All β Cells Contribute Equally to Islet Growth and Maintenance

In healthy adult mice, the β cell population is not maintained by stem cells but instead by the replication of differentiated β cells. It is not known, however, whether all β cells contribute equally to growth and maintenance, as it may be that some cells replicate while others do not. Understanding precisely which cells are responsible for β cell replication will inform attempts to expand β cells in vitro, a potential source for cell replacement therapy to treat diabetes. Two experiments were performed to address this issue. First, the level of fluorescence generated by a pulse of histone 2B–green fluorescent protein (H2BGFP) expression was followed over time to determine how this marker is diluted with cell division; a uniform loss of label across the entire β cell population was observed. Second, clonal analysis of dividing β cells was completed; all clones were of comparable size. These results support the conclusion that the β cell pool is homogeneous with respect to replicative capacity and suggest that all β cells are candidates for in vitro expansion. Given similar observations in the hepatocyte population, we speculate that for tissues lacking an adult stem cell, they are replenished equally by replication of all differentiated cells.


Introduction
Stem cells are defined by an ability to self-renew and differentiate into a variety of cell types. Some adult organs, including the intestine, skin, blood, and parts of the brain, are maintained by stem cells [1][2][3][4][5]. In cases where the differentiated cells are postmitotic, such as erythrocytes and olfactory neurons, tissue turnover depends entirely on stem cell differentiation.
To explain the mechanism of b cell maintenance and regenerative repair, it has been hypothesized that renewal occurs via an adult stem cell residing in the pancreatic ducts [6], acini [7], islets [8,9], spleen [10], or bone marrow [11]. In contrast, Dor et al. found that pre-existing b cells, rather than stem cells, are the major source of new b cells in healthy and pancreatectomized mice [12]. Furthermore, the forced cell cycle arrest of b cells severely restricts postnatal b cell mass [13], indicating that non-b cells (such as putative adult stem cells) cannot maintain b cell mass. Together, these results demonstrate that b cell mass is predominately, if not exclusively, sustained through the replication of b cells.
It remains unclear whether all b cells contribute equally to growth and maintenance. Two possible models might explain the expansion of b cells. The b cell population may be heterogeneous, comprised of both highly replicative cells and very slowly dividing, possibly postmitotic, cells. This would be consistent with the hypothesis that a subpopulation of insulin-expressing cells may maintain the entire pool, perhaps as unipotent adult stem cells [14] or by reversible dedifferentiation to a replicative state [15]. Alternately, the b cell population may be homogeneous, with all b cells contributing equally to growth.
Two approaches were used to address this issue ( Figure 1). First, a broad survey of the replicative potential of the entire b cell pool was performed by monitoring the dilution or disappearance of a fluorescent marker accompanying cell division. b cells were pulse labeled with a tetracycline-inducible histone 2B-green fluorescent protein (H2BGFP) [16] and, following a chase period, the level of fluorescence detectable within b cells was measured. Second, the clonal descendents of individual b cells were examined using a reporter system developed for mosaic analysis with double markers (MADM) [17]. Both assays are designed to assess whether b cells are a heterogeneous population. If b cells are heterogeneous, highly replicative b cells will lose the H2BGFP label quickly as they replicate and generate large clones, while slowly dividing b cells will retain the H2BGFP label and generate small clones. Alternately, if b cells are a homogenous population, all b cells would be expected to lose the H2BGFP label at similar rates, and all clones should be of comparable size. We observed uniform loss of the H2BGFP label with time, and detected only similarly sized clones in the chase population. The tetracycline-inducible H2BGFP and MADM systems are complementary approaches, both supporting the conclusion that all b cells contribute equally to b cell growth and maintenance.

H2BGFP Is Diluted with Cell Division In Vitro
Tumbar et al. engineered transgenic mice expressing H2BGFP from a tetracycline-responsive promoter (tetracycline-inducible promoter [tetO]-H2BGFP) to mark cells and assess their rates of division [16]. To verify that H2BGFP is diluted with cell division and distributed equally between daughter cells, we characterized the tetO-H2BGFP system in vitro. Rosa26 and CAGGs (constitutive promoters containing the CMV enhancer and the chicken b-actin promoter) are commonly used in mouse embryonic fibroblasts (mEFs) and mouse embryonic stem (mES) cells. We used these promoters to drive expression of the reverse tetracycline transactivator (rtTA) in the presence of doxycycline. Rosa26-rtTA; tetO-H2BGFP mEFs and CAGGs-rtTA; tetO-H2BGFP mES cells express H2BGFP within 12 h of doxycycline application (Figure 2A and 2C). Doxycycline was removed from the media, and the progressive dilution of H2BGFP protein resulting from cell division was measured by fluorescenceactivated cell sorter (FACS; Figure 2B and data not shown). A uniform loss of label was observed and the median GFP intensity of GFP-positive cells decreased with time. Given that mEFs divide every 24 h (unpublished data), and that H2BGFP fluorescence can no longer be detected after 5 d, H2BGFP fluorescence is no longer detectable above background in vitro by FACS after a population has undergone approximately five cell divisions ( Figure 2B). The standard deviation of fluorescent intensity within the labeled pulse population is too large to precisely measure the number of cell divisions within the chase population.
To verify that H2BGFP is segregated equally between daughter cells, CAGGs-rtTA: tetO-H2BGFP mES cells were cultured on the stage of a confocal microscope and imaged every 12 min. After the first division, total GFP fluorescence in the two daughter cells, measured as integrated pixel intensity, added up to the fluorescence of the original cell, and H2BGFP was split equally between daughter cells in the first and second divisions ( Figure 2D). In both dividing and nondividing cells, minimal bleaching was observed, despite imaging every 12 min over 18 h ( Figure 2D and unpublished data). Detection of H2BGFP fluorescence is dependent on the laser settings used. In this case, H2BGFP fluorescence was no longer detectable after three rounds of cell replication, though lower laser intensities were used than typically employed for fixed tissue sections.

H2BGFP Incorporation Is Replication Independent
Because b cells divide slowly, and some may be postmitotic, it was important to determine whether all cells, regardless of replicative activity, can be labeled by the inducible H2BGFP system. We found that H2BGFP labeling occurs independent of cell division in cultured Rosa26-rtTA; tetO-H2BGFP mEFs ( Figure S1). Cells treated with mitomycin C are irreversibly arrested in S phase, but still express H2BGFP, which incorporates into the nucleus within 12 h of the administration of doxycycline. Furthermore, cells reversibly arrested with either aphidicolin (G 0 /G 1 block) or nocadozole (G 2 /M block) express H2BGFP upon treatment with doxycycline. At 3 h after release from nocadozole, cells develop labeled mitotic spindles, indicating that the H2BGFP has been integrated into chromatin and not just added to a nuclear histone pool (unpublished data).

In Postmitotic Cell Populations, H2BGFP Is Stable for at Least Six Months
The Rosa26 locus is active in most cells of the mouse [18], suggesting that in the presence of doxycycline, Rosa26-rtTA should drive tetO-H2BGFP expression and label most mouse cells. Rosa26-rtTA; tetO-H2BGFP labeled diverse cell types, including but not limited to pancreas, intestine, fat, bone marrow, muscle, skin, and retina (Figures 3, 4, and unpublished data), but not cortical neurons, olfactory bulb, or spinal cord, possibly due to the inability of doxycycline to cross the blood-brain barrier (unpublished data). Under repressive conditions (without doxycycline), no expression was observed in these organs (n ¼ 4; Figures 3 and 4).
The  Two approaches were used to study replication of b cells. Pulse-chase analysis follows the loss of the H2BGFP label (green) with b cell division, while clonal analysis follows the generation of (yellow) clones from individual b cells. Model 1: the b cell population is heterogeneous, composed of fast-and slow-dividing subpopulations. This model predicts that highly replicative b cells will lose the H2BGFP label quickly and generate large clones, while slowly dividing b cells will fail to dilute the H2BGFP label and generate small clones. Model 2: the b cell population is homogeneous, and all b cells divide at the same rate. This model predicts that all b cells will lose the H2BGFP label uniformly, and that all clones identified through clonal analysis will be of comparable size. doi: 10

Author Summary
The b cells of the pancreas are responsible for insulin production and their destruction results in type I diabetes. b cell maintenance, growth, and regenerative repair is thought to occur predominately, if not exclusively, through the replication of existing b cells, not via an adult stem cell. It was previously unknown, however, whether all b cells divide at the same rate, or if multiple subpopulations of b cells exist, some highly replicative and others very slowly dividing, possibly postmitotic. We performed two types of experiments to determine whether all b cells are alike: label-retaining analysis and clonal analysis. Our results indicate that all b cells contribute equally to islet growth and maintenance. PLoS Biology | www.plosbiology.org [19]. Photoreceptor cells can be labeled by Rosa26-rtTA; tetO-H2BGFP ( Figure 4A). Following a chase of 6 mo ( Figure  3C), H2BGFP was detected in whole eyes and sectioned retinas at the same imaging settings used to collect pulse data. Staining with recoverin verified that label retention is restricted to the photoreceptor layer of the retina ( Figure  4A). Thus, the H2BGFP label is stable and retained in postmitotic cells. It is formally possible that the H2BGFP protein has a shorter half-life in pancreatic b cells than in postmitotic photoreceptor cells.

LRCs Can Be Identified in Known Stem Cell Populations Using H2BGFP
As a further validation that the tetO-H2BGFP system can identify heterogeneity in cell populations, we confirmed that we could detect nonuniform loss of H2BGFP in tissues where slow-cycling cells are known to exist. Stem cells are often proliferatively quiescent compared with neighboring transitamplifying cells. Because of their slow-dividing nature, stem cells tend to remain labeled in experiments that use DNA synthesis labels such as tritiated thymidine or bromodeoxyuridine. Label-retaining cells (LRCs) in Rosa26-rtTA; tetO-H2BGFP mice are clearly visible in the hair follicle bulge cells after a 2 mo chase, as previously shown by Tumbar et al. [16] ( Figure 4B). Furthermore, GFP-positive intestinal crypt cells can be identified up to 1 mo following the pulse, though the vast majority of the labeled cells in chase intestines seem to be slowly dividing mesenchymal cells and intestinal neurons ( Figure 4B). Finally, Lin low kit þ sca þ sorted hematopoietic stem cells within the bone marrow also retain GFP fluorescence to a much greater extent than whole bone marrow (unpublished data). Because pulse-chase experiments using tetO-H2BGFP can identify tissue heterogeneity in the form of slowly dividing stem cell populations, it should also be able to identify a subpopulation of slowly dividing b cells, should it exist.

No LRCs Are Detected in the b Cell Population
To determine whether all b cells divide at the same rate, we used the tetO-H2BGFP strategy in combination with promoters that express either tetracycline transactivator (tTA) or rtTA within the pancreas. Pdx1 expression in the postnatal pancreas is enriched in b cells, where it regulates insulin expression [20]. In tetO-H2BGFP animals, expression of tTA from the native Pdx1 locus (Pdx1-tTA) should label b cells in the absence of doxycycline, while the transgenic rat insulin promoter (RIP)-rtTA should label b cells in the presence of doxycycline. However, RIP-rtTA, but not Pdx1-tTA, labeled b cells even in unpulsed animals (unpublished data and Figure  3B), indicating that the RIP-rtTA transgene system is leaky. Therefore, Pdx1-tTA is the only b cell-specific driver suitable for these experiments. It should be noted that Pdx1-tTA animals are haploinsufficient; however, mice with one inactivated Pdx1 allele can be maintained in the heterozygous state and have normal pancreatic development and b cell maintenance, though they show modestly impaired glucose tolerance [20].
Pulsing Pdx1-tTA; tetO-H2BGFP animals for 6 wk after birth labeled 80% of b cells ( Figure 3B). In addition to labeling b cells, some labeled nuclei occurred outside the b cell pool ( Figure S2). Somatostatin and pancreatic polypeptide-expressing cells were frequently labeled in Pdx1-tTA; tetO-H2BGFP animals, consistent with the fact that Pdx1 was originally cloned from somatostatin-producing islet cell lines [21,22], and that Pdx1 expression can be detected in these cells [23]. Rare glucagon cells were labeled, while exocrine cells (amylase-positive) were consistently weakly labeled, and no H2BGFP expression was observed in the ducts (CK19positive; Figure S2). These observations are also consistent with the findings of Oster et al., who observed rare Pdx1immunoreactive nuclei in all pancreatic cell types [23].
The entire b cell pool was assayed for LRCs. A group of 38 Pdx1-tTA; tetO-H2BGFP animals (19 female, 19 male) were pulsed from birth until 6 wk of age by ceasing doxycycline administration at postpartum day 0 (P0; Figure 3C). Six of these animals (three male, three female) were euthanized following a 6-wk postnatal pulse. All animals collected in the pulse group showed 80% of b cells labeled with H2BGFP; labeling was consistent throughout the pancreas and between animals (n ¼ 6). The remaining mice were again administered doxycycline water to repress transcription of H2BGFP, and were euthanized after chase periods of 1 wk (n ¼ 4), 2 wk (n ¼ 4), 1 mo (n ¼ 6), 2 mo (n ¼ 6), 3 mo (n ¼ 6), and 6 mo (n ¼ 6). Sections of the pancreati were stained with insulin to identify b cells. We observed a uniform loss of label in b cells with time ( Figure 5A).
To confirm this finding, the experiment was repeated with Rosa26-rtTA; tetO-H2BGFP animals. A group of 40 animals (18 female, 22 male) were pulsed from conception until 6 wk of age by administration of doxycycline water. Eight animals (four male, four female) were euthanized at 6 wk of age. Pulse expression in these mice was more variable between animals than for Pdx1-tTA; tetO-H2BGFP mice, so all experiments were performed on sibling cohorts. The remaining mice were removed from doxycycline and were euthanized after chase periods of 1 wk (n ¼ 4), 2 wk (n ¼ 4), 1 mo (n ¼ 6), 2 mo (n ¼ 6), 3 mo (n ¼ 6), and 6 mo (n ¼ 6). Sections of the pancreati were stained with insulin to identify b cells. Again, we observed uniform loss of label in b cells with time ( Figure 5A).
To measure the relative intensity of GFP-positive cells, we used FACS analysis of dissociated islet cells. For these experiments, littermates were pulsed for at least 6 wk, and chases were structured so that all animals could be euthanized and analyzed by FACS on the same day. Importantly, comparisons of animals pulsed for 6 or 14 wk showed no significant increase in the median intensity of the GFPpositive population (unpublished data). By analyzing all timepoints in parallel, we were able to compare the relative GFP intensity between the pulse and chase populations. This experiment was repeated four times using Rosa26-rtTA; tetO-H2BGFP mice and twice using Pdx1-tTA; tetO-H2BGFP mice; the results were consistent each time. Exposure-matched photographs of whole islets taken prior to dissociation ( Figure 5B) and FACS plots of dissociated islets ( Figure 5C) indicate that the median intensity of the GFP fluorescence within the b cell pool decreases with time. No outlying population of LRCs in the b cell pool can be identified.
Hepatocytes, like b cells, are an endodermal cell type, and are thought to be maintained by self-replication [24]. It is unknown, however, whether all hepatocytes contribute equally to growth and maintenance. We examined livers from Rosa26-rtTA; tetO-H2BGFP pulse-chase animals and observed a uniform loss of label with time ( Figure S3). Similar to the results with b cells, no outlying population of LRCs in the hepatocyte pool was identified.

Clonal Analysis Demonstrates that b Cells Contribute Equally to Pancreas Expansion
To directly compare the replication capacity of individual b cells, a lineage-based clonal analysis in the pancreas was performed. RIP-CreER transgenic mice [12] drove expression of tamoxifen-dependent Cre recombinase specifically in b cells, while the MADM reporter system [17] was used to label individual b cells. The MADM system is a unique tool that allows low-frequency labeling of cells, a prerequisite for clonal analysis ( Figure 6A). It contains two alleles at the Rosa26 locus, Rosa26 GR and Rosa26 RG , each containing reciprocal parts of chimeric marker genes (GFP and RFP) interrupted by a loxP site. Neither allele generates an active fluorescent protein until Cre-mediated interchromosomal recombination restores functional expression of GFP and RFP. Cells are labeled differently depending on when recombination occurs during the cell cycle. Recombination at G 0 or G 1 results in double-colored cells (expressing both GFP and RFP). Alternately, recombination at G 2 results in two outcomes at equal frequency: either one red and one green daughter cell (single-colored cells), or one colorless and one double-colored daughter cell.
As expected, no labeling was observed in RIP-CreER; Rosa26 GR /Rosa26 RG mice in the absence of tamoxifen. We injected tamoxifen over 3 d into 23 RIP-CreER; Rosa26 GR / Rosa26 RG mice between 4 and 8 wk of age. Two animals were euthanized within 4 d of tamoxifen injection (the pulse group), and we found that 0.1%-0.5% of b cells were labeled. All clones observed were insulin positive, and 90% were single-cell clones (the remainder were composed of two cells).
In pulse and chase animals, all labeled cells expressed both GFP and RFP (RFP expression required antibody staining for detection; Figure S4A). The existence of only double-colored clones indicates that recombination occurred during G 1 or G 0 , which is expected for a slow-dividing cell population that spends little time in G 2 . Conversely, in experiments performed using Pdx1-Cre; Rosa26 GR /Rosa26 RG mice, where Cre recombinase was expressed in the rapidly dividing embryonic pancreas, single-colored clones were detected occasionally ( Figure S4B). These observations are consistent with those of Zong et al. [17], who found that single-colored cells are a minority, though their proportion to doublecolored cells increases when Cre recombinase is expressed in a rapidly dividing population. Given that all b cell clones are double-colored, all clones are shown only in green to allow for ease of presentation, and insulin staining is shown in red.
For the purpose of this experiment, clones are defined as clusters of labeled cells within a single islet; all cells within a clone are assumed to be derived from a single b cell. Tamoxifen-treated mice were euthanized 1 mo (n ¼ 12) or 2 mo (n ¼ 9) following the pulse. A total of 175 clones from 1mo chase animals and 122 clones from 2-mo chase animals were sampled using single random sections, and no large clones were detected. The absence of large clones qualitatively suggests an absence of fast-dividing b cells. To estimate the expansion of b cells, we analyzed serial sections. After 1 mo of chase, the average size of a clone was 5.1 6 5.4 cells (n ¼ 45 clones), whereas after 2 mo of chase, the average size of a clone was 8.2 6 6.9 cells (n ¼ 40 clones). In addition, the experiment was repeated using CAGGs-CreER; Rosa26 GR / Rosa26 RG mice, and again, no large b cell clones were detected (unpublished data). Labeled cells found within a single islet are of clonal origin: the probability that three or more labeled cells found within the same islet are not clonal is 1.8 3 10 À2 (based on Poisson distribution analysis assuming labeling frequency of 0.1% and islet size of 500-1,000 cells). This clonal analysis supports the model that the growth and maintenance of b cell mass in the adult pancreas is achieved by the replication of individual b cells that have similar replicative capacities.

Discussion
Our experiments were designed to address the mechanism of growth and maintenance The b cell mass is dynamic and can respond to environmental cues such insulin and glucose [25]. The b cell number increases dramatically in the first year of rodent life [12,26], up to 10-fold in cases of insulin resistance [27], and up to 1.5fold during pregnancy [28,29]. Recent experiments suggest that when not hindered by persistent autoimmune attack or the toxicity of high blood glucose levels [30], b cells have the capacity to regenerate. While the mechanism regulating b cell expansion remains unclear, our findings indicate that all b cells are capable of replication and are therefore viable targets for in vitro or in vivo expansion.
Seaberg et al. recently reported that single-cell clones derived from adult islets generated colonies of 2,000-10,000 cells that expressed markers of neural, glial, pancreatic endocrine, exocrine, and duct identities [31]. These clones were generated from ;0.02% of islet cells, though their identity and relationship to in vivo growth is yet to be determined. We cannot rule out the possibility that a rare type of b cell was missed in our examination of individual clones using the MADM marking experiments. However, because the rate of clonal expansion is sufficient to account for the growth of the b cell population during the chase period, a rare highly proliferative b cell did not contribute significantly to the expansion of b cell mass.
Published rates for b cell replication in adult mice (12 wk old) are highly variable, from 2% [32] to 15% per day [33]. Assuming 5% of b cells replicate per day, and that all b cells are equivalent, b cells should divide approximately every 20 d. This would dilute the H2BGFP label beyond detection (by completing up to five rounds of replication) within 100 d. In addition, clone size at 2 mo should be approximately eight cells. These straightforward calculations predict results that are entirely consistent with our findings. These estimates, of course, assume no b cell death over the duration of our experiments.
b cells have a finite lifespan, but the absolute b cell death rate is unknown. Based on b cell mass measurements and an estimate of b cell proliferation of 2% per day throughout adulthood, Finegood et al. calculated the b cell lifespan to be 52 d [32]. Recent findings demonstrate that b cell proliferation rates decline to less than 0.1% in 1-y-old mice [33], casting doubt on the often quoted rates for b cell turnover in mice. Furthermore, TUNEL analysis of wild-type b cells consistently fail to identify apoptotic cells [13,27,33,34]. Regardless of the true rate of b cell turnover, our findings of a uniform loss of label and a consistent clone size indicate that all b cells have equivalent replicative capacity.
Pancreatic b cells are not the only differentiated cell type capable of growth and maintenance without the support of an adult stem cell population. Hepatocytes are highly replicative and not thought to be supported by a facultative stem cell under normal conditions [35]. Pulse-chase analysis with the tetracycline-inducible H2BGFP label shows that all hepatocytes lose their label at the same rate. Therefore, like the b cell population, the hepatocyte population seems to be homogeneous. We do not know of an example of a mature differentiated cell type that has two populations (one replicative and the other not). We speculate that when tissues are without an adult stem cell, they are replenished by equal replication of all differentiated cells.
The demonstration that all b cells are equivalent, contributing equally to the growth and maintenance of the b cell population, has clinical implications if we assume that rodents and man use the same mechanism for pancreatic homeostasis. The destruction of b cells that causes type I diabetes has been counteracted by the transplantation of b cells. The clinical impact of this approach is currently limited, in part, by the scarcity of available pancreatic tissue [36]. A better understanding of adult b cell replication may help attempts to expand pancreatic b cells in vitro as a source of transplant material to treat diabetes.

Materials and Methods
Mice. Greater variation of H2BGFP expression was observed in Rosa26-rtTA; tetO-H2BGFP animals than with the other drivers; experiments were conducted on littermates to reduce variability. Notably, the highest-expressing animals were often runted, indicating that either Rosa26-rtTA activity or H2BGFP expression in some cell types is harmful. Breeding to a C57BL/6 inbred background reduced variability to some extent.
Doxycycline and tamoxifen. Doxycycline (Sigma, http://www. sigmaaldrich.com) was added to drinking water at 1 mg/ml and sweetened with sucrose (1%). Water bottles were changed weekly with freshly prepared solution. Tamoxifen (Sigma) was dissolved in corn oil at 20 mg/ml and mice were injected intraperitoneally (6 or 8 mg/d for 3 consecutive days).
Immunohistochemistry. Tissue was dissected from mice, fixed in 4% paraformaldehyde/PBS solution for two hours at 4 8C, washed in PBS, incubated in 30% sucrose/PBS solution overnight, embedded in OCT (Tissue-Tek; Electron Microscopy Sciences, http://www. emsdiasum.com) and stored at À80 8C. Frozen samples were sectioned at 10 lm for staining or up to 50 lm for serial analysis. www.hyclone.com], and 0.1 mg/ml ovalbumin trypsin inhibitor) was added, and islets were centrifuged, washed twice, and filtered through a 500 lm diameter wire mesh. Islets were centrifuged, washed twice in washing solution, resuspended in Histopaque 1077 (Sigma), and vortexed. The islet suspension was carefully overlaid with washing solution (without serum) and centrifuged for 20 min at 10 8C, separating islets from exocrine tissue. The islet layer was collected at the interface, pelleted, washed twice, and further purified by two rounds of gravity sedimentation. Finally, pure islets were handpicked under a dissecting scope. Islets were dissociated by incubation with 0.25% trypsin-EDTA (Gibco) at 37 8C for 5 min, washed, fixed for 15 min in 1% paraformaldehyde/PBS solution, resuspended in 5% donkey serum/PBS, and FACS sorted on a BD Aria (BD Biosciences, http://www.bdbiosciences.com).
CAGGs-rtTA; tetO-H2BGFP mES cells were obtained by electroporating a CAGGs-rtTA-IRES-puromycin plasmid into tetO-H2BGFP mES cells derived from blastocysts, selecting with 4 lg/ml puromycin in 10 lg/ml doxycycline and picking and expanding green colonies. mES cells were grown in standard mES conditions (knockout DMEM Dividing cells were imaged by culturing on the stage of a Zeiss LSM confocal microscope, and 15-image z-stacks were scanned every 12 min at 0.5% laser intensity. Images were quantified using MetaMorph software (Molecular Devices, http://www.moleculardevices.com), and values were recorded as integrated pixel intensity.
Cell-cycle experiments were performed as follows: 10 lg/ml mitomycin C (Sigma) was applied to cells for 3 h to irreversibly inhibit the cell cycle in S phase, 0.25 ng/ml aphidicolin (Sigma) or 100 ng/ml nocadozole (Sigma) was applied to cells for 6 h to reversibly arrest the cell cycle at G 0 /G 1 or G 2 /M, respectively. Figure S1. Labeling of Chromosomes with H2BGFP is Cell Replication Independent Rosa26-rtTA; tetO-H2BGFP mEFs express H2BGFP within 12 h of administration of doxycycline, even following cell-cycle inhibition.