Haploinsufficiency of Insm1 Impairs Postnatal Baseline β-Cell Mass Free

Pancreatic β-cell mass is regulated both during development and in the adult. Lineage tracing in animals indicates that embryonic β-cells mainly differentiate from pancreatic progenitor cells (1), whereas postnatal increases in β-cell mass arise through self-renewal (2,3). β-Cell replication slows considerably in adults, although variations in insulin demand can lead to adaptive changes in β-cell mass (4). Sufficient β-cell mass is essential for normal regulation of blood glucose levels. Loss of β-cell mass by an immune attack or metabolic stressors results in type 1 and type 2 diabetes, respectively. In addition, although β-cell mass varies in individuals, low β-cell mass is a risk factor for prediabetes and diabetes (5). At least two factors contribute to total β-cell mass: replication capacity and baseline β-cell mass. The replication rate of adult β-cells is low, and massive efforts have been made to restore diabetic β-cell loss by enhancing β-cell replication. In contrast, the establishment of the baseline β-cell mass is not well investigated, and it is not yet fully understood how postnatal β-cell expansion varies in different individuals.

The baseline β-cell mass is established in the early postnatal period in both mice and humans (2,6–8). Recent identified factors that modulate the postnatal β-cell expansion regulate the metabolic pathways or the cell cycle (9–11). The cyclin genes Ccnd1 and Ccnd2 are essential for postnatal β-cell growth and regulate the progression through G₁ through interaction with Cdk4 (12–14). Heterozygous mutations in Ccnd1 combined with complete knockout of Ccnd2 have dose-dependent effects on β-cell mass (14). Cell cycle inhibitors are another group of essential regulators in the postnatal β-cell mass expansion. p16^INK4a is a β-cell replication inhibitor specifically expressed in aging β-cells. Therefore, although mutation of p16^INK4a has no effect on the early postnatal β-cell mass, overexpression of p16^INK4a decreases postnatal β-cell mass (15). p21, p27, and p57 regulate the cell cycle in embryonic β-cells (16,17) as well as postnatal pancreatic β-cell proliferation (18).

Insm1 encodes a zinc finger protein that is essential for the development and function of mature β-cells. Null mutation of Insm1 interferes with the formation of insulin-positive β-cells (19,20), whereas deletion of Insm1 in adult β-cells leads to loss of mature β-cell function (21). In this study, we show that Insm1 haploinsufficiency impairs the early postnatal β-cell expansion, resulting in decreased baseline β-cell mass and impaired glucose tolerance. Mechanistically, the decreased dosage of Insm1 prolongs the cell cycle in part by targeting Ccnd1. Insm1 binds to the Ccnd1 locus with low affinity in β-cells, and binding is dramatically lost upon downregulation of Insm1. Our data demonstrate that Insm1 regulates postnatal baseline β-cell mass in a dose-dependent manner, indicating that decreased Insm1 expression is a potential risk factor for diabetes.

The Insm1^lacZ allele was reported previously (19). Male animals were used for all experiments. Wild-type littermates were used as controls. A high-fat diet (HFD) treatment was started at 3 months of age, and glucose tolerance and the β-cell mass were measured after 10 months on the HFD. The ob/ob mice were analyzed at 13 months of age. All animal experiments were approved by the Institutional Animal Care and Use Committee of Jinan University.

Blood glucose and blood insulin measurements were performed as previously described (21). Briefly, blood glucose levels from each animal were determined from at least three independent measurements on different days. Blood insulin levels from each animal were determined from average values obtained from at least two independent measurements on different days. For glucose tolerance and insulin secretion tests, glucose was injected intraperitoneally (2 g/kg body weight, unless otherwise indicated). Insulin and glucagon ELISA kits were used to detect insulin and glucagon levels in the blood (90080; Crystal Chem, Downers Grove, IL; and DGCG0, Quantikine ELISA; R&D Systems, Minneapolis, MN).

β-Cell mass was determined as previously described (21). In short, each pancreas was evenly flattened, fixed in 4% paraformaldehyde for 3 h, embedded in paraffin, and then sectioned at 8-µm thickness. For β-cell mass analysis, we collected sections at every 10th section in 2-, 7-, and 14-day-old animals and every 20th section in 2-month, 13-month, HFD, and ob/ob mice; that is, ∼8–10 pancreatic sections for each animal were used. β-Cell area was identified by immunofluorescence using anti-insulin antibodies. The total area of the pancreas, insulin-positive area, and pancreas weight were used to calculate β-cell mass: β-cell mass (milligrams per pancreas) = whole pancreas weight (milligrams) × insulin-positive area/pancreas area. ImageJ software was used to quantify the areas of pancreas and β-cell.

Immunofluorescence and Western blots were performed as previously described (21). Fluorescence was imaged on a Zeiss LSM 700 confocal microscope and processed using Adobe Photoshop software. Anti-Insm1, anti-insulin, anti-Glut2, anti-BrdU, and anti-ActB antibodies were used as previously described (21), and anti-Ccnd1 antibody was purchased from Abcam (ab-134175) and Santa Cruz Biotechnology (sc-753). Anti-Ki67 antibody was purchased from Dako (M7249) and Abcam (ab15580). Anti-p16 (ab51243; Abcam), anti-p21 (ab109199; Abcam), anti-p27 (ab92741; Abcam), anti-p57 (ab75974; Abcam), anti-CcnE1 (20808S; CST), and anti-CcnA2 (ab181591; Abcam) antibodies were used for Western blotting. Secondary antibodies (Jackson ImmunoResearch Laboratories) coupled to Cy3, Cy2, Cy5 (for immunofluorescence), or horseradish peroxidase (for Western blot) were used.

Pancreatic islets were isolated as previously described (21). The assay was performed by incubating islets for 30 min in secretion buffer containing 3.3 mmol/L or 16.7 mmol/L glucose with washing and preincubation steps between and before the assay (21). Released insulin was quantified using an ELISA kit (80-INSMSH; ALPCO, Salem, NH) as previously described (21).

Freshly isolated islets or cultured SJb cells were lysed, and total RNA was isolated using TRIzol reagent (Invitrogen). For quantitative RT-PCR analysis, cDNA from each animal was synthetized using the PrimeScript RT Reagent Kit with gDNA Eraser (Takara Bio, Bejing, China) and analyzed using the SYBR Fast qPCR Mix (Takara Bio) CFX96 RT-PCR system (Bio-Rad). Expression levels were determined using the 2^−ΔΔCt method. ActB was used as internal standard, and the results were displayed as the proportion of wild-type controls. Primers used for quantitative analysis are listed in Supplementary Table 1.

For chromatin immunoprecipitation PCR (ChIP-PCR), we used isolated pancreatic islets or the SJb β-cell line. Anti-Insm1 antibody was used for ChIP-PCR as described in our previous study (21). The PCR primers used for chromatin analysis are shown in Supplementary Table 1.

siRNAs against Insm1 and Ccnd1 mRNAs were delivered by electroporation with the Amaxa Nucleofector using the Amaxa Kit V and program G-16, according to the protocols provided by the kits. Mouse Insm1 and nontargeting control siRNAs were purchased from Dharmacon (J-049233-09/-11/-12 for Insm1 and D-001810-10 for control). Mouse Ccnd1 siRNAs were purchased from Ambion (s63513 and s201129).

Alcohol-fixed SJb cells were washed with PBS and incubated with PBS containing 0.1% Triton X-100, 3 mg/mL DNase-free RNase (EN0531; Thermo Fisher Scientific), and 20 µg/µL propidium iodide (PI) (Biofroxx; neoFROXX, Einhausen, Germany) for 15 min at 37°C. Cells were directly analyzed on an LSR Fortessa analyzer (Becton Dickinson, Franklin Lakes, NJ). Cell cycle distributions were calculated using the Dean-Jett model in FlowJo 10.0.7r2. A representative result from four biological independent experiments was shown.

To investigate the effects of Insm1 gene dosage on β-cell function, we used Insm1^+/lacZ mice that harbor one intact Insm1 allele and one mutant allele in which lacZ replaces the Insm1 coding sequence (19). In Insm1^+/lacZ islets, Insm1 protein levels were lower at 2 months or 2 weeks of age (Fig. 1A and Supplementary Fig. 1). Compared with Insm1^+/+ mice, blood glucose levels were significantly higher in Insm1^+/lacZ mice on a random feeding schedule, whereas the fasting blood glucose levels were comparable (Fig. 1B). Insm1^+/lacZ mice were glucose intolerant (Fig. 1C) and displayed defects in glucose-stimulated insulin secretion (Fig. 1D). Although the blood glucagon levels were comparable in Insm1^+/lacZ and Insm1^+/+ mice in both fasted and randomly fed conditions (Fig. 1E), the proportions of glucagon-producing α-cells and other endocrine cell types were not altered (Supplementary Fig. 2). Moreover, there was no change in peripheral insulin sensitivity in Insm1^+/lacZ mice using an insulin tolerance test (Fig. 1F). We isolated islets and performed glucose-stimulated insulin secretion ex vivo. We observed comparable levels of insulin secretion in islets from both genotypes at low and high glucose concentrations (Fig. 1G). In Insm1^+/lacZ mice, we observed reduced insulin secretion in vivo but normal secretion in isolated islets, indicating insufficient β-cell mass. Indeed, we observed that the β-cell mass in Insm1^+/lacZ mice was 70.9% of that of Insm1^+/+ mice at 2 months (P2M) (Fig. 1H). This decrease in β-cell mass was not caused by differences in β-cell size, as comparable cell sizes were observed in wide-type and Insm1^+/lacZ mice (Supplementary Fig. 3).

To further dissect the time course of the deficits, we measured β-cell mass in postnatal 2-day- (P2), 7-day- (P7), 14-day- (P14), and 13-month-old (P13M) mice (Fig. 1H). We observed comparable β-cell mass between newborn (P2) Insm1^+/+ and Insm1^+/lacZ mice. A small, but not significant, decrease in β-cell mass was detected at P7 in Insm1^+/lacZ mice. Significant decreases in β-cell mass were detected at P14, when the β-cell mass of Insm1^+/lacZ mice was 77.6% of that observed in Insm1^+/+ mice. A further decrease in β-cell mass was detected at P2M in Insm1^+/lacZ mice (70.9% at P2M vs. P77.6% at P14; P = 0.0058). However, there were no further differences in β-cell mass between Insm1^+/lacZ and Insm1^+/+ mice after P2M, as β-cell mass remained ∼70.9% at P2M and 65.9% at P13M (P = 0.7798) (Fig. 1H). These data indicated that the impaired β-cell expansion occurs in the postnatal period between 7 days and 2 months.

Because we first observed significant changes in β-cell mass at P14, we used this time point to analyze in vivo experiments, unless otherwise indicated. Increased TUNEL staining in Insm1^+/lacZ β-cells was detected at both P7 and P14, but not at P2 (Fig. 2A), whereas proliferation, indicated by the presence of Ki67, was comparable (Fig. 2B). Thus, increased apoptosis contributes to the low β-cell mass in Insm1^+/lacZ mice. However, we noticed that the apoptosis rate was much lower than the rate of proliferation (i.e., 1.5 apoptotic cells per 1,000 insulin-positive cells vs. 180 proliferated cells per 1,000 insulin-positive cells at P7) (Fig. 2A and B), suggesting that apoptosis might not be a major contributor to the decrease in β-cell mass.

We investigated cell cycle length using BrdU pulse-chase combined with Ki67 staining in P14 Insm1^+/lacZ and Insm1^+/+ mice. A single pulse of BrdU was administered to mice followed by a 23-h chase period, the approximate amount of time needed to complete one cycle of cell division (22). BrdU labeling was retained by the cells within the cell cycle as well as those that had reached a postmitotic state, whereas Ki67 only labeled the cells in the cell cycle. Thus, BrdU⁺Ki67⁻ cells had exited the cell cycle, whereas BrdU⁺Ki67⁺ cells had not yet completed the cell cycle (Fig. 2C). We found significantly decreased numbers of BrdU⁺Ki67⁻ cells and increased numbers of BrdU⁺Ki67⁺ cells in Insm1^+/lacZ mice (Fig. 2D and E), indicating that cell cycle length in Insm1^+/lacZ β-cells was prolonged. This prolonged cell cycle could be repeatedly detected at 3 weeks but not at 10 weeks in Insm1^+/lacZ mice (Supplementary Fig. 4A). We further counted cell numbers in different phases of the cell cycle using PI staining and flow cytometry analysis. Knockdown of Insm1 in SJb cells, a previously described pancreatic β-cell line that retains many of the in vivo characteristics of β-cells (21), resulted in a small, but significant, increase in cells in G₀/G₁ (Fig. 3A). Thus, haploinsufficiency of Insm1 in β-cells delays the cell cycle.

Because we observed delayed cell cycle and increased apoptosis in Insm1^+/lacZ mice, we asked whether abnormalities in cell cycle caused cell death. We performed TUNEL staining followed by immunofluorescence analysis using Ki67 staining. However, we did not detect any costaining of TUNEL and Ki67 in >1,000 Ki67⁺ cells/animal at postnatal day 14 (data not shown). Although decreases in Insm1 led to a longer cell cycle, proliferating cells did not directly enter cell death in Insm1^+/lacZ islets.

The cell cycle is marked and regulated by numerous cyclin genes (23,24). We investigated the expression of cyclin, cyclin kinase, and cyclin kinase inhibitor genes. We found that of the G₁/S phase cyclins, only Ccnd1 was downregulated in P14 Insm1^+/lacZ islets (Fig. 3B). Ccnd1 was further downregulated in SJb cells treated with Insm1 siRNA and in P7 Insm1^+/lacZ islets (Supplementary Fig. 4B and C). We further verified decreases in Ccnd1 protein but not other cell cycle regulators in islets isolated from P14 Insm1^+/lacZ mice (Fig. 3C and Supplementary Fig. 5A). Because Ccnd1 expression was consistently decreased in both early and postnatal Insm1^+/lacZ islets and Insm1-deficient SJb cells, we investigated the regulation of cell cycle length by Ccnd1.

Knockdown of Insm1 by siRNA in SJb cells resulted in fewer BrdU⁺Ki67⁻ cells and an increase in BrdU⁺Ki67⁺ cells (Fig. 3D). Similar changes were observed after knockdown of Ccnd1 expression (Fig. 3D). Next, we overexpressed Ccnd1 in SJb cells treated with Insm1 siRNA. Significantly increased numbers of BrdU⁺Ki67⁻ cells and decreased numbers of Brdu⁺Ki67⁺ cells were observed, demonstrating that the downregulation of Insm1 expression can be rescued by overexpressing Ccnd1 (Fig. 3D). PI labeling and flow cytometry cell cycle analysis showed that significant increases in cell number were detected in the G₀/G₁ phase in Insm1 or Ccnd1 knockdown cells, whereas overexpression of Ccnd1 in Insm1 knockdown SJb cells rescued the increasing of G₀/G₁ phase cell numbers (Fig. 3A). In summary, Ccnd1 is a downstream target of Insm1 and controls cell cycle length.

In our previous study, we observed downregulation of many metabolic genes and transcription factors in islets after conditional ablation of Insm1 (21), leading us to investigate the expression of these genes in Insm1^+/lacZ islets. We detected downregulation of Glut2 but no change in Pcx, Hk1, Pdx1, or MafA (Fig. 4A and B and Supplementary Fig. 5B), suggesting that Insm1 target gene Glut2 is regulated in a dose-dependent manner by Insm1.

To investigate the mechanisms of Insm1-dependent regulation of Ccnd1, we analyzed Insm1 ChIP sequencing (ChIP-seq) data (21). We observed binding of Insm1 on both the promoter and intron regions of Ccnd1 (Fig. 5A). The number of reads on the Ccnd1 gene observed in the ChIP-seq data set was lower than that detected in other Insm1-binding genes, such as the Pdx1 and Hk1 genes. However, Insm1 broadly binds to Pdx1, and these binding sites belong to a cluster of enhancers called superenhancers (25,26), which is therefore high-affinity binding (Fig. 5A). To determine how levels of Insm1 influenced binding to chromatin, we performed ChIP-PCR on control and Insm1 siRNA-treated SJb cells. Knockdown of Insm1 reduced Insm1 protein levels (Fig. 5B, left), and binding of Insm1 on Hk1, Pdx1, and Glut2 decreased by 1.4-, 1.7-, and 5.7-fold, respectively. The decrease of Insm1 binding to the promoter and intron regions of Ccnd1 was more pronounced, whereas an 8.8- and 10.3-fold decrease, respectively, was observed after Insm1 knockdown (Fig. 5B, right). We further verified the decreased binding of Insm1 on the Ccnd1 and Glut2 loci in P14 Insm1^+/lacZ islets (Fig. 5C). These data indicated that the low-affinity binding sites were prone to losing Insm1 binding in response to decreases in Insm1. In summary, Insm1 binds to Ccnd1 and Glut2 with relatively low affinity, and this binding is sensitive to changes in levels of Insm1.

To investigate the effect of Insm1 haploinsufficiency on the glycemic phenotype after metabolic stress, we introduced the Insm1^+/lacZ allele into the ob/ob background. We performed glucose tolerance tests (glucose was administered at 0.5 g/kg body weight) and observed severe glucose intolerance (Fig. 6A). The glucose tolerance of Insm1^+/lacZ mice in the ob/ob background was much lower than observed in age-matched wild-type animals (Fig. 6B). Similarly, Insm1^+/lacZ mice on an HFD showed further decreases in glucose tolerance (Fig. 6C). Although the fasted blood glucose levels in Insm1^+/lacZ mice were comparable to Insm1^+/+ mice on an HFD, these levels increased in an ob/ob background (Fig. 6D). Random blood glucose levels were consistently higher in Insm1^+/lacZ mice either on an HFD or in an ob/ob background (Fig. 6E).

To investigate whether the β-cell mass of Insm1^+/lacZ mice is further decreased during metabolic stress, we measured the β-cell mass after HFD treatment or in an ob/ob background. In Insm1^+/+ animals, β-cell mass increased to 7.96 mg and 11.92 mg, whereas the mass in Insm1^+/lacZ mice was 5.22 mg and 7.56 mg in HFD-treated or ob/ob background animals, respectively (Fig. 6F). The ratio of β-cell mass (Insm1^+/lacZ vs. Insm1^+/+) was similar in animals on an HFD (65.6%) or in an ob/ob background (63.5%) compared with age-matched animals on a normal diet (65.9%; unpaired t test, P = 0.94 for HFD and P = 0.84 for ob/ob). We investigated the expression of Ccnd1 and Glut2 in islets of 13-month-old mice and found comparable expression of Ccnd1 but decreased Glut2 expression in Insm1^+/lacZ compared with Insm1^+/+ mice (Fig. 6G). We performed ChIP-PCR in the aging islets; we found comparable Insm1 binding to the Ccnd1 locus but lower binding to the Glut2 locus in Insm1^+/lacZ mice (Fig. 6H). Together, our data indicate that the Insm1 dosage is particularly important for Ccnd1 expression and the establishment of β-cell mass specifically at early postnatal stages. Thus, decreases in β-cell mass and Glut2 expression in Insm1^+/lacZ animals contribute to impaired glucose handling during metabolic stress.

The Ccnd1 gene is regulated at the transcriptional and translational levels (27–29). We found that Ccnd1 transcription is dependent on Insm1 dosage, and the protein level is also decreased in β-cells of Insm^+/lacZ mice. Thus, Insm1 can transcriptionally regulate the cell cycle by targeting Ccnd1. Our previous work elucidated that the protein complex Insm1/NeuroD1/FoxA2 binds with high affinity to pancreatic β-cell genes that play an essential role in maintaining mature β-cell function (21). In this study, we identified that Insm1 dosage mainly affects the low-affinity binding sites (i.e., the Ccnd1 locus). Thus, we elucidated a novel mechanism of Insm1 function in pancreatic β-cells.

We show in this study that the decreased baseline β-cell mass established at early postnatal stages in the Insm1^+/lacZ mice was unchanged in the adult with or without metabolic stress. Thus, at least in the Insm1^+/lacZ mice, the baseline β-cell mass established in the early postnatal period determined the functional β-cells in the adult; it will be interesting to see whether this is a general phenomenon. In addition, we observed that Ccnd1 expression was not downregulated in the islets of the adult Insm1^+/lacZ mice. This suggests that Insm1 might not be involved in regulating Ccnd1 expression in adult β-cells or that one allele of Insm1 is sufficient to support the very low proliferation rate and Ccnd1 expression in adult β-cells. We did not observe decreases in β-cell mass in the newborn Insm1^+/lacZ mice, indicating that Insm1 dosage has little effect on β-cell development. Prenatal β-cells differentiate from stem/progenitor cells (1), whereas the postnatal β-cell expansion arises from the self-duplication of β-cells (3). There are different scenarios to describe regulation of β-cell mass in the two different developmental stages. For instance, Nkx6.1 is required for postnatal, but not prenatal, β-cell mass expansion (9); Ccnd1, identified as a target of Insm1 in our work, does not contribute to prenatal β-cell growth (13,14). We found that Insm1 dosage regulates Glut2 expression. Glut2, an gene essential for glucose sensing and metabolism, was markedly reduced in diabetic animals (30–32). We observed a decrease of ∼40% in Glut2 expression in the islets of Insm1^+/lacZ mice, which could contribute to the impaired glucose handling under physiological or metabolically stressed conditions.

We showed that changes in Insm1 expression only regulate a few Insm1 targets, which were identified in Insm1-null mutation (as showed in Figs. 3B and 4A). The cell cycle genes Cdkn1c (p57), Cdkn1b (p27), and Ccnd1 were downregulated in Insm1-null mutants (20,21), but only Ccnd1 was downregulated in Insm1 heterozygotes (Fig. 3B). The transcription factors and metabolic genes Pdx1, MafA, Glut2, Hk1, and Pcx were downregulated in Insm1 mutants (21). However, only Glut2 was downregulated in Insm1 heterozygotes (Fig. 4A). We found that the binding affinity of Insm1 to chromatin varied depending on the locus. Low-affinity binding sites were more susceptible to decreases in Insm1 levels and therefore more susceptible to potentially be affected by Insm1 dosage deficiency. Loci with no affinity or high affinity did not show dysregulation of the related genes in Insm1^+/lacZ islets (Figs. 3B, 4A, and 5A and Supplementary Fig. 5C). Thus, only genes that have relatively low Insm1 binding affinity are potentially regulated by Insm1 in a dose-dependent manner.

Osipovich et al. (20) observed increased Ccnd1 expression and sevenfold decreased proliferation in embryonic pancreatic β-cells in the absence of Insm1, as well as a slight increase in proliferation and decrease in Ccnd1 expression in mature Insm1-deficient β-cells (21). These studies investigated Insm1 function at embryonic or adult stages in the absence of Insm1 and did not study dose-dependent effects. Embryonic loss of Insm1 results in severe developmental defects (loss of β-cell development and no insulin production [19,33]), and conditional loss of Insm1 in adult islets causes functional defects (continuous leak of insulin). However, Ccnd1 expression is not only directly regulated by Insm1 but also indirectly regulated by these defects. In fact, insulin and glucose are essential regulators of β-cell proliferation (24,34). Severe developmental defects, abnormal blood insulin and glucose levels, or higher numbers of and/or more intense dysregulated genes other than Ccnd1 all contribute to the proliferation defects observed in complete or conditional mutants of Insm1. In our present study, we used Insm1^+/lacZ mice, which have mild defects in blood glucose and insulin levels. We found that Insm1 regulates Ccnd1 expression and cell cycle length in a dose-dependent manner.

Zhang et al. (35) showed that the Insm1 protein can interact with Ccnd1 and Cdk 4. Ectopic expression of Insm1 interrupts Ccnd1-Cdk4 function and negatively regulates the cell cycle (35,36). In the study by Zhang et al. (35), they defined Insm1 function in nonpancreatic β-cells. Interestingly, abnormally high Insm1 expression levels were observed in neuroendocrine tumors, including insulinoma cells, which are highly proliferative (37–44). Tissue-specific factors can play distinct roles in cell cycle control in different cell types, and one dramatic example is provided by the transcription factor RE1 silencing transcription factor (REST) (45). Additional investigation is required to determine whether Insm1 performs similar or different functions in regulating the Ccnd1–Cdk4 complex and cell cycle in pancreatic β-cells using additional Insm1 mutant models.

We detected increased apoptosis in pancreatic β-cells from Insm1^+/lacZ mice. Neurod1 and Pdx1 regulate β-cell apoptosis in prenatal or aging mice (46,47). Insm1 can bind the Neurod1 promoter and can also regulate the expression of Pdx1 (21,48). However, our current study showed that Insm1 dosage does not regulate the expression of these genes, illustrating a difference between haploinsufficiency and complete loss of function mutations. Furthermore, we did not find an association between delayed cell cycle and apoptosis in Insm1^+/lacZ β-cells. Thus, the increase in apoptosis in β-cells of Insm1^+/lacZ mice is likely independent of the observed cell cycle delay and requires further investigation.

Acknowledgments. The authors thank Bettina Brandt (Max Delbrück Center for Molecular Medicine) for technical assistance and Claudia Paeseler (Max Delbrück Center for Molecular Medicine) for animal husbandry. The authors also thank Dr. Michael Strehle (Max Delbrück Center for Molecular Medicine) for editing the language and style of the manuscript and Aimin Xu (The University of Hong Kong) for helpful discussion and critical reading of the revised manuscript.

Funding. This work was supported by the National Natural Science Foundation of China (grant 81770771 to S.J.), the Guangzhou Science and Technology Program (grant 201704020209 to S.J.), and the Natural Science Foundation of Guangdong Province (grant 2017A030313527 to S.J.).

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

Author Contributions. W.T. designed the study, performed experiments, and analyzed the data. Y.Z. and L.M. contributed to the animal experiments, data analysis, and figure preparation. C.D., H.D., X.L., and R.L. contributed to the in vitro experiments. S.L. and T.N. contributed to the figure preparation. W.C. and C.W. contributed to the data analysis. C.B. contributed to study design and edited the manuscript. S.J. supervised the project, analyzed the data, and wrote the article. S.J. is the guarantor of this work and, as such, had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.