Early and Late G1/S Cyclins and Cdks Act Complementarily to Enhance Authentic Human β-Cell Proliferation and Expansion

Complete or partial loss of functional β-cell mass is a major feature of type 1 and type 2 diabetes (1). Replacement or regeneration of lost β-cells is, therefore, a key goal of diabetes research. Thus, manipulating the regulation of the cell cycle in human β-cells holds great therapeutic potential. Expanding adult human β-cells is challenging since their basal proliferation level in vivo and in vitro is extremely low and they are resistant to the induction of replication (2–8). Recently, we made the unexpected observation that many key G1/S cell cycle activators are excluded from the nucleus in adult human β-cells, presumably contributing to their refractoriness to replication (7,8). Observations in neonatal human β-cells show that human β-cells replicate transiently during the first few years of life (9–13). The labeling index remains low compared with other tissues, however, in the range of 3%.

We, and others, have shown that it possible to directly manipulate the cell cycle and induce some cell cycle entry in adult human β-cells. For instance, the overexpression of cell cycle activators, such as cyclin-dependent kinase (cdk) 6 and cyclin D3 (5,14), or downregulation of inhibitors, such as p57 (15), lead to a substantial cell cycle entry in adult human β-cells. However, whether these replication levels are therapeutically relevant and whether this cell cycle entry actually leads to a true increase in β-cell number remains unknown.

Transition from the G1 to the S phase of the cell cycle requires the inactivation of the retinoblastoma protein (pRb) family (p107, p130) of cell cycle inhibitors at the key G1/S restriction point. pRb is inactivated in the nucleus by sequential phosphorylation of up to 16 serines and threonines orchestrated by multiple cdks and their cyclin partners (16,17). The “early” cyclin/cdk complexes, including one of the three d-cyclins bound to either cdk4 or cdk6, may mediate the initial pRb phosphorylation. Inactivation of pRb also may be performed by the “late” cyclins and cdks (complexes of cyclin A or E with either cdk1 or cdk2) (18). The regulation of cdk activity is multifactorial and can be controlled at the level of nuclear translocation, protein stability/abundance, cyclin binding, phosphorylation status, and activity of cdk inhibitors such as the Cip/Kip family (19,20). The relative importance of these in β-cells is unknown.

In mouse and man, the early G1/S cdk complexes play a crucial role in β-cell proliferation. The loss of either cdk4 or cyclin D2 in mice leads to a profound loss of β-cell mass and proliferation and severe diabetes (21,22). Growth factors and nutrients have been shown to induce cell cycle entry by activating early G1/S cyclins and cdks. For example, glucose stimulates mouse β-cell replication in part via an induction of cyclin D2 (23–25). c-Myc induces rodent β-cell replication through the induction of d-cyclins, cdk4 and cdk6 (26). We have shown that the overexpression of cdk6 or d-cyclins, individually or in combination, leads to a marked and sustained stimulation of cell cycle entry in human β-cells (5).

Recent studies (27) also underscore the importance of the late G1/S cyclin/cdk complexes in mediating β-cell proliferation as well, as follows: cyclin A has been shown to be necessary for exendin-4–induced proliferation in murine β-cells. The growth factor parathyroid hormone-related protein increases human β-cell proliferation via upregulation of cyclin E and cdk2 (28). Similarly, the induction of rat and human β-cell proliferation by the transcription factor Nkx6.1 requires the induction of cyclin E (29). Finally, the overexpression of cyclin E in combination with cdk2 induces a significant and marked stimulation of human β-cell proliferation (28). These data suggest that targeting the activation of either the early or the late G1/S cdk complexes is an efficient approach to inducing human β-cell proliferation.

While several studies have focused on either the early or the late G1/S cyclins and cdks, no prior study has examined the possibility that the activation of early plus late G1/S cyclin/cdk complexes may be additive or synergistic for inducing proliferation in adult human β-cells. We hypothesized that this may be possible. We further hypothesized that such a synergistic or additive effect may induce a therapeutically relevant level of human β-cell replication and lead to an authentic expansion of adult human β-cells.

Human islets were purchased from the National Institutes of Health– and JDRF-supported Integrated Islet Distribution Program (https://iidp.coh.org). Sixty-six different cadaveric preparations from donors without diabetes were used for these studies. The mean ± SEM age of the donors was 45.5 ± 1.6 years, the mean ± SEM BMI was 32.2 ± 0.8 kg/m², and the mean ± SEM purity was 85.1% ± 0.1. Thirty of the 66 preparations were from female donors, 32 were from male donors, and 2 had no specified donor sex. Only 2 of the 66 donors were under the age of 20 years, at 16 and 14 years. Upon arrival, islets were incubated in RPMI medium (Life Technologies) containing 5.5 mmol/L glucose, 1% penicillin, and streptomycin with 10% FBS until they were used for experiments. The use of human cadaver islets was approved in advance by the University of Pittsburgh School of Medicine and Icahn School of Medicine at Mount Sinai Institutional Review Boards.

Rat islets were isolated from 2-day-old, 2-month-old, 9- to 10-month-old, or 18- to 20-month-old Sprague Dawley rats (Charles River), as previously described (4). Islets were incubated in RPMI medium (Life Technologies) containing 5.5 mmol/L glucose, 1% penicillin, and streptomycin with 10% FBS until they were used for experiments. Rat islet isolations were performed with the approval of, and in accordance with, guidelines established by the Icahn School of Medicine at Mount Sinai and the University of Pittsburgh Institutional Animal Care and Use Committees.

Adenoviruses (Ad.), all under control of the cytomegalovirus promoter, were prepared using human cDNAs encoding human cdk6, human cdk1, human cdk2, and human cyclins D3, as described previously (4,5,7,14,30,31). Ad.human cyclin A2 was purchased from Vector Biolabs (Philadelphia, PA) and Ad.human cyclin E1 was purchased from Applied Biological Materials Inc. (Richmond, BC, Canada). Dispersed islets on coverslips were transduced with either Ad.LacZ or relevant control (CTL) adenoviruses for 2 h, cultured for 72 h as described in the figures, and immunolabeled as described below.

Glucose-stimulated insulin secretion was performed on transduced human islets three days after transduction as described previously (5). Five different human islet preparations were used. Insulin secretion was determined using a human insulin ELISA kit (ALPCO, Salem, NH).

Human islets or rat islets (400 islet equivalents) were washed twice in PBS and dispersed as previously described (5,7,14). Single cells were then plated on coverslips; transduced with Ad.cdk6 plus Ad.cyclinD3, Ad.cdk1 plus Ad.cyclin E, or Ad.cdk6 plus Ad.cyclinD3 plus Ad.cdk1 plus Ad.cyclin E (total of 400 or 1,000 multiplicities of infection [MOIs]); cultured for 72 h; then fixed in fresh 4% paraformaldehyde for 15 min at 25°C, as previously described (5,7,14). CTL cells were transduced with Ad.LacZ. Fixed cells were washed with PBS and incubated in blocking buffer (1.0% BSA, 0.5% Triton, and 5% normal goat serum in PBS) for 1 h at 25°C. Primary antisera were exposed to the fixed cells on coverslips overnight at 4°C in blocking buffer, and secondary antisera were exposed for 2 h at 25°C in secondary buffer (1% BSA and 0.5% Triton in PBS). Primary antisera and secondary antisera are described in Supplementary Table 1.

Islets were dispersed to single cells; transduced with Ad.LacZ, Ad.cdk6 plus Ad.cyclinD3, Ad.cdk1 plus Ad.cyclin E, or Ad.cdk6 plus Ad.cyclinD3 plus Ad.cdk1 plus Ad.cyclin E (total of 400 MOIs); and fixed as previously described above. After washing with PBS, the cells were immunolabeled for insulin as previously described (5,7,14). After washing with PBS, cells were incubated overnight at 4°C with a combination of a mouse antibody against the full-length pRb (catalog #9309; Cell Signaling Technology) and a rabbit antibody against either phospho-S780/S795-Rb (catalog #9307S and #9301P; Cell Signaling Technology), or against phospho-S811-Rb (catalog #9308S; Cell Signaling Technology). All the antibodies were diluted (1:100) in the antibody diluent provided by the manufacturer (Duolink; Sigma-Aldrich, St. Louis, MO). Negative CTLs omitted the primary antibody against pRb. After washing, the cells were incubated for 1 h at 37°C with the Duolink Proximity Ligation Assays Anti-Rabbit MINUS and Anti-Mouse PLUS proximity probes, both provided by the manufacturer. The proximity ligation and the amplification were performed using the Duolink In Situ Detection Reagent Kit according to the manufacturer protocol. Fluorescence was detected using a Leica SP5 confocal microscope. All of the experiments were repeated at least three times.

Islet extracts were resolved using 10% or 12% SDS-PAGE, and transferred to Immobilon-P membranes (Millipore, Bedford, MA). Primary antisera and secondary antisera are described in Supplementary Table 1. Each experiment shown is representative of three to six human islet preparations.

Human islets were dispersed and transduced with Ad.LacZ, Ad.cdk6 plus cyclin D3, Ad.cdk1 plus cyclin E, or Ad.cdk6 plus Ad.cyclinD3 plus Ad.cdk1 plus Ad.cyclin E (total of 400 MOIs) as described previously. Ten thousand transduced islets cells from each group were plated in a high-content imaging, glass bottom, 96-well plate (Corning), coated with poly-d-lysine (Sigma-Aldrich). Five thousand, 10,000, 15,000, and 20,000 untransduced dispersed human islet cells were plated as CTLs. One day later, 50 μL Matrigel (Corning, Corning, NY) was added to each well. Cells were fixed in 4% paraformaldehyde for 10 min at room temperature 5 days after plating. Cells were then immunolabeled for insulin and DAPI as previously described (5,7,14). Photomicrographs were taken to cover the entire surface of each well using an Olympus IX70 wide-field epifluorescence system. Images were then stitched together to create a single picture using MetaMorph. Each stitched picture was then analyzed using ImageJ to measure the total β-cell area per well. To measure the β-cell size, cells were immunolabeled with an antisera for E-cadherin and for insulin. The size of each β-cell was measured using ImageJ.

Statistical analysis for Figs. 1B and G, 4B, 5B, 6B, and 8B was performed using one-way ANOVA with Bonferroni post hoc correction. A Student paired, two-tailed t test was used in Figs. 1E; 5D, F, and H; 6D, F, and H; and 8D. All values are expressed as the mean ± SEM. P values <0.05 were considered to be significant.

Previous studies (1,5,7,8,14,26,32–41) have shown that rodent β-cells are more susceptible to replication than human adult β-cells. Here, we determined whether this species difference also existed when cell cycle molecules, such as cdk6 and cyclin D3, were directly overexpressed. As shown in Fig. 1A and B, basal levels of replication were extremely low in human β-cells. In dramatic contrast, basal levels of β-cell replication were ∼40% in 2-day-old rat β-cells but dropped below 1% in 2-month-old, 9- to 10-month-old (Fig. 1A and B, 10 months), and 18- to 20-month-old (Fig. 1A and B, 20 months) rat β-cells. When cdk6 and cyclin D3 (Fig. 1A and B, C6+D3) were overexpressed, 14.1 ± 0.9% of human β-cells entered the cell cycle, as assessed by Ki67 immunolabeling. In contrast, 49.3 ± 3.0% of 2-month-old rat β-cells entered the cell cycle under the same conditions. This was age independent since β-cells from 9- to 10-month-old and 18- to 20-month-old rats responded with a similar cell cycle entry level (50.9 ± 1.4% and 38.0 ± 6.7%, respectively). These high levels of cell cycle entry measured in rat β-cells in response to the overexpression of C6+D3 were similar to the levels of cell cycle entry measured in a rat insulinoma cell line, INS1, cultured in 5.5 mmol/L glucose (Fig. 1C and D). As shown in Fig. 1E, the efficiency of adenoviral transduction in rat and human islet cells appeared comparable, as measured by the adenoviral β-galactosidase labeling. Finally, since nuclear translocation of cdk6 and cyclin D3 is associated with the level of replication (5,7,8,14), we measured the percentage of the nuclear presence of cdk6 and cyclin D3 in human and rat β-cells after transduction. As shown in Fig. 1F and G, the percentage of nuclear cyclin D3 was similar in rat β-cells compared with human β-cells when cdk6 and cyclin D3 were overexpressed. The percentage of nuclear cdk6 was lower in rat β-cells.

Cell cycle entry begins with the inactivation of pRb and other pocket proteins either by the early G1/S cyclins and cdks, cdk6 and d-cyclins, or by the late cyclins and cdks, cdk1/2-cyclin A/E (3,16–19,42). To complete the progression through G1 and enter S, late cyclins and cdks must translocate to the nucleus after their synthesis in the cytoplasm (19,20). Thus, we queried whether the difference in cell cycle entry between rat and human β-cells by cdk6 and cyclin D3 could be explained by a difference in late cyclin and cdks nuclear translocation and activation. We, therefore, examined the trafficking of the late G1/S cyclins and cdks cyclin A and E, cdk1, and cdk2 under basal conditions, and, when cdk6 and cyclin D3 are overexpressed in adult human and rat β-cells. As shown in Fig. 2, cyclin A, cyclin E, cdk1, and cdk2 were expressed in rat and human β-cells under basal conditions and were predominantly cytoplasmic. All four late G1/S cyclins and cdks remained cytoplasmic in human β-cells when replication was induced by overexpression of cdk6 and cyclin D3 (Fig. 2). In contrast, in rat β-cells, cdk1 and cdk2 translocated to the nucleus when cdk6 and cyclin D3 were overexpressed. Moreover, in rapidly replicating rat INS1 cells, cdk1 and cdk2 were both cytoplasmic and nuclear. We also explored nuclear translocation of two cell cycle inhibitors, p16^Ink4A and p27^Kip1. As shown in Supplementary Fig. 1 and as previously described (14), p16^Ink4A and p27^Kip1 were predominantly cytoplasmic under basal conditions in human β-cells (6.74 ± 1.4% and 10.37 ± 2.83% in the nucleus, respectively) but were detected more often in the nucleus of β-cells transduced by cdk6 and cyclin D3 (13.75 ± 0.62% and 22.9 ± 6.15%, respectively). This nuclear translocation of p16^Ink4A and p27^Kip1 was more pronounced in rat β-cells in response to cdk6 and cyclin D3 (27.82 ± 3.17% and 61.08 ± 8.31%, respectively).

Next, we interrogated whether the apparent lack of nuclear translocation of the cdk1/2 and cyclin A/E reflected an intrinsic inability of these late cyclins and cdks to gain nuclear access in human β-cells. For that purpose, we determined whether the overexpression of cdk1/2 and cyclin A/E, individually or in combination, could lead to their nuclear translocation. As indicated in Fig. 3A and Supplementary Fig. 2, transduction with adenoviruses encoding for cdk1, cdk2, cyclin A, or cyclin E led to an increased level of expression of each of the late cell cycle molecules, in a dose-dependent manner. When overexpressed alone, cdk1 and cdk2 remained cytoplasmic (Fig. 3B). However, when combined with either cyclin A or E, cdk1 or cdk2 appeared in the nuclei of human β-cells (Fig. 3B). In contrast, the overexpression of cyclin A or cyclin E, alone or in combination with cdk1 or cdk2, led to their nuclear translocation in human β-cells.

We next determined whether the nuclear translocation of the late cyclins and cdks is associated with the cell cycle entry of human β-cells. As shown in Fig. 4, and as expected, the overexpression of cdk6 and cyclin D3 led to an increased BrdU incorporation (15.3 ± 1.6%). The overexpression of cyclin A, cdk1, and cdk2 alone did not induce significant cell cycle entry in human β-cells. In contrast, cyclin E, when overexpressed alone, induced cell cycle entry (1.9 ± 0.4%). In contrast to the individual cyclins and cdks, combinations of late cyclins and cdks induced a significant increase of BrdU incorporation, with cyclin E and cdk1 being the most effective combination (13.3 ± 1.9%).

Since both early and late cyclins and cdks can each induce significant cell cycle entry in human β-cells, we next inquired whether an early plus late combination could additively or synergistically increase their cell cycle entry. As shown in Fig. 5, the combination of cdk6 and cyclin D3 (Fig. 5, C6+D3) with cdk1 and cyclin E (Fig. 5, K1+E) at 250 MOIs each led to a remarkable 42.8 ± 2.2% BrdU incorporation; in parallel, 54.3 ± 3.1% of human β-cells were positive for Ki67. This was significantly higher than either of the early or late cyclin/cdk combinations (Fig. 5B). These high levels of cell cycle entry were coupled with a significant degree of DNA damage (∼25%), as measured by γ-phospho-H2AX (43). Surprisingly, this DNA damage was not coupled with increased β-cell death, as assessed by TUNEL assay (Fig. 5G and H).

Since the high MOI (250 MOIs each) induced DNA damage, we repeated these experiments with lower MOI (100 MOIs each) (Fig. 6). Under these conditions, the combination of early and late cyclins and cdks still led to a dramatic and high level of cell cycle entry, with 19.6 ± 1.5% of β-cells labeled with BrdU and 35.4 ± 4.2% labeled with Ki67. At this MOI, DNA damage was minimal (∼5%) and was not associated with any β-cell death (TUNEL). These data indicate that both early and late cyclins and cdks are independently capable of driving human β-cells to enter the cell cycle. Further, their combination is complementary; whether this reflects synergy or additivity cannot be ascertained because high MOIs prevent full dose-response curves.

To determine whether the transduction with early or late cyclins and cdks might result in the loss of β-cell function, we examined glucose-stimulated insulin secretion 3 days after transduction. As seen in Supplementary Fig. 3, glucose-stimulated secretion remained robust when β-cells were transduced with both early and late cyclins and cdks.

We next explored the mechanism through which early and late cyclins and cdks might complementarily enhance cell cycle entry in human β-cells. Figure 7A shows the 16 residues known to be phosphorylated by cdk4/6 and cdk1/2 in other cell types. As shown in Fig. 7B, phospho-pRb was undetectable in human islets transduced with CTL Ad.LacZ. In contrast, pRb was phosphorylated when either early cyclins (cdk6 and cyclin D3; Fig. 7B, C6+D3) or late cyclins (cdk1 and cyclin E; Fig. 7B, K1+E) and cdks were overexpressed. Importantly, pRb was “hyperphosphorylated” when the combination of early and late cyclins and cdks was overexpressed. More specifically, pRb was phosphorylated on serine 780 (Ser⁷⁸⁰) and Ser⁷⁹⁵, which are well-known phosphorylation sites for cdk4/6, only when cdk6 was overexpressed (cdk6 and cyclin D3 or cdk6, cyclin D3, cdk1, and cyclin E). In contrast, pRb was phosphorylated on Threonine 373 (Thr³⁷³), a well-known target for cdk1, only when cdk1 was overexpressed (cdk1 and cyclin E or cdk6, cyclin D3, cdk1, and cyclin E). The specificity of the antibodies was tested in HepG2 cells, transduced with or without an adenovirus-silencing pRb (shRb).

Since human islets are composed of only ∼50% β-cells (44), immunoblots of whole islets may not reflect events in β-cells. Thus, we confirmed these findings in β-cells using a proximity ligation assay (Fig. 7C and D). As has been seen by immunoblots, pRb was phosphorylated (Fig. 7C, nuclear red dots signal) on Ser⁷⁸⁰ and Ser⁷⁹⁵ only when cdk6 was overexpressed (cdk6 and cyclin D3 or cdk6, cyclin D3, cdk1, and cyclin E). pRb was phosphorylated on Ser⁸¹¹ (nuclear red dots signal in Fig. 7D and a well-known site for cdk1) only when cdk1 was overexpressed (cdk1 and cyclin E or cdk6, cyclin D3, cdk1, and cyclin E). Interestingly, pRb phosphorylation was detected on Ser⁷⁸⁰, Ser⁷⁹⁵, and Ser⁸¹¹ when cdk6 and cyclin D3 were overexpressed in rat β-cells, suggesting the activation of both early and late cdk complexes.

While many examples of induced cell cycle entry in human β-cell have been reported (4,5,8,14,15,26,37–39,43,45), it is not typically clear whether this apparent “proliferation” actually leads to increases in human β-cell numbers. Indeed, we, and others (43), have suggested that at least some of the apparent proliferation observed using BrdU, Ki67, proliferating cell nuclear antigen (PCNA), and other proliferation markers might actually reflect DNA damage and DNA repair. Given the markedly increased levels of human β-cell replication shown in Figs. 5 and 6, it was important to determine whether this apparent proliferation can translate into actual increases in β-cell numbers. To address this question, we developed a high content–imaging method to measure the numbers of human β-cells in vitro 5 days after adenoviral transduction and plating. As shown in Fig. 8A and B, the plating of increasing numbers of human islet cells led to an escalating dose-response curve, confirming that the method can detect differences in human β-cell numbers. More importantly, the β-cell area was significantly increased and readily observed when cells were transduced with the combination of early and late cyclins and cdks (129.5 ± 8.7%) compared with islet cells transduced with a CTL adenovirus (LacZ) (89.6 ± 2.7%). The β-cell mass expansion occurred more effectively with the combination of early and late cyclins and cdks than either alone (95.3 ± 3.2% for cdk6 and cyclin D3 and 104.8 ± 4.8% for cdk1 and cyclin E). Since the assay assesses β-cell area, the increase in β-cell area from the cdk/cyclin combination could reflect either an increased number of β-cells and/or an increase in the sizes of individual β-cells. Thus, we next determined the size of human β-cells using E-cadherin immunolabeling (23) (Fig. 8C and D). The sizes of human β-cells transduced with the combination of early and late cyclins and cdks was significantly smaller than those of CTLs. These data suggest that the combination of early and late cyclins and cdks not only induced a remarkable cell cycle entry in human β-cells but also increased the number of β-cells in vitro; that is, it was associated with completion of the cell cycle and the generation of daughter cells. Collectively, as summarized in Fig. 8E, these data suggest that the combination of early and late cyclins and cdks led to the expansion of human β-cells in vitro. It also provides a useful assay for quantifying human β-cell expansion.

This report contains a number of novel findings, some of which are summarized in the model in Fig. 8F. First, we show that early and late cyclins and cdks prompt human β-cells to enter the cycle independently. Second, we show that, in combination, they complementarily enhance cell cycle entry in the human β-cell. Third, we show, using two independent approaches, that the cumulative phosphorylation of distinct pRb residues is one mechanism through which this occurs. Fourth, we show that the expression and nuclear translocation of both early and late cyclins and cdks not only lead to an increase in surrogate markers for human β-cell replication but also actually lead to an authentic increase in human β-cell numbers. And, fifth, we provide a novel assay for measuring actual increases in human β-cell numbers.

Regarding the last two points, the ability to demonstrate increases in β-cell numbers represents an important advance, since prior studies on human β-cell “replication,” including our own, have used canonical markers of replication such as BrdU incorporation or Ki67, PCNA, or phospho-histone H3 immunolabeling. These approaches mark different phases of the cell cycle but do not distinguish between a completed cell cycle with daughter β-cells and a failed attempt at replication or DNA damage and repair. For example, Ki67 marks cells in late G1, S, or G2/M phases. BrdU and PCNA mark cells in the S phase. Phospho-histone H3 marks cells in G2/M. All can also reflect DNA damage and/or repair. Thus, since the markers of proliferation are not infallible, and coupled with the brief viability of mature β-cells in vitro, their very low replication levels, and their recalcitrance to proliferate, it has been particularly challenging to demonstrate authentic increases in the numbers of primary β-cells: definitive evidence of authentic increases in human β-cell numbers in vitro has not been reported (3,5,7,8,14,15,26,28,31,43).

Here, we report a simple method to address this important issue in vitro. Using this method, we have demonstrated that the activation of both early and late cyclins and cdks not only leads to robust cell cycle entry as assessed by BrdU incorporation and Ki67 labeling but also increases the β-cell area. This β-cell area expansion reflects increased β-cell numbers rather than an increased β-cell size, since β-cell size is smaller with the combination of early and late cyclins and cdks. This strongly supports the argument that the β-cell expansion observed is representative of authentic increased β-cell numbers.

The increased β-cell number is supported by the calculated expected β-cell number. If one estimates that 20–40% of β-cells were actively engaged in the cell cycle (Figs. 5 and 6); that half of these labeled cells represent new daughter cells and the other half represent the residual parent cells; that adenovirus transduction leads to cyclin/cdks expression after 24 h, leaving 4 days for β-cells to divide; and that, after replication, β-cells will divide only once during the 5 days of the experiment (25,46–48), one can estimate that the yield of new daughter cells should be in the range of 30%. This is not dissimilar from the 29.6% increase in β-cell mass with the combination of early and late cyclins and cdks.

The observation that early and late cyclins and cdks can independently induce pRb phosphorylation and cell cycle entry in human β-cells is important. A number of studies suggest that the sequential phosphorylation of pRb may not be necessary, that different mitogenic signaling pathways may activate either early or late cyclins and cdks, and that either is sufficient to initiate cell cycle progression. For example, phosphorylation of only a single residue in the C terminus of pRb is able to lead to its inactivation (49). Another mutational study has found that Thr³⁷³ is a critical phosphorylation site for pRb inactivation (50). Thus, the observation that early and late cyclins and cdks can independently phosphorylate pRb on distinct residues in human β-cells fits these recent models. The question of whether these early and late cdk phosphorylations occur on the same pRb molecule or two different molecules remains unclear (Fig. 8E). These findings provide a working model for the independent activation of the cell cycle in human β-cells by early and late cyclins and cdks, as well as for their complementarity.

In addition, several studies have demonstrated that late cyclins and cdks can induce the cell cycle without prior activation of early cyclins and cdks. Perhaps the best examples are in the studies by Kozar et al. (51), showing that most organs develop normally in the absence of all three d-cyclins, and Kozar and Sicinski (52), showing that cdk4 and cdk6 are dispensable for development in mammals. In another report, Keenan et al. (53) showed that ectopic expression of cyclin E dispenses with a need for cyclin D-cdk4 in pRb inactivation, E2f activation, and cell cycle progression. Another group has shown (54) that early and late cyclins and cdks can use distinct cell cycle re-entry programs in differentiated cells. Further, cyclin E has been shown to induce S phase without activation of the pRb/E2F pathway (55). Thus, it is also possible that, in addition to pRb phosphorylation and inactivation, late cyclins and cdks induce cell cycle entry complementarily without prior early cdk activation and pRb phosphorylation and/or independently of pRb in human β-cells. Finally, as noted in the introduction, several β-cell mitogens appear to act exclusively via the late cyclins and cdks (27–29).

The increase in β-cell numbers demonstrates that at least some of the proliferation observed in response to cyclins and cdks is authentic and supports the notion that the induction of proliferation by cyclins and cdks can occur without extensive DNA damage, β-cell death, or loss of function. Of course, DNA damage was observed (Fig. 6), and it is likely that some of the damaged β-cells may have died (43). On the other hand, recent reports (56) have revealed that cdks are actually involved actively in DNA repair and studies (57) in yeast and mammalian cells have revealed that cdk activity is essential for DNA resection and the progression of DNA repair during S and G2. Thus, it is possible that, even if DNA damage was induced, β-cells may be able to repair their DNA and complete the cell cycle. Finally, we cannot exclude the possibility that some of the increase in the number of β-cells may result from transdifferentiation from other islet cells (58,59).

The observation that the nuclear translocation of cdk1 and cdk2 is associated with high levels of cell cycle entry in rat and human β-cells is in accordance with the concept that cdks need to enter the nucleus in order to carry out their function as pRb kinases (19,60). In rapidly proliferating INS1 cells, cdk1 and cdk2 were mostly nuclear. On the other hand, under basal conditions, and in quiescent rat or human β-cells, late cyclins and cdks were predominantly cytoplasmic. The 14-3-3 proteins or the protein kinase MYT1 have been described in other systems to bind and retain cdks in the cytoplasm (19,20). But the exact nature of the scaffolding protein keeping cyclins and cdks in human β-cells remains unknown. When cdk1 or cdk2 was overexpressed alone in human β-cells, they remained cytoplasmic, and no cell cycle entry was observed; cell cycle entry occurred only when these cdks were provided in combination with cyclin A or E and in association with their translocation to the nucleus. The findings on Rb phosphorylation in Fig. 7 and Supplementary Fig. 3 support the same idea: early cyclins and cdks can induce the nuclear translocation and activity of late cyclins and cdks in rat β-cells but not in human β-cells.

Cyclin E was the only late G1/S molecule able to induce cell cycle entry when overexpressed alone in human β-cells. Correspondingly, overexpressed cyclin E was also found more often than other cyclins in the nuclear compartment. This likely reflects the fact that cyclin E has a classic nuclear localization signal (NLS) that targets it to the nucleus via the well-characterized importin-α/importin-β nuclear import pathway (61,62). Neither cyclin A nor cdk1 or 2 have a consensus NLS. Cyclin A is known, however, to bind a variety of proteins with a recognizable NLS, such as p107, E2F1, p21, and p27^Kip1, but cdks rely on their cyclins to traffic to the nucleus (62). Collectively, these observations make it likely that cdk1 and cdk2 can only translocate to the nucleus of human β-cells when complexed with cyclin A or cyclin E, subsequently phosphorylating pRb and driving β-cell replication.

Human β-cells are recalcitrant to the induction of proliferation compared with rodent β-cells (1,5,7,8,14,26,32–41). Here, we suggest that the nuclear translocation of cdk1 and cdk2 is associated with the difference in replication levels between rat and humans. An alternate explanation for this refractoriness may be age. Indeed, most studies on human β-cell proliferation use islets derived from human cadaveric donors in the age range of 40–60 years, reflecting access to cadaver human islets. In contrast, most studies on rodent β-cell proliferation are performed in young animals, when β-cells are the most responsive to the induction of proliferation. Several studies (24,63–70) have shown that rodent β-cell proliferation capacity declines with age. Surprisingly, we observe that the cell cycle entry induced by cdk6 and cyclin D3 was maintained with age in rat β-cells in vitro. In preliminary studies, we find no association between the ages of human islet donors and the levels of proliferation induced by cdk6 and cyclin D3 (Supplementary Fig. 5). Thus, aging may not be the most significant factor causing the difference in proliferation levels between human and rat in response to cdk6 and cyclin D3. Another possibility is that cell cycle inhibitors, such as p16^Ink4A or p27^Kip1, translocate to the nucleus in human β-cells but not in rat β-cells, limiting the cell cycle entry induced by cdk6 and cyclin D3. However, this is improbable since the higher levels of replication in rat β-cells and INS1 cells are associated with higher nuclear translocation of p16^Ink4A and p27^Kip1 compared with human β-cells (Supplementary Fig. 1). Finally, it is also possible that the amino acid sequences of rat and human cdk1 and/or cdk2 are different, allowing the nuclear translocation of rat cdk1 and cdk2 but not human cdk1 and cdk2. Again, this scenario is unlikely since rat and human cdk1 and cdk2 do not have a recognizable NLS (19,20) and since rat and human cdk1 and cdk2 sequences are highly homologous (97% and 98%, respectively).

These studies have limitations and raise new questions. Are human β-cells only capable of completing their cell cycle when a specific threshold of cell cycle entry is reached? Will only high level of cell cycle entry permit the detection of β-cell expansion in vitro? How high must this level be, and for how long must this replication be sustained? One limitation is that the majority of the study was performed using isolated and dispersed islets to the single cells, and human β-cells may not respond similarly in intact islets. Therefore, one question would be whether this remarkable induction of human β-cell replication and expansion can be achieved in intact islets in vivo. Will this high level of replication be therapeutically relevant to replacing lost β-cells in diabetes? Answers to these and other questions need to be addressed. Nonetheless, the findings demonstrate that it is actually possible to expand primary adult human β-cells, at least in vitro. The findings also provide a simple method to measure β-cell expansion in vitro. Finally, they suggest that drugs or small molecules designed to activate both early and late cdk nuclear translocation might be of particular therapeutic efficacy for inducing human β-cell expansion.

Acknowledgments. The authors thank Drs. Adolfo Garcia-Ocaña, Donald K. Scott, Andrew F. Stewart, and Rupangi C. Vasavada at the Icahn School of Medicine at Mount Sinai for many helpful discussions during the preparation of these studies. The authors also thank Drs. Peng Wang and Hainan Chen at the Icahn School of Medicine at Mount Sinai for making and testing the adenovirus to silence pRb. In addition, the authors thank Dr. Rumana Huq and the Microscopy Core at the Icahn School of Medicine at Mount Sinai for superb support.

Funding. This work was supported by American Diabetes Association Research Grant 7-12-BS-046 to N.M.F.-T. Human islets were generously supplied by the Integrated Islet Distribution Program, which was supported by the National Institute of Diabetes and Digestive and Kidney Diseases and JDRF.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. S.T., C.R., R.W., G.C., M.T., and K.K.T. researched the data. N.M.F.-T. researched the data, contributed to the discussion, and wrote the manuscript. S.T. and N.M.F.-T. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented at the 75th Scientific Sessions of the American Diabetes Association, Boston, MA, 5–9 June 2015.