Meflin-Positive Cancer-Associated Fibroblasts Inhibit Pancreatic Carcinogenesis

Tumors are not a homogenous mass of cancer cells, but also comprise many types of noncancer cells, including cancer-associated fibroblasts (CAF), myeloid cells, and lymphocytes (1, 2). These cells, together with tumor vessels and the extracellular matrix (ECM), shape the tumor microenvironment (TME), which is vital not only for the development and progression of cancer but also tumor immunity and resistance to anticancer therapies (1, 2).

CAFs constitute a major component of the TME and produce various types of ECM proteins, such as collagen, and soluble signaling molecules (3–8). CAFs promote the proliferation and invasion of cancer cells either directly via soluble signaling molecules or indirectly through the regulation of angiogenesis and immunity (3–9). Stromal fibrosis and stiffening driven by the ECM produced by CAFs, which are particularly conspicuous in aggressive cancers such as pancreatic ductal adenocarcinoma (PDAC), also increase cancer cell malignancy and therapeutic resistance (10–12). Many studies have shown positive correlations between CAF infiltration, monitored by analyzing standard CAF markers such as α-smooth muscle actin (αSMA), fibroblast-specific protein 1 (FSP1), fibroblast activation protein (FAP), and podoplanin, and poor outcome of patients with cancer (3, 4, 6, 8, 13–18). Thus, the biology of CAFs, including their origins, specific markers, functions, and clinical relevance, has gained increasing research interest in recent years (3–8). Clinical trials of therapeutics that target CAFs are also emerging, with investigations on the underlying mechanisms (4, 8, 19, 20).

One caveat in targeting CAFs, however, was pointed out by recent observations that the functions of individual CAFs are not necessarily the same (3–6, 8, 21–24). Genetic depletion of proliferating αSMA–positive (αSMA⁺) CAFs and conditional blocking of the Sonic Hedgehog (Shh) signaling pathway, which is crucial for promoting desmoplasia in PDAC (25), led to the progression of PDAC in mouse models, suggesting that certain populations of CAFs, if not all, may function to suppress cancer progression (21–24). Clinical studies that evaluated the use of CAF inhibitors in patients with PDAC were not successful, raising the question whether CAFs promote or restrain cancer progression (12). The current view is that there exist heterogeneous populations of CAFs that comprise cancer-promoting CAFs (pCAF) and cancer-restraining CAFs (rCAF; refs. 3–6, 8). However, marker protein(s) that specifically label rCAFs have not been identified to date. As noted above, αSMA is a candidate marker of rCAFs (3, 12, 22). However, αSMA is widely expressed by pericytes, smooth muscle cells that form the tunica media of vessels, myoepithelial cells, and smooth muscle fibers of various organs, and thus, is not specific to CAFs (3, 5, 6). In addition, some studies have shown that αSMA expression in CAFs correlates with unfavorable prognosis in some types of cancer (15–18). Thus, the role of αSMA⁺ CAFs in cancer progression remains controversial. Furthermore, it is unclear whether pCAFs and rCAFs represent distinct CAF populations with different origins or whether they undergo interconversion during tumor progression.

We previously identified Meflin, which is a glycosylphosphatidyl inositol (GPI)-anchored protein encoded by the immunoglobulin superfamily containing leucine-rich repeat (ISLR/Islr) gene, as a specific marker of mesenchymal stromal/stem cells (MSC) that are found in the perivascular space of multiple organs, including the bone marrow (BM; ref. 26). Meflin expression is limited to undifferentiated MSCs and is not detected in their differentiated lineages, such as mature osteoblasts, chondrocytes, and adipocytes (26). Forced overexpression of Meflin suppressed the expression of osteogenic and chondrogenic markers in an MSC cell line (26). Consistent with this, Meflin knockout (KO) resulted in accelerated osteoblastic differentiation of MSCs in the long bones. These data suggested that the primary function of Meflin is to maintain the undifferentiated state of MSCs (26). However, the histology of the other organs of Meflin-KO mice seemed to be indistinguishable from that of wild-type (WT) mice, leaving the role of Meflin in MSCs that exist throughout the body undetermined (26). A recent study revealed that Meflin is expressed by skeletal muscle stem cells and is crucial for skeletal muscle regeneration and the maintenance of the canonical Wnt signaling pathway (27). However, the mechanism underlying Meflin-mediated regulation of tissue stem cells remains to be precisely defined. It has been reported that Meflin interacts with disheveled segment polarity protein 2 (Dvl2), a cytoplasmic mediator of Wnt signaling (27). However, given that Meflin is a GPI-anchored cell-surface or -secreted protein, it is plausible that Meflin interacts with proteins in the lumen of the Golgi apparatus or with extracellular or membrane proteins.

In this study, we investigated the roles of Meflin and Meflin⁺ cells in the development of PDAC. We first examined the expression pattern of Meflin in the normal pancreas and tumor tissues of human PDAC and its correlation with the clinical outcome. The significance of Meflin in PDAC progression was confirmed by experiments using Meflin-KO mice and a PDAC mouse model. Next, we analyzed the effect of genetic ablation of Meflin⁺ cells on tumor progression and conducted a lineage-tracing experiment in mice. Then, we explored the molecular function of Meflin, and finally, investigated whether exogenous delivery of Meflin in the tumor stroma would suppress tumor development in mice.

Human PDAC samples were obtained at the time of surgery from patients who had provided written informed consent. This study was conducted in accordance with the Helsinki Declaration for Human Research and approved by the Ethics Committee of Nagoya University Graduate School of Medicine (approval number: 2017-0127).

Mouse PDAC cells (mT3, mT4, and mT5) were generously provided by David Tuveson (Cold Spring Harbor Laboratory) and Chang-Il Hwang (UC Davis College of Biological Sciences, Davis, CA). AsPC-1 and UE7T13 cells were purchased from the Japanese Cancer Research Resources Bank. U2OS, Saos-2, MBA-MB-231, and 578T cells were purchased from the ATCC. PANC-1, KLM-1, and KP-4 cells were obtained from the Cell Resource Center for Biomedical Research (Institute of Development, Aging, and Cancer, Tohoku University, Sendai, Japan). Cell lines were authenticated by routine morphologic and growth analyses. All cell lines were frozen at two passages and used in experiments within five passages after thawing. All cell lines used were routinely screened for Mycoplasma contamination by 4′, 6-diamidino-2-phenylindole staining. Human pancreatic stellate cells (PSC) were isolated from pancreatic tissues adjacent to primary PDAC tumors during surgery at Tohoku University Hospital (28, 29). This study was approved by the Ethics Committees of Tohoku University (approval number 2017-1-280) and Nagoya University Graduate School of Medicine (approval number 2015-0254). Immortalized human PSCs were established by retrovirus-mediated introduction of the simian virus 40 (SV40) large T antigen (29).

Animal protocols were approved by the Animal Care and Use Committee of Nagoya University Graduate School of Medicine (approval number: 30366). Detailed protocols for the generation of gene-targeted mice, in situ hybridization (ISH), IHC, cell biological, biochemical, and animal experiments and methods for statistical analysis are described in Supplementary Data.

We previously reported that Meflin is a marker of MSCs and is not expressed in pericytes, epithelial, endothelial, and smooth muscle cells, and blood and immune cells, and that it displays higher specificity for MSCs than other conventional MSC markers, such as CD90, CD105, and CD73 (26). Meflin⁺ MSCs are found in the BM (Supplementary Fig. S1) and the perivascular regions of multiple organs, as shown in our previous study (26). This is consistent with the notion that MSCs localize in multiple tissues as perivascular fibroblasts (30). ISH analysis of mouse and human pancreases showed that few Meflin⁺ cells are scattered in the perivascular, periductular, and periacinar regions (Fig. 1A and B). Meflin⁺ cells were abundantly present in the early postnatal pancreas, but their number decreased in the adult stage (Supplementary Fig. S1). To confirm Meflin expression in the pancreas, we generated a Meflin reporter–knockin mouse model [Meflin-ZsGreen-diphtheria toxin receptor (DTR)-Cre recombinase (Meflin-ZDC)], in which the expression of a green fluorescent protein ZsGreen, DTR, and Cre was driven by the Meflin gene (Islr) promoter (Fig. 1C; Supplementary Fig. S2A–S2C). In the Meflin-ZDC mice, ZsGreen⁺ cells expressed desmin (Fig. 1C), which is a known PSC marker (31). Immunofluorescence (IF) staining showed colocalization of Meflin and desmin at the protein level (Supplementary Fig. S1). These data suggested that Meflin is a new marker of PSCs. Meflin was expressed by PSCs isolated from the pancreas, which were positive for desmin but not for epithelial or endothelial markers, as indicated by Western blot analysis (Fig. 1D).

Consistent with our previous study on cultured MSCs (26), robust Meflin expression was observed in contact-inhibited PSCs cultured to a superconfluent monolayer, suggesting a shared mechanism for Meflin expression in MSCs and PSCs (Fig. 1E). PSC-specific expression of Meflin was also confirmed by the analysis of a recently published dataset of single-cell RNA sequencing (scRNA-seq) of healthy mouse pancreas, where Meflin was specifically detected in PSC populations but not in other cell populations (Supplementary Fig. S3A–S3C; ref. 32).

One of the hallmarks of PSCs is that they modulate their gene expression in response to vitamin D (33). A previous study that screened genes upregulated in PSCs treated with calcipotriol, a vitamin D analog, identified Islr as one of those genes (Supplementary Fig. S4A and S4B; ref. 33). Consistent herewith, most Meflin⁺ spindle fibroblastic cells expressed the vitamin D receptor (VDR) within the pancreatic stroma of human PDAC (Fig. 1F). The coexpression of Meflin and VDR was also observed in the stroma of the KPC pancreatic cancer mouse model (Supplementary Fig. S5A and S5B; ref. 34). Calcipotriol-induced Meflin expression was also observed in cultured human and mouse PSCs, where 25-hydroxyvitamin D3-24-hydroxylase (Cyp24A1) served as a positive control of calcipotriol activity (Fig. 1G and H). Intriguingly, the expression of Acta2, which encodes αSMA, was downregulated by calcipotriol treatment in WT, but not Meflin-deficient PSCs (Fig. 1H). Calcipotriol-dependent Meflin expression was confirmed by a luciferase reporter assay with a vector containing the putative promoter region of the human Meflin gene (ISLR; Fig. 1I and J). Together, these data suggested that Meflin is a vitamin D–responsive PSC marker.

Consistent with previous findings that PSCs are activated to become CAFs in PDAC (11, 33, 35–37), ISH analysis of human PDAC tissues revealed the presence of Meflin⁺ cells in the proximity of invading cancer cells, even in small invasive lesions (Fig. 2A). In advanced PDAC cases, Meflin⁺ cells were extensively distributed in the stroma of primary tumor lesions as well as in invasive lesions in peripancreatic adipose tissues (Fig. 2B). We also detected Meflin protein expression in stromal cells in some human PDAC samples by IHC using an anti-human Meflin mAb that we developed (Supplementary Fig. S6). However, IHC for Meflin often resulted in nonspecific staining of nonstromal cells that were negative for Meflin by ISH in other PDAC cases. Therefore, we adopted ISH to localize Meflin⁺ cells throughout the study. Meflin expression was neither detected in cancer cells in human PDAC tissues nor PDAC cell lines (Figs. 1D, 2A, and B; Supplementary Figs. S6 and S7). A major fraction of the cells that were positive for platelet-derived growth factor receptor α (PDGFRα), an established fibroblast marker, was also positive for Meflin, suggesting that Meflin is a new marker of CAFs (Supplementary Fig. S8).

Interestingly, CAFs in human PDAC exhibited variable expression of Meflin and other conventional CAF markers. Notably, approximately 10% of CAFs expressing αSMA protein were positive for Meflin mRNA in human PDAC as determined by alkaline phosphatase–based IHC following ISH (Supplementary Fig. S8). Similarly, dual-color chromogenic ISH suggested that a small fraction of ACTA2-expressing cells expressed Meflin mRNA (Supplementary Fig. S9A). In line herewith, highly sensitive detection of αSMA and Meflin mRNAs by fluorescence ISH and quantification by counting the dots per cell revealed that CAFs strongly positive for Meflin were weakly positive for αSMA, and there was an inverse correlation between Meflin and αSMA expression (Supplementary Fig. S9B). These data suggested that Meflin and αSMA expression are not mutually exclusive; rather, Meflin⁺ CAFs express a low level of αSMA.

Meflin was expressed by approximately 40% of FAP⁺ CAFs (Supplementary Fig. S8) and by the majority of CAFs expressing Gli1, a transcription factor involved in the Shh pathway crucial for PDAC progression (Supplementary Fig. S8; refs. 12, 38). These data suggested that Meflin defines a population of CAFs that is αSMA^low FAP^+/− PDGFRα⁺ Gli1⁺ in human PDAC.

Meflin expression was also detected in tumors from KPC PDAC mice (Fig. 2C). Meflin⁺ CAFs appeared in the early stage of acinar-to-ductal metaplasia and were present in greater numbers in the preinvasive lesion of the developed tumor, where rare αSMA⁺ CAFs were present in close proximity to p53 strongly positive cancer cells, and in the invasive front of the invasive lesion (Fig. 2C). Meflin expression was not detected in cancer cells throughout the tumorigenic process (Fig. 2C and D; Supplementary Fig. S7). As observed in human PDAC tissues, Meflin and αSMA expression are inversely correlated when analyzed at the mRNA level by fluorescence ISH, providing additional support for the finding that CAFs strongly positive for Meflin are weakly positive for αSMA (Fig. 2E and F). The inverse correlation between Meflin and αSMA expression in CAFs was also confirmed by the analysis of publicly available datasets of scRNA-seq analysis of CAFs isolated from the KPC model and a breast cancer mouse model (Fig. 2G; Supplementary Figs. S10 and S11; refs. 36, 39). The data showed that the CAF population that is highly positive for Meflin is distinct from populations enriched for αSMA and the PSC marker desmin.

Next, we assessed the effect of alterations in Meflin expression on global gene expression profiles in MSCs, a potential source of CAFs (8), using a gene expression microarray (Supplementary Fig. S12A–S12C). The data showed that Meflin depletion resulted in the upregulation of cytokines such as IL6 and C-C motif chemokine ligand 2 (CCL2) and several markers of activated CAFs, such as αSMA and Gli1. Furthermore, gene set enrichment analysis demonstrated the upregulation of previously reported activated PSC signature genes in Meflin-depleted MSCs (Supplementary Fig. S12D; ref. 33). These data suggested that alterations in Meflin expression in CAFs may be involved in gene expression changes associated with CAF activation.

When we investigated Meflin expression in surgically resected human PDAC tissues (N = 71), we found that the frequency of Meflin⁺ CAFs varies between patients with PDAC (Fig. 3A; Supplementary Table S1). We evaluated the number and ratio of Meflin⁺ cells in total stromal cells (visualized by hematoxylin counterstaining), excluding cells with the morphology of lymphocytes, erythrocytes, and endothelial cells. We divided the 71 cases into Meflin-high (≥20% Meflin⁺ stromal cells) and Meflin-low (<20% Meflin⁺ stromal cells) groups (Fig. 3A; Supplementary Table S1). The Meflin-high group exhibited better prognosis in Kaplan–Meier survival analysis and a more differentiated histology than the Meflin-low group, suggesting a unique feature of Meflin⁺ CAFs that is distinct from that of previously described pCAFs, which correlates with poor outcome (Fig. 3B; Supplementary Tables S1 and S2).

As we previously reported, Meflin-KO mice exhibit slight growth retardation and accelerated growth of long bones in the postnatal stages, but become indistinguishable from WT mice as they grow into young adults (26). These mice are viable and fertile, with grossly normal pancreatic tissue (Supplementary Fig. S13A and S13B). However, we found that tumors that developed in compound transgenic Meflin-KO KPC mice were significantly larger and more proliferative than those developed in WT KPC mice (Fig. 3C–F; Supplementary Fig. S13C and S13D), which was consistent with the poorer survival rate of Meflin-KO mice than that of WT mice (Fig. 3G). Reflecting the higher tumor burden with more necrosis in Meflin-KO KPC mice, the number of apoptotic cells was marginally but significantly increased in Meflin-KO KPC tumors compared with that in WT KPC tumors (Fig. 3F). These data further supported the notion that Meflin functionally defines a CAF population that is distinct from conventional pCAFs.

Further histologic analyses revealed that tumors in Meflin-KO KPC model mice were poorly differentiated compared with tumors in WT KPC mice (Fig. 4A and B). Meflin-KO tumors contained more αSMA⁺ CAFs and more collapsed tumor vessels than WT tumors (Fig. 4C and D). These data suggested that the loss of Meflin led to increased αSMA expression in CAFs and decreased tumor vessel perfusion. The expression of FAP, a marker of pCAFs (8, 11), was not significantly different between WT and Meflin-KO KPC tumors (Supplementary Fig. S13E and S13F). Next, organoids established from KPC tumors developed in Meflin-KO mice were orthotopically transplanted into the pancreases of WT and Meflin-KO mice (Fig. 4E). Histologic analysis showed that the tumors engrafted into WT pancreas exhibited features of well-differentiated tumors, with an acinar-like appearance, whereas those engrafted and developed in KO pancreas exhibited features of poorly differentiated tumors (Fig. 4E and F). These data suggested that Meflin shapes the TME, which blocks poor differentiation in PDAC.

To investigate the role of Meflin⁺ CAFs in tumor differentiation further, we subcutaneously transplanted syngeneic mT5 PDAC cells (40) into Meflin-ZDC; Rosa26-LSL (LoxP-stop-LoxP)-tdTomato mice, followed by intratumoral administration of diphtheria toxin to ablate cells that actively express Meflin (Fig. 5A). ISH analysis showed that endogenous Meflin was successfully depleted in tumors developed in diphtheria toxin–injected mice (Fig. 5B and C). The tumors depleted of Meflin⁺ cells exhibited no apparent changes in volume; however, they possessed a higher proliferative capacity and more poorly differentiated histology than tumors developed in the control groups (Fig. 5D–G). These data implied that Meflin⁺ CAFs include some CAF populations that block poor tumor differentiation. Furthermore, we found that the ablation of Meflin⁺ CAFs led to a decrease in the αSMA⁺ area (Fig. 5H and I), suggesting that Meflin⁺ CAFs might be involved in the infiltration or recruitment of αSMA⁺ CAFs into the tumor stroma. Alternatively, αSMA⁺ CAFs might be derived from Meflin⁺ CAFs during tumor progression.

To investigate whether Meflin⁺ CAFs yield αSMA⁺ CAFs in vivo, we generated a knockin mouse line in which the expression of a fusion protein of the Cre recombinase and the mutated ligand-binding domain of the human estrogen receptor (CreERT2) was driven by the Meflin (Islr) promoter (Meflin-CreERT2; Supplementary Fig. S14A and S14B). We generated Meflin-CreERT2; Rosa26-LSL-tdTomato mice to trace the lineage of Meflin⁺ cells by tamoxifen administration. Tamoxifen administration induced the expression of tdTomato in BM MSCs that are located around the trabecular bones, arterioles, periosteum, and sinusoids, as well as in PSCs with long cytoplasmic processes in the pancreas, supporting the notion that Meflin is a marker of these cells (Supplementary Fig. S14C and S14D). We subcutaneously transplanted syngeneic mT3 PDAC cells (40) following a 2-week washout period after tamoxifen administration (Fig. 6A). Meflin⁺ cells gave rise to 34% ± 15% of all αSMA⁺ CAFs on day 10 after transplantation (Fig. 6B). ISH analysis revealed that the number of tdTomato⁺ Meflin-lineage cells that expressed αSMA, but not those that expressed FAP, significantly increased between day 3 (46% ± 26%) and day 10 (93% ± 4%; Fig. 6C–F). Moreover, 42% ± 4% and 58% ± 4% (P = 0.0427) of tdTomato⁺ Meflin-lineage cells had lost endogenous Meflin expression on day 3 and day 10, respectively (Fig. 6G and H). These data suggested that Meflin⁺ cells yield CAFs that are positive for αSMA⁺, and negative or positive for Meflin.

Our previous study showed that forced overexpression of Meflin suppressed the expression of the osteoblast and chondrocyte differentiation markers in MSCs (26). Therefore, we hypothesized that exogenous expression of Meflin would suppress the myofibroblastic phenotype of CAFs and influence the TME and tumor progression. Exogenous Meflin expression suppressed αSMA expression in cultured MSCs (Supplementary Fig. S15), which was consistent with the observation that Meflin expression suppressed the capacity of MSCs to contract a collagen matrix (Fig. 7A–C). The ability of MSCs to form colony-forming unit fibroblasts (CFU-F) was also enhanced by Meflin expression, consistent with the role of Meflin in maintaining the undifferentiated state of MSCs (Fig. 7D and E; ref. 26). Subcutaneous xenografts comprising of AsPC-1 PDAC cells and immortalized human PSCs (29) transduced with Meflin exhibited tumor growth regression and decreased infiltration of αSMA⁺ CAFs compared with AsPC-1 cell and control human PSC xenografts in immunodeficient mice (Supplementary Fig. S16A–S16F). Furthermore, we engineered mT3 PDAC cells to produce lentiviruses that could infect surrounding stromal cells and implanted these cells into C57BL/6 mice (Fig. 7F–I). Transplantation of mT3 cells expressing a lentivirus encoding Meflin increased Meflin expression in the tumor stroma, which was accompanied by the suppression of tumor growth and αSMA⁺ CAF infiltration when compared with control mT3 cells expressing a lentivirus encoding the green fluorescent protein Azami Green. These data suggested that the transduction of Meflin into tumor tissues or the augmentation of Meflin expression in CAFs may provide a strategy to suppress tumor progression.

Finally, given our recent finding that Meflin is expressed in cardiac fibroblasts and suppresses cardiac fibrosis and stiffening of cardiac tissue in a chronic heart failure mouse model (41), we explored the involvement of Meflin in stromal collagen structural remodeling by second-harmonic generation (SHG) microscopic observations of WT and Meflin-KO KPC tumors (Fig. 7J–L; Supplementary Fig. S17A–S17C). The data revealed that the stroma of tumors developed in Meflin-KO KPC mice exhibited straighter and wider collagen structures than that of tumors in WT KPC mice. Differences in collagen signature such as altered curvature, total length, density, and alignment of fibers were observed between Meflin-high (≥20% stromal cells were positive for Meflin) and -low areas of the stroma of human PDAC tissues based on SHG measurements (Supplementary Fig. S17D–S17J). Given the significance of changes in collagen configuration in cancer progression (42–44), these data suggested that Meflin suppresses tumor progression in part by inhibiting the remodeling collagen structure (Fig. 7M).

It was first suggested over 50 years ago that the primary function of normal fibroblasts is to inhibit the growth of polyoma virus–transformed cells (45). Consistent with this landmark study, the presence of rCAFs in the TME has been postulated on the basis of the findings that genetic ablation of proliferating αSMA⁺ cells and genetic and pharmacologic inhibition of Shh signaling promoted cancer progression in mouse models of PDAC (22–24). However, the role of αSMA⁺ CAFs in cancer progression is controversial, and the precise nature of rCAFs has not been defined (3, 8, 11, 13, 15–18, 22). In this study, we showed that in the healthy pancreas, Meflin marks PSCs, whose relationship with MSCs had not been fully explored previously. Meflin also marks a population of CAFs in PDAC, with high Meflin expression correlating with favorable outcome in patients with PDAC and a PDAC mouse model. Meflin functions to suppress PDAC progression, presumably by suppressing αSMA expression in CAFs and changing collagen configuration in the tumor stroma. A lineage-tracing experiment showed that Meflin-lineage cells contained αSMA⁺ CAFs, some of which exhibited downregulated Meflin expression. Taking all data into consideration, we speculate that Meflin is a marker of rCAFs (αSMA^low PDGFRα⁺ Gli1⁺), which yield αSMA⁺ Meflin^low/− CAFs during PDAC progression. Given that CAFs expressing high levels of Meflin express low levels of αSMA, it is speculated that genetic depletion of Meflin⁺ CAFs also resulted in depletion of a subset of proliferating αSMA⁺ CAFs that were previously reported to have a cancer-restraining action (22).

In this study, the mechanisms underlying the downregulation of Meflin expression in CAFs remains to be unraveled. We previously reported that repeated passage and long-term culture of primary MSCs on cell culture plastic induced a significant decrease or loss in Meflin expression, in which the stiffness of their substrate is crucial (26). Immortalization of MSCs and human PSCs also resulted in the loss of Meflin expression. Other factors that significantly downregulate Meflin expression include aging, hypoxia, and transforming growth factor-β signaling, which has been shown to induce differentiation of PSCs into αSMA⁺ myofibroblastic CAFs (myCAF; refs. 41, 36). We speculate that some of these factors are involved in the downregulation of Meflin expression in Meflin⁺ PSCs and CAFs, which results in the generation of Meflin-negative or weakly positive CAFs and thus contributes to the generation of CAF heterogeneity during tumor progression. Recently, two distinct populations of CAFs derived from quiescent PSCs were identified in the tumor stroma of mouse PDAC, that is, αSMA⁺ myCAFs and IL6⁺ inflammatory CAFs (iCAF; refs. 12, 35, 36). Transcriptome analysis of these cells revealed that Meflin is expressed, in descending order, in quiescent PSCs (basemean value 2623.5), iCAFs (basemean value 665.7), and myCAFs [basemean value 125.0; Gene Expression Omnibus (GEO) accession no. GSE93313; ref. 35]. Further studies are needed to clarify the exact interrelationships between iCAFs, myCAFs, and Meflin⁺ CAFs. We have not yet determined whether Meflin⁺ CAFs originate from quiescent Meflin⁺ PSCs in the pancreas in vivo, it is possible that they are also derived from Meflin⁺ MSCs in the BM and other organs. Another limitation of this study is that we evaluated Meflin expression by ISH, not IHC; thus, how Meflin proteins are distributed in the tumor stroma and whether they act on non-CAFs, including tumor cells, remains to be resolved.

One interesting finding in this study was that PDAC that developed in Meflin-KO mice was poorly differentiated. This was confirmed by orthotopic transplantation of mouse PDAC organoids into the pancreases of WT and Meflin-KO mice and genetic ablation of Meflin⁺ CAFs in a xenograft tumor model. Given that Meflin is not expressed by pancreatic acini, ducts, the islets of Langerhans, vessels, blood cells, or tumor cells, the data support the notion that CAF phenotypes determine the behavior and differentiation of tumor cells. This is also supported by previous studies showing that depletion of proliferating αSMA⁺ CAFs or inhibition of the Shh pathway yielded predominantly undifferentiated tumor cells in mouse models of PDAC (22, 23). These findings are compatible with the concept that the TME influences the malignant phenotype, and a high degree of plasticity of differentiation status in cancer cells exists within tumors (7, 46–48). The detailed mechanisms of Meflin-mediated alterations of the tumor stroma have not been clarified in this study, but one potential explanation based on SHG observations of WT and Meflin-KO KPC tumors is that Meflin may be involved in the suppression of the activity of the lysyl oxidase (Lox) family of proteins and excessive ECM cross-linking. Considering the involvement of Lox activity in increased stiffness of the tumor stroma that is crucial for the activation of cancer cells and cancer progression (42–43), it will be interesting to investigate the role of Meflin in the regulation of the mechanical properties of the tumor stroma in future studies.

We also showed that exogenous expression of Meflin in the tumor stroma suppressed αSMA expression in CAFs and tumor progression. It is, therefore, tempting to investigate whether it would be possible to reverse the phenotype of pCAFs by restoring Meflin expression and to suppress cancer progression. In this respect, it was interesting to find that most Meflin⁺ cells found in the pancreas adjacent to PDAC were positive for VDR and that Meflin expression was upregulated by vitamin D. These data are consistent with the findings in previous studies that vitamin D or A treatments restored a quiescent state in PSCs, suppressed inflammation and fibrosis of the tumor stroma of PDAC, and increased tumor sensitivity to gemcitabine (12, 33, 49, 50). These studies have provided a rationale for the use of vitamin D analogs in combination with either chemotherapy or immune checkpoint inhibitors for the treatment of patients with PDAC in clinical trials (e.g., clinicaltrials.gov, NCT03331562 and NCT03519308). A critical issue that warrants a further investigation is how varied Meflin expression in CAFs is involved in the response of human PDAC to those combined treatments.

In conclusion, we found that Meflin is a marker of PSCs in the normal pancreas. Consistent with previous studies showing that PSCs can be activated to become CAFs in PDAC, Meflin expression was detected in CAFs in the tumor stroma of both human PDAC and a PDAC mouse model. Interestingly, Meflin expression in CAFs correlated with favorable outcomes in human PDAC and the mouse model. We also showed that Meflin has the capacity to suppress αSMA expression (myofibroblastic differentiation) in CAFs and ECM remodeling, which is crucial for cancer progression. Together, our data suggest that Meflin is a marker of a subpopulation of rCAFs in PDAC.

No potential conflicts of interest were disclosed.

Conception and design: Y. Mizutani, H. Kobayashi, Y. Hirooka, A. Enomoto, M. Takahashi

Development of methodology: Y. Mizutani, H. Kobayashi, T. Iida, L. Weng, M. Matsuyama, T. Kobayashi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Mizutani, H. Kobayashi, T. Iida, N. Asai, A. Hara, N. Esaki, S. Mii, K. Ando, S. Ishihara, S.M. Ponik, M.W. Conklin, T. Miyata, M. Matsuyama, T. Kobayashi, T. Fujii, S. Yamada, A. Enomoto

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Mizutani, H. Kobayashi, T. Iida, Y. Shiraki, S. Ishihara, S.M. Ponik, M.W. Conklin, T. Fujii, T. Wang, D. Worthley, T. Shimamura, A. Enomoto, M. Takahashi

Writing, review, and/or revision of the manuscript: Y. Mizutani, H. Kobayashi, S. Ishihara, S.M. Ponik, H. Haga, A. Nagasaka, T. Fujii, S.L. Woods, D. Worthley, Y. Hirooka, A. Enomoto, M. Takahashi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): N. Asai, A. Masamune, K. Ushida, H. Haga, A. Nagasaka, M. Matsuyama, T. Kobayashi, J. Yamaguchi

Study supervision: D. Worthley, M. Fujishiro, Y. Hirooka, A. Enomoto, M. Takahashi

We thank Patricia J. Keely (University of Wisconsin-Madison, Madison, WI), who sadly passed away in 2017, for her support in SHG microscopy; David Tuveson (Cold Spring Harbor Laboratory) and Chang-Il Hwang (UC Davis College of Biological Sciences, Davis, CA) for providing the mouse PDAC cell lines mT3, mT4, and mT5; Genichiro Ishii and Yasuhiro Matsumura (National Cancer Center), Masahiro Sokabe and Hiroyoshi Nishikawa (Nagoya University, Nagoya, Japan), Akira Orimo (Juntendo University, Tokyo, Japan), Satoru Kidoaki (Kyushu University, Fukuoka, Japan), Takashi Okada and Yoshiyuki Yamazaki (Nippon Medical School, Tokyo, Japan), Takuya Kato (Kitasato University, Tokyo, Japan), and Yumi Matsuzaki (Shimane University, Matsue, Japan) for helpful discussions; Kentaro Taki (Nagoya University, Nagoya, Japan) for help with mass spectrometry; Toshiro Sato (Keio University, Tokyo, Japan) for technical assistance in organoid culture; Takeshi Urano (Shimane University, Matsue, Japan) for technical assistance in protein purification; and Kozo Uchiyama (Nagoya University, Nagoya, Japan) for technical assistance. This work was supported by a Grant-in-Aid for Scientific Research (S) (26221304 to M. Takahashi) and a Grant-in-Aid for Scientific Research (B) (15H04719 and 18H02638 to A. Enomoto, and 15H04804 and 19H03631 to A. Masamune) commissioned by the Ministry of Education, Culture, Sports, Science and Technology of Japan; Nagoya University Hospital Funding for Clinical Research (to A. Enomoto); The Hori Sciences And Arts Foundation (to A. Enomoto); Kobayashi Foundation for Cancer Research (to A. Enomoto); AMED-CREST (Japan Agency for Medical Research and Development, Core Research for Evolutional Science and Technology; to A. Enomoto and H. Haga); the Project for Cancer Research and Therapeutic Evolution (P-CREATE) from AMED (to A. Enomoto); the Mochida Memorial Foundation for Medical and Pharmaceutical Research (to A. Enomoto); and the Smoking Research Foundation (to A. Masamune). H. Kobayashi is supported by the Japan Society for the Promotion of Science (JSPS) Overseas Challenge Program for Young Researchers and Takeda Science Foundation Fellowship.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.