50000), β-actin (Abcam, ab49900, HRP-linked,1/50000), c-Myc (Abcam, ab32072, 1/1000) and PD-L1 (Cell Signaling Technology, #13684, 1/1000), Rabbit secondary antibody (Cell Signaling Technology, #7074, 1/2000).

Immunohistochemistry: TIMELESS (Abcam, ab72458, 1/200), CD8a (Servicebio, GB13429, 1/200), PD-L1 (Servicebio, GB11339A, 1/500), Ki-67 (Servicebio, GB111141, 1/500), Mouse secondary antibody (Servicebio, GB23301, 1/200), Rabbit secondary antibody (Servicebio, GB23303, 1/200).

Immunofluorescence: TIMELESS (Abcam, ab109512, 1/100), c-Myc (Abcam, ab32072, 1/100), CD8a (Servicebio, GB13429, 1/5000), Granzyme B (Abcam, ab255598, 1/2000), active caspase 3 (Servicebio, GB11532, 1/500), Hoechst 33342 (Beyotime, C1027, 1/100), CY3-TSA (Apex Bio, K1051, 1/200), Donkey Anti-Rabbit IgG H&L Alexa Fluor^® 647 (Abcam, ab150075, 1/200).

Flow cytometry: PD-L1 (Cell Signaling Technology, #13684, 1/200), CD8a-PerCP/Cyanine5.5 (Clone 53-6.7, Biolegend 1007341/100), CD45-AF700 (Biolegend 103128, 1/80), Fixable Viability Dye eFluor™ 450 (Invitrogen, 1/500).

To knock down TIM, the small interference RNAs (siRNAs) siTIM, sic-Myc and its control of both humans and mice were purchased from Genepharma Shanghai, China, and transfected using Polyplus Transfection jetPRIME (France) according to the manufacturer’s instructions. After 48 h, cells were collected for the following research. To stably knock down TIM (shTIM), cells were infected with lentivirus vectors with short hairpin RNAs (shRNA) targeting mouse TIM sequences and empty plasmid as control. To stably overexpress TIM (oeTIM), the human TIM cDNA was cloned into the lentiviral expression vector GV341 to construct the vector GV341-hTIMELESS. The recombinant lentivirus was purchased from Genechem Shanghai, China, and used according to the protocol. After infection, puromycin (0.5 µg/ml, Solarbio) was used to select stably transduced cells. All the RNAs and plasmid sequences are listed in Additional file 1: Tables S1 and S2.

A cohort of 30 formalin-fixed paraffin-embedded human breast cancer specimens was used. Breast cancer tissues were collected at Renji Hospital during 2021–2022. IHC staining was performed using 5 μm sections of the formalin-fixed paraffin-embedded human breast cancer cutting onto glass slides. IHC for each protein was performed with the anti-TIM and anti-Ki-67 antibody on tumor cells, anti-CD8a antibody on lymphocytes, and PD-L1 on tumor and lymphocytes according to the standard protocol. The relative expression levels of TIM and Ki-67 were calculated by the percentage of the positive cells multiplied by the intensity of protein expression and finally normalized as percentage. The ratio of the positive cells was estimated for CD8a. Five areas for each slide were calculated into quantitative analysis.

Female BALB/c mice (aged five weeks) were obtained from Shanghai Charles River Animal Co., Ltd. Mice were grouped into four animals. They were maintained according to the criteria outlined in the Guide for the Care and Use of Laboratory Animals (23 ℃, 50% humidity, 12/12 h light/dark). 1 × 10⁶ treated 4T1 cells in 100 μL PBS were subcutaneously injected into the fat pad of the 4th mammary gland of the mice. The tumor diameters were monitored every five days. Tumor volumes (V) were measured using the long diameter (a) and the perpendicular diameter (b): V = a × b²/2. Mice were sacrificed at the end of the experiment or once the tumor tissue volume reached 1500 mm³. Then the tumor was isolated, weighed, and treated for the following experiments.

The total RNA of cells or tissues was extracted by SimplyP Total RNA Extraction Kit (BioFlux, USA). Total RNA was resuspended in RNase-free water in the kit. The RNA concentration was determined by spectrophotometry using a Nanodrop device (Thermo scientific Nanodrop 2000/2000c, wavelength between 230 to 340 nm). RNA expression levels were quantified by RT-qPCR. Reverse transcription was performed using 200 ng RNA with the HiScript II Q Select RT superMix for the qPCR kit (Vazyme, R233-01). RT-qPCR was performed on 2 ng of cDNA samples using Light Cycler 480 SYBR Green I Master Mix (Roche) supplemented with 0.625 μM forward primer and 0.625 μM reverse primer, and fluorescence was measured by Light Cycler 480 II instrument (Roche). For RNA expression levels, relative quantification was performed using the 2^−ΔΔCt method. β-actin was used for internal reference. Sequences of the primers can be found in Additional file 1: Table S3.

Cell samples were lysed in fresh RIPA Lysis Buffer 10X (Beyotime, P0013C) with protease inhibitor cocktail 100X (Sigma) and phosphatase inhibitor cocktail 100X (Sigma). Western blot membranes were incubated with the corresponding primary antibodies. β-actin was used as a loading control. Rabbit secondary antibody was used. Enhanced chemiluminescence (ECL) was performed by Immobilon Western Chemiluminescence HRP Substrate (Millipore, Billerica, USA) with ChemiDoc Touching Imaging System (Bio-rad, California, USA) and finally quantified by Image Lab software (Version 6.0.1, Bio-rad). During quantification, β-actin was used for internal reference.

CCK-8 and clonal formation assays were performed as previously described [22].

In the transwell migration assay, 2 × 10⁴ cells were suspended in 100 µl serum-free DMEM in the top compartment of the transwell chamber (Corning). In the lower compartment, 800 µl of DMEM with 10% Fetal Bovine Serum was added, and siRNA was added 24 h before adding cells to the chambers. Once the cells had been fixed with 4% paraformaldehyde for a further 24 h, they were stained with crystal violet, and the non-migrating cells were carefully removed. Images were taken using a light microscope at 20 × magnification. Transwell invasion assay was performed the same way as transwell migration assay, except matrigel was added to the upper chamber.

Cells were digested and regenerated into single-cell suspensions. Then cells were incubated with the corresponding first antibodies for 30 min and second antibodies for 30 min in the dark and detected by flow cytometry. 0.1% Triton X-100 (Sigma-Aldrich) was used for intranuclear staining. When detecting the immune cells in tumor tissue, after being separated from the mice, the tumor was minced with scissors and dissociated with collagenase IV (Sigma) at 1 mg/ml and collected by a 70 μm filter. After the digestion mixture was centrifuged and regenerated into suspensions by PBS, ficoll (GE healthcare, Ficoll-Paque PLUS) was added to the suspensions and centrifuged. Only the cells in the medium layer were picked for the following experiment. Cells were washed and data were acquired by Fortessa X-20 flow cytometry (BD Biosciences). All the flow cytometry data were processed with FlowJo software (version 10.4).

Cells, frozen sections, or paraffin sections were fixed with 4% paraformaldehyde (PFA) and washed with PBS 3 times. For tumor mass, 5 μm sections of the mass were cut. Cells or slides were stained with Hoechst stain and corresponding antibodies as shown in the Antibodies subsection in Materials and Methods section according to the manufacturers’ protocols. Fluorescent images were captured using a Leica DM2500 microscope.

BALB/c female mice aged 8 weeks were sacrificed and their spleens were used to separate CD8⁺ T cells by mouse CD8a (Ly-2) microbeads isolation kit (Miltenyi Biotec, USA) according to the manufacturer's protocol. For tumor-specific CD8⁺ CTL generation, the purified CD8⁺ T cells were cultured in 5 μg/ml CD3 (Biolegend, USA) and 5 μg/ml CD28 (Biolegend, USA) in RPMI 1640 with 10% heat-inactivated FBS at 37 ℃ in 5% CO₂ for 72 h to generate CTLs. Then the CD8⁺ T cells and CTLs were mixed with tumor cells at a ratio of 10:1 and were cultured in RPMI 1640 with 10% heat-inactivated FBS at 37 ℃ in 5% CO₂ for 48 h for the following experiments. Annexin V-AbFluor™-488 /PI apoptosis detection kit was purchased from Abbkine (USA) and detected according to the protocol.

The immunoprecipitation was done by a classic magnetic protein A/G IP/Co-IP Kit (YJ201, epizyme). Cells on 10 cm culture dishes were washed three times with PBS and lysed in 0.7 ml lysis buffer, and protease inhibitor mixture (Sigma). Insoluble material was removed after centrifugation at 4 ℃ for 15 min, 12000 rpm. Antibody for immunoprecipitation of each protein was added to the cell lysates and incubated for 1 h at room temperature. Then the antibody-antigen complex was incubated with protein‐A/G‐agarose beads for one hour at room temperature. Then the immunoprecipitated proteins were washed with lysis buffer and then denatured with loading buffer at 100 ℃ for 10 min. Then samples were used to make a Western blotting analysis as mentioned before.

The suspension of cells was collected, and the IFN-γ was assessed by the IFN-γ mouse uncoated ELISA kit (EK280/3-96, liankebio) according to the product instructions.

Cells were digested with trypsin without EDTA and resuspended into 1 × 10⁶ cells per group. Then cells were treated with the Annexin V-AbFluor™ 488/PI Apoptosis Detection Kit (KTA0002, Abbkine) according to the product instructions.

GSEA analysis was carried out with the software GSEA (version 4.2.3) and with the R package “clusterProfiler”. The Hallmark gene sets, Gene Ontology gene sets, REACTOME gene sets, BIOCARTA gene sets, and KEGG (Kyoto Encyclopedia of Genes and Genomes) gene sets were used as the molecular signatures databases “MSigDB Collections”. The core enrichment gene expression plot was also obtained by GSEA software. The circle plot of GSEA was drawn by R package “clusterProfiler”, “ggplot2”, “circlize”, “grid”, “graphics” and “ComplexHeatmap”. The R software version was R 3.6.3.

Estimation of immune fractions of TIM was determined through breast cancer tissue in the TCGA database using MCPcounter by R package “ebecht/MCPounter”. The ssGESA was also applied to explore the different infiltration degrees of the immune cells by R package “GSVA”. The data was tested by the Student’s t-test and visualized using R software.

Potential ICI response was predicted with the TIDE algorithm (http://​tide.​dfci.​harvard.​edu/​). The χ² test was calculated and visualized using R software.

Our previous study revealed that TIM interacts with specificity protein 1 (Sp1) and upregulates sphingolipid synthesis, which contributes to breast cancer growth and activates mitochondrial respiration [22]. To further elucidate the potential mechanisms of TIM in breast cancer progression, we performed GSEA using RNA-seq data from TIM-knockdown MCF-7 cells and the corresponding control cells. GSEA indicated that TIM may be involved in orchestrating lipid metabolism-related pathways, immune-related pathways, and several other pathways (Fig. 1a). Further in-depth GSEA analysis based on the immune pathways showed that TIM downregulation was related to the immune responses to the tumor cell, chemokine signaling pathway, antigen processing, and presentation (Fig. 1b). The expression of core immune-related genes was enriched in the siTIM group (Additional file 1: Fig. S1a). The expression profiles of the core enrichment genes in each top-scoring pathway are shown (Fig. 1c; Additional file 1: Fig. S1b). We also performed GSEA using the TIM profile of the TCGA breast cancer database. The lymphocyte infiltration signature score was significantly lower in the TIM high group (Fig. 1d). Taken together, these results demonstrate that TIM plays a vital role in immunosuppression, from antigen presentation to immune attack. GSEA also revealed that DNA repair and metabolism-related signaling pathways were significantly enriched (Fig. 1e). In addition to these two functions, which have already been discovered in existing research, immune-related pathways were also enriched in accordance with our RNA-sequencing data.

Moreover, interferon-gamma response and cytokine-cytokine receptor interactions were downregulated in the high TIM expression group (Additional file 1: Fig. S1c). GSEA showed that TIM suppresses tumor-extrinsic factors such as chemokines that regulate T cell recruitment. To deepen our understanding of the intrinsic immune profiles of TIM, an immunophenogram was constructed based on three categories with their representative genes or gene sets: major histocompatibility complex (MHC) molecules, immunostimulators, and immunoinhibitors (Fig. 1f). The immunophenogram showed that TIM expression was positively correlated with immunoinhibitors and inversely associated with immunostimulators and MHC molecules. Other studies have also revealed the suppressive role of TIM in the regulation of immune functions. For example, other intrinsic immune factors such as tumor mutant burden (TMB) and neoantigen load (NAL) also showed higher expression levels in the high TIM group (Additional file 1: Fig. S1d).

To further reveal the potential immunosuppressive mechanisms of TIM, we examined the immune infiltration landscape of multiple immune cell types in high- and low-TIM groups. Subsequently, we evaluated the immune correlation using the MCPcounter algorithm, which revealed that CD8⁺ T lymphocytes were most significantly and inversely correlated with TIM expression (Fig. 1g). The infiltration levels of T cells and myeloid dendritic cells and cytotoxicity scores also decreased considerably in the high TIM group (Additional file 1: Fig. S1e). TIM expression had a stronger inverse correlation with CD8⁺ T lymphocytes (r = − 0.160, p = 0.021) than with T lymphocytes (r = − 0.061, p = 0.043) in the scatter plot (Additional file 1: Fig. S1f). The ssGESA algorithm was used to confirm the role of TIM in the breast cancer immune microenvironment. Based on ssGESA, we found that T cell activation, proliferation, differentiation, and T cell-mediated immunity were lower in the high TIM group. This result was consistent with that of CD8⁺ T lymphocytes, with a more significant p-value (Fig. 1h). Furthermore, using the TISCH database, we found that TIM was expressed at the highest level in CD8⁺ T lymphocytes among all cell clusters in the single-cell sequence data of GSE161529 (Fig. 1i–k). Taken together, we hypothesized that oncogene TIM promotes breast cancer progression by inhibiting CD8⁺ T lymphocyte infiltration.

We then validated TIM expression using CD8⁺ T lymphocyte infiltration levels in the RENJI breast cancer patient cohort. Both protein levels determined by IHC (Fig. 2a) and mRNA levels determined by RT-qPCR (Fig. 2b) of tumor tissues showed that higher TIM expression was associated with lower CD8a expression. We also evaluated and compared the pathological and clinical features of the RENJI cohort. We found that TIM expression inversely correlated with CD8⁺ T lymphocyte infiltration (Fig. 2c), which is in line with our results (Fig. 1g–k). TIM expression was positively correlated with Ki-67 expression (Fig. 2c).

To investigate the effect of TIM on tumor immunity, we established a subcutaneous tumor model of breast cancer using the mouse breast cancer cell line 4T1 in female BALB/c mice with immune integrity. TIM expression was stably knocked down in mouse shTIM 4T1 cells (Fig. 3a). We subcutaneously injected shTIM and shNC 4T1 cells in female BALB/c mice aged five weeks. Tumor growth was significantly suppressed in shTIM syngeneic mice compared to that in the shNC group (Fig. 3b–c). IHC staining also suggested that Ki-67 levels were significantly decreased in TIM-knockdown tissues compared to those in control tissues (Fig. 3d). Surprisingly, we found that two mice in the shNC group and one mouse in the shTIM group developed lung metastases. The 4T1 shNC syngeneic mice showed multiple pulmonary metastases in each case, whereas the 4T1 shTIM syngeneic mice showed a single pulmonary metastasis by hematoxylin–eosin staining (Fig. 3e). TIM expression was positively correlated with lung metastasis, which was also confirmed by the gene expression levels in different metastatic sites of breast cancer from the GSE46141 dataset (Additional file 1: Fig. S2). The CD8⁺ T lymphocyte infiltration level of the tumor was evaluated by flow cytometry analysis, which indicated an increased proportion of CD45⁺CD8⁺ T cells in the shTIM tumor tissues compared to that in the shNC tissues (Fig. 3f), similar to the data in the RENJI cohort mentioned previously (Fig. 2a–c). As an essential effector of antitumor immunity, cytotoxic lymphocytes require granzyme B (GB) to eliminate tumor cells through granule-mediated apoptosis [26], and GB expression levels reflect the cytotoxic activity of CD8⁺ T lymphocytes. CD8⁺ T lymphocyte infiltration and activity were both examined by immunofluorescence, and we found that TIM downregulation not only increased the percentage of CD8⁺ T lymphocytes but also increased GB release (Fig. 3g). It has been reported that antitumor immunity is positively associated with apoptosis in tumor tissue [27]. Thus, the expression levels of the apoptotic biomarker cleaved caspase 3 (CCA3) were explored. A few cells in the tumor tissues of the control group underwent apoptosis. However, strongly clustered CCA3 signals were observed in the shTIM syngeneic mice (Fig. 3h). These results indicated that TIM may facilitate breast cancer progression by attenuating CD8⁺ T lymphocyte infiltration and impairing CD8⁺ T lymphocyte activity.

To further validate the above results in vitro, we co-cultured 4T1 cells and their specific CD8⁺ T lymphocytes with or without cytotoxic antitumor activity. We performed an apoptosis assay, proliferation assay, and ELISA for IFN-γ production following co-culture (Additional file 1: Fig. S3). The cell apoptosis data obtained by flow cytometry showed that breast cancer cells gained a higher percentage of apoptosis in the TIM knockdown group than in the control group (Fig. 3i). In the proliferation assay, TIM knockdown increased the sensitivity of 4T1 breast cancer cells to the antitumor activity of CTLs (Fig. 3j). The level of IFN-γ secretion by CTLs increased in the TIM-knockdown co-culture system (Fig. 3k). These results suggest that TIM compromises the antitumor activity of CTLs.

Programmed cell death ligand 1 (PD-L1) is usually expressed on the tumor surface, is the main ligand of Programmed cell death 1 (PD-1), and can compromise T cell proliferation and active functions [28, 29]. As mentioned above, TIM can attenuate CD8⁺ T lymphocyte infiltration and immune activity, and we hypothesized that TIM may modulate PD-L1 expression. In our previous study, we found that TIM expression was higher in estrogen receptor (ER)-positive breast cancer cell lines than in ER-negative breast cancer. Therefore, we used a TIM knockdown model in ER-positive cells and a TIM-stable overexpression model in ER-negative breast cancer cell lines. RT-PCR and western blotting analyses showed that PD-L1 expression decreased after MCF-7 and T-47D cells were transfected with two different TIM siRNAs (Fig. 4a, b). A reduction in PD-L1 was also observed on the membrane by flow cytometry (Fig. 4c). Similarly, the overexpression in MDA-MB-231 and BT-20 cells enhanced PD-L1 expression at the mRNA (Fig. 4d), protein (Fig. 4e), and PD-L1 membrane levels (Fig. 4f). In the in vivo syngeneic mouse model, a decrease in PD-L1 expression due to TIM knockdown was also demonstrated at the protein level in shNC and shTIM 4T1 tumor cells (Fig. 4g).

It has been demonstrated that TIM can promote breast cancer cell progression [30] and that PD-L1 can exert its intrinsic oncogenic function in breast cancer [31]. We postulated that TIM could affect breast cancer cell progression by influencing PD-L1 expression. Transwell migration and invasion assays confirmed that TIM knockdown reduced the migration and invasion of MCF-7 and T-47D cells (Fig. 5a). In contrast, TIM overexpression increased the migration and invasion of MDA-MB-231 and BT-20 cells (Fig. 5b). Through CCK-8 and clonal formation assays, we found that the overexpression of PD-L1 could partly rescue the decrease in proliferation ability caused by TIM knockdown (Fig. 5c, d). The transwell assay also revealed that the overexpression of PD-L1 reversed the inhibitory effects of TIM knockdown on migration and invasion ability (Fig. 5e).

Finding that TIM enhanced PD-L1 expression, we explored the mechanism by which TIM regulates PD-L1 expression. TIM interacts with several transcription factors to influence the expression of crucial oncogenic genes, thereby contributing to breast cancer progression [22]. Thus, we speculated that TIM may act as a transcriptional cofactor to regulate PD-L1 expression. According to previous studies, TIM can increase the activity of c-Myc [19], and PD-L1 can be transcribed by the recruitment of c-Myc in breast cancer [12]. We confirmed the proven results that c-Myc knockdown led to the downregulation of PD-L1 in MCF-7 and BT-20 cells (Fig. 6a). Meanwhile, the protein expression and mRNA levels of TIM were not altered by c-Myc knockdown (Fig. 6a, b). Next, we performed a co-immunoprecipitation (co-IP) assay to determine the relationship between TIM and c-Myc, which showed that TIM specifically binds to c-Myc in MCF-7 and BT-20 cells (Fig. 6c). Immunofluorescence analysis indicated that TIM co-localized with c-Myc in the nuclei of MCF-7 and BT-20 cells (Fig. 6d). Western blotting results showed that PD-L1 overexpression induced by TIM overexpression was partially inhibited by c-Myc knockdown (Fig. 6e). Taken together, our results suggest that TIM can, to some extent, enhance the transcriptional activity of c-Myc to upregulate PD-L1 expression.

Based on our findings, we hypothesized that TIM expression may be associated with the response to immune therapy in breast cancer. To further evaluate the value of TIM in predicting responsiveness to immunotherapy, we employed the Tumor Immune Dysfunction and Exclusion (TIDE) algorithm in the TCGA-BRCA dataset to calculate the immunotherapeutic response rates of patients with different TIM expression levels. Approximately 36.6% (201/549) of patients in the high TIM expression group responded to immunotherapy compared to 26.1% (143/548) in the low TIM expression group (Fig. 7a). There was a significant inverse correlation between TIDE score and TIM expression (r = − 0.17, p = 1.4e−06; Fig. 7b). Patients with high TIDE scores may benefit from immunotherapy [32]. To explore which type of immunotherapy is affected by the expression pattern of TIM, we calculated the correlation between well-known immune checkpoint genes and TIM. Most immune checkpoint genes were positively correlated with TIM, among which the most relevant was CD274, which encodes the well-known PD-L1 protein (Fig. 7c).

We further investigated the TISMO database using ICI-treated mouse models. The 4T1 transplanted breast cancer mouse model in GSE130472 was subjected to anti-PD-L1 and anti-cytotoxic T-lymphocyte-associated protein 4 (CTLA4) therapies (Fig. 7d). The TIM expression level at baseline was the same as that after immune therapy, illustrating that the TIM expression level did not change despite the different immune treatments. Upregulated TIM expression yields a better response to anti-PD-L1 therapy than to anti-CTLA4 therapy. Next, we examined the correlation between PD-L1 and TIM (Fig.S4). TIM showed the strongest correlation with PD-L1 expression in TNBC (r = 0.436, p = 1.89e−09).

To explore the relationship between TIM and PD-L1 expression in terms of prognosis, we evaluated the effect of PD-L1 on overall survival (OS) based on TIM expression levels using the TCGA dataset. The survival analysis revealed that PD-L1 expression was not significantly associated with OS in the entire study population (Fig. 7e). To integrate the TIM status into the survival analysis, we separated the patients into low (0–33%), medium (34–66%), and high (67–100%) subgroups. In the high TIM group, OS was significantly prolonged with increased PD-L1 expression. In the medium and low TIM groups, the difference was not statistically significant. The assays using tumor tissue from the RENJI cohort also showed higher levels of PD-L1 at both protein (Fig. 7f) and mRNA levels (Fig. 7g) in the high TIM expression group. These results implied that breast cancer patients with high TIM expression may be more sensitive to anti-PD-L1 therapy.