Anniina Färkkilä

Oxidative stress response (OSR) is suggested as a therapeutic target in triple negative breast cancer (TNBC). However, the cell type specificity and spatial distribution of OSR genes and proteins remain poorly defined. We re-analyzed public datasets, demostrating high OSR gene expression in TNBC tumor cells. We generated a multiplex immunofluorescence dataset of 345 TNBC samples that showed OSR proteins GCLC and NQO1 are differentially expressed across tumor, immune, and stromal cells, and elevated in inflamed regions. We identified a mesenchymal-like tumor cell state characterized by co-expression of GCLC and vimentin (GCLC⁺VIM⁺), with increased proliferation and reduced anastasis-associated signatures, suggesting a vulnerability to chemotherapy. Consistently, TNBC cell lines with higher GCLC expression showed increased sensitivity to cisplatin. Moreover, patients with higher proportions of GCLC⁺VIM⁺ tumor cells had longer overall survival. Our findings indicate that OSR proteins are expressed in spatially and phenotypically distinct tumor microenvironments, and that the GCLC⁺VIM⁺ tumor cell state is associated with favorable outcomes in TNBC.

High-grade serous ovarian carcinoma (HGSC) responds poorly to immune checkpoint blockade, partly due to a macrophage-dominated immunosuppressive microenvironment. We integrated single-cell spatial proteomics and spatial transcriptomics across 50 HGSC tumors and applied SPACEstat to resolve higher-order immune communities and their transcriptional programs. We identified six immune community types, with macrophage-dominated Myelonets representing the predominant spatial pattern of immune organisation. In chemotherapy-exposed tumors, Myelonets showed coordinated lipid metabolism–immunosuppression and inflammation–MHC-II macrophage transcriptional programs, with SPP1, C1Q, VEGF, MMPs, and CCL18 linked to immunosuppressive states and fibroblasts emerging as key mediators of macrophage communication. Chemotherapy contracted large Myelonets while increasing CD8+ T-cell organization into Lymphonets. Persistent macrophage dominance within Myelonets was associated with adverse outcomes among patients who achieved a complete response to treatment. Together, we identify Myelonets as clinically relevant, multicellular immunoregulatory niches sustained by spatiotemporally coordinated macrophage programs and stromal crosstalk.

Supplementary Figure S1: a Assignment of the tumors into the four different molecular profiles. Supplementary Figure S2: Consensus clustering of cancer cell - ConsensusClusterPlus result. Supplementary Figure S3: Cancer cell clusters. Supplementary Figure S4: a Heatmaps displaying the median of the median proportions of each immune population per tumor across molecular profiles. Supplementary Figure S5: a-b Kaplan-Meier plots display the probability of survival of HRDs (BRCAloss and HRD) and HRPs (CCNE1amp, HRP) groups. Supplementary Figure S6: a Box plots showing the estimated proportions of immune populations in OVA-TCGA. Supplementary Figure S7: a Flowchart representing the strategy used to assign the cells from whole slide t-CycIF images into different populations according to the expression of the indicated markers. Supplementary Figure S8: a Volcano plot displays differential gene expression results comparing tumors with high versus low cancer cell-specific RNA expression of the MHCII gene HLA_DPB1 in OVA-TCGA. Supplementary Figure S9: a Circos plots display ligand–receptor pairs predicted and prioritized by Multinichenetr – both immune–tumor and tumor–tumor cell interactions distinguishing high and low MHCII tumor expression in scRNA-seq dataset. Supplementary Figure S10: a,b Violin plots showing HLA-DR fluorescent intensity in cancer, immune, and myeloid populations within iPDCs, stratified by MHCII tumor status. Supplementary Figure S11: a Scatter plot displaying the proportion of cancer cells identified in t-CycIF images compared to the proportion annotated by an expert pathologist using H&E images from the adjacent slides in whole slide samples.

Table S1: Description of clinical information in the multidimensional cohort. For seven tumors, there was not matched clinical information. Table S2: "List of antibody panel used for t-CycIF in the multidimensional cohort (N = 265). Column description: Cycle = t-CycIF cycle in which used; Antibody = Target of the antibody; Company = Name of provider company; Cat = Catalog number by company; Conc = Concentration used; Purpose = Type of cell of function to identify in experiment." Table S3: The median of the median proportions of each immune population per tumor across molecular profiles. First, the median proportion of each immune population was calculated per tumor (across cores); then, for each molecular profile, the median of these tumor-level medians was computed. Only tumors with more than one core were used for the analysis. N of each group is reported in Supplementary Fig. S4a. Table S4: "Pearson correlation values of the median immune infiltration population across the tumors in the multidimensional cohort. Firstly, per population a median immune infiltration was calculated per tumor across the cores. Then correlation of the median infiltration values was calculated across tumors for each pair of immune populations. Only tumors with more than one core were used for the analysis (N = 251)" Table S5: False discovery rate (FDR) of the correlation test of the median immune infiltration correlation (Table S4). Table S6: Pearson correlation values of the median RCN proportion across the tumors in the multidimensional cohort. Firstly, per RCN a median proportion was calculated per tumor across the cores. Then correlation of the median proportion was calculated across tumors for each pair of RCN. Only tumors with more than one core were used for the analysis (N = 251). Table S7: False discovery rate (FDR) from the correlation test of the median RCN proportion across tumors (Table S6). Table S8: Top predictive markers (from Supplementary Fig. 5candd) for distinguishing long vs short overall survival (OS) HRDs and HRPs Cancer or Immune-Stromal cells, with reflected across corresponding predicted class median value of min-max scaled expression values after CEFIIRA preprocessing to highlight differences within the cell population. In addition, the number of cells for each comparison are in corresponding columns. P-values were calculated using the Mann-Whitney U test and adjusted using the Benjamini-Hochberg method. Table S9: Quantified variability of CEFIIRA input features for OS models within HRP-like and HRD-like tumors. Entropy indices in HRD and HRP tumors were calculated using Shannon entropy and Jensen–Shannon (JS) divergence. Each row represents the single-cell features used in the CEFIIRA models. The entropy difference between HRD-like and HRP-like tumors was assessed using a permutation test (50 folds), and the resulting p-value is reported. Table S10: List of overexpressed genes in OVA-TCGA when comparing tumors with high cancer-specific deconvoluted MHCII (HLA_DPB1) expression (n = 80) compared samples with to low cancer-specific deconvoluted MHCII (HLA_DPB1) expression (n = 79) using DESeq2. For each gene is show the mean value, fold change, p.value of test and FDR. Table S11: "Results of differential expression analysis comparing tumors with a high proportion of MHCII + cancer cells in t-CycIF (N = 66) to those with a low proportion of MHCII + cancer cells (N = 67) using the multidimensional cohort. Gene expression was profiled using NanoString technology. The full list of profiled genes (n = 335) is presented. Significance was assessed using a t-test, and p-values were adjusted for multiple testing using the false discovery rate (FDR). Genes were considered differentially expressed based on the following criteria: log2 fold change ≥1 or ≤ –1 and FDR <10⁻³." Table S12: "Summary table of analysis performed on deconvoluted bulk RNAseq for 35 genes enriched in immunopeptidomic samples with a high proportion of MHCII + cancer cells and filtered based on healthy tissue HLA-II ligandome across three patient groups: with low (four samples), mid (four samples), and high (four samples) proportion of MHCII + cancer cells. Includes: - Median values of scaled expression for tumor and immune-stromal fractions in high group. - Median z-scores for tumor and immune-stromal fractions in high group. - Specificity of gene expression to the tumor fraction in the high group, assessed by comparing the median scaled expression in the tumor fraction with the median scaled expression across both the tumor and immune–stromal fractions combined. This yields a value between 0 and 1, where values closer to 1 indicate expression predominantly in tumor fraction. - Fold change of tumor fraction expression in high vs. the mean tumor fraction expression in low and mid groups." Table S13: Granzyme B response of CD8⁺ T cells in iPDCs following treatment with pembrolizumab (PE, 10 µg/mL), tiragolumab (TI, 10 µg/mL), or their combination (PE + TI). Responses per sample were calculated as the mean difference between treated (days 2–5) and control (d0) samples and analyzed separately for MHCII-high and MHCII-low cancer sample groups. The mean log2FC across samples in each group is reported in the corresponding column. Statistical significance was assessed using the Wilcoxon rank-sum test. Averaged changes and pembrolizumab-induced (PE) changes in Granzyme B response are also shown. Table S14: "List of antibody panel used for t-CycIF in the whole slide validation cohort (N = 15). Column description: Cycle = t-CycIF cycle in which used; Antibody = Target of the antibody; Company = Name of provider company; Cat = Catalog number by company; Conc = Concentration used; Purpose = Type of cell of function to identify in experiment." Table S15: "Logic table to define cell phenotypes using TRIBUS. A value of one (1) is considered a marker expected in the cell...