Spatially resolved single-cell landscape of tumor immunotypes reveals the central role of interferon signaling and plasmacytoid dendritic cells in triple-negative breast cancer.

Spatially resolved single-cell landscape of tumor immunotypes reveals the central role of interferon signaling and plasmacytoid dendritic cells in triple-negative breast cancer - PMC Skip to main content An official website of the United States government Here's how you know Here's how you know Search Log in Primary site navigation Search Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice J Immunother Cancer . 2026 Jun 3;14(6):e014020. doi: 10.1136/jitc-2025-014020 Spatially resolved single-cell landscape of tumor immunotypes reveals the central role of interferon signaling and plasmacytoid dendritic cells in triple-negative breast cancer Yi Liu 1, 2, *, 0 , Jodi M Carter 3, 0 , Yaohua Ma 1 , Sachin K Deshmukh 4 , George W Sledge 4 , Justin M Balko 5 , Jennifer M Kachergus 2 , Ji Shi 2 , Mark Gregory 6 , Jingjing Gong 6 , Heikki Joensuu 7, 8 , Edith A Perez 9 , Daniel Stover 10 , Lisa Carey 11 , William M Sikov 12, 13 , Roberto Leon-Ferre 14 , Keith L Knutson 15 , Judy C Boughey 16 , James N Ingle 14 , Krishna R Kalari 17 , Fergus J Couch 18 , Yan W Asmann 1 , Matthew P Goetz 14 , E Aubrey Thompson 2, 9, 0 , Saranya Chumsri 19, ✉, 0 Author information Article notes Copyright and License information 1 Department of Quantitative Health Sciences, Mayo Clinic in Florida, Jacksonville, Florida, USA 2 Department of Pioneering Science, Office of Core Shared Services, Mayo Clinic in Florida, Jacksonville, Florida, USA 3 Department of Laboratory Medicine and Pathology, University of Alberta, Edmonton, Alberta, Canada 4 Caris Life Sciences Inc Phoenix, Phoenix, Arizona, USA 5 Department of Medicine, Vanderbilt University Medical Center, Nashville, Tennessee, USA 6 NanoString Technologies Inc, Seattle, Washington, USA 7 Department of Oncology, HUS Helsinki University Hospital, Helsinki, Finland 8 University of Helsinki, Helsinki, Finland 9 Department of Cancer Biology, Mayo Clinic in Florida, Jacksonville, Florida, USA 10 The Ohio State University Comprehensive Cancer Center Arthur G James Cancer Hospital and Richard J Solove Research Institute, Columbus, Ohio, USA 11 UNC School of Medicine, Chapel Hill, North Carolina, USA 12 Department of Medicine, Brown University Warren Alpert Medical School, Providence, Rhode Island, USA 13 Women & Infants Hospital of Rhode Island, Providence, Rhode Island, USA 14 Department of Oncology, Mayo Clinic Rochester, Rochester, Minnesota, USA 15 Department of Immunology, Mayo Clinic in Florida, Jacksonville, Florida, USA 16 Department of Surgery, Mayo Clinic Rochester, Rochester, Minnesota, USA 17 Department of Quantitative Health Sciences, Mayo Clinic Rochester, Rochester, Minnesota, USA 18 Department of Laboratory Medicine and Pathology, Mayo Clinic Rochester, Rochester, Minnesota, USA 19 Department of Hematology and Oncology, Mayo Clinic in Florida, Jacksonville, Florida, USA Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant. the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names, and drug dosages), and is not responsible for any errors and/or omissions arising from translation and adaptation or otherwise. Additional supplemental material is published online only. To view, please visit the journal online ( https://doi.org/10.1136/jitc-2025-014020 ). There is no relevant conflict of interest for SC, who receives research funding for her institution from Merck & Co., Pfizer, Salix Pharmaceuticals, and Rebiotix. She receives consulting fees to her institution from AstraZeneca, Daiichi Sankyo, Immunomedics, Biotheranostics, Novartis, Athenex, Syndax, Puma Biotechnology, and Eisai. RL-F reports receiving consulting fees paid to his institution from Gilead Sciences, AstraZeneca, and Lyell Immunopharma, outside the scope of this work. JMC reports consulting fees for advisory board participation from AstraZeneca, Agilent, and Roche outside the scope of this work. JCB receives research support paid to her institution from Eli Lilly and SymBioSis and is on a DSMB for CairnsSurgical. She has received honoraria for speaking for PER, PeerView, EndoMag, and contributed a chapter to UpToDate. MG and JG were employed by NanoString/Bruker at the time these experiments were performed. HJ has a consulting/advisory role in Orion Pharma, Neutron Therapeutics, and Maud Kuistila Foundation, has received research grants from Mersana Therapeutics and Defense Therapeutics, and has stock ownership in Orion Pharma and Sartar Therapeutics. MPG declares personal fees for CME activities from Medscape, MJH Life Sciences, and Research to Practice, personal fees for serving as a panelist for a panel discussion for Total Health Conferencing, and for serving as a moderator for Curio Science, consulting fees to Mayo Clinic from ARC Therapeutics Advisory Board, AstraZeneca, Biotheranostics Advisory Board, Biotheryx Advisory Board, Blueprint Medicines Advisory Board, Engage Health Media Roundtable interview for Expert Perspectives, Lilly General Consulting, Laekna Advisory Board, Novartis Advisory Board, RNA Diagnostics Clinical Advisory Board, Seattle Genetics Consulting, Sermonix Steering and Translational Committees, TerSera Therapeutics/Ampity Health Steering Committee (OFS Summit), grant funding to Mayo Clinic from AstraZeneca, ATOSSA Therapeutics, Lilly, Loxo, Pfizer, Sermonix, and travel support from Lilly (ESMO 10/2023). LC reports research funding from NanoString/Bruker, Seattle Genetics, Veracyte, Gilead, Novartis, Lilly, Genentech/Roche, and AstraZeneca with uncompensated relationships with Lilly, SeaGen, Novartis, Genentech/Roche, and GSK. All other authors declare no competing interests. ✉ Dr Saranya Chumsri; chumsri.saranya@mayo.edu * Dr Yi Liu; Liu.Yi1@mayo.edu 0 YL, JMC, EAT and SC contributed equally. Received 2025 Oct 23; Accepted 2026 Apr 20; Collection date 2026. Copyright © Author(s) (or their employer(s)) 2026. Re-use permitted under CC BY-NC. No commercial re-use. See rights and permissions. Published by BMJ Group. This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See https://creativecommons.org/licenses/by-nc/4.0/ . PMC Copyright notice PMCID: PMC13239575  PMID: 42236117 Abstract Background Tumor-infiltrating lymphocytes (TILs) are established determinants of clinical outcomes in breast cancer. 1–6 While stromal TILs (sTILs) are commonly evaluated due to easier scoring and higher interobserver reproducibility, intraepithelial TILs may provide additional insights into anti-tumor immunity. 7 We aimed to comprehensively characterize immunotype-associated cellular landscapes in triple-negative breast cancer (TNBC) and identify features associated with clinical outcomes. Methods 212 TNBC tumors were classified into three immunotypes based on immune infiltrate location and density: immune desert (ID), immune excluded (IE), and immune inflamed (IN). 8 Single-cell spatial transcriptomics was performed on 75 tumors to define the cellular composition and functional states of the tumor microenvironment across immunotypes. Findings were validated and integrated with bulk RNA sequencing data and clinical outcome datasets from FinXX, CALGB-40603, I-SPY 2, and Real-World Clinico-Genomic Data (RWCGD) cohorts. Results Patients with IN tumors had significantly improved outcomes compared with ID tumors. Despite high sTILs, IE tumors showed poor outcomes similar to ID tumors. Single-cell spatial analysis revealed that ID and IE tumors exhibited reduced major histocompatibility complex I/II expression and fewer tumor-resident plasmacytoid dendritic cells (pDCs) compared with IN tumors, which were enriched for interferon-alpha (IFNα) and interferon-gamma (IFNγ) responses. High IFN response scores were associated with favorable outcomes across multiple therapy types in independent datasets. Deconvolution of RWCGD bulk RNA-seq data confirmed that pDC abundance correlated with improved outcomes specifically in hormone receptor–negative subtypes. Conclusions Our study highlights pDCs and IFN signaling as hallmarks of effective anti-tumor immunity in TNBC. Immunotype-based profiling underscores the limited prognostic value of sTILs alone in immune-excluded tumors and supports pDCs and IFN pathways as potential biomarkers for prognosis and therapeutic development in TNBC. Keywords: Biomarker, Breast Cancer, Major histocompatibility complex - MHC, Tumor infiltrating lymphocyte - TIL, Immune modulatory WHAT IS ALREADY KNOWN ON THIS TOPIC Stromal tumor-infiltrating lymphocytes (sTILs) are established prognostic biomarkers in triple-negative breast cancer (TNBC). However, TIL density alone may not fully explain the marked heterogeneity in outcomes observed. Emerging evidence from other tumor types indicates that the spatial organization of immune infiltrates critically influences anti-tumor immunity, yet robust outcome-correlated spatial data in breast cancer remain limited. Moreover, the biological mechanisms that define and sustain distinct immunotypes in TNBC are incompletely understood. WHAT THIS STUDY ADDS Using high-resolution single-cell spatial transcriptomics integrated with multi-cohort clinical validation, this study demonstrates that spatial localization, rather than TIL quantity alone, defines prognostically distinct immunotypes in TNBC. We identify a subset of patients with immune-excluded tumors who exhibit poor outcomes despite high sTILs, underscoring the limitations of sTIL assessment alone. Furthermore, we establish tumor-resident plasmacytoid dendritic cells (pDCs) as central regulators of the immune-inflamed phenotype through coordinated activation of antigen presentation machinery and type I/II interferon signaling. The pDC-interferon axis is reproducibly associated with favorable outcomes across multiple independent TNBC cohorts, including contemporary datasets, and shows subtype specificity, conferring benefit only in hormone receptor-negative disease but not in hormone receptor-positive or HER2-positive tumors. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY These findings challenge the current reliance on sTIL quantification as a standalone immune biomarker in TNBC and support the incorporation of spatial immune architecture into prognostic frameworks. Immunotype classification combined with interferon gene signatures refines risk stratification and may improve prediction of therapeutic response, including immunotherapy. By identifying pDCs and interferon signaling as hallmarks of effective anti-tumor immunity specifically in TNBC, this study provides a strong biological rationale for the development of spatially informed biomarkers and for therapeutic strategies aimed at modulating the pDC-interferon axis. Introduction Despite recent therapeutic advances, triple-negative breast cancer (TNBC) remains one of the most challenging breast cancers to treat due to its limited molecular targets and its intrinsic tumorous heterogeneity. 1 In TNBC, tumor-infiltrating lymphocytes (TILs) are an instrumental component of anti-tumor immunity. 2 3 4 5 In TNBC, stromal TILs (sTILs) are prognostic in both neoadjuvant and adjuvant settings. 6 , 11 High sTILs are associated with pathological complete response (pCR) in patients who received neoadjuvant chemotherapy 11 12 as well as an immune checkpoint inhibitor in combination with chemotherapy. 13 14 More recently, higher sTILs scores in TNBC have been associated with better patient outcomes even in the absence of systemic adjuvant chemotherapy. 10 15 TILs encompass a diverse population of immune cells, and their dynamic interplay with tumor cells underlies the immunoarchitecture of TNBC. Spatially resolved single-cell analysis provides unprecedented insights into the complex interactions between immune cells and the tumor microenvironment. 16 This innovative technology allows interrogation of not only the quantification of phenotypic diversity but also the functional and activation status of individual cells, and “nearest neighbor” investigations while preserving the spatial context within tissues. In TNBC, single-cell spatial approaches enable detailed characterization of both intraepithelial TILs (iTILs) and stromal tumor microenvironments. However, quantifying iTILs is challenging and less reproducible, 7 9 which has led most studies to focus primarily on sTILs. 7 9 This approach offers only limited insight into the immune architecture within the tumor parenchyma. Emerging evidence suggests that not only the quantity but also the spatial distribution of immune cells is critical. Three distinct tumorous immunotypes have been described based on the quantity and location of immune infiltrates: immune inflamed (IN), immune excluded (IE), and immune desert (ID). 17 18 ID tumors contain minimal immune infiltrates, whereas IN tumors contain dense immune infiltration in both the tumor and stromal compartments. In contrast, immune infiltrates in IE tumors are “excluded” from the tumor parenchyma and thus confined to the peritumoral stroma. Although IE type tumors are as enriched in sTILs as those of the IN immunotype, patients with IE type tumors (urothelial, lung, and colon cancers) have worse outcomes. 19 , 22 These observations underscore the importance of spatial context when evaluating tumor immune microenvironments. Nevertheless, data regarding immunotype and clinical outcomes in breast cancer remain limited. To date, studies on mechanisms underlying immune defects in ID and IE tumors 23 , 25 have relied on bulk gene expression data from whole tumors, without spatial context. 22 Using spatially resolved multi-omics technology, the GeoMx Digital Spatial Profiling (DSP) platform, we previously described distinct spatial distributions of immune protein profiles in the tumor-immune microenvironment of PD-L1-positive tumors in an early-stage TNBC cohort. 26 Furthermore, we subsequently demonstrated that intratumoral, but not stromal or bulk RNA, expression of the antigen presentation-related proteins, HLA-DR, CD11c, and CD40, was associated with improved outcomes in TNBC, independent of sTILs in multivariate analysis. 27 Interferon (IFN) signaling, encompassing both type I (IFNα/β) and type II (IFNγ) responses, is a critical driver of anti-tumor immunity. IFN signaling enhances antigen presentation, promotes dendritic cell (DC) activation, and facilitates T-cell priming, all of which are essential for effective immune-mediated tumor clearance. 28 , 30 Plasmacytoid DCs (pDCs), as a primary source of type I IFNs, have been implicated in orchestrating tumor-directed immune responses, 31 32 yet their spatial distribution and clinical relevance in TNBC remain poorly defined. In this study, we used single-cell spatial transcriptomic profiling to characterize the cellular architecture and IFN-driven immune responses underlying distinct TNBC immunotypes. To validate the biological and clinical significance of our findings, we further integrated bulk RNA-sequencing and clinical outcome data from multiple independent datasets, including FinXX, CALGB-40603, I-SPY 2, and Real-World Clinico-Genomic Data (RWCGD). This comprehensive, multi-cohort approach allowed us to link spatially resolved immune features, particularly pDC abundance and IFN signaling activity—to patient outcomes, highlighting their potential as hallmarks of effective anti-tumor immunity in TNBC. Results Immunotypes stratify clinical outcomes in early-stage TNBC Using a tissue microarray (TMA) from the Mayo Clinic TNBC cohort and quantitative protein data generated by GeoMx DSP DSP, 26 27 33 212 TNBC tumors were classified into three immunotypes based on immune infiltrate density and spatial distribution: ID (n=116), IE (n=23), and IN (n=73). Representative images illustrating these immunotypes are shown in figure 1A , and clinicopathologic features are summarized in online supplemental table S1A . There were no significant differences in age, menopausal status, or nodal stage across immunotypes. Tumors classified as IN were more likely to exhibit high proliferative activity, as measured by Ki-67, and tumor size differed among the three groups ( online supplemental table S1B ). Survival analysis demonstrated clear stratification by immunotype. Patients with IN tumors had significantly improved recurrence-free survival (RFS) compared with those with ID tumors (HR 0.37, 95% CI 0.18 to 0.74, figure 1B ). Despite having higher sTILs, IE tumors exhibited RFS comparable to ID tumors and significantly worse than IN tumors (HR 0.84, 95% CI 0.36 to 1.98, figure 1B ). These findings indicate that immune infiltration alone is insufficient to confer favorable prognosis, and that spatial immune organization is a critical determinant of clinical outcome. Figure 1. Intratumoral immune architecture distinguishes IN tumors from ID and IE. ( A ) Representative GeoMx images illustrating the ID, IE and IN immunotypes. ( B ) Kaplan-Meier (KM) analysis of recurrence-free survival (RFS) by immunotype (ID=black, n=116; IE=blue, n=23; IN=orange, n=73). Log-rank p values are shown, and the number of patients at risk is indicated below each time point. (C–E) Forest plots of differential protein expression from DSP analysis in the intratumoral (panCK-positive, panCK + , or CKpos) compartment, comparing ID versus IE, ( C ), ID versus IN ( D ), and IE versus IN ( E ). Segment-level counts were averaged, and fold change (log₂FC) was estimated using a generalized linear model with a negative binomial distribution. Whisker bars indicate 95% CIs, with the square marking the mean log₂FC. Proteins with 95% CI not crossing zero (log₂FC=0) were considered significant (p<0.05). Significantly higher protein abundance is indicated as black for ID, blue for IE, and orange for IN. ( F ) Representative CosMx fields of view (FOV) for ID, IE, and IN samples. Left panels show the full FOV; right panels highlight the TN regions. (G, H) Distribution of 11 non-tumor cell types across immunotypes, summarized across samples ( G ) and shown per individual sample ( H ) in two distinct spatial neighborhoods (ID, n=22; IE, n=6; IN, n=28). DSP, Digital Spatial Profiling; EST, estimated; ID, immune desert; IE, immune excluded; IN, immune inflamed; mDC, myeloid dendritic cell; NK, natural killer; pDC, plasmacytoid DC; sTILs, stromal tumor-infiltrating lymphocytes; TMA, tissue microarray; TN, tumor neighborhood; TNBC, triple-negative breast cancer. Open in a new tab Intratumoral immune architecture distinguishes IN tumors from ID and IE To define the spatial immune features underlying these outcome differences, we analyzed multiplex protein expression profiles in intratumoral (panCK-positive, panCK + , or CKpos) and stromal (panCK-negative, panCK − , or CKneg) compartments using GeoMx DSP. 34 The average abundance of these proteins is given in online supplemental table S2A (panCK + ) and S2B (panCK − ). Differential expression analysis (Supplementary Data File S2) was carried out using a linear mixed model. While ID and IE subsets differed in sTILs abundance and stromal immune protein profiles, their intratumoral compartments were very similar ( figure 1C ). In contrast, the immune protein profile of the intratumoral compartment of IN tumors (orange) differed substantially from that of ID ( figure 1D ) and IE,( figure 1E ) tumors. Notably, these intratumoral differences were not paralleled by corresponding differences in the stromal compartment, indicating that stromal immune protein profiles do not necessarily reflect intratumoral immune states. These data reinforce prior observations that sTILs abundance and the stromal immune protein profile do not necessarily mirror intratumoral immune profiles. Specifically, intratumoral compartments of IN (which itself is defined by high intratumoral CD8 protein) were enriched in other T cell markers beyond CD8 markers (CD3, CD4) and T-cell activation markers (ICOS and Granzyme B), compared with ID and IE ( figure 1D,E , two-sided p<0.05 for all comparisons, Supplemental Data File S2). Immune checkpoint proteins (PD-L1, IDO-1, STING, Tim-3, and VISTA), and markers of other immune cells, CD20 (B cells), and CD68 (macrophages), were more abundant in IN; whereas CD56 (Natural Killer or NK cells) did not significantly differ ( figure 1D,E ). Corresponding to what we previously reported that intratumoral enrichment with antigen-presentation proteins HLA-DR, CD40, and CD11c is associated with reduced recurrence risk, 27 these three proteins were enriched in the intratumoral compartment of the IN samples compared with IE and ID ( figure 1D,E ). In ID samples, the stromal compartment exhibited elevated expression of tumor-promoting proteins, including B7-H3, CD34, fibronectin (FN), PanCK, and SMA. In contrast, expression of lymphocyte-associated proteins—such as B2M, CD3, CD40, CD44, CD45RO, ICOS, IDO1, PD-L1, and VISTA—was significantly reduced in the stroma of ID samples relative to non-ID samples ( online supplemental figure S1 ). To determine whether the intratumoral immune features distinguishing IN tumors from ID and IE were reproducible in an independent dataset, we evaluated TNBC samples from the FinXX trial. 35 Among 35 evaluable tumors classified as ID (N=15), IE (N=5), and IN (N=15), differential protein abundance patterns in both intratumoral and stromal compartments ( online supplemental table S2C,D ) closely mirrored those observed in the Mayo TNBC cohort ( online supplemental figure S2 ). Consistent with the primary cohort, IN tumors exhibited markedly higher expression of immune activation and antigen-presentation markers, including IDO1, HLA-DR, CD40, and CD11c, in the intratumoral compartment, with IDO1 showing approximately eight-fold higher expression in IN compared with ID tumors across both compartments. These findings support the robustness of intratumoral immune activation as a defining feature of the IN immunotype. One exception was intratumoral CD20, which did not show significant differential abundance between ID and IN tumors in the FinXX cohort, highlighting some cohort-specific variation while preserving the overall immunotype architecture. Spatially cellular composition of tumor neighborhoods differs by immunotype We used NanoString/Bruker CosMx Spatial Molecular Imaging (SMI) platform to quantify 960 RNA transcripts at single-cell resolution (CosMx 1K Prototype Panel, online supplemental table S3 ) with spatial context within user-defined fields of view (FOVs) to measure mRNA abundance, identify cell types, quantify cell numbers, and assess cellular activity in 75 TNBC TMA (1 FOV per Formalin-fixed, paraffin-embedded formalin-fixed, paraffin-embedded (FFPE) tissue core, totaling 75 FOVs) for which we also had multiplex proteomic data from the GeoMx DSP. Immunotypes were successfully assigned in 56 of 75 samples. No significant associations were observed between immunotype and menopausal status, age, Ki-67 group, or pathologic nodal stage ( online supplemental table S4 ). Using semi-supervised cell typing, we identified 12 cell types, including B cells, endothelial cells, fibroblasts, macrophages, myeloid DCs (mDCs), NK cells, pDCs, plasmablasts, T CD4 cells, T CD8 cells, regulatory T-cells (Treg), and tumor cells. The expression of selected cell type markers in these cell populations is shown in online supplemental figure S3A,B . UMAP of immunotypes is shown in online supplemental figure S3C . Our final Mayo TNBC TMA SMI dataset encompasses 230 815 CosMx single-cell transcript profiles from 22 ID samples, 6 IE samples, and 28 IN samples (as well as 19 NA samples) with comprehensive immune cell typing, cellular population distribution, cellular activity, and association between clinical outcomes and cell functionalities within the tumor and stroma spatial neighborhoods. Consistent with the immunotype classification from the protein-based GeoMx DSP data set, CosMx analysis revealed significantly fewer CD8 + T cells in ID samples compared with non-ID samples. Moreover, IE samples exhibited greater immune cell infiltration and enrichment of CD8 + T cells than ID samples, while IN samples demonstrated the highest levels of tumor infiltration and CD8 + T cell abundance ( figure 1F , left panel). Importantly, the CosMx data not only validated the immunotype-specific immune contexture observed with the GeoMx data set but also provided enhanced resolution by enabling cell-type identification and transcriptomic visualization at the single-cell level. Unlike GeoMx, which relies on PanCK-based compartmentalization, CosMx allows for precise spatial localization and molecular profiling of individual cells, offering a more comprehensive view of the tumor microenvironment. We further designated each individual cell into two compartments or neighborhoods based on their proximity to a tumor cell. Tumor neighborhood (TN) included any cells within 50 μm of a tumor cell, and stroma neighborhood (SN) composed of any cells that were not within 50 microns of a tumor cell. Representative CosMx images of TMA cores and TN are displayed in the right panel of figure 1F . The distribution of cell types in TN and SN across all samples is shown in online supplemental table S5 . As expected, tumor cells were the most abundant cell type within the TN (84.3% of all cells in TN), followed by fibroblasts (5.2%). In TN, macrophages were the most abundant immune cell type (3%), followed by T cells (2.1%), while plasmablasts were the most common immune cell type in the SN (25.1%), followed by T cells (14.2%). Notably, compared with TN, all immune cell types were more abundant in the SN, with the exception of macrophages. In immunotypes, plasmablasts, B cells, and T CD8 cells were enriched in IN and IE tumors compared with ID tumors in both TN and SN (Supplementary Data File S3). As expected, based on our method for classifying immunotypes, IN tumors had significantly more T cells in TN, including T CD4, T CD8, and Treg cells, compared with IE. Additionally, there were more macrophages in TN compared with IE and ID tumors. In contrast, ID tumors had significantly more fibroblasts and endothelial cells in both TN and SN compared with IE and ID tumors ( figure 1G,H and Supplementary Data File S3-S4). Together, single-cell spatial transcriptomic data indicates that the defining feature of the IN immunotype is robust intratumoral immune engagement, rather than stromal immune abundance, and highlights the importance of immune cell localization within the TN as a defining feature of immunotype. Tumor-intrinsic antigen presentation and interferon signaling define the IN phenotype We next examined tumor cell–intrinsic transcriptional programs associated with immunotype. Differential gene expression analysis of tumor cells within the TN revealed that IE tumors were transcriptionally similar to ID tumors and distinct from IN tumors ( figure 2A ). These results are similar to the protein expression data using DSP multiplex analysis described previously ( figure 1C ). Only RUNX3, a regulator of tissue-resident memory T CD8 cells, 36 was upregulated in IE compared with ID TNBC at Benjamini-Yekutieli adjusted (BY adj) p<0.05 (log2FC −0.341 compared with ID). In contrast, a comparison of tumor cell transcriptional profiles in the TN of ID vs IN samples identified 18 upregulated transcripts in IN (BY adj p<0.05) ( figure 2B ). Even though there were multiple transcripts that were differentially expressed more in IN tumors compared with IE at nominal p<0.05, none were significant at BY adj p<0.05 ( figure 2C ). There was a total of 19 genes that were differentially expressed in tumor cells within the TN across the three immunotypes (BY adj p<0.05, table 1 ). A heatmap comparing the average abundance of these 19 transcripts in ID, IE, and IN samples in the TN ( figure 2D , online supplemental figure S4A ) revealed a significant association of these genes with the major histocompatibility complex (MHC) class I and class II component expression in tumor cells in the three immunotypes. Differential expression of other MHC I and MHC II transcripts in tumor cells is shown in online supplemental figure S4B . Specifically, 13 out of 19 transcripts were involved in antigen presentation, including transcription factors CIITA and NLRC5, the MHC components HLA-A, HLA-B, HLA-C, HLA-DRA, HLA-DRB1, HLA-DPA1, HLA-DPB1, and HLA-E, the peptide antigen transporters TAP1 and TAP2, and the MHC II invariant subunit protein CD74 ( figure 2D ). While we have previously shown that intratumoral HLA-DR protein was associated with improved outcomes in TNBC 27 ; however, these data were generated with spatial DSP analysis (without single-cell resolution, we could not assign the HLA-DR protein to specific cell types). The differential expression among ID, IE, and IN samples in TN is shown in online supplemental table S6 . Figure 2. Tumor-intrinsic antigen presentation and interferon signaling define the IN phenotype. (A–C) Volcano plots of differential gene expression in tumor cells within the TN comparing ID versus IE ( A ), ID versus IN ( B ), and IE versus IN ( C ). Significantly higher transcript abundance is indicated as black for ID, blue for IE, and orange for IN. ( D ) Heatmap of 19 genes differentially expressed among the three immunotypes with BY-adjusted p<0.0001. ( E ) Gene set enrichment analysis (GSEA) of IN tumors compared with IE, ID, or the two combined in TN. Dot size indicates p value; dot color intensity reflects the FDR for each pathway. ( F ) Abundance of plasmacytoid dendritic cells (pDCs) per 100 tumor cells across immunotypes (ID=black, IE=blue, IN=orange) in TN. Group comparisons were performed using the Kruskal-Wallis one-way ANOVA on ranks. ANOVA, analysis of variance; FDR, false discovery rate; ID, immune desert; IE, immune excluded; IN, immune inflamed; TMA, tissue microarray; TN, tumor neighborhood; TNBC, triple-negative breast cancer. Open in a new tab Table 1. Significantly differential gene expression of tumor cells in tumor neighborhood from TNBC TMA SMI data set. Gene Tumor neighborhood ID vs IE ID vs IN IE vs IN Log2FC.95% CI P value Log2FC.95% CI P value Log2FC.95% CI P value CD74 −0.091 (−0.587 to 0.404) 0.721 −1.45 (−1.977 to −0.923) <0.0001 −1.359 (−2.312 to −0.406) 0.009 CIITA −0.027 (−0.163 to 0.108) 0.697 −0.323 (−0.476 to −0.17) <0.0001 −0.296 (−0.578 to −0.014) 0.048 CRYAB 0.448 (−0.221 to 1.118) 0.202 0.685 (0.358 to 1.013) <0.0001 0.237 (−0.166 to 0.641) 0.258 HLA_A −0.369 (−0.986 to 0.248) 0.253 −1.376 (−1.909 to −0.843) <0.0001 −1.007 (−1.912 to −0.102) 0.037 HLA_B −0.433 (−1.152 to 0.286) 0.25 −1.663 (−2.216 to −1.111) <0.0001 −1.23 (−2.169 to −0.291) 0.015 HLA_C −0.276 (−0.951 to 0.399) 0.431 −1.434 (−1.986 to −0.882) <0.0001 −1.158 (−2.132 to −0.184) 0.027 HLA_DPA1 0.045 (−0.118 to 0.207) 0.594 −0.644 (−0.91 to −0.379) <0.0001 −0.689 (−1.188 to −0.19) 0.011 HLA_DPB1 0.046 (−0.04 to 0.132) 0.305 −0.374 (−0.544 to −0.204) <0.0001 −0.42 (−0.743 to −0.098) 0.016 HLA_DRA −0.005 (−0.209 to 0.199) 0.964 −0.898 (−1.257 to −0.539) <0.0001 −0.893 (−1.575 to −0.212) 0.015 HLA_DRB1 −0.011 (−0.32 to 0.298) 0.946 −0.971 (−1.361 to −0.582) <0.0001 −0.961 (−1.685 to −0.236) 0.014 HLA_E 0.102 (−0.14 to 0.343) 0.418 −0.527 (−0.797 to −0.258) <0.0001 −0.629 (−1.112 to −0.146) 0.016 IGKC −0.386 (−0.862 to 0.091) 0.125 −0.931 (−1.408 to −0.454) <0.0001 −0.545 (−1.476 to 0.386) 0.26 ITGB1 0.305 (−0.125 to 0.735) 0.177 0.469 (0.238 to 0.701) <0.0001 0.165 (−0.126 to 0.455) 0.276 MGP 1.876 (0.8 to 2.952) 0.002 1.467 (0.749 to 2.184) <0.0001 −0.409 (−1.535 to 0.718) 0.483 NLRC5 0.021 (−0.119 to 0.161) 0.768 −0.256 (−0.39 to −0.123) <0.0001 −0.278 (−0.511 to −0.045) 0.026 SRGN 0.007 (−0.088 to 0.102) 0.893 −0.151 (−0.228 to −0.074) <0.0001 −0.158 (−0.301 to −0.015) 0.038 TAP1 0.009 (−0.138 to 0.157) 0.902 −0.492 (−0.703 to −0.281) <0.0001 −0.502 (−0.9 to −0.105) 0.019 TAP2 −0.017 (−0.17 to 0.135) 0.825 −0.374 (−0.564 to −0.184) <0.0001 −0.357 (−0.713 to −0.001) 0.059 RUNX3 −0.341 (−0.47 to −0.211) <0.0001 −0.095 (−0.157 to −0.032) 0.005 0.246 (0.099 to 0.393) 0.002 Open in a new tab ID, immune desert; IE, immune excluded; IN, immune inflamed; SMI, Spatial Molecular Imaging; TMA, tissue microarray; TNBC, triple-negative breast cancer. Next, we used Gene Set Enrichment Analysis (GSEA) analysis of the combined transcripts from all cell types in the TN to identify biological processes that were enriched in each immunotype. The IN immunotype showed significant enrichment in Gene Ontology Biological Process (GOBP) Antigen Processing and Presentation, interferon-alpha (IFNα), and interferon-gamma (IFNγ) responses (false discovery rate, FDR<0.001) in IN, compared with IE and ID tumors ( figure 2E and online supplemental table S7 ). Taken together, these observations support the hypothesis that antigen presentation, particularly with MHC class II components, and type I/II IFN responses specifically within tumor cells, are strongly associated with IN immunotype. pDCs are selectively enriched in IN TNs Given the strong association between interferon signaling and the IN immunotype, we examined the distribution of DC subsets within TNs. Among all immune cell types analyzed, B cells, NK cells, plasmablasts, Tregs, and T CD8 cells, but not T CD4 cells, were significantly more abundant in the TN of IN samples ( online supplemental figure S5A,B ) compared with ID and IE tumors. However, pDCs showed the most striking immunotype-specific pattern. pDCs were detected in the TN of the majority of IN tumors but were rare or absent in both ID and IE tumors ( figure 2F ). Only 6 of 22 ID tumors (27%) had any detectable pDCs in the TN (Fisher’s exact p=0.004 ID vs IN), and none of the six IE tumors had any detectable pDC within the TN. In contrast, 20 of 28 IN tumors (71%) had detectable pDCs in the TN (Fisher’s exact p=0.002, IE vs IN). However, mDCs were detected in the TN of most tumor samples, including 27 out of 28 IN tumors, 20 out of 22 ID tumors (Fisher’s exact p=0.58 for ID compared with IN), and all six IE tumors (Fisher’s exact p=1 compared with IN). These findings suggest that tumor-resident pDCs, but not mDCs, are associated with IN immunotype in TNBC. In contrast, conventional mDCs were present across all immunotypes, indicating that selective enrichment of pDCs, rather than DCs broadly, distinguishes IN tumors. These findings suggest that tumor-resident pDCs may contribute to the interferon-rich microenvironment characteristics of the IN immunotype and prompted further investigation into their functional relevance. Interferon signaling links the IN immunotype to clinical outcomes Given that pDCs are known to be the major producers of IFNα, 37 which plays a critical role in activating T cells and NK cells, 38 , 40 we further investigated the role of IFNαwithin the tumor immune microenvironment. We generated an eigengene composed of the 11 leading-edge genes ( online supplemental table S8 ) in the Hallmark IFNα response signature to evaluate the IFNα response. When comparing all three immunotypes, the IFNα response score was significantly higher in IN tumors, compared with ID and IE tumors ( figure 3A ). Further analysis of the IFNαresponse score (Supplementary Data File S5) across T cell and NK cell subsets within the TN also showed significantly higher IFNα response scores in T CD8 cells, Tregs, and NK cells, but not in T CD4 cells ( figure 3B ). Notably, the lower abundance of T CD4 cells relative to other subsets may have limited the ability to detect a statistically significant difference in this group. We further evaluated the influence of IFNα on T-cell activation using the 12-gene signature from the GeoMx DSP T-cell activation panel ( online supplemental table S8 ) to generate an activation score. Similar to the IFNαresponse score, the T-cell activation score was significantly higher in T CD8 cells, Tregs, and NK cells, but not in T CD4 cells ( figure 3C ). Although interferon signaling can contribute to T cell activation, the T cell inflammation signature used in this study does not include IFNA or IFNG genes themselves. Therefore, the observed T cell inflammation scores are not directly driven by IFNA or IFNG expression but rather reflect downstream T cell-associated effector and activation programs. This distinction allows interferon response and T cell inflammation to be evaluated as related but non-redundant biological processes within the tumor immune microenvironment. Figure 3. Effect of interferon alpha (IFNα) signaling in TNBC TMA and association with clinical outcome. ( A ) IFNα response scores calculated for all cells across immunotypes (ID=black, IE=blue, IN=orange) in TN in TNBC TMA samples. ( B ) IFNα response scores for NK and T cells in the TN. Immunotypes are colored as ID=black, IE=blue, and IN=orange. ( C ) T-cell activation scores for NK and T cells in the TN. Immunotypes are colored as ID=black, IE=blue, and IN=orange. (D, E) Kaplan-Meier analyses of recurrence-free survival (RFS), ( D ) and overall survival (OS), ( E ) stratified by IFNα response score in the FinXX trial, with log-rank p values and patients at risk shown for each time point. (F, G) In CALGB-40603, IFNα response scores were evaluated in relation to pathologic complete response (pCR), defined by two criteria: pCR_Breast (residual invasive disease in breast only, ( F ) and pCR_BrAx (residual invasive disease in both breast and axillary nodes ( G ). Scores are shown for pCR yes vs no groups ( pCR_Breast : yes n=210, no n=178; pCR_BrAx : yes n=186, no n=202). ( H ) Prevalence of IFNα response score in RCB-0 (n=192), RCB-I (n=49), RCB-II (n=99), and RCB-III (n=32) from CALGB-40603 data set. ( I, J ) In I-SPY 2, IFNα scores were associated with pCR in TNBC (n=364; no n=222, yes n=142) and evaluated in patients receiving investigational combinations with paclitaxel: paclitaxel+pembrolizumab (n=29; no n=10, yes n=19) and paclitaxel only (n=85; no n=68, yes n=17). (M, N) In I-SPY 2, IFNα scores were associated with pCR in TNBC (n=364) and evaluated in patients receiving investigational combinations with paclitaxel: paclitaxel+AMG 386 (n=53) and paclitaxel+pembrolizumab (n=29). (O, P) IFNγ scores showed similar associations with pCR in I-SPY 2. The Mann-Whitney U test was used to assess differences between pCR yes versus no groups. ANOVA, analysis of variance; FDR, false discovery rate; ID, immune desert; IE, immune excluded; IN, immune inflamed; RCB, residual cancer burden; TMA, tissue microarray; TN, tumor neighborhood; TNBC, triple-negative breast cancer. Open in a new tab To evaluate the prognostic role of the IFN signaling in TNBC, we evaluated IFN signatures in TNBC from the FinXX trial ( NCT00114816 ). Leveraging our previously published transcriptomic data comprising 770 genes from the NanoString/Bruker IO360 panel in TNBC samples (n=114) from the FinXX cohort, 41 we calculated IFNα and IFNγ response scores. A high IFNα signature score was significantly associated with improved RFS (HR=0.25, 95% CI 0.10 to 0.59, p=0.0006; figure 3D ) and overall survival (OS; HR=0.21, 95% CI 0.09 to 0.51, p<0.001; figure 3E ). Similarly, a high IFNγ signature score correlated with significantly better RFS (HR=0.36, 95% CI 0.16 to 0.79, p=0.0076; online supplemental figure S6A ) and OS (HR=0.27, 95% CI 0.12 to 0.63, p=0.0012; online supplemental figure S6B ). These findings support the prognostic utility of both IFNαand IFNγ response signatures in early-stage TNBC. In both univariate and multivariate Cox analyses, higher IFNα and IFNγ response scores were consistently associated with improved outcomes, reflected by HRs. Importantly, the association remained statistically significant after adjustment for tumor size, nodal involvement, and grade, indicating that the IFN signature provides prognostic information beyond these established clinicopathologic variables ( online supplemental table S9 ). We next analyzed these IFN signatures in a separate cohort using the bulk RNA sequencing (RNA-seq) data from 388 patients enrolled in the CALGB-40603 (Alliance) trial ( NCT00861705 ). This trial is a 2×2 randomized phase II trial evaluating the addition of carboplatin or bevacizumab to standard-of-care neoadjuvant chemotherapy in patients with stage II-III TNBC, and the primary endpoint in this trial was pCR. Both IFNα and IFNγ response scores were positively associated with pCR in both the breast and axillary lymph nodes ( figure 3F,G , online supplemental figure S6C,D ). Furthermore, higher IFNα and IFNγ response scores were significantly associated with lower rates of residual cancer burden classes II and III following neoadjuvant therapy ( figure 3H , online supplemental figure S6E ). In terms of long-term outcomes, a high IFNγ signature score was significantly associated with improved event-free survival (EFS; HR=0.59, 95% CI 0.38 to 0.91, p=0.016; online supplemental figure S6F ). While a higher IFNα response score showed a favorable trend toward improved EFS, this association did not reach statistical significance (HR=0.79, 95% CI 0.54 to 1.14, p=0.20; online supplemental figure S6G ). Nonetheless, the IFNγ gene signature remained significantly associated with improved EFS in this cohort, further supporting its relevance as a biomarker of therapeutic response and prognosis in TNBC. To assess whether the IFNα and IFNγ gene signatures retained prognostic or predictive value in the context of modern therapeutic regimens, including immune checkpoint inhibitors, we analyzed these signatures in the publicly available dataset from the I-SPY 2 trial. 42 I-SPY 2 is a multicenter, phase II adaptive platform trial enrolling patients with stage II–III breast cancer. For this analysis, we focused on a subset of 364 patients with hormone receptor-negative and HER2-negative tumors. This dataset included eight distinct treatment arms: single-agent paclitaxel (n=85); paclitaxel plus the PARP inhibitor ABT-888 and carboplatin (n=39); paclitaxel plus the angiopoietin inhibitor AMG-386 (n=53); paclitaxel plus the HSP90 inhibitor ganetespib (n=45); paclitaxel plus the IGF-1R antibody ganitumab (n=49); paclitaxel plus the allosteric AKT inhibitor MK-2206 (n=32); paclitaxel plus neratinib (n=32); and paclitaxel plus the PD-1 inhibitor pembrolizumab (n=29). Across these treatment arms, the IFNα response score was significantly associated with higher rates of pCR (p<0.001; figure 3I ). In arm-specific analyses, elevated IFNα scores were significantly associated with pCR in the paclitaxel plus pembrolizumab but not paclitaxel only ( figure 3J ; online supplemental figure S6H ). Similar trends were observed for the IFNγ response signature: higher IFNγ scores correlated with increased pCR rates in the overall TNBC cohort ( online supplemental figure S6I ), as well as within the pembrolizumab treatment arm ( online supplemental figure S6I ). These findings suggest that both IFNα and IFNγ signaling pathways may serve as predictive biomarkers of response in immunotherapy-based and targeted treatment combinations for TNBC. pDC-associated neighborhood enriched for IFNα signaling To further assess the functional context of pDCs in IE tumors, we used CosMx single-cell spatial transcriptomics to define pDC-associated regions within intratumoral areas ( figure 4A ). Across 65 analyzed FOV, pDCs were detected in 23 FOVs (18 IN and 5 ID), comprising 101 pDCs localized within tumor nests. Based on spatial proximity, we defined two intratumoral microenvironments for comparison: a pDC neighborhood, consisting of all cells within 50 µm of a pDC, and a non-pDC neighborhood, comprising intratumoral cells located more than 50 µm away from any pDC. Figure 4. Role of plasmacytoid dendritic cells (pDCs) in TNBC TMA and RWCGD data. ( A ) Representative CosMx fields of view (FOV) showing pDC neighborhoods (blue) and non-pDC neighborhoods (red) within the tumor neighborhood (TN) in TNBC TMA samples. ( B ) Percentage of cell types among ID (n=5) and IN (n=18) samples in two pDC separate spatial neighborhoods. ( C ) Composition of major cell types in pDC neighborhoods, summarized for ID (n=5) and IN (n=18) samples in two distinct spatial neighborhoods. ( D ) Kaplan-Meier analysis of event-free survival (EFS) stratified by pDC abundance in the RWCGD cohort. (E–H) High-pDC tumors were associated with higher IFNα abundance ( E ), IFNγ abundance ( F ), IFNγ response score ( G ), and T cell-inflamed score ( H ) in the RWCGD dataset. ID, immune desert; IFNγ, interferon alpha; IN, immune inflamed; RWCGD, Real-World Clinico-Genomic Data; TMA, tissue microarray; TNBC, triple-negative breast cancer. Open in a new tab As anticipated, the pDC neighborhood was significantly enriched for pDCs, constituting 2.0% of all cells ( table 2 ). The most abundant immune cell types in the pDC neighborhood were macrophages (7.6%) and B cells (3.4%). In contrast, the most abundant immune cell types in the non-pDC neighborhood were macrophages (4.3%) and plasmablasts (2.1%) ( figure 4B ). Tumor cell density was lower in the pDC neighborhood compared with the non-pDC neighborhood (67.7% vs 81.1% of all cells, respectively). Notably, tumor cells within the pDC neighborhood exhibited significantly higher IFNα response scores than those in the non-pDC neighborhood ( figure 4C ). These findings indicate that tumor cells in proximity to pDCs were influenced by IFNα signaling, supporting a role for pDCs in modulating the local IFNα-enriched tumor microenvironment. Table 2. Cellular distribution of pDC_neighborhood versus non-pDC_neighborhood at cell level. Terms Non-pDC_neighborhood_TN (N=45 289) pDC_neighborhood_TN (N=4999) Total (N=50 288) P value B cell 839 (1.9%) 168 (3.4%) 1007 (2.0%) <0.001 Endothelial 828 (1.8%) 116 (2.3%) 944 (1.9%) 0.015 Fibroblast 2132 (4.7%) 404 (8.1%) 2536 (5.0%) <0.001 Macrophage 1927 (4.3%) 382 (7.6%) 2309 (4.6%) <0.001 NK 346 (0.8%) 89 (1.8%) 435 (0.9%) <0.001 mDC 118 (0.3%) 30 (0.6%) 148 (0.3%) <0.001 pDC 0 (0.0%) 101 (2.0%) 101 (0.2%) <0.001 Plasmablast 965 (2.1%) 142 (2.8%) 1107 (2.2%) 0.001 T CD4 54 (0.1%) 9 (0.2%) 63 (0.1%) 0.249 T CD8 537 (1.2%) 60 (1.2%) 597 (1.2%) 0.928 Treg 822 (1.8%) 113 (2.3%) 935 (1.9%) 0.027 Tumor 36 721 (81.1%) 3385 (67.7%) 40 106 (79.8%) <0.001 Open in a new tab mDC, myeloid dendritic cell; NK, natural killer; pDC, plasmacytoid DC; TN, tumor neighborhood. Validation of pDC prognostic significance in a real-world cohort As our data identified the potential prognostic significance of pDCs and interferon signaling in early-stage TNBC across the Mayo Clinic cohort, 33 FinXX, 35 43 CALGB-40603, 44 and I-SPY 2 (47) trial ( online supplemental table S10 ), we sought to validate these associations in a larger, real-world dataset. In collaboration with Caris Life Sciences (Phoenix, AZ), we used the CODEai data platform and identified a cohort of 2899 patients with TNBC with available genomic and clinical outcome data (demographic characteristics summarized in online supplemental table S11 ). Within the CODEai data platform, we applied a transcriptomic deconvolution approach to estimate pDC abundance from whole transcriptome sequencing (WTS) data (NovaSeq; Caris Life Sciences, Phoenix, AZ). Consistent with our previous findings, TNBC with higher pDC levels, defined using the median as a cut-off, were associated with significantly improved clinical outcomes, with a median EFS of 24.18 months (95% CI 22.6 to 25.89), compared with 18.35 months (95% CI 17.07 to 19.41), reflecting a median difference of approximately 5.82 months (HR 0.73, 95% CI 0.68 to 0.79, p<0.00001; figure 4D ). In line with our hypothesis, tumors enriched in pDCs also showed significantly elevated expression of IFNα ( figure 4E ), IFNγ ( figure 4F,G ), and a higher T cell-inflamed gene expression score (88 vs −64; figure 4H ). These findings suggest that pDC-enriched tumors are characterized by a more inflamed tumor-immune microenvironment. To determine whether the prognostic relevance of pDC abundance is specific to TNBC or extends to other breast cancer subtypes, we expanded our analysis to include HR−/HER2+ (n=560), HR+/HER2+ (n=780), and HR+/HER2− (n=6094) tumors. Notably, elevated pDC levels were associated with significantly improved EFS specifically in the HR-/HER2+subset (HR 0.79, 95% CI 0.63 to 0.99, p=0.047; online supplemental figure S7A ). In contrast, no significant association was observed in HR+/HER2+tumors (HR 1.0, 95% CI 0.82 to 1.23, p=0.95; online supplemental figure S7B ). Interestingly, in HR+/HER2− breast cancer, higher pDC abundance was associated with worse EFS (HR 1.18, 95% CI 1.11 to 1.27, p<0.00001; online supplemental figure S7C ), suggesting a subtype-specific role of pDCs in modulating the tumor immune microenvironment. Discussion Using high-resolution single-cell spatial transcriptomics, our study advances the understanding of the tumor immune microenvironment in TNBC by characterizing its spatial architecture. Merging evidence in melanoma, colorectal, and lung cancers 19 , 22 shows that not only immune cell quantity but also spatial localization influences outcomes and therapy response. 17 45 This has led to the concept of immunotypes—IN, IE, and ID 17 —which reflect distinct immune cell density and organization patterns within the TME. Understanding these patterns may inform strategies to improve responses in immunologically cold tumors. In breast cancer, while earlier studies assessed both stromal and intratumoral TILs, current standards emphasize sTILs due to better reproducibility. 7 However, the added value of incorporating iTILs, particularly in assessing immune exclusion, remains unclear. Most prior studies have relied on immunohistochemistry or immunofluorescence to quantify T CD8 cells, 19 22 46 while these low-plex approaches can define diverse immune subsets or their functional transcriptional states, these approaches remain limited in multiplexing depth and in capturing functional transcriptional states, thereby constraining insight into immune dynamics within the TME. Our current study highlights the significance of spatially resolved immune profiling in TNBC. Leveraging high-plex single-cell spatial transcriptomics, we demonstrate that not only the density but also the localization and cellular composition of immune cell infiltrates have significant prognostic value. These results extend our previous work, which identified intratumoral, rather than stromal, expression of antigen-presenting molecules HLA-DR and co-stimulatory receptor CD40 as predictors of outcomes in early-stage TNBC, 27 independent of sTILs and established clinicopathologic parameters. Consistent with our findings, other studies have highlighted the prognostic relevance of antigen presentation machinery in TNBC and other malignancies. 41 42 Using bulk RNA profiling and immunohistochemical analyses, prior reports have demonstrated that increased expression of MHC molecules, particularly MHC class II, is associated with improved clinical outcomes in TNBC. However, these approaches lacked the spatial and single-cell resolution required to discern whether MHC expression on specific cell types determined its prognostic role. Our study addresses this gap by leveraging spatially resolved single-cell analyses to pinpoint that MHC class II expression specifically within tumor cells, not immune infiltrates, is the critical determinant of favorable prognosis in TNBC. This tumor-intrinsic antigen-presenting phenotype likely enhances visibility to T CD4 cells, thereby fostering more effective anti-tumor immune surveillance and highlighting the functional importance of tumor cell immunogenicity in shaping tumor-immune architecture and patient outcomes. To further evaluate distinct tumor immunotypes, we performed differential gene expression analysis of tumor cells stratified by immunotype. Acknowledging the limited resolution of a 960-transcript panel, our transcriptomic analyses demonstrated that IE tumors, despite having high levels of sTILs similar to IN tumors, exhibited gene expression profiles that were more closely aligned with ID tumors. Both ID and IE tumor cells showed significantly reduced expression of MHC class I and II transcripts compared with IN tumors, implicating impaired antigen processing and presentation pathways. In addition, IN tumors showed robust enrichment of the antigen presentation machinery as well as IFNα and IFNγ signaling. Given that IE and IN tumors diverge primarily within the intratumoral compartment, we examined the immune cellular architecture specifically within the TN. As anticipated, IN TNs had significantly higher T CD8 cells, B cells, and plasmablasts, as well as pDC, compared with those in ID and IE tumors. pDCs are a subset of DCs that secrete high levels of type I interferon, classically characterized in antiviral responses, and promote cytotoxic T-cell and NK-cell activation 38 , 40 as well as antigen presentation via MHC class II. 47 More recently, tumor-resident DCs have received increasing attention for their role in sustaining immune surveillance within the tumor microenvironment. 17 39 47 48 In preclinical models, the lack of pDC has been implicated as a potential driver of immune evasion, particularly in the ID phenotype. 17 49 50 While pDCs are well established as key effectors in antiviral defense and autoimmune pathogenesis, 38 51 their role in cancer immunity remains context-dependent, with conflicting data on clinical outcomes in different contexts and tumor types. 52 , 55 Transcriptomic analysis in TNBC from the Cancer Genome Atlas and the Molecular Taxonomy of Breast Cancer International Consortium databases suggested that high levels of pDCs confer improved outcomes compared with conventional DCs. Furthermore, a high level of pDCs was also associated with increased T CD8, T CD4 cells, IFN-signaling, and cytolytic activity. 56 Our data suggest that tumor-resident pDCs may play an important role in maintaining the IN state. To validate the prognostic relevance of pDCs beyond our discovery cohorts, we examined transcriptomic and clinical data from a large real-world cohort of TNBC patients through the Caris Life Sciences CODEai platform. Consistently, higher inferred pDC abundance was associated with significantly longer OS and was positively correlated with both IFNα/IFNγ signaling and IN gene expression scores. Interestingly, this prognostic association was confined to hormone receptor–negative subtypes, including HR−/HER2+, while high pDC levels were paradoxically associated with worse outcomes in HR+/HER2− disease, suggesting that the role of pDCs may be subtype-specific and context-dependent. Nevertheless, there are limitations in our study that warrant consideration. First, the single-cell spatial transcriptomic analysis was conducted on a relatively limited number of cases, which may restrict the generalizability of our findings. In addition, we used TMAs for spatial profiling, potentially underrepresenting the full spatial immune architecture and heterogeneity of the tumor microenvironment. Furthermore, our transcriptomic panel was limited to 960 genes, which constrains the resolution and breadth of cellular and molecular characterization. Nevertheless, to mitigate these limitations, we validated our key findings in multiple independent cohorts using bulk tumor gene expression datasets, including FinXX, 35 43 CALGB-40603, 44 I-SPY2(47), and the Caris CodeAI clinico-genomic database. In addition, the Mayo Clinic TNBC cohort represents an earlier patient population treated prior to the incorporation of ICIs into standard breast cancer management. Most patients (70%) received conventional adjuvant chemotherapy alone, while the remainder did not receive adjuvant systemic therapy. Consequently, the IFN signature evaluated in this cohort reflects intrinsic tumor–immune biology rather than a biomarker of response to immunotherapy or targeted therapies. This distinction is critical for interpretation, as the observed IFN-associated immune features likely represent baseline immune states that influence disease biology and prognosis independent of treatment-induced immune modulation. Given that ICIs are now an integral component of systemic therapy for TNBC, we further assessed the clinical relevance of IFN signaling and pDCs in more contemporary cohorts, including patients treated with ICIs. These complementary analyses support the broader applicability of our findings and underscore the importance of IFN signaling and pDCs as fundamental components of TNBC tumor–immune biology across diverse therapeutic contexts. In summary, our study highlights the critical role of spatial immune architecture in defining immunotypes and predicting outcomes in TNBC. We identify intratumoral pDCs as key orchestrators of a favorable immune contexture via IFN signaling and provide rationale for the development of pDC-targeted therapies and spatially informed biomarkers. Further studies are warranted to define the upstream regulators of pDC recruitment and function within the tumor, and to explore the therapeutic potential of modulating the pDC–interferon axis in TNBC. Materials and methods Study design TNBC samples from the Mayo Clinic cohort were analyzed using NanoString/Bruker GeoMx through Mayo Clinic Spatial Biology Core Facility (RRID: SCR_026548) to obtain spatial protein expression profiling and CosMx to obtain single cell RNA transcript profiling through NanoString Technologies (RRID:SCR_023909). This cohort is composed of 605 patients (female only) with early stage, centrally verified TNBC (ER/PR<1% and HER2-negative) who underwent surgery at Mayo Clinic (1985–2012). Tumor samples were collected prior to any systemic therapy. A detailed clinicopathologic review of the Mayo TNBC cohort has been previously published and the corresponding TMA is composed of 2–3×1 mm tissue cores from representative FFPE tumor blocks, as previously reported. 33 GeoMx data were used to define TNBC immunotypes and identify protein expression patterns associated with immunotype. The TMA samples represent a random selection for CosMx experiment. FFPE sections from tumor tissue cores of 56 unique patients were evaluated using CosMx platform. All tumors were conventional invasive ductal carcinomas (NOS), with no special histologic subtypes included. FOV (500 μm × 500 μm) were selected based on histopathologic assessment and PanCK immunostaining to ensure representation of both tumor and stromal compartments. Specifically, one FOV per core was chosen, preferentially capturing regions containing approximately balanced tumor and adjacent stromal areas when feasible. This approach was intended to enable analysis of tumor–microenvironment interactions rather than enrichment for a single cellular component. CosMx data were used to identify and quantify immune and other cell types, to define neighborhood associations, and to assess transcript profiles in several cell types as a function of immunotype. Results from the single cell transcriptome analysis were evaluated in TNBC samples from the FinXX ( NCT00114816 ), CALGB-40603 ( NCT00861705 ), I-SPY 2 ( NCT01042379 ) trials, and Caris Life Sciences (Phoenix, AZ, USA) data. 15 247 breast cancer samples underwent molecular testing at Caris Life Sciences (Phoenix, AZ, USA), a College of American Pathologists/Clinical Laboratory Improvement Amendments-certified laboratory. H&E-stained FFPE slides of the patient’s tumor underwent review by a board-certified pathologist. For WTS, only those samples were included that had a minimum of 10% tumor nuclei in the area of microdissection, with a minimum 10 mm 2 dissection area. Definition of immunotype The conventional definition of immunotype is based on histopathological evaluation of sTILs, scored per International Immuno-Oncology Biomarker Working Group guidelines 7 and/or the abundance of CD8-positive T-cells in whole-mount sections (reviewed in Chen and Mellman). 18 We used whole mount sTILs 33 in combination with CD8 protein abundance based on GeoMx DSP of user-defined regions of interest as a surrogate for CD8-positive T-cells. TNBC TMA samples were segmented into intratumoral (cytokeratin-positive, CK pos ), and stromal (cytokeratin-negative, CK neg ) compartments based on cytokeratin immunofluorescence, as previously described. 26 27 Both IE and Inflamed (IN) samples were defined by sTILs >30% and differed only in their intratumoral CD8 protein abundance, with IE samples defined by ≤median CD8 protein expression and IN samples defined by CD8 protein>median. ID samples were defined by whole mount sTILs ≤30% and total CD8 protein (from CK pos intratumoral and CK neg stroma compartments combined) ≤median. Data normalization GeoMx DSP raw counts from each AOI could be read by GeoMx software (RRID: SCR_023424), and they were scaled to the 75th percentile of expression (Q3) following the NanoString/Bruker GeoMx Data Analysis User Manual (MAN-10154-04). The expected background of each AOI was estimated as the geometric mean of the negative control probes, multiplied by the geometric SD to the second power 2 . Sample preparation for CosMx SMI RNA assay To prepare the tissue samples for the CosMx SMI platform (RRID: SCR_023909; NanoString/Bruker Technologies Inc, Seattle, WA), TMA sample slides were baked overnight at 60°C to ensure FFPE tissue core adherence to the glass slides. Samples underwent deparaffinization, proteinase K digestion, and heat-induced epitope retrieval (HIER) procedures to expose target RNAs and epitopes using the Leica Bond Rx system. For TNBC TMA samples, 3 µg/mL of proteinase K (Thermo Fisher) incubated at 40°C for 30 min and HIER at 100°C for 15 min in Leica buffer ER1 conditions were used. The TMA samples were rinsed with diethylpyrocarbonate (DEPC)-treated water (DEPC H2O) twice before incubating in 1:1000 diluted fiducials (Bangs Laboratory) in 2X SSCT (2X saline sodium citrate (SSC), 0.001% Tween 20) solution for 5 min at room temperature. Excessive fiducials were removed by rinsing the samples with 1X phosphate-buffered saline (PBS), followed by fixation with 10% neutral buffered formalin for 5 min at room temperature. Fixed samples were rinsed with Tris-glycine buffer (0.1M glycine, 0.1M Tris-base in DEPC H2O) and 1X PBS for 5 min each before blocked using 100 mM N-succinimidyl acetate (NHS-acetate, Thermo Fisher) in NHS-acetate buffer (0.1M NaP, 0.1% Tween pH 8 in DEPC H2O) for 15 min at room temperature. Prepared samples were rinsed with 2X SSC for 5 min, and then an Adhesive Secure Seal Hybridization Chamber (Grace Bio-Labs) was placed to cover the TMAs. 980-plex RNA ISH probes were denatured at 95°C for 2 min and then placed on ice before preparing ISH probe mix (1 nM ISH probes, 1X Buffer R, 0.1 U/µL SUPERaseIn in DEPC H2O). The ISH probe mix was pipetted into the hybridization chamber and the chamber was sealed using adhesive tape. Hybridization was performed at 37 °C overnight to prevent evaporation. After the overnight hybridization, samples were washed with 50% formamide (VWR) in 2X SSC at 37 °C for 25 min twice, rinsed with 2X SSC for 2 min for twice at room temperature, and then blocked with 100 mM NHS-acetate for 15 min. After blocking, the samples were washed twice using 2X SSC for 2 min at room temperature. A custom-made slide cover was attached to the sample slide to form a flow cell. SMI instrument run The 980-plex RNA assay was designed to detect 960 target genes ( online supplemental table S3 ) and 20 negative probe controls. Negative control probes are modeled after synthetic sequences from the External RNA Controls Consortium. 57 RNA target readout on the SMI instrument was performed following published protocols. 58 In brief, TMA samples in the assembled flow cell were loaded onto the SMI instrument and washed with Reporter Wash Buffer to rinse the samples and remove air bubbles. The entire flow cell was scanned (SMI preview scan) and 15–25 FOVs were placed on the TMA cores for RNA readout, with 1 FOV per core. The RNA readout cycle started by flowing 100 µL of Reporter Pool 1 into the flow cell and incubating for 15 min. After incubation, 1 mL of Reporter Wash Buffer was flowed into the flow cell to wash out the unbound reporter probes, followed by replacing the Reporter Wash Buffer with Imaging Buffer prior to imaging. Nine Z-stack images (0.8 µm step size) of each FOV were acquired, and then fluorophores on the barcoded reporter probes were released at photocleavable linkers by UV illumination and washed off with Strip Wash Buffer. This fluidic exchange and imaging procedure were repeated for the 16 reporter pools, and the 16th round of reporter hybridization-imaging cycle was repeated multiple times to increase RNA detection sensitivity. After RNA readout, TMA samples were incubated with a 4-fluorophore-conjugated antibody cocktail against CD298, PanCK, CD45, and CD3 proteins and DAPI satin in the SMI instrument for 1 hour. Nine Z-stack images for five channels (four antibodies and DAPI) were captured after unbound antibodies and DAPI stain were washed with Reporter Washing Buffer, and the flow cell was filled with Imaging Buffer. SMI image processing and feature extraction Raw image processing and feature extraction were performed using an in-house SMI data processing pipeline 58 which includes registration, feature detection, and localization. 3D rigid image registration was made using fiducials embedded in the TMA samples matched with the fixed image reference established at the beginning of the SMI run to correct for any shift. Second, RNA image analysis algorithm was used to identify reporter signature locations in X, Y, and Z axes along with the assigned confidence. The reporter signature locations and the associated features were collated into a single list. Lastly, the XYZ location information of individual target transcripts was extracted and recorded in a table by secondary analysis algorithm, as described in He et al . 58 Cell segmentation analysis The Z-stack images of immunostaining plus DAPI were used for drawing cell boundaries on the samples. 58 A cell segmentation pipeline using a machine learning algorithm 59 60 was used to accurately assign transcripts to cell locations and subcellular compartments. The transcript profile of individual cells was generated by combining target transcript location and cell segmentation boundaries. Cell type annotation We employed the InSituType R function to carry out cell typing using a Bayesian classifier from reference datasets that supports unsupervised and semisupervised cell typing through an Expectation Maximization (EM) algorithm. 61 Cell type annotation began by partitioning the cells into two categories, CKpos and CKneg cells. This determination was based on both the observed protein and RNA expression for cytokeratins, specifically the RNA expression of “KRT5”, “KRT7”, “KRT8”, “KRT17”, “KRT18”, and “KRT19”. CKpos cells were defined as having both elevated cytokeratin protein and RNA expression, and similarly, CKneg cells were defined as having no or low expression of both protein and RNA. The CKpos cells were grouped by TMA core of origin, and each group was further subsetted to remove outliers, as defined as cells not near the group centroid in two dimensional UMAP space. The mean expression was then calculated for each group over all genes to generate a CosMx derived expression profile for each TMA core’s tumor cell population. The CKneg cells were assigned cell types by direct comparison of observed RNA expression profiles in each cell to expression profiles of immune cell types using a negative binomial model within the InSituType package ( https://github.com/Nanostring-Biostats/InSituType ; RRID: SCR_025976). The cell profiles were derived from single-cell RNA-seq data https://github.com/Nanostring-Biostats/InSituType/blob/main/data/ioprofiles.RData . The mean expression was calculated for each cell type over all genes to generate a CosMx data-derived expression profile for all panCKneg cell populations. Finally, the InSituType package was used in semisupervised mode to make single-cell assignments using both the CKpos and CKneg CosMx derived cell profiles. Cell assignments were verified using comparison to known marker gene expression, protein expression, and location within tissue spatial organization. 61 Spatial neighborhood annotation The cell type composition of cell populations within a 50 μm radius in 2D physical space of each cell was calculated and clustered into two groups using MClust ( https://mclust-org.github.io/mclust/ ; RRID: SCR_027153). The resulting neighborhood types, TN and SN, were labeled based on cell type composition and spatial location within the tissue structure. TNBC TMA SMI data analysis The SMI data included two neighborhoods (TN and SN) with 12 cell types (tumor, B-cell, endothelial, fibroblast, macrophage, mDC, NK, mDC, plasmablast, T-CD4, T-CD8, Treg). There were 240 311 CosMx single-cells recorded in initial 75 SMI-FOV samples. Based on immunotype defined by TNBC TMA DSP data, there were 56 TNBC TMA SMI samples matched by clinic-ID and immunotypes (22 ID, 6 IE, 28 IN). Unassigned samples failed to meet both of the criteria used to define immunotype (sTILs AND CD8 protein). The eigengene score represents the principal component analysis of the expression matrix 62 of IFN response leading edge genes or T-cell activation signature genes ( online supplemental table S8 ). The eigengene score was calculated using the built-in R prcomp and princomp functions. Samples from the FinXX trial were stratified based on an optimized cut-off value of IFNα response score, calculated using the R survival functions survfit and coxph. 63 Whole transcriptome sequencing Tumor-specific RNA was extracted from micro-dissected FFPE specimens with at least 10% tumor content using a Qiagen RNeasy FFPE Kit (Germantown, MD). RNA quality and quantity were determined using the Agilent TapeStation (Santa Clara, CA). Biotinylated RNA baits were hybridized to the synthesized and purified cDNA targets, and the bait-target complexes were amplified in a post-capture PCR reaction. WTS on Illumina Novaseq 6500 (Illumina, San Diego, CA) generated an average of 60M reads. Raw data were demultiplexed by Illumina Dragen BioIT accelerator, trimmed, counted, PCR-duplicates removed, and aligned to the human reference genome hg19 by STAR aligner (v.2.7.8a; RRID: SCR_004463). Transcripts per million were calculated with the Salmon expression pipeline. 64 Differentially expressed genes were determined by calculating log2 fold-change between groups. Real-world OS analysis Caris CODEai clinico-genomic database containing insurance claims data were used to calculate real-world OS from tissue collection as a surrogate for diagnosis until last contact. Patients without contact/claims data for a period of at least 100 days were presumed deceased. Conversely, patients with documented clinical activity within 100 days prior to the latest data update were censored in the analysis. Kaplan-Meier estimates were calculated for molecularly defined patient cohorts. HRs were determined by Cox Proportional Hazards model and p values by log-rank test. Significance was determined as p<0.05. Gene set enrichment analysis The GSEA software is publicly available from the Broad Institute of MIT and Harvard University ( http://software.broadinstitute.org/gsea/index.jsp ). Gene counts from CosMx SMI RNA assay were utilized for Hallmark and GOBP analysis (version v2023.1). GSEA performed using the gene permutation option, and gene sets smaller than 15 or larger than 500 were excluded. The FDR with less than 0.05 (FDR<0.05) was considered as significantly enriched. Additionally, GSEA was conducted using gene set signatures from the MsigDB database, with enrichment scores (NES) calculated to elucidate biological pathways associated with TNBC. Statistical analysis There were a total of 212 samples with 608 segments (tumor or stroma). The samples were identified as 116 ID, 23 IE, and 73 IN. All differential expression analyses were carried out using R Statistical Software (V.4.2.2; R Foundation for Statistical Computing, Vienna, Austria). For segment-level analysis of DSP data, the linear mixed model was used to calculate the differential abundance of proteins as a function of immunotypes (eg, ID vs IE, ID vs IN, or IE vs IN) based on tumor or stroma segments separately. Differential abundance (listed as log2 fold change (log2FC) in all figures and tables) was estimated from the model with significance defined as two-sided p<0.05. For survival analysis, segments were averaged to estimate the HR of recurrence using the Cox regression model. Kaplan-Meier survival curves were generated for RFS as a continuous variable vs immune cohorts. Statistical significance was determined using log-rank test with two-sided p<0.05. The linear mixed model was used for differential expression analysis of CosMx data from specific cell types within the spatially defined neighborhoods. Log2FC with 95% CI were estimated for gene expression grouped by immunotypes. False detection rate was estimated using Benjamini-Yekutieli estimation, 65 with significance defined as two-sided p<0.05. Distribution of pDC cells in ID, IE, and IN samples was analyzed using Fisher’s Exact test, with significance defined and two-sided p<0.05. In Caris data, Kaplan-Meier estimates were calculated for pDC-low and pDC-high cohorts from tissue collection to last contact. A p<0.05 was considered statistically significant. Statistical significance of molecular characteristics between pDC-stratified cohorts was determined by Fisher’s exact or χ² and Mann-Whitney U tests. Where shown, p values were adjusted for multiple comparisons to control the FDR using the Benjamini-Hochberg method ( q <0.05 was considered statistically significant, and p<0.05 was considered a trend). Caris data availability The deidentified sequencing data is owned by Caris Life Sciences and cannot be publicly shared due to the data usage agreement in place. These data will be made available to researchers for replication and verification purposes through our letter of intent process, which is generally fulfilled within 3 months. For more information on how to access this data, please contact Joanne Xiu at jxiu@carisls.com. Supplementary material online supplemental file 1 jitc-14-6-s001.pdf (4.3MB, pdf) DOI: 10.1136/jitc-2025-014020 online supplemental file 2 jitc-14-6-s002.jpg (2.7MB, jpg) DOI: 10.1136/jitc-2025-014020 Acknowledgements We wish to express our most grateful thanks to our patients, without whose willingness to donate tissue samples these experiments would have been impossible. Additional thanks to Brandi Edenfield and Laura Lewis-Tuffin, Ph.D. for help with H&E staining and Aperio scanning. Footnotes Funding: Research reported in this publication was supported by the Breast Cancer Research Foundation (BCRF 22-161) to EAT; Florida Department of Health Bankhead Coley (22B08) to EAT.; National Cancer Institute of the National Institutes of Health Mayo Clinic Breast Cancer SPORE P50CA116201 to MPG; and US Department of Defense, Breast Cancer Research Program—W81XWH-15-1-0292 to SC. Provenance and peer review: Not commissioned; externally peer reviewed. Patient consent for publication: Not applicable. Ethics approval: All patients provided written informed consent for the use of deidentified tissues for research, and all studies were carried out under IRB protocol 18-008231 approved by the Mayo Clinic Institutional Review Board. This study was conducted in accordance with guidelines of the Declaration of Helsinki, Belmont Report, and US Common Rule, and in keeping with 45 CFR 46.101(b)(4), was performed using retrospective, deidentified clinical data. Per WCG IRB, it was considered an institutional review board (IRB) exempt with waiver of patient consent. Data availability free text: The clinical trials datasets CALGB 40603 trial ( NCT00861705 ), and I-SPY2 trial ( NCT01042379 ) analyzed during this study are available in the Gene Expression Omnibus (GEO) repository under accession number GSE154524 and GSE194040 . The deidentified sequencing Real-World Data is owned by Caris Life Sciences and cannot be publicly shared due to the data usage agreement in place. The data of TNBC TMA GeoMx data and TNBC TMA CosMx are available as a supplement to the manuscript (Supplementary Data and Github ( liuyi901/pDC_JITC: Supplementary Data )). The multiplex protein expression profiles and differential expression analysis in intratumoral and stromal compartments using GeoMx DSP are given in Supplementary Data Files 1 and 2. The cell identity, location, and gene counts using CosMx SMI are given in the Supplementary Data Files 3 and 4. The IFNa response score and T cell activation score in selected cell types in TN is given in the Supplementary Data File 5. Data generated or analyzed are available on reasonable request. Raw data supporting the figures and graphs as well as tables are available from the authors on request. Data availability statement Data are available on reasonable request. Data may be obtained from a third party and are not publicly available. References 1. Zagami P, Carey LA. Triple negative breast cancer: Pitfalls and progress. NPJ Breast Cancer. 2022;8:95. doi: 10.1038/s41523-022-00468-0. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 2. Galon J, Costes A, Sanchez-Cabo F, et al. Type, density, and location of immune cells within human colorectal tumors predict clinical outcome. Science. 2006;313:1960–4. doi: 10.1126/science.1129139. [ DOI ] [ PubMed ] [ Google Scholar ] 3. Gooden MJM, de Bock GH, Leffers N, et al. The prognostic influence of tumour-infiltrating lymphocytes in cancer: a systematic review with meta-analysis. Br J Cancer. 2011;105:93–103. doi: 10.1038/bjc.2011.189. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 4. Fridman WH, Pagès F, Sautès-Fridman C, et al. The immune contexture in human tumours: impact on clinical outcome. Nat Rev Cancer. 2012;12:298–306. doi: 10.1038/nrc3245. [ DOI ] [ PubMed ] [ Google Scholar ] 5. Brummel K, Eerkens AL, de Bruyn M, et al. Tumour-infiltrating lymphocytes: from prognosis to treatment selection. Br J Cancer. 2023;128:451–8. doi: 10.1038/s41416-022-02119-4. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 6. Leon-Ferre RA, Jonas SF, Salgado R, et al. Tumor-Infiltrating Lymphocytes in Triple-Negative Breast Cancer. JAMA. 2024;331:1135–44. doi: 10.1001/jama.2024.3056. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 7. Salgado R, Denkert C, Demaria S, et al. The evaluation of tumor-infiltrating lymphocytes (TILs) in breast cancer: recommendations by an International TILs Working Group 2014. Ann Oncol. 2015;26:259–71. doi: 10.1093/annonc/mdu450. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 8. Loi S, Drubay D, Adams S, et al. Tumor-Infiltrating Lymphocytes and Prognosis: A Pooled Individual Patient Analysis of Early-Stage Triple-Negative Breast Cancers. J Clin Oncol. 2019;37:559–69. doi: 10.1200/JCO.18.01010. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 9. El Bairi K, Haynes HR, Blackley E, et al. The tale of TILs in breast cancer: A report from The International Immuno-Oncology Biomarker Working Group. NPJ Breast Cancer. 2021;7:150. doi: 10.1038/s41523-021-00346-1. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 10. Park JH, Jonas SF, Bataillon G, et al. Prognostic value of tumor-infiltrating lymphocytes in patients with early-stage triple-negative breast cancers (TNBC) who did not receive adjuvant chemotherapy. Ann Oncol. 2019;30:1941–9. doi: 10.1093/annonc/mdz395. [ DOI ] [ PubMed ] [ Google Scholar ] 11. Denkert C, von Minckwitz G, Darb-Esfahani S, et al. Tumour-infiltrating lymphocytes and prognosis in different subtypes of breast cancer: a pooled analysis of 3771 patients treated with neoadjuvant therapy. Lancet Oncol. 2018;19:40–50. doi: 10.1016/S1470-2045(17)30904-X. [ DOI ] [ PubMed ] [ Google Scholar ] 12. Denkert C, Loibl S, Noske A, et al. Tumor-associated lymphocytes as an independent predictor of response to neoadjuvant chemotherapy in breast cancer. J Clin Oncol. 2010;28:105–13. doi: 10.1200/JCO.2009.23.7370. [ DOI ] [ PubMed ] [ Google Scholar ] 13. Wood SJ, Gao Y, Lee J-H, et al. High tumor infiltrating lymphocytes are significantly associated with pathological complete response in triple negative breast cancer treated with neoadjuvant KEYNOTE-522 chemoimmunotherapy. Breast Cancer Res Treat. 2024;205:193–9. doi: 10.1007/s10549-023-07233-2. [ DOI ] [ PubMed ] [ Google Scholar ] 14. Sharma P, Stecklein SR, Yoder R, et al. Clinical and Biomarker Findings of Neoadjuvant Pembrolizumab and Carboplatin Plus Docetaxel in Triple-Negative Breast Cancer. JAMA Oncol. 2024;10:227. doi: 10.1001/jamaoncol.2023.5033. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 15. Leon-Ferre RA, Salgado R, Jonas SF. Prognostic Value of Tumor-Infiltrating Lymphocytes for Patients With Triple-Negative Breast Cancer-Reply. JAMA. 2024;332:337–8. doi: 10.1001/jama.2024.10483. [ DOI ] [ PubMed ] [ Google Scholar ] 16. Vandereyken K, Sifrim A, Thienpont B, et al. Methods and applications for single-cell and spatial multi-omics. Nat Rev Genet. 2023;24:494–515. doi: 10.1038/s41576-023-00580-2. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 17. Mellman I, Chen DS, Powles T, et al. The cancer-immunity cycle: Indication, genotype, and immunotype. Immunity. 2023;56:2188–205. doi: 10.1016/j.immuni.2023.09.011. [ DOI ] [ PubMed ] [ Google Scholar ] 18. Chen DS, Mellman I. Elements of cancer immunity and the cancer-immune set point. Nature New Biol. 2017;541:321–30. doi: 10.1038/nature21349. [ DOI ] [ PubMed ] [ Google Scholar ] 19. Joyce JA, Fearon DT. T cell exclusion, immune privilege, and the tumor microenvironment. Science. 2015;348:74–80. doi: 10.1126/science.aaa6204. [ DOI ] [ PubMed ] [ Google Scholar ] 20. Herbst RS, Soria J-C, Kowanetz M, et al. Predictive correlates of response to the anti-PD-L1 antibody MPDL3280A in cancer patients. Nature New Biol. 2014;515:563–7. doi: 10.1038/nature14011. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 21. Hegde PS, Karanikas V, Evers S. The Where, the When, and the How of Immune Monitoring for Cancer Immunotherapies in the Era of Checkpoint Inhibition. Clin Cancer Res. 2016;22:1865–74. doi: 10.1158/1078-0432.CCR-15-1507. [ DOI ] [ PubMed ] [ Google Scholar ] 22. Mariathasan S, Turley SJ, Nickles D, et al. TGFβ attenuates tumour response to PD-L1 blockade by contributing to exclusion of T cells. Nature New Biol. 2018;554:544–8. doi: 10.1038/nature25501. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 23. Mlynska A, Gibavičienė J, Kutanovaitė O, et al. Defining Melanoma Immune Biomarkers-Desert, Excluded, and Inflamed Subtypes-Using a Gene Expression Classifier Reflecting Intratumoral Immune Response and Stromal Patterns. Biomolecules. 2024;14:171. doi: 10.3390/biom14020171. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 24. Ayers M, Lunceford J, Nebozhyn M, et al. IFN-γ-related mRNA profile predicts clinical response to PD-1 blockade. J Clin Invest. 2017;127:2930–40. doi: 10.1172/JCI91190. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 25. Roller A, Davydov II, Schwalie PC, et al. Tumor-agnostic transcriptome-based classifier identifies spatial infiltration patterns of CD8+T cells in the tumor microenvironment and predicts clinical outcome in early-phase and late-phase clinical trials. J Immunother Cancer. 2024;12:e008185. doi: 10.1136/jitc-2023-008185. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 26. Carter JM, Polley M-YC, Leon-Ferre RA, et al. Characteristics and Spatially Defined Immune (micro)landscapes of Early-stage PD-L1-positive Triple-negative Breast Cancer. Clin Cancer Res. 2021;27:5628–37. doi: 10.1158/1078-0432.CCR-21-0343. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 27. Carter JM, Chumsri S, Hinerfeld DA, et al. Distinct spatial immune microlandscapes are independently associated with outcomes in triple-negative breast cancer. Nat Commun. 2023;14:2215. doi: 10.1038/s41467-023-37806-0. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 28. Bian Y, Walter DL, Zhang C. Efficiency of Interferon-γ in Activating Dendritic Cells and Its Potential Synergy with Toll-like Receptor Agonists. Viruses. 2023;15:1198. doi: 10.3390/v15051198. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 29. Duong E, Fessenden TB, Lutz E, et al. Type I interferon activates MHC class I-dressed CD11b+ conventional dendritic cells to promote protective anti-tumor CD8+ T cell immunity. Immunity. 2022;55:308–23. doi: 10.1016/j.immuni.2021.10.020. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 30. Slaney CY, Möller A, Hertzog PJ, et al. The role of Type I interferons in immunoregulation of breast cancer metastasis to the bone. Oncoimmunology. 2013;2:e22339. doi: 10.4161/onci.22339. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 31. Ngo C, Garrec C, Tomasello E, et al. The role of plasmacytoid dendritic cells (pDCs) in immunity during viral infections and beyond. Cell Mol Immunol. 2024;21:1008–35. doi: 10.1038/s41423-024-01167-5. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 32. Greene TT, Jo YR, Zuniga EI. Infection and cancer suppress pDC derived IFN-I. Curr Opin Immunol. 2020;66:114–22. doi: 10.1016/j.coi.2020.08.001. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 33. Leon-Ferre RA, Polley M-Y, Liu H, et al. Impact of histopathology, tumor-infiltrating lymphocytes, and adjuvant chemotherapy on prognosis of triple-negative breast cancer. Breast Cancer Res Treat. 2018;167:89–99. doi: 10.1007/s10549-017-4499-7. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 34. liuyi901 . GitHub; pDC_JITC. https://github.com/liuyi901/pDC_JITC Available. [ Google Scholar ] 35. Joensuu H, Kellokumpu-Lehtinen P-L, Huovinen R, et al. Adjuvant Capecitabine in Combination With Docetaxel, Epirubicin, and Cyclophosphamide for Early Breast Cancer: The Randomized Clinical FinXX Trial. JAMA Oncol. 2017;3:793–800. doi: 10.1001/jamaoncol.2016.6120. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 36. Milner JJ, Toma C, Yu B, et al. Runx3 programs CD8+ T cell residency in non-lymphoid tissues and tumours. Nature New Biol. 2017;552:253–7. doi: 10.1038/nature24993. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 37. Kvedaraite E, Ginhoux F. Human dendritic cells in cancer. Sci Immunol. 2022;7:eabm9409. doi: 10.1126/sciimmunol.abm9409. [ DOI ] [ PubMed ] [ Google Scholar ] 38. Reizis B. Plasmacytoid Dendritic Cells: Development, Regulation, and Function. Immunity. 2019;50:37–50. doi: 10.1016/j.immuni.2018.12.027. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 39. Murphy TL, Murphy KM. Dendritic cells in cancer immunology. Cell Mol Immunol. 2022;19:3–13. doi: 10.1038/s41423-021-00741-5. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 40. Koucký V, Bouček J, Fialová A. Immunology of Plasmacytoid Dendritic Cells in Solid Tumors: A Brief Review. Cancers (Basel) 2019;11:470. doi: 10.3390/cancers11040470. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 41. Asleh K, Brauer HA, Sullivan A, et al. Predictive Biomarkers for Adjuvant Capecitabine Benefit in Early-Stage Triple-Negative Breast Cancer in the FinXX Clinical Trial. Clin Cancer Res. 2020;26:2603–14. doi: 10.1158/1078-0432.CCR-19-1945. [ DOI ] [ PubMed ] [ Google Scholar ] 42. Wang H, Yee D. I-SPY 2: a Neoadjuvant Adaptive Clinical Trial Designed to Improve Outcomes in High-Risk Breast Cancer. Curr Breast Cancer Rep. 2019;11:303–10. doi: 10.1007/s12609-019-00334-2. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 43. Joensuu H, Kellokumpu-Lehtinen P-L, Huovinen R, et al. Adjuvant Capecitabine for Early Breast Cancer: 15-Year Overall Survival Results From a Randomized Trial. J Clin Oncol. 2022;40:1051–8. doi: 10.1200/JCO.21.02054. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 44. Shepherd JH, Ballman K, Polley M-YC, et al. CALGB 40603 (Alliance): Long-Term Outcomes and Genomic Correlates of Response and Survival After Neoadjuvant Chemotherapy With or Without Carboplatin and Bevacizumab in Triple-Negative Breast Cancer. J Clin Oncol. 2022;40:1323–34. doi: 10.1200/JCO.21.01506. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 45. Hegde PS, Chen DS. Top 10 Challenges in Cancer Immunotherapy. Immunity. 2020;52:17–35. doi: 10.1016/j.immuni.2019.12.011. [ DOI ] [ PubMed ] [ Google Scholar ] 46. Emens LA, Goldstein LD, Schmid P, et al. The tumor microenvironment (TME) and atezolizumab + nab -paclitaxel (A+nP) activity in metastatic triple-negative breast cancer (mTNBC): IMpassion130. JCO. 2021;39:1006. doi: 10.1200/JCO.2021.39.15_suppl.1006. [ DOI ] [ Google Scholar ] 47. Gardner A, Ruffell B. Dendritic Cells and Cancer Immunity. Trends Immunol. 2016;37:855–65. doi: 10.1016/j.it.2016.09.006. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 48. Marciscano AE, Anandasabapathy N. The role of dendritic cells in cancer and anti-tumor immunity. Semin Immunol. 2021;52:101481. doi: 10.1016/j.smim.2021.101481. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 49. Ortiz-Muñoz G, Brown M, Carbone CB, et al. In situ tumour arrays reveal early environmental control of cancer immunity. Nature New Biol. 2023;618:827–33. doi: 10.1038/s41586-023-06132-2. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 50. Spranger S, Bao R, Gajewski TF. Melanoma-intrinsic β-catenin signalling prevents anti-tumour immunity. Nature New Biol. 2015;523:231–5. doi: 10.1038/nature14404. [ DOI ] [ PubMed ] [ Google Scholar ] 51. Ye Y, Gaugler B, Mohty M, et al. Plasmacytoid dendritic cell biology and its role in immune-mediated diseases. Clin Transl Immunology. 2020;9:e1139. doi: 10.1002/cti2.1139. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 52. Le Mercier I, Poujol D, Sanlaville A, et al. Tumor promotion by intratumoral plasmacytoid dendritic cells is reversed by TLR7 ligand treatment. Cancer Res. 2013;73:4629–40. doi: 10.1158/0008-5472.CAN-12-3058. [ DOI ] [ PubMed ] [ Google Scholar ] 53. Vermi W, Bonecchi R, Facchetti F, et al. Recruitment of immature plasmacytoid dendritic cells (plasmacytoid monocytes) and myeloid dendritic cells in primary cutaneous melanomas. J Pathol. 2003;200:255–68. doi: 10.1002/path.1344. [ DOI ] [ PubMed ] [ Google Scholar ] 54. Hartmann E, Wollenberg B, Rothenfusser S, et al. Identification and functional analysis of tumor-infiltrating plasmacytoid dendritic cells in head and neck cancer. Cancer Res. 2003;63:6478–87. [ PubMed ] [ Google Scholar ] 55. Zou W, Machelon V, Coulomb-L’Hermin A, et al. Stromal-derived factor-1 in human tumors recruits and alters the function of plasmacytoid precursor dendritic cells. Nat Med. 2001;7:1339–46. doi: 10.1038/nm1201-1339. [ DOI ] [ PubMed ] [ Google Scholar ] 56. Oshi M, Newman S, Tokumaru Y, et al. Plasmacytoid Dendritic Cell (pDC) Infiltration Correlate with Tumor Infiltrating Lymphocytes, Cancer Immunity, and Better Survival in Triple Negative Breast Cancer (TNBC) More Strongly than Conventional Dendritic Cell (cDC) Cancers (Basel) 2020;12:3342. doi: 10.3390/cancers12113342. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 57. Moffitt JR, Hao J, Wang G, et al. High-throughput single-cell gene-expression profiling with multiplexed error-robust fluorescence in situ hybridization. Proc Natl Acad Sci USA. 2016;113:11046–51. doi: 10.1073/pnas.1612826113. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 58. He S, Bhatt R, Brown C, et al. High-plex imaging of RNA and proteins at subcellular resolution in fixed tissue by spatial molecular imaging. Nat Biotechnol. 2022;40:1794–806. doi: 10.1038/s41587-022-01483-z. [ DOI ] [ PubMed ] [ Google Scholar ] 59. Stringer C, Wang T, Michaelos M, et al. Cellpose: a generalist algorithm for cellular segmentation. Nat Methods. 2021;18:100–6. doi: 10.1038/s41592-020-01018-x. [ DOI ] [ PubMed ] [ Google Scholar ] 60. Greenwald NF, Miller G, Moen E, et al. Whole-cell segmentation of tissue images with human-level performance using large-scale data annotation and deep learning. Nat Biotechnol. 2022;40:555–65. doi: 10.1038/s41587-021-01094-0. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 61. Danaher P, Zhao E, Yang Z, et al. Insitutype: likelihood-based cell typing for single cell spatial transcriptomics. Bioinformatics. 2022 doi: 10.1101/2022.10.19.512902. Preprint. [ DOI ] 62. Danaher P, Paul D, Wang P. Covariance-based analyses of biological pathways. Biometrika. 2015;102:533–44. doi: 10.1093/biomet/asv013. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 63. Budczies J, Klauschen F, Sinn BV, et al. Cutoff Finder: a comprehensive and straightforward Web application enabling rapid biomarker cutoff optimization. PLoS ONE. 2012;7:e51862. doi: 10.1371/journal.pone.0051862. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 64. Patro R, Duggal G, Love MI, et al. Salmon provides fast and bias-aware quantification of transcript expression. Nat Methods. 2017;14:417–9. doi: 10.1038/nmeth.4197. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 65. Benjamini Y, Yekutieli D. The control of the false discovery rate in multiple testing under dependency. Ann Statist. 2001;29:1165–88. doi: 10.1214/aos/1013699998. [ DOI ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Supplementary Materials online supplemental file 1 jitc-14-6-s001.pdf (4.3MB, pdf) DOI: 10.1136/jitc-2025-014020 online supplemental file 2 jitc-14-6-s002.jpg (2.7MB, jpg) DOI: 10.1136/jitc-2025-014020 Data Availability Statement Data are available on reasonable request. Data may be obtained from a third party and are not publicly available. Articles from Journal for Immunotherapy of Cancer are provided here courtesy of BMJ Publishing Group ACTIONS View on publisher site PDF (5.0 MB) Cite Collections Permalink RESOURCES Similar articles Cited by other articles Links to NCBI Databases Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top

Related documents

• Identification of cellular determinants of responseto systemic treatment in triple-negative breast carcinomas#1166380
• Efficacy and Safety of Extended-Dose Interval Immunotherapy Versus Standard-Dose Interval Immunotherapy for Advanced Triple-Negative Breast Cancer#1112187
• The Intestinal Microbiome in Triple Negative Breast Cancer Treated with Immunotherapy#303723
• Image 2_Identification of a novel inflammatory mast cell subpopulation in triple-negative breast cancer.tif#1145924
• Resistance mechanisms and countermeasures in CLDN18.2-positive tumors: from molecular networks to clinical practice.#678536