• Open Access

Mesenchymal stem cells regulate epithelial–mesenchymal transition and tumor progression of pancreatic cancer cells


To whom all correspondence should be addressed.

E-mail: higuchi@a2.keio.jp


Cancer-associated fibroblasts contribute to cancer progression that is caused by epithelial–mesenchymal transition (EMT). Recently, mesenchymal stem cells (MSCs) were found to be the major candidate involved in the development of tumor-promoting cancer stroma. Here we report that α-smooth muscle actin-positive myofibroblast-like cells originating from MSCs contribute to inducing EMT in side population cells of pancreatic cancer. More importantly, MSC-derived myofibroblasts function to maintain tumor-initiating stem cell-like characteristics, including augmenting expression levels of various stemness-associated genes, enhancing sphere- forming activity, promoting tumor formation in a mouse xenograft model, and showing resistance to anticancer drugs. Furthermore, both γ-secretase inhibitor and siRNA directed against Jagged-1 attenuated MSC-associated E-cadherin suppression and sphere formation in pancreatic cancer side population cells. Thus, our results suggest that MSC-derived myofibroblasts play important roles in regulating EMT and tumor-initiating stem cell-like properties of pancreatic cancer cells through an intermediating Notch signal.

During tumor progression, epithelial–mesenchymal transition (EMT) contributes considerably to the malignant characteristics of tumors such as local invasion and distant metastasis.[1, 2] Epithelial–mesenchymal transition has recently been reported as the key phenomenon that tightly regulates the stem cell-like characteristics of both normal and malignant cells.[3, 4] Side population (SP) technology has been widely used to isolate the stem cell-enriched fraction in a variety of tissue. Side population cells are detected by their own ability to efflux Hoechst33342 dye through an ATP-binding cassette membrane transporter. We recently found that SP cells from pancreatic cancer cells are highly responsive to transforming growth factor-β (TGF-β)-mediated EMT, invasion, and metastasis.[5] Our results suggest that SP cells are enriched with cells that undergo mesenchymal–epithelial transition (MET) after TGF-β-associated EMT. Thus, our results indicated that an EMT/MET conversion is tightly linked to malignant potential in pancreatic cancer, such as invasion/metastasis. However, the mechanisms by which the EMT/MET status is regulated within a tumor in vivo remains undetermined.

The tumor microenvironment consists of various stromal cells, including tumor-associated fibroblasts, endothelial cells, pericytes, adipocytes, and immune cells.[6] Among these cell types, cancer-associated fibroblasts (CAFs) and/or myofibroblasts have been recently implicated in regulating tumor progression, invasion, and metastasis.[7, 8] Cancer-associated fibroblasts and myofibroblasts secrete a number of important inflammatory mediators, including MMP-2, -3, and -9, that can alter the stromal ECM and potentiate invasion, cell motility, and metastasis.[9, 10] Recently, bone marrow-derived α-smooth muscle actin (α-SMA)-positive myofibroblast-like cells have been reported to contribute to cancer progression within tumor tissue.[11] Using a mouse model of inflammation-induced gastric cancer, Quante et al.[11] showed that at least 20% of CAFs originate from bone marrow and are derived from mesenchymal stem cells (MSCs). Thus, although the origin of CAFs and/or myofibroblasts remains unknown, these recent studies suggest that MSCs derived from either bone marrow or local tissues may partly contribute to the origin of CAFs or tumor-associated myofibroblasts.

In this study, we investigated whether MSCs enhance tumor formation using pancreatic cancer cell–MSC co-implantation models. We hypothesized that MSCs and/or MSC-derived myofibroblast-like cells regulate both EMT/MET and tumor-initiating stem cell-like characteristics of pancreatic cancer. We performed both in vitro co-culturing experiments and in vivo co-injection experiments to identify the roles of MSCs in pancreatic cancer progression. We found that MSCs contributed to the regulation of both EMT status and maintenance of so-called tumor-initiating stem cell (TISC)-like characteristics among pancreatic cancer cells. We focused on pancreatic cancer cells because pancreatic cancer is one of the aggressive cancers characterized by relatively large amounts of stroma within tumor tissue. Although some mechanisms and mediators are known to contribute to cancer cell–stromal cell interactions, we found that the Notch-associated signal appeared to contribute to the regulation of EMT/stemness by MSCs. The interactions between cancer cells and MSCs and/or MSC-derived myofibroblast-like cells could be an important target to prevent tumor progression, invasion, and metastasis in pancreatic cancer.

Materials and Methods

Cells and cell culturing

Pancreatic cancer PANC-1 cells were obtained from ATCC (Manassas, VA, USA). The MIAPaCa2 cell lines were obtained from the Health Science Research Resources Bank (Osaka, Japan). These cell lines are tested and authenticated by short tandem repeat profiling analysis. The MSCs were isolated from human bone marrow (Lonza, Walkersville, MD, USA). The PANC-1 and MIAPaCa2 cells were grown in DMEM (Sigma, St. Louis, MO, USA). All media were supplemented with 10% FBS and penicillin. Isolated MSCs were cultured in prime DMEM (Invitrogen, Carlsbad, CA, USA) supplemented with 20% FBS, penicillin, and 10 ng/mL basic fibroblast growth factor. Surgically resected pancreatic tissues (pancreatic cancer tissues or adjacent non-tumor tissues) were chopped into fragments and disrupted with 2 mg/mL collagenase L (Nitta Gelatin, Osaka, Japan) for 2 h at 37°C. Subsequently, cells were washed three times with Hanks' balanced salt solution containing 2% FBS. To exclude epithelial cells, cultured cells were labeled with anti-Ber-EP4 (Dako, Glostrup, Denmark) and anti-mouse IgG MicroBeads (Miltenyi Biotec, Auburn, CA, USA). Non-epithelial cells were collected as stromal cells using MACS technology (Miltenyi Biotec) according to instructions. Purified stromal cells were cultured in DMEM. This study was carried out with the approval of the Ethics Committee of the Keio University School of Medicine (No. 20040034; Tokyo, Japan).

Flow cytometry

Isolation methods for SP cells and main population (MP) cells that include high Hoechst33342 fluorescence have been described previously.[5] In brief, after the Hoechst3342 staining procedure, the cell fraction with high fluorescent intensity was identified as a majority of total cells, or MP cells. Side population cells were also identified as the cells that exclude Hoechst33342 dye by their enhanced ATP-binding cassette transporter activities.[5] To isolate MSCs, mononuclear cells from bone marrow were labeled with CD271 and CD90 antibodies. Labeled cells were analyzed using a Moflo flow cytometer (Beckman, Brea, CA, USA), and double-positive cells were sorted.

Xenograft experiments

Stromal cells and cancer cells were mixed, resuspended in 100 μL saline, and injected s.c. into 6-week-old male NOD/SCID mice (Charles River Laboratories International, Kanagawa, Japan) under anesthesia. Tumor diameters were measured weekly using a caliper. Tumor volumes were determined by the following formula: volume = 0.52 × length × width2.

Coculturing with MSCs

Indirect coculture

Prior to coculturing, MSCs were pre-treated with TGF-β for 3 days. We used Transwell chambers (Corning, Tewksbury, MA, USA). Transforming growth factor-β-treated stromal cells were plated into the upper chamber, and cancer cells were plated into the lower chamber.

Direct coculture

To allow re-isolation of cancer cells and stromal cells, PANC-1 cells were labeled with GFP by retroviral infection. Then, 1 × 105 TGF-β-treated cells (stromal cells) and 5 × 104 cells (cancer cells) per well in a 6-well plate were cultured for an appropriate time period. Subsequently, cocultured cells were resorted into GFP-positive cancer cells and GFP-negative stromal cells.

Notch reporter gene analysis

A Notch reporter system was constructed as described previously.[12] The constructs with tandem repeat of RBJ-binding sequences were followed by the dVenus gene. The constructed reporter vector was transfected to PANC-1 cells using Lipofectamine 2000 reagent (Invitrogen) according to the manufacturer's instructions. Forty-eight hours after transfection, Geneticin (100 μg/mL; Roche, Mannheim, Germany) was added. Transfected PANC-1 (Notch-PANC-1) cells were grown in the presence of Geneticin. To distinguish cancer cells and MSCs, TGF-β-treated MSCs (Tb-MSCs) were labeled with PKH26 dye (Sigma) according to instructions. Notch-PANC-1 SP cells or MP cells and PKH26 dye-labeled Tb-MSCs were cocultured directly for 3 days. The Notch-associated dVenus fluorescence was observed by flow cytometry.

Statistical analysis

Results are given as the mean ± SD from at least three experiments. Statistical comparisons were by Student's t-test. Significant P-values are denoted by asterisks.


Transforming growth factor-β treated MSCs enhance pancreatic cancer cell tumor progression

We first evaluated the effects of co-incubation with MSCs on the tumor-forming activity of pancreatic cancer cell lines. The MSCs were isolated from human bone marrow using CD90 and CD271 surface markers (Mabuchi et al., submitted).[13, 14] We compared the in vivo tumor volumes in the dorsal regions of mice after injecting either cancer cells alone or cancer cells that had been cocultured with MSCs. Unexpectedly, although coculturing with naïve MSCs (untreated) modestly enhanced tumor formation of pancreatic cancer cells, there were no dramatic differences between cancer cells alone and cancer cells plus naïve MSCs (data not shown). However, pretreatment of MSCs with TGF-β dramatically enhanced tumor volumes after co-injecting these cells into mice as compared with injecting either cancer cells alone or naïve MSCs plus cancer cells. Naïve MSCs or Tb-MSCs (1 × 105cells) and cancer cells (5 × 104 cells) were implanted into NOD/SCID mice, after which tumor volumes were measured weekly. After 8 weeks, s.c. tumors injected with Tb-MSCs showed a high rate of tumor growth compared with those injected with naïve MSCs (Fig. 1a). These results suggested that MSCs or MSC-derived cells enhanced tumor formation in pancreatic cancer cells. Transforming growth factor-β may alter the phenotypes of MSCs, and may make them suitable as cancer-associated stromal cells. Indeed, when MSCs were treated with TGF-β for 3 days, they showed morphological changes and became spatulate in appearance (Fig. 1b). To test whether Tb-MSCs attained a phenotype of cancer-associated stromal cells, we examined the expressions of cancer-associated stromal cells markers such as α-SMA, tenascin C, and podoplanin (PDPN) by immunoblotting (Fig. 1c). The Tb-MSCs expressed α-SMA, tenascin C, and PDPN proteins, whereas naïve control cells did not. Interestingly, the expression profiles of these stromal cell-associated proteins were similar to those of cancer-associated stromal cells isolated from surgically resected pancreatic cancer tissue samples (Fig. 1c).

Figure 1.

(a) Pancreatic cell lines were injected into NOD/SCID mice along with naïve (untreated) mesenchymal stem cells (MSCs) or MSCs pretreated with 7.5 ng/mL transforming growth factor (TGF)-β (Tb-MSCs) (n = 3). (b) Morphological appearance of naïve MSCs and Tb-MSCs. Scale bar = 200 μm. (c) Markers of tumor-associated stromal cells were detected by immunoblotting. Stromal cells isolated from tumor tissues were labeled as “Tumor”. Stromal cells isolated from non-tumor regions of surgical specimens were labeled as “Normal”. (d) Immunohistochemistry results for α-smooth muscle actin (α-SMA) and epithelial–mesenchymal transition-associated proteins in pancreatic cancer tissues from clinical specimens. Alpha-SMA high expression areas or low expression areas were marked by white squares. Left squares are α-SMA-low expression areas and right squares are α-SMA high expression areas (×4). PDPN, podoplanin.

We next evaluated the expression and localization of α-SMA-positive cells in clinical pancreatic cancer specimens. Positive staining of stromal cells for α-SMA was primarily localized in the area where cancer ducts lost well-defined glandular structures. Interestingly, the intensity of α-SMA staining showed a heterogeneous pattern, which we identified as α-SMA-low expression areas and α-SMA-high expression areas, as shown in Figure 1(d). In the α-SMA-high expression areas, cancer cells with liner membranous E-cadherin expression were rarely observed, whereas these areas were rich in vimentin-positive cells (Fig. 1d). The E-cadherin-negative/vimentin-positive cancer cells, which were solitary infiltrating cancer cells that had undergone EMT, were predominantly localized in the α-SMA-high expression areas (Fig. 1d, black arrows). Based on these results, we hypothesized that α-SMA-positive myofibroblast-like cells are associated with solitary infiltrating cancer cells that have undergone EMT.

Mesenchymal stem cells and/or MSC-derived myofibroblast-like cells induce EMT predominantly in pancreatic cancer SP cells

There is an increasing body of evidence that cancer-associated stromal cells have the potential for inducing EMT and are involved in micro-invasion and metastasis. Thus, we hypothesized that Tb-MSC-mediated tumor progression is associated with inducing EMT. We previously reported that SP cells, a fraction that is enriched with so-called TISCs, predominantly undergo EMT/MET in association with invasion/metastasis.[5] Thus, we investigated whether MSCs and/or MSC-derived myofibroblast-like cells could induce EMT in pancreatic cancer SP cells. We first used an indirect coculturing system with Transwell culture chambers. A total of 1 × 105 Tb-MSCs were placed in the upper chamber and 1 × 104 PANC-1 SP or MP cells were placed in the lower chamber. After 5 days of culturing, the shape of SP cells changed to spindle-like and their cell–cell contacts became very loose. These changes were not observed in MP cells (Fig. 2a). In addition, real-time PCR showed that coculturing with Tb-MSCs resulted in a significant reduction in E-cadherin expression and other EMT-associated gene alterations, including increased expressions of Slug and MMP-9 (primer sequences are shown in Table 1). Interestingly, these EMT-associated alterations were greater in SP cells than in MP cells (Fig. 2b). These results suggested that factors released from MSCs and/or MSC-derived myofibroblasts might contribute to inducing EMT in pancreatic cancer cells. As expected, the potential to undergo EMT was greater in SP cells.

Table 1. Primer sequences used in this study
Sequence 5′–3′Sequence 5′–3′
Figure 2.

(a) Pancreatic cancer PANC-1 cells were cocultured with transforming growth factor-β treated mesenchymal stem cells (Tb-MSCs) for 5 days using Transwell chambers. Representative images are shown for the morphologies of cocultured cancer cells. (b) Epithelial–mesenchymal transition-associated mRNA expression changes were evaluated by quantitative real-time PCR analysis. All results are the mean ± SD from at least three independent experiments. *< 0.05. (c) Side population (SP) and main population (MP) cells were isolated from GFP-labeled PANC-1 cells. Cancer cell cohesion with Tb-MSCs showed a mesenchymal cell-like appearance (black arrow), and it grew toward the inner space surrounded by Tb-MSCs with an epithelial appearance. (d) E-cadherin (Alexa555, red) and α-smooth muscle actin (α-SMA) (Alexa350, blue) expression were assessed by immunofluorescent staining. Cancer cells are stained green (GFP).

We further evaluated the cell–cell interactions between cancer cells and Tb-MSCs using a direct coculturing system, with 1 × 105 Tb-MSCs and 5 × 104 GFP-labeled PANC-1 (GFP-PANC-1) cells cultured in 6-well plates. Cancer cells were located between or alongside Tb-MSCs (Fig. 2c). During the coculturing period, the cancer cells appeared to grow toward the inner cavity surrounded by Tb-MSCs. Interestingly, the cancer cells located close to the Tb-MSCs possibly adopted spindle shapes (Fig. 2c, black arrow), whereas those cancer cells located away from the Tb-MSCs appeared to be epithelial. Thus, we carried out immunostaining for E-cadherin as described in Data S1. As expected, GFP-labeled cancer cells that directly contacted Tb-MSCs lost E-cadherin expression (Fig. 2d, white arrow), whereas E-cadherin-positive staining was detected for only those cancer cells that were located away from the Tb-MSCs. Interestingly, these findings were more obvious in GFP-PANC-1 cells cocultured with Tb-MSCs compared to GFP-PANC-1 cells cocultured with naïve MSCs.

Coculturing with Tb-MSCs enhances so-called TISC-like properties in pancreatic cancer SP cells

It is believed that there is an association between inducing EMT and acquiring stemness phenotypes. To evaluate the expressions of stemness-associated genes by coculturing with Tb-MSCs, we isolated SP cells or MP cells from GFP-PANC-1 cells directly cocultured with Tb-MSCs. After 3 days of culturing, GFP-positive SP or MP cells were resorted and collected (Fig. 3a). The SP cells showed twofold higher levels of CD133 mRNA expression compared with untreated MP cells. Coculturing with Tb-MSCs enhanced CD133 mRNA levels predominantly in SP cells, as the CD133 mRNA level was threefold greater in SP cells compared with that in MP cells. Similarly, mRNA expressions of other stemness genes including LGR5, Oct4, Nanog, and KLF4 were also enhanced by coculturing with Tb-MSCs (Fig. 3b).

Figure 3.

(a) GFP-positive cancer cells and negative transforming growth factor (TGF)-β treated mesenchymal stem cells (Tb-MSCs) were resorted and isolated for the following experiments. (b) mRNA expression changes of stemness-associated genes were evaluated by quantitative real-time PCR analysis. Results are the mean ± SD. *< 0.05. (c) Sphere-forming activity was evaluated. Results are the mean ± SD. *< 0.01, cocultured main population (MP) cells versus cocultured side population (SP) cells. (d) The SP cells or MP cells were injected into NOD/SCID mice (n = 4) along with naïve (untreated) MSCs or MSCs pretreated with 7.5 ng/mL TGF-β. Xenograft growth was evaluated by calculating tumor volumes (mm3). Data from each mouse (numbered 1–4) were expressed for the SP cells plus Tb-MSC group and the MP cells plus Tb-MSCs group. (e) Susceptibility to antitumor agent TNF-related apoptosis-inducing ligand (TRAIL) was evaluated by annexin V staining.

To further evaluate the stem-like characteristics of PANC-1 SP cells, we carried out sphere formation assays as described in Data S1. When SP cells and MP cells were plated without Tb-MSCs, they showed only modest sphere-forming activity (approximately 50 spheres from 1 × 104 cells). When either SP cells or MP cells were cocultured with Tb-MSCs, sphere-forming activity was dramatically enhanced in both. Interestingly, the extent of sphere formation was significantly greater in SP cells than MP cells (Fig. 3c). No spheres were observed when Tb-MSCs were cultured alone, suggesting that Tb-MSCs do not show sphere-forming activity. Thus, all spheres observed in this coculturing system were considered to be cancer cell-associated spheres. In addition, we investigated whether Tb-MSCs enhance tumor formation predominantly in SP cells. Either SP cells or MP cells (5 × 104 cells) were mixed with Tb-MSCs (1 × 105 cells) and injected into NOD/SCID mice. Notably, only Tb-MSCs dramatically enhanced the tumor-forming activity of SP cells isolated from PANC-1 cells; Tb-MSCs had negligible effects on tumor formation by PANC-1 MP cells (Fig. 3d).

Finally, we tested if sensitivity to an antitumor drug was regulated by coculturing with Tb-MSCs. The protocol is described in Data S1. Both PANC-1 SP cells and MP cells were sensitive to TNF-related apoptosis-inducing ligand (TRAIL) at 75 ng/mL, as approximately 80% of the cells died either by apoptosis or necrosis. After coculturing with Tb-MSCs, SP cells became relatively resistant, as 68.6% of SP cells remained alive after incubation with TRAIL. This effect was also observed in MP cells, although only 47.1% of MP cells survived after coculturing with Tb-MSCs (Fig. 3e). Thus, these results suggested that Tb-MSCs protected SP cells from TRAIL-associated cytotoxicity.

Transforming growth factor-β-treated MSCs regulate EMT and sphere formation in pancreatic cancer cells through a Notch-dependent mechanism

We next focused on the mechanisms by which Tb-MSCs might regulate cancer cell EMT or stemness properties. There are some MSC-derived molecules that could regulate EMT or stemness within the niche, including Notch-associated molecules, interleukin-6 and hepatocyte growth factor. Thus, we evaluated the expression profiles of these molecules before and after coculturing with cancer cells. Among the genes tested, Notch ligand Jagged-1 was most significantly upregulated within the MSCs after coculturing with pancreatic cancer SP cells. Jagged-1 mRNA levels were upregulated approximately sixfold after direct coculturing with cancer cells (Fig. 4a). We next evaluated if the Notch-associated signal is transduced within PANC-1 cells. Side population or MP cells obtained from the Notch-PANC-1 cells were cocultured with Tb-MSCs labeled by PKH26 dye. Coculturing was carried out under direct coculturing conditions for 3 days. The percentage of Notch signal-associated dVenus fluorescence-positive cells was 7.11% in MP cells and 14.9% in SP cells before coculturing with Tb-MSCs. Interestingly, coculturing with Tb-MSCs increased dVenus-positive cells to 19.2% in SP cells, whereas Tb-MSCs did not increase dVenus-positive cells in MP cells (8.35%) (Fig. 4b).

Figure 4.

(a) Expression levels of Jagged-1 mRNA were assessed by real-time PCR. Results are the mean ± SD from three independent experiments. *< 0.01 for transforming growth factor (TGF)-β treated mesenchymal stem cells (Tb-MSCs) versus MSCs cocultured with either side population (SP) or main population (MP) cells. (b) Notch signal-associated dVenus fluorescence was observed by flow cytometry. The percentage of dVenus positive cells was increased in SP cells by coculturing with Tb-MSCs. (c) Tb-MSC-mediated E-cadherin suppression was attenuated by inhibiting Notch signals either by a γ-secretase inhibitor (DAPT) (left) or by Jagged-1-specific siRNA (right). Results are the mean ± SD from three independent experiments. *< 0.01 for SP cells plus Tb-MSCs versus SP cells plus Tb-MSCs plus either DAPT or siRNA directed against Jagged-1. (d) SP cells or MP cells were cocultured with Tb-MSCs in the presence or absence of DAPT and sphere-forming activity was evaluated. Results are mean ± SD from three independent experiments. *< 0.01 for SP cells plus Tb-MSCs versus SP cells plus Tb-MSCs plus DAPT.

To test if the EMT of Tb-MSC-mediated cancer cells was Notch signal-dependent, we evaluated the effects of γ-secretase inhibitor IX (DAPT). Interestingly, Tb-MSC-mediated downregulation of E-cadherin was suppressed by DAPT only in SP cells. In contrast, only a minimal effect was found in MP cells (Fig. 4c). In addition, we used siRNA to suppress MSC-derived Jagged-1 (Fig. 4c). These results were similar to those using DAPT, as siRNA directed against Jagged-1 abrogated Tb-MSC-mediated E-cadherin suppression in SP cells, but only a minimal effect was detected in MP cells. These results suggest that Tb-MSC-associated EMT induction is Notch signal-dependent.

To determine whether Tb-MSC-mediated enhancement of SP cell sphere formation was Notch-dependent, we used DAPT and carried out the sphere formation assay (Fig. 4d). As expected, DAPT significantly suppressed sphere formation in SP cells cocultured Tb-MSCs. Our results suggest that Notch-associated EMT and/or sphere formation is predominantly observed in SP cells. Because SP cells have been reported to be enriched with cells that have so-called TISC-like properties,[15-17] our results may suggest that SP cells are enriched with cells that are more responsive to the so-called stemness signal Notch.[18, 19]


Epithelial–mesenchymal transition mediated by Tb-MSCs was clearly identified in SP cells isolated from pancreatic cancer cell lines, whereas EMT induction in the MP cell fraction was modest. We have reported that SP cells isolated from pancreatic cancer cells were highly responsive to TGF-β-mediated EMT.[5] Side population cells show better capability than MP cells for not only undergoing EMT but also returning to epithelial phenotype from a mesenchymal phenotype, the so-called MET, after removing TGF-β. The present study demonstrated that SP cells were more responsive to Tb-MSC-mediated EMT compared with MP cells (Fig. 2). In addition, taking our previous findings into account, controlling EMT in SP cells appears to be the key to preventing pancreatic cancer invasion and metastasis. Preventing cell–cell interactions between SP cells and MSC-derived stromal cells appears to be important for inhibiting SP cells from undergoing EMT.

Recently, EMT was reported to be linked with the expression of the so-called TISC properties.[3, 4] In the present study, we hypothesized that coculturing with MSCs and/or MSC-derived myofibroblast-like cells regulated the so-called TISC-like phenotypes of pancreatic cancer SP cells. As expected, Tb-MSCs dramatically enhanced TISC-like properties, based on the following observations: (i) the expression of certain TISC markers, including CD133, LGR5, Oct4, Nanog and KLF4; (ii) sphere formation within the context of an anchorage-independent ultra-low attachment culture system; (iii) in vivo tumor-forming activity; and (iv) resistance to an apoptosis-inducing stimulus, TRAIL. These results suggest that Tb-MSCs regulate TISC-like properties in pancreatic cancer cells. The MSCs and/or MSC-derived myofibroblast-like cells might provide an appropriate microenvironment.

In this study, we focused on the roles of α-SMA-positive myofibroblast-like cells in cancer cell EMT regulation and cancer progression. Our results indicated that α-SMA-positive myofibroblast-like cells were enriched in those areas in which cancer cells had undergone EMT, and were frequently identified in human pancreatic cancer specimens (Fig. 1d). Alpha-SMA-positive cells originating from bone marrow showed superior activity to enhance tumor formation in pancreatic cancer cell lines (Fig. 1a). In addition, we found that pretreatment with TGF-β induced MSCs to express certain molecules, including α-SMA, tenascin C, and PDPN, which are enriched within cancer stroma.[9] Thus, taken together with previous studies showing the roles of myofibroblasts in tumor progression,[8, 11, 20] α-SMA-positive myofibroblast-like cells play important roles in cancer progression in pancreatic cancer, especially in regulating EMT status and TISC-like characteristics. Because MSCs are well known to migrate from the bone marrow into tumor tissue,[21-23] it is reasonable that bone marrow-derived MSCs are potential sources of α-SMA-positive cells within the tumor stroma.

There is an increasing body of evidence that Notch-associated signal transduction contributes to the carcinogenic processes in pancreatic cancer.[19, 24, 25] Notch appears to enhance pancreatic cancer progression, as Wang et al.[26, 27] reported that inhibiting Notch signaling with either siRNA or γ-secretase inhibitors reduced pancreatic cancer cell growth and invasion/metastasis. Mullendore et al.[28] reported that Notch ligands as well as their target genes, (i.e., HES, HEY, and VEGF) were upregulated within pancreatic cancer tissues. De La O JP et al. reported that Notch signals cooperated with a K-Ras-dependent pathway to mediate pancreatic carcinogenesis.[29, 30] Furthermore, Notch signals may contribute to pancreatic TISC regulation as forced activation of Notch signals accelerates self-renewal in pancreatic TISCs.[18, 19] In this study, we detected that Jagged-1, one of the major Notch ligands, was upregulated after coculture with pancreatic cancer cells in Tb-MSCs (Fig. 4a). In addition, both the γ-secretase inhibitor DAPT and siRNA directed against Jagged-1 could effectively suppress Tb-MSC-mediated PANC-1 cell EMT and sphere formation. These results are consistent with previously reported findings suggesting the importance of Notch signaling in pancreatic cancer progression. Our results suggest that Notch ligands, such as Jagged-1, are provided by MSCs and/or MSC-derived myofibroblast-like cells during the interaction between pancreatic cancer cells and MSCs. Our results also suggest that pancreatic cancer cells undergo EMT through Notch-dependent mechanisms during the interactions between cancer cells and MSCs. Another finding suggested by the present study was that coculturing with Tb-MSCs mediated apoptosis resistance, one of the important characteristics of so-called TISCs, in pancreatic cancer cells. Although detailed mechanisms by which Tb-MSCs mediate cell death resistance is unknown, possible cross-talk between the Notch signaling pathway and cell death cascade has been suggested by previous studies. Notch signals may affect the expression profile of apoptosis-regulating molecules, that is, upregulation of anti-apoptotic molecules including Bcl-2, Mcl-1, and XIAP.[31-33] Furthermore, the Notch-associated signal also suggested enhancing so-called cell survival signals including nuclear factor-κB and the PI3K/Akt pathway.[34, 35]

The contribution of Notch-associated mechanisms on EMT induction under Transwell culture conditions (Fig. 2a,b) remains unclear. Although we identified that Tb-MSCs expressed certain Notch ligands, such as Jagged-1 or DLL4, we did not determine if soluble forms of such ligands were released into the culture medium through the Transwell chamber. However, in another series of experiments, we identified that SP cells upregulated Notch ligands after culturing with Tb-MSC under Transwell conditions (data not shown). Therefore, we could hypothesize that other soluble factors released from Tb-MSCs may affect SP cells, resulting in Notch signal transduction and/or induce EMT directly within the SP cells.

In conclusion, our results suggest that MSCs and/or myofibroblast-like cells originating from MSCs have the capability to regulate cancer cell EMT status. Mesenchymal stem cells also regulate TISC-like characteristics, such as the expression of so-called TISC markers, the ability to form spheres in anchorage-independent culturing conditions, and the potential to form tumors in vivo. Mesenchymal stem cells cocultured with pancreatic cancer cells express certain Notch ligands, such as Jagged-1, to mediate EMT or sphere formation in pancreatic cancer cells through Notch-dependent mechanisms. These results are consistent with a scenario in which MSC-derived myofibroblast-like cells may function as a TISC niche by providing a microenvironment for regulating TISC maintenance and EMT status. Targeting the MSC-associated microenvironment may be an attractive strategy to prevent cancer progression and invasion/metastasis. The signal transduction processes of the Notch pathway are a potential target for preventing MSC–cancer cell interactions.


This work was supported by a grant-in-aid for scientific research from the Japanese ministry of education, scientific and culture of Japan (A.K.-N. is supported by grant 23-5599; H.H. is supported by grant 24591017). The authors acknowledge Dr Minoru Kitago for his management to obtain surgical specimens, Mr Sadafumi Suzuki for his technical support with flow cytometry and Mr Toru Igarashi for his technical assistance.

Disclosure Statement

The authors have no conflict of interest.