Lineage Tracing Evidence for In Vitro Dedifferentiation but Rare Proliferation of Mouse Pancreatic β-Cells

The following transgenic mouse strains were used: Insulin-CreER (3), Insulin-Cre (16) (a gift of Mark Magnuson), ROSA26-LSL-YFP (17) (a gift of Frank Costantini), Z/AP (18) (a gift of Corrinne Lobe), and MIP-GFP (19) (a gift of Manami Hara). Mice were genotyped by PCR on DNA extracted from ear punch biopsies using the following primers: for Cre, 5′-tgccacgaccaagtgacagc-3′ and 5′-ccaggttacggatatagttcatg-3′ (produces a 600-bp PCR product); and for yellow fluorescent protein (YFP), 5′-ctggtcgagctggacggcgacgtaaac-3′ and 5′-atgtgatcgcgcttctcgttgggg-3′ (produces a 600-bp PCR product). The Z/AP transgene was identified by X-gal staining on tail biopsies. To activate cre recombinase in transgenic mice carrying the Insulin-CreER transgene, tamoxifen (20 mg/ml in corn oil; Sigma) was injected subcutaneously (five injections of 8 mg each, spread over a period of 2 weeks). In all mouse procedures, we followed the institutional guidelines of the Hebrew University.

Islets were prepared from 2- to 6-month-old mice by ductal perfusion of collagenase P (Roche). Hand-picked islets were plated, either intact or dissociated to single cells with trypsin, on gelatin-coated coverslips in 24-well plates. Culture medium was CMRL 1066 (Invitrogen) containing 5.6 mmol/l glucose, supplemented with 10% FBS, 2 mmol/l l-glutamine, 100 units/ml penicillin, and 100 μg/ml streptomycin. Medium was changed twice a week.

Expression of YFP was detected using either endogenous green fluorescence or indirectly using a rabbit anti-GFP antibody (1:250; Invitrogen). Expression of human placental alkaline phosphatase (HPAP) was detected using a monoclonal mouse anti-HPAP antibody (1:500; Sigma). Other primary antibodies used were guinea pig anti-insulin (1:200; Dako), goat anti–C-peptide (1:100; Linco), guinea pig anti-pdx1 (1:2,500; a gift of Chris Wright), rabbit anti-Glut2 (1:100; Chemicon), rabbit anti–prohormone convertase 1/3 (anti-PC1/3) (1:100; Chemicon), mouse anti-bromodeoxyuridine (anti-BrdU) (1:300; Amersham), rabbit anti-glucagon (1:200; Dako), rabbit anti-somatostatin (1:200; Zymed), guinea pig anti-pancreatic polypeptide (1:100; Dako), hamster anti-Muc1 (1:100; Lab Vision), rabbit anti-carboxypeptidase A (1:200; Rockland), and rabbit anti-Ki67 (1:200; NeoMarkers). BrdU staining was preformed using a Cell Proliferation kit (Amersham). Secondary antibodies conjugated to CY2, CY3, or CY5 were from Jackson Immunoresearch (1:200). Cultured islet cells were fixed with cold 4% paraformaldehyde at room temperature for 10 min. For Ki67 staining, cells were permeabilized with 0.1% Triton X-100 (Sigma) in PBS at room temperature for 5 min. Cells were incubated with primary antibodies diluted in 1% BSA in PBS at room temperature for 30 min, washed with PBS, and incubated with secondary antibodies for 30 min. For nuclear counterstain, we used DAPI (1:100; Sigma), propidium iodide (1:1,000; Sigma), or toto3 (1:500; Invitrogen). Images were taken using a Nikon 90i C1 confocal microscope. Parallel islet cell cultures from wild-type mice and samples from transgenic mice incubated in the absence of primary antibodies were used as negative controls. To detect YFP expression in the pancreas, tissue was fixed for 3 h in cold 4% paraformaldehyde, incubated for 48 h in 30% sucrose, and embedded in OCT. Ten-micrometer frozen sections were refixed with cold 4% paraformaldehyde, permeabilized with methanol:acetone (1:1) at −20°C for 20 min, blocked with Casblock solution (Zymed), and stained with primary and secondary antibodies diluted in Casblock.

To identify recombination events in islet cultures, genomic DNA was extracted using proteinase K (Roche) in 10× PCR buffer (Promega), incubated at 55°C for 10 min, inactivated at 95°C for 20 min, and amplified (35 cycles) as follows. for cultures of ROSA26-LSL-YFP mice, we used primer 1, 5′-ctggtcgagctggacggcgacgtaaac-3′; primer 2, 5′-atgtgatcgcgcttctcgttgggg-3′; primer 3, 5′-GGCCGCTCTAGAACTAGTGGATCCG-3′; primer 4, 5′-CACCCCGGTGAACAGCTCCTCGCCC-3′; primer 5, 5′-GCCGCTTGGGTGGAGAGGCTATTC-3′ and primer 6, 5′-GAGCAAGGTGAGATGACAGGAGATC-3′. Primers 1 plus 2 amplify a 600-bp fragment from the YFP transgene, confirming the presence of amplifiable transgenic DNA. Primers 3 plus 4 amplify a 200-bp fragment from the recombined locus, indicating the presence of cells that underwent Cre-mediated excision. Nonrecombined genomes should produce a longer 3-kb fragment, which was not observed under the PCR conditions used. Primers 5 plus 6 amplify a 280-bp fragment, present only in the nonrecombined locus and absent after recombination. For cultures of Z/AP mice, we used primer 7, 5′-accattggcttgagtgcagccgcccgc-3′; primer 8, 5′-cagggcagcctctgtcatctccatcag3′; primer 9, 5′-CCTACAGCTCCTGGGCAACGTGCTGG-3′; primer 10, 5′-CTCCTCCTCAACTAGGATGATGCCC-3′; primer 11, 5′-caacagttgcgcagcctgaatgg- 3′; and primer 12, 5′-aaatcgctgatttgtgtagtcggt-3′. Primers 7 plus 8 amplify a 600-bp fragment from the HPAP transgene. Primers 9 plus 10 amplify a 1,500-bp fragment from the recombined locus; a nonrecombined genome should produce a longer, 5.7-kb fragment, which was not observed under the PCR conditions used. Primers 11 plus 12 amplify a 520-bp fragment from the nonrecombined locus, which is lost upon recombination. Fragments from the β-actin or HPRT genes were amplified as additional controls for the presence of genomic DNA.

To ask whether insulin-negative cells seen in islet cultures originate from differentiated β-cells, we used a genetic lineage tracing approach (3), which allows identification of differentiated β-cells and their progeny via Cre-mediated recombination in transgenic mice. We first adapted the protocols established for adherent human islet cultures (10,11) to mouse islets. Wild-type mouse islets were hand-picked after mild collagenase digestion of the pancreas and cultured either intact or after dissociation to single cells as described in research design and methods. Similarly to what has been reported for human islets, we found that mouse islet cultures gradually lose insulin staining and are overtaken by insulin-negative cells. When mouse islets are dissociated to single cells before culture (10), most insulin staining is lost within 2 weeks, and extensive proliferation of insulin-negative cells is observed (Fig. 1A; data not shown). When intact mouse islets are plated (11), insulin-negative, proliferating mesenchymal cells migrate out of islets, and a gradual loss of insulin staining is seen as well (Fig. 1B; data not shown). Similar results were obtained when we plated islets of MIP-GFP mice (19): 2 weeks after plating, all green fluorescence disappeared (Supplementary Fig. S1, which is detailed in the online appendix [available at http://dx.doi.org/10.2337/db06-1654). Cultures of dissociated islets (10) manifested a more rapid and uniform loss of insulin staining and an efficient expansion of insulin-negative cells, and therefore this protocol was used in most of the experiments described below, unless indicated otherwise.

After confirming that insulin staining is gradually lost from cultured mouse islets, we designed genetic lineage tracing experiments to permanently label β-cells and determine their fate in such cultures. If β-cells undergo dedifferentiation, cells should be found in cultures that lack markers of differentiated β-cells, such as pdx1 and insulin, but express the lineage markers of β-cells (because expression of these reporters is driven by ubiquitous promoters insensitive to dedifferentiation). In contrast, death of β-cells and their progeny should result in the absence of lineage markers in cultures. Such a result would mean that islet cultures are derived from non–β-cells. The logic of the system is described schematically in Fig. 2A. To label differentiated β-cells, we used either Insulin-Cre (16) or Insulin-CreER mice (3), the latter requiring tamoxifen injection for the activation of Cre. As a reporter for Cre-mediated recombination, we used two different mouse strains: Z/AP mice, in which Cre-mediated recombination leads to the permanent expression of HPAP (18), and ROSA26-LSL-YFP mice, in which Cre-mediated recombination leads to the permanent expression of YFP under the ROSA26 promoter (17). We have previously shown that HPAP expression in tamoxifen-injected Insulin-CreER mice is specific to β-cells (3). To demonstrate that recombination in Insulin-Cre;ROSA26-LSL-YFP and Insulin-CreER;ROSA26-LSL-YFP mice is also specific to β-cells, frozen sections from such double-transgenic mice were stained for insulin or other pancreatic markers. Confocal images showed that as expected, YFP fluorescence is limited to insulin-expressing cells, with ∼80% efficiency of β-cell labeling in both Cre driver lines (Fig. 2B). On the basis of these results, the experiments described below were carried out using different combinations of double-transgenic mice (Insulin-Cre or Insulin-CreER and Z/AP or ROSA26-LSL-YFP transgenes). All double-transgenic combinations gave similar results (see below), ruling out transgene-specific artifacts.

Islets were isolated from 2- to 6-month-old double-transgenic mice, dissociated or left intact, and cultured as detailed in research design and methods. As expected, isolated islets contained abundant labeled cells at their cores (Fig. 3A, a). When intact islets were cultured, cells with a mesenchymal morphology started to migrate out of the islets after a few days, as described previously (6–11). Importantly, the vast majority of these cells did not express the β-cell lineage label (Fig. 3A, c), suggesting that they were pre-existing non–β-cells. After dissociation, ∼30% of total islet cells were labeled with YFP or HPAP (Fig. 3A, b). This figure reflects the abundance of recombined β-cells (80% of total β-cells) among other islet endocrine cells, nonrecombined β-cells, islet endothelial cells and fibroblasts, and contaminating exocrine cells. After 7–14 days of culture of dissociated islet cells, numerous YFP⁺ or HPAP⁺ cells were observed, representing cells that likely had expressed insulin in vivo. Many of these cells had a mesenchymal morphology different from the typical epithelial appearance of β-cells (Fig. 3B, a–g). Importantly, many labeled cells were completely devoid of insulin, as determined by immunofluorescence, whereas others contained low levels of insulin and appeared degranulated (Fig. 3C, a–d). Most YFP⁺ or HPAP⁺ cells also lacked expression of the key β-cell transcription factor pdx1 (Fig. 3C, f), and many had undetectable levels of the two other β-cell markers analyzed, Glut2 (Fig. 3C, e) and PC1/3 (Fig. 3C, g). The frequency of label-positive, insulin-negative cells increased with time. Although no such cells were observed at the beginning of cultures, 27% of label-positive cells were insulin-negative at day 8 of culture, and 74% of label-positive cells were insulin-negative after 21 days in culture (Fig. 3D). These results demonstrate that mouse β-cells can lose their differentiated phenotype, yet remain viable, under these culture conditions.

We then examined the proliferation status of β-cells and their dedifferentiated progeny. Immunostaining for Ki67, a marker of proliferating cells, showed multiple proliferating cells in islet cultures, however, the vast majority of these cells were negative for both insulin and the lineage labels of β-cells (Fig. 4A). Similarly, BrdU incorporation during a 24-h period before immunostaining resulted in the accumulation of multiple BrdU⁺ cells, but only rare cells among them were labeled as β-cell progeny (Fig. 4B). Figure 4C shows examples of rare dividing dedifferentiated β-cells, identified by their lineage label. The proliferation rate of dedifferentiated β-cells was ∼1:4,000 or 0.024% (n = 37,600 labeled cells). The results indicate that under these culture conditions, dedifferentiated mouse β-cells are largely postmitotic.

Consistent with observed minimal proliferation of dedifferentiated β-cells, the proportion of β-cells and their progeny in islet cultures gradually declined. Although the frequency of β-cells at the beginning of cultures was 30%, labeled cells consisted of only 0.55% of the cells in islet cultures after 21 days. After 4 weeks of culture (which involved 1–3 passages), lineage-labeled β-cells were virtually undetectable (Fig. 5A). Cultures derived from Insulin-CreER;Z/AP mice were almost homogenously lacZ⁺, providing an independent evidence that the cultures were dominated by nonrecombined cells (and therefore non–β-cells) (Fig. 5B). To rule out the possibility that the failure to observe label-positive cells reflected silencing of the reporter transgenes, we analyzed genomic DNA from these cultures. PCR reactions were carried out to amplify specifically the recombined DNA (which should reflect the presence of β-cell progeny regardless of reporter transgene expression). As shown in Fig. 5C and D, long-term cultures had no evidence for the presence of the recombined DNA. Finally, we performed immunostaining to characterize the cells that eventually dominated islet cultures. As shown in Fig. 5E, the majority of cells in long-term islet cultures expressed vimentin, suggesting that these are the progeny of islet-associated mesenchymal cells. Endothelial cells (identified by platelet-endothelial cell adhesion molecule staining Fig. 5E) and pancreatic exocrine cells (identified by amylase staining for acinar cells and Muc1 for ductal cells; data not shown) were virtually absent from long-term islet cultures. These results indicate that long-term mouse islet cultures are dominated by the progeny of non–β-cells and that dedifferentiated β-cells are gradually eliminated from such cultures.

Efforts to induce proliferation of differentiated β-cells in vitro have traditionally resulted in the disappearance of such cells from cultures and the expansion of cells with a mesenchymal phenotype. Recent studies raised the exciting possibility that proliferating mesenchymal cells in cultures of human islets were in fact the progeny of β-cells that underwent dedifferentiation into a mesenchymal, proliferative state. It is widely assumed that such cells might maintain some of the chromatin structure associated with their previous differentiation program and thus represent good candidates for redifferentiation into functional β-cells after expansion. To directly examine this provocative hypothesis, we used genetic lineage tracing, in which differentiated β-cells were heritably labeled in vivo, and their fate was followed in islet cell cultures. The results demonstrate that cultured β-cells can undergo dedifferentiation, including the loss of multiple β-cell markers, such as pdx1, insulin, Glut2, and PC1/3, and a shift to a mesenchymal morphology. However, dedifferentiated mouse β-cells only rarely proliferated and were eliminated from the cultures after several weeks. Thus, long-term cultures of mouse islets, in the growth conditions used here, are dominated by proliferating cells that originate from non–β-cells. Most of these cells have a mesenchymal morphology and express vimentin, suggesting that they are the progeny of islet-associated mesenchymal cells.

Why does the transition from the in vivo environment to culture conditions induce such a rapid shut-off of β-cell differentiation? Our results indicate that cell division is not a prerequisite for in vitro dedifferentiation. Dedifferentiation is also unlikely to reflect a general cellular stress, because dedifferentiated β-cells remain viable for days and weeks. We suggest that it is the disruption of islet three-dimensional organization that induces β-cell dedifferentiation. Specifically, the detachment of β-cells from their supporting bed of endothelial cells and endothelial cell–derived extracellular matrix is a likely triggering event for dedifferentiation. Consistent with this idea, cultured intact islets retain the differentiated β-cell phenotype for longer periods (weeks), compared with dissociated islet cells (days), and the addition of endothelial cell–derived matrix improves islet yield and maintenance in culture (20–25). Islet vasculature was recently proposed to be important for β-cell survival and for proper expression of the insulin gene (26,27). We speculate that not only insulin expression but the entire differentiation program of β-cells depends on signals from islet vasculature. Future experiments using lineage-labeled β-cells co-cultured with endothelial cells will allow testing of this hypothesis. Finally, cell-to-cell contacts among β-cells and between β-cells and other islet cells may play an important role in maintaining β-cell differentiation and in facilitating β-cell redifferentiation. The lineage tracing system described here provides a unique platform for the identification of factors that modulate the extent of β-cell dedifferentiation (as opposed to β-cell death) and the ability of dedifferentiated β-cells (otherwise not identifiable) to proliferate, survive, and redifferentiate.

What is the relationship between β-cell dedifferentiation and β-cell proliferation? Our results demonstrate that β-cell dedifferentiation does not necessarily lead to proliferation. It is possible that transient dedifferentiation is nevertheless a prerequisite for β-cell proliferation (13,14), for example, because the high density of secretory vesicles might interfere with cell mitosis. The idea that β-cell division involves transient dedifferentiation remains to be carefully tested using in vivo models of β-cell proliferation, even though the occurrence of human and rodent insulinomas with a high insulin content (28,29) would argue against this possibility.

We note two potential caveats in our experiments. First, we have selected islet culture conditions based on published protocols that proposed dedifferentiation and expansion of β-cells (10,11). Although we show that under these conditions, dedifferentiated β-cells are not proliferative, other culture conditions still to be defined may be more conducive for expansion of dedifferentiated β-cells. Second, the recent proposition of β-cell dedifferentiation was based on studies of human islets (10,11), whereas our lineage tracing experiments were performed in mice. Human islets may behave differently (for example, dedifferentiated human β-cells may have a higher proliferation index), and lineage tracing of human β-cells remains an important future challenge (30).

In conclusion, we provide definitive lineage tracing evidence that mouse β-cells can dedifferentiate in vitro into an insulin-negative state. Under present culture conditions, these cells do not proliferate significantly and are eventually eliminated. Future experiments will seek to identify the determinants of β-cell dedifferentiation, the potential of these cells to proliferate, and their ability to differentiate into functional β-cells or other fates. The idea that transient dedifferentiation of β-cells may occur in vivo under certain conditions also merits a thorough investigation.

Y.D. has received grants from the Beta Cell Biology Consortium, the Juvenile Diabetes Research Foundation (JDRF), and the Israel Science Foundation. S.E. has received grants from the Juvenile Diabetes Research Foundation (JDRF) and the Israel Science Foundation.

We thank Frank Costantini, Corrinne Lobe, Mark Magnuson, Manami Hara, and Chris Wright for sharing mice and antibodies; Antonello Pileggi for advice on islet preparation; Marvin Gershengorn for discussions; and Gilad Grader for assistance.