All patients with HNC were recruited from Xuzhou Central Hospital. The eligibility criteria were as follows: patients with primary head and neck cancer who were scheduled for biopsy or lesion resection. We excluded those who had received any preoperative neoadjuvant therapy, including radiotherapy, chemotherapy, targeted therapy, immunotherapy, or other molecular treatments. The pathological types of all HNC samples were identified by the Pathology Department of Xuzhou Central Hospital, and the samples were then processed and tested. This study was approved by the Ethical Review Committee for Biomedical Research of Xuzhou Central Hospital. (Approval Number: XZXY-LK-20250115-0011) Before collecting specimens, the patients and their families provided signed informed consent.

The reagents and materials used in this study are listed in Table 1. All compounds were obtained commercially according to the specifications detailed in the table.

Fresh HNC tissues were obtained through biopsy or surgical resection, stored in 20% fetal bovine serum (FBS)-DMEM, and transported on ice to our laboratory for ALI organoid culture within 2 h. The tumor tissue was cut into < 0.3 cm diameter fragments, mixed with recombinant collagen solution without enzymatic digestion, and evenly spread over pre-cured collagen gel in the internal Transwell insert with a permeable membrane at the bottom. The insert was placed in the cell culture dish and culture medium was added between the culture dish and the insert to form a dual-disc ALI culture system (Figure 1). In this way, the upper part of the tumor tissue and collagen mixture was directly exposed to air, and the culture medium penetrated the bottom of the collagen through micropores, forming an ALI culture model. ALI-HNCO collagen gel was mixed on ice with Cellmatrix I A solution, 10× Ham’s F-12 concentrated sterile medium, and sterile recombination buffer solution at a ratio of 8:1:1, containing final concentrations of 200 mmol/L HEPES, 0.05 mol/L NaOH, and 2.2 g NaHCO₃ per 0.1 L (15).

The culture medium contained 1× Glutamax additive, 10 mmol/L HEPES, 1 mmol/L N-acetylcysteine, 10 mmol/L nicotinamide, 1× Pen-Strep glutamine, 1× B27, 10 μmol/L SB202190 (a p38MAPK inhibitor), 0.5 μmol/L A8301, 0.05–0.25 μg/mL R-spondin1 recombinant protein, 0.05–0.2 μg/mL recombinant human noggin protein, 50 ng/mL EGF, 10 nmol/L gastrin, 0.05–0.2 μg/mL human Wnt-3A, and 500 IU/mL IL-2 in Advanced DMEM/F12 basic culture medium.

The mature ALI-HNCO was collected and fixed with 4% paraformaldehyde at 4°C for 30 min. The HNCO precipitate was blown to even distribution, followed by pre-embedding with liquid agarose, cooling to a gel-like state, and then subjected to gradient dehydration, transparency, paraffin-embedding, sectioning, and hematoxylin and eosin (H&E) staining and immunohistochemical (IHC) staining. After sealing, it was analyzed and compared with the corresponding original tumor tissue sections.

After the paraffin sections prepared in Section 2.4 were deparaffinized, antigen-retrieved, and subjected to three rounds of antibody staining. Each round included serum blocking, primary antibody incubation at 4°C overnight, HRP-labeled secondary antibody, and TSA-based fluorescent tyramide covalent labeling. The primary antibodies included CK5 (Servicebio, GB111246; 1:10000) for SCC, CK7 (Servicebio, GB12225; 1:2000) for MEC and DC, CD3 (Servicebio, GB12014; 1:2,000), and α-SMA (Servicebio, GB15044; 1:10000). The nuclei were counterstained with DAPI, and the sections were treated to quench autofluorescence, followed by sealing. Finally, the sections were sealed, and images were collected. This TSA-based method enables triple-labeling for protein co-localization analysis.

PD-1 is a member of the B7/CD28 costimulatory receptor family and can be expressed on the surface of activated CD8+ T cells and B cells. It primarily regulates CD8+ T cell activity through binding to its ligands, PD-L1 and PD-L2, transmembrane proteins on tumor cells. Pembrolizumab has a high affinity for PD-1 on the surface of T cells and acts by blocking PD-1/PD-L1 cell channels, thereby facilitating cancer cell killing by the immune system. Since pembrolizumab is currently the first-line immunotherapy for HNC, the agent was chosen to explore the feasibility of using ALI-PDO models to evaluate the efficacy of immunotherapeutic drugs. IgG 4 (10 μg/mL) was the control group, and pembrolizumab (10 μg/mL) was the experimental group. We administered two consecutive doses of medication (changed every 3 days) to the cultured ALI organoids and cultured them for 7 days.

The release of IFN-γ during T-lymphocyte activation was measured using the Human IFN-γ ELISA Kit. Cells treated with pembrolizumab and ICI were designated the experimental group, and IgG4 was designated the control group. The concentration of IFN-γ in culture medium was determined after 7 days. The average IFN-γ concentration was calculated by comparing the OD450 value of the sample measured on the microplate reader to the standard curve.

The inner matrix gel was collected in a 15 mL centrifuge tube using precision tweezers, digested with 300 U/mL collagenase IV at 37°C for 30 min, followed by washing. The cell clusters were resuspended in 2 mL of Liberase TL (25 U/mL) and digested at 37°C for 15 min to prepare single cells. The single cells were washed once with MRS (5 mM EDTA/PBS) and FACS solution (PBS containing 2% FBS), respectively. Then, the cells were filtered, and the cell pellet obtained by centrifugation was resuspended in an appropriate amount of FACS solution. The sample was prepared by adding the following antibodies: Fixable Viability Stain 780, anti-CD45-percp-cy5.5, CD3-PE-Cy7, anti-CD4-PE, anti-CD8-APC, and anti-CD69-APC-R700. After staining, it was incubated on ice in the dark for 30 to 45 min. The sample was washed with 1 mL of FACS solution and resuspended in an appropriate amount of FACS solution or organoid growth medium (usually 100–500 μL). A total of 100,000 cells were collected using the LSRF Fortessa and analyzed using FlowJo software (version 10.8.1, Treestar).

Statistical analyses were carried out with GraphPad Prism, employing an independent samples t-test for two-group comparisons and one-way ANOVA for multi-group comparisons. Results are presented as follows: ns (not significant) for P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

We established 28 HNCO models (28/35, with a success rate of 80%) using patient-derived HNC resection specimens. Table 2 summarizes the detailed clinical and pathological data of the patients corresponding to HNCO. We successfully established an immune microenvironment model for HNC (Figure 1). Among the 35 cultured cases, 28 grew successfully, for an overall success rate of 80%. Within about 2 weeks of culture, the growth of organoids was recorded by bright-field photography using an inverted microscope (Figure 2A).

H&E staining and IHC assays were employed to analyze and compare the morphological and histological characteristics of HNCO and the corresponding tumor specimens. The nuclear atypia exhibited by HNCO was highly similar to that of homologous parent tumor tissues. CK5, EGFR, P63, and Ki67 staining of the SCC organoids was positive, which was consistent with those of the parents. CK9, EMA, P63, and Ki67 staining of the DC organoids were all positive, which was consistent with those of the parents. Among them, the staining results of CK7, E-cad, and P63 in one MEC organoid HNCO23 were consistent with those of its parental tissue/cells and showed positive signals, while Ki67 was consistent with the parents and showed negative results. That is, the results of typical IHC organoid markers were consistent with those of the parents, which can prove their homology at the histopathological level. The results also demonstrated the successful construction of HNCOs (Figures 2B, C).

Immune cells, fibroblasts, and other non-tumor cells in the TME have a significant impact on tumor occurrence and development. Unlike the previous stromal cell scaffold method, the ALI-HNCO culture can maintain other components of the TME for a period of time. For this purpose, we performed immunofluorescence staining of parental tumor tissues and ALI-HNCOs cultured for 14 days to analyze the T-cell immune marker CD3 and the fibroblast immune marker SMA. The results showed that CD3+ cells and α-SMA+ cells could be preserved in ALI-HNCOs, confirming that the ALI-HNCO model can preserve T cells and fibroblasts in the TME (Figures 3A–C).

CTLs release effector cytokines, such as IFN-γ, which participate in the initiation and differentiation of CTLs and directly kill tumor cells (27). During tumor progression, T cells enter a state of exhaustion. The characteristics of exhausted T cells include the continuous high expression of multiple inhibitory receptors (e.g., PD-1) and the inability to secrete IFN-γ (28). Pembrolizumab effectively blocks the “brakes,” such as immune checkpoints (PD-1/PD-L1), enabling T cells to regain the ability to secrete IFN-γ. After 14 days of culture, the experimental group was treated with pembrolizumab, and the control group was treated with IgG4. Pembrolizumab mainly activated cytotoxic T cells (CD8+ T cells), while IFN-γ was mainly secreted by activated CD8⁺ CTLs. Therefore, we validated T-cell activation by measuring IFN-γ concentrations in the culture medium using ELISA. The ELISA results showed that IFN-γ concentrations in the six ALI-PDTO experimental groups increased significantly, revealing a remarkable difference compared with the control groups (Figure 4).

Pembrolizumab can rescue exhausted T cells. Due to the release of inhibition, antigen-specific CD8+ T cells undergo clonal proliferation, resulting in a significant increase in cell numbers. In contrast, the total number of CD4+ T cells remained relatively stable. Thus, the ratio of CD8+ to CD4+ naturally increased. An increase in the proportion of CD8+/CD4+ T cells was observed in five ALI-PDTO groups treated with pembrolizumab compared with the control group treated with IgG4, indicating that ICB rescued exhausted CD8+ T cells and reversed the immune depletion state of the TME. The IFN-γ concentrations in these five ALI-PDTO experimental groups increased significantly. However, in the remaining 23 groups, no significant change in the CD8+/CD4+ T-cell ratio was seen between the pembrolizumab-treated group and the IgG4-treated group, indicating the failure of CD8+ T-cell expansion after anti-PD-1 treatment (Figure 5A).

CD69 is an early marker of T-cell activation. Therefore, the expression of CD69 is a key signal for the successful activation of T cells and their transition from a quiescent state to a functional state (29, 30). Analysis revealed that in the five ALI-PDTO groups with elevated CD8+ T cells, the proportion of cytotoxic CD8+ and CD69+ cell populations among all T cells was significantly increased in the pembrolizumab-treated group compared with the IgG4-treated control group. In other organoid groups, no significant difference was seen in the proportion of CD69+ cells between the experimental and control groups (Figure 5B).

FVS staining of the cells after organoid digestion indicated that the organoid group, with a significantly increased proportion of cytotoxic CD69+ T cells, showed increased numbers of dead tumor cells, demonstrating anti-PD-1-dependent tumor killing activity (Figure 5C). This indicated that T cells were activated to exert a killing effect on tumor cells.

In conclusion, our results demonstrated the heterogeneity of responses to ICB. Among the 28 groups of organoids, five responded to pembrolizumab, while the remaining 23 were resistant to it. The OOR to the immunotherapy drugs was 17.86% (5/28), which was consistent with existing studies on OORs in HNCs, including squamous cell carcinoma and salivary gland cancer.