Integrated immune profiling of chordomas reveals spatially organized niches and functional heterogeneity

Abstract

Background

Chordomas are locally aggressive notochordal tumors with no systemic therapy options. As an ultra-rare cancer type, our understanding of its immune landscape is limited. While tumor-associated macrophages (TAMs) and T cells are critical components of the immune landscape, their functional states and interactions remain poorly understood.

Methods

We conducted an integrative analysis of 35 chordoma samples and six paired tumor-PBMC samples using single-cell RNA sequencing (scRNA-seq), T-cell receptor (TCR) profiling, and multiplex immunofluorescence. Immune cell phenotypes, spatial distribution, TCR motif diversity, and functional states were assessed using unbiased co-expression network analysis and predictive modeling.

Results

Chordomas exhibited remarkable immune cell heterogeneity, ranging from highly infiltrated to immune-­desert tumors. Tumor-associated macrophages dominated the tumor microenvironment (TME) and were enriched for antigen-processing pathways. T-cell receptor profiling revealed clonal overlap between tumor-infiltrating and peripheral T cells, suggesting systemic anti-tumor responses. Exhausted CD8+ T cells exhibited restricted clonality and tumor-specific amino acid motifs. Weighted gene co-expression network analysis (WGCNA) identified gene modules associated with immune activation and suppression, underscoring the dual roles of immune cells in the TME. Spatial analysis revealed fibrous septa as immune interaction hubs, where immune cell clustering was significantly higher than in tumor regions.

Conclusions

This study advances understanding of the chordoma immune landscape by integrating spatial, transcriptomic, and TCR data. The findings highlight systemic and local immune dynamics, reveal tumor-specific TCR motifs, and identify potential therapeutic targets. These insights provide a foundation for developing personalized immunotherapies to overcome immune suppression and enhance anti-tumor immunity in chordomas.

Key Points.

1. Chordomas exhibit a heterogeneous TME, with TAMs enriched in antigen-processing pathways.

2. Exhausted CD8+ T cells show clonal overlap with peripheral T cells, highlighting potential systemic anti-tumor reactivity.

3. Fibrous septa serve as immune interaction hubs, facilitating immune cell clustering.

Importance of the Study

This study provides critical insights into the immune landscape of chordomas, a rare and aggressive tumor with limited therapeutic options. By integrating single-cell RNA sequencing, TCR profiling, and spatial analysis, we uncover the functional heterogeneity of immune cells in the tumor and periphery. Our findings highlight dendritic cells and TAMs as key players enriched in antigen presentation pathways, while exhausted CD8+ T cells exhibit restricted clonality and clonal overlap with peripheral T cells, suggesting systemic tumor-specific immune responses. The identification of fibrous septa as immune interaction hubs further emphasizes the spatial organization of immune cells within the TME. By uncovering tumor-specific TCR motifs and functional immune cell states, this study advances the understanding of chordoma immunobiology and establishes a foundation for developing personalized immunotherapies targeting tumor-specific antigens and overcoming immune suppression in these tumors.

Chordomas are rare, locally aggressive notochordal remnant tumors that occur primarily in the axial skeleton, with common sites, including the skull base, sacrococcygeal region, and various spine segments. Chordomas are particularly challenging to treat due to their aggressive nature, complex location, and resistance to conventional therapies.¹ While en bloc surgical resection remains the standard of care,² the intricate location of these tumors often makes complete resection unachievable, leading to poor long-term outcomes. Despite advancements in surgical and radiotherapeutic interventions, chordomas are notorious for their high recurrence rates, metastases, and limited responsiveness to conventional therapies, including chemotherapies.^2–4 There are currently no FDA-approved systemic therapies for chordomas. For these reasons, new pharmacotherapies are urgently needed to advance the treatment paradigm for chordoma patients.

Recent studies have highlighted the complex and immunological nature of the chordoma TME,⁵ which plays an important role in tumor progression and therapeutic resistance.⁶ The TME comprises a heterogeneous population of immune cells, stromal components, and extracellular matrix elements that interact dynamically with each other. In this context, immune checkpoint blockade therapies and other immunotherapies have been tested in chordomas with mixed responses.^7–12 Understanding these immune infiltrates’ composition and functional states is crucial for developing and tailoring effective immunotherapeutic strategies.

In chordomas, single-cell RNA sequencing (scRNA-seq) studies have revealed diverse immune cell populations, including T cells, macrophages, dendritic cells, and mast cells, each exhibiting unique functional states.^13,14 Notably, myeloid populations such as tumor-associated macrophages (TAMs) are dominant players in the chordoma tumor microenvironment.^15,16 Tumor-associated macrophages in many tumors, including chordomas, are frequently characterized by immunosuppressive phenotypes that facilitate tumor progression by dampening anti-tumor immune responses.^13–15 However, the functional diversity of myeloid cells, including their potential roles in immune activation, remains elusive in the context of chordomas.

T cells represent another critical component of the immune system with the potential to mediate effective anti-tumor immunity. However, the T cell functional states and T-cell receptor (TCR) clonality in chordomas remain unexplored. Investigating TCR diversity and clonal relationships between tumor-resident and peripheral T cells provides a unique opportunity to identify systemic immune responses and tumor-specific antigen targets. This could inform the development of personalized immunotherapies such as adoptive T cell transfer or TCR-engineered therapies.

In this study, we present a comprehensive characterization of the chordoma immune landscape by integrating ­single-cell transcriptomics, TCR profiling, and multiplex immunofluorescence. Our analysis encompasses 35 chordoma samples and paired tumor and blood samples from six patients, creating an extensive single-cell atlas of the chordoma TME. We explore the spatial organization, phenotypic diversity, and functional states of immune cells within chordomas while identifying shared TCR sequences between tumor-infiltrating and peripheral T cells. These findings offer crucial insights into the immunobiology of chordomas and provide a framework for developing targeted immunotherapies tailored to this rare malignancy.

Methods

Human Sample Acquisition

Tumor tissues and blood samples were collected under an IRB protocol (STU00095863) approved by Northwestern University. Patients provided informed written consent for the use of their samples in research. Tumor tissues were surgically resected at our institution, with histological review conducted by CMH and PJ to confirm good quality and sufficient tumor tissue for analysis.

Tumor and Immune Cell Isolation, Processing, and Sequencing

Six newly diagnosed chordoma samples and their corresponding PBMCs were collected for scRNA-seq as detailed in Supplementary material.

Single-Cell Data Processing, Quality Control, and Cell Type Annotation

Detailed single-cell sequencing processing is included in Supplementary material.

Weighted Gene Correlation Analysis (WGCNA) and Functional Enrichment

scRNA-seq data from myeloid and T cells were analyzed to explore transcriptional heterogeneity and functional pathways in the TME of chordomas. The dataset, stored in anndata format, was imported into R via the reticulate package for integration with Seurat. Raw count matrices were extracted, transposed, and normalized. A Seurat object was constructed with metadata and cell-type annotations for clustering and visualization.

PHATE embedding, precomputed in Python, was integrated into Seurat using a custom function. Data were normalized via LogNormalize, highly variable features identified, and PCA applied for dimensionality reduction. Harmony was used for batch correction, and UMAP visualized the corrected latent space.

Weighted Gene Correlation Analysis, implemented with hdWGCNA, identified co-expressed gene modules. Features were selected based on expression in ≥5% of cells, and metacells were generated for robust module detection. The optimal soft thresholding power was determined via TestSoftPowers, and the topological overlap matrix enabled gene clustering. Module organization was assessed using gene dendrograms. As per WGCNA convention (Langfelder and Horvath, 2008), modules were labeled using color-based identifiers (eg black, turquoise, yellow) to facilitate consistency and reproducibility.

Eigengenes (hMEs) and connectivity scores (kME) ranked hub genes within each module. Functional enrichment analysis, using Gene Ontology (GO) terms, was performed with clusterProfiler and msigdbr. Gene Ontology terms with adjusted P values <.05 were considered significant, and results were visualized as dot plots. UCell calculated enrichment scores,¹⁷ and Seurat tools were used for visualization.

TCR Analysis and Motif Identification

T-cell receptor sequences were analyzed using Grouping of Lymphocyte Interactions by Paratope Hotspots 2 (GLIPH2) to identify shared binding motifs in the CDR3 region.¹⁸ The algorithm identified statistically enriched motifs indicative of tumor-specific antigen recognition. Motif analysis revealed distinct TCR signatures in tumor-resident and peripheral T cells.

Cell-Cell Interaction Analysis

Ligand-receptor interactions were quantified using Liana,¹⁹ integrating multiple statistical methods to measure interaction strength and specificity. A graph-based network of interactions was generated using Graph-Tool. Dimensional reduction and hierarchical block modeling were performed to reveal the structure of cell-cell interaction networks.

Multiplex Immunofluorescence of Human Chordoma Samples

Thirty-five chordoma samples (23 newly diagnosed and 12 recurrent tumors) were used for multiplex immunofluorescence using the Akoya Biosciences platform as detailed in Supplementary Methods.

Sequential Multiplex Immunofluorescence (seqIF)

Six tumor specimens were used for sequential multiplex immunofluorescence using the Lunaphore’s COMET platform as detailed in Supplementary material.

Methods for k-Nearest Neighbor (kNN) Analysis of Immune Cell Interactions

k-Nearest Neighbor analysis was performed to investigate spatial interactions among immune cells in multiplex immunofluorescence images. Cell coordinates (“Cell X Position”, “Cell Y Position”) were extracted, and distances to the 40 nearest neighbors were computed. Neighbor indices were matched with cell IDs across samples, and missing phenotype values were cleaned. CD3+CD8+ T cells, CD3+CD8− T cells, CD163+ macrophages, and CD20+ B cells were categorized based on marker expression.

To ensure comparability, distances were normalized by the global median kNN distance and rescaled to microns. Self-interactions were excluded to focus on intercellular interactions. CD3+CD8+ T cells were used as the reference population, and kernel density estimation with bootstrapping (100 replicates) was applied to assess spatial clustering. Density plots, stratified by tissue compartments (tumor regions and fibrous septa), were generated using “ggplot2,” with confidence intervals highlighting spatial proximity patterns. Analyses characterized interactions between CD163+ macrophages, CD20+ B cells, CD3+CD8− T cells, and CD3+CD8+ T cells in the TME.

Statistical Analysis

Statistical comparisons were performed using Python and R. One-way ANOVA with Tukey’s multiple comparison test was used to compare group means, while paired t-tests analyzed paired data. Statistical significance was set at P < .05. Replicate numbers for each experiment are provided in the figure legends. Data analysis and figure generation were conducted using Python, Prism v. 9 (GraphPad), R v. 4.0.3, RStudio, and Adobe Illustrator v. 29.2.1.

Results

Multiplex Immunofluorescence and Single-Cell RNA Sequencing Reveal Immune Heterogeneity in Chordomas

Chordomas are known for their complex and immunosuppressive TME, yet the immune landscape within these rare tumors remains poorly understood. To bridge this gap, we combined multiplex immunofluorescence and scRNA-seq to analyze cellular heterogeneity and immune dynamics in chordomas. The study included 35 tumor samples for multiplex immunofluorescence and 6 paired tumor-PBMC samples from sacral chordoma patients for scRNA-seq. (Figure 1A). Demographic and clinical characteristics of our chordoma cohort are listed in Table S1.

Figure 1.

Multimodal analysis of the chordoma immune microenvironment. (A) Schematic overview of the analytical workflow, highlighting the use of multiplex immunofluorescence and single-cell RNA sequencing to characterize the immune landscape of chordomas. Samples from the skull base and sacrum are shown. Modified from images available at www.anatomystandard.com (CC BY-NC 4.0). (B) (Top) Heatmap showing the hierarchical clustering of chordoma samples based on immune cell percentages, using markers CD163, CD20, CD3+CD8–, and CD3+CD8+ cells. Annotations include age, sex, anatomical location, diagnosis, group, and histological subtype. (Bottom) Representative multiplex immunofluorescence images using the Akoya Biosciences system of tumor tissue sections stained for CD3 (T cells), Brachyury (notochordal cells), CD163 (macrophages), CD20 (B cells), CD8 (cytotoxic T cells), and nuclei (DAPI). Insets highlight specific immune cell populations within the tissue. (C) PHATE plot illustrating immune cell populations within the tumor microenvironment, with each group labeled by marker expression. (D) PHATE plot showing immune cell populations grouped by sample of origin (tumor or PBMC). (E) Dot plot of marker gene expression across cell types. Dot size indicates the fraction of cells expressing each marker, and shading intensity represents mean expression levels. Cell types include B cells, T cells, NK cells, monocytes, macrophages, dendritic cells, mast cells, and malignant cells. Key markers are labeled. Expression levels are represented on a continuous scale from low to high; dot size scale: 20%-100%.

We quantified the overall percentages of major immune cell types using multiplex immunofluorescence. Figure S1A shows the distribution of macrophages (CD163+), B cells (CD20+), cytotoxic T cells (CD3+CD8+), and helper T cells (CD3+CD8−) across the entire cohort. Macrophages constituted the dominant immune population in most tumors (7.6% of all cells; 95% CI of mean: 6.1%-9%), followed by lower percentages of CD3+CD8+ T cells (1.4% of all cells; 95% CI of mean: 1%-1.8%), CD3+CD8- T cells (3.3% of all cells; 95% CI of mean: 1.8%-4.8%), and B cells (1% of all cells; 95% CI of mean: 0.3%-1.8%). In a subset of 6 samples, we added CD11c+ to identify dendritic cells, confirming macrophages as the largest immune subset and revealing a lower, consistent level of CD11c+ dendritic cells (Figure S1B and C).

Multiplex immunofluorescence analysis of our chordoma cohort revealed an important heterogeneity in the immune cell composition across samples. To showcase this immunological diversity, we performed unsupervised hierarchical clustering, which delineated three illustrative groups. One group exhibited robust T cell-dominant environments, with markedly higher CD3⁺CD8⁺ and CD3⁺CD8⁻ fractions than the other samples (P < .0001 and P < .05, respectively; Fig. S1D). A second group comprised “immune-desert” tumors with minimal lymphoid or myeloid presence, and a third displayed intermediate infiltration driven predominantly by myeloid cells (P < .0001; Fig. S1D). These visual groupings underscore the breadth of immune profiles in chordomas, and although they span extremes of infiltration, they did not correlate with survival, demographics, tumor site, or disease status (Fig. S1E and F). When the same clustering approach was applied to an independent chordoma cohort, we again observed three analogous patterns: high T cell infiltration, immune-desert, and myeloid-enriched profiles (Fig. S2). Of note, these descriptive groupings serve to underscore the substantial variability in immune landscapes across chordomas, rather than to define discrete functional subtypes.

To characterize immune heterogeneity at single-cell resolution, we integrated scRNA-seq data from paired tumor and PBMC samples collected from chordoma patients. This joint analysis confirmed extensive cellular heterogeneity and revealed diverse populations, including T cells, macrophages, monocytes, dendritic cells, mast cells, and malignant cells (Figure 1C, Figure S3A). Copy number variant patterns distinguished malignant from non-malignant cells (Figure S3B). Tumor samples contained higher fractions of macrophages (supporting multiplex immunofluorescence findings), as well as more monocytes and T cells, while PBMCs comprised circulating immune subsets such as B cells, T cells, monocytes, and NK cells (Figure 1C and D). Established markers validated these phenotypes (Figure 1E), underscoring the immunological diversity in local tumor and systemic compartments of chordoma patients.

Functional and Spatial Characterization of Myeloid Cells in the TME

Given the abundance of myeloid cells, particularly macrophages, in the chordoma TME (Figure S1A and C), we aimed to elucidate their diversity, localization, and functional states across tumor and peripheral compartments. We examined the spatial distribution of CD163+ macrophages by delineating tumor regions and fibrous septa, finding that macrophages were highly enriched in fibrous septa compared to tumor regions (Figure 2A). Quantitative analysis confirmed significantly higher CD163+ cell percentages in fibrous septa (P < .0001, paired t-test, Figure 2B), suggesting that structural TME features influence macrophage recruitment and function.

Figure 2.

Spatial, transcriptional, and functional analysis of myeloid cells in chordoma samples. (A) Representative multiplex immunofluorescence image using the Akoya Biosciences platform showing CD163+ macrophages within fibrous septa, with tumor regions identified with brachyury staining. Nuclei are counterstained with DAPI. Scale bar: 200 μm. (B) Quantification of percentages of CD163+ macrophages within tumor regions and fibrous septa. n = 35 chordomas. Statistical significance determined by paired t-test (P < .0001). (C) PHATE embedding of scRNA-seq data illustrating distinct myeloid cell subsets, including DCs, macrophages, monocytes, mast cells, and plasmacytoid DCs (pDCs). (D) PHATE embedding showing the distribution of myeloid cells according to the sample of origin (tumor or PBMC). (E) Stacked bar plots showing the composition of myeloid cell subsets across tumor and PBMC compartments for individual patients, highlighting patient-specific variability in cell populations. (F) WGCNA dendrogram of co-expressed gene modules in myeloid cells, with module colors assigned to clusters of genes with similar expression patterns. (G) Dot plot showing the correlation of module eigengenes with specific myeloid cell types. (H) Visualization of enrichment scores for hub genes of the pink, turquoise, red, and blue modules, showing their activity across myeloid cell populations in PHATE embeddings. (I) GSEA of module eigengenes, identifying biological pathways associated with each module. (J) Dot plot displaying key hub genes with the highest kME values for each module. CD163+ tumor-associated macrophages and DCs expressed antigen presentation genes (HLA-DPA1, HLA-DPB1, and HLA-DQA1)), complement-related genes (C1QC, C1QA, C1QB), and inflammatory genes (FCN1, S100A8, and S100A9). Mast cells expressed tissue remodeling genes such as TPSAB1, CPA3, and TPSB2, while pDCs were characterized by immune surveillance markers (JCHAIN, GZMB, and MZB1).

To probe whether septal macrophages also differ functionally from those in the tumor core, we quantified co-­expression of CD205, CD206, and TIM-3 in CD163⁺ cells by multiplex immunofluorescence. While CD205 levels were similar between compartments, septal macrophages trended toward higher CD206 (P = .0688) and were significantly enriched for TIM-3 (P = .0461; Figure S4D), pointing to enhanced tissue-remodeling and immunoregulatory programs in these niches.

Next, our integrated scRNA-seq dataset of paired tumor and PBMC samples revealed clusters corresponding to DCs, macrophages, monocytes, mast cells, monocyte-macrophage intermediates (Mono-mac), and plasmacytoid DCs (pDCs) (Figure 2C and D, Figure S4A). Tumor samples consistently showed high macrophage and DC abundance (Figure 2E, Figure S4B). Mono-macs originated mostly from one tumor sample. PBMCs, by contrast, were enriched in monocytes, with few macrophages or mast cells (Figure S4C).

To identify groups of co-expressed genes, we performed WGCNA in the myeloid cell population (Figure 2F). For each group, we focused on the most representative hub genes, those most strongly connected within the group, and evaluated how enriched these genes were in each cell (Figure 2H). To focus on the most biologically relevant patterns, we selected modules that showed enrichment in at least 25% of myeloid cells. Modules were associated with specific myeloid cell types and activation states; for example, the black and turquoise modules were linked to dendritic cells and macrophages, while the yellow module was enriched in monocytes (Figure 2G).

Gene set enrichment analysis (GSEA) of each module’s hub genes (Figure 2I) indicated that the black module was enriched for antigen presentation and adaptive immune responses in DCs and macrophages, while the turquoise module was associated with complement activation and immune response. These specialized functional programs among myeloid subsets are highlighted in Figure 2J. Macrophages and especially DCs showed strong expression of antigen-presentation genes (HLA-DPA1, HLA-DQA1, HLA-DPB1). Macrophages and DCs also expressed ­complement-related genes (C1QC, C1QA, C1QB), suggesting roles in immune regulation, while monocytes were marked by inflammatory mediators (FCN1, S100A8, S100A9).

Given that myeloid cells can both present antigen and inhibit T cells, we asked whether genes involved in antigen presentation and inhibition co-occur in the same cells. We found that less than 50% of DCs, and to a lesser extent, macrophages expressed high levels of HAVCR2 (TIM-3) and MRC1 (CD206) alongside MHC-II genes. In contrast, LY75 (CD205), IDO1, and CD274 (PD-L1) are each expressed in only a small fraction of myeloid cells (Figure S4E). This co-expression points to immunoregulatory hubs that may limit anti-tumor responses.

Overall, this analysis underscores the pivotal roles of chordoma-associated myeloid cells, revealing their functional specialization and diverse contributions to tumor immunity.

Spatial Localization and Functional Heterogeneity of CD8+ T Cells Highlight Exhaustion and Immune Dysregulation in Chordomas

Given T cells’ central role in antitumor immunity, we next characterized their diversity, localization, and functional states in both tumor and peripheral compartments. Multiplex immunofluorescence showed significantly higher percentages of CD3+ T cells (both cytotoxic and helper) in fibrous septa compared to tumor regions (Figure 3A, Figure S5A). Quantitative analysis confirmed these findings, with elevated CD3+ T cell percentages in fibrous septa (P < .0001, t-test; Figure 3B, Figure 5B). This distribution parallels that of myeloid cells, suggesting that structural TME features strongly influence T cell localization.

Figure 3.

Spatial localization, heterogeneity, and functional analysis of CD8+ T cells in chordoma samples. (A) Representative multiplex immunofluorescence image using the Akoya Biosciences platform showing CD3+ T cells and CD8+ T cells within chordoma samples, with tumor regions identified by brachyury staining. Nuclei are counterstained with DAPI. Scale bar: 200 μm. (B) Quantification of CD3+ T cells, including CD3+CD8+ (cytotoxic T cells) and CD3+CD8− (helper T cells), in fibrous septa and tumor regions. n = 35 chordomas. Statistical significance determined by paired t-test (P < .0001). (C) PHATE embedding of scRNA-seq data showing distinct CD8+ T cell subsets, including naive, central memory, effector memory, early effector, cytotoxic effector, MAIT, and exhausted cells. (D) PHATE embedding showing the distribution of CD8+ T cell subsets according to sample of origin (tumor or PBMC). (E) Expression of exhaustion markers (TOX, PDCD1 (PD-1), and HAVCR2 (TIM-3) across CD8+ T cell subsets, visualized on PHATE embeddings. (F) Proportion of CD8+ T cell subsets across tumor and PBMC compartments, stratified by individual patients. (G) WGCNA identifying modules of co-expressed genes associated with CD8+ T cell subsets. The turquoise module correlates with cytotoxic effector and exhausted T cells, the brown module with effector memory cells, and the yellow module with naive and central memory cells. (H) Enrichment scores for module hub genes, visualized on PHATE embeddings, highlighting module activity across CD8+ T cell subsets. (I) GSEA of the turquoise module, identifying pathways related to intracellular receptor signaling, negative regulation of the MAPK cascade, and cytokine production. (J) Expression of key hub genes within each module. The turquoise module is enriched for NR4A2, CXCR4, and FOSB in exhausted and cytotoxic effector subsets, while the yellow module includes CCR7 and LEF1, associated with naive and central memory T cells.

Figure 5.

Spatial interactions and inferred cell-cell communication in chordoma samples. (A) Schematic representation of spatial analyses. Distances of immune cell populations to CD3+CD8+ T cells, calculated iteratively for k = 1 to k = 40, where k represents the k-th nearest neighbor. (B) Density plot showing the clustering of CD163+ macrophages, CD3+CD8− T cells, and CD20+ B cells to CD3+CD8+ T cells around CD3+CD8+ T cells calculated based on k-nearest neighbors (k = 1 to 40) across all tumor samples (n = 35). Colored lines represent distinct cell populations, indicating the density of cells surrounding CD3+CD8+ T cells. (C) Representative multiplex immunofluorescence image using the Akoya Biosciences platform illustrating immune cell localization in chordoma tissues. CD3+ T cells (cyan), CD8+ T cells (red), CD163+ macrophages (yellow), and CD20+ B cells (green) are shown relative to tumor (brachyury+, magenta) and DAPI (blue). Scale bar: 100 μm. (D-G) Density plot showing the clustering of immune cells (CD163+ macrophages (D), CD20+ B cells (E), CD3+CD8− T cells (F), and CD3+CD8+ T cells (G)) around CD3+CD8+ T cells is shown for tumor regions (blue) and fibrous septa (pink). Lines represent the average number of cells within increasing kk-nearest neighbors, with shaded areas indicating the 95% CI. (H) Network analysis of inferred cell-cell interactions derived from scRNA-seq data. Nodes represent immune and stromal cell populations, with edge thickness corresponding to interaction strength. Three major components are identified: Immune Regulatory, TAM & Tumor, and Inflammatory components. (I) Breakdown of cell type contributions within the Immune Regulatory Component, highlighting dendritic cells, exhausted CD8+ T cells, and mast cells. (J) Cell type composition of the TAM & Tumor Component, dominated by tumor-associated macrophages (TAMs), stromal cells, and CD8+ effector T cells. (K) Composition of the Inflammatory Component, including matrix-remodeling TAMs, CD4/CD8 effector cells, and NK cells.

scRNA-seq further revealed the heterogeneity of T cells in tumor and PBMC samples (Figure S4C and D). We identified naive, central memory, effector memory, early effector, cytotoxic effector, MAIT (mucosal-associated invariant T), and exhausted CD8+ T cells (Figure 3C). Tumor-resident T cells were enriched for exhausted and effector subsets, while PBMCs showed higher proportions of naive and central memory T cells (Figure 3D). Tumor-infiltrating CD8+ T cells prominently expressed TOX, PDCD1 (PD-1), and HAVCR2 (TIM-3) (Figure 5E), reflecting their dysfunctional phenotype. Conversely, effector memory and cytotoxic effector cells expressed cytotoxic genes (GZMB, GZMA, GZMH, and PRF1) (Figure S5E). The relative abundance of these subsets varied among patients in both tumor and PBMC compartments (Figure 3F, Figure 5F and G).

Pseudotime analysis (Figure S6A and B) traced a continuum from naive and central memory T cells in PBMCs through effector states to exhaustion in the TME, marked by increasing TOX, PDCD1, and HAVCR2 along this trajectory (Figure S6C and D). Weighted gene co-expression network analysis identified co-expressed gene modules in CD8+ T cells (Figure 3G). Calculating module enrichment scores with genes showing positive kME values quantified these modules’ activity (Figure 3H). The turquoise module was strongly associated with exhausted and cytotoxic effector T cells, while the brown module characterized effector memory and cytotoxic effector subsets. The yellow module was linked to central memory T cells, and the green module to naive T cells.

Functional enrichment (GSEA) revealed significant pathways only in the turquoise module (Figure 3I), highlighting its importance in CD8+ T cell function. Key hub genes within this module (NR4A2, CXCR4, FOSB) were predominantly expressed in exhausted, cytotoxic effector, and early effector T cells, implicating them in T cell differentiation and dysfunction (Figure 3J). NR4A2 regulates T cell exhaustion, CXCR4 mediates T cell trafficking, and FOSB can influence activation and exhaustion. Effector memory T cells showed high expression of NKG7, FGFBP2, and GZMH, preserving cytotoxic potential. Naive and central memory T cells expressed CCR7 and LEF1, reflecting proliferative and migratory capacities. Overall, exhausted and cytotoxic effectors prevailed within the tumor, indicating potential mechanisms of immune evasion in chordomas.

CD8+ TCR Analysis Reveals Clonal Relationships and Peripheral Anti-Tumor Reactivity

To assess CD8+ T cell binding specificity and functional diversity in tumor and peripheral compartments (Figure 4A), we used the Grouping of Lymphocyte Interactions by Paratope Hotspots 2 (GLIPH2) algorithm.¹⁸ GLIPH2 identified shared TCR binding motifs in the CDR3 region, revealing that many TCRs from exhausted tumor-infiltrating CD8+ T cells shared motifs with peripheral effector memory, MAIT, and, to a lesser extent, cytotoxic effector T cells (Figure 4B). This pattern suggests a clonal relationship stemming from chronic antigen stimulation in the TME, with peripheral cells possibly retaining anti-tumor reactivity post-resection.

Figure 4.

T cell receptor analysis of CD8+ T cells in chordomas. (A) Percentage of shared TCRs between exhausted CD8+ T cells and other CD8+ T cell subsets in the tumor and PBMC compartments. (B) PHATE embedding of CD8+ T cells depicting TCR clonality. (C) Predicted amino acid motifs derived from the CDR3 regions of TCR sequences, comparing motif frequencies between tumor and PBMC compartments. (D) TCR clone frequency across CD8+ T cell subsets in tumor and PBMC compartments.

We calculated a clonality score based on TCR cluster diversity and cell count. PHATE visualization indicated higher clonality in effector memory and MAIT cells, with particularly elevated clonality among tumor-resident exhausted T cells (Figure 4C). This restricted TCR repertoire likely reflects continuous antigen exposure and tumor-specific recognition. Amino acid motif analysis showed distinct motif frequencies between tumor-resident and peripheral TCRs (Figure 4D). Tumor-resident T cells displayed enriched motifs, consistent with antigen-driven selection, whereas PBMCs exhibited broader motif diversity.

Analysis of TCR clone frequencies showed that exhausted, early effector, and cytotoxic effector T cells made up the most expanded clones in tumor samples, while naive and memory subsets predominated in PBMCs (Figure 4E). These findings underscore tumor-specific antigen recognition within the TME, as evidenced by shared motifs and constrained clonality in exhausted T cells.

Spatial Relationships and Cell-Cell Interactions in the Chordoma TME

To investigate spatial interactions and cellular organization in chordoma, we combined multiplex immunofluorescence and scRNA-seq data to analyze immune cell localization and potential interactions within the TME (Figure 5A). We applied a k-nearest neighbor approach, examining up to 40 neighbors around CD3+CD8+ T cells to capture immune cell density in their immediate vicinity. CD20+ B cells and T cell subsets (CD3+CD8−, CD3+CD8+) peaked closer to CD3+CD8+ T cells, suggesting potential direct interactions. In contrast, macrophages showed clustering peaks at greater distances, indicating a more diffuse distribution (Figure 5B).

Spatial comparisons revealed significant differences between tumor regions and fibrous septa (Figure 5C). Notably, CD163+ macrophages (Figure 5D), CD20+ B cells (Figure 5E), CD3+CD8− T cells (Figure 5F), and CD3+CD8+ T cells (Figure 5G) clustered more densely around CD3+CD8+ T cells in fibrous septa, underscoring the role of tumor architecture in promoting immune cell aggregation.

To complement these spatial findings, we performed a nested block model analysis of scRNA-seq data (Figure 5F), identifying three major components of cell-cell interactions: the Immune Regulatory Component, the TAM & Tumor Component, and the Inflammatory Component. The Inflammatory Component, comprising Matrix Remodeling TAMs, CD4/CD8 effector cells, and NK cells, demonstrated a dual role in antitumor immunity and tissue-destructive processes (Figure 5G). The TAM & Tumor Component included TAMs, stromal cells, and effector CD8+ T cells, highlighting key macrophage-effector interactions within the tumor niche (Figure 5H). The Immune Regulatory Component featured DCs, exhausted CD8+ T cells, and mast cells, reflecting an interplay of immune suppression and antigen presentation (Figure 5I).

Overall, these data highlight the complexity of the TME, where immune activation, suppression, and tumor-supportive niches coexist. The predominance of suppressive and tumor-supportive interactions likely contributes to immune evasion and tumor progression in chordoma.

Discussion

Chordomas, being rare and clinically challenging malignancies, have historically been difficult to study owing to their unique histological and molecular profiles. While prior studies indicate that chordomas often harbor more active immune microenvironments than other sarcomas,^6,20 they nevertheless exhibit considerable heterogeneity. Using hierarchical clustering as a visualization aid, we identified three characteristic patterns of immune composition. One pattern is marked by pronounced infiltration of both myeloid and lymphoid cells, reflecting an immunologically active tumor microenvironment; another pattern is largely devoid of immune cells, resembling an “immune-desert” phenotype; and a third pattern shows moderate, predominantly myeloid infiltration. These observations build on previous reports, which identified chordomas with both high and low immune activity,^6,21 and extend them by integrating T cells, B cells, and macrophages for a more complete depiction of immune infiltration patterns. Larger single-cell datasets, balanced across immune cell infiltration profiles and complemented by spatial transcriptomics, will be essential to uncover distinct immune cell recruitment patterns within each tumor microenvironment.

By examining both tumor and peripheral blood compartments, this study underscores the immunological heterogeneity and systemic interplay that characterize the chordoma immune milieu. A key finding was the predominance of myeloid cells, particularly macrophages and dendritic cells, within chordoma immune infiltrates, corroborating prior research emphasizing the vital influence of macrophages on the chordoma TME.^13,15 Although macrophages are classically regarded as immunosuppressive drivers of immune evasion, our analyses reveal a more nuanced role. We characterized TAM heterogeneity based on function-specific transcriptional programs derived from co-expression network analysis. Distinct TAM subsets exhibited gene expression profiles reflective of both antitumorigenic and protumorigenic functions. For example, TAMs enriched in the black module expressed genes associated with antigen presentation (HLA-DPA1, HLA-DPB1, HLA-DQA1), suggesting a capacity to support T cell activation. In contrast, the turquoise module was enriched in complement pathway genes (C1QA, C1QB, C1QC), often associated with immune regulation and tissue remodeling. This duality underscores the inherent plasticity of macrophages and dendritic cells, which appear to switch between immune suppression and stimulation depending on local signals such as cytokines, extracellular matrix (ECM) interactions, and crosstalk with other immune cells. Dissecting this balance between antitumor and protumor TAM programs offers a framework for developing therapeutic strategies that modulate macrophage function to favor immune-mediated tumor control in chordomas.

The spatial enrichment of CD163+ macrophages in fibrous septa, rather than central tumor regions, emphasizes how structural features of the TME govern the localization and functional specialization of macrophages. Fibrous septa seemingly act as focal points for immune cell recruitment, retention, and antigen presentation, encouraging close interactions among myeloid cells, T cells, and other immune subsets. This observation parallels evidence from other tumor types, where spatial immune cell arrangements shape functional outputs.^22–24 In chordomas, fibrous septa may support specialized immunologic niches that facilitate myeloid cell activity. Given that the ECM, a hallmark of sarcomas, including chordomas, is known to affect immune cell migration and function,²⁵ our data imply that fibrous septa could serve as additional regulatory structures by hosting myeloid-T cell interactions with implications for both suppression and activation.

Examining the spatial organization of immune cells within chordomas provides valuable insights into the influence of structural tumor features on immune behavior. The clustering of CD3+CD8+ T cells alongside CD163+ macrophages, CD20+ B cells, and CD3+CD8– T cells in the fibrous septa suggests that these anatomical regions function as immune interaction hubs. Depending on local cytokine signals and cellular composition, these interactions may foster T cell activation or reinforce immunosuppression. Moreover, fibrous septa-restricted T cells may have limited direct engagement with tumor cells in central regions, hindering antitumor immune responses. Although we are currently underpowered to stratify functional programs by immune cluster using single-cell data, our spatial analysis highlights fibrous septa as likely sites of immune modulation, an observation that can be further validated through future integrative studies combining spatial and single-cell transcriptomics. Thus, the physical architecture of chordomas emerges as a pivotal factor for devising immunotherapies that target specific immune subsets.

Our data also reveal marked functional heterogeneity within the CD8+ T cell compartment in chordomas, pinpointing transcriptional programs linked with exhaustion, cytotoxicity, and memory. Key hub genes—NR4A2, CXCR4, and FOSB—play critical roles in T cell differentiation and dysfunction.²⁶ Among these, NR4A2 and CXCR4 emerge as central regulators of the shift between functional and exhausted T cell states. NR4A2, a transcription factor known to act in concert with TOX to drive T cell exhaustion,²⁶ can heighten inhibitory receptor expression and curtail cytokine production, thus sustaining a chronically stimulated state. Studies in murine models indicate that disrupting NR4A2 boosts T cell stemness and diminishes exhaustion, improving tumor clearance.²⁷ Consequently, NR4A inhibition holds promise as a means to restore cytotoxic T cell function in chordomas, mirroring its potential in other solid tumors.²⁸ Similarly, CXCR4 has been recognized as a mediator of both T cell migration and exhaustion.²⁹ In PD-1^{^high} exhausted T cells, CXCR4 blockade has shown synergy with PD-1 inhibition, restoring cytotoxic function by enhancing T cell trafficking and preventing the functional-to-exhausted transition via JAK2–STAT3 modulation. Our discovery of CXCR4 as a major hub gene in chordoma-infiltrating CD8+ T cells suggests that similar therapeutic strategies could curb the emergence of exhausted populations and improve T cell access to tumor regions. Notably, while exhaustion signatures prevail in many tumor-infiltrating T cells, a subset retains cytotoxic markers (NKG7, FGFBP2, GZMH), signaling residual effector potential. Furthermore, CCR7 and LEF1 expression in naïve and central memory T cells suggests that a fraction of tumor-infiltrating T cells may preserve proliferative and migratory capacities, which could be exploited in immunotherapeutic settings. The coexistence of exhausted and cytotoxic T cells underscores the dynamic, immunosuppressive nature of chordomas and highlights potential strategies to prevent T cell exhaustion while preserving effector function.

Additionally, TCR clonality analysis unveiled overlapping motifs between exhausted tumor-infiltrating T cells and peripheral effector memory subsets, indicating systemic connectivity in tumor-specific immune responses. This relationship suggests that circulating T cells could serve as reservoirs for tumor-reactive clones, potentially useful for adoptive T cell therapies or other immunotherapeutic interventions. Overall, these data illuminate diverse T cell states in chordomas and point to multiple opportunities to modulate T cell exhaustion without compromising their cytolytic capabilities.

We identified 3 principal immune interaction components within chordomas: Immune Regulatory, TAM & Tumor, and Inflammatory. These encompass a wide array of immune functions, from suppression and antigen presentation to tissue-destructive inflammation. Previous research underlines the significance of TGF-β signaling in chordomas, particularly regarding its influence on interactions among fibroblasts, macrophages, tumor cells, and CD4+ T cells.^13,14 Our findings expand on this model by illustrating how fibrous septa can serve as hot spots for immune modulation through myeloid-T cell engagements.

Taken together, these results carry considerable implications for immunotherapy in chordomas. The detection of antigen-presenting pathways in TAMs and distinct TCR motifs that may be tumor-specific supports the development of interventions aimed at heightening antigen presentation and activating T cell–mediated responses. Furthermore, the evidence of systemic connectivity between tumor and peripheral immune compartments underscores the potential of leveraging peripheral blood for adoptive T cell strategies, combined with therapies that alleviate T cell exhaustion.

In conclusion, this study provides an in-depth analysis of the chordoma TME, highlighting the multifaceted roles of macrophages, the systemic nature of T cell responses, and the spatial orchestration of immune cells. By depicting how immune activation and suppression coexist in chordomas, our findings advance the current understanding of this rare malignancy and offer a springboard for targeted immunotherapeutic approaches.

Supplementary Material

Supplementary material is available online at Neuro-­Oncology (https://academic.oup.com/neuro-oncology).

Author Contributions

Conceptualization: J.P.W., D.H.H., and C.L.C. Methodology: V.A.A., J.K.B., J.P.W., D.H.H., and C.L.C. Investigation: V.A.A., J.K.B., R.D., K.J.H., J.Z., H.N., S.W., D.H., J.L.K., G.I.V., B.C., N.K., P.J., C.M.H., A.M.S., R.S., A.B.M., M.S.L., J.P.C., J.P.W., D.H.H., and C.L.C. Resources: E.V.N., L.S.S., L.J., C.M.H., A.B.H., M.S.L., J.M.M., J.P.W., and C.L.C. Data Curation: V.A.A., R.D., and D.H.H., Writing—original draft preparation: V.A.A., J.K.B., R.B., D.H.H., and C.L.C. Writing—review and editing: V.A.A., J.K.B., R.D., G.I.V., J.P.C., D.H.H., and C.L.C. Visualization: V.A.A., J.K.B., K.J.H., H.N., M.E.C., and S.P. Supervision: J.P.W., D.H.H., C.L.C. Project administration: Y.H. and C.L.C. Funding acquisition: J.P.W. and C.L.C.

Conflict of Interest Statement

All the authors declare that they have no competing interests.

Funding

This work is supported by the Chordoma Foundation, the Cancer Research Institute (grant #CR13733), and the Lou and Jean Malnati Brain Tumor Institute from Northwestern University. C.L.C. is supported by the National Cancer Institute (NCI, R37CA258426, P50CA221747) and the Cancer Research Institute (CR68036). V.A.A. is supported by the Society for Immunotherapy of Cancer-Diversity, Equity, and Inclusion in Cancer Immunotherapy Fellowship.

Ethics Approval

The study protocol was approved by the Institutional Review Board at Northwestern University. Informed consent was obtained from each patient.

Data Availability

The single-cell RNA sequencing dataset generated and analyzed in this study are available in the Figshare repository under the accession https://doi.org/10.6084/m9.figshare.30326650.v1. All other data supporting the findings of this study are available within the article and its supplementary information or from the corresponding author upon reasonable request.