Single-cell analysis reveals diverse stromal subsets associated with immune evasion in triple-negative breast cancer

Abstract

44The tumour stroma regulates nearly all stages of carcinogenesis. Stromal heterogeneity in 45 human triple-negative breast cancers (TNBCs) remains poorly understood, limiting the 46 development of stromal-targeted therapies. Single cell RNA-sequencing of five TNBCs 47 revealed two cancer-associated fibroblast (CAF) and two perivascular-like (PVL) 48 subpopulations. CAFs clustered into two states, the first with features of myofibroblasts and 49 the second characterised by high expression of growth factors and immunomodulatory 50 molecules. PVL cells clustered into two states consistent with a differentiated and immature 51 phenotype. We showed that these stromal states have distinct morphologies, spatial 52 relationships and functional properties in regulating the extracellular matrix. Using cell-53 signalling predictions, we provide evidence that stromal-immune crosstalk acts via a diverse 54 array of immunoregulatory molecules. Importantly, the investigation of gene signatures from 55 inflammatory-CAFs and differentiated-PVL cells in independent TNBC patient cohorts 56 revealed strong associations with cytotoxic T-cell dysfunction and exclusion, respectively. 57Such insights present promising candidates to further investigate for new therapeutic 58 strategies in the treatment of TNBCs. 59 60 61Heterotypic interactions between stromal, immune and malignant epithelial cells play 62 important roles in solid tumour progression and therapeutic response. Cancer-associated 63 fibroblasts (CAFs) play an integral part in the tumour microenvironment (TME), and can 64 influence many aspects of carcinogenesis including extracellular matrix (ECM) remodelling, 65 angiogenesis, cancer cell proliferation, invasion, inflammation, metabolic reprogramming 66 and metastasis [1]. Recent studies have described roles for CAFs in mediating immune 67 suppression and chemo-resistance, establishing CAFs as novel and attractive targets for 68 anti-cancer therapies in advanced breast cancer [2-6]. Despite their well-described roles in 69 cancer biology, CAFs remain enigmatic: limited studies suggest phenotypic heterogeneity, 70 plasticity and functional diversity, with both tumour-promoting and tumour-suppressive 71 properties [1]. The multi-faceted nature of CAFs suggests that they are comprised of diverse 72 subpopulations, and an improved understanding of stromal heterogeneity may explain how 73 CAFs contribute to the dynamic complexity and functional malleability of the tumour 74 ecosystem. 75 76CAFs of the tumour parenchyma are routinely studied using a handful of markers including 77 a-smooth muscle actin (a-SMA), fibroblast activation protein (FAP), CD90 (THY-1), platelet 78 derived growth factor receptor a and b (PDGFRa and PDGFRb), podoplanin (PDPN) and 79 fibroblast specific protein 1 (FSP-1, also named S100A4) [1, 7-9]. However, these markers 80 are not necessarily co-expressed, nor specific to the fibroblast lineage [4]. For instance, α-81 SMA not only identifies CAFs with a myofibroblast morphology but also serves as a genera...

Tumour Stroma Ratio Assessment Using Digital Image Analysis Predicts Survival in Triple Negative and Luminal Breast Cancer

“…Detailed single cell RNA-seq (scRNA-seq) data are now emerging regarding the heterogenous nature of the stromal cell population, which provides cellular resolution not previously available from bulk sequencing studies [ 37 , 38 ]. One study of human TNBC identified two distinct populations of CAFs: myofibroblastic (myCAFs) and inflammatory (iCAFs) and two populations of peri-vascular-like cells (PVL) [ 38 ]. Through analysis of large RNA-Seq datasets, the authors showed that PVL cells were associated with TILs exclusion, perhaps explaining our observation of a positive correlation between TSR and TIL.…”

DeCompress: tissue compartment deconvolution of targeted mRNA expression panels using compressed sensing

“…Then, we assess performance of using compartment-specific proportions in downstream analyses of breast cancer outcomes and gene regulation. Using TCGA breast cancer (TCGA-BRCA) expression as a training set, we iteratively searched for compartment-specific features using TOAST + NMF ( 32 ) (Step 1 in Figure 1 ) and included canonical compartment markers for guidance using a priori knowledge ( 29 , 96 , 97 ) (see Materials and Methods). After expanding the targeted CBCS expression to these genes using a compressed sensing model based on elastic net, lasso, or ridge regression, we estimated compartment proportions.…”

“…Previous groups have emphasized reliance on a priori knowledge for deconvolving well-studied tissues, such as blood and brain ( 113 , 114 ). However, diseased tissues, like bulk cancerous tumors, especially in understudied subtypes or populations, are more difficult to deconvolve due to the similarity between compartments, many of which may be rare or reflect transient cell states ( 29 , 94 , 115 , 116 ). For this reason, we included several data-driven approaches for estimating the number of compartments from variation in the gene expression and recommended applying prior domain knowledge about the tissue of interest.…”

“…Given the advent of single-cell technologies and studies into cell trajectories, the concept of cell types in bulk tissue has been debated ( 28 ). Especially in perturbed or diseased tissues, like cancerous tumors, individual cells may be present in different states or various cells of possibly different identities may contribute, in aggregate, to the same biological process and have similar molecular profiles ( 29–31 ). While previous reference-free methods rely on searching the feature space for compartment-specific molecular features from the entire transcriptome and thus require a large feature space ( 22 , 25 , 32 ), reference-free deconvolution methods can, with fewer assumptions, identify tissue compartments or isolated units of a tissue that represent either a biological process or a cell type ( 33 ).…”

DeCompress: tissue compartment deconvolution of targeted mRNA expression panels using compressed sensing

“…Previous groups have emphasized reliance on a priori knowledge for deconvolving well-studied tissues, such as blood and brain (106,107). However, diseased tissues, like bulk cancerous tumors, especially in understudied subtypes or populations, are more difficult to deconvolve due to the similarity between compartments, many of which may be rare or reflect transient cell states (30,91,108,109). For this reason, we included several data-driven approaches in estimating the number of compartments from variation in the gene expression and recommended applying prior domain knowledge about the tissue of interest.…”