In the present study, two publicly available datasets, namely the Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) cohort and The Cancer Genome Atlas (TCGA)-BRCA provisional cohort, were used. The METABRIC dataset on breast cancer (21,22) and the TCGA dataset on breast invasive carcinoma [TCGA, PanCancer Atlas (23)] were obtained from cBioportal (https://www.cbioportal.org/) in August 2023. PD-L1 expression analysis was conducted on the METABRIC (total n=1,905, basal n=296) and TCGA (total n=1,084, basal n=171) datasets according to subtype and TN [tumor size (T) and lymph node status (N)] stage. Subsequently, survival analysis was performed on patients with TNBC (METABRIC, n=233; TCGA, n=163) by dividing them into high and low-PD-L1 expression groups. PD-L1 mRNA expression was quantified using z-scores, which represent the number of standard deviations a value is from the mean expression across all samples. Patients were classified as high (z-score ≥0) or low (z-score <0) PD-L1 expression accordingly. Overall survival was estimated using the Kaplan-Meier method, and statistical differences between groups were assessed using the log-rank test.

The mouse TNBC cell line, 4T1 [cat. no. CRL-2539; American Type Culture Collection (ATCC^®)], and the mouse colon carcinoma cell line, CT26 (cat. no. CRL-2638; ATCC^®), were used in this study. CT26 cells were used as a source of mouse wild-type PD-L1 cDNA, which was amplified by PCR for subsequent cloning and overexpression in 4T1 cells. The cells were cultured in RPMI 1640 medium (Welgene, Inc.) and Dulbecco's Modified Eagle's Medium (Welgene, Inc.) supplemented with 10% fetal bovine serum (FBS; Welgene, Inc.) and 1% penicillin-streptomycin (10,000 U/ml; Gibco; Thermo Fisher Scientific, Inc.). The cells were maintained at 37°C in a humidified atmosphere of 95% air and 5% CO₂.

The murine colorectal carcinoma cancer cell, CT26, derived from BALB/C mice, was used to establish the PD-L1 overexpressing stable cell line. Mouse PD-L1 RNA was extracted from the CT26 cells with TRIzol (FAVORGEN Biotech Corp.), and the mouse PD-L1 gene (RefSeq: NM_021893.3) was amplified through reverse transcription-polymerase chain reaction (RT-PCR). Reverse transcription was performed using PrimeScript™ RT Reagent Kit (cat. no. RR037A; Takara Bio, Inc.) according to the manufacturer's instructions. The following two primers, incorporating a Kozak sequence for efficient translation and NheI and NotI restriction sites for cloning, were used: 5′-GCTAGCGCCACCATGAGGATATTTGCTGGCATT-3′ (forward) and 5′-GCGGCCGCTTACGTCTCCTCGAATTGTGT-3′ (reverse). The PCR product was ligated into the pMD20-T vector (Takara Korea Biomedical Inc.) to create restriction enzyme sites. The pMD20-T vector was transformed into Escherichia coli DH5α competent cells (cat. no. CP011; Enzynomics, Inc.). The amplified PD-L1 gene with restriction enzyme sites was inserted using the NheⅠ and NotⅠ sites in a PiggyBac vector (System Biosciences, LLC). The constructed PiggyBac vector containing the mouse PD-L1 gene and an empty PiggyBac control vector were transfected into 4T1 cells using JetPrime (Polyplus-transfection SA). Transfection was performed in 100 mm dishes using 10 μg of plasmid DNA per dish at 37°C for 24 h, according to the manufacturer's instructions. After transfection, stable PD-L1 overexpressing cell lines were established by treating the cells with 5 μg/ml puromycin starting 48 h later. Puromycin was used at 5 μg/ml for selection and 1 μg/ml for maintenance in both PD-L1-overexpressing and control (empty PiggyBac vector) 4T1 cells. Following antibiotic selection, GFP-positive cells were collected and PD-L1 overexpression was confirmed by both surface staining using APC-conjugated anti-mouse PD-L1 antibody (1:100; cat. no. 124312; BioLegend, Inc.) and western blot analysis with anti-PD-L1 antibody (1:1,000; cat. no. 60475; Cell Signaling Technology, Inc.). To isolate a subpopulation of cells with intermediate PD-L1 expression, cells were subsequently sorted using a FACSAria III cell sorter (BD Biosciences) operated with FACSDiva software (version 9.2; BD Biosciences), based on GFP and PD-L1 fluorescence intensity. For surface PD-L1 staining, cells were incubated with APC-conjugated anti-mouse PD-L1 antibody (1:100; cat. no. 124312; BioLegend, Inc.) in FACS buffer (PBS containing 2% FBS) for 30 min at 4°C in the dark. After staining, cells were washed twice with PBS and filtered through a 40 μm cell strainer prior to sorting. Cells were gated into subpopulations with low, medium or high PD-L1 expression levels based on GFP and PD-L1 fluorescence intensity. The sorted sublines were subsequently expanded, and PD-L1 expression levels were validated by western blot analysis using an anti-PD-L1 antibody (1:1,000; cat. no. 60475; Cell Signaling Technology, Inc.) as described below. For experimental consistency, the 4T1 control population with no PD-L1 overexpression was designated as 4T1-Ctl, the PD-L1 overexpression medium-expression group as 4T1-PD-L1 O/E (Medium level) and the highest-expression group as 4T1-PD-L1 O/E.

Cells were lysed in RIPA buffer (MilliporeSigma) mixed with a protease inhibitor cocktail (Thermo Fisher Scientific, Inc.) and 0.5 M Tris-EDTA (Thermo Fisher Scientific, Inc.). Protein concentrations were determined using the BCA Protein Assay Kit (Thermo Fisher Scientific, Inc.), and 20 μg of total protein was loaded per lane. The prepared proteins were separated by 10% sodium dodecyl sulfate/polyacrylamide gel electrophoresis and transferred onto Immobilon-P transfer membranes (Merck KGaA). Membranes were blocked with 5% skim milk in TBST (TBS containing 0.1% Tween-20) for 1 h at room temperature. Membranes were incubated with the following primary antibodies overnight at 4°C: Anti-PD-L1 (1:1,000; cat. no. 60475, Cell Signaling Technology, Inc.) and anti-β-actin (1:5,000; cat. no. sc-4778, Santa Cruz Biotechnology, Inc.). After washing, membranes were incubated with the following HRP-conjugated secondary antibodies for 1 h at room temperature: Goat anti-rabbit IgG-HRP (1:5,000; cat. no. sc-2004, Santa Cruz Biotechnology, Inc.) and goat anti-mouse IgG-HRP (1:5,000; cat. no. sc-2005, Santa Cruz Biotechnology, Inc.). After membrane incubation with specific antibodies, the signal was enhanced with chemiluminescence reagents (Thermo Fisher Scientific, Inc.) and measured using an Amersham Imager 680 (GE Healthcare). The relative values of the bands observed by western blotting were analyzed using ImageJ software (version 1.53j; National Institutes of Health).

Total RNA was extracted from the cells using the Tri-RNA reagent (FAVORGEN Biotech Corp.). cDNA was synthesized from the extracted RNA using the PrimeScript™ RT Reagent kit (cat. no. RR037A; Takara Korea Biomedical Inc.; Takara Bio Inc.), according to the manufacturer's instructions. qPCR was conducted on a QuantStudio™ 3 Real-Time PCR System (Applied Biosystems; Thermo Fisher Scientific, Inc.) using SYBR Green PCR master mix (cat. no. 4367659; Applied Biosystems; Thermo Fisher Scientific, Inc.). The thermocycling conditions were as follows: Initial denaturation at 95°C for 10 min, followed by 40 cycles of denaturation at 95°C for 15 sec and annealing/extension at 60°C for 1 min. The relative mRNA levels were measured using the 2^−ΔΔCq method (24), with Gapdh as the control. The following primer sequences were used: Pd-l1, (forward) 5′-GCGGACTACAAGCGAATCACG-3′ and (reverse) 5′-CTCAGCTTCTGGATAACCCTCG-3′; Gapdh, (forward) 5′-CATCACTGCCACCCAGAAGACTG-3′ and (reverse), 5′-ATGCCAGTGAGCTTCCCGTTCAG-3′.

The cells were harvested using 0.25% trypsin and washed with PBS. The prepared cells were suspended in 1% BSA or PBS. Staining was conducted using a mouse PD-L1 antibody conjugated to phycoerythrin (cat. no. 124307; BioLegend, Inc.) and 7-AAD (cat. no. 00-6993-50; Invitrogen; Thermo Fisher Scientific, Inc.) for 20 min at room temperature in the dark. The mean fluorescence intensity was detected using a BD FACSCanto™ Flow Cytometer (BD Bioscience)., and the data were analyzed using BD FACSDiva software (version 8.0.1; BD Biosciences; https://www.bdbiosciences.com/).

For migration assays, 1×10⁵ cells were seeded in an 8.0-μm pore size insert (Falcon^®; Corning Life Sciences) with serum-free medium. For invasion assessment, an insert pre-coated with 1 mg/ml Matrigel (by incubation for 20 min at 37°C) was used, with 1×10⁵ cells in serum-free medium. A medium containing 10% FBS was added to the lower chambers. After incubating for 12 h at 37°C, cells that migrated were fixed with 4% paraformaldehyde for 10 min at room temperature and stained with a 0.1% crystal violet for 2 min at room temperature. Stained cells were imaged using a light microscope equipped with a CCD camera (Leica Microsystems, Inc.) and quantified with ImageJ software.

For the gap closure assay, 5×10⁴ cells were seeded in a culture-insert in a 2-well μ-dish (Ibidi GmbH) and incubated overnight at 37°C. After reaching confluency, a gap was made by removing the insert, the plate was washed with PBS and 500 μl medium containing 10% FBS (25) was added. Gap closure was observed from 0 to 12 h using a light microscope equipped with a CCD camera (Leica Microsystems, Inc.).

The CellTiter-Glo^® Luminescent Cell Viability Assay Kit (Promega Corporation) was used to assess cell proliferation. A total of 3×10³ cells were seeded into each well of a 96-well plate and incubated at 37°C for 24, 48 and 72 h. Afterward, 100 μl of the CellTiter-Glo solution was added to the cells following the manufacturer's guidelines. Luminescence was measured using a Luminiskan ascent (Thermo Fisher Scientific, Inc.).

Tumors were digested in collagenase and hyaluronidase for 30 min in a 37°C shaking incubator, followed by an RBC lysis buffer treatment (Thermo Fisher Scientific, Inc.). The cells were filtered through 40- and 75-μm cell strainers, and 1×10⁶ cells were suspended in the staining buffer (BioLegend, Inc.). Cells were pre-incubated with 0.25 μg of the TruStain FcX™ PLUS (anti-mouse CD16/32) antibody (BioLegend, Inc.) for 5-10 min. Antibodies were then incubated on ice for 20 min in the dark. Fluorescence intensity was detected using a BD FACSCanto™ Flow Cytometer (BD Bioscience), and the data were analyzed using BD FACSDiva software (version 8.0.1; BD Biosciences; https://www.bdbiosciences.com/). The following antibodies from BioLegend, Inc. were used for lymphoid panel staining: PE anti-mouse CD3 (cat. no. 100206), Brilliant Violet 421 anti-mouse CD4 antibody (cat. no. 100437), Brilliant Violet 510 anti-mouse CD45 antibody (cat. no. 103137) and PerCP/Cyanine5.5 anti-mouse CD8a antibody (cat. no. 100733).

In vivo experiments were conducted at the animal facility of Seoul National University Hospital (Seoul, South Korea), following guidelines and obtaining prior approval from the Institutional Animal Care and Use Committee (IACUC; approval no. 23-0157-S1A0). Animals were maintained in the facility-accredited AAALAC International (no. 001169) in accordance with Guide for the Care and Use of Laboratory Animals 8th edition, NRC (2010). In total, 54 6-week-old female BALB/c mice (body weight, 18-22 g) were obtained from Koatech. Mice were housed under specific pathogen-free conditions at 22±2°C with 55±10% humidity, under a 12 h light/dark cycle, with ad libitum access to food and water. Animal health and behavior were monitored daily. Syngeneic mouse models were established by injecting 1×10⁵ control 4T1, PD-L1-overexpressing 4T1 (medium level), or PD-L1 overexpressed 4T1 cells into the mammary gland fat pads. Tumor size was measured twice weekly using digital calipers and calculated with a modified ellipsoidal formula [volume=1/2 (length x width²)]. In this experiment, two groups of 4T1 tumor-bearing mice were used: One injected with control 4T1 cells and the other with 4T1 cells overexpressing PD-L1 at a high level. In a subsequent analysis, an additional group bearing 4T1 tumors with moderate PD-L1 overexpression was included, resulting in three experimental groups: Control, medium PD-L1 expression and high PD-L1 expression. Each group was treated with either control IgG or anti-PD-L1. The PD-L1 inhibitor, a monoclonal anti-PD-L1 antibody (cat. no. BP0101; 150 μg/100 μl) and immunoglobulin G (IgG; cat. no. BP0090; 150 μg/100 μl) were administered intraperitoneally a total of three times, starting when tumor volumes reached ~100 mm³. The experiments were concluded after three intraperitoneal injections of either IgG or anti-PD-L1, on days 6, 9 and 12. Additionally, any mouse with a tumor volume reaching 1,000 mm³ was scheduled for immediate euthanasia. The study was designed with an appropriate sample size to account for biological variability in accordance with IACUC guidelines, and the animal was included in the analysis. The experiments lasted for a total of 17 or 13 days, with the shorter duration resulting from earlier tumor progression in some mice necessitating humane endpoint euthanasia (tumor volume reached 1,000 mm³). At the end of the study, mice were euthanized following deep anesthesia using isoflurane (2-5% for induction, 1-3% for maintenance). CO₂ inhalation was used to complete euthanasia at a displacement rate of 30-70% of the chamber volume per minute, following NIH and AVMA (2020) guidelines. Death was confirmed by the cessation of respiration, which was monitored continuously for at least 10 min following CO₂ exposure to ensure complete loss of vital signs. After euthanasia, tumors, lungs and spleens were resected and then preserved in 4% paraformaldehyde at 4°C overnight or Bouin's solution at room temperature overnight for further analysis.

Control and PD-L1-overexpressing 4T1 murine breast cancer cells were seeded on 8-well chamber slides (cat. no. 154534; Nunc; Thermo Fisher Scientific, Inc.) and cultured in RPMI-1640 medium supplemented with 10% FBS and 1% penicillin-streptomycin at 37°C in a humidified incubator with 5% CO₂. Cells were fixed with 4% paraformaldehyde in PBS for 15 min at room temperature and permeabilized with 0.2% Tween-20 in PBS for 10 min. Non-specific binding was blocked with 1% BSA (cat. no. AC1025-100-00; Biosesang) in PBS for 1 h at room temperature. Cells were incubated overnight at 4°C with a fluorescence-conjugated anti-PD-L1 antibody (1:200; cat. no. MAB90782AF647; Novus Biologicals, LLC) diluted in blocking buffer. The next day, cells were washed three times with PBS. Nuclei were counterstained with DAPI (1 μg/ml; cat. no. S7113; Merck KGaA) for 5 min at room temperature. Slides were mounted using fluorescence mounting medium (cat. no. S3023; Dako; Agilent Technologies), and representative images were captured using a fluorescence microscope (ECLIPSE Ni-E; Nikon Corporation).

The primary tumors were fixed in 4% buffered paraformaldehyde at 4°C overnight, embedded in paraffin and sectioned into 4-μm slices. After being deparaffinized in xylene and rehydrated in graded ethanol and water solutions, antigen retrieval was performed using citrate (pH 6.0) or 10 mM Tris/1 mM EDTA (pH 9.0) at 95°C for 20 min. Endogenous peroxidase activity was blocked with 3% H₂O₂ for 20 min at room temperature. The sections were then blocked with 10% normal goat serum (cat. no. S-1000; Vector Laboratories, Inc.) for 1 h at room temperature to prevent non-specific binding, followed by overnight incubation at 4°C with primary antibodies. The antibodies used were as follows: CD45 (1:100; cat. no. AF114; R&D Systems, Inc.), CD4 (1:200; cat. no. ab183685; Abcam), CD8 (1:200; cat. no. ab209775; Abcam) or PD-L1 (rabbit polyclonal; 1:200; cat. no. ab10558; Abcam). After washing, the sections were incubated with species-appropriate HRP-conjugated secondary antibodies for 1 h at room temperature. For rabbit and mouse primary antibodies, a ready-to-use EnVision+ System-HRP (Rabbit/Mouse) (cat. no. K5007; Dako; Agilent Technologies) was used. For the goat-derived CD45 antibody, rabbit anti-goat IgG(H+L)-HRP (1:500; cat. no. SA007-500; GenDEPOT, LLC) was used. Immunoreaction was visualized using a DAB chromogen kit included in the EnVision system (cat. no. K5007; Dako; Agilent Technologies, Inc.). Nuclei were counterstained with hematoxylin (cat. no. S330930-2; Dako; Agilent Technologies, Inc.) for 2 min at room temperature. Histological images were captured using a light microscope equipped with a CCD camera (Leica Microsystems, Inc.).

RNA was extracted from tumors with confirmed PD-L1 inhibitory effects, and transcriptome sequencing was performed by Macrogen, Inc. (Seoul, Korea). RNA quality and integrity were assessed using the 2100 Bioanalyzer System (cat. no. G2939BA, Agilent Technologies, Inc.) with a DNA 1000 Kit (cat. no. 5067-1504, Agilent Technologies, Inc.). The SureSelectXT RNA Direct Library Preparation Kit (cat. no. G7550A; Agilent Technologies, Inc.) was used for library preparation, followed by quality control of the raw sequencing reads. Sequencing was carried out on the NovaSeq 6000 platform (cat. no. 20012850, Illumina, Inc.) using paired-end 101-bp reads. Raw sequencing data were first evaluated using FastQC (version 0.11.7; http://www.bioinformatics.babraham.ac.uk/projects/fastqc/) to assess read quality. Adapter sequences and low-quality bases were trimmed using Trimmomatic (version 0.38; http://www.usadellab.org/cms/?page=trimmomatic). The cleaned reads were aligned to the Mus musculus reference genome (mm10) using HISAT2 (version 2.1.0; https://ccb.jhu.edu/software/hisat2/), with internal use of Bowtie2 (version 2.3.4.1). Transcript assembly was performed using StringTie (version 2.1.3b; https://ccb.jhu.edu/software/stringtie/). Differential expression analysis was conducted using DESeq2 (version 1.36.0; https://bioconductor.org/packages/release/bioc/html/DESeq2.html). Expression profiles were generated, and DEGs were identified based on a |fold-change|>2 and raw P<0.05. The P-values were adjusted for multiple testing using the Benjamini-Hochberg method, with a false discovery rate threshold of 0.05. DEGs were then functionally annotated using the Database for Annotation, Visualization and Integrated Discovery (https://david.ncifcrf.gov). Gene Ontology (https://geneontology.org/) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway (https://www.genome.jp/kegg/) analyses were subsequently performed.

To evaluate the data, graphs were constructed to display the mean ± standard deviation derived from a minimum of three independent experiments. Statistical comparisons between two groups were performed using unpaired t-tests and the Mann-Whitney U test. For experiments involving two independent variables (such as treatment, expression status or time), data were analyzed using two-way ANOVA followed by Tukey's post hoc test. In survival analysis, the Kaplan-Meier method with a two-sided log-rank test was used. Statistical analyses were performed using GraphPad Prism v9.2.0 (Dotmatics). P<0.05 was considered to indicate a statistically significant difference.

To understand the relationship between PD-L1 expression and clinicopathology in breast cancer subtypes, TCGA and METABRIC datasets were analyzed. It was observed that TNBC had a high level of PD-L1 expression in both TCGA and METABRIC datasets, similar to the HER2 subtype in TCGA dataset (Fig. 1A). The difference in overall survival between patients with TNBC divided into the high and low PD-L1 mRNA expression groups in the TCGA and METABRIC datasets was not statistically significant (Fig. 1B). Similarly, PD-L1 expression was not significantly associated with T stage and nodal involvement (Fig. S1). These results confirm that, although PD-L1 expression is elevated in TNBC, it may not be directly associated with poor patient prognosis based on public datasets.

To further elucidate the role of PD-L1 upregulation in breast cancer, a 4T1 murine breast cancer cell line that stably overexpressed PD-L1 was established. PD-L1 overexpression was confirmed at the protein and mRNA levels in the 4T1 PD-L1 overexpressed cell line (Fig. S2A and B). The cell surface expression of PD-L1 was further validated using flow cytometry and immunocytochemistry (Fig. S2C and D). To determine the phenotypic importance of PD-L1-overexpression, the proliferation, migration and invasion of the cells was examined. The result showed that PD-L1 O/E induced enhanced the proliferative capacity of 4T1 cells compared with the control cells (Fig. 2A). Additionally, the overexpression of PD-L1 led to significant increases in the migratory and invasive capabilities, as evidenced by the migration, invasion and gap closure assays (Fig. 2B and C). To investigate the therapeutic implications of PD-L1 overexpression, PD-L1-overexpressing 4T1 cells were treated with anti-PD-L1 antibodies. The effects of anti-PD-L1 treatment were not pronounced, with only modest changes in cell proliferation observed. Specifically, there was a slight reduction in proliferation at 24 h, followed by an increase at 48 h, though these changes were not statistically significant (Fig. 2D). This suggests that the interaction between PD-L1 signaling and tumor cell survival mechanisms may be complex, potentially involving compensatory pathways that mitigate the effects of PD-L1 blockade.

ICIs targeting PD-L1 could be applied when the PD-L1 expression level on tumor or immune cells in the tumor tissue is above a specific value. Most clinical benefits occur in patients with high levels of PD-L1 expression (12). However, not all patients with high PD-L1 levels show consistent treatment effects, suggesting that the relationship between PD-L1 expression and treatment efficacy is not yet fully understood (12). To further investigate the efficacy of anti-PD-L1 depending on PD-L1 expression level, 4T1 breast cancer cell lines with different PD-L1 expression levels were established and transplanted into syngeneic breast tumor mouse models to observe the effect of anti-PD-L1 on tumors. It was observed that anti-PD-L1 treatment had no antitumor effect on tumors generated from 4T1-PD-L1 O/E cell lines. In the 4T1-Ctl group with a very low PD-L1 expression level, an antitumor effect of anti-PD-L1 (Fig. 3A and B) was observed. Tumor weight measurements confirmed that PD-L1 overexpression reduced the anti-PD-L1 antitumor effect compared with the 4T1-Ctl group (Fig. 3C). Notably, in the control + anti-PD-L1 group, 1 animal died on day 10 of the study. Necropsy revealed no apparent abnormalities, and the animal was included in the analysis. Compared with tumors derived from 4T1 control cells with low PD-L1 expression, tumors with high PD-L1 expression exhibited a markedly weaker response to anti-PD-L1 treatment, indicating that PD-L1 overexpression reduced the antitumor efficacy of the therapy (Fig. 3A-C).

To clarify this observation, the anti-PD-L1 effect was further tested by additionally including a group with medium-PD-L1-expression [PD-L1 O/E (Medium)], which had a lower PD-L1 expression level than the previously described overexpression group. The medium-level group was collected and confirmed by GFP sorting and immunoblotting (Fig. 3D and E). Of note, anti-PD-L1 treatment was most effective in the 4T1-PD-L1 O/E (Medium level) group and the treatment outcomes were similar to those observed in the 4T1-Ctl group; in some tumors, treatment exhibited improved antitumor effects (Fig. 3F and G). The improved antitumor effect in the 4T1-PD-L1 O/E (Medium level) group was also reflected in the final tumor weight measurements (Fig. 3H).

These findings demonstrate that PD-L1 expression is essential in determining the efficacy of anti-PD-L1 therapies. Excessive PD-L1 levels diminish treatment response, while the medium-level exhibited the greatest sensitivity to tumor treatment. This suggests that there may be an optimal PD-L1 expression range to maximize the effect of anti-PD-L1 therapy. This research emphasizes the need to further refine patient selection criteria for anti-PD-L1 therapy, as PD-L1 expression levels above a certain level alone may not be sufficient to predict treatment success.

PD-L1 expression and treatment responses are related to T cell infiltration into tumors (26). In this regard, based on the aforementioned results, it was evaluated whether there was a difference in immune cell infiltration according to anti-PD-L1 treatment between the control group and the PD-L1-overexpressing tumor group. It was first confirmed whether PD-L1 was expressed at a high level in tumors overexpressing PD-L1 (Fig. 4A). To understand the relationship between the anti-PD-L1 effect and immune cell infiltration, immunohistochemistry staining was performed for CD45, a total immune cell marker, on tumor sections. It was observed that the control tumors, in which anti-PD-L1 treatment exhibited a notable antitumor effect, were infiltrated with immune cells (Fig. 4B). By contrast, the PD-L1-overexpressing group, in which anti-PD-L1 treatment exhibited a weak antitumor effect, had fewer infiltrated immune cells than the control tumors (Fig. 4B). Additionally, high levels of immune cell infiltration were also observed in PD-L1 overexpressing tumors treated with IgG. This result suggests that PD-L1 overexpression was not related to total immune cell infiltration.

The group differences in the distribution of CD4⁺ and CD8⁺ T cells were analyzed, which are associated with the anti-PD-L1 antitumor effect among immune cells. In the control tumors, where anti-PD-L1 treatment was highly effective, infiltrated CD4⁺ and CD8⁺ T cells were more numerous that in the other groups (Fig. 4C). By contrast, tumors overexpressing PD-L1 exhibited only a slight increase in CD4⁺ and CD8⁺ T cell infiltration levels after anti-PD-L1 treatment compared with the control (Fig. 4C). Notably, in these tumors, anti-PD-L1 treatment lowered the infiltration of both types of T cells compared with the control anti-PD-L1 treatment group. For validation of the aforementioned results, flow cytometry was performed. The control group treated with anti-PD-L1 had higher CD4⁺ and CD8⁺ T cell infiltration levels, whereas the tumors treated with anti-PD-L1 that overexpressed PD-L1 had reduced T cell distribution (Figs. 4D and S3).

To understand the underlying mechanism for these observations, the tumor samples were subjected to bulk RNA sequencing and group specific DEGs were applied to the pathway analysis. Using the identified DEGs and pathways, the DEGs were compared between the control tumors treated with IgG and those treated with anti-PD-L1. This method led to the identification of pathways specifically upregulated by anti-PD-L1, independent of PD-L1 overexpression. To assess downregulation in primary resistance, DEGs were also compared between control tumors and PD-L1-overexpressing tumors, both treated with anti-PD-L1. Analysis of the upregulated genes in the control + anti-PD-L1 group showed enhanced signaling related to T cell activation, with increased PD-L1 and PD-1 immune check point pathway activities, as well as upregulation of Th17 and Th1/Th2 cell differentiation KEGG pathways (Fig. 4E). By contrast, the PD-L1 O/E + anti-PD-L1 group exhibited significant downregulation of genes involved in T cell signaling, the PD-L1-PD-1 pathway and various immune cell-related signaling pathways, compared with the control treatment group (Fig. 4F). These results suggest that while PD-L1 upregulation contributes to immune cell infiltration within tumors, excessive PD-L1 levels may impair T cell activation and overall immune responses, thereby compromising the efficacy of anti-PD-L1 therapy. These findings highlight the importance of PD-L1 expression levels in determining the success of the immune checkpoint blockade and suggest that excessively high PD-L1 expression levels may lead to an immunosuppressive microenvironment that impedes the therapeutic effect of PD-L1 inhibitors.