Cancer-associated fibroblast (CAF) pro– and anti–pancreatic cancer functional dichotomy has been at the center of numerous studies. In this issue of Cancer Discovery, Helms and colleagues demonstrate that although pancreatic stellate cell–derived CAFs constitute a desmoplastic cell minority, these cells play a protumorigenic role via microenvironmental mechanomodulation.
See related article by Helms et al., p. 484.
Pancreatic ductal adenocarcinoma (PDAC) is a deadly disease characterized by a unique cancer-associated fibroblast (CAF)–based stroma expansion, known as desmoplasia (ref. 1; Fig. 1). Desmoplastic CAFs are considered “double-edged swords.” This is because although desmoplasia sequesters PDAC cells within a poorly irrigated tumor microenvironment, some CAFs protect cancer cells via direct nurture, induction of immunosuppression to evade immune surveillance, and amelioration of therapy-induced toxicity (1–3). Notably, clinical attempts to eliminate desmoplasia in PDAC have failed to provide benefit and have occasionally been harmful (3). As an alternative approach, investigators redirected research into “reprogramming” desmoplastic CAFs to harness their tumor-suppressive functions while obstructing their tumor-stimulating roles (3, 4). Thus, it became clear that defining the distinct CAF subtypes within this heterogenous population (5, 6), namely through lineage tracing of CAF ancestry (7–10), could be an effective way of explaining the seemingly paradoxical pro- and anti-PDAC CAF functions and garner a better understanding of these cells' interchangeable functional plasticity (2).
Pancreatic stellate cells (PSC) have been developed as a model to study CAFs. PSCs represent a population of pancreas-resident fibroblasts, which are easily harvested from normal pancreata and can be “activated” ex vivo through the modulation of culturing conditions to simulate various CAF types (5). In vitro use of the PSC-to-CAF transition revealed the same dichotomy of pro- and antitumor functions found in desmoplasia. For example, PSCs in vitro can be induced to present at least two identifiable sets of “functional” traits, designating these myofibroblastic or inflammatory, with transcription signatures akin to the ones observed in primary PDAC CAFs (5). These transcriptionally diverse CAFs, first considered distinct molecular types (5), are now thought as interchangeable (2), and their functional diversity is proposed as progeny of a common ancestral resident fibroblast (6). Myofibroblastic CAFs, classically represented by TGFβ-driven α-smooth muscle actin (α-SMA) expression and high levels of interstitial extracellular matrix (ECM) production (e.g., collagen I), are perhaps the epicenter of most uncertainties. There is strong evidence to support that α-SMA+ CAFs are tumor suppressive. This is because their elimination and/or the exclusion of key molecules such as collagen I from α-SMA+ CAFs accelerates PDAC tumorigenesis (7, 10). Then again, others demonstrated that α-SMA+ CAFs are the product of more than one PDAC CAF cell sublineage (8). Hence, the pending question has been whether all CAFs are the result of activation, followed by the expansion, of PSCs during PDAC and whether PSC-generated CAFs play a pro- or antitumor role.
This complicated question was addressed by Helms and colleagues (9). Using transcriptional data obtained from primary quiescent PSCs (4), Helms and colleagues categorized the fatty acid binding protein 4 (Fabp4) as a PSC biomarker and used a Fabp4 promoter-dependent fluorescent reporter to elegantly trace PSC fate during pancreatic tumorigenesis (9). The team, led by Dr. Sherman, engineered a clever model in which Cre+ PSCs permanently express GFP, whereas all other cells were tagged by the denoted red protein (9). Using this Fabp4-Cre;Rosa161mTmG model as a host, the researchers performed an orthotopic allograft of PDAC cells into the pancreas. During tumorigenesis, the GFP+ PSC cell population expanded and expressed known CAF markers (e.g., podoplanin and α-SMA), accounting for about 10% of all CAFs. The team then compared the transcriptional signature of GFP+ PSC-derived CAFs versus other CAFs and found marked differences. This suggested that specific CAF lineages (and not just exposure of a uniform CAF progenitor to distinct growth factor gradients) might dictate unique transcriptional signatures and CAF functions (Fig. 1). Notably, the signature of PSC-derived CAFs included expression of the receptor tyrosine kinase Tie1. Hence, α-SMA+/Tie1+/GFP+ cells were used to follow PSC-generated CAFs. The unique transcription signature was highly informative and confirmed the expected common Lrrc15 fibroblastic ancestry (6). Interestingly, no significant differences were noted between PSC-derived and other CAFs in expression of known receptors for ligands, including hedgehog, TGFβ, and IL1, which are associated with canonical myofibroblastic or inflammatory CAF functions. This suggested that PSC progeny could generate both pro- and anti-inflammatory CAFs.
Next, to test the functional contribution of PSC-derived CAFs to desmoplasia, the team used a retrograde ductal injection of viral particles to selectively eliminate, using a toxin-based strategy, Fabp4-Cre+ cells from established PDACs. Although the resulting tumors incorporated similar amounts of collagen into the desmoplastic ECM with significant α-SMA+ remaining CAFs, a modulation in the mechanoregulatory machinery, with decreased expression of tenascin-C and activated MLC2 and FAK, was noted. Because reduction in stromal perlecan expression was also detected, and because this ECM protein has been linked to a commonly seen PDAC cell gain-of-function mutation in p53, the team hypothesized that the PDAC genotype could modulate PSC-generated CAF prevalence. Indeed, results suggest that p53 mutant, but not p53 loss, prompts for a PSC-to-CAF expansion, demonstrating that the PDAC genotype can dictate the type of desmoplasia. Notably, the team also found that the ECM signature associated with abundant PSCs correlated with reduced overall survival of human patients with PDAC compared with that of patients with a low PSC-ECM signature.
This seminal study thus revealed that although PSC-generated CAFs can indeed express α-SMA, the total number of PSC-generated CAFs is limited. This fact also, for the first time, suggests that most α-SMA+ CAFs in PDAC desmoplasia are not PSC derived. The study hence explains that although the specific PSC-generated CAF population is tumor supporting, most myofibroblastic α-SMA+ CAFs are not the result of PSC activation and can remain assigned with tumor-suppressive roles. This led to two main conclusions: first, PSCs are not the main source of desmoplastic CAFs in PDAC, even though PSC-to-CAF activation (at least in vivo) presents features of protumoral myofibroblastic cells. Second, the article suggests that differences in progenitors contribute to CAF heterogeneity but leaves open the question of the lineages that are not PSC derived, such as the Gli+ and Hoxb6+ CAFs reported in other studies (8). The study by Helms and colleagues (9) also supports the possibility of functional plasticity and spatial localization/distribution as added explanations for functional CAF heterogeneity.
Harnessing tumor-suppressive while targeting tumor-promoting CAF functions constitutes the most sought-after goal in the field of PDAC microenvironmental investigation. This study brings us a step closer to achieving it!
E. Cukierman reports personal fees from Phenomic and Takeda outside the submitted work.
The author is currently supported by the following sources: the Pancreatic Cancer Cure Foundation, the Concetta Greenberg Pancreatic Cancer Institute at the Fox Chase Cancer Center, the 5th AHEPA Cancer Research Foundation, the Worldwide Cancer Research, and NIH/NCI grants R21CA231252, R21CA252535, R01CA232256, and CA06927.