Evaluation of β-Cell Replication in Mice Transgenic for Hepatocyte Growth Factor and Placental Lactogen: Comprehensive Characterization of the G₁/S Regulatory Proteins Reveals Unique Involvement of p21^cip

Single-transgenic RIP-HGF and RIP-mPL1 transgenic mice that express HGF and mPL-1 under the control of the rat insulin II promoter (13) have been described in detail previously (3,4,26,27). RIP-HGF and RIP-PL mice were crossed to generate normal mice, mice singly transgenic for placental lactogen or HGF, and double-transgenic mice overexpressing HGF and placental lactogen in the pancreatic β-cell. All studies were performed on mice between 8 and 12 weeks of age, and mice were ∼50% male and ∼50% female. No significant differences were observed between males or females in these studies. The original backgrounds of the RIP-mPL-1 and RIP-HGF mice were CD-1, and the crosses thus remain on a CD-1 background. p21^cip-null mice (29) and controls on a C57B6 background were purchased from The Jackson Laboratories (Bar Harbor, ME). All procedures were approved by, and in compliance with, The University of Pittsburgh Institutional Animal Care and Use Committee.

These procedures were performed as described in detail previously (3–5,26,27) and in the figure legends. In vitro β-cell replication rates were assessed using tritiated thymidine incorporation as described in detail previously (5). Briefly, p21−/− and normal mouse islets were freshly isolated and placed in RPMI medium containing 5 mmol/l glucose and 10% fetal bovine serum in 24-well plates, with 50 islets per well. HGF or prolactin was added 24 h later in the concentrations shown in the legend of Fig. 8 (for HGF, for both p21+/+ and p21−/−, 11 mmol/l glucose was used), and the islets were incubated for an additional 48 h. Tritiated thymidine (0.5 μCi/well) was added 24 h later in RPMI medium containing the growth factors but without fetal bovine serum, and the islets were allowed to incubate for an additional 24 h (for a total of 48 h with HGF or prolactin and 24 h with tritiated thymidine). The remainder of the procedure was as described previously (5).

The adenoviral cyclin D₁ and cdk-4 studies were performed using adenoviruses, conditions, and reagents described in detail previously (5). Briefly, murine and human islets were transduced for 1 h with 500 multiplicity of infection of adenoviruses expressing cyclin D₁ and cdk-4 under the control of the cytomegalovirus promoter.

Mouse islets were isolated as described previously (3–5,26,27). Human islets were provided by the University of Pittsburgh Human Islet Isolation Facility as described in detail previously (5). Semiquantitative PCR was performed as described in detail previously (3–5,26,27) using the primer pairs shown in Supplemental Table 1, which is detailed in the online appendix (available at http://diabetes.diabetesjournals.org) (see table legend for full details).

Islet extracts were prepared using nuclear extraction buffer as described previously (5). Antisera and dilutions used are shown in Supplemental Table 2, which is detailed in the online appendix (available at http://diabetes.diabetesjournals.org). Each immunoblot in Figs. 4–7 was performed a minimum of three times on a minimum of three different islet preparations, and each was accompanied by an appropriate housekeeping protein control, tubulin, or actin. In cases in which minor changes appeared possible, the immunoblots were repeated for a total of five to eight times, and quantitative densitometry of digitalized blots was performed and analyzed statistically. Thus, the immunoblots and text descriptions are representative of multiple experiments. Because of space limitations, only representative immunoblots and action/tubulin controls are shown.

For all studies, Student’s two-tailed unpaired t-test was performed using Microsoft Excel. P values <0.05 were considered significant. Bars in the figures indicate SE.

Our original goal was to determine, by crossing RIP-PL with RIP-HGF mice, whether overexpression of both mitogenic growth factors in combination would lead to a further increase in islet mass and function. Supplemental Fig. 1, which is detailed in the online appendix, confirms, using RT-PCR, that expression of both transgenes occurs in islets of mice singly and doubly transgenic for placental lactogen and for HGF. Supplemental Fig. 2 of the online appendix shows fasting and postprandial glucose and insulin concentrations in the four mouse genotypes. The two single-transgenic lines, RIP-PL and RIP-HGF, reproduce their original phenotypes (3,4,26,27) in this second independent study. However, overexpression of placental lactogen and HGF in the double transgenics did not lead to an exaggeration either of fasting or postprandial hypoglycemia. Intraperitoneal glucose tolerance tests (Supplemental Fig. 3 of the online appendix) were also reflective of events described earlier (3,4,26,27): The RIP-HGF mice displayed mildly improved glucose tolerance, and the RIP-PL mice did not. Surprisingly, the double transgenics were significantly impaired compared with the normals and their single-transgenic littermates.

The RIP-HGF and RIP-PL mice displayed an obvious increase in β-cell mass (Fig. 1A), paralleling that described previously (3,4,26,27). However, the combined overexpression of the two growth factors had no additional effect on islet number or mass. This was confirmed using quantitative islet histomorphometry (Fig. 1B), which showed that islet cell volume was increased approximately threefold in both single transgenics, as reported previously (3,4,26,27). However, there was no further increase in islet mass in their double-transgenic littermates. Similarly, islet number was increased in the RIP-HGF mice, but the combined overexpression of placental lactogen and HGF did not further increase islet number. Islet morphology was normal, although there appeared to be a redistribution of α-cells and somatostatin cells from the periphery into more central portions of the islet in the RIP-PL and double-transgenic mice (Supplemental Fig. 4, online appendix).

Prior studies using RIP-HGF and RIP-mPL1 mice have demonstrated that HGF and mPL1 increase rates of β-cell replication in vivo by a factor of ∼2 (3,4,26,27). Figure 2 confirms these findings in this second study. However, mice doubly transgenic for placental lactogen and HGF did not display a further increase in β-cell replication rates.

The failure of combined overexpression of two potent β-cell mitogenic growth factors to further increase β-cell replication or mass suggested that a regulatory system(s) may exist that represses cell cycle activity in response to excessive mitogenic stimuli. We therefore explored in a comprehensive fashion all of the key cell cycle regulatory molecules known to be involved in the G₁/S transition (Fig. 3). Because many, if not most, cell cycle regulatory proteins are known to be regulated at the posttranslational level, for example by ubiquitination and proteosamal degradation (22–24), we studied these molecules at the protein rather than RNA level.

The retinoblastoma protein, pRb, is the principal checkpoint protein of the G₁/S transition. Figure 4A demonstrates for the first time that pRb is present in murine islets and can be phosphorylated (inactivated) to phospho-pRb in murine islets by adenoviral overexpression of cyclin D₁ and cdk-4 (Fig. 4B). Surprisingly, pRb is not apparently phosphorylated by either HGF or placental lactogen alone or in combination. pRb is one member of a family of tumor suppressor proteins, the “pocket proteins,” that includes pRb, p107, and p130. Figure 4A also demonstrates for the first time in any species that p107 and p130 are present in the islet. None of the pocket protein family members appear to be regulated by HGF, placental lactogen, or the combination.

Seven E2F transcription factors are believed to regulate the myriad genes required for G₁/S transition (Fig. 3). Figure 4C demonstrates that mRNAs encoding the seven E2F proteins are expressed in the pancreatic islet. E2F1, a G₁/S transition agonist, and E2F4 and E2F6, G₁/S transition transcriptional repressors, are expressed most highly, whereas E2Fs 2, 3, 5, and 7 are expressed at lower levels. We studied E2Fs 1, 2, 4, and 6 further, because they are expressed at the highest levels (E2Fs 1, 4, and 6) or have been implicated in β-cell function (E2Fs 1 and 2) (18,19). Figure 4D demonstrates that each of these four E2F proteins is present in the murine islet.

Figure 5A shows that all of the three members of the cyclin D family are present in the islets of CD-1 mice. Of particular note is cyclin D₃, which some authors have suggested is not abundant in the islet (16,17,30). This observation was confirmed by RT-PCR using two different sets of primers (not shown). Figure 5B demonstrates that cdk-4 is present in murine islets, whereas cdk-6 is completely absent. Figure 5C demonstrates that cyclin A, cyclin E, and cdk-2 are present in murine islet. Figure 5D demonstrates that all four members of the INK family of kinase inhibitors are present in the murine islet. Figure 5E demonstrates that p27 and p57, two members of the Cip-Kip-Waf family of kinase inhibitors, are also present in the murine islet. Figure 5F demonstrates the presence of c-Myc (assessed by RT-PCR in the absence of reliable antisera), MDM2, and Egr-1. p53 appears to be reproducibly increased in RIP-HGF islets. Importantly, none of the other proteins shown in Fig. 4 or 5 is differentially regulated by HGF, placental lactogen, or the combination.

Figure 6A reveals that p21^cip is markedly, consistently, and very specifically upregulated in RIP-HGF and RIP-PL islets. Interestingly, although p21^cip is upregulated in the RIP-PL and RIP-HGF islets, it is not further upregulated in the doubly transgenic HGF-PL islets. As shown in Fig. 6B, RIP-PTHrP islets, which also display increased β-cell replication (26,28), also demonstrate a marked increase in p21^cip expression. Furthermore, p21^cip is upregulated in mPL1 and double-transgenic RIP-PTHrP and RIP-mPL1 islets (28). Figure 6C demonstrates that p21^cip expression is also increased in response to cell cycle activation by adenoviral overexpression of cyclin D₁ and cdk-4 (5). Figure 6D and E demonstrates that exogenous prolactin, a surrogate for placental lactogen, directly upregulates p21 protein expression in normal isolated murine islets and in the rat insulinoma cell line, Ins-1.

All of the studies described thus far apply to murine islets. We wanted to determine whether these observations might apply to human islets. Figure 7A demonstrates that human islets express p21^cip protein as well and that it is dramatically increased in response to cell cycle activation by either cyclin D₁ alone or in combination with cdk-4 (5). p21^cip is present primarily in the cytoplasm of nonstimulated human β-cells but traffics into the nuclear compartment on mitogenic stimulation (Fig. 7B).

p21^cip-null mice are dwarf and tumor prone and develop renal failure by 6 months of age because of glomerulonephritis (29,31), complicating the evaluation of their phenotype in vivo. We therefore turned to islets isolated from these mice (29), stimulated in vitro with HGF or placental lactogen. As can be seen in Fig. 8, the addition of exogenous HGF or prolactin, a surrogate for placental lactogen, to normal murine islets had no effect on [³H]thymidine incorporation. In contrast, when administered to p21^cip−/− islets, ³H incorporation increased substantially. These observations suggest that p21^cip inhibits β-cell replication in response to β-cell mitogens.

These studies make six significant and novel observations: 1) Combined overexpression in transgenic mice of two well-characterized and potent in vivo β-cell mitogens, placental lactogen and HGF, does not further enhance β-cell replication rates nor further enhance islet mass or function. 2) We provide the first comprehensive catalog defining the cell cycle regulatory proteins that are likely to govern the G₁/S transition in the murine islet. This “catalog” or “G₁/S β-cell cycle proteome” has been characterized under basal and mitogenic conditions. 3) p21^cip is markedly and uniquely subject to regulation by three different β-cell growth factors—HGF, placental lactogen, and PTHrP—as well as by a key G₁/S transition regulatory kinase complex, cyclin D₁/cdk-4. 4) p21^cip traffics to the nucleus of β-cells under mitogenic stimulation. 5) p21^cip serves to inhibit growth factor–driven replication in murine islets. 6) These observations extend to human islets, where p21^cip is present, is upregulated, and translocates to the nucleus of human β-cells in association with activation of cell cycle progression.

The metabolic and histological findings in the RIP-PL and RIP-HGF mice confirm in an independent study that these two factors are potent and effective β-cell mitogens, increasing β-cell replication and aggregate β-cell mass in vivo, while retaining the differentiated functions of the β-cell. In contrast, the results of combined or simultaneous overexpression of placental lactogen and HGF were surprising: We had anticipated that because HGF and placental lactogen act via different signaling pathways—JAK2/STAT5 for placental lactogen (10,32) and phosphatidylinositol 3-kinase and mitogen-activated protein kinase for HGF (3,9)—they would have an additive or synergistic effect on islet mass. This lack of synergy is present in two additional double-transgenic systems: the cross of the RIP-HGF mouse with the RIP-PTHrP mouse (33) and the cross of the RIP-mPL-1 and RIP-PTHrP mouse (28). These findings suggest that intrinsic gatekeeper or rate-limiting pathway(s) may exist that prevent β-cell replication rates from exceeding certain predetermined limits.

We focused on the molecular control of the cell cycle, in part, because of our demonstration (3–5), as well as that of others (6–10), that β-cell can replicate and, in part, because of the essential absence of a “cell cycle roadmap” relating to the β-cell. Despite the extensive characterization of cell cycle regulatory mechanisms in fibroblasts, cancer cell lines, and developmental systems, relatively little is known about the molecular and cellular control of replication machinery in the β-cell. We selected the G₁/S transition point because, in general, growth factors activate cell cycle progression at G₁/S (22–24) and also because virtually all mouse genetic models had suggested that this was likely to be an important pathway for β-cell replication (11–21,25).

Under basal conditions, many, but not all, of the key G₁/S regulatory molecules are present in the murine pancreatic islet. For example, we show for the first time that pRb, the prototypical G₁/S gatekeeper in most systems, is present in the murine islet, as are its two pocket protein homologs, p107 and p130 (34,35). However, we see no evidence of pRb phosphorylation in these islets. E2F proteins mediate the transcriptional effects of pRb on downstream cell replication machinery. In the current study, we demonstrate for the first time that most E2F family members are present within the murine pancreatic islet and that, among these, E2Fs1 and 4 are most abundant.

All three members of the cyclin D family are present in the murine islet. This is significant because Martin et al. (31) and Kushner et al. (17) have suggested that cyclin D₃ is not abundant in the islet, and Georgia et al. (16) and Kushner et al. (17) indicate that null mutations for cyclin D₂ result in islet hypoplasia. All three of these prior studies were performed using C57B6 mice, whereas ours were performed using CD-1 mice. We recognize that the best current commercial antisera may cross-react among cyclin D isoforms, and therefore we confirmed that cyclin D₃ mRNA is present in the CD-1 murine islet using RT-PCR with two different sets of primers (not shown). Thus, β-cell cycle control may prove to be species and strain specific. As with the cyclin D₃ immunoblot data shown in Fig. 5A, cyclin D₃ mRNA levels were not different in the islets of normal littermates compared with the RIP-HGF, RIP-mPL-1, or double-transgenic mice (not shown). We cannot exclude the possibility that a portion of the cyclin D₃ mRNA observed may be derived from contaminating exocrine or non–β-islet cells.

Null mutations for cdk-4 result in marked islet hypoplasia, hypofunction, and diabetic ketoacidosis (20,21). The studies reported here make it clear why: Surprisingly, murine islets specifically lack cdk-6, a functional analog of cdk-4. Thus, loss of cdk-4 in cells that already lack cdk-6 might be predicted to lead to islet failure.

Murine islets were shown to contain pRb, phospho-pRb, p107, p130, cyclins D₁,₂,₃, cyclin A, cyclin E, cdk-2, cdk-4, p15^Ink, p16^Ink, p18^Ink, p19^Ink, p27^kip1, p57^cip2, MDM2, Egr-1, and c-Myc. In contrast to our expectations, overexpression of HGF, placental lactogen, or the combination had essentially no effect on any of these cell cycle control molecules. This may reflect the fact that these molecules are not subject to regulation by growth factors in the islet, that their associations with one another and trafficking are more important than their amounts, or that increases or decreases in these proteins in a small subpopulation of proliferating β-cells are “diluted out” by the larger aggregate mass of G₀/G₁-arrested mass of β-cells. In contrast to all of these apparently “stable” cell cycle regulatory proteins, HGF, but not placental lactogen, reproducibly stimulated p53. Understanding the significance of this observation will require further study.

In contrast to the lack of change in the 20 molecules itemized above, p21^cip was dramatically and specifically upregulated in response to HGF, placental lactogen, PTHrP, and cyclin D₁ alone or in combination with cdk-4. Furthermore, p21^cip was upregulated by cell cycle activation not only in murine islets, but also in human islets. Importantly, this upregulation was observed to occur in the β-cell and to be associated with translocation from the cytoplasmic to the nuclear compartment. p21^cip has previously been shown to be present in rat and human β-cells and in rat β-cell lines (36–38). However, it never has been implicated in β-cell cycle control. Kaneto et al. (36) have shown that p21^cip is upregulated in rat islets in response to oxidative stress induced by adding H₂O₂ and is also upregulated in isolated islets of obese Zucker diabetic rats. Forced overexpression of p21^cip led to a reduction in insulin content in rat islets, but no cell cycle studies were reported (36). Zhang et al. (37) have reported that induction of apoptosis in the insulinoma cell line RIN m5F leads to upregulation of p21^cip. Maitra et al. (38) reported that p21^cip mRNA is reduced in human insulinomas and that p21^cip is present in normal human β-cell nuclei but is downregulated in nuclei of human pancreatic endocrine tumors. Although no information regarding p21^cip involvement in β-cell replication was provided in any of these studies, they collectively suggest that p21^cip may be upregulated by β-cell injury or stress and that loss of p21^cip may be associated with islet tumorigenesis.

In some tissues and organs, p21^cip is an inhibitor of cell cycle progression, acting as a classical INK or KIP/CIP family member (22–24,39,40), whereas in other tissues or cells, it can serve as an agonist for cell cycle progression (31,41,42). Therefore, one might speculate that the increase in p21^cip expression in response to mitogenic stimuli serves as the rate-limiting factor, or the molecular “brakes,” on β-cell cycle progression that we had hoped to identify. On the other hand, one might equally reasonably speculate that p21^cip is a mediator of cell cycle progression, or a molecular “accelerator,” that augments β-cell cycle progression. The observation that p21^cip-null islets are more responsive to mitogenic stimulation by HGF and prolactin, used here as a surrogate for placental lactogen, supports the first possibility. We hypothesize, therefore, that the marked upregulation of p21^cip in the single and double-transgenic islets may prevent excessive β-cell replication in the single transgenics and that the exuberant β-cell replication and increase in islet mass we had anticipated might occur in the double transgenics.

The ability of p21^cip-null islets to proliferate basally and in response to growth factors indicates that one standard role of p21^cip—assembly of cdk-cyclin D and E complexes in the cytosol and use of the nuclear localization signal in p21^cip to transport these cdk-cyclin complexes into the nucleus where they can phosphorylate pRb—is not essential or is complemented by another protein that can serve the same function. One obvious candidate is p27^kip, which is believed to subserve the same function (22–25). p27^kip has recently been reported by Uchida et al. (25) to be critical to the reduction of β-cell mass and function in IRS-2−/− and leptin receptor–null mice. Interestingly, although p27^kip was upregulated in IRS-2−/− islets and trafficked to the nuclear compartment, there was no change in p21^cip in these “metabolically” abnormal islets. These events are the mirror image of those observed here, in which mitogenically challenged islets display upregulation and nuclear translocation of p21^cip, with no change in p27^kip. It seems possible that p27^kip may respond to metabolic signals such as glucose/lipotoxicity, whereas p21^cip may respond to mitogenic signals such as growth factors.

These studies have a number of limitations. First, as alluded to above, most cell cycle proteins shuttle between the nuclear and cytoplasmic compartments and within these compartments, shuttle among complexes with one another, yet we have provided no information on this trafficking of any molecule except for p21^cip. Further immunocytochemical and coimmunoprecipitation studies will be required to define the subcellular distributions and key interactions of the proteins shown in Fig. 3. Second, kinase activity and phosphorylation status are often more relevant to cell cycle activity than the absolute level of a particular cell cycle protein, yet the current studies do not provide information on kinase activity of the cdks. Third, growth factor–driven cell cycle protein changes likely occur exclusively or predominantly in a subset of cells in G₁ and S phases, but we have examined only whole islets. Thus, important single-cell changes in many of these proteins could be masked or diluted by a lack of change in the larger component of noncycling β-cells. Fourth, islets are complex mixtures of α-, β-, δ-, PP, ductal, endothelial, fibroblast, and other cells, and studies using immunoblots or whole islets may not reflect events in β-cells. Fifth, it is likely that other members of the G₁/S cell cycle proteome are present in the β-cell (for example, the DPs, the Ids), and additional yet unidentified cell cycle regulatory molecules are certain to be identified in the future. Sixth, these studies may have identified p21^cip as the molecular “β-cell brake” in HGF- and placental lactogen–stimulated cells, but they do not elucidate the cell cycle pathways that lead to increased β-cell proliferation. Thus, this is a preliminary “G₁/S proteome” database, and for all of the above reasons, it will be critical to build further upon this database, defining and refining the cellular and subcellular localization of key proteins, evaluating their phosphorylation status and kinase activity, and defining their intermolecular interactions and complexes and their functions. Finally, these studies have made no attempt to define the signaling pathways downstream of HGF and placental lactogen, such as phosphatidylinositol 3-kinase, mitogen-activated protein kinase, protein kinase C, JAK2-STAT5, etc., that lead to increases in p21^cip. Nor have we attempted here to define the cellular mechanisms responsible for p21^cip upregulation: transcription, translation, protein stability, etc. These are important goals of ongoing studies.

In summary, these studies confirm the potent islet growth factor activity of HGF and placental lactogen and suggest that growth factor–driven β-cell replication is likely intrinsically limited by molecular brakes. The studies define a preliminary β-cell G₁/S proteome and reveal a remarkable stability in cell cycle control molecules in response to HGF and placental lactogen. p21^cip is a novel, reproducible, readily measurable, highly selective, in vivo marker of cell cycle activation in the murine and human β-cell, and inhibits islet replication in response to growth factors. p21^cip likely represents the “molecular brakes” that prevent more exuberant proliferative responses to HGF and placental lactogen overexpression. p21^cip may be a therapeutic target for therapeutic manipulation of β-cell replication.

We thank Drs. Adolfo Garcia-Ocaña, Nathalie Fiaschi-Taesch, and Rupangi Vasavada for their repeated and thoughtful input during the evolution of these studies.