Cell Cycle Regulation of the Pdx1 Transcription Factor in Developing Pancreas and Insulin-Producing β-Cells

Understanding the expansion and function of insulin-producing pancreatic β-cells is of paramount relevance to therapeutic efforts to address all forms of diabetes. Adult β-cell mass expansion occurs primarily via proliferation of existing β-cells (1). In many cell types, proliferation and differentiation are mutually exclusive states, with terminally differentiated cells acquiring their specific functional states after exiting the cell cycle (2). Therefore, cell proliferation often needs to be coordinated with cell differentiation and function. Previous studies have demonstrated that >1,000 cell cycle–specific genes are not expressed in quiescent cells (3). β-cells induced to proliferate by c-Myc upregulation (4) or partial pancreatectomy (5) show decreased expression of many β-cell lineage identity genes, including Pdx1, although β-cell identity genes may be longer lived, and some can still be detected in proliferating cells (6,7). Likewise, gene expression analyses of sorted proliferating β-cells from transgenic mice expressing GFP in cycling cells revealed decreased expression of genes involved in β-cell function (8). How β-cells coordinate cell proliferation and differentiation is largely unknown. The transcription factor Pdx1 is essential for both β-cell proliferation and function (9–16). Thus, regulation of Pdx1 activity might be important to coordinate cell proliferation and differentiation. However, whether Pdx1 is dynamically regulated during mitosis has not been characterized.

In nondividing cells, nuclei compartmentalize into different functional domains (17). Once cells enter cell division, both nuclear and cytoplasmic architectures undergo substantial reorganization, with a large number of proteins showing cell cycle–dependent regulation of localization and/or protein level (18). It has been shown that some transcription factors remain associated with the condensed chromosomes as a means of bookmarking to transmit gene expression programs through mitosis, whereas other transcription factors dissociate from the chromosomes (19–21). In this study, we thoroughly examined the regulation of Pdx1 using three systems: the glucose-responsive β-cell line INS-1 832/13 (hereafter referred to as INS-1) (22), maternal mouse islets during pregnancy, and embryonic pancreas at embryonic day (E) 11.5. Our results demonstrate that Pdx1 protein localization and levels are tightly regulated during the cell cycle and that only cells with relatively optimal Pdx1 levels can progress through S phase and enter G2/M.

All experiments were approved by the Institutional Animal Care and Use Committee of Vanderbilt University Medical Center. All mice were on a mixed genetic background and between ages 3 and 6 months. Mouse pregnancies were timed by checking for vaginal plugs. The day of the vaginal plug was conserved gestational day (GD) 0.5. Virgin females or pregnant females at GD14.5 were euthanized, and 1 mg/mL collagenase P was injected into the pancreas through the common bile duct. Pancreata were digested for 15 min. Islets were then manually picked under a dissection microscope. The purified islets were cultured at 37°C overnight in RPMI containing 11 mmol/L glucose and 1% penicillin-streptomycin. To examine the multipotent pancreatic progenitor cells in E 11.5 embryos, posterior foregut tissues including the pancreatic buds and digestive organs were isolated.

mCherry was amplified from pmCherry-C1 (Clontech). The Mus musculus Pdx1 cDNA (NM_008814.3) was used as the template. NEBuilder HiFi DNA assembly was used to clone the mCherry-Pdx1 into the previously modified pSKII vector (23), using the following primers: forward, GTGCTGTCTCATCATTTTGGCAAAGAATTCATGGTGAGCAAGGGCGAGGAG; and reverse, CAGCCTGCACCTGAGGAGTGAATTCTCACCGGGGTTCCTGCGGTCGCA. The integrity of the mCherry-Pdx1 sequence was checked by sequencing. In this vector, Pdx1 expression is driven by the Hsp minimal promoter and the 1-kb PB islet enhancer from the Pdx1 upstream regulatory sequences (23). To drive mCherry localization specifically to the nucleus, the cDNA for nuclear localization signal (NLS)–mCherry was amplified from NLS-mCherry (#108881; Addgene) and subcloned into the same pSKII vector. mCherry-Pdx1 driven by the cytomegalovirus promoter was used to express Pdx1 in RPE1 cells. For this construct, the same mCherry-Pdx1 cDNA was subcloned into the pcDNA3 vector.

Human retinal pigment epithelial (RPE1) cells were cultured with DMEM/F12K supplemented with 10% heat-inactivated FBS and 1% penicillin-streptomycin. INS-1 832/13 cells were cultured as described (22). Briefly, cells were maintained in RPMI 1640 medium supplemented with 10% heat-inactivated FBS, 10 mmol/L HEPES, 2 mmol/L L-glutamine, 1 mmol/L sodium-pyruvate, 1% penicillin-streptomycin, and 0.05 mmol/L 2-mercaptoethanol. Lipofectamine 3000 was used to transfect INS-1 cells following the manufacturer’s instructions. Fugene 6 (Promega) was used to transfect RPE1 cells following the manufacturer’s instructions. To reduce levels of methylated histone 3, INS-1 cells were incubated with 500 nm GSK126 in DMSO (CayMan Chemical) for 72 h. Control cells were treated with DMSO.

INS-1 cells were synchronized using a modified published protocol (24). Briefly, cells were starved in medium with 0.1% FBS for 3 days and then cultured in 10% FBS with 2 μg/mL aphidicolin for 12 h to arrest cells at the G1/S boundary. Cells were then released into medium with 10% FBS to allow cell cycle progression. For BrdU labeling, cells were released into the growth medium containing 10 μmol/L BrdU. Cells were harvested at 0, 4, and 12 h.

Tissues were fixed in 4% paraformaldehyde overnight at 4°C and then incubated 48 h at 4°C in PBS supplemented with 0.2% Triton X-100 (Sigma), 0.1% Tween (Sigma), 1% BSA (Sigma), and 5% donkey serum (Jackson ImmunoResearch). Samples were then labeled with guinea pig anti-Pdx1 (1:500; a gift from Dr. Christopher V.E. Wright, Vanderbilt University) and rabbit anti-Ki67 (1:400; Abcam). The corresponding secondary antibodies were raised in donkey and conjugated with Cy3 or Cy5. All antibodies were diluted in blocking buffer and incubated for 48 h at 4°C. To label DNA and/or F-actin, Alexa Fluor 488 phalloidin (1:1,000; Invitrogen) and/or 1 μg/mL DAPI (Thermo Scientific) was mixed together with the secondary antibody. After labeling, islets were mounted with ProLong Gold antifade mounting medium (Invitrogen) and No. 1.5 coverslips. Embryonic tissues were dehydrated in an increasing methanol series and then cleared with benzoic acid/benzyl benzoate (Sigma). Before imaging, embryonic tissues were mounted in benzoic acid/benzyl benzoate using iSpacer (Sunjin Lab) together with No. 1.5 coverslips.

Cells were grown on No. 1.5 coverslips coated with 10 μg/mL collagen IV (Corning Life Sciences) for at least 24 h before the experiments. Cells were then fixed in 4% paraformaldehyde supplemented with 0.2% Triton X-100 at room temperature for 10 min. Fixed cells were blocked with PBS supplement with 1% BSA (Sigma) and 5% donkey serum (Jackson ImmunoResearch) at room temperature for 1 h and then immunolabeled with goat anti-Pdx1 (1:5,000), rabbit anti-Ki67 (1:400), mouse anti-Aurora B (1:100; #3094S; Cell Signaling), mouse anti-BrdU (1:100; #G3G4; Developmental Studies Hybridoma Bank), and rabbit anti–histone H3 lysine 27 (H3K27me3) (1:100; #AB2561020; Active Motif) overnight at 4°C. Cells were then incubated with the corresponding secondary antibodies conjugated with different fluorophore for 1 h. Prolong Gold was used to mount the coverslips. To label BrdU, cells were treated with 1.5 N hydrochloric acid for 40 min immediately before blocking. The ApoAlert DNA Fragmentation Assay Kit (#630107; Clontech) was used to detect the dying cells. Briefly, INS-1 cells were cultured for 72 h after transfection. TUNEL labeling was performed following the manufacturer’s instruction.

To measure cell cycle progression by fluorescence activated cell sorting (FACS), cells were harvested and fixed in 70% ethanol overnight at 4°C and then labeled with 1 μg/mL DAPI overnight at 4°C. To analyze Pdx1 intensity by FACS, INS-1 cells were harvested and fixed for 30 min with BD Fixation/Permeabilization buffer (#5547114; BD Biosciences). After washing, cells were permeabilized with BD Permeabilization Buffer Plus (#561651; BD Biosciences) for 10 min on ice and refixed with BD Fixation/Permeabilization buffer. Fixed cells were then immunolabeled with goat anti-Pdx1 (1:5,000), rabbit anti-MafA (1:1,000; #79737; Cell Signaling), rabbit anti-FoxM1 (1:500; #sc-500; Santa Cruz Biotechnology), and rabbit anti-NKX6.1 (1:100; #NBP1–49672; Novus Biologicals) overnight at 4°C and Cy3 donkey anti-goat or anti-rabbit secondary antibody (Jackson ImmunoResearch Laboratories) for 1 h at 4°C. Cells were then labeled with 1 μg/mL DAPI. All reagents were diluted with BD Perm/Wash buffer (#5547114; BD Biosciences).

INS-1 cells expressing mCherry-Pdx1 were first sorted into the culture medium by FACS and then harvested and lysed with radioimmunoprecipitation assay buffer. To extract nuclear and cytoplasmic proteins, the NE-PER Nuclear and Cytoplasmic Extraction Kit (#78833; Thermo Scientific) was used according to the manufacturer’s instructions. Proteins were separated by SDS-PAGE and immunoblotted with the following antisera: PDX1 (D59H3; Cell Signaling Technology), GAPDH (D16H11; Cell Signaling Technology), PARP (9542; Cell Signaling Technology), and cyclin E (sc-481; Santa Cruz Biotechnology). Ponceau S (P3504; Sigma) was used to stain for total protein. Bio-Rad Image Lab software was used for quantifications.

Both confocal and Airyscan images were obtained using a 60× NA1.4 lens with a Zeiss 880 imaging scope. Airyscan images were reconstructed using the built-in function in Zeiss Zen microscope software. For all images, brightness, contrast, and γ settings for each fluorescence channel were adjusted to make small structural features visible. To analyze Pdx1 intensity in INS-1 cells, nuclei were first detected based on DAPI labeling. Pdx1 intensity was then quantified from each nucleus. All intensity values were normalized between the range [0,1] using a customized function in Python. To quantify Pdx1 intensity in relation to H3K27me3 in INS-1 cells, H3K27me3⁺ areas in each cell were detected using the Image J particle analysis function. Size and Pdx1 intensity in each positive area were measured directly in Image J. For each cell, H3K27me3⁻ areas were created by removing the positive areas from the entire nuclear region. The average Pdx1 intensity in the H3K27me3⁻ areas was then measured. The intensity ratio of Pdx1 at each H3K27me3⁺ area to the negative area of the same cell was then calculated and used to generate the plots in Fig. 2D3 and 4. All plots were created in Python.

Data are presented as means ± SDs. For continuous variables, two-tailed Student t tests were then performed. Normality was based on the central limit theorem. For categorical data, χ² test was used. For Fig. 6, Wilcoxon-Mann-Whitney tests were used to compare each cell cycle stage using G2 as a baseline. The significance level after Bonferroni correction was P = 0.0125. All statistical analyses were performed using Python.

All data generated or analyzed during this study are included in the published article (and its Supplementary Material). Novel plasmid constructs generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.

To assess a large number of cells at each stage to determine the consistency of our observations, we first examined Pdx1 localization in INS-1 cells by high-resolution immunofluorescence. In addition to using chromosome configuration as an indicator, we also labeled cells with two other markers, Ki67 and Aurora B, to precisely determine cell-cycle stage. Localization and expression of these two proteins are tightly regulated by cell cycle progression. Ki67 is a widely used marker to identify proliferating cells and can be observed at all stages of the cell cycle except G0. At G1, Ki67 is at its lowest expression level. As cells progress through the cell cycle, Ki67 levels increase and peak at G2/M (25). In interphase cells, Ki67 is mainly localized to the nucleoli. After chromosome condensation, Ki67 coats the chromosomes, helping organize the chromosome periphery compartment (26). Aurora B protein is first detected during S phase and peaks at G2/M (27,28). At mitotic entry, Aurora B localizes along the chromosomes and later to the centromeres. At anaphase, Aurora B is localized to the spindle midzone between the separated chromosomes (28). By labeling INS-1 cells simultaneously for Aurora B, Ki67, DAPI, and Pdx1, we were able to dissect Pdx1 localization and protein level at different cell cycle stages.

Using high-resolution imaging, we found that in interphase INS-1 cells, Pdx1 was nonuniformly distributed throughout the nuclei. Pdx1 was excluded from the nucleoli, which were Ki67 rich (Fig. 1A, white arrow). Pdx1 was also absent from the nuclear periphery within areas of constitutive heterochromatin (Fig. 1A, zoomed images in white box), which could be identified by the strong DAPI signal (29). These findings are not surprising. Constitutive heterochromatin is generally transcriptionally inert (30,31), and the main function for the nucleolus is ribosome biogenesis (32). The lack of Pdx1 protein in these two regions is consistent with its known roles as a transcription factor.

The nonuniform distribution of Pdx1 likely reflects its biological functions. Although Pdx1 is best known for its role as an activator of β-cell gene expression (33–35), it also represses a subset of α-cell genes (36). Therefore, Pdx1 should localize to some regions of inactive/repressed promoters. To test this hypothesis, we labeled INS-1 cells for trimethylation of H3K27me3, a marker associated with transcriptional repression (37). As shown in Fig. 2A and D1, a majority of H3K27me3⁺ regions, in particular the largest areas, had low Pdx1 levels, whereas a few H3K27me3⁺ regions, mainly the smaller regions, showed high Pdx1 levels (Fig. 2A and D2–D4). H3K27me3 is maintained by enhancer of zeste gomolog2 (EZH2), a histone methyltransferase (38). The reduction in H3K27me3 signal observed with the EZH2 inhibitor GSK126, which reduces levels of methylated histone 3 (39), confirms the specificity of the H3K27me3 signal (Fig. 2B and C). We predict those smaller regions in which Pdx1 and H3K27me3 colocalize to be regions of active Pdx1 repression of α-cell–specific genes, but this will require further analysis. These Pdx1⁺/H3K27me3⁺ regions may also represent promoters that are bivalently marked, allowing for rapid regulation of gene expression.

We next analyzed Pdx1 localization in mitotic INS-1 cells. In prophase, when chromosomes have begun to condense (DAPI) (Fig. 1B) but the nuclear envelope is still intact, Pdx1 was confined to the nucleus and excluded from the DNA (Fig. 1B, arrows). In contrast, at this stage, both Aurora B and Ki67 were associated with the DNA, as indicated by immunolabeling (Fig. 1B) and intensity line scan analysis (Fig. 1F1 and G1). At early prometaphase (Fig. 1C), when the nuclear envelope has broken down, Pdx1 protein could be detected in the cytoplasm (compare DAPI and Pdx1 in Fig. 1C). At this stage, the chromosomes are organized in a circular configuration, and Pdx1 was enriched at the regions adjacent to the condensed chromosomes. By late prometaphase (Fig. 1D), Pdx1 was uniformly distributed in the cytoplasm, and there was no longer any obvious enrichment of Pdx1 surrounding the chromosomes (Fig. 1F2 and G2). Interestingly, in late anaphase/telophase, when chromosomes have started to decondense but the nuclear envelope is not yet reformed, Pdx1 was again enriched in the vicinity of the chromosomes (Fig. 1E).

In contrast to islet β-cells in vivo, INS-1 cells grow as a monolayer and do not have the three-dimensional (3D) architecture or microenvironment found in islets. In addition, they divide quickly with a reduced G0/G1 population (40). To understand whether the Pdx1 localization pattern we observed in INS-1 cells also occurs in vivo, we examined Pdx1 localization in islets from female mice during pregnancy, a time when β-cell proliferation increases in response to gestational insulin resistance (41). Chromosome condensation and spindle alignment make it difficult to identify mitotic cells in sectioned tissues. Therefore, whole islets were isolated at GD14.5, the peak of maternal β-cell proliferation, immunolabeled, and examined with high-resolution 3D imaging using Airyscan.

Similar to what we described in INS-1 cells at interphase, the distribution of Pdx1 in the nuclei of GD14.5 islets was not uniform in G0 and interphase cells. Pdx1 was not observed at peripheral constitutive heterochromatin (Fig. 3A, globular DAPI-labeled structures). In interphase, Pdx1 was excluded from the nucleoli (Fig. 3A, Ki67⁺ region). In prophase, Ki67 associated tightly with the chromosomes, whereas Pdx1 was localized around the chromosomes externally to Ki67 protein labeling (Fig. 3B). Pdx1 was found diffusely in the cytoplasm in prometaphase cells (Fig. 3C). The fact that the Pdx1 localization pattern is similar in proliferating INS-1 cells and GD14.5 islets suggests a mechanistic link between cell cycle stage and Pdx1 localization. Furthermore, these results suggest that changes in Pdx1 localization during cell division are β-cell intrinsic and are not dependent on the 3D islet structure or specific aspects of the islet microenvironment.

To determine whether the observed dynamic Pdx1 localization pattern is specific to β-cells or a common feature of Pdx1 during the cell cycle, we next analyzed Pdx1 localization in multipotent progenitors (MPCs) in developing pancreatic buds at E 11.5. To identify the mitotic cells, we used high-resolution 3D imaging after optical clearing and whole-mount immunolabeling of dissected embryonic posterior foregut organs. To locate the cells of interest, we first identified the pancreatic buds under low-magnification imaging (Fig. 4A1) and then switched to 3D high-resolution imaging (Fig. 4A2–4). We found that in E 11.5 MPCs, the subnuclear localization of Pdx1 was also cell cycle dependent. Pdx1 was excluded from the nucleoli and constitutive heterochromatin (Fig. 4B and C, Ki67⁻ cells). We were unable to identify cells in prophase; however, in some prometaphase cells, we could observe Pdx1 enrichment around the chromosomes, similar to our observations in INS-1 cells and GD14.5 islets (Fig. 4D). We also observed the enrichment of Pdx1 around the chromosomes again at telophase (Fig. 4E).

In both yeast and human cells, >10% of proteins show cell cycle–dependent changes in expression level (18,42). Therefore, we next analyzed whether Pdx1 protein levels are also regulated during cell cycle progression. Both Aurora B and Ki67 intensities increase throughout interphase and peak at G2/M (25,27,28). We therefore quantified Ki67 and Aurora B intensity, along with Pdx1, in immunolabeled asynchronously dividing INS-1 cells from all stages of the cell cycle (Fig. 5A). The cell cycle stage was inferred based on the intensities of each of these two proteins. In Fig. 5B, a clear positive correlation trajectory of increasing Ki67 and Aurora B can be identified (Pearson correlation coefficient r = 0.654; P = 4.5e−55). Such a trajectory represents pseudo-time progression toward G2/M. Interestingly, among cells expressing the highest levels of Pdx1, only 19.6% progressed toward G2/M. In contrast, 46.3% of cells expressing low levels of Pdx1 progressed toward G2 (n = 441; χ² test P = 1.04e−7) (Fig. 5B), indicating that cells with the highest levels of Pdx1 are less likely to progress toward G2/M. Initially, this result seems to be in conflict with previous studies from our group and others showing that Pdx1 is required for β-cell proliferation (10,43). However, as shown in Fig. 5C1 and 2, as cells progressed toward G2/M, their Pdx1 levels converged to an optimal middle range. In these graphs, increasing Aurora B or Ki67 levels represent the pseudo-time progression of cells toward G2/M. Cells with extremely low levels of Pdx1 also showed low levels of Aurora B or Ki67. In fact, among cells with Pdx1 levels in the lowest 10th percentile, no cells had an Aurora B level greater than the half maximum level, supporting the concept that Pdx1 is required for β-cell proliferation. Thus, consistent with previous reports (10,43,44), cells with very low Pdx1 did not progress through G2. We also analyzed Pdx1 intensity at different stages during mitosis and saw that mean Pdx1 intensity was maintained at this optimal level until prometaphase/metaphase (Supplementary Fig. 1), with potentially a slight decrease in Pdx1 levels at anaphase. However, it is difficult to draw a firm conclusion on this because of the dramatic change in cell morphology and Pdx1 redistribution during cell cycle progression.

To further examine whether Pdx1 protein levels change specifically at S or G2 phase, we used FACS to analyze Pdx1 levels during the cell cycle (Fig. 5E). Consistent with our immunolabeling results, we observed that when cells were in S phase, Pdx1 levels converged within a tighter optimal range. Interestingly, Pdx1 levels were more divergent in G1 and again at G2/M. To determine whether this phenomenon is unique to Pdx1, we also used FACS on INS-1 cells labeled with antibodies against other transcription factors, including Nkx6.1, required for β-cell identity and proliferation (45), the essential regulator of adult β-cell proliferation, FoxM1 (46,47), and the β-cell identity transcription factor MafA (48,49). Cells were sorted at different cell cycle stages based on DNA content. Both Nkx6.1 and FoxM1 showed similar patterns to Pdx1 in that cells in S phase had a narrower range of protein level, whereas cells in G1 (100K) and G2 (200K) showed a broader range of protein expression (Supplementary Fig. 2). The data are less clear for MafA but follow a similar trend.

Together, our data support the conclusion that only cells with optimal levels of Pdx1 protein can enter the cell cycle and progress toward G2/M. It is unlikely that in the continuously dividing clonal INS-1 cell model, low levels of Pdx1 expression represent β-cell heterogeneity. Instead, we hypothesize that the levels of Pdx1 protein are actively regulated throughout the cell cycle. To determine whether cell cycle arrest affects Pdx1 protein levels, we used serum starvation to arrest INS-1 cells. INS-1 cells were cultured in 0.1% FBS for 72 h, and Pdx1 protein levels were then examined by Western blot (Fig. 5F). We did not observe any statistically significant changes in Pdx1 levels or in the ratio of cytoplasmic to nuclear Pdx1 content after serum starvation, although there was a trend toward higher levels of total Pdx1 protein in cells grown under normal serum conditions.

After cell cycle arrest by serum starvation, INS-1 cells were cultured in 10% FBS and aphidicolin to synchronize cells at the G1/S boundary (Supplementary Fig. 3B). Cells were then released into regular growth medium. Four hours after release from aphidicolin, most cells expressed Aurora B at different levels (Supplementary Fig. 3C), indicating the cells were in S phase. Twelve hours after release of cell-cycle arrest, most of the cells expressed Aurora B at high levels. Some cells also exhibited typical mitotic chromosome configurations (Supplementary Fig. 3D, arrow), indicating they were at G2/M. Interestingly, cells with high Pdx1 levels were Aurora B⁻ even after 12 h of recovery from cell cycle arrest (Supplementary Fig. 3D, white boxes), indicating slowing down of the cell cycle. To confirm this finding, we released cells into growth medium containing BrdU. Consistently, we found that at the individual cell level, cells with higher Pdx1 levels were BrdU⁻ even after 12 h of recovery (Fig. 5G). Western blotting failed to show any differences in average Pdx1 protein levels across the entire population of cells (Fig. 5F). Therefore, we conclude that high levels of Pdx1 block individual cells from progressing through S phase.

To determine whether Pdx1 overexpression is sufficient to block cell cycle progression, we expressed wild-type mCherry-tagged full-length Pdx1 protein in INS-1 cells and examined whether cell cycle progression was affected by an increase in Pdx1 level. For this purpose, mCherry-Pdx1 was subcloned into a modified PSKII vector, where expression of Pdx1 is driven by an Hsp minimal promoter and regulated by the 1-kb islet enhancer region from the Pdx1 promoter (23). As Fig. 6A shows, mCherry-Pdx1 localized specifically to cell nuclei and was recognized by the Pdx1 antibody. Western blotting confirmed that mCherry-Pdx1 was expressed at a level comparable to endogenous Pdx1 (Fig. 6B), thus avoiding nonspecific effects resulting from protein overexpression.

Only 2.0% of INS-1–expressing mCherry-Pdx1 was mitotic (n = 410 cells). In contrast, 7.8% of cells expressing the empty mCherry vector were mitotic (n = 399 cells; χ² test P = 0.0006), suggesting that elevations in Pdx1 block mitotic entry. To understand whether overexpressing Pdx1 reduces the number of cells in S and G2, we immunolabeled INS-1 cells overexpressing mCherry-Pdx1 for Aurora B and then quantified Aurora B intensity in interphase cells (Fig. 6C). In cells expressing only mCherry, we were able to identify cells with different levels of Aurora B, indicating there are cells at different stages of interphase; 38.1% of these cells had Aurora B levels above threshold (Fig. 6C, horizontal line indicating threshold for Aurora B level). However, in cells overexpressing mCherry-Pdx1, we did not observe cells with high Aurora B intensity, and only 17.4% of cells expressed Aurora B at levels above threshold (χ² test P = 0.0002; n = 195 for mCherry-expressing cells; n = 121 for mCherry-Pdx1–expressing cells), suggesting a reduction in the number of cells progressing to S and G2.

As an alternative, to assess whether Pdx1 overexpression inhibits cell cycle progression, we analyzed DNA content in INS-1 cells expressing mCherry-Pdx1 using FACS (Fig. 6D1–6). Compared with control cells expressing NLS-mCherry with 69.8% cells at G1 and 18.9% cells at S, 77.6% of cells expressing mCherry-Pdx1 were at G1, and only 10.6% cells were at S (Fig. 6D1 and D2). The slowing down of cell cycle progression was more prominent in synchronized INS-1 cells (Fig. 6D3–6) 4 h after releasing cells into growth medium, where 77.8% of cells expressing NLS-mCherry were at S phase. However, in cells overexpressing mCherry-Pdx1, only 52.0% of cells were at S phase. Consistent with the slowing down at S phase, 12 h after releasing cells, only 36.5% of cells expressing NLS-mCherry were at G1. In contrast, 47.5% of cells expressing mCherry-Pdx1 were at G1. To determine whether the sensitivity of the cell cycle to Pdx1 levels was specific to β-cells, we also overexpressed Pdx1 in a human pancreatic duct cell line (HPDE6) (50) and found that Pdx1 overexpression resulted in decreased cell numbers (data not shown). We also overexpressed Pdx1 in RPE1 cells and found that Pdx1 overexpression in RPE1 cells actually stimulated progression to G2/M phase (Supplementary Fig. 4), in contrast to its effect in β-cells. Therefore, we conclude that the effects of Pdx1 overexpression are pancreas or β-cell specific but are not generalizable to all cell types.

One possible outcome of cell cycle arrest is apoptosis. TUNEL labeling was used to assay for apoptosis after increased expression of Pdx1. We found that 6.78% of cells expressing mCherry-Pdx1 had an aberrant nuclear morphology, and these were TUNEL⁺, suggesting ongoing apoptosis (Fig. 7B). In comparison, only 0.87% of nuclei in control cells expressing NLS-mCherry alone were TUNEL⁺ (Fig. 7A) (Fisher exact test P = 0.0007; n = 329 for control; n = 251 for mCherry-Pdx1 cells). These observations lead us to suggest that the increase in cell death in Pdx1-overexpressing cells is likely due to the failure of cell cycle progression. Interestingly, previous research indicates that Pdx1 overexpression increased cell proliferation without triggering DNA damage by γH2AX labeling (43,51). Consistent with such observations, when we labeled Pdx1-expressing cells with γH2AX, we noticed cells with highly condensed nuclei were often γH2AX⁻ (data not shown), but all of the condensed nuclei were TUNEL⁺. Such results are not only consistent with previous reports (43,51) but also suggest that the cell cycle arrest that occurs with Pdx1 overexpression might not involve DNA double-strand breaks.

In this study, we examined the cell cycle regulation of Pdx1 in immortalized β-cells, adult β-cells, and MPCs and found that Pdx1 exhibits a dynamic cell cycle–regulated protein localization pattern that is consistent across all three systems. In nondividing cells, Pdx1 distributes nonuniformly in the nuclei. We think such a localization pattern reflects compartmentalization of Pdx1 target genes. High-resolution microscopy allowed us to discern that Pdx1 protein excluded from the chromosomes and enriched at the vicinity of the chromosomes at prophase, when chromosomes start to condense, and also at telophase, when chromosomes start to decondense. The fact that such enrichment persists after nuclear envelope breakdown at early prometaphase and reappears before the reformation of the nuclear envelope at the end of cell division suggests that such a localization pattern of Pdx1 is not simply due to physical confinement of Pdx1 by the nuclear membrane. We consider two possible mechanisms underlying the dynamic regulation of Pdx1 localization. First, Pdx1 might associate with cytoskeletal structures that are involved in chromatin dynamics (52–54). Further study is required to understand whether Pdx1 is associated with the cytoskeleton, either directly or indirectly resulting in enrichment around the chromosomes.

Alternatively, Pdx1 localization might be regulated through phase separation. Recent studies reveal that membrane-less structures of different morphology can be formed within cells and nuclei through phase separation (55,56). Ki67 organizes the chromosome periphery compartment through phase separation during mitosis (26,57). A recent study found that the Pdx1 C-terminus contains an intrinsically disordered protein region (IDPR) (58). In general, IDPRs lack a fixed or ordered protein structure and play key roles in modulating flexible protein-protein interactions, protein localization, and complex formation by promoting phase separation (59). Thus, with its C-terminal IDPR, Pdx1 might localize to different subnuclear structures at different cell cycle stages through phase separation.

The alteration of Pdx1 localization pattern coincided with chromosome condensation and decondensation, suggesting that Pdx1 is important for chromosome conformational change. Pdx1 may also be involved in bookmarking certain chromosomal regions. It remains to be determined how this transient localization pattern is regulated and what its function might be in regulating Pdx1 target gene expression. Pdx1 subnuclear localization pattern might be important for gene expression in interphase and be involved in the process of chromosome condensation/decondensation during mitosis.

Another interesting finding in the current study is that cell cycle–dependent regulation of Pdx1 protein levels is required for cells to progress through S/G2, specifically in INS-1 cells. To ensure the successful completion of the cell cycle, multiple cell cycle checkpoints are used to monitor the cell cycle progression. At the G1/S checkpoint, DNA damage is monitored. Failure of checkpoints can lead to apoptosis (60). In our study, we observed that cells with high Pdx1 levels did not show proper chromosome duplication. Cells with high Pdx1 also had a low mitotic index and enhanced apoptosis. Therefore, we propose that high levels of Pdx1 impede the proper reorganization of chromosomes during this process, which might trigger the activation of G1/S checkpoints. Detailed analysis is required to understand how Pdx1 affects cell cycle checkpoints. Our data suggest that this is a common feature of transcription factors involved in β-cell proliferation, because Nkx6.1 and FoxM1 showed similar dynamic levels of protein expression during cell cycle stages.

It is unknown how Pdx1 levels are regulated during the cell cycle. Previous research has identified an evolutionarily conserved motif within the C-terminus (aa224–238) that mediates interaction with speckle-type BTB/POZ protein (SPOP/PCIF1), a substrate adaptor for the E3 ubiquitin ligase cullin 3 (CUL3), to promote proteasomal degradation of Pdx1 (61–63). Future studies will determine whether SPOP mediates protein degradation and plays a role in the dynamic regulation of Pdx1 during β-cell proliferation.

Taken together, we demonstrate here the dynamic regulation of Pdx1 localization and protein levels during the cell cycle. To successfully progress through the cell cycle, Pdx1 levels must be maintained at an optimal range. Further research is necessary to address the molecular mechanisms behind such dynamic regulation and uncover the biological function of Pdx1 during mitosis.

Acknowledgments. The authors would like to thank members of the Gannon laboratory for helpful discussions and reading of the manuscript. The authors also thank Jennifer Dunn from the Gannon laboratory for technical assistance.

Funding. Confocal and high-resolution microscopy was performed in the Vanderbilt University Cell Imaging Shared Resource, which is supported in part by the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) (DK20593). Flow cytometry experiments were performed in the Flow Cytometry Shared Resource, which is supported by the Vanderbilt Ingram Cancer Center (P30 CA068485) and the Vanderbilt Digestive Disease Research Center (DK058404). X.Z., D.A.S., and M.G. were supported by NIDDK, National Institutes of Health (NIH) (R01 DK105689). X.Z. and M.G. were also supported by a VA Merit award (1 I01 BX003744-01) to M.G. M.A.G. was supported by the NIDDK, NIH (F31 DK122761). S.A.S. was supported by the JDRF (CDA-2016-189, SRA-2018-539, and COE-2019-861), NIDDK, NIH (R01 DK108921), and Department of Veterans Affairs (I01 BX004444).

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

Author Contributions. X.Z. conceived and performed experiments, analyzed data and generated figures, and wrote the manuscript. A.O. performed experiments and edited the manuscript. M.A.G. generated unique reagents used in the experiments and read the manuscript. S.A.S. helped with experimental design and data interpretation, provided unique reagents used in the experiments, and edited the manuscript. D.A.S. helped with experimental design and data interpretation and edited the manuscript. M.G. helped with experimental design and data analysis and edited the manuscript. M.G. 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.