DNA Damage Does Not Cause BrdU Labeling of Mouse or Human β-Cells

β-Cell replication is the principal process generating new β-cells; therapeutic induction of β-cell replication is an important goal for diabetes research (1–3). The incorporation of nucleoside analogs has been a high-value tool in quantifying cell proliferation behavior for decades, allowing measurement of cumulative S-phase entry during a defined exposure period, in vitro and in vivo. Nucleoside analogs such as BrdU and 5-ethinyl-2′-deoxyuridine (EdU) (4) have been used not only in β-cell biology, but also broadly across developmental and cell biology.

Under certain conditions, BrdU incorporation in β-cells has been observed to colocalize with markers of DNA damage (5,6). In other cell types, BrdU exposure has been shown to activate a DNA damage response (7–9), but in β-cells the reasons for this colocalization are not well understood. The observation has been widely discussed and rapidly incorporated into the thinking in the field, with a range of impacts. In the most discussed work (5), the conclusion drawn by the originating authors was that some BrdU-labeled human β-cells, particularly the subset with atypical punctate BrdU staining, fail to complete S-phase, instead showing evidence of DNA damage, DNA rereplication, and failure to divide. In other words, BrdU did label cells that transitioned into S-phase, but BrdU-labeled cells could not be assumed to progress through successful mitosis. However, concern in the field has extended beyond this concept; in many quarters, the question has been raised as to whether BrdU labeling, counted as evidence of S-phase entry, could in fact be due to a completely unrelated process: nucleotide incorporation during DNA damage repair. If this hypothesis were true, then nuclei might label for BrdU in the absence of S-phase entry, invalidating some prior results and diminishing the value of nucleoside analogs in the study of cell replication.

The field of β-cell regeneration has been impacted by uncertainty of the correct interpretation of BrdU-labeled nuclei. There is an urgent need to clarify the reasons for co-occurrence of BrdU and DNA damage labeling in β-cells. The goals of this study were to explore the possible causes of BrdU and γ-phosphorylated H2A histone family member X (γH2AX) colocalization in nuclei of mouse and human β-cells, and to specifically test whether there are conditions in which BrdU labeling can be induced by DNA damage-related cellular processes rather than cell cycle entry.

All mouse procedures were approved by the UMass Medical School Institutional Animal Care and Use Committee. Young (10- to 12-week-old) or old (50- to 60-week-old) C57BL/6J male and female mice, from an in-house breeding colony, had continuous access to normal mouse chow, on a 12-h light/dark cycle. Islets were isolated by pancreatic ductal collagenase injection and Ficoll (Histopaque-1077; Sigma-Aldrich) gradient (10). Islets from multiple mice were pooled and mixed before experiments. Whenever possible, all control and experimental comparisons were performed in parallel on islets from the same mice. Each combined pool of islets was considered one biological replicate.

To study proliferative conditions in vivo, pancreas sections were analyzed from experiments previously published on 10- to 12-week-old male mice fed a high-fat diet for 7 days (11), or 10- to 12-week-old male mice with continuous hyperglycemia achieved by intravenous infusion of glucose (10,12).

Whole islets were cultured overnight in islet medium (RPMI medium, 10% FBS, penicillin/streptomycin, 5.0 mmol/L glucose) after isolation. For dispersed-cell experiments, islets were hand picked and digested with single-use-apportioned 0.05% trypsin, and 50 islet equivalents per well were plated on uncoated glass coverslips (Fisherbrand; Thermo Fisher Scientific, Waltham, MA) in islet medium (12). Mouse islet cell experiments were 72 h in duration, starting the day after dispersion, with adenovirus, glucose and/or harmine (5 μmol/L) (catalog #286044; Sigma-Aldrich) exposure for 72 h, and BrdU (10 μg/mL) added 24 h prior to fixation. For whole-islet experiments, islets were cultured for 72 h in islet medium with additives including glucose and/or mitomycin C, with BrdU added for the final 24 h, fixed in 10% formalin for 30 min, washed with PBS, and embedded in paraffin for sectioning (13). For virus transduction, dispersed cells were transduced with adenovirus expressing bacterial Cre recombinase (control virus) or human cyclin D2 protein, at a multiplicity of infection of 5, as described previously (12). Mitomycin C (0.5 μmol/L) (Sigma-Aldrich) was added at the start of the 72-h experiment. Ultraviolet (UV) irradiation of islet cell cultures was administered by placing the culture dish for 10 min on a UVP Benchtop Transilluminator equipped with a 302-nm filter (Cole-Parmer) after 24 h of glucose treatment.

Human islets received from the Integrated Islet Distribution Program (IIDP) at City of Hope (Duarte, CA) (Supplementary Table 1) were cultured in islet medium overnight after shipment, then dispersed and plated on coverslips, as described above for mouse islets. After 1 day of recovery from plating, human islet cell experiments were performed exactly as described above, except that the experiment duration was 96 h and BrdU was included for the entire 96 h of culture.

Islet cell cultures were fixed in 4% paraformaldehyde for 10 min at room temperature. Whole islets (13) and pancreas sections were obtained from formalin-fixed, paraffin-embedded tissue blocks, as described in the publications where these samples originated (10–12). For BrdU, γH2AX, and pHH3 (phosphohistone H3) staining, fixed cells or rehydrated paraffin sections were submerged in 1 N HCl for 20 min at 37°C, blocked for 2 h in goat serum–based block with 0.1% Tween 20, labeled with primary antibodies then secondary antibodies (Invitrogen), and mounted on glass slides with Fluoroshield mounting media containing DAPI (Sigma-Aldrich). The primary antibodies used were guinea pig anti-insulin (catalog #A0564; Dako); rabbit anti-γH2AX (catalog #9718; Cell Signaling Technology); mouse anti-γH2AX (catalog #05–636; EMD Millipore); rat anti-BrdU (catalog Ab6326; Abcam); rabbit anti-pHH3 (catalog #3377; Cell Signaling Technology); and rabbit anti-Ki67 (catalog RM-9106; Thermo Fisher Scientific). Fluorescent images were acquired using a Nikon fluorescent microscope or a Solamere CSU10 Spinning Disk confocal system mounted on a Nikon TE2000-E2 Inverted Microscope. Image data were quantified by an unbiased laboratory member (B.G.) using Cell Profiler automated counting software from the Broad Institute (Cambridge, MA) (14,15) with manual image checking postquantification. A cell was considered to be γH2AX(+) if the nucleus was labeled with multiple intense nuclear foci. The number of cells counted for each experimental condition are included in Supplementary Table 2.

Data were analyzed and graphed using GraphPad Prism 7. In all column plots, each sample is represented by one data point. Venn diagrams were generated using EulerAPE 3.0.0 from the University of Kent (Canterbury, U.K.) (16). Data were compared using a two-tailed Student t test; P < 0.05 was considered to be significant.

Consistent with the reports of human β-cell colabeling (5), we have observed BrdU-γH2AX colabeling in primary mouse islet cell cultures (17) (Fig. 1). We reasoned that these two markers could label the same nucleus if 1) DNA damage leads to BrdU incorporation, 2) BrdU incorporation leads to DNA damage, 3) an upstream process leads to both labels, or 4) BrdU incorporation and DNA damage occur stochastically and occasionally occur in the same cell but are unrelated (Supplementary Table 3).

We first tested the simplest explanation for colocalization: that BrdU labeling and DNA damage each occur in a fraction of cultured primary β-cells, unrelated to one another, leading to occasional co-occurrence in the same cell. If this model were true, γH2AX and BrdU labels should both be detectable in standard culture conditions, and the presence of one label in a cell should not influence the likelihood of the presence of the other label. We cultured dispersed mouse islet cells under standard, unstimulated conditions. When the cells were fixed and immunostained for insulin, BrdU, and γH2AX, a small but consistent fraction of insulin-containing cells was labeled for each marker (Fig. 1A, A′, and B). To test whether colabeling could be due to random co-occurrence of two unrelated processes, we calculated the predicted likelihood of random colabeling based on the frequency of each individual label (predicted fraction colabeling was defined as the product of the observed fractions of each single label) and compared this predicted frequency with the actual observed frequency of colabeling. The observed frequency, although low, was higher than the predicted random co-occurrence frequency (Fig. 1C). Confocal microscopy confirmed nuclear colabeling in some β-cells (Fig. 1D and D′). As reported previously (5), in some BrdU-γH2AX–colabeled nuclei, the BrdU labeling was punctate in appearance. Thus, in unstimulated culture conditions, the colabeling of BrdU and γH2AX occurred more frequently in mouse β-cells than predicted by chance, suggesting that one process influences the other or that both are influenced by a common upstream process.

Since the overexpression of HNFA8 or cyclin D3/Cdk6 in combination (5) increased the number of colabeled cells, we reasoned that proliferation drivers increase the frequency of double-labeled cells. Since the originating work was in human β-cells, which are considerably older than young-adult mouse β-cells, we also considered that β-cells from older mice might be more vulnerable. We tested whether the mitogenic stimuli frequently used in our laboratory, 15 mmol/L glucose and adenoviral-delivered cyclin D2 (12,17), altered the likelihood of BrdU and γH2AX colabeling in β-cells from younger (10- to 12-week-old) and older (50- to 60-week-old) mice. In 15 mmol/L glucose, both BrdU and γH2AX (Fig. 2A and B) labeling increased. Supporting the concept that BrdU labeling reflects S-phase entry, β-cells from younger mice had a higher frequency of BrdU incorporation than β-cells from older mice. Perhaps surprisingly, high-glucose β-cells from older mice did not label more frequently for γH2AX, or both markers, than β-cells from younger mice (Fig. 2C).

To test whether directly forcing cell cycle entry increases BrdU-γH2AX colabeling, we overexpressed cyclin D2. In normal 5 mmol/L glucose, Ad-cyclin D2 produced similar results to 15 mmol/L glucose stimulation (Fig. 2D–F): BrdU and γH2AX were individually modestly increased, the frequency of double-labeled cells increased, and older mouse age did not increase the frequency of β-cells with γH2AX or double labeling. However, when cyclin D2 overexpression was combined with 15 mmol/L glucose (Fig. 2G–I), BrdU, γH2AX, and BrdU-γH2AX–labeled β-cells were markedly increased over baseline. In all proliferation-stimulated conditions (Fig. 2C, F, and I), the observed frequency of colabeled cells was significantly higher than predicted if colabeling was due to chance. To test a third, unrelated proliferative stimulus, we treated the cultures with harmine, a Dyrk1a inhibitor that increases mouse and human β-cell proliferation (18). Harmine increased BrdU labeling synergistically with glucose and increased γH2AX labeling in both 5 and 15 mmol/L glucose, and, similar to the other stimuli, the observed frequency of colabeling was higher than predicted (Fig. 2J–L). Thus, the frequency of BrdU-γH2AX double labeling is increased by proliferative stimuli, especially when stimuli are combined in aging mouse β-cells.

To explore whether DNA damage and BrdU incorporation influence each other, we next tested whether BrdU incorporation was increased during conditions that induce DNA damage. These were perhaps the most critical experiments in this study, because if BrdU is spuriously incorporated as part of a DNA damage response, then BrdU labeling cannot be used to quantify S-phase cell cycle entry. To address this question we cultured β-cells in DNA damage–inducing conditions and quantified BrdU- and double-labeled cells. In islet cell cultures exposed to a sublethal dose of mitomycin C, which induces DNA damage by cross-linking DNA (19–21), almost all β-cells had γH2AX labeling (Fig. 3A), confirming DNA damage. To test whether DNA damage caused spurious BrdU incorporation, we stained for BrdU. The results showed the opposite; mitomycin C strongly suppressed BrdU labeling frequency in both young and old β-cells, and the observed frequency of double-labeled cells was suppressed to the predicted frequency if colabeling were caused by random co-occurrence of unrelated processes (Fig. 3B and C). These results suggest that DNA damage repair following mitomycin C exposure did not cause spurious BrdU incorporation in mouse β-cells that could be misconstrued as proliferation.

To thoroughly explore this important question, we repeated the experiment using a second DNA damage insult: UV irradiation, which damages DNA by joining adjacent thymine nucleotides into pyrimidine dimers (22–24). We selected a UV dose that caused moderate DNA damage based on trial experiments (data not shown). Cellular response to UV irradiation was variable, with two experiments showing a high proportion of cells labeling for γH2AX (Fig. 3D–F, circles), and two experiments, despite being performed identically, showing fewer cells labeling for γH2AX (Fig. 3D–F, triangles). Similar to the mitomycin C result, UV irradiation–induced DNA damage suppressed BrdU incorporation rather than increasing it, and reduced the observed frequency of double-labeled β-cells to that predicted if colabeling was due to random co-occurrence of unrelated events (Fig. 3E). Reduced BrdU incorporation even in the cultures with only modest γH2AX labeling frequency (Fig. 3E, triangles) suggests that the suppression of BrdU label was not due to the overwhelming toxicity of the DNA damage insult. Taken together, these data suggest that BrdU labeling in cultured mouse β-cells is not increased by DNA damage.

Under most conditions, we noticed that although many BrdU-labeled cells were not γH2AX labeled, the converse was not true: most γH2AX-labeled cells were BrdU labeled. Stated another way, the population of γH2AX-labeling β-cells seemed to be a subset of the BrdU-labeled population. Expressing the data using quantitative Venn diagrams confirmed this observation; especially under proliferation-stimulated conditions, the majority of γH2AX-labeled β-cells were within the BrdU-labeled population (Fig. 4A and B). In contrast, under DNA-damaging conditions, Venn analysis showed that the vast majority of γH2AX-labeling β-cells did not label with BrdU (Fig. 4C). We used confocal microscopy to identify the pattern of BrdU labeling in double-labeled cells. We specifically asked whether BrdU labeled nuclei uniformly, as is characteristic of S-phase DNA replication, or in punctate fashion overlapping with foci of the γH2AX DNA damage label. Images showed that many of the double-labeled nuclei were uniformly labeled with BrdU (Fig. 4D and E), suggesting the possibility that γH2AX labeling occurred after DNA replication. Some nuclei had punctate BrdU, as described previously (Fig. 4F) (5). Consistent with the notion that DNA damage prevents successful cell division, the few mitotic nuclei visualized had no γH2AX label (Fig. 4G). On the other hand, mitotic figures did label for BrdU (Fig. 4G), and many BrdU-labeled nuclei costained for the mitotic marker pHH3 or appeared as doublets (Supplementary Fig. 1), confirming that some BrdU-labeled β-cells do enter mitosis. We tested whether γH2AX-BrdU colabeling of β-cells occurs in vivo under conditions in which β-cell mass increases, such as high-fat feeding (11) and hyperglycemia (10,12). Careful examination of many islets in sections from multiple mice with elevated BrdU incorporation revealed very rare γH2AX labeling in either condition: in eight pancreata from the high-fat diet experiment, only four islet-related nuclei were found that stained for γH2AX (Supplementary Fig. 2), and in four pancreata from the hyperglycemia experiment only four islet-related nuclei stained for γH2AX (Supplementary Fig. 3). Of the eight islet nuclei labeling for γH2AX across both experiments, six were β-cell nuclei, but only two costained for BrdU; interestingly, these two were found in the same islet (Supplementary Fig. 2B). To assess whether γH2AX labeling is induced by ex vivo islet culture in general, or by dispersed islet cell cultures specifically, whole intact islets were cultured under proliferative conditions and assessed for BrdU and γH2AX colabeling (Supplementary Fig. 4). β-Cells colabeling for both markers were very rare in intact islets (Supplementary Fig. 4D). Taken together, these results suggest that in ex vivo dispersed islet cell culture, but not intact islet culture or in vivo pancreas, conditions that promote cell cycle entry or BrdU incorporation increase β-cell risk of γH2AX labeling, whereas conditions that promote DNA damage do not increase the likelihood of BrdU incorporation.

Since the data suggested that cell cycle entry, or possibly BrdU incorporation itself, increased the likelihood of γH2AX labeling, and nucleoside analogs are known to induce a DNA damage response in other cell types (7–9), we next asked whether BrdU exposure leads to DNA damage in β-cells. To test this, we cultured mouse islet cells with or without BrdU and assessed the insulin(+) fraction labeling for γH2AX. Omitting BrdU from the culture medium eliminated all BrdU labeling in the cultures (data not shown). In the absence of BrdU exposure, we observed a subtle reduction in γH2AX labeling in β-cells from younger mice that was not observed in β-cells from older mice (Fig. 5A). Hypothesizing that γH2AX labeling might increase over time following BrdU exposure, we compared cells exposed to BrdU during the first 24 h (condition X), the last 24 h (condition Y), or the entire 72-h glucose exposure (condition Z) (Fig. 5B). γH2AX labeling was not different among these conditions (Fig. 5C). To test whether BrdU exposure increased DNA damage under more pronounced proliferative stimuli, we cultured islet cells with or without BrdU in the presence of Ad-cyclin D2 in low-glucose (Fig. 5D) or high-glucose (Fig. 5E) conditions. Intriguingly, BrdU exposure increased γH2AX labeling only in the low-glucose condition. To test whether γH2AX labeling was more related to cycling β-cells or to BrdU incorporation itself, we stained proliferation-stimulated cultures with pHH3 and γH2AX, predicting that if γH2AX labeling was due to BrdU toxicity, there would not be colabeling of γH2AX with pHH3 in the absence of BrdU exposure (Fig. 5F–H). In fact, β-cell nuclei frequently colabeled with both pHH3 and γH2AX, at levels much higher than expected if colabeling was due to random chance (Fig. 5G and H). Similar results were observed with experiments in which costaining was performed for Ki67 and γH2AX (data not shown). These data suggest that although BrdU exposure or BrdU incorporation into DNA under certain conditions can cause double-strand DNA breaks or nucleotide excision/repair that label with γH2AX (23), this toxicity is not the major cause of BrdU-γH2AX double-labeled cells in islet cell cultures.

The goal of studies in β-cell regeneration is to find approaches to expand human β-cell numbers (25). To learn more about BrdU-γH2AX colabeling in human β-cells (5,6), dispersed human islet cells were cultured under the same conditions as in the mouse experiments, and tested for BrdU incorporation, γH2AX labeling, and colabeling for both markers. Under basal 5 mmol/L glucose conditions, the frequency of human β-cells labeling for BrdU was low, and the frequency of γH2AX labeling was also rare (Fig. 6A and B). To determine whether DNA damage increased in human β-cells during conditions of forced proliferation, we compared basal conditions with pro-proliferative stimuli 15 mmol/L glucose without or with Ad-cyclin D2. Similar to prior published results, insulin(+) cells in human islet cultures were relatively resistant to proliferative stimuli, showing only a marginal increase in the fraction labeling with BrdU when exposed to 15 mmol/L glucose (Fig. 6A–C), Ad-cyclin D2 (Fig. 6D–F), or a combination of the two (Fig. 6G–I). Surprisingly, only a small fraction of human β-cells labeled for γH2AX under basal or stimulated conditions (Fig. 6B, E, and H). The fraction of human β-cells labeling for both labels remained low and was not significantly higher than predicted by the random chance of co-occurrence of two rare events (Fig. 6C, F, and I). These data suggest that under these growth conditions, human β-cells do not frequently colabel for γH2AX and BrdU.

To test whether human β-cells incorporate BrdU during DNA damage repair, DNA-damaging conditions were applied to human islet cultures. Mitomycin C exposure caused γH2AX labeling in the majority of β-cells (Fig. 7A). Despite widespread DNA damage in the cultures, the percentage of insulin(+) cells labeling for BrdU decreased (Fig. 7B) and the observed fraction of double-labeled cells was similar to predicted if colabeling was by random chance. When human islet cultures were exposed to UV irradiation (Fig. 7D–F), the DNA damage response was variable (Fig. 7D), but BrdU incorporation was suppressed (Fig. 7E) and there was no increase in the proportion of β-cells labeling for both markers (Fig. 7F). In sum, human β-cells behaved similarly to mouse β-cells: DNA damage did not cause a spurious increase in BrdU labeling.

Quantitative Venn analysis of the human β-cell BrdU- and γH2AX-labeling populations revealed the interesting result that under basal and proliferation-stimulated conditions, insulin(+) cells labeling for BrdU were mostly not the same cells that labeled with γH2AX (Fig. 8A). In fact, without Ad-cyclin D2, no double-labeled cells were identified. Similar to mouse islet cultures, however, DNA damage stimuli caused the majority of human β-cells to label for γH2AX, and only a tiny fraction of those cells also labeled for BrdU (Fig. 8B). Overall, fewer human β-cells labeled for γH2AX than mouse. As in mouse β-cells, DNA damage conditions did not induce spurious BrdU incorporation in human β-cells. Neither BrdU nor γH2AX were meaningfully increased by proliferative stimuli in human β-cells, and in these conditions, the observed double-labeled fraction was not increased over that predicted by random chance.

These studies explore the frequency and causes of mouse and human β-cell colabeling for BrdU and the γH2AX DNA damage marker. The results suggest that in mouse β-cells, colabeling occurs more frequently than predicted if the association is due to random chance. Evidence does not support the hypothesis that colabeling is due to BrdU incorporation during DNA repair. Instead, the results suggest that γH2AX labeling can be triggered by something related to the proliferative process, either by an upstream proliferation stimulus or by cellular events associated with the cell cycle itself.

These experiments are important, because the uncertainty of the fidelity of BrdU as a label of S-phase in β-cells has impacted the field. For some, conflating the concepts of cell cycle and DNA damage has rekindled long-standing doubts that β-cells can replicate at all, despite a large quantity of evidence to the contrary (3,14–16). Investigators have been at times required to use multiple duplicative tools to measure the same outcome, slowing progress. The results of these current studies are reassuring. Labeling for γH2AX in human β-cells was rare, even under proliferation-stimulating conditions (25), and most BrdU(+) human β-cells were not γH2AX(+). In vivo mouse pancreata rarely labeled with γH2AX, even under conditions when many β-cells were BrdU(+). Overall, the results are consistent with the conclusion that, even in mouse islets, ex vivo BrdU authentically labels β-cells that enter S-phase.

The data show that DNA damage does not cause BrdU labeling that could be misconstrued as S-phase entry, in either mouse or human β-cells. Using two different DNA damage–inducing conditions, one of which caused widespread damage in essentially all β-cells (mitomycin C), and the other of which caused more moderate damage (UV irradiation), BrdU incorporation was reduced rather than increased. Furthermore, the observed frequency of double-labeled nuclei was suppressed in DNA damage conditions to that expected if the occurrence was due to random chance. Thus, β-cell colabeling for BrdU and γH2AX in these cultures was not caused by spurious BrdU incorporation as part of the cellular response to DNA damage.

The data show a clear association between γH2AX labeling and proliferative stimuli in mouse β-cells. γH2AX labeling, and γH2AX-BrdU or γH2AX-pHH3 colabeling, occurred occasionally under unstimulated conditions but were markedly increased in 15 mmol/L glucose with or without the overexpression of cyclin D2 or the addition of harmine. Intriguingly, the γH2AX-labeled cells were mostly a subset of the BrdU-labeled population. Also, many γH2AX-labeled nuclei were evenly and completely BrdU labeled, giving the impression that S-phase had occurred smoothly. Taken together with the observations that DNA damage suppressed new S-phase entry and that γH2AX foci were absent from the few mitotic spindles observed, we speculate that the γH2AX foci may mostly occur in a cell cycle window from late S-phase through G2 phase or in postmitotic daughter cells. γH2AX is best known for labeling double-strand breaks or nucleotide excision repair, recruiting DNA repair effectors (23,24). γH2AX also labels DNA replication stress, such as replication fork collapse, defects in chromatin assembly, cell cycle arrest, or prolonged mitosis (26–29). This raises the possibility that β-cells forced into the cell cycle may have weakened replication forks or other defects in completing mitosis. In some cell types, γ-phosphorylation of H2AX may occur intrinsically during S- or G2-M phases of the cell cycle (30,31). It is possible that chromatin handling during DNA replication or G2 may predispose β-cells to γH2AX phosphorylation; whether this is a sign of damaged DNA or another process remains uncertain.

A toxic effect of nucleoside analogs, including both BrdU and EdU, has been reported (7–9). In the β-cell field, EdU is also commonly used, especially for high-throughput screening; whether EdU performs similarly to BrdU has not been carefully tested. The final concentration of BrdU used in our study (33 μmol/L) was similar to those in other published studies (10–100 μmol/L) (5,7,8). BrdU exposure may or may not induce sister chromatid exchanges, a marker of genome instability (32). Triphosphate nucleoside analogs such as 5-fluorouracil act as replication fork blocks; however, these induce γH2AX not at the time of initial incorporation, but rather during S-phase of a subsequent cell cycle event (33). Extensive colabeling of γH2AX and pHH3 in the absence of BrdU exposure argue against a primary role for BrdU toxicity in the observed γH2AX population, but rather that γH2AX staining is associated with cycling cells. β-Cells with smooth BrdU label and γH2AX foci may be cells that have re-entered the cell cycle following a successful initial cell division.

Surprisingly, β-cells from older mice were not spontaneously more likely to label for γH2AX. However, when both high-glucose conditions and cyclin D2 overexpression were combined, β-cells from older mice had a higher frequency of γH2AX labeling, and a lower fraction of BrdU-labeled β-cells that did not show evidence of DNA damage. This may be related to the fact that β-cells from older mice required a combination of both glucose and cyclin D2 to enter the cell cycle, whereas β-cells from younger mice responded to either stimulus. The human islet donors used for this study were from a narrow range of ages, 41–56 years, precluding assessment of the impact of age on DNA damage markers in human β-cells.

Also contrary to expectations, in these studies human β-cells did not have a high frequency of γH2AX labeling. If anything, human β-cells were less likely than mouse β-cells to label for γH2AX, especially in the presence of proliferative stimuli. However, the pro-proliferation stimuli used in these studies did not result in a high rate of cell cycle entry in human β-cells. As such, the low rate of human γH2AX labeling is consistent with the hypothesis that DNA damage occurs in later stages of the cell cycle, or during cell cycle re-entry following successful mitosis. We have not tested conditions that induce rapid proliferation in human β-cells.

This study does not address the extent to which BrdU, marking S-phase entry, predicts completion of the cell cycle and production of two daughter cells. Evidence supporting the conclusion that some BrdU-labeled β-cells do successfully complete cell division include the many studies showing that BrdU labeling correlates with other proliferation markers such as Ki67, PCNA (proliferating cell nuclear antigen), and pHH3 (17,34,35), the documented increase in β-cell number under conditions when BrdU labeling is increased in vitro (17,36,37) and in vivo (33), and our current data showing BrdU(+) nuclei in mitosis (Fig. 4G) and colabeling with pHH3 (Supplementary Fig. 1). We have not formally tested for β-cell endoreduplication, but, similar to prior observations (38), we did not detect systematically increased nuclear size in double-labeled cells. On the other hand, the fate of β-cells labeled for both BrdU and γH2AX remains unknown. Although getting β-cells to traverse the G1-S transition is an important and challenging goal, studies are also needed to find ways to increase the frequency of successful mitosis in β-cells that do enter the cell cycle. Taken together, the results of the current studies support the use of BrdU labeling as a faithful measure of S-phase entry and a useful tool in the overarching goal of the pursuit of β-cell regeneration strategies as a therapy for diabetes.

Acknowledgments. The authors thank the Beta Cell Biology Group at the University of Massachusetts Medical School for helpful discussions.

Funding. Human pancreatic islets were provided by the IIDP at City of Hope, which was funded by the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK). This work was supported by National Institutes of Health/NIDDK grants R01-DK-114686 (L.C.A.), R01-DK-113300 (L.C.A.), R01-DK-095140 (L.C.A.), U2C-D-093000 (National Mouse Metabolic Phenotyping Center Analytical and Functional Core; L.C.A.), and 2UC4-DK-098085 (IIDP); American Diabetes Association grant 1-18-IBS-233 (L.C.A.) in collaboration with the Order of the Amaranth; and the George F. and Sybil H. Fuller Foundation.

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

Author Contributions. R.B.S. performed and analyzed the experiments, with assistance from C.D. and X.Z. (immunostaining and microscopy) as well as B.G. (unbiased image analysis). R.B.S. and L.C.A. devised and planned the project. L.C.A. wrote the manuscript. All authors had the opportunity to edit and approve of the manuscript. L.C.A. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the 77th Scientific Sessions of the American Diabetes Association, San Diego, CA, 9–13 June 2017.