Islet-Derived Fibroblast-Like Cells Are Not Derived via Epithelial-Mesenchymal Transition From Pdx-1 or Insulin-Positive Cells

For mouse experiments, the following strains were used: C57BL/6J, Ins2cre, Z/EG (The Jackson Laboratories, Bar Harbor, ME), Pdx1cre (Pedro Herrera, University of Geneva), and MIP-GFP (Manami Hara, University of Chicago). Transgenic offspring were genotyped by PCR, as described in the supplemental research methods (see online appendix, available at http://diabetes.diabetesjournals.org). All transgenic animals were bred and maintained under specific pathogen-free conditions. University of Minnesota housing and husbandry guidelines were adapted from requirements in the Animal Welfare Act and the Guide for the Care and Use of Laboratory Animals (4).

Adult mice (10–24 weeks of age) were anesthetized with 250 mg/kg Avertin, the abdomen was opened, the pancreas was perfused, and the islets were purified as detailed in the supplemental research methods. Islet fractions were then seeded in a low-serum–, platelet-derived growth factor BB–, epidermal growth factor–, and leukemia inhibitory factor–supplemented culture media formulation (see supplemental research methods for detailed culture protocol) or as previously described (2,3). Bone marrow mesenchymal stem cells (BM-MSCs) were obtained from Dr. Darwin Prockop (Tulane University) and cultured as described (5). Methods for differentiation of Sca1⁺ fibroblast-like progenitors are described in the supplemental research methods.

Cells were harvested using 0.25% trypsin (Mediatech, Herndon, VA), washed with PBS plus 2% fetal bovine serum and 1 mmol/l EDTA (Ambion, Austin, TX), and filtered through a 40-μm cell strainer (BD Biosciences, San Jose, CA) to remove remaining cell clusters. Cells from total islet cultures were stained with anti–CD45-APC (30-F11) and Sca1-PE (E13-161.7) (BD Biosciences), and CD45⁻/Sca1⁺ cells were selected using a FACSAria (BD Biosciences). Alternatively, 1 × 10⁵ Sca1⁺–cultured cells were harvested and stained with anti–CD45-APC (30-F11), Sca1-PE (E13-161.7), and CD44-FITC [fluorescein isothiocyanate] (IM7) (BD Biosciences) and analyzed using the FACSCalibur flow cytometer (BD Biosciences). Anti-rat IgG_2bκ-APC (A-95-1), IgG_2bκ-FITC (A-95-1), and IgG_2aκ-PE (R35-95) antibodies (BD Biosciences) were used as isotype controls.

Total RNA (1 × 10⁵–1 × 10⁶ cells) was isolated using an RNeasy Micro Kit (Qiagen, Valencia, CA). RNA was treated with Turbo DNA-free (Ambion) according to the manufacturer’s protocol and reverse transcribed into cDNA using Superscript III with random hexamers (Invitrogen, Carlsbad, CA). For real-time RT–quantitative PCR (RT-qPCR), cDNA was amplified using SYBR Green PCR Master mix with an ABI Prism 7700 (Applied Biosystems, Foster City, CA), as described in the supplemental research methods.

Genomic DNA was obtained from trypsinized and filtered day 10–14 Pdx1cre-Z/EG (ZPC) cultures using the QiaAmp DNA Micro Kit (Qiagen). PCR detection of excision was performed using excision (chicken β-actin promoter/cytomegalovirus enhancer [CAAGS]-GFP)–and nonexcision (CAAGS-LacZ)–specific primers (Table 1 of online appendix). Standard PCR or real-time qPCR was performed using Platinum TaqDNA Polymerase (Invitrogen) or SYBR Green PCR Master mix, respectively, as described in the supplemental research methods. As a relative (100%) control, tail genomic DNA was obtained from a double-transgenic Actbcre-Z/EG animal. This animal provides expression of both Cre recombinase under control of the β-actin promoter and the Z/EG reporter.

Mouse islet fractions from transgenic Ins2cre-Z/EG (ZI) mice were cultured in 24-well culture plates (BD Biosciences) in culture media as described. Once confluent, fibroblast-like cells were fixed in 0.2% glutaraldehyde/2% formaldehyde for 5 min and stained overnight at 37°C with X-gal stain as described (6).

Consistent with previous reports on human islet cultures (2,3), we observed a population of fibroblast-like cells appearing when mouse islet–enriched fractions (Fig. 1A, a) were cultured in medium containing 2% fetal bovine serum, epidermal growth factor, platelet-derived growth factor BB, and leukemia inhibitory factor (Fig. 1A, b). Within this population, ∼10–20% of the cells were CD45⁻ and expressed the stem cell antigen Sca1 (Fig. 1B, a). The CD45⁻/Sca1⁺ cells, purified by fluorescence-activated cell sorter (FACS) analysis, could be maintained for >40 passages in the same media (Fig. 1A, c and d). The expanded Sca1⁺ cells coexpressed CD44 but not CD45 (Fig. 1B, b) and expressed characteristic epithelial and mesenchymal transcripts (Fig. 2A), a phenotype similar to that of islet-derived fibroblast-like cells described by others (2).

When switched to serum-free differentiation conditions (with or without individual or combinations of cytokines, including activin, bone morphogenetic protein-4, growth and differentiation factor-11, betacellulin, exendin-4, and/or hepatocyte growth factor), the mouse Sca1⁺ pancreas-derived cells did not acquire a more mature pancreatic phenotype (data not shown). We observed, however, a robust differentiation to lipid-filled adipocytes (Fig. 2B).

As the CD44⁺/Sca1⁺ phenotype and adipocyte differentiation potential of the Sca1⁺ pancreatic progenitors was reminiscent of BM-MSCs, we compared Sca1⁺ pancreatic progenitors with BM-MSCs. Like pancreas-derived Sca1⁺ cells, murine BM-MSCs were Sca1⁺ and CD44⁺ by flow cytometry (Fig. 1C). Real-time RT-qPCR demonstrated that both cell types expressed low levels of most epithelial markers (Fig. 2A), with pancreas-derived Sca1⁺ cells expressing high levels of CK18 and -19 and BM-MSCs expressing moderate levels of CK19. Likewise, undifferentiated human MSCs have been found to express both CK18 and -19 (7). Both mouse Sca1⁺ pancreas-derived cells and BM-MSCs expressed very low or undetectable levels of the endoderm/pancreas markers Afp, Sox17, Hnf1β, Hnf3β, and Pdx1 but expressed high levels of the mesenchymal markers Vim, Acta2, Cdh2, Csf1, Zeb, Snail1, and Snail2 (Fig. 2A).

To conclusively determine whether fibroblast-like cells in our pancreas cultures could be explained by EMT, we performed studies using two Cre-Lox recombination-based genetic lineage-tracing models. Islet-enriched fractions were cultured from ZI and ZPC murine pancreata using the expansion culture conditions described above. In these animals, the Pdx-1 or insulin-2 promoters, respectively, drive Cre recombinase expression, causing excision of the βgeo gene in the Z/EG reporter. This event replaces βgeo expression with permanent green fluorescent protein (GFP) expression (8) (Fig. 3A). In agreement with previously published reports (8,9,10), ZI animals displayed robust islet-specific GFP labeling (Fig. 3B, a), while ZPC animals have permanent GFP labeling of all endodermal pancreatic lineages (Fig. 3B, e). In cultures from ZI or ZPC islet-enriched fractions, few, if any, fibroblast-like cells expressed GFP, as evaluated by immunofluorescence microscopy (Fig. 3B, b–d, f, and g) or flow cytometry (Fig. 3C). Individual fibroblast-like GFP-positive cells were very rarely observed (Fig. 3B, g). Standard PCR (Fig. 3D, a) and real-time qPCR (Fig. 3D, b) on genomic DNA isolated from nonclustered cells within ZPC bulk cultures revealed minimal evidence of lineage tracing, as indicated by a lack of amplification of the CAAGS-GFP–excised Z/EG construct. The low level of excision detectable only by real-time qPCR was likely due to the persistence of a small number of individual mature pancreatic endocrine or exocrine cells remaining in the filtered, analyzed cell population. That the reporter construct was not silenced during culture is shown by persistent X-gal staining of fibroblast-like cells from ZI islet fractions (Fig. 3B, h). To rule out the possibility that the media formulation selectively expanded nonendodermal fibroblast-like cells, we also performed identical lineage-tracing studies using media as previously described (2,3). When islet-enriched fractions from ZI mice (Fig. 3E, a) were cultured for 14 days in this media, we again did not observe GFP-positive fibroblast-like cells (Fig. 3E, b–d).

These lineage-tracing studies provide evidence, under the tested conditions, that endodermal pancreatic cells or β-cells do not undergo EMT and do not significantly contribute to the proliferating fibroblast-like cell population. When islet fractions from MIP-GFP mice (11) were cultured under our conditions, some GFP-positive cells were found to attain a fibroblast-like morphology (Fig. 1 of online appendix). As we did not stain these cultures for mesenchymal markers, we do not know whether these cells coexpressed β-cell and mesenchymal markers, as has been previously observed (2,3). However, as revealed through the lineage-tracing studies, these insulin-expressing cells did not significantly contribute to the proliferative fibroblast-like population. This lack of expansion precluded testing whether these cells could redifferentiate into mature β-cells. It is unknown whether differences between our results and those recently published (2,3) are due to undetermined differences in culture protocols or reflect the use of mouse pancreatic tissue rather than human tissue.

Here we demonstrate that in two different media formulations, fibroblast-like cells can be cultured from mouse islet–enriched pancreatic fractions. Lineage-tracing studies conclusively demonstrate that these fibroblast-like cells are not derived from a β-cell or endoderm pancreatic origin and therefore do not result from β-cells undergoing EMT. By contrast, the phenotype and differentiation potential of the proliferating fibroblast-like cells appears similar to that of BM-MSCs; hence, these cells likely represent mesenchymal cells from within the pancreas.

These studies were supported by National Institutes of Health award U19 DK61244, a grant from the Juvenile Diabetes Research Foundation, and a research grant from Athersys (Cleveland, OH).