2.1 Cell culture

Human NSCLC cell lines A549 and PC9 and human embryonic kidney cell line 293T were acquired from the American Type Culture Collection (Manassas, VA, USA), and cisplatin (DDP)-resistant A549^R and PC9^R cells from the Tianjin Medical University Cancer Institute and Hospital (Tianjin, China). All cells were verified to be mycoplasma negative via a GMyc-PCR mycoplasma test kit (Yeasen, Shanghai, China). A549 and PC9 cells were grown in RPMI-1640 (Gibco, Waltham, MA, USA), containing 10% fetal bovine serum (FBS) (Gibco) and 1% penicillin/streptomycin (ThermoFisher Scientific, Waltham, MA, USA). 293T cells were cultured in Dulbecco's Modified Eagle Medium (DMEM) (Gibco), containing 10% FBS, 1% sodium pyruvate (Gibco), 1% non-essential amino acids (NEAA) (Gibco) and 2% glutamine (Gibco). A549^R and PC9^R cells were cultured in RPMI-1640, containing 10% FBS, 1% penicillin/streptomycin and 2.5 μmol/L DDP (Selleck, Houston, TX, USA).

2.2 Reagents and drugs

PD-L1 neutralizing antibody avelumab (Selleck), the dosages were 2 μg/mL in vitro and 5 mg/kg in vivo; the NF-κB inhibitor Bay 11-7085 (TOPSCIENCE, Shanghai, China), the dosage was 5 μmol/L in vitro; the protein synthesis inhibitor cycloheximide (CHX) (Selleck), the dosage was 100 μg/mL in vitro; the chemotherapeutic drug DDP, the dosages were 5 μmol/L in vitro and 5 mg/kg in vivo; the proteasomal inhibitor MG132 (Selleck), the dosage was 20 μmol/L in vitro; the lysosomal inhibitor chloroquine (CQ) (Selleck), the dosage was 20 μmol/L in vitro.

2.3 Lentivirus-mediated knockdown and expression

Specific short hairpin RNAs (shRNAs) for PD-L1 and USP51 underwent annealing and subcloning into the pLV-H1-EF1α-puro vector (Biosettia, San Diego, CA, USA). The full-length PD-L1 and USP51 as well as their deletion constructs (PD-L1-ET, PD-L1-TC, Flag-PD-L1-E, Flag-PD-L1-C, Flag-PD-L1-C1, Flag-PD-L1-C2, Flag-PD-L1-C3, Myc-USP51-N and Myc-USP51-C) (Sangon Biotech, Shanghai, China) underwent subcloning into the pLV-EF1-MCS-IRES-Bsd vector (Biosettia). The individual mutants of PD-L1 (K263R, K270R, K271R, K280R and K281R) (Sangon Biotech) also underwent subcloning into the pLV-EF1-MCS-IRES-Bsd vector. The constructed and packaged plasmids including gag polyprotein (Gag-Pol), regulator of expression of virion protein (Rev) and vesicula stomatitis virus glycoprotein G (VSVG) (Biosettia) were then co-incorporated into 293T cells with Lipofectamine 2000 reagent (Invitrogen, Waltham, MA, USA) to produce lentivirus particles. The details of primers used can be found in Supplementary Table S1.

2.4 RNA isolation and quantitative RT-PCR (qRT-PCR)

Total RNA was isolated with Trizol (Yeasen), followed by the first-strand cDNA generation using TransScript SuperMix (TransGen, Beijing, China). The chemoresistant genes were specifically amplified using quantitative PCR with SYBR Green Master Mix (Yeasen). The following PCR parameters were used: 95°C for 10 min, then 40 cycles of 95°C for 15 s, and 60°C for 1 min. The value of ^△△Ct was employed to determine chemoresistant gene relative expression. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) served as a normalization control. The primer details can be found in Supplementary Table S1.

2.5 Western blotting analysis

Total protein isolation and Western blotting was conducted as reported previously [24]. Protein samples were collected in lysis buffer for radio immunoprecipitation assay (RIPA) (Yeasen) with protease and phosphatase inhibitors (Yeasen), then quantified with a bicinchoninic acid (BCA) kit (ThermoFisher Scientific). Protein separation was performed with sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) prior to transfer to polyvinylidene fluoride (PVDF) membranes (Millipore, Billerica, MA, USA), then blocked for 1 h in 5% skimmed milk (Yeasen) at room temperature, before overnight incubation with primary antibodies at 4°C with subsequent 1 h incubation in secondary antibodies at room temperature. Lastly, protein visualization employed enhanced chemiluminescence (ECL)-based chemiluminescence (Yeasen). A summary of the employed antibodies is available in Supplementary Table S2.

2.6 Deubiquitination assay

Cells were incorporated with haemagglutinin (HA)-ubiquitin and the previously described plasmids, then treated with MG132 (Yeasen) for 6 h to inhibit ubiquitin-proteasome-dependent dysregulation of protein. Cell lysates were made in RIPA buffer (Yeasen), prior to overnight incubation in anti-Flag antibody or IgG at 4°C. The resulting immunoprecipitates were assayed using an anti-HA-Tag antibody-mediated Western blotting. The details of antibodies used can be found in Supplementary Table S2.

2.7 Protein expression and purification for PD-L1 proteins

Glutathione S-transferase (GST)-tagged PD-L1 and its corresponding truncated mutants (GST-PD-L1-E, GST-PD-L1-IgV, GST-PD-L1-IgC2 and GST-PD-L1-C) (Sangon Biotech) underwent cloning into the pGEX-6p-1 vector (Biosettia) prior to expression in E. coli BL21 cells (WeiDi, Shanghai, China). Cell induction employed 0.5 mmol/L isopropyl β-D-thiogalactoside (IPTG) (Yeasen) in Luria-Bertani (LB) medium (Sigma, St. Louis, MO, USA) at 16°C for 24 h to induce PD-L1 protein expression. The cell pellet was dissolved in lysis buffer (50 mmol/L Tris, 300 mmol/L NaCl), and 2 mg/mL lysozyme (Yeasen) was added and incubated at 4°C overnight, followed by ultrasonication using Ultrasonic Crusher (SCIENTZ, Ningbo, Zhejiang, China), and a 20 min centrifugation at 18,000 × g at 4°C. Following removal of supernatant, the inclusions were collected in solubilization buffer [200 mmol/L Tris, 5 mmol/L dithiothreitol (DTT), 8 mol/L urea, pH = 8.0]. Insoluble material was centrifuged at 18,000 × g for 20 min at 4°C and removed, and the supernatant was placed onto Glutathione Sepharose 4B beads (ThermoFisher Scientific) and equilibrated with binding buffer (50 mmol/L Tris, 300 mmol/L NaCl) overnight. The protein was eluted with elution buffer (50 mmol/L Tris, 300 mmol/L NaCl, 10 mmol/L glutathione, pH = 7.5-8.0) and stored at -80°C.

2.8 Coimmunoprecipitation and GST pulldown assays

Cell lysates were made in RIPA buffer, precleared using desired antibody or IgG at 4°C overnight, with subsequent incubation with protein G Resin (GenScript, Nanjing, Jiangsu, China) for 4 h. For GST pulldown assay, we incubated the cell lysates in GST-tagged proteins and Glutathione Sepharose 4B beads overnight at 4°C. Finally, the immunoprecipitates were separated with Western blotting with corresponding antibodies. The details of antibodies used can be found in Supplementary Table S2.

2.9 In vitro binding assays

The purified proteins of histidine (His)-USP51 and GST-PD-L1 (full-length PD-L1 and its corresponding truncated mutants) were incubated at 4°C overnight. Glutathione Sepharose 4B beads were introduced and maintained for 4 h, before Western blotting of the bound proteins with USP51 antibody (H00158880-A01, Abnova, Taiwan, China).

2.10 In vitro USP51 activity assay

C-terminal conjugate of ubiquitin with 7-amino-4-methylcoumarin (Ub-AMC) was used to measure the deubiquitinase activity of USP51. Briefly, the reaction system contained 50 mmol/L 2-hydroxyethyl, 0.5 mmol/L ethylene diamine tetraacetic acid, 100 mmol/L NaCl,1 mmol/L DTT, 0.1 mg/mL bovine serum albumin, 64 nmol/L USP51 (MERRYBIO, Nanjing, Jiangsu, China) and 5 μmol/L DHM (T2998, TOPSCIENCE) and was incubated at 37°C for 1 h. Then, 250 nmol/L Ub-AMC (U-550, BostonBiochem, Emeryville, CA, USA) was added, and the fluorescence intensity was detected using a microplate reader (BMG Labtech, Offenburg, Germany) with a 345 nm excitation/460 nm emission optic module for 3 h. All experiments had three biological replicates, and the mean value was analyzed.

2.11 Surface plasmon resonance (SPR)

The binding kinetic evaluations were conducted on a Biacore T200 (GE Healthcare, Boston, MA, USA). Briefly, His-tagged USP51 were immobilized in Acetate 4.0 (BR-1003-49, GE Healthcare) on a Biacore Series S NTA chip (CM5) using N-terminal His tag capture, then amine-coupling with 1-(3-dimethylaminopropyl)-3-ethylcarbodiimide hydrochloride/N-hydroxy succinimide/ethanolamine (Sigma) as per kit guidelines. A 7500 response unit immobilization was attempted. Compounds were examined in two-fold dilution series in immobilization buffer containing 5% (V/V) dimethylsulfoxide (DMSO), at 30 μL/min, with 60 s contact time and 120 s dissociation time. We also included a solvent correction curve between 4.5%-5.8% (V/V) DMSO. Ten-point dilution curves before and after compounds were used to confirm surface integrity throughout the experiment. The Biacore T200 Evaluation software (version 3.0, GE Healthcare) was employed with a kinetic fit.

2.12 Cellular thermal shift assay

Cells underwent a 5 h DHM treatment at specified concentrations, then, they were accumulated for 3 min in PBS with a complete protease inhibitor cocktail, and respectively incubated at 40°C, 43°C, 46°C or 49°C. Next, cells were freeze-thawed in liquid nitrogen, and the supernatant was collected. Finally, proteins were assessed using Western blotting assay with USP51 antibody (SAB1305451, Sigma).

2.13 Cell viability assessment

In all, 3,000 cells/well were plated in 96-well plates and treated with different drugs for different time points. Cell survivability was assessed via cell counting kit-8 (Yeasen), as per kit guidelines, using absorbance at 450 nm via a microplate reader (BMG Labtech).

2.14 Colony formation experiment

In all, 750 cells/well were plated in 6-well plates before treatment with different drugs for 10 days. Cells underwent a 30 min fixation in 4% paraformaldehyde (Sigma), and 15 min crystal violet staining. The number of colonies was counted (colony formation rate = colony number/plated cell number).

2.15 Flow cytometric analysis of cell apoptosis

Cells were treated with different drugs for 48 h before apoptotic assessment with the fluorescein isothiocyanate (FITC) Annexin-V/prodium iodide (PI) apoptosis detection kit (Yeasen) as per kit directions. Apoptotic cell percentage was evaluated using flow cytometry (LSR, BD Biosciences, San Diego, CA, USA).

2.16 Immunofluorescence assay

Cells underwent 24-hour treatment with varying drugs before fixation in 4% paraformaldehyde, followed by overnight incubation with phosphorylated histone H2AX (γH2AX) antibody (9718S, Cell Signaling Technology, Boston, MA, USA) at 4°C. The next day, cells were treated for 1 h with DyLight 594-conjugated secondary antibody (ab96885, abcam, Cambridge, Cambs, UK) before a 3 min staining with 4’,6-diamidino2-phenylindole (DAPI) (1μg/mL). The labeled nuclei were detected by Confocal FV1000 (Olympus, Tokyo, Japan). The percentage of cells with ≥10 foci was quantified. At least 100 cells were counted per well.

2.17 Tumor xenograft experiment

All animal protocols received ethical approval from the Medical College of Nankai University (2021-SYDWLL-000040, Tianjin, China). We obtained 6-week-old male NOD-SCID mice from Beijing HFK Bioscience LTD (Beijing, China), and injected cells into their bilateral groins. Once tumors achieved a size of 50 mm³, mice were treated with different drugs for three weeks. When the tumor size reached to approximately 800 mm³, the mice were euthanized following intraperitoneal anesthesia with ketamine and promethazine. The tumor weight was weighed, and volume was calculated as 1/2 (length × width²).

2.18 Tissue microarray and immunohistochemistry (IHC)

Microarrays of human lung adenocarcinoma tissues (HLugA150CS03 and HlugA180Su04) were obtained from Shanghai Outdo Biotech Co., Ltd. (Shanghai, China). Samples underwent USP51, PD-L1 (17952-1-AP, Proteintech, Wuhan, Hubei, China) and ITGB1 (34971T, Cell Signaling Technology) antibody staining for IHC assays with the Envision Kit (Dako) in accordance as per kit protocols. Several pathologists then blindly scored the immunostaining, as described previously [24]. This protocol received ethical approval from the Nankai University (NKUIRB2021010).

2.19 Bioinformatic analysis

The gene expression profiling interactive analysis (GEPIA) database (http://gepia.cancer-pku.cn/) was used to analyze the effect of ITGB1 on clinical survival. JASPAR (http://jaspar.genereg.net/) analysis was employed to identify a series of canonical NF-κB response elements in the promoter region of chemoresistance-related genes.

2.20 Statistical analysis

The experimental data were analyzed by GraphPad Prism 8.0 (GraphPad Software, San Diego, CA, USA) and SPSS 17.0 (IBM, Chicago, IL, USA) software (Chicago, IL, USA). Data from all experiments are expressed as mean ± standard deviation (SD) of three distinct replicates. Gene profile correlation in various samples were assessed via Spearman's rank correlation test. Inter-group comparison was performed using one-way analysis of variance (ANOVA). Unpaired data assessments were conducted using Student's t-test, as appropriate. P value < 0.05 was set as the significance threshold.

3.1 The extracellular domain of PD-L1 was required for the regulation of chemosensitivity in a cell-intrinsic manner

Consistent with previous report showing that abnormal PD-L1 content is involved in chemoresistance development in human cancers [10], our Western blotting and flow cytometry assay results revealed a significant increase in PD-L1 levels in DDP-resistant A549^R cells compared to the wild-type A549 cells (Figure 1A). Thus, to further investigate the significance of cancer cell-intrinsic PD-L1 in chemoresistance, specific shRNAs targeting PD-L1 were introduced in A549^R followed by treatment with DDP. Based on our cell viability (Figure 1B), colony formation (Figure 1C) and cell apoptosis analyses (Figure 1D), PD-L1 depletion strongly promoted DDP-induced cell proliferation retardation and cell apoptosis. Similarly, the anti-apoptosis protein Bcl-2 was strongly diminished, whereas the pro-apoptosis proteins Bax and cleaved Caspase-3 were enhanced in shPD-L1/A549^R cells in response to DDP treatment (Supplementary Figure S1A). We also explored the effect of PD-L1 on DDP-induced γH2AX expression, which is a marker for DNA damage repair [25]. The results confirmed that PD-L1 knockdown significantly enhanced DDP-induced formation of γH2AX nuclear foci (Figure 1E) and the level of γH2AX expression (Supplementary Figure S1B). Consistently, we obtained the similar results in shPD-L1/PC9^R cells (Supplementary Figure S2), indicating that cell-intrinsic PD-L1 potentially modulates chemosensitivity in NSCLC cells.

Further, we determined whether aberrant PD-L1 content influenced tumor responsiveness to chemotherapeutic treatment in vivo. To do so, shPD-L1/A549^R or shCtrl/A549^R cells were subcutaneously injected into BALB/c nude mice to develop a tumor xenograft model, which was then treated with DDP. Based on our observation, DDP strongly suppressed tumor development in BALB/c nude mice with shPD-L1/A549^R tumors as compared to shCtrl/A549^R tumors (Figure 1F-G). Immunohistochemical staining further demonstrated that DDP treatment resulted in decreased Ki67 expression but increased cleaved Caspase-3 and γH2AX expression in shPD-L1/A549^R tumors as compared to shCtrl/A549^R tumors (Figure 1H). Thus, the above observations confirmed the involvement of aberrant PD-L1 in NSCLC chemoresistance, independent of the immune system in vitro and in vivo.

Next, to elucidate the functional domain of PD-L1 that contributes to its cell-intrinsic role in chemoresistance, the full-length PD-L1 (PD-L1-FL) and two deletion constructs PD-L1-ET (containing the extracellular and transmembrane domains of PD-L1) and PD-L1-TC (containing the transmembrane and cytoplasmic domains of PD-L1) were generated and overexpressed in A549 cells (Figure 1I). The cell viability (Figure 1J), colony formation (Figure 1K), cell apoptosis (Figure 1L and Supplementary Figure S3A), and DNA damage detection assays (Figure 1M and Supplementary Figure S3B) confirmed that overexpression of PD-L1-FL strongly reduced cell proliferation retardation and cell apoptosis and decreased the formation of γH2AX nuclear foci upon DDP treatment. Notably, we found that overexpression of PD-L1-ET showed a similar effect to attenuate cancer cell responsiveness to DDP. We observed the same alterations in PC9 cells (Supplementary Figure S4), indicating that the PD-L1 extracellular domain conferred the regulation of chemosensitivity through a cell-intrinsic mechanism in NSCLC cells.

3.2 ITGB1 mediated PD-L1-induced chemoresistance

In addition, we moved to identify the downstream signaling pathway that mediated PD-L1-induced chemoresistance in NSCLC cells. Surprisingly, treatment with the PD-L1 neutralizing antibody avelumab did not show a chemosensitizing effect by the analysis of cell viability (Supplementary Figure S5A), colony formation (Supplementary Figure S5B), cell apoptosis (Supplementary Figure S5C-D), and DNA damage detection (Supplementary Figure S5E-F) in A549^R cells. Considering that a complex structure analysis revealed an IgV-dominated binding pattern between PD-L1 and its neutralizing antibody [26], these results highlight that the cancer cell-intrinsic role of PD-L1 in chemoresistance might be independent of its interaction with the PD-1 receptor through the IgV domain. Consequently, we performed an endogenous Co-IP combining with mass spectrometry and identified several cell membrane-bound receptors that could potentially interact with PD-L1 in A549^R cells (Supplementary Figure S6A). Quantitative RT-PCR further demonstrated a significant upregulation of ITGB1, CD98 and CD44 expression in A549^R cells (Supplementary Figure S6B). Of note, the GEPIA database analysis revealed that elevated ITGB1 levels were closely related to shorter overall survival (OS) (Supplementary Figure S6C) and disease-free survival (Supplementary Figure S6D) in NSCLC patients.

Importantly, the Co-IP assay demonstrated an increased physical interaction between PD-L1 and ITGB1 in A549^R cells relative to A549 cells (Figure 2A). Moreover, cell viability (Figure 2B), colony formation (Figure 2C), cell apoptosis (Figure 2D and Supplementary Figure S7A), and DNA damage detection assays (Figure 2E and Supplementary Figure S7B) demonstrated that either PD-L1 depletion or ITGB1 neutralization significantly promoted cancer cell chemosensitivity to DDP in A549^R cells; however, the combinatorial treatment did not elicit a synergistic anticancer activity. Similarly, we conducted the same investigations in PC9^R cells and achieved comparable results (Supplementary Figure S8).

Next, we identified the PD-L1 functional domain that mediates its association with ITGB1. We constructed three GST fusion proteins for the extracellular domain of PD-L1 (PD-L1-E, PD-L1-IgV and PD-L1-IgC2). The results of GST-pulldown assay confirmed that PD-L1 physically interacted with ITGB1 through its extracellular domain (Figure 2F). Of note, the deletion variant PD-L1-IgC2, but not PD-L1-IgV, was able to interact with ITGB1. Taken together, our data proposed that ITGB1 specifically functioned as a membrane-binding receptor for PD-L1 and thus mediates its cell-intrinsic role in chemoresistant NSCLC cells.

3.3 PD-L1 conferred chemoresistance by triggering the ITGB1/NF-κB signaling

Next, we moved to identify the molecular mechanism of PD-L1/ITGB1-regulated cancer cell responsiveness to chemotherapy in NSCLC. Of note, TCGA database analysis revealed that the transcript levels of multiple chemoresistance-related genes, such as baculoviral IAP repeat containing 5 (BIRC5) [27], breast cancer gene 1 (BRCA1) [28, 29], BRCA2 [30, 31], glutathione synthetase (GSS) [32], hypoxia inducible factor 1 alpha (HIF1α) [33], and solute carrier family 31 member 2 (SLC31A2) [34], were positively correlated with those of PD-L1 (Supplementary Figure S9A) and ITGB1 (Supplementary Figure S9B) in NSCLC patients. Quantitative RT-PCR further verified that expression of these chemoresistance-related genes was significantly downregulated by either PD-L1 depletion or ITGB-1 neutralization in A549^R cells (Figure 3A). However, the combinatorial treatment did not exhibit a synergistic inhibition, which is consistent with our notion that PD-L1-promoted chemoresistance is mainly mediated through its binding to the ITGB1 receptor on NSCLC cells. Interestingly, analysis using the JASPAR database identified a series of canonical NF-κB response elements in the promoter region of all above chemoresistance-related genes (Supplementary Figure S9C), indicating a potential role of the NF-κB signaling in PD-L1/ITGB1-regulated chemoresistance [35]. Thus, expression of the NF-κB components was assessed via Western blotting in shPD-L1/A549^R cells with or without ITGB1 neutralizing antibody. We confirmed that either PD-L1 knockdown or ITGB1 neutralization could reduce the expression of phosphorylated IκBα (Figure 3B), with a simultaneous decrease in the nuclear transfer of the P65 subunit (Figure 3C). However, a synergistic inhibition in the NF-κB signaling was not achieved by the combinational treatment.

In line, the quantitative RT-PCR (Figure 3D), cell viability (Figure 3E), colony formation (Figure 3F), cell apoptosis (Figure 3G and Supplementary Figure S10A), and DNA damage detection (Figure 3H and Supplementary Figure S10B) assays further revealed that either depletion of PD-L1 or treatment with an NF-κB inhibitor Bay 11-7085 increased A549^R cell chemosensitivity to DDP; however, we did not observe a synergistic suppression with the combinational treatment. Similar results were also obtained in PC9^R cells (Supplementary Figure S11), confirming that the PD-L1/ITGB1 axis promoted NSCLC cell sensitivity to chemotherapy by triggering the NF-κB signaling.

3.4 USP51 promoted chemoresistance through deubiquitination of PD-L1

To further explore if the abnormal PD-L1 content is related to posttranslational modulation in chemoresistant NSCLC cells, we next treated A549^R and A549 cells with the protein synthesis inhibitor CHX. Western blotting analysis demonstrated that the stability of PD-L1 protein was strongly enhanced in A549^R cells relative to A549 cells (Supplementary Figure S12A). Moreover, exposure to the proteasomal inhibitor MG132 (Supplementary Figure S12B) and not the lysosomal inhibitor CQ (Supplementary Figure S12C) led to remarked upregulation of PD-L1 protein. The ubiquitination assay data revealed that the polyubiquitination level of PD-L1 protein was strongly decreased upon MG132 treatment in A549^R cells compared to that in A549 cells (Supplementary Figure S12D). Also, we confirmed these results in PC9^R and PC9 cells (Supplementary Figure S13). These data indicated a ubiquitin proteasome-dependent dysregulation of PD-L1 content in chemoresistant NSCLC cells.

Notably, the analysis of endogenous Co-IP combining with mass spectrometry identified two deubiquitinases [USP51 and ubiquitin C-terminal hydrolase L1 (UCHL1)] and two E3 ubiquitin ligases [ring finger protein 31 (RNF31) and tripartite motif containing 21 (TRIM21)] that could potentially interact with PD-L1 in A549^R cells (Supplementary Figure S14A). We further validated the physical interactions of PD-L1 with USP51 (Figure 4A), UCHL1 (Supplementary Figure S14B), RNF31 (Supplementary Figure S14C), and TRIM21 (Supplementary Figure S14D) in PD-L1-overexpressing A549 cells. However, Western blotting and flow cytometry assays demonstrated that the specific interference of USP51 (Figure 4B), but not UCHL1 (Supplementary Figure S14E), RNF31 (Supplementary Figure S14F) and TRIM21 (Supplementary Figure S14G), altered PD-L1 levels in A549^R cells. Western blotting data revealed that the expression of USP51 was significantly elevated in A549^R cells (Figure 4C). These observations demonstrated a predominant function of USP51 in the posttranslational modulation of PD-L1 in chemoresistant NSCLC cells. In line, the PD-L1 protein stability was strongly decreased upon CHX treatment in shUSP51/A549^R cells (Figure 4D). However, USP51-interfered PD-L1 content decrease was strongly weakened with MG132 addition (Figure 4E). Based on our ubiquitination assay, USP51 depletion strongly enhanced PD-L1 protein polyubiquitination in A549^R cells (Figure 4F), collectively suggesting that USP51, a deubiquitinase, targeted PD-L1 protein for deubiquitination and stabilization. Similar results were further obtained in PC9^R cells (Supplementary Figure S15).

Next, we examined whether USP51 depletion could chemosensitize NSCLC cells through the regulation of PD-L1. To do so, PD-L1 was overexpressed in shUSP51/A549^R cells (Figure 4G). The cell viability (Figure 4H), colony formation (Figure 4I), cell apoptosis (Figure 4J and Supplementary Figure S16A), and DNA damage detection assays (Figure 4K and Supplementary Figure S16B) demonstrated that cell responsiveness to chemotherapy was increased in shUSP51/A549^R cells; nevertheless, this outcome was strongly diminished in PD-L1-overexpressed cells. We observed comparable results in PC9^R cells (Supplementary Figure S17). In line, using BALB/c nude mice xenograft model, we confirmed that USP51 knockdown severely inhibited tumor growth upon DDP treatment, which was suppressed in mice with PD-L1-harboring tumors (Figure 4L-M). Immunohistochemical staining further indicated that the Ki67 content was diminished but the cleaved Caspase-3 and γH2AX contents were increased in tumors from shUSP51/A549R mice when treated with DDP, and this outcome was suppressed in tumors with PD-L1 rescue (Figure 4N). These observations together revealed that USP51 deficiency chemosensitized NSCLC cells via PD-L1 regulation both in vitro and in vivo.

3.5 USP51 deubiquitinated PD-L1 at lysine residues K280/K281

Consequently, we conducted an endogenous Co-IP to validate the USP51 and PD-L1 relationship. The results revealed that USP51 directly associated with PD-L1 in A549 cells, which was further increased in A549^R cells (Figure 5A). Importantly, an in vitro binding assay proved that the purified proteins of His-USP51 and GST-PD-L1 could directly associate with one another in cell-free environment (Figure 5B). Moreover, a series of deletion mutants of USP51 (Myc-USP51-N and Myc-USP51-C) and PD-L1 (Flag-PD-L1-E and Flag-PD-L1-C) were generated and overexpressed in A549 cells. Our Co-IP data verified that the deletion variant Myc-USP51-N interacted with the PD-L1 cytoplasmic domain (Figure 5C-D). In addition, three deletion constructs of PD-L1-C were generated to demonstrate that the deletion variant PD-L1-C3, but not PD-L1-C1 and PD-L1-C2, was able to interact with USP51 (Figure 5E), highlighting that the PD-L1C-terminal region (280-290aa) is critical for the USP51-PD-L1 association.

Next, we examined whether USP51 could deubiquitinate the lysine residues of PD-L1-C3. As shown in Figure 5F, the wild-type and individual mutants of PD-L1 (K263R, K270R, K271R, K280R and K281R) were respectively transfected into shUSP51/A549^R cells. The ubiquitination assay indicated that the polyubiquitination levels of either K280R or K281R was significantly reduced compared to wild-type PD-L1, however this effect was not shown for the K263R, K270R and K271R mutants. In addition, the simultaneous mutant K280R+K281R totally abolished the polyubiquitination levels of PD-L1 (Figure 5G), identifying lysine residues K280 and K281 as major ubiquitination sites on PD-L1. Indeed, USP51 was able to deubiquitinate PD-L1 at either K280 or K281 lysine residue (Figure 5H). The interference of USP51 led to a strongly reduced stability of wild-type PD-L1 protein; however, this effect was deprived in cells carrying the simultaneous mutant K280R+K281R (Figure 5I). The same experiments were also conducted in PC9^R cells and produced comparable results (Supplementary Figure S18). Taken together, our data demonstrated that lysine residues K280 and K281 on the PD-L1 C-terminal region were essential for its deubiquitination and stabilization by USP51 in chemoresistant NSCLC cells.

3.6 The USP51, PD-L1 and ITGB1 contents were positively associated with worse prognosis among NSCLC patients

To further validate the pathological relationship between the USP51/PD-L1/ITGB1 axis and chemoresistant features in human NSCLC, we conducted immunohistochemical staining for USP51, PD-L1 and ITGB1 in 167 samples of primary NSCLC (Figure 6A). We observed direct relationships between USP51, PD-L1 and ITGB1 (Figure 6B). It has been reported that Cyclin D1 [36, 37] and breast cancer gene 1 (BRCA1) [38, 39], which mediate cell proliferation and DNA damage response, respectively, were strongly elevated in NSCLC tissues resistant to genotoxic drugs. Indeed, we revealed that the USP51, PD-L1 and ITGB1 contents were intricately linked to Cyclin D1 (Figure 6C) and BRCA1 (Figure 6D), highlighting the involvement of aberrant USP51/PD-L1/ITGB1 expression in the cellular mechanisms that mediated NSCLC chemoresistance.

In addition, we observed increased USP51, PD-L1 and ITGB1 levels in high-grade tumors (Figure 6E) and advanced tumor-node-metastasis (TNM) stage (Figure 6F). Of note, the cancer patients with substantially elevated USP51, PD-L1 and ITGB1 levels in tumors exhibited shorter OS relative to patients with reduced USP51, PD-L1 and ITGB1 contents (Figure 6G). We also examined OS of the triple-high group (USP51^highPD-L1^highITGB1^high) compared to the rest (Figure 6H). The results confirmed that patients with concomitantly high expression of USP51, PD-L1 and ITGB1 in tumors had significantly shorter OS (Figure 6I). Taken together, these data revealed that dysregulation of the USP51/PD-L1/ITGB1 axis potentially contributes to the malignant progression and may be employed in the prediction of worse NSCLC patient prognosis.

3.7 Dihydromyricetin acted as a USP51 inhibitor and conferred chemosensitization in NSCLC

Next, we moved to identify the small molecules that could potentially alter USP51 activity. To do so, we used the fluorescent Ub-AMC assay to examine an in-house library containing 188 natural compounds (Figure 7A and Supplementary Table S3). Five compounds, dihydromyricetin (DHM), (-)-epigallocatechin (EGC), gallic acid, linoleic acid, and tannic acid, were shown to significantly inhibit the deubiquitinase activity of USP51 at a concentration of 1 μmol/L (Supplementary Table S4). Further, we conducted a computational molecular dynamics simulation and revealed that three flavonoid compounds DHM, EGC and gallic acid could potentially bind to the catalytic domain of USP51 (Figure 7B). Of note, the analysis of half maximal inhibitory concentration (IC₅₀) indicated that DHM performed a stronger inhibition in USP51 activity than EGC and gallic acid (Figure 7C). In line with this, the SPR binding assay validated a significantly increased binding affinity between USP51 and DHM (Figure 7D). Taken together with the cellular thermal shift assay showing that DHM treatment effectively stabilized USP51 protein in A549^R (Figure 7E) and PC9^R cells (Supplementary Figure S19), our results indicated that DHM might target USP51 directly.

In addition, to examine whether DHM exerted its biological function through a USP51-dependent mechanism, we introduced the full-length USP51 into A549 cells followed by treatment with DHM (Supplementary Figure S20A). Based on our Western blotting results, the PD-L1 content was upregulated in USP51-OE/A549 cells, whereas this effect was significantly abolished by DHM treatment. Consistently, the ubiquitination assay revealed that USP51 deubiquitinase inhibition by treatment with DHM led to increased polyubiquitination of PD-L1 proteins in USP51/A549 cells (Supplementary Figure S20B). The cell viability (Supplementary Figure S20C), colony formation (Supplementary Figure S20D), cell apoptosis (Supplementary Figure S20E-F), and DNA damage detection assays (Supplementary Figure S20G-H) further confirmed that overexpression of USP51 strongly reduced cell proliferation retardation and cell apoptosis but promoted DNA damage repair in response to DDP; however, USP51-induced chemoresistant phenotypes were weakened by the addition of DHM. Similar results were also obtained in PC9 cells (Supplementary Figure S21), highlighting a role of DHM in the regulation of chemosensitization via targeting USP51.

Indeed, treatment with DHM strongly promoted DDP-induced cell proliferation retardation (Figure 7F-G) and cell apoptosis (Figure 7H and Supplementary Figure S22A) in A549^R cells, while this effect was impaired by USP51 interference. We also examined the effect of DHM on DDP-induced DNA damage response with or without USP51 depletion (Figure 7I and Supplementary Figure S22B), revealing that DHM conferred responsiveness to chemotherapeutic treatment in a USP51-dependent manner in NSCLC cells. These experiments were further performed in PC9^R cells and obtained the similar results (Supplementary Figure S23).

3.8 DHM treatment or ITGB1 neutralization chemosensitized NSCLC

Our study revealed a cell-intrinsic role for USP51/PD-L1/ITGB1 signaling in mediating chemotherapeutic resistance in NSCLC. Therefore, treatment with DHM or anti-ITGB1 might suppress the chemoresistant phenotypes of NSCLC cells and further enhance the efficacy of chemotherapy. To achieve this, A549^R cells were subcutaneously injected into BALB/c nude mice to allow the establishment of primary tumors, then they were treated with DHM, anti-ITGB1 or anti-PD-L1 (avelumab) in the presence of DDP (Figure 8A). We noticed that treatment with DHM or anti-ITGB1 effectively promoted the cell responsiveness to DDP, as evidenced by a reduction in tumor volume (Figure 8B) and weight (Figure 8C) by approximately 75%-80%; however, the combinatorial treatment did not exhibit a synergistic inhibition. Of note, the chemosensitizing efficacy was not shown for anti-PD-L1 treatment, which is consistent with our notion that the cell-intrinsic role of PD-L1 in chemoresistance might not achieve by its interaction with PD-1. The immunohistochemical staining also revealed significant reduction in Ki67 content but increase in cleaved Caspase-3 and γH2AX expression in tumors with DHM and/or anti-ITGB1 treatment (Figure 8D). No obvious undesirable side effects, namely reduced body weight (Figure 8E) and toxic pathological alterations occurred in the heart, liver, spleen, lung, and kidney (Supplementary Figure S24). Overall, these results implied that disruption of USP51/PD-L1/ITGB1-deployed juxtacrine interaction by treating cancer patients with the USP51 inhibitor (e.g. DHM) or ITGB1 neutralization might inhibit their chemoresistant properties and thus reduce cancer risk.

AUTHOR CONTRIBUTIONS

Conception and design: JL, XX, GY and SY; Development of methodology: JL, XX, YO, LC, MG, HW, GY and SY; Analysis and interpretation of data (for example, statistical analysis, biostatistics, computational analysis): JL, XX, CQ, ZW, YL, QS, PS, YS, GY and SY; Writing, review of the manuscript: JL, XX, PS, YS, GY and SY.