Part V · Immunology and the tumor microenvironment · Chapter 22

Cellular architecture of the tumor microenvironment

The fibroblast compartment suppresses immunity and organises it. Which one depends on state, not on abundance.

1 · Cancer-associated fibroblast subsets and their opposing functions

Fibroblast is the name of a compartment rather than of a cell type with one function. Flow cytometry using a six-marker panel separated fibroblasts in human breast cancer into four subsets, designated CAF-S1 through CAF-S4¹. Two of them, CAF-S1 and CAF-S4, are myofibroblastic. Both accumulate preferentially in triple-negative disease.

Only one of them is immunosuppressive, and the mechanism is stepwise. CAF-S1 secretes CXCL12, which attracts CD4-positive CD25-positive T lymphocytes. It retains those cells through OX40L, PD-L2 and JAM2. It promotes their differentiation towards a CD25-positive FOXP3-positive phenotype through B7H3, CD73 and DPP4. It then increases the capacity of the resulting regulatory cells to inhibit effector T cell proliferation¹.

CAF-S4 does none of this, despite being myofibroblastic and despite being enriched in the same subtype¹.

That contrast is the argument of the chapter. Fibroblast abundance is not the variable that matters. Fibroblast state is. A strategy that depletes fibroblasts as a class removes CAF-S1 and CAF-S4 together, and only one of them was the problem.

One distribution point deserves naming early. The most immunosuppressive fibroblast subset is enriched in the subtype with the most immune infiltrate. The luminal tumour has far fewer lymphocytes to suppress², and its stromal problem is a different one. Fibroblasts and endocrine resistance, rewiring the ER response describes it.

2 · Named CAF states, myCAF, iCAF, vCAF, mCAF, and apCAF

A second naming system exists, derived from single-cell transcriptomics rather than from surface marker flow cytometry. The two systems are frequently used in the same sentence. They do not map onto one another cleanly, and the reason is methodological rather than biological.

Cross-species single-cell analysis of pancreatic ductal adenocarcinoma corroborated myofibroblastic and inflammatory states and defined their gene signatures³. The same work described a third population expressing MHC class II and CD74 without classical costimulatory molecules. That population activated CD4 T cells in an antigen-specific model system, which is why it was named antigen-presenting.

In breast cancer, single-cell RNA sequencing of 768 mesenchymal transcriptomes from a genetically engineered mouse model defined three subpopulations with distinct spatial origins⁴. The origins were the perivascular niche, the mammary fat pad, and the transformed epithelium itself. The corresponding gene profiles held independent prognostic value in human cohorts through their association with metastatic disease.

The measurement point is the whole difficulty. One taxonomy comes from surface protein flow cytometry on human breast tumours. Another comes from transcriptomes of mouse mammary mesenchyme and of human pancreatic tissue. States defined by different assays, on different tissues, in different species are not the same objects, even where the labels overlap.

Read the names as descriptions of observed states rather than as a settled lineage classification. A paper reporting myCAF and a paper reporting CAF-S1 may or may not be reporting the same cells, and no cross-walk between the two systems has been established.

3 · Fibroblast-mediated immune exclusion

Suppression and exclusion are different mechanisms and merging them causes trouble. Suppression means a T cell reaches the tumour and cannot function there. Exclusion means it never arrives.

Single-cell analysis of more than 19,000 CAF-S1 fibroblasts from breast cancer identified eight clusters⁵. Two of them were indicative of primary resistance to immunotherapy. Cluster 0 was characterised by extracellular matrix proteins and cluster 3 by TGF-β signalling, and both were myofibroblastic. The matrix cluster upregulated PD-1 and CTLA-4 protein on regulatory T lymphocytes.

Exclusion itself was demonstrated most clearly outside breast cancer. In a large cohort of patients with metastatic urothelial cancer treated with atezolizumab, lack of response was associated with a signature of TGF-β signalling in fibroblasts⁶. In those tumours CD8 T cells sat in collagen-rich peritumoural stroma rather than in tumour parenchyma. In a mouse model reproducing that phenotype, co-administration of TGF-β blockade with anti-PD-L1 allowed T cells into the tumour centre and produced regression.

Transferring that to breast cancer requires care. The human cohort was urothelial and the interventional experiment was murine. What breast cancer contributes is the fibroblast state data above, which is descriptive rather than causal.

The distinction predicts different failures in clinic. Releasing an inhibitory receptor on a cell that never entered the tumour accomplishes nothing, whatever the receptor is doing elsewhere.

4 · Fibroblasts and endocrine resistance, rewiring the ER response

Here the stroma acts on the luminal tumour, and it acts on the receptor rather than on the immune infiltrate.

Human oestrogen-receptor-positive breast tumours contain fibroblasts separable by CD146 expression⁷. In coculture, CD146-negative fibroblasts reduced oestrogen receptor expression in tumour cells, reduced oestrogen-dependent proliferation, and reduced sensitivity to tamoxifen. CD146-positive fibroblasts sustained receptor expression and preserved tamoxifen sensitivity.

Two things follow, and the second is the more useful.

The biological point is that endocrine sensitivity is not entirely a property of the tumour cell. It is a property of the tumour cell in a particular stromal context. The receptor-level mechanisms of endocrine resistance are set out in Estrogen receptor signaling and endocrine resistance. This is a route to the same clinical outcome that requires no alteration in the tumour genome at all.

The measurement point is that a resistance mechanism living in the stroma is invisible to every assay in routine use. Tumour cell sequencing will not find it, because there is nothing to find in the tumour cell. Receptor immunohistochemistry may capture its consequence, because receptor expression falls, without indicating the cause. A falling receptor score at progression therefore has at least two explanations, and Receptor discordance and conversion sets out how they are told apart.

This is also the form the luminal deficit takes in this chapter. The luminal tumour's stromal problem is not that its fibroblasts suppress an immune response. It is that they modulate the dependency the treatment is aimed at.

6 · Neutrophil extracellular traps and their role in awakening tumor cells

Neutrophils release decondensed chromatin studded with granule proteins. The resulting extracellular traps are an antimicrobial mechanism. They have a second effect that matters to this book.

In mouse models, sustained lung inflammation produced either by tobacco smoke exposure or by nasal instillation of lipopolysaccharide converted disseminated dormant cancer cells into aggressively growing metastases⁸. Trap formation was required for the conversion. Two trap-associated proteases, neutrophil elastase and matrix metalloproteinase 9, were the effectors.

The mechanism is worth stating precisely, because it explains how an inflammatory stimulus with no relation to the tumour can act on a cancer cell. The proteases remodel laminin in the extracellular matrix. The remodelled protein then engages integrin signalling in the dormant cell and returns it to proliferation⁸.

That is a route by which an infection, a period of chronic inflammation or a sustained exposure can end dormancy. Dormancy, residual disease, and late recurrence develops the consequences for late recurrence, where it is one of the few mechanisms that explains why a recurrence happens at the moment it happens.

Two limits belong with it. The work is murine throughout. And it establishes that traps can awaken dormant cells, not that they are the usual reason human breast cancer recurs after many years.

8 · Adipocytes and the mammary adipose environment

The breast is largely fat, so an invasive breast tumour meets adipose tissue almost immediately. The adipocytes it meets do not stay passive.

Coculture of tumour cells with mature adipocytes produced adipocytes that lost lipid and lost adipocyte markers⁹. The same cells gained expression of proteases including matrix metalloproteinase 11, and of proinflammatory cytokines including interleukin 6 and interleukin 1β. The altered adipocytes increased the invasive capacity of the tumour cells, and interleukin 6 was required for that effect. The same altered phenotype was identified in human breast tumours by immunohistochemistry and quantitative PCR. Interleukin 6 levels in tumour-surrounding adipocytes were highest in larger tumours and in tumours with nodal involvement⁹.

Two further properties of this interface are developed elsewhere. Adipose tissue is the principal site of oestrogen synthesis after the menopause, which makes the stroma a source of ligand rather than only a mechanical environment. Obesity alters the compartment systemically, and Host and tumor interactions takes that up.

The interface therefore does different work in the two subtypes. In triple-negative disease it supplies an inflammatory, invasion-promoting environment. In luminal disease it does that and also supplies the hormone the tumour depends on.

9 · Extracellular matrix composition and tissue mechanics

Stiffness is a signal rather than a background property.

In mouse mammary models, collagen crosslinking stiffened the matrix, promoted focal adhesion formation, enhanced PI3 kinase activity and induced invasion by an oncogene-initiated epithelium¹⁰. Reducing lysyl oxidase-mediated crosslinking prevented fibrosis in the MMTV-Neu model, decreased PI3 kinase activity and lowered tumour incidence. Stiffness in those experiments was not a consequence of malignancy. It was upstream of it.

Organisation matters as much as amount. One arrangement has a name: collagen in straightened bundles oriented perpendicular to the tumour boundary is tumour-associated collagen signature 3. In biopsied tissue from 196 patients it carried hazard ratios between 3.0 and 3.9 for disease-specific and disease-free survival¹¹. The association was independent of grade, size, oestrogen and progesterone receptor status, HER2 status, nodal status and subtype.

The measurement point explains why this is not on a pathology report. The signature was scored by second harmonic generation imaging, and picrosirius staining is an alternative. Neither is part of routine practice, and no reproducibility exercise comparable to the lymphocyte ring studies has established how consistently observers score it.

The connection to immunology is direct rather than analogical. A dense, aligned, crosslinked matrix is the physical substrate of the exclusion phenotype in Fibroblast-mediated immune exclusion. It is also the tissue property that raises interstitial pressure and impedes drug delivery, which Functional states of the microenvironment develops.

11 · Tumor and stroma communication

The strongest evidence that stromal signalling carries independent information is that stromal gene expression alone predicts outcome.

Laser capture microdissection of tumour stroma from 53 primary breast tumours produced a stroma-derived prognostic predictor¹². It stratified outcome independently of standard clinical prognostic factors and of published expression-based predictors, and it identified poor-outcome patients across several clinical subtypes including node-negative disease. The genes carrying its prognostic weight represented immune responses, angiogenesis and hypoxia. None of them came from a tumour cell.

Single-cell and spatially resolved profiling has since shown that these interactions are organised rather than diffuse. Stromal and immune cells occupy defined niches within the tumour, and deconvolution of large cohorts using single-cell signatures separates tumours into ecotypes with distinct cellular compositions and distinct outcomes¹³. Spatial organization and what it adds takes up what that resolution does and does not add.

Which returns the chapter to its tension. The compartment that recruits and sustains regulatory T cells¹, and that excludes cytotoxic cells behind a matrix⁵, is also the compartment that builds lymphoid tissue. In a murine melanoma model, fibroblasts with lymphoid tissue organiser characteristics organised tertiary lymphoid structures, and their expansion depended on CXCL13-mediated recruitment of B cells¹⁴. Those structures are the ones associated with better outcome in B cells and tertiary lymphoid structures.

References

1. Costa A, Kieffer Y, Scholer-Dahirel A, et al. Fibroblast heterogeneity and immunosuppressive environment in human breast cancer. Cancer Cell 2018 33:463-479. PMID 29455927
2. Stanton SE, Adams S, Disis ML. Variation in the incidence and magnitude of tumor-infiltrating lymphocytes in breast cancer subtypes: a systematic review. JAMA Oncol 2016 2:1354-1360. PMID 27355489
3. Elyada E, Bolisetty M, Laise P, et al. Cross-species single-cell analysis of pancreatic ductal adenocarcinoma reveals antigen-presenting cancer-associated fibroblasts. Cancer Discov 2019 9:1102-1123. PMID 31197017
4. Bartoschek M, Oskolkov N, Bocci M, et al. Spatially and functionally distinct subclasses of breast cancer-associated fibroblasts revealed by single cell RNA sequencing. Nat Commun 2018 9:5150. PMID 30514914
5. Kieffer Y, Hocine HR, Gentric G, et al. Single-cell analysis reveals fibroblast clusters linked to immunotherapy resistance in cancer. Cancer Discov 2020 10:1330-1351. PMID 32434947
6. Mariathasan S, Turley SJ, Nickles D, et al. TGF-beta attenuates tumour response to PD-L1 blockade by contributing to exclusion of T cells. Nature 2018 554:544-548. PMID 29443960
7. Brechbuhl HM, Finlay-Schultz J, Yamamoto TM, et al. Fibroblast subtypes regulate responsiveness of luminal breast cancer to estrogen. Clin Cancer Res 2017 23:1710-1721. PMID 27702820
8. Albrengues J, Shields MA, Ng D, et al. Neutrophil extracellular traps produced during inflammation awaken dormant cancer cells in mice. Science 2018 361:eaao4227. PMID 30262472
9. Dirat B, Bochet L, Dabek M, et al. Cancer-associated adipocytes exhibit an activated phenotype and contribute to breast cancer invasion. Cancer Res 2011 71:2455-2465. PMID 21459803
10. Levental KR, Yu H, Kass L, et al. Matrix crosslinking forces tumor progression by enhancing integrin signaling. Cell 2009 139:891-906. PMID 19931152
11. Conklin MW, Eickhoff JC, Riching KM, et al. Aligned collagen is a prognostic signature for survival in human breast carcinoma. Am J Pathol 2011 178:1221-1232. PMID 21356373
12. Finak G, Bertos N, Pepin F, et al. Stromal gene expression predicts clinical outcome in breast cancer. Nat Med 2008 14:518-527. PMID 18438415
13. Wu SZ, Al-Eryani G, Roden DL, et al. A single-cell and spatially resolved atlas of human breast cancers. Nat Genet 2021 53:1334-1347. PMID 34493872
14. Rodriguez AB, Peske JD, Woods AN, et al. Immune mechanisms orchestrate tertiary lymphoid structures in tumors via cancer-associated fibroblasts. Cell Rep 2021 36:109422. PMID 34289373