Unveiling the immune microenvironment in sinonasal intestinal type adenocarcinoma: new arguments for immunotherapy

ABSTRACT

Sinonasal intestinal-type adenocarcinoma (ITAC) is a rare malignancy strongly associated with occupational exposure to wood dust that shares morphological and molecular similarities with colorectal adenocarcinoma. Standard treatments offer limited benefit for advanced or recurrent disease. It is therefore necessary to develop other strategies, such as immunotherapies, which require a better understanding of the tumor immune microenvironment, are needed. In this study, we investigated the immune microenvironment of ITACs (n=59), focusing on tumor-infiltrating T lymphocytes (TILs) and macrophages, in situ using immunohistochemistry, ex vivo by multiparametric flow cytometry, as well as by transcriptomic profiling. High CD8⁺ T-cell densities were associated with significantly improved disease-free and overall survival and emerged as an independent prognostic factor. These TILs express immune checkpoints (IC), the most highly expressed being PD1, TIGIT, and TIM-3. Furthermore, TIM-3 was not only expressed by TILs but also by immunoregulatory macrophages, and may represent an important immune escape mechanism in ITACs, since TIM-3 blockade in patient-derived explants led to increased tumor cell apoptosis. Tumor cells strongly expressed CD155 (TIGIT ligand) but not PDL1 (PD1 ligand); however expressed by myeloid cells. Finally, transcriptomic analyses identified one cluster displaying features of “hot tumors” characterized by adaptive immune activation, a better prognosis, and a higher immunotherapy response score, and another cluster with features of “cold tumors”. These first exploratory and descriptive findings in ITACs provide a rationale for further investigations using ex vivo and in vivo models to evaluate the therapeutic potential of immunotherapies in these rare cancers.

Introduction

Sinonasal intestinal-type adenocarcinoma (ITAC) is a rare entity limited to the nasal cavity and paranasal sinuses. ¹ It is considered an occupational disease, typically associated with exposure to wood dust. ^2-4 The disease may occur decades after exposure. ⁵ ITACs display morphological features closely resembling those of adenocarcinomas of the gastrointestinal tract. Histologically, ITACs can be subdivided into several morphologic subtypes: papillary or well-differentiated, colonic or moderately differentiated, solid or poorly differentiated, and mucinous with or without signet ring cells. Mixed patterns, particularly those combining mucinous and other subtypes, can be observed. Mucinous ITACs with signet ring cells are associated with the poorest prognosis. ^{6 , 7}

In addition to their morphological similarities, ITACs share some genetic features with colorectal adenocarcinoma (CRC), such as recurrent mutations in TP53, APC, CTNNB1, KRAS, and BRAF genes. ^8-10 Moreover, microsatellite instability, which occurs in approximately 15%–20% of CRC, ^{11 , 12} appears to be exceptional in ITACs. ^13-15

Overall, patients with recurrent, unresectable, or metastatic disease derive limited benefit from chemotherapy, and the prognosis remains poor. No dedicated clinical studies evaluating immune checkpoint inhibitors have been conducted in ITACs. While immunotherapy has become an established treatment modality for head and neck squamous cell carcinomas, ¹⁶ no data are available on the efficacy of such therapies in ITACs. Moreover, to our knowledge, no comprehensive analysis of the ITAC immune microenvironment has been published. To date, only one case of partial response to nivolumab combined with chemotherapy has been reported. ¹⁷

Our objective was therefore to characterize the ITAC’s immune microenvironment at the protein and transcriptomic levels, in relation to the clinicopathological and survival data, with a particular emphasis on immune checkpoints (IC) and their ligands, in order to identify potential targets for immunotherapy and provide a rationale for future clinical trials.

Materials and methods

Patient clinical data collection

Clinical data were retrospectively collected from 59 patients diagnosed with sinonasal intestinal-type adenocarcinoma (ITAC) and treated at Nantes University Hospital since 2000. The most recent follow-up update was performed on May 1 st, 2025. The pTNM stage was assessed according to the 8^th UICC TNM classification.

Ethical considerations

This study was conducted in accordance with the Declaration of Helsinki and approved by a certified French Ethics Committee (Comité de Protection des Personnes, CPP, authorization 2021-A01662-39). All patients provided written informed consent for the use of their samples and clinical information for research purposes.

Immunohistochemistry (IHC)/Immunofluorescence

Immunostaining was performed on 3 µm formalin-fixed, paraffin-embedded sections of a representative block from surgical resections of 59 ITACs. The following primary antibodies were used: CD3 (polyclonal rabbit, Agilent), CD8 (monoclonal mouse, clone C8/144B, Agilent), T-BET (monoclonal mouse, clone 4B10, Santa Cruz Biotechnology), PD-1 (monoclonal mouse, clone NAT105, Abcam), TIGIT (monoclonal rabbit, clone E5Y1W, Cell Signaling Technology), TIM-3 (monoclonal rabbit, clone D5D5R, Cell Signaling Technology), Lag3 (monoclonal mouse, clone 17B4, Biotechne), NKG2A (polyclonal rabbit, Abcam), CD68 (monoclonal mouse, clone PG-M1, Agilent), PD-L1 (clone22C3, Agilent), CD155 (monoclonal rabbit, clone D8A5G, Cell Signaling Technology), galectin-9 (polyclonal rabbit, Abcam), HLA-DR (monoclonal mouse, clone SPM289, Abcam), HLA-E (monoclonal mouse, clone MEM-E/02, Bio-Rad), PMS2 (monoclonal rabbit, clone EP51, Agilent), MSH6 (monoclonal rabbit, clone EP49, Agilent), MLH1 (monoclonal mouse, clone ES05, Agilent) and MSH2 (monoclonal mouse, clone FE11, Agilent).

Immunostaining was performed using an autostainer (AutostainerLink48, Agilent) according to a standard protocol, including antigen retrieval. The immunological reaction was visualized using a peroxidase/diaminobenzidine Envision system (Agilent), and the sections were counterstained with hematoxylin.

On digitized whole slides (scanned with HT2, Hamamatsu), densities in T cells (CD3⁺, CD8⁺, and T-BET⁺) and in cells expressing the various ICs were assessed per mm², using the open-source software QuPath (0.3.2). A pathologist (CL) manually annotated representative areas within the tumor and in the peritumoral stroma. For mucinous ITACs with signet-ring cells, in which cells diffusely infiltrate the stroma, two representative areas containing both tumor cells and their stroma were selected. For all the cases, the largest possible amount of tissue was selected for each area, with at least 1000 cells in the tumor and 3000 cells in the stroma. Regions of interest were selected in proportion to the surface of each representative area to fully capture tumor heterogeneity, excluding areas that were entirely necrotic or lacked tumor tissue.

Concerning the IC ligands, a semi-quantitative score was performed, taking into account the percentage of positive tumor cells (score 0: no positive cells; score 1: ≤10% positive cells; score 2: >10% - <50% positive cells; score 3: ≥50% positive cells). PD-L1 scoring was also assessed using the combined positive score (CPS).

Multiplex immunofluorescence staining was performed on 3 μm paraffin sections of 3 ITACs using a mixture of anti-CD8 (mouse IgG1,κ), anti-CD68 (mouse IgG3, κ), and anti-TIM-3 (rabbit IgG) antibodies and then with goat anti-mouse IgG1-AF568, IgG3-AF488, and goat-anti-rabbit-AF647. Multichannel acquisition was performed using the Axio Observer 7 inverted microscope (Zeiss) and the Zen software.

Ex vivo multiparametric flow cytometry assay

Fragments of fresh tumors from 14 ITACs were minced in RPMI 1640 medium and mechanically dissociated (GentleMACS tissue Dissociator, Miltenyi). To exclude dead cells, staining with the Fixable Viability stain (FVS780, BD Biosciences) was performed for each panel.

The first multiparametric panel, designed to study IC in T cell subpopulations, used the following monoclonal antibodies: CD3-BUV395 (BD Biosciences), CD4-BUV496 (BD Biosciences), CD8-APC (BioLegend), TCRαβ-BB700 (BD Biosciences), PD-1-BV421 (BD Biosciences), TIGIT-PECy7 (BioLegend), TIM-3-PE (BioLegend), Lag3-BV650 (BioLegend), and NKG2A-FITC (REA, Miltenyi Biotec). The second multiparametric panel was designed to assess IC ligands in the Epcam⁺ tumor cells as well as in the CD11b⁺ myeloid cells, using the following antibodies: Epcam-FITC (BioLegend), CD11b-BV421 (BD Biosciences), PDL-1-BUV395 (BD Biosciences), CD155-APC (BioLegend), Galectin-9-PE (BioLegend), HLA-DR-BV650 (BioLegend), and HLA-E-PECy7 (BioLegend). Finally, to assess the phenotype of TIM-3⁺ macrophages, a staining using anti-CD11b -BUV395 (BD Bioscience), anti-TIM-3-BV421 (BD Biosciences), and anti-CD163-PE (BD Biosciences) antibodies was performed. Ig isotype controls were performed for each panel.

Cells were acquired from the viable cell gate on a BD LSR Fortessa X-20 flow cytometer and analyzed using FlowJo 10.10.0 software.

Patient-derived tumor explant cultures

Fresh fragments of tumor (40 mg) from 5 ITAC were pinned onto Sylgard-coated Petri dishes and maintained in culture for 24 h in a 95% O₂/5% CO₂ humid atmosphere on a rocking platform at low speed. The following medium was used: RPMI 1640 (Eurobio) with glutamax 2 mM (Gibco) supplemented with human serum AB (8%, Pan Biotech), antibiotics (penicillin/streptomycin/gentamycin, Gibco), amphotericin B (Gibco) and IL-2 (proleukin, Novartis, 50 U/ml). These explant cultures were performed in the presence of an anti-TIM-3 blocking antibody or isotype control (10 μg/ml each, BioLegend), in triplicate or quadruplicate per condition. After 24 h of culture, the explants were formalin-fixed and paraffin-embedded. Tumor cell apoptosis was assessed by immunohistochemistry using a cleaved caspase-3 antibody (polyclonal rabbit, Cell Signaling Technology). The percentage of active caspase-3⁺ tumor cells was calculated using the QuPath software.

RNA isolation and gene expression

Three consecutive 10 μm paraffin curls were cut from each FFPE block. Total RNAs were isolated using the miRNeasy RNeasy FFPE kit with the deparaffinization solution (Qiagen).

The nCounter Human PanCancer IO 360 Panel (NanoString Technologies) was used with the nCounter SPRINT Analysis System according to the manufacturer’s instructions. The RNA input was set to 75 ng per sample corrected with the DV200 percentage with a maximum input of 1000 ng. Initial quality control assessment was performed using nSolver 4.0 Analysis Software (NanoString Technologies). The digital counts were standardized using positive-control probes and normalized using the GeNorm algorithm from the nSolver advanced analysis module (v. 2.0.134). The data were analyzed using the nSolver 4.0 Advanced Analysis module, using default settings to generate pathway signature scores.

The T Cell Inflamed GEP (TIS) score, including 18 genes known to be associated with the response to IC blockade, was established, as described by Cristescu et al. ¹⁸ in their Supplementary materials.

Tumor mutational burden (TMB)

Three consecutive 10 μm paraffin curls were cut from each FFPE block, and total DNA was isolated from paraffin curls using the QIAamp DNA FFPE Advanced UNG Kit with the deparaffinization solution (Qiagen). The library preparation was done using the QIAseq Targeted DNA Panel Library Prep Kit - Human TMB Panel (DHS-6600Z, QIAGEN). Sequencing was performed on a NovaSeq X Plus (Illumina Inc.) instrument (2 × 150, 2 × 10). The raw data was de-multiplexed, and FASTQ files for each sample were generated using the bcl2fastq software v2.20.0.422 (Illumina Inc.). All primary analysis was carried out using CLC Genomics Server 23.0.5.

Statistical and bioinformatic analyses

Statistical and bioinformatic analyses were performed using RStudio (version 2025.05.1 Build 513) unless otherwise specified. Survival analyses, including Kaplan–Meier curves, reverse Kaplan–Meier estimation for median follow-up time, log-rank tests, and multivariate Cox proportional hazards models, were conducted using the survival and survminer R packages.

Gene expression data were visualized using heatmaps generated with the ComplexHeatmap package. Statistical significance for differential gene expression analysis was assessed using a negative binomial model, and p-values were adjusted using the Benjamini‒Yekutieli (BY) procedure, with a significance threshold set at an adjusted p-value <0.05.

Comparative analyses of quantitative variables were carried out using box plots; normally distributed data were analyzed with Student’s t-test, while non-normally distributed data were evaluated using the Mann–Whitney U test. Box plots were designed and statistically analyzed using GraphPad Prism (version 9). Statistical significance was set at p < 0.05.

Regarding bioinformatic analysis of somatic variants, sequencing reads from the TMB panel underwent a pipeline. The resulting VCFs underwent SBS96 mutational signature analysis with SigProfilerMatrixGenerator v1.3.6 (PASS variants only), de novo signature extraction (SigProfilerExtractor v1.2.6, 1-8 signatures, NMF optimization), and COSMIC v3.4 assignment (SigProfilerAssignment v1.1.1; cosine similarity >0.02).

Results

Patient characteristics and outcomes

A total of 59 patients were included: 58 men and 1 woman. All patients in our cohort were exposed to wood dust. The clinical characteristics of the patients are detailed in Table 1. In our cohort, the median follow-up time was 143 months, with a minimal follow-up without events of 24 months. The median time to relapse was 29 months (range 5–212), with 2 patients developing secondary lymph node metastases, and 6 developing secondary distant metastases. In terms of histological subtypes, moderately differentiated ITACs were the most common (39%), followed by mucinous ITACs without signet ring cells (20%). Mucinous ITACs with signet ring cells and well-differentiated ITACs each represented approximately 15% of the cohort, while poorly differentiated ITACs were the least represented (8.5%). All patients had a microsatellite stable (MSS) phenotype, as assessed by immunohistochemistry.

Table 1.

Patient characteristics according to histological subtypes.

	Well differentiated N = 9	Moderatly differentiated N = 23	Poorly differentiated N = 5	Mucinous w/o signet ring cell N = 12	Mucinous w/signet ring cell N = 10	p-value	Overall N = 59
Age at Diagnosis, Median, [Min–Max]	66, [42–78]	69, [43–90]	70, [65–75]	70, [53– 85]	69, [57– 78]	> 0.9 2	69, [42 - 90]
Tobacco use, n (%)	2 (25%)	5 (29%)	3 (60%)	2 (18%)	4 (44%)	0.5 3	16 (32%)
Missing data	1	6	0	1	1		9
TNM, n (%)						0.001 3
T4a–T4b	4 (50%)	6 (26%)	5 (100%)	7 (58%)	9 (90%)		31 (53%)
T1–T3	4 (50%)	17 (74%)	0 (0%)	5 (42%)	1 (10%)		27 (47%)
Missing data	1	0	0	0	0		1
Initial treatment, n (%)						0.6 3
Surgery + adj radiotherapy	7 (78%)	20 (91%)	5 (100%)	11 (92%)	8 (89%)		51 (89%)
Chemo or radiotherapy alone	1 (11%)	2 (9.1%)	0 (0%)	0 (0%)	1 (11%)		4 (7.0%)
Surgery alone	1 (11%)	0 (0%)	0 (0%)	1 (8.3%)	0 (0%)		2 (3.5%)
Missing data	0	1	0	0	1		2
Relapse, n (%)	4 (50%)	4 (20%)	1 (20%)	6 (50%)	7 (88%)	0.012 3	22 (42%)
Missing data	1	3	0	0	2		6
Status, n (%) 1						0.6 3
Alive	4 (44%)	12 (52%)	2 (40%)	6 (50%)	2 (20%)		26 (44%)
DOD	2 (22%)	5 (22%)	0 (0%)	2 (17%)	4 (40%)		13 (22%)
UCD	2 (22%)	3 (13%)	1 (20%)	3 (25%)	4 (40%)		13 (22%)
NaD	1 (11%)	3 (13%)	2 (40%)	1 (8.3%)	0 (0%)		7 (12%)

¹ DOD: Died of disease, UCD: Unknown cause of death, NaD: Natural death.

² Kruskal‒Wallis rank sum test.

³ Fisher's exact test.

Noticeably, univariate analyses showed that patients with mucinous ITACs with signet ring cells had a worse prognosis, since they had a significantly lower disease-free survival (DFS) than other subtypes (p = 0.0031, Supplementary Figure 1A); they also relapsed more quickly with the lowest median DFS (22.5 months) compared to other subtypes (197 months). Considering the TNM stage, patients with T4a/b tumors also had a worse prognosis (Supplementary Figure 1C). In addition, multivariate analysis using the Cox regression model for DFS showed that TNM stage was an independent prognostic factor in ITACs (HR 8.28, p = 0.003 for T4a/b stages) (Supplementary Table 1).

Assessment of the tumor immune microenvironment in situ and ex vivo

High densities of CD3 and CD8 lymphocytes are associated with better outcome

Quantitative immunohistochemical analyses (n = 59) of CD3, CD8 and the Th1-specific transcription factor T-BET, both inside the tumor and in the peritumoral stroma, showed an inter-patient heterogeneity of T-cell densities, with, however, about 30% of ITACs displaying high T-cell infiltration (Figure 1A). In all cases, higher T-cell densities were observed in the peritumoral stroma than inside the tumor (Figure 1A/B). Based on median T cell densities, tumors were stratified into CD3⁺ and CD8⁺ high/low subgroups. Interestingly, mucinous tumors with signet ring cells, which had the worst prognosis, had the lowest densities of CD3 and CD8 in the stroma with median densities of 169 cells/mm² for CD3⁺ cells and 101 cells/mm² for CD8⁺ cells. In contrast, the other subtypes all had a median density above 400 cells/mm² for CD3⁺ cells and 188 cells/mm² for CD8⁺ cells.

Figure 1.

T-cell densities in sinonasal ITACs and their impact on patient survival. (A) In situ assessment of CD3, CD8, and T-BET infiltrating ITACs by immunohistochemistry/quantitative analysis, both in the stroma and tumor. (B) Representative CD8 immunostaining of an ITAC with high T cell infiltrate, showing numerous CD8⁺ cells (brown) in the stroma (star) and a few intra-tumoral CD8⁺ cells (arrows). C, D- Impact of CD8 density (high vs low) on patient survival (DFS, OS, Kaplan–Meier curves).

We then analyzed the impact of T-cell density on patient survival, focusing on CD8⁺ cells that play a key role in various solid tumors. ^{19 , 20} As shown in Figure 1, the highest densities of CD8⁺ cells in the stroma were significantly associated with longer DFS in univariate analysis (197 months in “CD8 high” patients vs 29 months for “CD8 low”) (Figure 1C). The same trend was observed for overall survival (poorer OS rate in ITACs with low CD8 density (76 months) vs high CD8 density (143 months)) (Figure 1D). In addition, multivariate analysis using the Cox regression model for DFS showed that stromal CD8⁺ T-cell density was an independent prognostic factor for sinonasal ITACs (HR 3.57, p = 0.041 for patients with low CD8⁺ T-cell density) (Supplementary Table 2).

Expression of immune checkpoints (IC) and their ligands

IHC showed that IC was expressed at higher levels in the stroma than inside the tumor. The most highly expressed ICs were PD-1, TIGIT, and TIM-3 (Figure 2A/B and Figure 3A). NKG2A and Lag3 were weakly expressed. Regarding the different histological subtypes, mucinous ITACs with signet ring cells exhibited the lowest levels of immune cell infiltration (Figure 2B); however, no statistically significant differences were observed between all the histological subtypes.

Figure 2.

Expression of immune checkpoints (IC) in sinonasal ITACs, in situ by IHC/quantitative analysis (A, B), and ex vivo by flow cytometry (C, D). (A) PD1 and TIGIT are the most highly expressed IC, mainly present in the stroma (A). (B) Radar plot of the mean values of IC densities by histologic subtype. (C) Frequency of IC expression ex vivo by flow cytometry after mechanical dissociation of the tumor, among CD3⁺ TCRαβ⁺ cells, as well as in the CD4 and CD8 subsets. (D) Representative case analyzed by flow cytometry (isotype control in blue, specific staining in red).

Figure 3.

TIM3 expression in ITACs assessed by immunohistochemistry/immunofluorescence (A, B, C) and flow cytometry (D, E). High numbers of TIM3 +  cells (brown) were observed in the stroma (A and B showing a representative case) in the areas where many CD68 +  macrophages were present (B TIM3 and CD68 immunostaining on serial sections). (C) Immunofluorescence multistaining showing CD68 +  macrophages (green) co-expressing TIM3 (purple), localized near CD8 T lymphocytes (red). (D, E) Frequency of TIM3 +  cells determined by flow cytometry, among CD11b +  myeloid cells and among CD11b + CD163 +  macrophages (dot plots (D), n = 10 ITACs, and E, histogramsof a representative case, isotype control in blue, specific staining in red).

Multiparametric flow cytometry was performed ex vivo on fresh tumor samples (n = 14), confirming and extending the IHC data. First, the vast majority of CD3⁺ cells bear the T-cell receptor αβ (TCRαβ; mean 92% ± 2%). Among these CD3⁺ TCRαβ⁺ cells, a standard CD4/CD8 ratio was observed, with an average of 67% CD4⁺ cells (range 35%–85%) and 27% CD8⁺ cells (range 15%–60%), respectively. As shown in Figure 2 C/D, the highest IC expression found in CD3⁺ cells, whether CD4⁺ or CD8⁺ cells, was that of PD-1, TIGIT, and TIM-3. LAG-3 was barely expressed, and NKG2A was expressed at low levels only in the CD8⁺ subset (Figure 2C).

In addition, IHC revealed a high density of TIM-3⁺ cells in the stroma, and to a lesser extent in the tumors of some patients (Figure 3A/B). In these cases, numerous TIM-3⁺ cells have a macrophage morphology, a finding corroborated by CD68 immunostaining on serial sections (Figure 3B) and by multiplex immunofluorescence labeling, revealing that cells co-expressing TIM-3 and CD68, often located near CD8⁺ T cells (Figure 3C). Finally, flow cytometry analyses (n = 10) confirmed the presence of CD11b⁺ TIM-3⁺ myeloid cells (10 to 50% of CD11b⁺ myeloid cells expressing TIM3, Figure 3D/E). In addition, the majority of the CD11b⁺ CD163⁺ macrophages (anti-inflammatory M2 phenotype) expressed TIM-3 (Figure 3D/E).

Expression of IC ligands by tumor cells and myeloid cells in ITACs

As shown in Figure 4A/B, PD-L1 was not expressed by tumor cells in the vast majority of ITACs, but was expressed by stromal myeloid cells. We also measured the combined positive score (CPS). Thirty samples (50%) had no positive internal control and were therefore excluded from the analysis; for the remaining 29 samples, 48% of tumors had a CPS score > 1 (min 3; max 11). Interestingly, unlike PD-L1, tumor cells from the majority of patients showed strong and diffuse expression of CD155 (TIGIT ligand). Tumor cells also expressed, at lower levels, Galectin-9 (TIM-3 ligand) as well as HLA-E (NKG2A ligand). The tumor cells did not express HLA-DR (LAG-3 ligand), except in two cases, one of which presented the highest IC densities.

Figure 4.

Expression of immune checkpoint ligands in ITACs by immunohistochemistry (A, B) and flow cytometry (C, D). (A) Percentages of ITACs scoring positive for the various IC ligands (semi-quantitative scores are specified in the Materials and Methods section). (B) Representative immunostainings of CD155 and PD-L1, showing high CD155 expression (brown) by tumor cells and conversely no PD-L1 expression. (C and D) Frequency of IC ligand expression by flow cytometry in the Epcam⁺ (tumor cells) and CD11b⁺ (myeloid cells) compartments and representative dotplots (isotype control is shown in blue, and positive staining is shown in red).

Flow cytometry confirmed the immunohistochemical data. As shown in Figure 4C/D, epithelial cells (EpCAM+), devoid of PD-L1, expressed CD155, galectin-9, and expressed lower levels of HLA-E and HLA-DR. Myeloid cells (CD11b+) expressed PD-L1, galectin-9, and at lower levels CD155, HLA-E, and HLA-DR (Figure 4C, D).

Effect of TIM3 blockade on tumor explant cultures

Since ITACs can be rich in immune cells expressing TIM-3, we assessed the impact of an anti-TIM-3 blocking antibody on tumor cells in patient-derived tumor explants maintained in culture for 24  h (n = 5). As shown in Figure 5 (A-C), TIM-3 blockade significantly increased tumor cell apoptosis, as assessed by active caspase-3 immunostaining, compared to untreated control cultures (mean value 22.8% vs 4.5%, p = 0.01 with a mean fold increase of 17.7).

Figure 5.

Effect of an anti-TIM-3 antibody on 24 h explant cultures of ITAC. Quantitative assessment of apoptotic tumor cells using IHC with an anti-active caspase-3 antibody (A) and representative immunostaining of untreated (B) or anti-TIM3-treated (C) explants (apoptotic cells in brown, nuclei in blue).

Molecular data

Transcriptomic profiling uncovers three molecularly distinct clusters

We first characterized ITACs (n = 26) using scores of predefined biological pathways. As shown in Figure 6A, we identified 3 distinct clusters of ITACs. Cluster 1 (n = 6), which includes ITACs with high CD8 T-cell densities (Figure 6B), was characterized by (i) the overexpression of genes in pathways related to an adaptive immune response (cytokine/chemokine signaling, interferon signaling, cytotoxicity, and antigen presentation) and (ii) by a decreased expression of genes related to cancer cell biology (cell proliferation and metabolic stress). In contrast, Cluster 3 (n = 8) showed decreased expression of genes involved in immune responses, and the overexpression of genes linked to cell proliferation and metabolic stress. Cluster 2 (n = 12) was a heterogeneous mixture with no significant over- or under-expressed pathways.

Figure 6.

Transcriptomic clusters of ITACs. (A) Heatmap of biological pathway scores with the identification of 3 clusters. (B) Stromal CD8⁺ density determined by IHC in cluster 1 versus cluster 3. (C) Patient overall survival for the 3 clusters of ITACs (Kaplan–Meier curves). (D) Representative genes involved in tumor immune responses and cancer cell progression among the 3 clusters, identified through differential gene expression analysis between clusters 1 and 3 (BY-adjusted p-value).

When examining differentially expressed (DEG) genes between clusters 1 and 3 (Supplementary Table 3), genes involved in modulating anti-tumor immune responses, such as granzymes, perforin, some cytokines (IL-6), and the gene encoding CD11c (ITGAX), were significantly higher in cluster 1 compared to cluster 3 (Figure 6D). Conversely, genes involved in tumor cell proliferation and cancer progression in solid tumors, such as WNT11 (WNT/β-catenin pathway), CDC20 (cell division cycle 20) and CENPF (centromer protein F), were significantly downregulated in cluster 1 compared to cluster 3 (Figure 6D).

Given the histological subtypes, it was difficult to assign a particular histological subtype to a given cluster. The mucinous subtype without signet ring cells was mainly present in cluster 1, while the mucinous subtype with signet ring cells was only present in cluster 2. The only poorly differentiated carcinoma was found in cluster 3, together with the majority of moderately differentiated carcinomas, while well-differentiated carcinomas were distributed across clusters 1 and 2 (Supplementary Table 4).

Finally, we assessed the difference in patient survival between the clusters. There was a trend toward a better prognosis for patients in cluster 1 (n = 6), with all patients alive during the analysis period, while patients in cluster 3 (n = 8) had a median overall survival of 62 months (Figure 6C). In terms of disease-free survival, only one patient in cluster 1 recurred (197 months after initial treatment), compared with 3 in cluster 3 (not statistically significant). The median follow-up time for the three clusters was 71 months.

Evaluation of molecular immunotherapy response biomarkers

Evaluation of T-cell-inflamed GEP (TIS) Score

Ayers et al. developed an immunotherapy response score called the T-cell inflamed GEP score (TIS), which was validated in several cohorts of various cancers, including colon cancer. ^{18 , 21-23} The cut-off score proposed in various publications is −0.318, with patients whose score is above this value showing a better response to immunotherapy than those below. In our cohort, 10 patients (38%) had a score above −0.318, and 16 had a score below −0.318 (62%). Patients with a high TIS score had higher CD8 densities in the stroma and tumor than patients with a low TIS score (Supplementary Figure 2).

Low proportion of high TMB tumors

We analyzed the TMB score for 24 patients, and found a relatively low median of 5 mutations/Mb (Min = 0.86; max = 65.09), with 3 patients (12%) having a score greater than 10 mutations/Mb. Interestingly, these 3 cases displayed a high CD8 infiltrate in the stroma. Likewise, patients displaying a high CD8⁺ T-cell infiltrate had a higher average mutational burden compared with those with low infiltration (230 vs. 142 mutations, respectively). Most mutations were missense (76%), followed by frameshift variants (19%) (Supplementary Figure 2).

From these sequencing data, we determined the single-base substitution (SBS) signatures for these 24 patients. Across the entire set of mutations, SBS102 was the most represented signature (accounting for 42% of all mutations). The SBS1 signature was also detected, albeit to a much lesser extent (accounting for 4.3% of all mutations). SBS1 is considered a clock-like signature. Notably, the sole ITAC lacking the SBS102 signature was a patient with a smoking history exceeding 50 pack-years and was the only ITAC harboring the SBS5 signature, which is thought to be potentially associated with tobacco exposure. ²⁴

Discussion

To our knowledge, this is the only study to provide a multiparametric characterization of the immune microenvironment of sinonasal ITACs in relation to clinicopathological features.

First, our cohort of 59 patients with ITAC is comparable to previously published series in terms of clinicopathological features. Consistent with the literature, we observed poorer outcomes in patients with mucinous ITAC with signet ring cells as well as in those with tumors diagnosed at an advanced TNM stage ^{6 , 7} in univariate models for DFS and OS. However, in our multivariate Cox model for DFS, only the TNM stage emerged as an independent adverse prognostic factor, not the histological subtype, likely due to the limited size of our cohort.

Second, our assessment of T-cell infiltrates in situ by IHC on FFPE sections and ex vivo by flow cytometry demonstrated, on the one hand, varying infiltration profiles of stromal CD8⁺ T-cells, with some tumors exhibiting dense infiltrates. On the other hand, we confirmed previous literature findings that a high CD8⁺ T-cell infiltrate correlates with a better patient prognosis. ^{25 , 26} Indeed, patients with high CD8⁺ T-cell density had a longer DFS and a lower risk of recurrence in our multivariate Cox model.

IHC and multiparametric flow cytometry allowed us to demonstrate that T lymphocytes (CD8 and CD4) express mainly PD-1, TIGIT, as well as TIM-3 at a lower level. These ICs can interact with their ligands on tumor cells/immune cells (PDL-1, CD155, galectin-9, respectively), and could be responsible for immune escape mechanisms in sinonasal ITACs. LAG-3 and NKG2A were also detected but at much lower levels, and NKG2A was found only in CD8⁺ T cells, similarly to colorectal cancer. ²⁷ Interestingly, PD-L1, expressed by myeloid cells, is hardly expressed by tumor cells; however, it strongly expresses CD155, a feature also found in colorectal cancer cells. ²⁸ Our findings thus highlight some similarities in terms of immune contexture and IC/IC ligands expression between sinonasal ITACs and CRC. Indeed, CD8 high/low subgroups as well as IC high/low subgroups have been described in CRC, ^{29 , 30} and the TIGIT/CD155 axis is also highly present in CRC, in addition to the PD-1/PD-L1 axis. ²⁸

In addition, we also show herein that ICs are not only expressed by T lymphocytes in sinonasal ITACs, but also by tumor-associated macrophages (TAM). Our findings indicate that a subset of patients displays a substantial infiltration by TIM-3⁺ macrophages. Importantly, our flow cytometric analysis revealed that a high proportion of immunoregulatory CD163⁺ macrophages, which are associated with poor outcomes in several cancers, ^31-33 exhibit co-expression of TIM-3. This finding suggests that TIM-3 could be a potential immune escape mechanism in ITACs. Indeed, in solid tumors such as head and neck squamous cell carcinoma, ³⁴ hepatocellular carcinoma, ³⁵ or CRC, ³⁶ it has been reported that TIM-3 can be involved in the development of tumor-promoting M2 macrophages. Our preliminary experiments using explant cultures of fresh sinonasal ITACs, showing a significantly higher tumor cell apoptosis in explants treated with anti-TIM-3 antibodies versus controls, also point to this hypothesis in ITACs, even though the number of tumors analyzed was limited (n = 5) due to the rarity of ITACs. Explant cultures were chosen because this approach has already been successfully employed in previous studies to investigate treatment responses. ^37-39 Interestingly, Vanmeerbeek et al. also found a pro-apoptotic effect of anti-TIM-3 antibodies in vitro when co-culturing non-immunogenic Lewis lung carcinoma cells with TIM-3 + VISTA +  macrophages. These data suggest, in the absence of effector T lymphocytes, that proinflammatory activation of TIM-3⁺ VISTA⁺ TAMs upon TIM-3 blockade created TAM-driven anticancer cytotoxicity via TRAIL signaling. ⁴⁰ Although the experimental conditions differ, these findings provide insight into the potential mechanism of action of anti-TIM3 blocking antibodies in our explant model. The respective contribution of TIM3-expressing T lymphocytes and TIM3-expressing macrophages and the mechanism of apoptosis resulting from TIM3 blockade remains to be elucidated in more comprehensive ex vivo and in vivo models.

Our transcriptomic analysis confirms and extends our in situ and ex vivo data, identifying for the first time three distinct clusters of ITACs according to their gene pathway signature. Although these results should be interpreted with caution, given the limited number of tumors in some clusters, this is the first time such an analysis has been performed in ITACs. It is difficult to compare these clusters with the molecular classification of CRC (CMS), ⁴¹ since ITACs do not exactly share CRC molecular alterations that have been described in MSS CRC. However, an analogy can likely be drawn between our cluster 3 and CMS2, as both clusters show overexpression of cell cycle-related genes and underexpression of immune-related genes. Conversely, our cluster 1 most closely resembles CMS1, as both are characterized by an overexpression of immune related-genes. Our cluster 2 represents a heterogeneous group that may be defined as a mixed or indeterminate cluster. Our pathway analysis is well supported by the differential gene expression between cluster 1 and cluster 3. Interestingly, our SBS signature analysis performed on a subset of our ITAC cohort identified the SBS102 signature as predominant, a signature also observed in MSS CRC. ⁴²

Finally, we sought to explore in our cohort the potential prediction of response to immunotherapy, using biomarkers established in other cancer types. Several companion diagnostic tests are available to guide the use of anti-PD1/PD-L1 therapies, such as a PD-L1 CPS ≥ 1 ⁴³ or a high tumor mutational burden (TMB ≥ 10). ^{44 , 45} Moreover, we were able to calculate an immunotherapy response score named T-cell inflamed GEP (TIS). This transcriptomic signature, predictive of response to anti PD-1/PD-L1 immunotherapy, was validated in several tumors, including CRC. ^{18 , 21-23} Overall, across these three biomarkers of immunotherapy response (CPS, TMB, and TIS), a subset of patients in our cohort displayed scores compatible with potential eligibility for immunotherapy. These findings should be interpreted with caution, given the observational nature of the study. Indeed, these biomarkers (CPS, TMB, and TIS) have not yet been validated in ITACs.

Altogether, our molecular, immunohistochemical, and flow cytometry data collectively demonstrate the existence of immunologically “hot” and “cold” tumors among sinonasal ITACs. The identification of the PD-L1/PD-1, TIGIT/CD155, and TIM-3/Galectin-9 immune checkpoint axes in ITACs contributes to a more comprehensive characterization of the immune landscape of these tumors. In addition, preliminary findings from our ex vivo explant experiments suggest that ITACs may be sensitive to TIM-3 blockade. These data provide a rationale for further investigations using ex vivo and in vivo investigations to evaluate the therapeutic potential of immunotherapies in these cancers, particularly for patients with advanced or refractory disease.

Supplementary Material

Supplementary_Material.docx

Supplementary_Material.docx

SupplementaryTable3DE_cluster1VScluster3.xlsx

SupplementaryTable3DE_cluster1VScluster3.xlsx

Funding Statement

This work was supported by grants from CHU de Nantes (AOI 2024), “Entraide Cancer 44”, “Ligue contre le cancer Grand Ouest” (CD44, CD35, and CD72), and the French Society of Pathology (SFP).

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Data availability statement

Datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.

Consent for publication

All authors give their consent for the submission and publication of this manuscript.

Supplementary material

Supplemental data for this article can be accessed at https://doi.org/10.1080/2162402X.2026.2725500.