Article Text
PDF
Abstract
Background Although the roles of PD-L1 in promoting tumor escape from immunosurveillance have been extensively addressed, its non-immune effects on tumor cells remain unclear.
Methods The spatial heterogeneity of PD-L1 staining in human papillomavirus (HPV)-positive head and neck squamous cell carcinoma (HNSCC) tissues was identified by immunohistochemistry. Three-dimensional (3D) specific cell-led invasion assay and 3D cancer spheroid model were used to investigate the roles of PD-L1hileader cells in collective invasion. The impact of M1 macrophages on specific PD-L1 expression in leader cells and its mechanisms were further studied. Finally, the effect of combination therapy of anti-PD-L1 and CDK4 inhibitor on HPV-positive tumors were evaluated on a mice model.
Results Here, we observed a distinctive marginal pattern of PD-L1 expression in HPV-positive HNSCC tissues. By mimicking this spatial pattern of PD-L1 expression in the 3D invasion assay, we found that PD-L1hi cells led the tumor collective invasion. M1 macrophages induced specific PD-L1 expression in leader cells, and depletion of macrophages in tumor-bearing mice abrogated PD-L1hileader cells and collective invasion. Mechanistically, TNF-α secreted by M1 macrophages markedly increased the abundance of PD-L1 via CDK4/ubiquitin-specific peptidase 14-mediated deubiquitination of PD-L1. We also found that suppression of CDK4 enhanced the efficacy of anti-PD-L1 therapy in an E6/E7 murine model.
Conclusions Our study identified TNF-α/CDK4/ubiquitin-specific peptidase 14-mediated PD-L1 stability as a novel mechanism underlying M1 macrophage-induced PD-L1hileader cells and collective tumor invasion, and highlighted the potential of the combination therapy of anti-PD-L1 and CDK4 inhibitor for HPV-positive HNSCC.
- immune checkpoint inhibitors
- tumor microenvironment
- macrophages
- drug therapy, combination
- head and neck neoplasms
Data availability statement
All data relevant to the study are included in the article or uploaded as online supplemental information.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
- immune checkpoint inhibitors
- tumor microenvironment
- macrophages
- drug therapy, combination
- head and neck neoplasms
WHAT IS ALREADY KNOWN ON THIS TOPIC
The response rate of the patients with human papillomavirus (HPV)-positive head and neck squamous cell carcinoma (HNSCC) to PD-1/PD-L1 blockade is only ~20%. Although it is well illustrated that PD-L1 participates in tumor escape from immunosurveillance, little is known about tumor-intrinsic roles of PD-L1 in modulating tumor biology. In addition, several studies have identified the heterogeneity of PD-L1 expression in tumor nest, but its underlying mechanisms and clinical implication remain unclear.
WHAT THIS STUDY ADDS
We demonstrated that PD-L1-positive leader cells at the periphery of the tumor nests triggered collective tumor invasion. Mechanistically, M1 macrophages induced PD-L1hi leader cells via CDK4/ubiquitin-specific peptidase 14-mediated deubiquitination of PD-L1. We also found that suppression of CDK4 enhanced the efficacy of anti-PD-L1 therapy in an E6/E7 murine model.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
This study increases our knowledge of the mechanisms of the spatial heterogeneity of PD-L1 expression and the tumor-intrinsic effects of PD-L1 in HPV-positive HNSCC. These findings indicated that the combination therapy of anti-PD-L1 and CDK4 inhibitor is a potential treatment for HPV-positive HNSCC.
Background
Head and neck squamous cell carcinoma (HNSCC) is the sixth most common human cancer, with over 600,000 new cases reported worldwide each year.1 Human papillomavirus (HPV) infection is the third most common risk factor for HNSCC, after smoking and alcohol consumption.2 Although patients with HPV-positive HNSCC have a better prognosis than those with HPV-negative HNSCC, ~ 25% of HPV-positive cases are lethal. Therefore, novel therapeutic approaches to improve the outcome of patients with HPV-positive HNSCC are necessary.
PD-1/PD-L1 blockade is an essential part of therapy in multiple cancers.3 4 Patients with recurrent HPV-positive HNSCC derive more clinical benefit from immunotherapy than those with non-HPV-associated HNSCC; further, most patients who respond to therapy show a durable response and significant improvement in overall survival.5 However, the response rate to PD-1/PD-L1 blockade is only ~20%, which impedes its extensive use.4 Most of the research on this pathway has focused on the tumor-extrinsic effects of PD-L1, particularly its effects on T cells. However, recent studies have identified tumor-intrinsic roles for PD-L1 in modulating cancer stemness, epithelial-mesenchymal transition (EMT), tumor glucose metabolism, and resistance to therapy.6 7 These data provide a new perspective for understanding the reasons for the constraints of anti-PD-L1/PD-1 therapy and offer unique opportunities for innovative therapies.
Invasion of malignant cells into the basement membrane and extracellular matrix is a critical step in cancer metastasis. In most epithelial cancers tumor cells disseminate via multicellular groups that retain cell–cell connections.8–10 These tumor cell clusters with collective motility are composed of heterogeneous populations of tumor cells that polarize into “leader cells” and “follower cells”.11 12 Leader cells are located at the invasive front of these multicellular groups and steer their movement.11 Previous studies by us and others have identified key markers of tumor leader cells, such as cytokeratin-14, p63, cathepsin B, and integrin β.13–15 Recent studies have reported PD-L1-positive tumor cells to be limited to the periphery of tumor nests at the tumor–host interface in HPV-positive HNSCC, indicating a potential role of PD-L1 in tumor leader cells.16 17 This distinct architectural pattern of PD-L1 expression has attracted attention, but its clinical implications and underlying mechanisms have not been elucidated.16 17
Here, we observed a distinctive marginal pattern of PD-L1 expression in HPV-positive HNSCC using immunohistochemistry (IHC). To recapitulate this architecture pattern, we used a specific cell-led three-dimensional (3D) invasion assay and found that PD-L1hi leader cells promoted collective invasion of tumor cells via partial EMT. We further demonstrated that M1 macrophages induced specific expression of PD-L1 in leader cells via the CDK4/ubiquitin-specific peptidase 14 (USP14)-mediated deubiquitination of PD-L1. Of note, suppression of CDK4 enhanced the efficacy of anti-PD-L1 therapy in an E6/E7 murine tumor model, highlighting the potential of this combination for treating HPV-positive HNSCC.
Materials and methods
Additional information is provided in online supplemental materials.
Supplemental material
Animals and experimental design
For the 4NQO-induced HNSCC model, Rosa26-E6-E7 constitutive knock-in C57BL/6 mice (ID: TOS150814BA1) were purchased from Chengdu Dashuo Biological Technology (Sichuan, China). The mice were maintained at 22°C–25°C and 30%–70% relative humidity under a 12 hours light/dark cycle. After 1 week of acclimation, mice were administered either water or 100 µg/mL 4NQO in drinking water for 10 weeks and then switched to distilled water for 10 weeks according to the procedures described in our previous studies.18 19 After that, tumor-bearing mice were matched by tumor size and then randomly divided into four groups (n=6 each group): control (vehicle), palbociclib, anti-PD-L1 antibody, and combination treatment. Control IgG and anti-PD-L1 antibody treatments were administered by intraperitoneal injection (200 µg/mouse in PBS buffer) twice a week for 4 weeks. Vehicle and palbociclib (100 mg/kg in 50 mM sodium lactate, pH=4) were administered daily by oral gavage for 4 weeks. Mice were sacrificed and tongue tissues were collected for analysis at the end of the experiment.
For macrophage depletion model, 2×106 SCC47 cells suspended in 200 µL 25% Matrigel/PBS (BD Biosciences) were subcutaneously injected into the flanks of 5-week-old female nude mice (6 mice/group). After the formation of palpable tumors, mice were administered 100 µL clodronate liposomes or control liposomes by intraperitoneal injection twice a week. Tumor size was measured every 3 days, and the tumor volume was calculated using the formula V=(width2×length)/2. After 5 weeks, mice were sacrificed, and tumor tissues were collected for further analysis. The invasive front of the tumor was examined using microscopy after H&E and panCK staining, and the density of panCK-positive collective invasion packs (CIPs) ranging in size from 5 to 120 cells was quantified by two pathologists.
For the animal survival model, 2×106 SCC7 cells were subcutaneously inoculated in C3H mice. The indicated treatments were initiated on day 7 after inoculation, when the diameter of the tumors reached 2–4 mm. The mice survival of each treatment was monitored and recorded as the percentage of survivors.
Cell lines
HPV+ HNSCC cell lines (SCC47 and SCC090), HPV− cell lines (Cal27, SCC7), and the THP-1 cell line were obtained from the State Key Laboratory of Oral Diseases of Sichuan University. SCC47 and Cal27 cells were cultured in DMEM with 10% fetal bovine serum. SCC090 cells were cultured in MEM with 1% non-essential amino acids, 2 mM L-glutamine, and 10% fetal bovine serum. Conditioned medium (CM) was collected from tumor cells incubated in serum-free medium for 48 hours. THP-1 cells were cultured in RPMI 1640 supplemented with penicillin/streptomycin and 10% fetal bovine serum.
Patient samples
We included 73 HPV-positive and 23 HPV-negative HNSCC samples and 10 normal oral tissues resected from patients between January 2005 and December 2015 at the West China Hospital of Stomatology of Sichuan University (Chengdu, China). The HPV infection status of all samples was evaluated by p16 staining before inclusion.
Immunohistochemistry
Tumor samples were sliced into 4 µm sections and used for routine immunohistochemical staining procedures.13 Briefly, the samples were deparaffinized and rehydrated, and antigen retrieval was performed by heating the slides in citrate or EDTA solution at 99–100°C for 20 min. After blocking with 3% hydrogen peroxide and goat serum albumin, the slides were incubated with primary antibodies at 4°C overnight. The following primary antibodies and dilutions were used: from Proteintech, anti-CD68, 1:200; anti-CD163, 1:600; anti-CD80, 1:200; and anti-PD-L1, 1:300; and from Cell Signaling, anti-panCK, 1:300; anti-F4/80, 1:400; anti-TNF-α, 1:400; and CD8a, 1:300. Immunostaining for p16ink4a was performed using the CINtec p16ink4A Histology kit (Dako, Denmark) according to the manufacturer’s instructions. After staining with diaminobenzidine and counterstaining with hematoxylin, the slides were dehydrated in graded ethanol, vitrificated by dimethylbenzene, and mounted.
Immunostaining was analyzed by two independent pathologists blinded to the sample identification. HPV infection status was evaluated by p16 staining; cytoplasmic staining in >70% of tumor cells was defined as p16(+). A 5% threshold of cell surface expression of PD-L1 in tumor cells was interpreted as positive.20 CD68-positive, CD80-positive, and CD163-positive cells in five random high power fields at 200×magnification were counted. CIPs were identified as panCK-positive epithelial cancer cell clusters in size from 5 to 120 cells invading from primary site. The tumor–host interfaces were examined and the number of CIPs in 10 random high power fields at 200×magnification were counted and averaged in each field. The evaluation was performed by two investigators blinded to grouping.
Spheroid formation and invasion assays
Briefly, 200 µL of tumor cells resuspended in DMEM at a density of 1–2×104 /mL were seeded into 96-well ultra-low attachment plates (Corning, Corning, New York, USA). After 48 hours, multicellular spheroids of tumor cells were formed. To embed spheroids, 170 µL of cell culture medium was removed from each well, and 70 µL of type I collagen solution was added. After incubation at 37°C for 45 min, 100 µL of growth medium was added to the well. siRNA-transfected or plasmid-transfected cells were used for spheroid generation 48 hours after transfection. Spheroid invasion was evaluated using images taken at 0 and 24–48 hours postembedding using an Olympus IX51 microscope. Invasive area was quantified by measuring both the total spheroid area and the inner spheroid core area using Image-Pro Plus V.6.0 and assessed using the following formula: 100×(total spheroid area−inner spheroid core)/well total spheroid area. The number of chains was quantified by counting invasive chains per spheroid and used to assess the ability of leader cells to drive collective invasion of the spheroids.15 The evaluation was performed by two investigators blinded to grouping.
Specific cell-led spheroid invasion assay
One thousand tumor cells to be used as follower cells (for example, non-fluorescent SCC47 cells) were seeded in 96-well ultra-low attachment plates (Corning) to form a primary tumor spheroid. After 24 hours, 1000 tumor cells of another type (GFP-labeled SCC47 cells) were added to each well as leader cells that would be located at the peripheral area of the tumor spheroids. After 24–48 hours, one final spheroid was generated in each well and embedded in collagen solution for the spheroid invasion assay.
Real-time RT-PCR analysis
Total RNA was extracted using TRIzol reagent (Roche, Indianapolis, Indiana, USA), and 1 µg of total RNA was reverse transcribed into cDNA using a Maxime RT PreMix kit (iNtRON Biotechnology, Gyeonggi, Korea) according to the manufacturer’s instructions. Quantitative PCR was performed using PCR premix (Bioneer Co., Daejeon, Korea) with primers specific for the target genes (online supplemental table S1). PCR reactions were run in triplicate on 96-well plates using a QuantStudio 3 Real Time PCR system (Applied Biosystems). Gene expression was normalized to that of GAPDH.
Statistical analysis
Statistical analysis was performed using Prism V.6.07. Statistical significance of differences between two treatment groups was tested using Student’s t-test. Multiple comparisons among groups were performed using one-way or two-way analysis of variance. Survival analysis of mice was determined by the log-rank test.
Results
HPV-positive HNSCC exhibited a unique PD-L1 expression pattern
We evaluated PD-L1 immunostaining in HPV-negative and HPV-positive HNSCC, which were distinguished by p16 staining. Positive PD-L1 staining was observed in 83.6% (61/73) of the HPV-positive carcinomas, whereas only 43.5% (10/23) of the HPV-negative carcinomas were PD-L1-positive (p<0.001). We observed two different architectural patterns of PD-L1 expression: diffuse and marginal patterns (figure 1A). The tumor cells formed multicellular nests with a typical epithelial morphology. In 90% (9/10) of the HPV-negative HNSCC samples with PD-L1 expression, PD-L1 staining was diffuse throughout the tumor nests, and the marginal pattern was observed in only one sample (online supplemental table S2). However, 78.7% (48/61) of the PD-L1-positive HPV-positive HNSCC samples showed a distinctive marginal staining pattern, in which, PD-L1-positive tumor cells were predominantly located at the periphery of the tumor nests or tumor buds adjacent to the tumor–stroma interface.
PD-L1hi leader cells promote cancer cell collective invasion in HPV-positive HNSCC. (A) Representative images of PD-L1 immunostaining patterns in HPV-negative, HPV-positive HNSCC, and normal tissue. There was no PD-L1 staining in normal tissue, but PD-L1 was prominently positive in HNSCC tissue with two different architectural patterns, diffuse and marginal staining. Representative images of negative and diffuse staining from HPV-negative cases and marginal staining from two HPV-negative cases are provided. Scale bars represent 100 µm. (B–D) A mixed spheroid three-dimensional (3D) invasion assay was performed to determine whether PD-L1hi cancer cells can polarize to the leading position. (B) Scheme of the mixed spheroid 3D invasion assay. Mixed spheroids were generated by coculturing PD-L1-overexpressing mCherry-positive cells and control GFP-positive cells or by coculturing control mCherry-positive cells and PD-L1-overexpressing GFP-positive cells. (C) Representative images of mixed spheroid 3D invasion assay after 48 hours. Higher magnification images of the delineated regions are shown in the zoomed panels. Arrowheads indicate leader cells in collective invasion strands. Scale bars represent 1000 µm. (D) The percentages of GFP-positive or mCherry-positive cells leading collective invasion strands were quantified. Graphical values illustrate the mean±SD of n=10/group. ***p<0.001, assessed by t-test. (E, F): Scheme of a specific cell-led 3D invasion assay by two-stage spheroid formation. (E) One thousand non-fluorescent cells were seeded in 96-well ultra-low attachment plates for primary spheroid formation (as the core of a secondary spheroid), and after 24 hours, GFP-labeled SCC47 cells were added to each well as leader cells for final spheroid formation. Scale bars represent 500 µm. (F) PD-L1hi cell-led and control cell-led spheroids were generated by two-stage spheroid formation. (G–H) A 3D invasion assay was performed to detect the invasion ability of PD-L1hi cell-led and control cell-led spheroids. Scale bars represent 500 µm. (H) Quantification of invasive area and the number of chains per spheroid from each group. Graphical values illustrate the mean±SD of n=10/group. **p<0.01, ***p<0.001, assessed by t-test. HNSCC, head and neck squamous cell carcinoma; HPV, human papillomavirus; PD-L1, programed death ligand 1.
PD-L1hi leader cells promoted collective invasion of cancer cells in HPV-positive HNSCC
Marginal PD-L1 expression at the invasive front of HPV-positive HNSCCs indicated that PD-L1-expressing cells could be the leader cells in a collective invasion. To investigate this hypothesis, we assessed the invasion pattern in HPV-positive HNSCC samples using IHC. At the tumor–host interface, panCK-stained epithelial cancer cells invaded the adjacent glands, vessels, muscles, and nerve tissues as multicellular units (online supplemental figure 1). In vitro, PD-L1 knockdown in cancer cells resulted in a significant reduction in the expression of the mesenchymal markers, N-cadherin (CDH2) and vimentin (VIM), whereas E-cadherin (CDH1) expression was retained (online supplemental figure 2A–D). These data indicate that PD-L1 expression in cancer cells induces partial EMT, which is crucial for collective invasion of cancer cells.21
To determine whether PD-L1hi cancer cells could polarize to the leading position, we performed a mixed spheroid 3D invasion assay of PD-L1low and PD-L1hi cancer cells (figure 1B). After embedding in collagen for 24–48 hours, mCherry-labeled or GFP-labeled PD-L1hi SCC47 cells were observed at the leading tip of the invasive strands in 70.9% and 71.4% of spheroids, respectively, whereas there were only a few GFP-labeled or mCherry-labeled PD-L1low SCC47 cells at the leading tip (figure 1C,D). Similar results were observed for SCC090 spheroids (online supplemental figure 2E,F). These results suggest that PD-L1hi cancer cells can polarize to the leading edge of multicellular clusters. To further explore the roles of PD-L1-positive tumor cells at the leading edge, we developed a specific cell-led 3D invasion assay using two-stage spheroid formation (figure 1E). PD-L1low cell-leading tumor spheroids showed indolent behavior without protrusions, whereas PD-L1hi cell-led tumor spheroids extensively invaded into the collagen, exhibiting markedly protrusive borders (figure 1F–H and online supplemental figure 2G,H). Spheroids composed of PD-L1hi cells invaded the collagen more efficiently and exhibited more invasive strands than did spheroids composed of PD-L1low cells (online supplemental figure 2I–L). In addition, wound healing and Transwell assays revealed that PD-L1hi cells were more motile and had higher invasion efficiency than PD-L1low cells (online supplemental figure 3). Overall, these data revealed that PD-L1-positive tumor cells at the invasive front led to the collective invasion of HPV-positive HNSCC cells.
Specific PD-L1 expression in cancer leader cells was induced by M1 macrophages
Based on the spatial proximity of tumor and stromal cells at the invasive front, we hypothesized that marginal PD-L1 expression in HPV-positive HNSCC was induced by inflammatory cells. Macrophages have been reported to induce PD-L1 expression in cancer cells.22 To identify whether macrophages contribute to the marginal expression of PD-L1, we examined the expression of the pan-macrophage marker CD68, M1 macrophage marker CD80, and M2 macrophage marker CD163 in HPV-positive HNSCC samples. IHC of serial sections revealed abundant macrophage infiltration at the tumor–stroma interface, adjacent to PD-L1-positive tumor cells (online supplemental figure 4A). Notably, the infiltration of CD68+ and CD80+ macrophages was significantly higher in samples with a peripheral PD-L1 pattern than in samples with a diffuse pattern, whereas there was no difference in the infiltration of CD163+ macrophages (online supplemental figure 4B,C). In addition, there was greater infiltration of CD68+ and CD80+ macrophages in HPV-positive HNSCC than in HPV-negative HNSCC. Our analysis of TCGA datasets also revealed that PD-L1 expression was higher in HPV-positive HNSCC than in HPV-negative HNSCC, and PD-L1 was positively correlated with CD80 (r=0.478) (online supplemental figure 4D,E). Overall, these data indicate that peripheral expression of PD-L1 in tumor cells is associated with increased tumor-infiltrating M1 macrophages.
The roles of macrophages in PD-L1-expressing leader cell-driven collective invasion were further investigated. In spheroids treated with CM from M1 macrophages, PD-L1-positive cells were predominantly distributed at the leading edge, and follower cells were lack of PD-L1 expression. Conversely, spheroids treated with control medium had low and more uniform PD-L1 expression (figure 2A and online supplemental figure 5A). Further analysis revealed that CM from M1 macrophages increased PD-L1 protein levels, but not mRNA levels (online supplemental figure 5B). Moreover, treatment with CM from M1 macrophages promoted the collective invasion of tumor spheroids, increasing both the number of invasive chains and invasion distance (figure 2B,C and online supplemental figure 5C,D). PD-L1 knockdown inhibited the invasive phenotypes induced by M1 macrophage CM, suggesting that PD-L1 expression is required for M1 macrophage-mediated collective invasion of cancer cells (figure 2D,E and online supplemental figure 5E).
M1 macrophages promote collective invasion by inducing specific PD-L1 expression in cancer leader cells. (A) Representative IF of PD-L1 in M1 CM-stimulated SCC47 spheroids and control-stimulated spheroids. Scale bars represent 500 µm. (B, C) Three-dimensional (3D) invasion assay of SCC47 spheroids treated with CM from M0, M1, or M2 macrophages or vehicle. Scale bars represent 1000 µm. (C) Quantification of invasive area and the number of chains per spheroid from each group. Graphical values illustrate the mean±SD of n=10/group. *p<0.05, **p<0.01, ***p<0.001, assessed by one-way ANOVA. (D, E) SCC47 spheroids were transfected with PD-L1 siRNA (siPD-L1) or control siRNA (siCtrl). Then, spheroids were treated with M1 macrophage CM for a 3D invasion assay. Scale bars represent 1000 µm. (E) Quantification of invasive area and the number of chains per spheroid from each group. n=10/group, ***p<0.001, assessed by t test. (F–G) SCC47 tumor-bearing mice were treated with clodronate or control liposomes for 5 weeks (n=6/group). (F) Representative immunostaining of PD-L1, F4/80, and TNF-α in serial sections from three mice per group. Scale bars represent 50 µm. (G) Representative images of H&E staining and panCK immunostaining from six mice/group. Arrowheads indicate collective invasion packs (CIPs) marked with panCK staining. Scale bars represent 100 µm. (H) Q Quantification of the density of CIPs. Graphical values illustrate the mean±SD **p<0.01, assessed by t-test. ANOVA, analysis of variance; CM, conditioned media; IF, immunofluorescence; PD-L1, programed death ligand 1; panCK, pancytokeratin.
To address whether macrophages are necessary for the induction of PD-L1 in cancer leader cells in vivo, SCC47 tumor-bearing mice were injected with clodronate liposomes to deplete macrophages. Macrophage depletion was confirmed by the reduction in pan-macrophage infiltration (F4/80) and TNF-α levels in tumor tissues (figure 2F). Consistent with previous observations,23 macrophage depletion inhibited tumor growth in mice (online supplemental figure 5F,G). The tumor tissues of control mice displayed a marginal PD-L1 pattern with strong staining, whereas macrophage-depleted mice showed a diffuse PD-L1 pattern with weak staining (figure 2F). To determine the effect of macrophage depletion on collective tumor invasion, we assessed the density of CIPs of cancer cells, which are positive for the epithelial marker panCK and vary in size ranging from 5 to 120 cells.24 We observed that macrophage depletion reduced the number of CIPs disseminating from the primary tumor into the surrounding stroma (figure 2G,H). These data indicate that macrophages are necessary for the induction of PD-L1 in cancer cells and the resulting invasive nature.
TNF-α-induced PD-L1 expression promoted collective invasion of HPV-positive HNSCC
To determine which major cytokines secreted by macrophages are involved in the upregulation of PD-L1 in HPV-positive HNSCC, the cytokine levels in CM from macrophages preconditioned by Cal27 and Cal27-E6/E7 cells were analyzed using a cytokine array. The levels of TNF-α, MIP-1α, CCL5, and IL-1ra in the media from macrophages preconditioned by Cal27-E6/E7 cells were more than 2-fold higher than those in the media from macrophages preconditioned by Cal27 cells, and TNF-α exhibited the highest increase (21.4-fold) (figure 3A,B). ELISA showed that the TNF-α level in the CM from macrophages preconditioned by Cal27-E6/E7 cells was 11.2-fold higher than that in CM from control macrophages (figure 3C).
M1 macrophages promote cancer cell collective invasion via TNF-α-induced PD-L1 expression in leader cells. (A, B) The levels of 36 cytokines in macrophage supernatant preconditioned by Cal27 or Cal27-E6/E7 cells for 72 hours were analyzed using a cytokine array. (B) The pixel densities of indicated cytokines in the array were quantified and normalized to reference spots. (C) The level of TNF-α in macrophage supernatant was analyzed by ELISA. ***p<0.001, assessed by t-test. (D–E) The mRNA (D) and protein level (E) of PD-L1 in SCC47 and SCC090 cells treated with recombinant TNF-α or vehicle for 48 hours were detected by RT-PCR and western blot, respectively. The quantification is based on data from three independent experiments. *p<0.05, **p<0.01. (F–G) Three-diemnsional (3D) invasion assay of SCC47 spheroids treated with recombinant TNF-α or vehicle. Scale bars represent 1000 µm. (G) Quantification of invasive area and the number of chains per spheroid from each group. Graphical values illustrate the mean±SD of n=10/group. ***p<0.001, assessed by t-test. (H–I) 3D invasion assay of M1 CM-stimulated SCC47 spheroids treated with a neutralizing antibody against TNF-α or a control isotype antibody. Scale bars represent 1000 µm. (I) Quantification of invasive area and the number of chains per spheroid from each group. Graphical values illustrate the mean±SD of n=10/group. ***p<0.001, assessed by t test. (J) Representative images of mixed spheroid 3D invasion assay of recombinant-TNF-α-treated and vehicle-treated SCC47 cells for 48 hours. A higher magnification image of the delineated region is shown in the zoomed panels. Arrowheads indicate leader cells in collective invasion strands. Scale bars represent 1000 µm. (K) The percentages of GFP-positive or mCherry-positive cells leading collective invasion strands were quantified. Graphical values illustrate the mean±SD of n=10/group. ***p<0.001, assessed by t-test. CM, conditioned media; PD-L1, programed death ligand 1.
We further tested whether M1 macrophages enhance the expression of PD-L1 in HPV-positive HNSCC by secreting TNF-α. Recombinant human TNF-α increased the protein levels of PD-L1 in HPV-positive HNSCC cells but not the mRNA levels (figure 3D,E). TNF-α treatment also induced PD-L1 in cancer cells at the leading edge (online supplemental figure 6A,B). Recombinant TNF-α also significantly promoted protrusive collective invasion of SCC47 and SCC090 cells, whereas neutralizing antibodies against TNF-α abrogated the invasive phenotype induced by CM from M1 macrophages (figure 3F–I and online supplemental figure 6C–F). Furthermore, TNF-α-treated cells were more frequently observed at the leading tip of the invasive strands in mixed spheroids of TNF-α- and vehicle-treated cancer cells (figure 3J,K and online supplemental figure 6G,H). Overall, these data suggested that TNF-α was the key cytokine involved in M1 macrophage-induced collective invasion in HPV-positive HNSCC.
CDK4 regulated TNF-α-mediated deubiquitination of PD-L1 through USP14
As M1 macrophages increased PD-L1 but not PD-L1, we investigated whether the increase in PD-L1 was mediated by a post-translational mechanism. CM from M1 macrophages and TNF-α treatment significantly increased the half-life of PD-L1 and decreased the abundance of ubiquitinated PD-L1 (online supplemental figure 7A–D). We further evaluated whether the cyclin D1-CDK4/6 pathway, a known proteasome-mediated PD-L1 degradation pathway,25 was involved in TNF-α-mediated stabilization of PD-L1. TNF-α did not alter the CCND1 or CDK6 levels, but it did decrease CDK4 in SCC47 and SCC090 cells (figure 4A,B and online supplemental figure 7E,F). Similarly, CM from M1 macrophages decreased CDK4 protein levels in SCC47 and SCC090 cells (figure 4C,D). Moreover, CDK4 knockdown significantly increased PD-L1 and promoted protrusive collective invasion of cancer cells (figure 4E–J and online supplemental figure 7G,H). Overall, these data indicate that CDK4 downregulation by TNF-α is a key event in TNF-α-mediated stabilization of PD-L1 in HPV-positive HNSCC.
TNF-α regulates CDK4 to promote PD-L1 expression and tumor cell invasion. (A, B) Western blot of cyclin D1, CDK4, and CDK6 in SCC47 and SCC090 cells treated with recombinant-TNF-α or vehicle. (B) The quantification is based on data from three independent experiments. **p<0.01. (C, D) Western blot of cyclin D1, CDK4, and CDK6 in SCC47 and SCC090 cells stimulated with M0, M1, and M2 macrophage CM or vehicle for 48 hours. (D) The quantification is based on data from three independent experiments. *p<0.05, **p<0.01. (E, F) Western blot of PD-L1 in HNSCC cells transfected with CDK4 siRNA or control siRNA for 48 hours. (F) The quantification is based on data from three independent experiments. *p<0.05, **p<0.01. (G–J) Three-dimensaional invasion assay of SCC47 and SCC090 spheroids previously transfected with CDK4 siRNA (siCDK4) or control siRNA (siCtrl) for 48 hours. Scale bars represent 500 µm. (H, J) Quantification of invasive area and the number of chains per spheroid from each group. Graphical values illustrate the mean±SD of n=10/group, **p<0.01, ***p<0.001, assessed by t-test. CM, conditioned media; HNSCC, head and neck squamous cell carcinoma; PD-L1, programed death ligand 1.
To identify the E3 ubiquitin ligases and deubiquitinating enzymes that regulate PD-L1 stability, PD-L1-binding proteins were analyzed by coimmunoprecipitation and liquid chromatography-tandem mass spectrometry analysis. A total of 837 PD-L1 binding proteins were identified, including three E3 ubiquitin ligases, SYVN1, STUB1, and NEDD4L, and two deubiquitinases, USP14 and OTUB1 (online supplemental table S3). CDK4 knockdown increased the expression of USP14, but the expression of the other four candidate enzymes remained unchanged (figure 5A).
USP14 deubiquitinates PD-L1. (A) RT-PCR of SYVN1, STUB1, NEDD4L, USP14, and OTUB1 in SCC47 and SCC090 cells 24 hours after transfection of CDK4 siRNA (siCDK4) or control siRNA (siCtrl). ***p<0.001, assessed by t test. (B) SCC47 cell lysates were subjected to immunoprecipitation with control IgG, anti-PD-L1, or anti-USP14 antibodies. The immunoprecipitates were then used for immunoblotting with the indicated antibodies. (C) Western blot of USP14 in SCC47 and SCC090 cells 48 hours after transfection of CDK4 siRNA (siCDK4) or control siRNA (siCtrl). The quantification is based on data from three independent experiments. *p<0.05 and **p<0.01. (D) Western blot of PD-L1 in SCC47 and SCC090 cells 48 hours after transfection of USP14 siRNA (siUSP14) or control siRNA (siCtrl). The quantification is based on data from three independent experiments. *p<0.05, **p<0.01. (E, F) Analysis of ubiquitination of PD-L1 in SCC47 and SCC090 cells after CDK4 and USP14 knockdown. The quantification is based on data from three independent experiments. *p<0.05, **p<0.01. (G, H) Three-dimensional (3D) invasion assay of SCC47 spheroids 48 hours after transfection with USP14 siRNA (siUSP14) or control siRNA (siCtrl). Scale bars represent 500 µm. (H) Quantification of invasive area and the number of chains per spheroid from each group. (I, J) 3D invasion assay of CDK4-knockdown SCC47 spheroids with or without knockdown of USP14 siRNA (siUSP14). Scale bars represent 1000 µm. (J) Quantification of invasive area and the number of chains per spheroid from each group. (H, J) Graphical values illustrate the mean±SD of n=10/group, **p<0.01, ***p<0.001, assessed by t-test. NEDD4L, NEDD4 Like E3 Ubiquitin Protein Ligase; OTUB1, Otubain-1; PD-L1: programed death ligand 1; P14, ubiquitin-specific peptidase 14; SYVN1, synovial apoptosis inhibitor 1; STUB1, STIP1 homology and U-box containing protein 1; USP14: ubiquitin-specific peptidase 14.
We then examined the effect of USP14 on PD-L1. Endogenous USP14 bound to PD-L1, and USP14 knockdown reduced the expression of PD-L1 and restored its ubiquitination after CDK4 depletion, suggesting that USP14 deubiquitinates PD-L1 (figure 5B–F). USP14 knockdown in SCC47 and SCC090 cells inhibited collective invasion and abrogated the invasive phenotype induced by CDK4 depletion (figure 5G–J and online supplemental figure 7I–L). These data indicate that TNF-α downregulates CDK4 to promote PD-L1 deubiquitination via USP14, thus enhancing the collective invasion of HPV-positive HNSCC cells.
CDK4/6 inhibitors enhanced the therapeutic effects of anti-PD-L1 therapy in an E6/E7 murine tumor model
As CDK4/6 knockdown increased the expression of PD-L1 in tumor cells and high PD-L1 is a good predictor of the response to anti-PD-1/PD-L1 therapy,23 the potential synergistic effect of combination therapy using CDK4/6 inhibitors and anti-PD-L1 antibodies for HPV-positive HNSCC was investigated. An E6/E7 murine tumor model was established by 4NQO administration, and the tumor-bearing mice were subsequently treated CDK4/6 inhibitor (palbociclib), PD-L1 mAb, palbociclib plus PD-L1 mAb, or vehicle (figure 6A). Consistent with the findings of a previous study,26 HPV-positive HNSCC was non-responsive to palbociclib monotherapy, as there were no differences in tumor size or number between the palbociclib and vehicle groups. Similar to our in vitro observations, treatment with palbociclib increased the expression of PD-L1 in tumor tissues (online supplemental figure 8). Combining palbociclib with PD-L1 mAb therapy resulted in a significant decrease in tumor number, size and tumor grade compared with either treatment alone (figure 6B–E). In addition, we further evaluate the therapeutic efficacy of the combined therapy using an immunocompetent mouse SCC7-E6E7 xenotransplantation model. We found that combined therapy prolonged survival of the tumor-bearing hosts compared with either treatment alone (figure 6H).
Palbociclib enhances the therapeutic effects of PD-L1 blockade in an E6/E7 murine tumor model. (A) Protocol diagram of the 4NQO-induced HNSCC mouse model and drug treatments. Rosa26-E6-E7 constitutive knock-in C57BL/6 mice were subjected to oral administration of 4NQO for 10 weeks and were then switched to distilled water for another 10 weeks. After 20 weeks, mice were randomly administered control (vehicle), palbociclib, anti-PD-L1 antibodies, or combination treatment for 4 weeks. At the end of the experiment, the mice were sacrificed, and the tongues were harvested. (B) Images of the tongue lesions from mice in each treatment group (n=6/group). Higher magnification of the delineated region is shown in the zoomed panels. (C, D) Tumor number and size per mouse in each group. (E) The H&E-stained sections of mice tongue tissues were scored by a pathologist without knowledge of the experimental group, and the tissue was determined to be normal epithelium, mild-moderate dysplasia, severe dysplasia, or carcinoma. (F) Representative H&E staining and panCK immunostaining of mouse lesions. Scale bars represent 50 µm. (G) Quantification of the density of collective invasion packs (CIPs). Graphical values illustrate the mean±SD, *p<0.05, ***p<0.001, assessed by one-way ANOVA. (H) Survival curves of mice bearing E6/E7-expressing SCC7 tumor after administration of vehicle, palbociclib, anti-PD-L1 antibodies, or combination treatment (n=6). ***p<0.001, assessed by log-rank test. (I) Schematic depicting how M1 macrophages regulate PD-L1 expression in HPV-positive HNSCC to drive tumor cell collective invasion. 4NQO, 4-Nitroquinoline N-oxide; ANOVA, analysis of variance; HNSCC, head and neck squamous cell carcinoma; panCK: pancytokeratin; PD-L1, programed death ligand 1.
We further investigated the effects of combination therapy on collective invasion by assessing the density of CIPs in the mouse tumors. The addition of anti-PD-L1 antibody to palbociclib significantly reduced the density of CIPs in each invasive field (figure 6F–G). Moreover, palbociclib treatment decreased the number of CD8+ tumor-infiltrating lymphocytes, whereas the combination of anti-PD-L1 antibody and palbociclib restored tumor-infiltrating lymphocytes (online supplemental figure 8). These data indicate that although palbociclib monotherapy is not an effective therapy for HPV-positive HNSCC, the combination of palbociclib and anti-PD-L1 therapy enhances the therapeutic efficacy in HPV-positive HNSCC.
Discussion
PD-L1 facilitates tumor development by constructing a permissive immune microenvironment and potentially modulating EMT and a cancer stem cell-like phenotype.27 28 Our study revealed an additional dimension: the mechanism underlying the architectural pattern of PD-L1 expression in clustered tumor cells of HPV-positive HNSCC. M1 macrophages induced PD-L1hi leader cells via a CDK4/USP14-mediated deubiquitination mechanism. Of note, palbociclib, a CDK4/6 inhibitor, enhanced the efficacy of anti-PD-L1 therapy in an E6/E7 murine tumor model, which offers a potential therapeutic approach for HPV-positive HNSCC.
Leader cells at the tumor invasive front exhibit distinct phenotypes, including invasive morphology and higher invasion abilities compared with follower cells.11 Although the marginal PD-L1 expression in HPV-positive HNSCC have been reported in previous studies, the biological significances and underlying mechanisms of this phenomenon are first uncovered in our study. Using a specific cell-led 3D invasion assay, we showed that PD-L1hi cells could polarize to the leading edge of organoids and enhance the collective invasion of PD-L1low tumor cells. These involved with the phenotypes of partial EMT, but not complete EMT, which is consistent with previous descriptions of the activation of Ras/Erk/EMT signaling on PD-L1 overexpression in glioblastoma multiforme.7 Additionally, leader cells and collectively invading tumor cells typically retain E-cadherin and exhibit a partial EMT phenotype.13 29 30
PD-L1 expression in tumor cells is regulated by two major mechanisms, including tumor cell-intrinsic and cell-extrinsic mechanisms.31 We provide direct evidence supporting M1 macrophages as the critical cells for marginal expression of PD-L1 in HPV-positive HNSCC. Specifically, M1 macrophages induced specific expression of PD-L1 in leader cells via TNF-α secretion, whereas depletion of macrophages by clodronate liposomes diminished the marginal expression of PD-L1. These findings demonstrated a novel mechanism for the architectural pattern of PD-L1 expression in light of the critical roles of inflammatory cells in inducing specific molecular expression in tumor cells via paracrine mechanisms. Consistently，abundant M1 macrophage infiltration was observed at the tumor–stroma interface, adjacent to PD-L1-positive tumor cells. In addition, peripheral macrophage sheaths have been reported in HPV-positive HNSCC by Lyford-Pike et al,16 which can be attributed to why marginal PD-L1 expression was exclusively observed in HPV-positive HNSCC. There are several post-translational mechanisms regulating PD-L1 expression in tumor cells, including protein glycosylation, phosphorylation, and ubiquitination.32–34 In this study, we showed that CDK4 regulates M1 macrophage-induced expression of PD-L1 in tumor cells via USP14-mediated deubiquitination. Due to the low response rate to PD-1/PD-L1 blockade, identifying novel targets and combination strategies has become a new focus. CDK4/6 inhibitors have been approved for the treatment of different types of cancers.35 36 CDK4/6 inhibitors showed a good response in HPV-negative HNSCC but a poor response in HPV-positive HNSCC in clinical trials.26 37 Our data indicate that although palbociclib monotherapy is not an effective therapy for HPV-positive HNSCC, the combination of palbociclib and anti-PD-L1 therapy has high therapeutic efficacy. Previous studies reported that CDK4/6 inhibitor showed limited tumor-suppressive activities on HPV-positive carcinomas because of the absence of pRB, the key target of CDK4/6 inhibitor in these carcinomas.38 39 We hypothesis that CDK4/6 inhibitors enhanced the response of HPV-positive murine tumors to anti-PD-L1 therapy by upregulating PD-L1 expression in tumor cells via USP14-mediated deubiquitination.
Several limitations in the current study should be acknowledged. First, our work provided new insights for understanding the mechanisms underlying M1 macrophages-induced PD-L1 expression in leader cells, whereas how PD-L1hi cell lead collective invasion in HPV-positive HNSCC needs further investigation. Second, clodronate liposome treatment is not specific for M1 macrophages but depletes all macrophage populations, the influence of other macrophage populations was unable to determine.40 Third, although 4NQO-induced HNSCC model is a well-validated model for investigating the mechanisms of head and neck carcinogenesis and identifying novel therapies for HNSCC, some human HNSCC cases are not related to carcinogen exposure.
In conclusion, we demonstrated that PD-L1hi leader cells promoted collective invasion of tumor cells via partial EMT. We also showed that M1 macrophages induce specific expression of PD-L1 in cancer leader cells via the CDK4/USP14-mediated deubiquitination of PD-L1 protein (figure 6I). This work provides new insight into the roles and mechanisms of the spatial heterogeneity of PD-L1 expression in tumor cells, which represents a conceivable tumor-targeting strategy.
Data availability statement
All data relevant to the study are included in the article or uploaded as online supplemental information.
Ethics statements
Patient consent for publication
Ethics approval
This study involves human participants and clinical samples and data were obtained with written consent from each patient and used in accordance with protocols approved by the Institutional Ethics Committee of West China Hospital of Stomatology of Sichuan University (no. WCHSIRB-CT-2021-035). Participants gave informed consent to participate in the study before taking part. All animal studies were conducted according to the protocol approved by the Subcommittee on Research and Animal Care of Sichuan University (WCHSIRB-D-2021–-055, Chengdu, China) and under the guidelines of the Institutional Animal Care.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Contributors YT and XL conceived the project. JW, XP, XYang, MZ and XYu performed the in vitro experiments. JW, BC, HF and HW performed the mouse experiments. JW, XP and XYu performed HandE staining and immunostaining experiments of the patient tissues. JW, XP and XYang performed the data analysis. JW, YT and XL wrote the paper with input from all authors. XL is responsible for the overall content as guarantor.
Funding This work was financially supported by the National Natural Science Foundation of China (grant nos. 81972542, 82073000 and 82173326), Science and Technology Foundation of Sichuan Province (grant no. 2022YFS0289) and Clinical Research Project of West China Hospital of Stomatology, Sichuan University (grant no. LCYJ2019-8).
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.