Professional Documents
Culture Documents
*Correspondence to: PAW Edwards, Department of Pathology, University of Cambridge, Tennis Court Road, Cambridge CB2 1QP, UK.
E-mail: pawe1@cam.ac.uk
Abstract
Much of tumour cell biology and the tumour microenvironment may be normal wound-healing responses as a
consequence of the disruption of tissue structure. This is why tumours resemble wounds, and many features of the
tumour microenvironment, such as the epithelial–mesenchymal transition, cancer-associated fibroblasts, and
inflammatory infiltrates, may largely be normal responses to abnormal tissue structure, not an exploitation of
wound-healing biology.
© 2023 The Author. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Brit-
ain and Ireland.
Keywords: tumour microenvironment; wound healing; epithelial–mesenchymal transition; cancer-associated fibroblast; tumour-infiltrating
lymphocytes; malignancy; cancer
Figure 1. Epithelial tumours viewed as wounded epithelia. Epithelial tumours are disrupted epithelium, and will therefore mount at least
some of the responses that a normal epithelium would when disrupted. Left: intact epithelium is a continuous sheet with an electrical poten-
tial across it [8], salt differences (not shown), and ligands produced at one face (shown green on apical face) that cannot reach receptors on
the other face (shown yellow on basolateral face) [9]. Grey, basement membrane. Right: in malignant epithelial tumours, the epithelial sheet
is disrupted, and the basement membrane may or may not be breached [10]. This activates wound-healing responses in the epithelium, which
include signals (orange arrows) to stromal cells, represented to the right, in blue, by a macrophage, lymphocyte, and fibroblast; signals from
the alerted stroma back to the epithelium (blue arrows); electric currents; and interaction between ligands and receptors [9,11].
have suggested that tumours exploit wound-healing response to breach of the epithelial sheet, and other
programmes to achieve properties such as invasiveness, changes in structure or behaviour.
rather than the wound-healing programme being an
inevitable consequence of the disruption of the epithe-
lium. In particular, Dvorak, who in 1986 proposed that
The ‘epithelial–mesenchymal transition’ (EMT)
a cancer is “a wound that does not heal”, also “proposed
that tumours co-opted the wound-healing response to
induce the stroma they required” (my italics), and The ‘epithelial–mesenchymal transition’ is a set of
“tumours disguise themselves as wounds and call upon phenotypic changes in epithelial tumour cells towards
the host to ‘heal’ them” [4]. A recent review [14], for a more mesenchymal phenotype, such as loss of polar-
example, states “Cancer-associated fibroblasts (CAFs) ity and expression of the mesenchymal protein
vimentin [1]. It is part of normal wound healing but
… can adopt … [a phenotype] comparable to … fibro-
has, controversially, also been invoked as a mecha-
blasts in … wound-healing”, and apparently does not
nism – indeed a driver – of malignancy, invasion,
consider that these fibroblasts are indeed wound-healing
and metastasis [1,12].
fibroblasts because the tissue is mounting a wound-
We can re-interpret EMT as an inevitable conse-
healing response. quence of disrupting epithelium, not something that
These contrasting concepts are not, of course, mutu- tumour cells exploit to invade and metastasize [1,12].
ally exclusive: some wound-healing responses may be
consequences of disruption, while others may be
invoked by cancer mutations.
Motility, proteases, and angiogenesis
Applying the idea It is often assumed that invasion and metastasis require
enhanced motility [15], but normal epithelial cells at a
Perhaps, then, we could understand the behaviour of wound are highly motile [12] – skin keratinocytes at a
malignant tumour cells by reference to what normal cells wound edge can move around 50 μm/day [16], much
would be doing in similar circumstances – which we can faster than a typical tumour infiltrates tissue. The
call ‘wound healing’. In an epithelial tumour, the epithe- observed high motility of some tumour cells may be their
lial cells and stroma will be changing behaviour in normal response to disorganisation.
© 2023 The Author. The Journal of Pathology published by John Wiley & Sons Ltd J Pathol 2023; 260: 1–4
on behalf of The Pathological Society of Great Britain and Ireland. www.pathsoc.org www.thejournalofpathology.com
10969896, 2023, 1, Downloaded from https://pathsocjournals.onlinelibrary.wiley.com/doi/10.1002/path.6070 by Cochrane Canada Provision, Wiley Online Library on [17/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Re-interpreting tumour behaviour 3
Epithelial wound healing also includes secretion of whether there were differences between EMT in wound
matrix metalloproteases to break down clot, and initiates healing and in cancers.
angiogenesis [11,12]. This argument should also help us to understand how
mutations lead to cancer, by separating out the compo-
nents of tumour behaviour that are truly abnormal, and
hence the product of mutation. This would also inform
Dense stroma and cancer-associated
therapy, which might otherwise target components of nor-
fibroblasts (CAFs) mal wound healing without much benefit to the patient.
It also generates new questions and redirects old ones.
Fibrosis, i.e. scarring, the laying down of collagen-rich A recurring question about the similarity between
tissue by fibroblasts, occurs at the breach of an epithe- tumours and wounds has been why wound-associated
lium (think of scars forming where skin is cut). The changes are transient in wounds but persist in tumours
dense, collagen-rich stroma seen in many tumours, (although reversing malignancy can reverse the changes
called ‘desmoplasia’, resembles chronic fibrosis [4,17], [19]). This becomes the wrong question if changes are
so could simply be the response to broken epithelium. caused by disorganisation, which persists.
The various fibroblast-like cells found in tumour stroma, This argument also focuses attention on what drives
‘cancer-associated fibroblasts’, are similar to cells disorganisation, and, more precisely, how mutations dis-
observed in wounds [2]. rupt organisation. Is unlimited survival and proliferation
alone enough to bulldoze through normal homeostasis
and disrupt tissue architecture, or does malignancy also
Lymphocyte and macrophage infiltration require alteration of signals that control 3D organisation?
An important question is whether cancer driver muta-
tions directly modify the wound-healing response, for
There is much interest in inflammatory cells in example by enhancing EMT or sending signals to the
tumours, notably ‘tumour-infiltrating lymphocytes’ stroma. This seems likely. A component of EMT is
(TILs), because of their possible role in an immune down-regulation of E-cadherin/CDH1, and inactivation
response to the tumour [3]. But any breach in an epithe- of CDH1 is a driver mutation, for example in lobular
lium threatens infection, so some of the inflammatory breast cancer. Oncogenic signalling pathways including
cells (neutrophils, macrophages, or lymphocytes) could the Wnt, TGF-β, and Notch pathways apparently
be summoned to defend against infection, rather than directly affect EMT, even in normal tissues [20]. Acti-
the tumour. Indeed, some TILs are primed to viruses that vating KRAS can modify the ability of epithelial cells
infect epithelia [18]. to correct tissue damage [21]. But whether driver muta-
tions directly modify stromal responses is more difficult
to pin down, because we must distinguish between direct
Differences and variability effects and the indirect effects of breaking the rules of
tissue organisation. Nevertheless, it seems that onco-
Changes seen in tumours will not, of course, exactly genic mutations can cause inflammatory signalling
match those in a wound caused by mechanical damage. [22]. For example, activating KRAS in pancreatic epithe-
For example, tumours do not usually cause lium can reprogramme neighbouring fibroblasts [23],
haemorrhage, which contributes to damage signalling and activated RAS genes can induce senescence, which
via complement and platelet activation [11]. And matrix results in signalling to neighbouring cells [24].
structures such as basement membranes may or may not Dvorak’s [4] original concept, that tumours exploit
be disrupted [10]. wound healing, could still prove to be insightful. Could
The tumour microenvironment will also vary between turning on wound-healing responses permit cells to pro-
tumours because of variations in their growth pattern, liferate and move into inappropriate environments?
anoxia, and necrosis, etc. (as well as any differences in Could, for example, the constitutive activation of
mutations that affect wound healing). For example, lob- Erbb2/Her2 seen in some carcinomas be mimicking the
ular breast cancer cells invade in single file, while many proliferative wound signal, access of apical Nrg1 ligand
other breast cancers are solid masses of cells. to basolateral Erbb2 [9]?
This line of argument may be useful because it may pro- In conclusion, we need to know how much of malignant
vide a framework for understanding tumour behaviour. tumour cell biology is simply programmed in as a nor-
For example, cancer-associated fibroblasts are probably mal wound-healing response, consequent on disruption
best understood by comparison with fibroblasts in of normal tissue architecture. And we need to understand
wounds, as already suggested [2]. We might better how cancers become disorganised, in fact how they
understand the role of EMT in invasion if we knew become malignant.
© 2023 The Author. The Journal of Pathology published by John Wiley & Sons Ltd J Pathol 2023; 260: 1–4
on behalf of The Pathological Society of Great Britain and Ireland. www.pathsoc.org www.thejournalofpathology.com
10969896, 2023, 1, Downloaded from https://pathsocjournals.onlinelibrary.wiley.com/doi/10.1002/path.6070 by Cochrane Canada Provision, Wiley Online Library on [17/05/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
4 PAW Edwards
Acknowledgements 11. Shaw TJ, Martin P. Wound repair at a glance. J Cell Sci 2009; 122:
3209–3213.
12. MacCarthy-Morrogh L, Martin P. The hallmarks of cancer are also the
My work was funded principally by Cancer Research hallmarks of wound healing. Sci Signal 2020; 13: eaay8690.
UK and Breast Cancer Now. 13. Ge Y, Gomez NC, Adam RC, et al. Stem cell lineage infidelity drives
wound repair and cancer. Cell 2017; 169: 636–650.e14.
14. Menezes S, Okail MH, Jalil SMA, et al. Cancer-associated fibroblasts
References in pancreatic cancer: new subtypes, new markers, new targets.
J Pathol 2022; 257: 526–544.
1. Yang J, Antin P, Berx G, et al. Guidelines and definitions for research
15. Welch DR, Hurst DR. Defining the hallmarks of metastasis. Cancer
on epithelial–mesenchymal transition. Nat Rev Mol Cell Biol 2020;
Res 2019; 79: 3011–3027.
21: 341–352.
16. Park S, Gonzalez DG, Guirao B, et al. Tissue-scale coordination of
2. Biffi G, Tuveson DA. Diversity and biology of cancer-associated
cellular behaviour promotes epidermal wound repair in live mice.
fibroblasts. Physiol Rev 2021; 101: 147–176.
Nat Cell Biol 2017; 19: 155–163.
3. Paijens ST, Vledder A, de Bruyn M, et al. Tumor-infiltrating lympho-
17. Cox TR. The matrix in cancer. Nat Rev Cancer 2021; 21: 217–238.
cytes in the immunotherapy era. Cell Mol Immunol 2021; 18: 842–859.
18. Simoni Y, Becht E, Fehlings M, et al. Bystander CD8+ T cells are
4. Dvorak HF. Tumors: wounds that do not heal – redux. Cancer
abundant and phenotypically distinct in human tumour infiltrates.
Immunol Res 2015; 3: 1–11.
Nature 2018; 557: 575–579.
5. Rous P. An experimental comparison of transplanted tumor and a
19. Sodir NM, Kortlever RM, Barthet VJA, et al. MYC instructs and
transplanted normal tissue capable of growth. J Exp Med 1910; 12:
maintains pancreatic adenocarcinoma phenotype. Cancer Discov
344–366.
2020; 10: 588–607.
6. Janossy G, Francis GE, Capellaro D, et al. Cell sorter analysis of
20. Dongre A, Weinberg RA. New insights into the mechanisms of
leukaemia-associated antigens on human myeloid precursors.
epithelial–mesenchymal transition and implications for cancer. Nat
Nature 1978; 276: 176–178.
Rev Mol Cell Biol 2019; 20: 69–84.
7. Edwards PA. Heterogeneous expression of cell-surface antigens in
21. Alonso-Curbelo D, Ho Y-J, Burdziak C, et al. A gene–environment-
normal epithelia and their tumours, revealed by monoclonal anti-
induced epigenetic program initiates tumorigenesis. Nature 2021;
bodies. Br J Cancer 1985; 51: 149–160.
590: 642–648.
8. Kucerova R, Walczysko P, Reid B, et al. The role of electrical signals
22. Mantovani A, Allavena P, Sica A, et al. Cancer-related inflammation.
in murine corneal wound re-epithelialization. J Cell Physiol 2011;
Nature 2008; 454: 436–444.
226: 1544–1553.
23. Velez-Delgado A, Donahue KL, Brown KL, et al. Extrinsic
9. Vermeer PD, Einwalter LA, Moninger TO, et al. Segregation of
KRAS signaling shapes the pancreatic microenvironment through
receptor and ligand regulates activation of epithelial growth factor
fibroblast reprogramming. Cell Mol Gastroenterol Hepatol 2022;
receptor. Nature 2003; 422: 322–326.
13: 1673–1699.
10. Carter RL, Burman JF, Barr L, et al. Immunohistochemical localiza-
24. Olan I, Narita M. Senescence: an identity crisis originating from
tion of basement membrane type IV collagen in invasive and meta-
deep within the nucleus. Annu Rev Cell Dev Biol 2022; 38:
static squamous carcinomas of the head and neck. J Pathol 1985;
219–239.
147: 159–164.
© 2023 The Author. The Journal of Pathology published by John Wiley & Sons Ltd J Pathol 2023; 260: 1–4
on behalf of The Pathological Society of Great Britain and Ireland. www.pathsoc.org www.thejournalofpathology.com