2.1 Patient information and tissue specimens

HNSCC diagnosis was based on histopathological features, including depth of invasion, pathological stage, and lymph node metastasis. Tumor tissues and adjacent normal tissues from 69 patients with HNSCC were collected and frozen immediately in liquid nitrogen within 30 min after surgery and stored at -80°C between June 2020 and January 2021. Analyses were performed on these samples using gas chromatography-mass spectrometry (GC-MS), liquid chromatography-mass spectrometry (LC-MS), and RNA-sequencing (RNA-seq). A tissue microarray (TMA) was prepared using 70 HNSCC and 18 adjacent normal tissues from patients who had undergone surgery between January 2007 and December 2008 and was used for immunohistochemical analysis. All enrolled primary HNSCC patients met our inclusion criteria: (1) treatment with surgical resection; (2) pathologically confirmed by two independent experts; and (3) no local or systemic treatment prior to the surgery. Exclusion criteria were as follows: (1) patients with history of previous malignancies; (2) patients refused to utilize their surgical specimens for research purposes. The clinical characteristics of HNSCC patients for metabolomics and RNA-seq analyses are listed in Supplementary Table S1, and those for TMA immunohistochemical analysis are listed in Supplementary Table S2. All patients had signed informed consent prior to surgery. The use of human specimens in the present study was approved by the Institutional Research Ethics Committee of Shanghai Ninth People's Hospital, Shanghai, China (approval ID: SH9H-2021-T135).

2.2 Cell lines and cell cultures

Human tongue squamous cell carcinoma cell lines (SCC9, SCC25, and Cal27) were purchased from the Type Culture Collection of the Chinese Academy of Sciences (Shanghai, China). Human tongue squamous cell carcinoma cell lines (HN4 and HN6) and human pharyngeal squamous cell carcinoma cell line HN30 were obtained from the University of Maryland Dental School (Baltimore, MD, USA). Mouse squamous cell carcinoma cell line SCCVII was donated by Prof. Zhuang Liu (Soochow University, Suzhou, Jiangsu, China). The human embryonic kidney cell line 293T and human T-cell leukemia cell line Jurkat were purchased from the American Type Culture Collection (Manassas, VA, USA). All cultures were tested routinely using MycoAlert^TM PLUS Mycoplasma Detection Kit (Lonza, Basel, Switzerland) and found to be negative for mycoplasma. All cancer cells were cultured in high-glucose Dulbecco's modified Eagle's medium (DMEM, Gibco, Grand Island, NY, USA) or DMEM/F-12 (Gibco) comprised of 0.584 g/L glutamine and supplemented with 10% fetal bovine serum (FBS, Gibco). Jurkat and SCCVII cells were cultured in Roswell Park Memorial Institute (RPMI)-1640 (Gibco), supplemented with 10% FBS and 1% penicillin/streptomycin (Gibco). All cell lines were maintained at 37°C in 5% CO₂. Short tandem repeat (STR) profiling was used to test the authenticity of the cell lines. Aryl hydrocarbon receptor (AhR) inhibitors, BAY-218 (S8842) and CH-223191 (S7711), were purchased from SelleckChem (Houston, TX, USA). System L amino acid transporter (Kyn transporter) inhibitor 2-aminobicyclo-(2,2,1)-heptane-2-carboxylic acid (BCH, S6894) and polyhydroxyalkanoate (PHA, HY-N7038) were purchased from MedChemExpress (Monmouth Junction, NJ, USA). Dimethyl sulfoxide (DMSO, 0.1%, 67-68-5, Sigma, St. Louis, MO, USA) was used as solvent control.

2.3 Mice

All animal experiments were approved by the Animal Care and Use Committee of Ninth People's Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China (approval ID: SH9H-2022-A031-SB). Female 4-6 weeks old C3H/He mice and BALB/c nude mice were purchased from Shanghai Bikai Keyi Biotechnology Co., Ltd (Shanghai, China). All mice were maintained under pathogen-free conditions in the animal care facilities of the Ninth People's Hospital, Shanghai Jiao Tong University School of Medicine. All animals were maintained at room temperature (22-25°C) and with free access to food and water with a 12-h light/dark cycle. All mice used in experiments throughout the study exhibited normal health. All mice were used after 1 week of acclimatization to the facility. All animal studies were performed in accordance with the Guide for Care and Use of Laboratory Animals (The Ministry of Science and Technology of China, 2006).

2.4 LC-MS analysis

The metabolic profiles of HNSCC and paired adjacent normal tissues were analyzed using LC (Dionex U3000 UHPLC, Thermo Fisher, Waltham, MA, USA)-tandem MS (Q-Exactive, Thermo Fisher). Briefly, prepared samples were loaded separately on an Acquity UPLC BEH C18 column (1.7 µm, 2.1 mm × 100 mm, Waters, Milford, MA, USA). Each sample was eluted using a gradient elution system consisting of mobile phase A (99.9% H₂O, 0.1% formic acid, v/v) and mobile phase B (99.9% acetonitrile, 0.1% formic acid, v/v). Elution gradient solutions were added as follows: from 100% to 95% mobile phase A in 2 min; from 95% to 70% mobile phase A in 2 min; from 70% to 50% mobile phase A in 4 min; from 50% to 20% mobile phase A in 2 min; from 20% to 0% mobile phase A in 4 min; holding at 0% mobile phase A for 1 min; from 0% to 95% mobile phase A in 5 s; holding at 95% mobile phase A for 55 s. The flow rate was 0.35 mL/min, column temperature was 45°C, and injection volume was 2 µL. Sample signals were collected continuously in the mass spectrometer at a capillary temperature of 320°C and a spray voltage of 3,800 V. Principal component analysis (PCA) and orthogonal partial least-squares discriminant analysis (OPLS-DA) were carried out to visualize the metabolic alterations among experimental groups, after mean centering and Pareto variance scaling, respectively. The Hotelling's T2 region, shown as an ellipse in score plots of the models, defines the 95% confidence interval of the modeled variation. Seven-fold cross-validation was applied with 1/7 of the samples being excluded from the mathematical model in each round, in order to guard against overfitting. In addition, 200 response permutation testing (RPT) was performed to evaluate the risk of model overfitting. Linear regression was conducted with R2Y and Q2Y of the original model, and the obtained intercept values between the regression line and Y-axis were R2 and Q2, respectively, evaluating whether the OPLS-DA model was overfitting. Variable importance in the projection (VIP) ranks the overall contribution of each variable to the OPLS-DA model, and those variables with VIP > 1 are considered relevant for group discrimination. Metabolics was annotated and analyzed using the Kyoto Encyclopedia of Gene and Genome (KEGG) database (https://www.genome.jp/kegg). Area under the curve (AUC) was calculated by summing the area under the receiver operating characteristic (ROC) curve, which shows the classifier's performance at varying decision thresholds.

2.5 GC-MS analysis

All the prepared samples were analyzed using an Agilent gas chromatograph (7890B, Agilent Technologies, Santa Clara, CA, USA) and an Agilent mass spectrometer (5977A, Agilent Technologies). A DB-5MS (30 m × 0.25 mm × 0.25 µm, Agilent Technologies) fused-silica capillary column (Agilent Technologies) was utilized for the GC-MS analysis. The column operating conditions were as follows: helium (> 99.999%) was used as carrier gas with a flow rate of 1.0 mL/min, inlet volume of 1 µL, and split-less inlet temperature of 260°C. The initial oven temperature was 60°C and held for 0.5 min; increased to 125°C at a rate of 8°C/min, 210°C at a rate of 5°C/min, 270°C at a rate of 10°C/min, and 305°C at a rate of 20°C/min; and finally held at 305°C for 5 min. The temperature of the MS quadrupole was set to 150°C, and electron impact was set to 230°C. The collision energy was 70 eV, and data were acquired in full-scan mode (m/z: 50-500). Data processing was the same in LC-MS.

2.6 RNA-seq and data analysis

Total RNA extraction was performed using a TRIzol reagent according to the manufacturer's protocol (Takara, Dalian, Liaoning, China). mRNA library construction was performed on an Illumina HiSeq X Ten platform (OE Biotech Co., Ltd., Shanghai, China). Raw data (raw reads) in fastq format for each sample were generated and processed using Trimmomatic [26]. Clean reads for each sample were retained by removing low-quality reads and mapped to the human genome (GRCh38) using HISAT2 for subsequent analyses [27]. The patients were separated into two groups based on the median values of the Kyn level (Kyn-high or Kyn-low) or CD274 expression (CD274-high or CD274-low). Gene set enrichment analysis (GSEA) was performed with a gene list ordered based on the log₂ (Ratio) value, the absolute value of the log₂ (Ratio), and the network-based score, respectively, using the GSEA-preranked function, from which we computed enrichment scores, normalized enrichment scores, P values, and false discovery rates for each gene set based on a modified K–S test. Gene Expression Profiling Interactive Analysis (GEPIA, http://gepia.pku.cn/) is a newly developed interactive web server for analyzing the bulk gene expression datasets in The Cancer Genome Atlas (TCGA) and the Genotype-Tissue Expression (GTEx) projects. In the present study, SIGLEC15 expression and survival analysis were performed using the GEPIA HNSCC dataset.

2.7 LC-MS targeted metabolomics analysis

Samples were ground in 100% methanol for 10 min and centrifuged (16,000 ×g) at 4°C for 10 min. The supernatant was filtered through a 0.22-µm membrane filter and analyzed using a liquid chromatograph (UltiMate 3000 RS, Thermo Fisher). The samples were loaded separately on a Thermo Hypersil GOLD HILIC column (1.9 µm, 2.1 mm × 100 mm, Thermo Fisher). Samples were eluted with an elution including phase water (containing 0.1% formic acid, v/v) and phase acetonitrile (containing 0.1% formic acid, v/v). Elution gradient solutions were added as follows: from 5% to 95% water in 2.5 min, holding at 95% water for 1 min, from 95% to 5% water in 18 s, holding at 5% water for 1 min. The flow rate was 0.40 mL/min, column temperature was 40°C, and inlet volume was 2 µL. Tryptophan (Trp), Kyn, kynurenic acid, indole-3-carboxaldehyde and serotonin standards (Sigma) were used to construct a standard curves for quantifying the concentrations. Briefly, all standard samples were dissolved in 0.1% formic acid and methanol for a stock solution of 2 mg/mL, then a series of standard curve solutions were diluted with pure methanol in concentration gradients of 20,000, 5,000, 2,000, 500, 200, 50, 20, and 5 ng/mL in turn. The compounds’ chromatogram collection and integration were processed using Xcalibur 3.0 software (Thermo Fisher). Linear regression was performed with a weighting coefficient of 1/x² to calculate the content of metabolites in the samples.

2.8 Cell counting kit-8 (CCK8), colony formation assay, 5-ethynyl-2’-deoxyuridine (EdU) assay

Cells were seeded at an initial density of 1,000 cells/well in 96-well plates and treated with 0.1% DMSO, 50 or 100 µmol/L of Kyn for 5 days to detect cell proliferation using a CCK8 kit (Dojindo, Kumamoto, Japan).

Cells were seeded in 6-well plates at 1,000 cells/well and treated with 0.1% DMSO, 50 or 100 µmol/L of Kyn for 10 days. Subsequently, cells were prefixed with 4% paraformaldehyde (P0099, Beyotime, Shanghai, China) for 10 min at room temperature and stained by crystal violet staining solution (C0121, Beyotime) for 15 min to analyze colony formation.

Cells were seeded in 6-well plates at 5,000 cells/well and treated with 0.1% DMSO, 50 or 100 µmol/L of Kyn for 24 h. Subsequently, cells were stained using a Cell-Light EdU Apollo567 In Vitro Kit (C10371-1, RiboBio, Guangzhou, Guangdong, China) according to the manufacturer's protocol. Cells were viewed under a Zeiss inverted light microscope (Axio vert A1, Zeiss, Oberkochen, Germany), and the percentage of EdU-positive cells was analyzed using ImageJ software (NIH, Bethesda, ML, USA).

2.9 Cell migration and invasion assay

Wound healing and transwell assays using uncoated polycarbonate inserts (Millipore, Billerica, MA, USA) were performed to determine cell migration. Transwell invasion assays were performed using Bio-Coat™ inserts (BD Biosciences, San Jose, CA, USA). Cells were pretreated with 1 µg/mL mitomycin C (Sigma) for 1 h to exclude the effect of cell proliferation. Medium without FBS (2 × 10⁴ cells/200 µL) were added into the upper portion of a migration (uncoated insert) or invasion (matrigel-coated insert) chamber, with 500 µL DMEM containing 10% FBS added into the lower chamber. After staining with crystal violet, the number of colonies was counted and analyzed statistically. The experiment was performed in triplicate.

2.10 Western blotting analysis

Western blotting was performed as described in our previous study [28]. The cell protein was extracted in sodium dodecyl sulfate (SDS) lysis buffer (Beyotime) with protease inhibitor cocktail (4693132001, Roche, Basel, Switzerland). The protein samples were electrophoresed on 4%-20% polyacrylamide gels (Genscript, Nanjing, Jiangsu, China) and transferred to polyvinylidene fluoride (PVDF) membranes (Millipore). Membranes were then incubated overnight at 4°C with primary antibody. The antibodies used in this study included anti-E-cadherin Ab (1:1,000 dilution, 3195, Cell Signaling Technology, Danvers, MA, USA), anti-Vimentin Ab (1:1,000 dilution, 5741, Cell Signaling Technology), anti-Snail Ab (1:1,000 dilution, 3879, Cell Signaling Technology), anti-programmed death-ligand 1 (PD-L1) Ab (1:1,000 dilution, 85164, Cell Signaling Technology), anti-Siglec-15 Ab (1:1,000 dilution, 37925, Signalway Antibody, Greenbelt, ML, USA), primary rabbit anti-human AhR Ab (1:1,000 dilution, 83200, Cell Signaling Technology), and anti-programmed cell death protein 1 (PD-1) Ab (1:5,000 dilution, 66220-1-Ig, Proteintech, Rosemont, IL, USA). An anti-β-actin antibody (1:2,000 dilution, 8457, Cell Signaling Technology) was used as an internal control. After washing with phosphate buffered saline (PBS), the membrane was probed with horseradish peroxidase (HRP)-conjugated secondary antibody (Cell Signaling Technology) at 1:2,000 dilution for 1 h, and the signals were detected with enhanced chemiluminescence (ECL) detection reagents (Millipore). Immunoreactive bands were analyzed using ImageJ software.

2.11 RNA interference and lentiviral transfection

Lipofectamine 2000 (11668019, Thermo Fisher) was used for transfection with siRNAs (siAhR #1-3 for homo sapiens and siSIGLEC15 [siS15] #1-3 for mouse, RiboBio), at a final concentration of 100 nmol/L. Mouse SIGLEC15 cDNA was cloned into the PGMLV-CMV-MCS-EF1-ZsGreen1-T2A-Puro vector (Genomeditech, Shanghai, China). Polybrene was used for the lentiviral transfection. Puromycin (1 µg/mL) was used for selection for 7 days to establish a stable SCCVII^Siglec-15 cell line. Vector lentivirus was used to establish SCCVII^Vector cell line. The sequences are listed in Supplementary Table S3.

2.12 Preparation and characterization of NH₂-modified mesoporous silica (NH₂-MSN) nanoparticles and NH₂-MSN-siRNA (NP-siRNA)

NH₂-MSNs were synthesized according to previous reports [29, 30]. Briefly, 1.5 g of cetyltrimethyl ammonium bromide (CTAB, H5882, Sigma) was dissolved and stirred, and different amounts of diethanolamine (D8885, Sigma) were added and stirred. Subsequently, 12 mL of tetraethyl orthosilicate (131903, Sigma) was added dropwise to the mixture under intense stirring. The products were dried, collected, and refluxed in methanol/HCl. The steps were repeated to remove CTAB and obtain MSNs. Subsequently, the MSNs were redispersed and stirred with 3-aminopropyltriethoxysilane (A3648, Sigma). After refluxing, the obtained products were cleaned and dried successively, as described above. The amino-functionalized samples were named NH₂-MSNs. The positively charged NH₂-MSNs (10 µg/mL) were mixed with the siRNA (siS15 or siScramble [siScr]) solution in PBS at pH 6.0 and shaken on a bench-top shaker for 1 h to complete the binding between siRNA and NH₂-MSNs via electrostatic interactions. The prepared samples were termed NP-siRNA (NP-siS15 or NP-siScr) herein. The morphologies of NH₂-MSNs and NP-siS15 were examined under transmission electron microscopy (JEM-1400, JEOL, Tokyo, Japan). Zeta potential measurements of NH₂-MSNs and NP-siS15 were performed using the Zetasizer Nanoseries (Nano ZS90, Malvern Instrument Ltd., Worcestershire, UK).

2.13 in vivo experiments

Orthotopic transplants were performed as described [31]. Required number of SCCVII cells were collected by centrifuging (300 ×g) at 4°C for 3 min, and supernatant was discarded before washed twice by pre-cold PBS. Subsequently, cell pellets were resuspended in Matrigel (356234, Corning, NY, USA) at concentration of 2,000 cells/µL and mixed gently on ice. C3H/He mice were anesthetized, and the tongues were gently grasped and pulled out from the mouth using forceps. Using a 1-mL syringe attached to a 29 gauge needle, 25 µL cells (5 × 10⁴) were injected slowly into the right lateral side of each tongue (∼1.5 mm from the tip of the tongue) so that a bulbous mass formed in the center of the tongue. Small white tumors were visible 10 days after injection. The mice received intraperitoneal injections of either PBS or Kyn (50 mg/kg) every 3 days for 3 cycles. The mice were euthanized at 20 days and analyzed. For tongue samples taken thereafter, the entire tongue was removed by cutting the tongue at the back of the mouth and bisecting it down the middle from tip to back. Generally, the entire tongue was used for immunohistology analysis. For the Siglec-15 overexpression experiment, SCCVII^Vector or SCCVII^Siglec-15 cells were injected into the tongue as orthotopic transplants.

Female BALB/c nude mice were used to establish lung metastasis model. A total of 5 × 10⁵ HN30 cells in 100 µL PBS were injected intravenously into the lateral tail vein. After 10 days of tumor cell injection, 4 mg/100 µL Kyn or 100 µL PBS was administered by gavage twice a week for 4 weeks. Mice were euthanized 5 weeks after treatment. Whole lung tissues were removed and fixed in Bouin's fixative (HT10132, Sigma) diluted 1:5 with neutral-buffered formalin for metastasis nodule observation.

Subcutaneous tumorigenesis models were established using C3H/He mice. A total of 1 × 10⁵ SCCVII^Vector or SCCVII ^Siglec-15 cells in 100 µL PBS were inoculated subcutaneously into both flanks, and small tumors were visible 10 days after injection. The length (L) and width (W) of the tumors were measured every 2 days from the 10th day after injection, and the volume was calculated as (L × W² × π/6). On day 20, the mice were euthanized, and the tumors were extracted carefully for flow cytometry and immunohistochemical analyses. For the Kyn stimulation experiment, SCCVII cells (1 × 10⁵) were inoculated subcutaneously into C3H/He mice as before. After 10 days of tumor transplantation, mice received oral gavage of PBS, Kyn (100 mg/kg), and/or CH-223191 (10 mg/kg) once every 2 days. The mice were euthanized at day 20 and analyzed as previously described. The vital organs were fixed, embedded in paraffin, sectioned, and stained with hematoxylin and eosin (H&E), and blood samples were obtained for hepatic and renal function testing. For anti-PD-1 and anti-Siglec-15 experiments, SCCVII subcutaneous tumorigenesis models were established as before and separated randomly into 5 groups. After 10 days of tumor transplantation, the anti-PD-1 Ab (10 mg/kg, BE0146, BioXCell, West Lebanon, NH, USA) or isotype IgG (10 mg/kg, BE0089, BioXCell) were administered by intraperitoneal injection, while NP-siScr (0.5 nmol siScr:4 µL NH₂-MSNs) and NP-siS15 (0.5 nmol siS15:4 µL NH₂-MSNs) were administered by intravenous injection. Tumor volume was measured every 2 days, and mice were euthanized at the end of the experiment. Overall survival probability curves were used to evaluate treatment efficacy. For anti-PD-L1 and anti-Siglec-15 experiments, SCCVII subcutaneous tumorigenesis models were established as before and separated randomly into 4 groups. After 10 days of tumor transplantation, the anti-PD-L1 (5 mg/kg, 2115, Selleck) or isotype IgG (5 mg/kg, A2116, Selleck) were administered by intraperitoneal injection, whereas NP-siS15 (0.5 nmol siS15: 4 µL NH₂-MSNs) were administered by intravenous injection. Tumor volume measurement and analysis were performed as previously described.

Tribromethyl alcohol (HY-B1372, MedChemExpress) was used to anesthetize mice. Mice were euthanized by cervical dislocation after anesthesia in the following conditions: near death or immobility, significant weight loss, inability to feed or drink, at the end of experiment, or tumor volume > 2,000 mm³.

2.14 Immunofluorescence

Cells with the indicated treatments were fixed for 20 min with 4% paraformaldehyde and blocked for 1 h with 10% goat serum at 37°C. Subsequently, they were incubated with primary anti-AhR antibody (1:200, 67785-1-Ig, Proteintech, USA) overnight at 4°C. The cells were stained with Alexa Fluor 647-conjugated secondary antibody (4410, Cell Signaling Technology), and nuclei were counterstained with 4’,6-diamidino-2-phenylindole (DAPI) (D21490, Thermo Fisher). SCCVII cells were transfected with NH₂-MSNs that contained siS15 (RiboBio) labeled with Cyanine (Cy) 5 (NP-siS15-Cy5). The cytoskeleton and nuclei were counterstained using fluorescein isothiocyanate (FITC)-phalloidin (40735ES75, Yeasen, Shanghai, China) and DAPI. The cellular uptake of NP-siS15-Cy5 was observed using Cytation 5 (Biotek, Winooski, VT, USA).

2.15 Chromatin immunoprecipitation (ChIP) assay

The SimpleChIP^® Enzymatic Chromatin IP kit (9003, Cell Signaling Technology) was used for the ChIP assay, as previously described [21]. Briefly, after the indicated treatment, Cal27 and HN30 cells were cross-linked by formaldehyde (Millipore) at 1% final concentration. Chromatin was digested with micrococcal nuclease and sonicated into 150-900 bp fragments. An aliquot of the cross-linked protein complexes was immunoprecipitated by incubation with either primary rabbit anti-human AhR antibody or IgG antibody (2729, Cell Signaling Technology) overnight at 4°C with rotation. Chromatin-antibody complexes were isolated from solution by incubation with ChIP-Grade Protein G Magnetic Beads (9003, Cell Signaling Technology) for 1 h at 4°C with rotation. The bead-bound immune complexes were then washed and eluted from the beads with elution buffer. Eluates were heated at 65°C overnight to reverse the formaldehyde cross-linking, and DNA was extracted. DNA samples from chromatin immunoprecipitation preparations were analyzed by real-time quantitative polymerase chain reaction (RT-qPCR). The promoter primers for SIGLEC15 and CD274 were synthesized by Sangon Biotech (Shanghai, China) and are listed in Supplementary Table S4.

2.16 Fluorescent multiplex immunohistochemistry staining and analysis

For multiple-immunofluorescence staining, paraffin-embedded tissues were sequentially stained with indicated primary antibodies overnight at 4°C and HRP-conjugated secondary antibodies for 1 h at room temperature, followed by microwave treatment (700 w for 5 min and then 100 w for 30 min) and another round of staining with three different primary antibodies. Anti-CD3 rabbit antibody (1:200 dilution, ab16669, Abcam, Cambridge, MA, USA) and anti-CD8 rabbit antibody (1:200 dilution, ab209775, Abcam) for SCCVII tumors, anti-CD8 (D8A8Y) rabbit antibody (1:200 dilution, 85336, Cell Signaling Technology) for HNSCC tissues, and anti-PD-1 rabbit antibody (1:200 dilution, 66220-1-Ig, Proteintech) were used as primary antibodies. Cy3, FITC, Cy5, and DAPI dyes were used for staining. The slides were mounted with the nuclear dye DAPI. Images of the samples were captured and collected using a slice scanner (Pannoramic MIDI, 3DHISTECH Ltd., Budapest, Hungary). Analyses of the different immunofluorescence images differed in each experiment and are indicated in the figure legends.

2.17 RT-qPCR assay

The total RNA of related cells was isolated using a TRIzol reagent (15596018, Invitrogen, Carlsbad, CA, USA). Total RNA (500 ng) was reversely transcribed into first-strand cDNA using PrimeScript™ RT Master Mix (RR036A, TaKaRa, Tokyo, Japan). RT-qPCR was conducted to quantify mRNA expression using a StepOnePlus Real-time PCR system (Applied Biosystems, Foster City, CA, USA). The reaction conditions were denaturation at 95°C for 30 s, followed by 40 cycles of 5 s of denaturation at 95°C, and annealing and extension for 30 s at 60°C. The primer sequences used in the present study are listed in Supplementary Table S5. Relative quantification of mRNA levels compared to the internal control gene was performed using the 2-ΔΔct method. All samples were assayed in triplicate.

2.18 Luciferase reporter assay

293T cells were transfected with luciferase reporter (firefly and renilla luciferase) plasmids containing the promoter of SIGLEC15 or CD274. Transfected 293T cells were then treated with Kyn at a concentration or time gradient. To examine the transcription of AhR for SIGLEC15 or CD274, 293T cells were co-transfected with siAhR or siScr and luciferase reporter (firefly and renilla luciferase) plasmids, including the SIGLEC15 or CD274 promoter. After 48 h of incubation, a dual-luciferase reporter assay kit (RG027, Beyotime) was used to detect luciferase activity as previously described [32].

2.19 TIL analysis

Tumors established using the indicated treatments were removed from C3H/He mice, dissociated, and digested, as previously described [32]. The digested tissues were washed twice with PBS and filtered through a 70-µm strainer. The obtained single-cell suspensions were stained with anti-mouse CD45 PerCP-Cy5.5 (BioLegend, San Diego, CA, USA) to label the immune cells and stained with viability dye 450 (Peprotech, Cranbury, NJ, USA) at 4°C in dark for 30 min to distinguish between live and dead cells.

2.20 Isolation of CD8⁺ T cells

Human CD8⁺ T cells were isolated from peripheral blood of 4 healthy donors at physical examination using a human CD8 MicroBeads kit (130-045-201, Miltenyi Biotec, Bergisch Gladbach, Germany), according to the manufacturer's recommendations. All donors gave informed consent. Enriched CD8⁺ T cells (1 × 10⁵) were stimulated in 6-well plates precoated with 2 µg/mL anti-human CD3 (01121-25, BioGems, Westlake Village, CA, USA) and cultured in RPMI-1640 with 2 µg/mL anti-human CD28 (10311-25, BioGems) and 20 ng/mL recombinant human interleukin-2 (IL-2, 200-02, Pepro Tech, Rocky Hill, NJ, USA) for 48 h. Cells were then transferred to new wells with a new culture medium for follow-up experiments. PHA (1 µg/mL) was used to sitmulate Jurkat T cells for 24 h, and BCH (5 mmol/L) was applied to block the transport of kyn to cytoplasm in CD8⁺ T and Jurkat T cells for 48 h.

2.21 Flow cytometry assay

For TIL analysis, samples were incubated with membrane antibodies containing anti-mouse CD3 allophycocyanin (APC)-Cy7, anti-mouse CD8-FITC, anti-mouse PD-1 phycoerythrin (PE)-Cy7, and incubated with transcription factor fixation/permeabilization working solution (92550-00, BioGems) for 30 min at 4°C, and washed with permeabilization buffer (BioGems). Samples were then stained with intracellular or intranuclear antibodies, including anti-mouse interferon-gamma (IFN-γ) APC, anti-mouse lymphocyte activation gene 3 (LAG-3) PE, anti-mouse Ki-67 BV605, anti-mouse granzyme PE, anti-mouse perforin APC for 30 min at 4°C in the dark. The samples were then washed with PBS and analyzed using a BD Fortessa cytometer (BD Biosciences). For T cell analysis, human CD8⁺ T and Jurkat cells were detected using EdU Cell Proliferation Kit with 10 µmol/L Alexa Fluor 555 (Beyotime) for 2 h, 1 µmol/L carboxyfluorescein succinimidyl ester (CFSE, 65-0850-84, Thermo Fisher) for 10 min, and surface PD-1 expression (Anti-Human PD-1 PE, 31831-60, BioGems) assays, with the indicated treatments. All antibodies against immune cells were purchased from PeproTech and analyzed using a BD Fortessa cytometer (BD Biosciences). Flow cytometry data were analyzed using FlowJo 10.8.1 (BD Biosciences).

2.22 Enzyme-linked immunosorbent assay (ELISA)

The concentration of cytokines in C3H/He mice were detected by ELISA. Mouse serum was collected at the experimental endpoint and analyzed using a mouse tumor necrosis factor-alpha (TNF-α) ELISA kit (YC113, Shycbio, Shanghai, China) and an IFN-γ ELISA kit (YC111, Shycbio) according to the manufacturer's instructions.

2.23 Drug combination studies

In vitro drug combination studies were conducted similar to the steps in our previous study using the CCK8 assay [33]. The molar ratio of equipotent doses of the two agents (at the ratio of their half maximal inhibitory concentration [IC₅₀] values) was applied. Briefly, cells were seeded at 3,000 cells/well in 96-well plates, and then treated with serial dilutions of epacadostat (an indoleamine 2,3-dioxygenase 1 [IDO1] inhibitor, S7910, SelleckChem) and CH-223191 for 72 h. The cell viability was measured using the CCK8 assay. The dimensionless value at fraction affected (Fa) = 0.5 was calculated from Fa–CI curve. The combination index (CI) was used to analyze the synergistic inhibitory effects of drug combinations using CompuSyn software 1.0 (CompuSyn Inc., Paramus, NJ, USA) according to the Chou–Talalay equation [34].

2.24 Immunohistochemistry (IHC)

Tumor tissues were stained with H&E, Ki-67 (AF1738, Beyotime), and TdT-mediated dUTP-biotin nick end labeling (TUNEL, C1098, Beyotime), according to the recommended protocols. The tissues of the patient were embedded and sectioned into 4-µm slides. Subsequently, slides were deparaffinized, rehydrated, antigen repaired, and incubated with primary rabbit anti-human Siglec-15 Ab (dilution 1:100; 37925, Signalway Antibody), primary rabbit anti-human AhR Ab (dilution 1:100) overnight at 4°C. The immunoreactive score (IRS) was calculated by multiplying the intensity score (0  =  negative, 1  =  weak, 2  =  intermediate, 3  =  strong) and the fraction score (percentage of positive tumor cells, range  =  0-100) [32]. Images were acquired and analyzed using the CaseViewer software 2.4 (3DHISTECH Ltd., Budapest, Hungary).

2.25 Statistical analysis

A power analysis to determine sufficient sample size was not conducted. Samples were randomized before LC-MS, GC-MS, and RNA-seq analyses. R sofeware 4.2.0 (https://www.r-project.org/) was used for the analysis of metabolomic and transcriptomic data. The R packages pheatmap v1.0.12 (https://www.rdocumentation.org/packages/pheatmap/versions/1.0.12/topics/pheatmap) and ggplot2 v3.4.2 (https://ggplot2.tidyverse.org/index.html) were used for data visualization. For animal experiments, mice injected with tumor cells were randomized before allocation to cages. All other experiments were not randomized, and the investigators were not blinded to the allocation during the experiments or outcome assessment. Statistical analysis and plotting were performed using the GraphPad Prism 9.0 (GraphPad Software, San Diego, CA, USA). Two-sided Student's t-test or one-way analysis of variance (ANOVA) was used to compare two or more groups. Tukey's test was used as the post-hoc after ANOVA. The relationship between Siglec-15 or AhR expression and clinicopathological characteristics was determined using two-sided χ² tests. Data are expressed as mean ± standard error of mean. Statistical significance was set at P < 0.05.

3.1 Untargeted metabolomics revealed metabolic landscape and Trp/Kyn pathway alterations in HNSCC

Metabolic reprogramming is a critical characteristic of malignant tumors, which plays an important role in tumor progression, metastasis, and immune escape [3]. However, a comprehensive analysis of metabolic reprogramming in HNSCC remains elusive. To reveal the metabolic profile of HNSCC, 69 pairs of tumor and adjacent normal tissues were collected (Supplementary Table S1), and their metabolic and transcriptional alterations were analyzed using GC/LC-MS and RNA-seq (Figure 1A). Nine human papillomavirus (HPV)-negative oropharyngeal samples were included in this cohort. As HPV-negative HNSCCs have similar biological characteristics [35, 36], our research conclusions would not be affected by the diversity of sites. PCA and OPLS-DA model score plots showed a perfect separation between HNSCC and adjacent normal tissues (Supplementary Figure S1A-B). Seven-fold cross-validation and 200 RPT indicated the accuracy of the OPLS-DA model (Supplementary Figure S1C). Collectively, 178 differential metabolites were identified using LC-MS, out of which 121 had elevated levels, and 57 had decreased levels (Figure 1B, Suppementary Figure S1D). They were further classified as carbohydrates (n = 22), lipids and lipid-like molecules (n = 42), amino acids (n = 56), and others (n = 58) (Figure 1C, Supplementary Figure S1E). Amino acids were the major category of differential metabolites, indicating that altered amino acid metabolism may be involved in HNSCC development.

Similarly, KEGG metabolic pathway enrichment analysis revealed that several amino acid pathways, such as those of glycine, serine, threonine, and Trp, were enriched significantly (Supplementary Figure S1F). We identified the key metabolites or molecules in the metabolic and gene networks. The most significant hub network had 59 nodes and 771 interactions, containing 25 genes and 34 metabolites (Supplementary Figure S1G). Top 20 altered metabolites were obtained using GC-MS (Supplementary Table S6) and LC-MS analyses (Supplementary Table S7). Furthermore, we analyzed the contents of all human amino acids and their metabolites and found that Kyn had the most significant fold change (Figure 1D, Supplementary Figure S1H), indicating a potential association between Kyn and HNSCC progression. In addition, Kyn is the only molecule included in the top 10 altered metabolites and the hub network, suggesting that its reprograming may associated with HNSCC progression.

Trp is an essential amino acid involved in the biosynthesis of cellular proteins and formation of the cytoskeleton, and Kyn is a major metabolite of Trp catabolism [37]. Consequently, we focused on the Trp degradation trends, and Kyn, serotonin, quinolinic acid, 5-hydroxy-L-tryptophan, and N-methyltryptamine were identified by GC-MS (Figure 1E) and LC-MS analyses (Supplementary Figure S2A) as immediate products. The most significantly altered molecule in the Trp pathway was Kyn (Supplementary Figure S2B-C). Indeed, Kyn was the only differential metabolite involved in the Trp pathway that was detected by both GC-MS and LC-MS analyses (Figure 1F), and the mass spectrogram of Kyn is shown in Figure 1G. Correlation analysis showed good concordance between LC-MS and GC-MS data with regard to Kyn content (Supplementary Figure S2D).

The gene expression of metabolic enzymes involved in the Trp metabolic pathway was analyzed, and 13 differentially expressed genes were observed, among which kynurenine 3-monooxygenase (KMO), kynureninase (KYNU), interleukin-4-induced-1 (IL4I1), IDO1, and tryptophan 2,3-dioxygenase 2 (TDO2) were increased significantly in HNSCC (Supplementary Figure S2E-F). A schematic representation of the dysregulated enzymes and metabolites in the Trp pathway was drawn according to the metabolomics and transcriptomics data (Figure 1H). Considering the impact interconnected systems between metabolites, we found that blocking the Kyn pathway by epacadostat resulted in Trp accumulation and serotonin generation (Supplementary Figure S2G).

To confirm the metabolomics results, 5 pairs of fresh HNSCC tissues were collected for determination of Trp and Kyn contents. The HNSCC tissues had lower Trp levels and higher Kyn levels and Kyn/Trp ratios than the adjacent normal tissues (Figure 1I). Furthermore, the HNSCC cell lines had higher Kyn content than normal primary oral mucosal epithelia (Supplementary Figure S2H). Indeed, the Kyn content was increased in 69 patients with HNSCC (Figure 1J, Supplementary Figure S2I). Increased levels of Kyn were associated with advanced TNM stage, poor pathological grade, and greater depth of invasion (Figure 1K, Supplementary Figure S2J) but were not associated with age or sex (Supplementary Figure S2K-L).

ROC analysis was performed to further characterize the predictive value of Kyn. In GC-MS, Kyn had a high AUC value of 0.8895, indicating a good diagnostic efficiency in HNSCC (Supplementary Figure S2M). Similar results were obtained for LC-MS, with an AUC of 0.9053 (Supplementary Figure S2N). The results indicate that upregulated Kyn levels may play a role in HNSCC progression.

3.2 Kyn promoted HNSCC proliferation and metastasis

To investigate the biological role of Kyn in HNSCC, a series of in vitro and in vivo assays were conducted. The CCK8 and colony formation assays indicated that 50 and 100 µmol/L Kyn enhanced proliferation, colony formation ability and the percentage of EdU-positive cells compared with the control in Cal27 and HN30 cells (Figure 2A-C). Cells after Kyn stimulation migrated more quickly than DMSO control in the would healing assay(Figure 2D), and the number of migratory and invasive cells increased in the transwell assay (Figure 2E-F). Moreover, Kyn decreased the expression of E-cadherin significantly but increased that of Vimentin and Snail (Figure 2G, Supplementary Figure S3A), indicating that Kyn promotes epithelial-to-mesenchymal transition in HNSCC cell lines. The metastasis model showed that Kyn induced lung metastasis significantly, as shown by H&E staining. The number of metastatic nodules increased after Kyn treatment (Supplementary Figure S3B). Considering the limitations of the metastasis model, SCCVII orthotopic transplant models were established in C3H/He immune-competent mice (Figure 2H). Tumor volume and Ki-67 analysis showed that Kyn accelerated tumor progression (Figure 2I, Supplementary Figure S3C-D). More importantly, Kyn promoted neck lymph node metastases after tongue injection (Figure 2J).

3.3 Kyn promoted CD8⁺ T cell dysfunction and Siglec-15 expression in tumor cells

To explore the effects of Kyn for HNSCC immunity, the expression of immunecheckpoint molecules was detected by RNA-seq in the Kyn-high group and Kyn-low group. Notably, most classical immune checkpoint molecules were significantly higher in the Kyn-high group than in the Kyn-low group (Figure 3A). In addition, GSEA revealed that the T cell receptor signaling pathway was enriched significantly in Kyn-high patients (Figure 3B). The data suggested that Kyn may impede HNSCC immunity by affecting T cell signaling. To further investigate the role of Kyn in T cells, the primary human CD8⁺ T cells and Jurkat T cell line were used for follow-up experiments. EdU and CFSE staining showed that Kyn suppressed T cell proliferation (Figure 3C, Supplementary Figure S4). PD-1 is well-known as a T cell-exhaustion marker [38]. Therefore, PD-1 in T cells was detected after stimulation with Kyn, and PD-1 was upregulated significantly in CD8⁺ T and Jurkat T cells (Figure 3D). Western blotting analysis also confirmed that Kyn induced PD-1 expression in primary CD8⁺ T and Jurkat T cells, with or without PHA stimulation (Figure 3E), indicating that Kyn could promote T cell dysfunction.

We further examined the mRNA expression of effector molecules in Jurkat T cells and found that the cytotoxic effector molecules were down-regulated while the inflammatory molecules were up-regulated after Kyn stimulation (Figure 3F), indicating that Kyn may reduce the secretion of cytotoxic effector molecules and killing function of T cells. To test whether the process depends on Kyn intake, Kyn transport was blocked using the system L transporter SLC7A5 inhibitor BCH [39]. BCH reversed Kyn-mediated T cell dysfunction in both human CD8⁺ T cells and Jurkat T cells (Figure 3G). The data confirmed that Kyn could inhibit the T cell vitality and promote the PD-1 expression.

To investigate whether Kyn affects HNSCC immunity through targeting tumor cells, the expression of immune checkpoint ligand in tumor and adjacent normal tissues was analyzed by RNA-seq. Strikingly, SIGLEC15, a freshly reported immune checkpoint ligand molecule [40], exhibited the maximum fold change of expression (Figure 3H). Subsequently, the levels of SIGLEC15 in the CD274-low (a classical immune checkpoint ligand, encoding PD-L1) and CD274-high groups were detected via RNA-seq data. SIGLEC15 had higher levels in the CD274-high group than in the CD274-low group (Figure 3I). Based on the above findings, SIGLEC15 and CD274 mRNA and protein were measured by RT-qPCR and Western blotting after Kyn treatment. The mRNA levels of SIGLEC15 and CD274 were elevated after Kyn stimulation (Supplementary Figure S5A). As expected, Western blotting analysis verified that Kyn increased the expression of Siglec-15 and PD-L1 (Figure 3J, Supplementary Figure S5B-C). In addition, based on immunofluorescence staining, Kyn promoted the nuclear translocation of AhR (Figure 3K), a well-known feature of AhR activation by Kyn [41]. Furthermore, two AhR inhibitors, BAY-218 and CH-223191, reversed Kyn-induced Siglec-15 and PD-L1 expression upregulation (Figure 3L-M, Supplementary Figure S5D-E), and similar results were obtained following knockdown of AhR via siRNA transfection (Figure 3N, Supplementary Figure S5F-H). Therefore, we demonstrated that Kyn could induce CD8⁺ T cell dysfunction and promote Siglec-15 and PD-L1 expression through AhR in HNSCC.

3.4 Kyn induced Siglec-15-mediated immune escape

To test whether Kyn-induced expression of Siglec-15 depends on AhR transcriptional activity, ChIP-PCR assays were performed using two specific primers. The AhR antibody was bound to the promoter region of SIGLEC15 or CD274 after Kyn stimulation in the Cal27 and HN30 cell lines (Figure 4A). Dual-luciferase assays showed that Kyn increased the promoter activity of SIGLEC15 and CD274 significantly (Figure 4B-C). Conversely, forced knockdown of AhR in 293T cells reduced the luciferase activity of SIGLEC15 or CD274 (Figure 4D), which was also confirmed in the Cal27 cell line (Supplementary Figure S6). The data suggested that Kyn can transcriptionally activate the expression of Siglec-15 and PD-L1 through AhR in HNSCC.

To investigate the immunoregulatory functions of Siglec-15 in HNSCC, SIGLEC15-overexpressing subcutaneous tumorigenesis models were established in C3H/He mice (Figure 4E, Supplementary Figure S7A-B). Tumor growth curves and weight analysis showed that the tumors grew faster in the Siglec-15 group than in the vector group (Figure 4F-G). SCCVII tumors were confirmed using H&E staining (Figure 4H). TUNEL and Ki-67 staining analyses showed that Siglec-15 overexpression reduced apoptosis and promoted tumor cell proliferation (Figure 4I-J).

We further analyzed immune cell infiltration within the TME and observed that Siglec-15 reduced CD3⁺CD8⁺ T cell infiltration significantly (Figure 4K, Supplementary Figure S7C). The percentage of Ki-67⁺ T cells was reduced significantly in the Siglec-15 group (Figure 4L, Supplementary Figure S7D), indicating that Siglec-15 suppresses T cell proliferation. Regarding immune-exhaustion markers on T cells, there was a significant increase in the percentage of PD-1⁺ T cells in the xenograft TME (Figure 4M, Supplementary Figure S7E), whereas LAG-3⁺ showed no difference (Supplementary Figure S7F).

The effector molecules were analyzed, and the positive expression of IFN-γ, granzyme B, and perforin in tumor-infiltrating T cells was decreased notably in the Siglec-15 group (Figure 4N-P, Supplementary Figure S7G-I). Fluorescent multiplex immunohistochemistry showed that Siglec-15 decreased CD8⁺ T cell infiltration in xenograft tumors significantly (Figure 4Q, Supplementary Figure S7J). Furthermore, SIGLEC15-overexpressing orthotopic transplant models were established and showed that Siglec-15 promoted tumor progression and neck lymph node metastases (Figure 4R, Supplementary Figure S8A-B). In addition, the proportions of PD-1⁺CD8⁺ T cells in the Siglec-15 group were increased significantly based on fluorescent multiplex immunohistochemistry analysis results (Figure 4S, Supplementary Figure S8C). Furthermore, Kyn promoted PD-1⁺CD8⁺ T cell infiltration in orthotopic transplant models (Supplementary Figure S8C-D). These results suggested that Kyn promotes HNSCC progression through Siglec-15-mediated immune escape.

3.5 Targeting AhR reversed Kyn-Siglec-15-mediated immune escape in vivo

To confirm the role of Kyn and AhR in CD8⁺ T cell dysfunction, subcutaneous tumorigenesis C3H/He immunocompetent mouse models were established. Kyn promoted tumor growth, whereas the AhR inhibitor CH-223191 reversed the effect (Figure 5A), which was verified by tumor weight analysis (Figure 5B). Western blotting analysis also confirmed that CH-223191 could block Kyn-induced Siglec-15 expression in fresh tumor tissues (Figure 5C). SCCVII tumors were examined using H&E staining to confirm squamous cell carcinoma (Figure 5D), and TUNEL and Ki-67 staining analyses showed that Kyn reduced apoptosis and promoted tumor proliferation, whereas CH-223191 effectively reversed the effect (Figure 5E-F).

Serum TNF-α and IFN-γ levels were decreased in the Kyn treatment group but were increased in the CH-223191 group (Figure 5G). Flow cytometry analysis showed that the proportion of CD3⁺CD8⁺ T cells in the TME was reduced following Kyn treatment, and this was reversed by treatment with CH-223191 (Figure 5H, Supplementary Figure S9A), similar findings were observed with regard to Ki-67⁺ T cell percentage (Figure 5I, Supplementary Figure S9B).

The PD-1⁺ T cells induced by Kyn were blocked partially by CH-223191 (Figure 5J, Supplementary Figure S9C), indicating that Kyn can exhaust tumor-infiltrating CD8⁺ T cells by activating AhR. Moreover, IFN-γ, granzyme B, and perforin analyses showed that Kyn reduced the secretion of cytotoxic effector molecules in CD8⁺ T cells, whereas CH-223191 reversed IFN-γ and granzyme B expression, but not perforin (Figure 5K-M, Supplementary Figure S9D-F). Fluorescent multiplex immunohistochemistry observed that Kyn reduced CD8⁺ T cell infiltration, whereas CH-223191 reversed the effects (Figure 5N).

Lastly, in vivo drug toxicity was analyzed, and no significant alterations in the levels of alanine transaminase, aspartate aminotransferase, blood urine nitrogen, and creatinine were observed, as well as no pathohistological changes in the main organs (Supplementary Figure S9G-H). Moreover, we investigated the effect of a combination of epacadostat and CH-223191, and the CI value was < 0.5 (Cal27: 0.010; HN30: 0.400), which suggested a strong synergistic effect between epacadostat and CH-223191 (Supplementary Figure S10).

The results indicate that Kyn can promote HNSCC progression by inducing CD8⁺ T cell dysfunction, and the AhR inhibitor CH-223191 can reverse this effect.

3.6 Siglec-15 specific siRNA delivered by NH₂-MSN nanoparticles enhanced immunotherapy efficacy in vivo

Owing to a lack of Siglec-15 inhibitor, siRNA delivery through nanoparticles for mRNA interference is a potential solution for enhanced immunotherapy efficacy. Therefore, a nanoparticle delivery system (herein, NH₂-MSNs) was established to encapsulate siS15 for targeted delivery (Figure 6A). NH₂-MSNs loaded with siS15 were near transparent liquid (Figure 6B), and the morphology and zeta potential of a NH₂-MSNs and NP-siS15 were determined. The average size of NH₂-MSNs was 81 nm, with an electric potential of 31 mV. After siS15 loading, the diameters of the nanoparticles increased, and their edges became blurred, with a rapid decrease in zeta potential (Figure 6C-E). These findings indicate that siS15 was adsorbed on the surface of NH₂-MSNs.

Immunofluorescence analysis showed that NP-siS15 labeled with Cy5 could be effectively internalized and released in the cytoplasm of SCCVII cells at different mixing ratios (Figure 6F). Western blotting analysis confirmed that NP-siS15 effectively degraded the Siglec-15 protein in SCCVII compared with NP-siScr (Figure 6G). The results confirmed the successful establishment of NP-siS15, which can achieve efficient siRNA delivery into cancer cells and degrade the immune checkpoint molecule Siglec-15.

To investigate the effect of NP-siS15 treatment in vivo, subcutaneous tumorigenesis C3H/He mouse models were established (Figure 6H). The anti-tumor effect between the NP-siS15 and anti-PD-1 groups was not significantly different. More importantly, the mice treated with both anti-PD-1 and NP-siS15 exhibited prolonged survival compared with those treated with a single agent (Figure 6I, Supplementary Figure S11A), indicating a superior antitumor effect of anti-PD-1 and NP-siS15. The combination treatment group had significantly increased tumor cell apoptosis and decreased proliferation capacity, as indicated by TUNEL and Ki-67 staining (Figure 6J, Supplementary Figure S11B-C). TIL analysis showed that the NP-siS15 group could more significantly promote CD8⁺ T cell infiltration than the anti-PD-1 group, and the number of CD8⁺ T cells were the highest in the combination group (Figure 6K, Supplementary Figure S11D). Interestingly, the percentage of PD-1⁺CD8⁺ T cells decreased notably in the combination treatment group (Figure 6L, Supplementary Figure S11E). Fluorescent multiplex immunohistochemistry confirmed that NP-siS15 or anti-PD-1 promoted CD8⁺ T-cell infiltration, while combined NP-siS15 and anti-PD-1 had a better effect on immunotherapy efficacy (Figure 6M, Supplementary Figure S11F). Additionally, we applied the combination therapy strategy and observed that NP-siS15 enhanced the efficiency of anti-PD-L1 therapy (Figure 6N-P, Supplementary Figure S11G).

Overall, our results indicated that the newly-established NP-siS15 can effectively restore the killing function of CD8⁺ T cells and enhance anti-PD-1/PD-L1 therapy efficacy.

3.7 Siglec-15 expression positively correlated with CD8⁺ T cell exhaustion in HNSCC patients

An HNSCC tissue microarray was performed to determine the clinical correlation between AhR and Siglec-15 in CD8⁺ T cells. Siglec-15 expression was positively correlated with AhR expression (Figure 7A-B). Furthermore, high levels of AhR and Siglec-15 expression were associated with advanced TNM stage, while only AhR expression was associated with tumor location, which may be due to the large variation in the location of TMA samples (Supplementary Table S2). In particular, the percentage of Siglec-15-positive cases was 44.3% in HNSCC tissues, which was significantly higher than that of PD-L1-positive cases, at 22.8% (Figure 7C). Higher SIGLEC15 levels indicated poorer prognosis in the GEPIA HNSCC dataset (Figure 7D). The results indicated that Siglec-15 may play a more important immunosuppressive role than PD-L1 in HNSCC. More importantly, high Siglec-15 expression was closely associated with a high infiltration of exhausted CD8⁺PD-1⁺ T cells (Figure 7E). The results suggested that tumor-derived Kyn promoted CD8⁺ T cell exhaustion via the AhR/Siglec-15 pathway in HNSCC (Figure 7F).