High Mobility Group Protein B1 Decreases Surface Localization of PD-1 to Augment T-cell Activation

Damage-associated molecular patterns (DAMP) are “danger signals” released by necrotic cells or tumor tissues under stress (1). A lack of appropriate “danger signaling molecules” in the tumor microenvironment (TME) to stimulate an effective antitumor immune response and maintain tumor immune surveillance is a key factor for tumor immune evasion and the failure of tumor immunotherapy (2). Chemotherapy or radiotherapy can cause necrosis and apoptosis of tumor cells, which promote the transformation of tumor cells from nonimmunogenic to immunogenic states or increase the release of tumor antigens by tumor cells, thereby stimulating an antitumor immune response in vivo (3). Therefore, there is great interest in identifying appropriate “danger signaling molecules” as a means to activate effective antitumor immune responses and to enhance tumor surveillance.

HMGB1, the best characterized DAMP, is a regulator of tumor development (4, 5). During chemotherapy or radiotherapy, HMGB1 released by cells under stress conditions induces the release of cytokines and maturation of antigen-presenting cells (6). The combination of HMGB1 and nucleic acids in the TME activates the initial immune response to inhibit tumor progression. The binding of HMGB1 to the surface of dendritic cells (DC) and receptor for advanced glycation end products (RAGE) on T cells surface promotes the maturation of DCs, T-cell proliferation and the presentation of tumor antigens. In addition, HMGB1 can recruit DCs and cytotoxic T cells directly during radiotherapy or chemotherapy (7). In the late stage of chemotherapy, HMGB1 released into the TME has been shown to bind with TLR4 on the surface of DCs to activate the TLR4–MyD88–NFκB signaling pathway. This promotes cross-presentation of tumor antigens to CD8⁺ T cells, to induce the killing effect of CTLs on tumors, mediate immunogenic cell death (ICD), and increase ICD-related antitumor immune response (8). Thus, HMGB1 plays an important role in the TME; but how HMGB1 directly regulates the function of CD8⁺ T cells remains unclear, and the mechanism of regulating effective T-cell function still needs further study.

T cells in the TME often exhibit immune exhaustion via the expression of PD-1, leading to a decreased ability to kill tumor cells (9). The binding of PD-1 to PD-L1 inhibits T-cell proliferation and activity (10). In addition, downregulation of PD-1 enhances CTL-mediated antitumor immunity (11, 12). It has been reported that suppressing the PD-1 pathway through rapid internalization and degradation of PD-1 can lead to enhanced CTL function (13). Increasing evidence has shown that internalization and degradation of PD-1 affects the function of T cells. F-Box Protein 38 (FBXO38) is an E3 ligase that mediates ubiquitination and proteasome degradation of PD-1, thereby regulating PD-1 expression and antitumor immunity (14). Expression of other proteins on the surface of T cells is regulated by internalization and degradation, and this can lead to the cells having altered function. For example, upon stimulation with IL7, CD127 (the IL7 receptor α chain) undergoes clathrin-mediated endocytosis and subsequent degradation by the proteasome in primary human CD8⁺ T cells. CD127 shows increased colocalization with the endosomal marker early endosome antigen 1 (EEA1), and CD127 staining is associated with the late endosomal marker RAB7 and proteasomal 20S subunit (15). In addition, chloroquine (CQ), which inhibits lysosome activity, has been identified as an inhibitor of CTLA-4 degradation in human CD8⁺ T cells (16).

In the current study, we aimed to understand whether HMGB1 affects PD-1 expression on CD8⁺ T cells. We found that PD-1 expression on CD8⁺ T cells was negatively correlated with HMGB1 release induced by chemotherapy in patients with non–small cell lung cancer (NSCLC). Further analysis revealed that HMGB1 bound to PD-1 to induce dynamin-mediated internalization, leading to early endocytosis in the cytoplasm and subsequent degradation in the lysosomes. This cascade of events led to downregulation of PD-1 expression. In addition, we explored the antitumor effect of HMGB1 on chimeric antigen receptor (CAR) T cells and laid a theoretical foundation for a new immunotherapy combination.

Peripheral blood and tumor tissue specimens were obtained from 56 patients with NSCLC at the First Affiliated Hospital of Zhengzhou University (Zhengzhou, P.R. China) from 2016 to 2019. Detailed characteristics of the patients are summarized in Supplementary Tables S1–S6. None of the patients had received radiotherapy before sampling, and patients with comorbidities such as infectious diseases, autoimmune diseases, or multiprimary cancer were excluded. Three healthy donors were also recruited for the study. Fresh lung tissue specimens from surgery were placed in sterile medium, and tumor-infiltrating lymphocytes (TIL) were isolated within 2 hours. This study was conducted in accordance with International Ethical Guidelines for Biomedical Research Involving Human Subjects. The procedures were performed according to the protocols (2016-LW-208) that were approved by the Ethics Committee of the First Affiliated Hospital of Zhengzhou University (Zhengzhou, P.R. China). Written informed consent was obtained from all participants.

A549, H322, Jurkat T cells, and 293T cells were authenticated by and obtained from the National Collection of Authenticated Cell Cultures (Shanghai, P.R. China) in 2015. A549 and H322 cells were cultured in DMEM (SH30023.FS, Hyclone) supplemented with 10% FBS (10099-141, GIBCO) and 100 U/mL penicillin/streptomycin (SV30010, Hyclone) in an incubator at 37°C with 5% CO₂. Jurkat T cells were cultured in RPMI1640 (SH30809.01, Hyclone) supplemented with 10% FBS and 100 U/mL penicillin/streptomycin. 293T cells were cultured in DMEM (D6429, Sigma) containing 10% FBS and 100 U/mL penicillin/streptomycin. All cell lines were used within 16 passages of thawing. Mycoplasma contamination was regularly tested by PCR method.

Protein quantification for HMGB1 by ELISA was performed using the HMGB1 ELISA Kit (CSB-E08223h, CUSABIO BIOTECH) according to the manufacturer's instructions.

Peripheral blood lymphocytes (PBL) and TILs were obtained as described previously (17). CD8⁺ T cells were purified from PBLs or tumor tissue using magnetic beads (480108, BioLegend) following the manufacturer's instructions. To increase PD-1 expression, CD8⁺ T cells were stimulated with CD3/CD28 dyna-beads (11132D, Thermo Fisher Scientific) for 24 hours in RPMI1640 containing 10% heat-inactivated FBS. In functional analyses, PD-L1 Fc protein (156-B7-100, R&D Systems) was added to active PD-1 signaling in T cells, unless otherwise described.

For surface staining, cells were suspended in PBS (14200166, Gibco) containing 2% FBS and incubated with fluorescence-conjugated antibodies for 30 minutes at 4°C. For intracellular staining, T cells were treated with Brefeldin A (BFA) for 4 hours, then fixed (Fixation Buffer, 420801, BioLegend), permeabilized (Intracellular Staining Perm Wash Buffer, 421002, BioLegend), and stained with fluorescence-conjugated antibodies for 30 minutes at 4°C. Following that, samples were analyzed using a BD Canto II flow cytometer (Becton Dickinson). Results were analyzed with FlowJo_V10 software. To detect nonspecific background signals, isotype-matched nonspecific antibodies at similar concentration to primary antibodies were used. Details of the antibodies used for flow cytometry in this study can be found in Supplementary Table S7.

For image cytometry analysis, Jurkat T cells were transduced with PD-1–mCherry fusion protein-coding sequence and H322 cells were transduced to express a fusion protein of HMGB1 and GFP via lentivirus. For immunofluorescence (IF) assay, Jurkat T cells were transduced with lentivirus coding PD-1–GFP fusion protein. Jurkat T cells expressing PD-1–GFP fusion protein (PD-1–GFP Jurkat T) and H322 cells expressing HMGB1–RFP fusion (HMGB1–RFP H322) were independently isolated using a MoFlo XDP cytometer (Beckman Coulter Inc.) after lentivirus infection. PD-1–Flag 293T cells were also transduced to express a fusion protein of PD-1 and Flag via lentivirus. All the constructs used here were synthesized by Sangon Biotech and inserted into lentiviral pCDH vector (18). Lentivirus was produced with 293T cells as described previously (19). HMGB1-specific short hairpin (shRNA) was inserted into hU6-MCS-Ubiquitin-eGFP-IRES-puromycin lentivirus vector (Genechem), which contained two promoters separately driving shRNA and eGFP expression. Lentivirus was produced using 293T cells and used to infect A549 cells. Then, shRNA-expressing A549 cells were purified using a MoFlo XDP cytometer (Beckman Coulter) according to GFP expression. All the sequences used for these studies are shown in Supplementary Table S8.

PD-1 knockout CD8⁺ T cells were generated by using electroporation-based recombinant clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated protein 9 (CRISPR–Cas9) gene editing too. In brief, oligonucleotides containing T7 promoter and 20 bp targeting sequences were used as forward primer (5′-GAAATTAATACGACTCACTATAGGGTCTGGGCGGTGCTACAACTGTTTT AGAGCTAGAAATAGC-3′) and an single-guide RNA (sgRNA) backbone reverse primer (5′-AAAAGCACCGACTCGGTGCC-3′) was used to amplify the sgRNA-coding fragment using pX458 plasmid (48138, Addgene plasmid) as template. The T7-sgRNA PCR products were gel-purified and then used as the template for generating PD-1 sgRNAs through in vitro transcription with the MEGAshortscript T7 kit (AM1354, Thermo Fisher Scientific). RNAs were purified using RNA Clean & Concentrator-25 kit (R1017, Zymo Research) and eluted in RNase-free water. A total of 1 × 10⁶ CD8⁺ T cells were stimulated for 24 to 48 hours and then electroporated with 5 μg TrueCut Cas9 Protein v2 (A36498, Thermo Fisher Scientific) and 5 μg PD-1 sgRNA using the Lonza 4D-Nucleofector System with the P3 Primary Cell 4D-Nucleofector X Kit (V4XP-3024, Lonza) according to the manufacturer's instructions. After electroporation, CD8⁺ T cells were resuspended in prewarmed T-cell medium (RPMI1640 (SH30809.01, Hyclone) containing 10% FBS, 100 U/mL penicillin/streptomycin, and 100 IU/mL IL2) and transferred into a 24-well cell plate for incubation. Cell culture medium was half replaced using fresh complete medium every 48 to 72 hours.

RNA was extracted from whole cells using RNAiso Plus (9108, Takara). The RNA was reverse-transcribed to obtain cDNA using the Prime Script RT reagent Kit (RR047A, Takara). For quantitative PCR, we used a Light-Cycler System (Roche) and SYBR Green (6924204001880, Roche), according to the manufacturer's instructions. The primers for the gene encoding PD-1 were 5′-CAGTTCCAAACCCTGGTGGT-3′ and 5′-GGCTCCTATTGTCCCTCGTG-3′, and the primers for the housekeeping gene GAPDH were 5′-GCACCGTCAAGGCTGAGAAC-3′ and 5′TGGTGAAGACGCCAGTGGA-3′. The 2^−ΔΔC_t method was used to calculate relative gene expression.

Cells were lysed in ice-cold RIPA lysis buffer (P0013B, Beyotime) plus protease inhibitor cocktail (P8340-1ML, Sigma) and protein concentration was determined using Pierce BCA Protein Assay Kit (23225, Thermo Fisher Scientific). Western blotting was performed using SDS-PAGE gel and then transferred onto nitrocellulose transfer membrane using Bio-Rad Blotting System (Bio-Rad). Details of the antibodies used for Western blotting in this study can be found in Supplementary Table S7.

X-ray irradiation (10 Gy) or chemical treatment was used to stimulate HMGB1-GFP H322 cells to release the HMGB1-GFP fusion protein. Supernatant containing HMGB1-GFP fusion protein was collected and added to 1 × 10⁵ PD-1–mCherry Jurkat T cells. After 6 hours, the interaction between HMGB1-GFP and PD-1–mCherry was detected using ImageStreamX Mark II image cytometer (Luminex). The yielded histogram reflected the extents of colocation of two stained proteins (20). The measure of colocalization is achieved by the Bright Detail Similarity parameter.

Tissue sections were incubated with primary antibodies specific for HMGB1 (dilution 1:200, ab77302, Abcam) at 4°C overnight. All steps of the IHC process were performed as published previously (17). Images were recorded using a microscope (Leica).

CD8⁺ T cells purified from PBLs or tumor tissue were stimulated with CD3/CD28 dyna-beads for 24 hours. In the CD8⁺ T-cell culture system, inhibitors such as dynasore (HY-15304, MedChem Express), anti-RAGE (ab89911, Abcam), MG132 (S2619, Selleck), and CQ (HY-17589A, MedChem Express) were preadded for 30 minutes, and then recombinant HMGB1 (rHMGB1; 200 ng/mL, 557804, BioLegend) was added. Suspended CD8⁺ T cells were collected according to the corresponding time.

A HER2-specific CAR comprising an extracellular single-chain variable fragment (scFv) specific for human HER2, a CD8 hinge region and transmembrane segment, a CD28 intracellular domain, and a CD3ζ activation domain was synthesized by Sangon Biotech are reported previously (18). Then the CAR coding sequence was inserted into the pCDH lentivirus vector (Systembio). Lentivirus production and CD8⁺ T-cell transduction were carried out as reported previously (19). After transduction, T cells were cultured in RPMI medium supplemented with IL2 (200 U/mL; 200-02, PEPROTECH).

Immunostaining for PD-1 (dilution 1:500, ab52587, Abcam) and EEA1 (dilution 1:200, ab2900, Abcam) was performed using tissue-fixed slides incubated with nonlabeled primary antibodies for 1 hour and subsequently, an appropriate Alexa Fluor-conjugated secondary antibody for 1 hour. The slides were then counterstained with 4, 6-diamidino-2-phenylindole dihydrochloride (D1306, Thermo Fisher Scientific) in mounting medium for 10 minutes. Images for the expression of PD-1 and EEA1 were collected using a Zeiss LSM800 confocal microscope.

For immunofluorescence analyses, 2 × 10⁶ HMGB1–RFP H322 cells were irradiated (10 Gy) for 24 hours and then added to 2 × 10⁵ PD-1–GFP Jurkat T cells for coincubation for 6 hours. After coincubation, the PD-1–GFP Jurkat T cells were examined using a Leica DMI 6000 B fluorescence microscope.

A total of 2 × 10⁶ HMGB1–Myc H322 cells were irradiated (10 Gy) for 24 hours, and then 2 × 10⁵ PD-1–Flag Jurkat T cells were coincubated with the stimulated HMGB1–Myc H322 cells. After 6 hours, Jurkat T cells expressing PD-1–Flag were washed and disrupted in a tube containing cell lysis buffer (FNN0021, Thermo Fisher Scientific) including protease inhibitor (P8340-1ML, Sigma), left on ice for 30 minutes, and then centrifuged at 12,000 rpm at 4°C for 30 minutes. For Western blot analysis, a small amount of the supernatant was collected and incubated with anti-Flag (2368, Cell Signaling Technology) or anti-Myc (2278, Cell Signaling Technology) at 4°C overnight with gentle shaking. Ten-microliter of pretreated protein A/G agarose beads (abs955, Absin) was added to the incubated cell lysate for 2 hours to induce immunoprecipitation and the mixture was centrifuged at 3,000 rpm for 5 minutes at 4°C. The supernatant was discarded gently and the agarose beads sedimented to the bottom of the tube before washing three times with cell lysis buffer (abs955, Absin). Finally, 20–40 μL 1× SDS sample buffer solution was added and the tube was placed in boiling water for 10 minutes prior to Western blotting.

Serum from 3 patients with NSCLC for whom we had prechemotherapy and postchemotherapy blood samples was collected and measured for the difference of cytokine expression levels using Customized Cytokine ELISA Strip (Cus-EA-0002, Signosis).

A CM5 chip (28953828, Cytiva) was activated for 420s using the Amine Coupling Kit (BR-1000-50, GE Healthcare), followed by CM5 coupling, and the target protein was treated with 10 mmol/L sodium acetate (pH 4.0; BR-1003-49, GE Healthcare), which was diluted to 30 μg/mL, coupled at a flow rate of 10 μL/minute for 500 seconds. The chip was blocked with ethanolamine (BR-1000-50, GE Healthcare) and the analyte (PD-1, Z03424, GenScript; HMGB1, HM1-H5220, ACRO; RAGE, 11629-HCCH, Sino Biological) was injected at a flow rate of 30 μL/minute for binding analysis. The testing equipment was Biacore 8K (GE Healthcare).

The cytoplasm and the membrane were separated with Mem-PER Plus Membrane Protein Extraction Kit (89842, Thermo Fisher Scientific). The Kit contains cell wash solution, permeabilization buffer, and solubilization buffer. Briefly, CD8⁺ T cells were harvested, centrifuged in the presence of cell wash solution, and incubated with permeabilization buffer for 10 minutes at 4°C. After centrifugation at 12,000 rpm at 4°C, the supernatants containing cytosolic proteins were collected. The pellets were suspended in solubilization buffer and recentrifuged at 12,000 rpm for 15 minutes at 4°C. Then the supernatants containing membrane proteins were collected. The expression level of PD-1 in the extracted cytoplasm and membrane proteins was detected by Western blotting.

Surface-bound proteins were depleted from CD8⁺ T cells by acid wash as described previously (21). In brief, 2 × 10⁵ CD8⁺ T cells from patients with NSCLC were washed with PBS once and transferred into ice-cold acid wash buffer (DMEM/HEPES, 10 mmol/L MES, pH 5.0; Thermo Fisher Scientific) for 20 minutes. Following that, cells were rinsed with acid wash buffer twice. Then cells were extensively washed with PBS and used in further assays.

NOD/SCID female mice were obtained from Charles River. The experimental animals were raised under sterile conditions in containers supplied with filtered air at 20°C–25°C. NOD/SCID mice were divided into four groups (n = 6): Sh-NC A549 tumors treated with HER2–CAR T cells, Sh-NC A549 tumors only, Sh-HMGB1 A549 tumors treated with HER2–CAR T cells, and Sh-HMGB1 A549 tumors only. NOD/SCID mice at 5 to 8 weeks of age, were engrafted subcutaneously with 2 × 10⁶ Sh-HMGB1 A549 cells or Sh-NC A549 cells. On day 7 after tumor engraftment, HER2–CAR T cells (6 × 10⁶) which were resuspended in 40 μL normal saline were injected intravenously according to groups; the control group mice were administered the same volume of normal saline. Three days later, half the number of the mice in each group were sacrificed, the tumor tissues were removed and weighed. Approximately 0.2 g tumor tissue from each mouse was ground to make a single-cell suspension. The expression of PD-1 and IFNγ by HER2–CAR T cells in each group was detected by using FACS analysis. Tumors of the remaining mice were measured using bioluminescence imaging, beginning at day 7 after engraftment and thereafter daily until day 28. Mice were intraperitoneally injected with d-luciferin (40902ES01, Yeasen Biotechnology), and images were acquired 10 minutes after the luciferin injection using an IVIS Spectrum (Perkin Elmer). All animal studies were approved by the Animal Care and Ethics Committee of the First Affiliated Hospital of Zhengzhou University (Zhengzhou, P.R. China).

Statistical analysis was performed using GraphPad Prism 7. Student t test was used to determine statistical significance between two groups. A one-way ANOVA was conducted to compare the differences among three or more variables. Spearman test was used to determine the correlation between HMGB1 and PD-1 levels. According to the median ranking of HMGB1 protein levels, patients were divided into low and high expression groups, as described previously (17). Statistical significance was set at P < 0.05.

All data supporting the findings of this study are available within the article and its Supplementary Data files or from the corresponding author upon reasonable request.

Mounting evidence suggests that chemotherapy exerts severe immune-modulating effects (22, 23). Thus, it is vital to detect the functional surface markers expressed on T cells in patients with NSCLC treated with or without chemotherapy. The percentage of CD8⁺ T cells expressing PD-1 on the surface in the peripheral blood of the chemotherapy group (n = 12) was markedly lower than that in the control group (n = 5), whereas the percentage of CD8⁺ T cells expressing Tim-3, CD69, and CD27 was no different between the groups (Fig. 1A). Compared with Tim-3, CD69, and CD27 expression in the prechemotherapy and postchemotherapy groups, the percentage of cells expressing PD-1 decreased considerably after chemotherapy (n = 3; Fig. 1B). CD69⁺ T cells have been reported to display the characteristics of activated T cells (24–26). We found that the percentage of CD8⁺CD69⁺ T cells expressing PD-1 in the chemotherapy group was markedly lower than that in the control group (Fig. 1C). Peripheral blood samples of healthy donors or patients with NSCLC who underwent chemotherapy were collected, CD8⁺ T cells were sorted and then reactivated using CD3/28 dyna-beads. We found that chemotherapy did not affect the reactivation ability of CD8⁺ T cells (Fig. 1D). Furthermore, we used a multi-factor screening assay to analyze the difference in cytokine expression levels in the serum of the 3 patients with NSCLC for whom had the prechemotherapy and postchemotherapy blood samples. Results showed that HMGB1 levels were increased in the postchemotherapy samples (Fig. 1E). HMGB1 acts as an important mediator of various inflammatory responses (5); therefore, we further evaluated the expression of HMGB1 in the cohort of patients from the chemotherapy group (n = 12) and control group (n = 5). The expression of HMGB1 in the peripheral blood of the chemotherapy group was significantly higher than for the control group (Fig. 1F). There was a negative correlation between the level of HMGB1 and expression of PD-1 (Fig. 1G). In summary, these results showed that in patients with NSCLC, postchemotherapy serum levels of HMGB1 were increased and expression of PD-1 on CD8⁺ T cells was decreased.

The relationship between HMGB1 release and the level of expression of PD-1 on CD8⁺ T cells was further examined. To test the hypothesis that HMGB1 regulates the expression level of PD-1, we used rHMGB1 to treat CD8⁺ T cells derived from either peripheral blood or tumor tissues of patients with NSCLC and found that PD-1 expression decreased (Fig. 2A). Even when blocking RAGE, which is a classical receptor of HMGB1 (27, 28), rHMGB1 decreased PD-1 expression in CD8⁺ T cells (Fig. 2B). HMGB1 had no effect on the expression of PD-1 mRNA (Fig. 2C). Similar results were obtained using PD-1–mCherry Jurkat T cells (Supplementary Fig. S1A and S1B). Therefore, we examined whether HMGB1 interferes with PD-1 downregulation. PD-1–mCherry Jurkat T cells were treated with HMGB1–GFP, which was secreted from HMGB1–GFP H322 cells. Image cytometry analysis showed that PD-1 was colocated with HMGB1 (Fig. 2D and E). Coimmunoprecipitation analysis further demonstrated that HMGB1 was physically associated with PD-1 (Fig. 2F; Supplementary Fig. S2). The association of HMGB1 to PD-1 further suggested that HMGB1 interferes with PD-1 downregulation. However, the Biacore assay confirmed that HMGB1 and PD-1 proteins did not to interact directly (Fig. 2G and H).

To further examine the biological activity of HMGB1 indirect binding to PD-1, we assessed the level of PD-1 on permeabilized CD8⁺ T cells and unpermeabilized CD8⁺ T cells. Permeabilized CD8⁺ T cells treated with rHMGB1 expressed similar PD-1 levels as control cells, only unpermeabilized CD8⁺ T cells treated with rHMGB1 had lower PD-1 levels (Fig. 3A). In addition, CD8⁺ T cells treated with HMGB1 for 3 or 6 hours expressed similar levels of total PD-1 as the control, as assessed by Western blotting (Fig. 3B). We inferred that PD-1 was internalized after binding with HMGB1 indirectly, and the expression of PD-1 on the surface of the membrane decreased without damaging the membrane (Fig. 3C).

Dynamin is a key factor in endocytosis and synaptic vesicle recycling (29, 30). It plays a key role in these processes by recruiting various proteins to participate in the endocytosis of synaptic vesicles through protein–protein and protein–lipid interactions (31). Pretreatment of cells with dynasore, an inhibitor of dynamin, partially restored the expression of PD-1 on the membrane of CD8⁺ T cells treated with rHMGB1 for 6 hours (Fig. 3D). Proteins in the cytoplasm and membrane were extracted, and the expression level of PD-1 was detected by Western blotting. Results indicated that treatment with HMGB1 in combination with dynasore led to a markedly higher PD-1 level in the membrane and lower PD-1 level in the cytoplasm compared with that in HMGB1 monotherapy (Fig. 3E). Image cytometry analysis using the PD-1–mCherry Jurkat T cells showed that PD-1 internalization was inhibited by dynasore (Fig. 3F). The expression of PD-1 in PD-1–GFP Jurkat T cells treated with HMGB1 showed a change in the aggregation and localization of PD-1 in the cells, as detected using IF (Supplementary Fig. S3). In addition, PD-1–GFP Jurkat T cells treated with HMGB1 and dynasore showed inhibited internalization but not aggregation of PD-1 (Supplementary Fig. S4). These results support the idea that HMGB1-induced PD-1 internalization is mediated by dynamin.

EEA1 is a protein antigen that participates in the formation of early endosomes (32). Early endosomes are cellular compartments involved in accepting and sorting of endocytosed proteins for vesicular transport via late endosomes and lysosomes or returning the endocytosed proteins to the plasma membrane (33, 34). The internalization mechanism of PD-1–HMGB1 was elucidated by intracellular staining of EEA1 and PD-1 proteins (35). Colocalization of PD-1 and EEA1 occurred at different timepoints after HMGB1 treatment (Fig. 4A). Proteins in the cytoplasm and membrane were extracted after 12 hours HMGB1 treatment for the detection of PD-1 expression levels. Results indicated that HMGB1 treatment led to markedly lower levels of PD-1 in the membrane and higher levels in the cytoplasm compared with levels detected in the control group (Fig. 4B). In addition, results of different timepoints showed that HMGB1 treatment for 24 hours led to a markedly lower overall PD-1 expression than that in the control group (Fig. 4C). Similar data were obtained with CD8⁺ TILs derived from 2 patients with NSCLC; treatment of these cells with HMGB1 reduced PD-1 levels compared with levels in untreated CD8⁺ TILs (Fig. 4D). Compared with HMGB1 monotherapy, combination HMGB1 and CQ treatment led to a markedly higher PD-1 level of T cells in the cytoplasm and lower level in the membrane (Fig. 4E). To further confirm the lysosomal degradation of PD-1 resulted from internalization but not HMGB1 chronic stimulation, T cells from PBLs were subjected to acid wash after 3-hour incubation with HMGB1. As expected, the degradation of internalized PD-1 was still blocked by CQ when surface-bound HMGB1 was depleted (Fig. 4F). These results indicated that CQ exhibited an inhibitory effect on the degradation of PD-1 in lysosomes. In addition, we found that when the proteasome was inhibited, surface expression of PD-1 was not recovered after HMGB1 incubation, suggesting that proteasome may not affect the degradation of internalized PD-1 induced by HMGB1 (Supplementary Fig. S5).

We found that after blocking RAGE rHMGB1 could still cause an increase in IFNγ release by the CD8⁺ T cells (Fig. 5A). Even though treated with HMGB1, knockout of the gene-encoding PD-1 in CD8⁺ T cells derived from PBLs using the CRISPR gene editing tool still inhibited T-cell function upregulation (Fig. 5B and C). On the basis of multiplex IF analysis of NSCLC tissues, we found that sections with high expression of HMGB1 had low expression of PD-1, whereas sections with low expression of HMGB1 exhibited high expression of PD-1 (Supplementary Fig. S6). On the basis of the expression level of HMGB1, patients were divided into high and low HMGB1 expression groups. Patients with NSCLC that had higher levels of HMGB1 in tissue specimens, had decreased expression of PD-1 and increased secretion of IFNγ in CD8⁺ T cells derived from PBLs (Fig. 5D and E). These results further confirmed that HMGB1 increases T-cell function by downregulating PD-1.

The function of HER2-targeted CAR T cells was increased with treatment of rHMGB1 (Supplementary Fig. S7). HMGB1 expressions in tumor cells were significantly higher than that in CD8⁺ T cells (Supplementary Fig. S8). Furthermore, after coincubation of A549 tumor cells modified to express Sh HMGB1 or the control (Sh-NC) with HER2-targeted CAR T cells, the level of HMGB1 in the culture supernatant was detected by ELISA. The release of HMGB1 was upregulated in Sh-NC–A549 tumor cells treated with HER2-targeted CAR T cells. The release of HMGB1 by Sh-HMGB1–A549 cells was reduced whether or not the cells were treated with HER2-targeted CAR T cells (Fig. 6A). We also assessed whether HMGB1 knockdown impacted tumor-cell proliferation. The results showed that Ki67 expression and proliferation of tumor cells were not affected (Supplementary Fig. S9). To explore the antitumor effect of HMGB1 in vivo, we established a xenograft mouse model. HMGB1 levels in tumor tissue were detected by performing IHC assays for the different treatment groups. The expression of HMGB1 was the highest in Sh-NC–A549 tumors from mice treated with HER2-targeted CAR T cells (Fig. 6B). Sh-HMGB1–A549 tumor cells harboring HER2 were subcutaneously injected into mice. In mice treated with HER2-targeted CAR T cells, those that had tumors with HMGB1 knockdown had increased tumor load, enhanced PD-1 expression, and reduced IFNγ and granzyme B expressions compared with those that had control tumors (Fig. 6C–F). Taken together, these results demonstrated that knockout of HMGB1 upregulated PD-1 expression, leading to decreased CAR T-cell function in vivo.

Numerous studies have shown that the TME plays an important role in regulating the immune response of tumors (36). The abundant TILs in “hot tumors,” which indicates a good antitumor immune response in the early stage of cancer treatment, is often short-lived due to upregulation of immune checkpoint expression or an increase in the number of immunosuppressive cells (37–39). A “hot tumor” does not guarantee a prolonged antitumor effect due to progressive loss of function by CD8⁺ T cells. Therefore, maintaining the antitumor capability of CD8⁺ T cells after tumor infiltration remains an immensely challenging task. In this study, we demonstrated that chemotherapy induced the upregulation of HMGB1 expression. Our results confirmed that in the TME, the function of CD8⁺ T cells was markedly enhanced in tissues that express high levels of HMGB1. In addition, the expression level of PD-1 on CD8⁺ T cells decreased markedly after treatment with HMGB1. Therefore, we inferred that HMGB1 can enhance immune-cell function by downregulating PD-1 expression.

Under normal conditions, the release of endogenous nucleic acids from dead cells triggers the secretion of HMGB1 into the TME, which signifies a mechanism of cellular damage control (40). In the TME, the combination of HMGB1 and nucleic acids activates the initial immune response and inhibits tumor progression (41). It has been reported that Tim-3 binding to HMGB1 suppresses the formation of HMGB1–DNA complexes, which negatively regulates HMGB1-mediated nucleic acid localization in DCs, thereby suppressing the antitumor immune response mediated by tumor-related DCs (42). However, the mechanism by which HMGB1 affects the function of CD8⁺ T cells in the TME of lung cancer without relying on DCs remains elusive. Our studies focused on seeking to understand how HMGB1 affects the secretion of cytokines by CD8⁺ T cells and their ability to kill tumor cells. HMGB1 can promote the maturation of DCs and the proliferation of T cells by binding to RAGE on the surface of DCs and T cells (28, 43). Therefore, in this study, we blocked RAGE on the surface of T cells (and confirmed that HMGB1 could enhance the function of T cells by binding to PD-1 indirectly. However, HMGB1 and PD-1 proteins seemed not to interact directly. In contrast, PD-L1 and PD-1, and HMGB1 and RAGE, showed strong positive signals in Biacore tests. We speculated that other surface proteins were involved in the interaction between HMGB1 and PD-1 on the cell membrane. But the exact proteins remain to explore in the future.

A study reported that the release of HMGB1 from tumors after neoadjuvant chemoradiotherapy treatment had a better clinical outcome due to the maturation of DCs and the subsequent recruitment of PD-1⁺ TILs (44). In our previous studies, we also found that HMGB1 induces the maturation of DCs (45) and verified that HMGB1 increases the infiltration of TILs to tumor sites by promoting the secretion of chemokine CXCL11 (17). These findings support our hypothesis that HMGB1 induces an increase in antitumor function by reducing the expression of PD-1 in TILs.

CAR T-cell therapy has an outstanding therapeutic effect on hematologic malignancies; however, its efficacy in solid tumors remains dismal (46). HER2 is highly expressed in lung cancer cells and is an oncogene target for targeted therapy (trastuzumab) that has yielded a good therapeutic effect. Therefore, we selected HER2 as the CAR T-cell target antigen for the NSCLC model (47, 48). HER2-targeted CAR T cells demonstrated a killing effect both in vitro and in vivo. In the tumors treated with HER2-targeted CAR T cells, HMGB1 knockdown resulted in increased tumor volume, as opposed to that in the group without HMGB1 knockout. Results clearly indicated that HMGB1 can upregulate the killing ability of CAR T cells on tumor cells.

Many cells, including CD8⁺ T cells and tumor cells, are able to produce HMGB1. To verify which type of cells is the main origin of HMGB1, we compared the levels of HMGB1 in CD8⁺ T cells and tumor cells. We found that HMGB1 expression in tumor cells was significantly higher than that in CD8⁺ T cells. Moreover, HMGB1 expressions in tumor cells were significantly upregulated after coculture with HER2-targeted CAR T cells. Hence, tumor cells may play a more prominent role in HMGB1 secretion than CD8⁺ T cells.

In summary, we have demonstrated that HMGB1 binds to PD-1 to induce dynamin-mediated internalization, followed by forming early endocytosis in the cytoplasm, which were transferred into the lysosomes for degradation. Concurrently, HMGB1 enhanced CD8⁺ T-cell function by downregulating the expression level of PD-1 (Fig. 7).

No disclosures were reported.

Q. Gao: Data curation, writing–original draft. S. Wang: Data curation, investigation, methodology. F. Li: Conceptualization, methodology. J. Lian: Methodology. S. Cheng: Data curation. D. Yue: Resources. Z. Zhang: Investigation. S. Liu: Methodology. F. Ren: Methodology. D. Zhang: Methodology. S. Wang: Supervision. L. Wang: Resources. Y. Zhang: Project administration, writing–review and editing.

This study was supported by grants from the National Natural Science Foundation of China (grant nos. 81902924, U1804281, 91942314), and the National Major Scientific and Technological Special Project (grant no. 2020ZX09201009).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.