Spatial dynamics of tertiary lymphoid aggregates in head and neck cancer: insights into immunotherapy response

Abstract

Background: Recurrent/metastatic head and neck squamous cell carcinoma (R/M HNSCC) generally has a poor prognosis for patients with limited treatment options. While incorporating immune checkpoint inhibitors (ICIs) has now become the standard of care, the efficacy is variable, with only a subset of patients responding. The complexity of the tumor microenvironment (TME) and the role of tertiary lymphoid structures (TLS) have emerged as critical determinants for immunotherapeutic response.

Methods: In this study, we analyzed two independently collected R/M HNSCC patient tissue cohorts to better understand the role of TLS in response to ICIs. Utilizing a multi-omics approach, we first performed targeted proteomic profiling using the Nanostring GeoMx Digital Spatial Profiler to quantify immune-related protein expression with spatial resolution. This was further characterized by spatially resolved whole transcriptome profiling of TLSs and germinal centers (GCs). Deeper single-cell resolved proteomic profiling of the TLSs was performed using the Akoya Biosciences Phenocycler Fusion platform.

Results: Our proteomic analysis revealed the presence of T lymphocyte markers, including CD3, CD45, and CD8, expressing cells and upregulation of immune checkpoint marker PD-L1 within tumor compartments of patients responsive to ICIs, indicative of 'hot tumor' phenotypes. We also observed the presence of antigen-presenting cells marked by expression of CD40, CD68, CD11c, and CD163 with upregulation of antigen-presentation marker HLA-DR, in patients responding to ICIs. Transcriptome analysis of TLS and GCs uncovered a marked elevation in the expression of genes related to immune modulation, diverse immune cell recruitment, and a potent interferon response within the TLS structure. Notably, the distribution of TLS-tumor distance was found to be significantly different across response groups (H = 9.28, p = 0.026). The proximity of TLSs to tumor cells was found to be a critical indicator of ICI response, implying that patients with TLSs located further from tumor cells have worse outcomes.

Conclusion: The study underscores the multifaceted role of TLSs in modulating the immunogenic landscape of the TME in R/M HNSCC, likely influencing the efficacy of ICIs. Spatially resolved multi-omics approaches offer valuable insights into potential biomarkers for ICI response and highlight the importance of profiling the TME complexity when developing therapeutic strategies and patient stratification.

Fig. 1

Study scheme. (1) Formalin-fixed paraffin-embedded (FFPE) tissue samples were collected prior to therapy from two independent R/M patient cohorts at the Princess Alexandra Hospital (PAH) and Royal Brisbane & Women’s Hospital (RBWH). In each clinical site, 17 samples were determined to be suitable for subsequent spatial analysis. (2) Tumor tissue serial sections and hematoxylin & eosin (H&E) staining were provided by the Pathology Queensland. (3) Using the Nanostring GeoMx DSP targeted spatial proteomics and transcriptomics were performed across the cohorts. (5) Spatial phenotyping of a sub-cohort was performed using the Akoya Biosciences PhenoCycler-Fusion. (4, 6) Data analysis consisted of probe quality control (QC), principal component analysis (PCA), differential expression (DE) and gene set enrichment analysis (GSEA) were conducted and followed by spatial analyses, including cell phenotyping and mapping

Fig. 2

Differential protein expression in patients with different response groups. (A) Spatial proteomics profiling was conducted on tissue samples from R/M HNSCC tumor tissues. Demarcation of tissues was achieved through masking PanCk + and PanCk- regions to delineate tumor and stromal compartments, respectively. The morphology markers included PanCk (green), CD45 (red), and SYTO 13 (blue) for the tumor cells, immune cells, and nucleus, respectively. Segmentation of tumors focused on regions of interest (ROIs) to distinctly identify the Tumor mask in green and the Stromal mask in yellow. (B) Utilizing the PanCk+/- feature, masks were generated to liberate barcodes for digital counting via the Nanostring nCounter platform. (C) MA plots of Mean Expression (AveExpr) vs. fold change (logFC) visualize the expression of protein biomarkers within tumor compartments in patients with complete response (CR) versus patients with progressive disease (PD). (D) MA plots of Mean Expression (AveExpr) vs. fold change (logFC) visualize the expression of protein biomarkers within tumor compartments in patients with disease control status (CR/PR/SD) versus patients with PD. Color coding represents markers that are not differentially expressed (gray), significantly upregulated (red), and downregulated (blue), based on a false discovery rate (FDR) of < 0.05 following multiple testing adjustments

Fig. 3

Visualization of significantly enriched gene sets different comparisons. (A) Spatial transcriptomics profiling was conducted on an R/M HNSCC tissue sample. Demarcation of the tissue was achieved through masking PanCk + and PanCk- regions to delineate tumor and stromal compartments, respectively. The morphology markers included PanCk (green), CD45 (red), and SYTO 13 (blue) for the tumor cells, immune cells, and nucleus, respectively. Utilizing the CD45+/- feature, masks were generated to liberate barcodes for sequencing via the Illumina’s NovaSeq 6000 platform. (B) The differential gene expression in TLS regions against GCs visualised as mean transcript expression (AveExpr, in log2) versus fold change (logFC, in log2). Color represents markers that are not differentially expressed (gray), significantly upregulated (red), or downregulated (blue), based on a false discovery rate (FDR) of < 0.05 following multiple testing adjustments. (C, E, F, and G) Gene-set clusters identifying dominant biological themes (E-G) upregulated or (C) downregulated in the comparison TLS vs. GCs with gene-set names depicted as representative wordclouds. Clusters 1,2,4, and 6 representing distinct biological themes were identified in each comparative analysis. (D, H, I, and J) The corresponding gene statistics (i.e. fold change in log2) within the gene sets clusters are plotted against the number of gene sets in the cluster to which the gene is differentially expressed

Fig. 4

Cell types. (A, B) Multiplex immunofluorescence imaging of key immune cell markers within TLS at single-cell resolution. The morphology markers included CD31 (yellow) and CD34 (green) for blood vessels, CD45RO (cyan) for memory T cells, Ki67 (red) for cell proliferation, HLA-DR (cyan) for antigen presentation, PanCk (green) for tumor cells, CD20 (brown) and CD21 (red) for B cells, CD4 (blue) for CD4⁺ T cells, CD8 (cyan) for CD8⁺ T cells, CD68 (purple) for macrophages, and CD11c (grey) for dendritic cells. (C) Representative image of cell types in the TLS structure. (D) Annotated clusters were merged into 7 cell types. (E) Heatmap indicating clustering of cell types based on markers. (F) Representative image of cell types in the GC structure. (G, H, I, and J) Representative images of cell proportions within the tumor and the stromal compartments of patients with different response group, including, CR, PR, SD, and PD, respectively

Fig. 5

Cellular interaction and distance analysis. (A) Representative field of view of enriched TLS: tumor interactions. Morphology markers included CD20 (orange) for B cells and PanCk (green) for tumor cells. (B) Violin plot representing the mean distance of TLS to the closest tumor cell in each response group. (C) Combined strip and point plots visualizing the number of cell types per response group. Responder group representing patients with CR/PR/SD, and Non-responder representing a patient with PD