Nestin-Lineage Cells Contribute to the Microvasculature but Not Endocrine Cells of the Islet

The Institutional Animal Care and Use Committees of the University of Michigan approved all animal procedures. C57BL/6J-Gtrosa26^tm1Sor mice (Jackson Laboratories) were mated with nestin-cre mice (a gift from Rudy Jaenisch, Massachusetts Institute of Technology, and Gail Martin, University of California, San Francisco [28]) that carry a 1.8-kb nestin promoter/enhancer fragment from the nestin second intron that drives the expression of cre-recombinase. F1 mice were killed at 3 weeks or 3 months of age, and pancreata were fixed for immunohistochemistry or β-galactosidase activity (29). F1 mouse embryos were harvested at embryonic day 12.5 from pregnant dams and immediately fixed for sectioning and immunohistochemistry as previously described (29).

Mouse insulin promoter (MIP-GFP) mice were generated by pronuclear injection of a construct containing the mouse insulin I promoter (MIP)–green fluorescent protein (GFP) containing an 8.5-kb fragment of MIP that includes a region from −8.5 kb to 12 bp (relative to the transcriptional start site), the coding region of enhanced GFP (0.76 kb) (Clontech, Palo Alto, CA), and a 2.1-kb fragment of the human growth hormone gene cassette (30).

Ducts were isolated by collagenase digestion (P2; Boehringer Mannheim) using a protocol described by Bonner-Weir et al. (31). The ducts were minced, plated onto tissue culture dishes (Primeria; Falcon), and cultured in Dulbecco’s modified Eagle’s medium/F12 medium (1:1) with 10% fetal bovine serum and antibiotics at 37°C. After 7–8 days of growth, the cultures were overlaid with 10% Matrigel, which was removed after 24 h, and the cultures were grown for an additional 14–21 days in the same medium.

Fixed tissues were sectioned and stained for nestin (1:100 dilution; Pharmingen, San Diego, CA) and β-galactosidase (1:400; Abcam, Cambridge, U.K.) using a Vectastain ABC-Peroxidase Kit (Vector Laboratories, Burlingame, CA). Biotinylated secondary antibodies and horseradish peroxidase–conjugated anti-biotin were used to detect staining and peroxidase activity visualized by applying diaminobenzidine solution. Sections were then counterstained with hematoxylin, dehydrated, cleared, and mounted. Antisera for the detection of cre-recombinase (1:100; Novagen, Madison, WI) and von Willebrand factor (1:200; Abcam) was incubated with whole mount embryos or pancreatic sections, respectively, treated with Vector Antigen Unmasking Solution for 15 min in a microwave oven. After blocking with 5% donkey serum in PBS with 0.1% Triton X-100, the sections were incubated with the antibodies overnight at 4°C. After rinsing, fluorescein-conjugated donkey anti-rabbit secondary antibody (1:200; Molecular Probes, Eugene, OR) was applied for 90 min at room temperature. After washing with PBS, the second step of immunofluorescent staining with mouse primary antibodies was performed using the Mouse on Mouse (M.O.M.) Immunodetection Fluorescein Kit (Vector).

Explant cultures were fixed in 4% paraformaldehyde at room temperature and blocked in 10% normal goat serum, 0.3% Triton X-100. Monoclonal antibodies to nestin (1:100), rabbit anti–β-galactosidase (1:400), guinea pig anti-insulin (1:2,000; Linco Research, St. Charles, MO), rabbit anti-glucagon (1:1,000; Linco), and rabbit anti-NeuroD/BETA2 (1:200; Abcam) were added for 2 h, and binding was detected with fluorescent secondary antibodies (1:400, Alexa-fluor; Molecular Probes). Nuclei were counterstained with 4′,6-diamidino-2-phenylindole dihydrochloride in VectaShield (Vector Laboratories).

Pancreata were fixed for 15 min in 0.5% glutaraldehyde solution with PBS, 2 mmol/l MgCl₂, and 1.25 mmol/l EGTA for 4 h at 4°C. Tissues were mounted in Opti-tec and frozen, and 15 μmol/l sections were placed on glass slides and stained for 18 h in X-gal solution (PBS containing 5 mmol/l potassium ferrocyanide, 5 mmol/l potassium ferricyanide, and 1 mg/ml X-Gal). Duct cultures were fixed in 0.5% glutaraldehyde for 1 h at room temperature and stained with X-gal solution for 18 h.

To directly determine the fate of cells within the pancreas that at some time expressed nestin, we analyzed offspring of mice resulting from crossing Rosa26 mice (32) with mice that express a nestin promoter/enhancer-driven cre-recombinase (28) (nestin-cre^+/tg;R26R^loxP/+) (Fig. 1A). In embryonic day 12.5 embryos, we found that nestin and cre-recombinase were coexpressed in the developing heart and small intestine (not shown) and in the ventricular zone of the brain (Fig. 1B, arrow heads). In the developing pancreas, antisera to cre-recombinase showed primarily nuclear staining with a small amount of cytoplasmic staining, whereas antisera to nestin showed fibrous cytoplasmic staining (Fig. 1C). These findings suggest faithful expression of the cre-recombinase driven by the nestin promoter/enhancer element in these tissues. In the pancreatic mesenchyme, nucleated erythrocytes (which autofluoresce at both 514 and 543 nm) were present in vascular channels lined by nestin-positive cells (Fig. 1C, inset, arrow), suggesting that nestin-positive cells comprise part of the vascular endothelium in the developing pancreas. In the absence of primary antisera to cre-recombinase, no specific peri-epithelial staining was observed; however, the ductal epithelium continued to show fluorescence (Fig. 1D).

To determine the postnatal fate of nestin-positive cells, frozen sections of pancreas from Rosa26 mice were developed with 5-bromo-4-chloro-3-indolyl β-d-galactoside (X-Gal). The tissues showed no specific X-gal staining (Fig. 2A), confirming the absence of endogenous β-galactosidase activity. X-gal staining in pancreatic sections of nestin-cre^+/tg;R26R^loxP/+ mice was intense in small flattened cells dispersed throughout the islet (Fig. 2B and C, arrows) and in a few cells scattered between acini throughout the exocrine pancreas (Fig. 2B, C, F, and G). We examined >50 islets in animals from 3 weeks to 6 months of age from two independent nestin-cre^+/tg;R26R^loxP/+ crosses and only observed specific X-gal staining in the flattened cells of the islet and never in the larger rounded endocrine cells. The X-gal⁺ cells comprised ∼7% of the cells in the islet (63/873). β-Galactosidase expression was also detected in cells in the muscularis layer of arterioles near islets (Fig. 2C, arrows) and in the muscularis layer of larger arteries (Fig. 2D and E). The orientation of X-gal stained cells was in a circumferential pattern in the arteries (Fig. 2D), and in some sections, the X-Gal staining appeared in a spiral pattern around the vascular structure as it entered the islet (Fig. 2B and C, arrow heads). An occasional X-gal⁺ vascular-like structure was noted between acinar lobes of the exocrine pancreas (Fig. 2F), but the majority of the exocrine pancreas vasculature did not stain with X-gal, indicating that nestin-positive precursors do not contribute significantly to smaller vascular structures outside of the islet. X-gal⁺ cells could be seen in the subepithelial stroma of ducts (Fig. 2E and G, arrows), the identity of which is uncertain but may be important for generation of nestin-positive cells from the in vitro duct culture (see below). X-gal staining was absent from all other structures within the pancreas, including the acini and large and small pancreatic ducts.

Previous work has demonstrated nestin expression in cells within the islet (2), pericytes (33), and muscle precursor cells (16). Immunohistochemical staining of nestin-cre^+/tg;R26R^loxP/+ pancreas with nestin antisera showed specific staining of cells in the islet (Fig. 3A and B, arrow) and small arteries (Fig. 3A, arrow) identical to that defined by X-gal staining. To define the characteristics of the nestin-positive cells in the islet, we again exploited the autofluorescence of red blood cells (RBCs) to define the vascular channels in the islets. Confocal images of sections stained with anti-nestin antibodies showed fluorescent nestin-positive cells lining channels containing the RBCs (Fig. 3C– F). Electron microscopy showed that in the majority of sections, only one cell separates the granule containing β-cells and erythrocytes (Fig. 3G) and could represent the nestin-positive cell population. Nestin also stained occasional cells in the walls of small arteries in the pancreas (Fig. 3H, arrow), similar to that seen observed with X-gal staining. Although the endothelial cells of larger arteries of the pancreas could be easily detected with antisera to CD31 (not shown) or von Willebrand factor (Fig. 3I, arrows), these antisera only stained an occasional vessel near the base of the islet (Fig. 3I, asterisk). These data suggest that in the islet, nestin-positive cells may represent endothelial cells with unique antigenic properties.

Although these data suggest that nestin lineage cells contribute to the microvasculature of the islet and not endocrine cells, this does not exclude the possibility that nestin-positive cells cultured in vitro may give rise to alternative cell types, as has been suggested previously (3,4). To characterize the in vitro development of endocrine cells from postnatal mouse pancreata, we prepared primary cultures of islets and pancreatic ducts, per the protocol of Bonner-Weir et al. (31), from 8-week-old transgenic mice that express GFP under the control of an 8.5-kb fragment of MIP (30). Primary islet cultures gave rise to cellular outgrowths that demonstrated fluorescence only in the islets themselves (Fig. 4A and B). After 7 days of growth, examination of living cells from MIP-GFP–derived duct cultures showed no GFP fluorescence in the cells growing from the duct fragment, indicating that contaminating β-cells are not present in the initial duct cultures (Fig. 4C and D). The cells were then treated with Matrigel to induce differentiation, and during the next 7–10 days, the cultures coalesced into bands of densely packed cells (Fig. 4E– H) and eventually formed nodes of even higher cellular density (Fig. 4E– H). The cells in the nodes showed GFP fluorescence (Fig. 4B and D), indicating activation of the insulin promoter. After 14–21 days in culture, GFP⁺ cell masses in the nodes, termed “islet-like structures,” migrated into the relatively acellular areas or grew on top of the dense cell masses (Fig. 4I– J). Frequent observation of the cultures showed that multiple islet-like structures appear to bud from individual nodes. Examination of Matrigel-differentiated duct-derived cultures from wild-type mice showed no autofluorescence (not shown). These results show that, under the culture conditions used in these studies, differentiation of presumptive endocrine precursors occurs within nodes of higher cell density in a stereotypic manner.

To further define the relationship between nestin-positive cells and endocrine cells developing in vitro, we performed a series of costaining experiments. Duct explant cultures from wild-type mice gave rise to cells, of which ∼30–50% were positive for nestin (not shown). These cultures formed identical dense cell structures after Matrigel treatment and islet-like structures within and extruding from the condensed bands of cells (Fig. 5). The islet-like structures in the nodes immunostained for both insulin and glucagon (Fig. 5A and B) with the insulin-positive cells (Fig. 5B, insert, arrow head) surrounding the lower number of glucagon-positive cells (Fig. 5B, insert, arrow). Nestin expression was always observed in the condensed cell areas and formed a lattice around the developing endocrine cells (Fig. 5C and D, arrow); however, no costaining of nestin and insulin was detected. When stained with antisera to the endocrine cell transcription factor NeuroD/BETA2, differentiated duct-derived cultures showed nuclear immunoreactivity only in nodes (not shown) and in islet-like structures (Fig. 5E–H). Nestin antisera stained elongated cells that were intermeshed between the NeuroD/BETA2⁺ cells, but, again, there did not appear to be overlap in staining (Fig. 5H).

To further define the role of nestin-positive cells in the generation of insulin-producing cells in vitro, we prepared primary duct cultures of either Rosa26 or nestin-cre^+/tg;R26R^loxP/+ mice. During the initial growth phase, filamentous nestin staining could be detected in all β-galactosidase–positive cells grown from nestin-cre^+/tg;R26R^loxP/+ (Fig. 6A and B), indicating that under these culture conditions, there is accurate recombination of the β-galactosidase gene by nestin-driven cre-recombinase in vitro. Nestin-cre^+/tg;R26R^loxP/+ and Rosa26-derived duct cultures could be differentiated to form condensed cell areas, nodes, and islet-like structures in a manner identical to cultures from wild-type and MIP-GFP mice (Fig. 6 C–F). When nodes containing islet-like structures were stained for β-galactosidase and insulin, the β-galactosidase–positive cells could be seen coursing through the islet-like structure (Fig. 6C, arrow), but there was no costaining with insulin, showing that the insulin-positive cells did not arise from the nestin-positive cell population. Identical results were observed using sensitive X-gal staining of cultures. β-Galactosidase–positive cells were scattered through the condensed cell areas (Fig. 6D) and in a small number of cells in the islet-like structure (Fig. 6F) in the same pattern as seen in duct-derived cultures stained for nestin (Fig. 5) and β-galactosidase (Fig. 6C). No X-gal staining was observed in cultures from Rosa26 mice (Fig. 6E).

Our data indicate that, in vivo, the intermediate filament protein nestin is expressed only in cells destined to become endothelial cells of the islet vasculature, a subset of cells residing in the wall of arteries and scattered cells in the stroma surrounding the ductal/vascular bundles of the pancreas. This result depends on accurate recombination by the nestin-driven cre-recombinase during embryogenesis and postnatal development. Although we cannot completely rule out alternative nestin promoter/enhancer use in potential endocrine precursor cells, several lines of evidence suggest that the lineage tracing that we observe accurately reflects the in vivo situation. First, we observe transgenic cre-recombinase coexpression in nestin-positive cells in the pancreatic mesenchyme and in the central nervous system (CNS), heart, and intestine, similar to what has been described for nestin expression (1,14–18). Second, identical patterns of expression of nestin and β-galactosidase are seen in the pancreas of mature nestin-cre^+/tg;R26R^loxP/+ mice. Third, in primary pancreatic duct cultures, there is coexpression of nestin and β-galactosidase in the cells grown from these cultures. Thus, we believe that the lack of β-galactosidase expression in pancreatic cells outside of the observed nestin-positive cells, together with earlier studies (12), provide direct evidence that, in vivo, the endocrine cells of the islet arise from non–nestin-positive cells.

Nestin is expressed in the vascular endothelium of a wide variety of tissues during fetal development (23) and is expressed in a subset of cells in the adult animal, including pericytes (33) and astrocytes (34) in the brain, and in stellate cells of the pancreas (5) and liver (35). Nestin expression rises in response to damage in the CNS (23), the liver (35), and the pancreas (36) and is associated with regenerative responses in these tissues. These observations, along with the close association of mesenchymal nestin-positive cells with the pancreatic epithelium during development and our finding of an association during the generation of endocrine cells in vitro, suggest that there is a functional relationship between nestin-positive cells and the growth and differentiation of the endocrine cells in the pancreas. The vascular nature of the nestin-positive cells in the pancreatic mesenchyme and in the mature pancreatic islet is intriguing in context of a recent study demonstrating important inductive signals arising from vascular endothelial cells that are important for endocrine cell development (24). In this study, misexpression of the vascular endothelial growth factor results in localized hypervascularization and ectopic production of insulin-expressing cells. It is evident that the vascular endothelium is important in the growth and differentiation in a number of other tissues including the liver (25), fat (26), and CNS (37). In the latter study, hormonal stimulation of neuronal precursor cells in the avian CNS results in secretion of vascular growth factors. The expanded vasculature in turn secretes factors important for the growth and development of neurons. The finding that nestin is expressed exclusively in the islet microvasculature raises the possibility that these cells participate in sensing the environment of the islet and generate signals to induce the growth of the islet in response to physiological stimuli such as insulin resistance or pregnancy (38). An increase in the islet vasculature has been observed in association with conditions resulting in an expanded islet mass (39,40). In recent studies, we have observed a marked increase in the number and proliferation of nestin-positive cells in the islets of insulin-resistant mice and rats associated with islet hyperplasia but a decline in association with islet failure (M.K.T., C.L.D.-L, C.F.B., unpublished data).

In our studies, CD34 or von Willebrand factor antiserum inconsistently stains the vasculature of the islet, whereas the vasculature of the exocrine pancreas is easily detected by these antisera (data not shown). This result has been reported by others (41,42), providing further evidence for islet microvasculature specialization. Although this study suggests that nestin-positive cells line the vascular channels in the embryonic mesenchyme and in the adult islet, it is possible that nestin also marks other cells, such as pericytes (33,36), within the pancreas. The role of the nestin-positive cells in the walls of the arteries and arterioles is more difficult to understand; however, these cells too may subserve a specific role in islet physiology. The vasculature of the islet apparently responds to environmental influences in a manner that is distinct from that of the exocrine pancreas and regulates blood flow independent of that for the entire pancreas (43). An alternative explanation for these findings is that the vascular nestin-positive population may serve as a reservoir for neovascular growth of the islet during times of increased physiological demand. Further study will determine the exact nature of these cells.

These studies demonstrate a reproducible and inducible developmental process in primary duct cultures that results in the production of both glucagon- and insulin-expressing cells. The generation of these islet-like structures occurs in the context of a network of nestin-positive cells, but nestin-positive cells are not sufficient for islet formation. The finding that nestin-positive cells do not give rise to endocrine cells in vitro is in contrast to findings in earlier studies on nestin-enriched cultures (3,4). Others have suggested that these cells can develop into insulin-secreting cells along a route similar to neurons in the CNS that can express insulin (7,12). The finding of an intimate relationship between cultures enriched for nestin-positive cells and the development of endocrine cells in vitro may be due to the need for signals from the nestin-positive population to allow endocrine cells to develop from precursors present in the cultures. However, we cannot rule out the possibility that under different culture conditions, the nestin-positive population may show more inherent plasticity than under the conditions used in our system.

Finally, a recent publication has generated considerable interest in the possibility of artifacts in identifying “β-cells” grown in in vitro systems (44). We believe that the present study provides strong evidence that there is reliable endocrine differentiation occurring in our system. First, and most importantly, the identification of insulin gene activation in duct-derived cultures from MIP-GFP mice does not rely on antibodies for identification, eliminating the possibility of staining artifact or the possibility of the cells accumulating insulin from the media. This 8.5-kb promoter fragment appears to not be “leaky,” in that only β-cells in vivo show GFP expression. Second, we also observe nuclear staining of NeuroD/BETA2 only in the cellular areas that show insulin staining: in the nodes and in the islet-like structures (and this staining is nuclear and not cytoplasmic). Third, insulin staining is only observed in the same structures that show GFP fluorescence. Finally, we can find islet-like structures that stain for both insulin and glucagon, making nonspecific staining unlikely.

In summary, we show that both in vivo and in vitro, endocrine cell development does not depend on a nestin-positive lineage. However, on the basis of these and earlier studies, we propose that the nestin-positive cells are specialized vascular cells that are important for islet formation in vivo and in vitro by providing signals to true endocrine precursors, similar to what has been proposed to occur in vivo (24) for the pancreas and other tissues (25,26,37,45). Understanding the regulation of growth and development of the nestin-positive population will provide insights into the physiological regulation of islet growth in vivo and provide a tool for use in the development of insulin-producing cells in vitro.

This work was supported by a grant from the American Diabetes Association (C.F.B.) and U.S. Public Health Service Grants DK-20595 and DK-61245 (M.H.) and KO8HD40288 (D.M.M.).

The authors thank Lanjing Zhang for technical assistance.