Bookmarking by specific and nonspecific binding of FoxA1 pioneer factor to mitotic chromosomes

While most transcription factors are released from chromatin during mitosis, a subset remains, bookmarking genes for rapid reactivation upon mitotic exit. Here, Zaret and colleagues investigate pioneer factor FoxA1 mitotic binding in hepatic cells. Genomic mapping and mutagenesis studies reveal specific and nonspecific modes of FoxA1 binding that contribute to target gene reactivation post-mitosis. This study reveals an unexpected diversity in the mechanisms by which cell-specific transcriptional programs are activated after each round of mitosis.

Abstract

While most transcription factors exit the chromatin during mitosis and the genome becomes silent, a subset of factors remains and “bookmarks” genes for rapid reactivation as cells progress through the cell cycle. However, it is unknown whether such bookmarking factors bind to chromatin similarly in mitosis and how different binding capacities among them relate to function. We compared a diverse set of transcription factors involved in liver differentiation and found markedly different extents of mitotic chromosome binding. Among them, the pioneer factor FoxA1 exhibits the greatest extent of mitotic chromosome binding. Genomically, ∼15% of the FoxA1 interphase target sites are bound in mitosis, including at genes that are important for liver differentiation. Biophysical, genome mapping, and mutagenesis studies of FoxA1 reveals two different modes of binding to mitotic chromatin. Specific binding in mitosis occurs at sites that continue to be bound from interphase. Nonspecific binding in mitosis occurs across the chromosome due to the intrinsic chromatin affinity of FoxA1. Both specific and nonspecific binding contribute to timely reactivation of target genes post-mitosis. These studies reveal an unexpected diversity in the mechanisms by which transcription factors help retain cell identity during mitosis.

When cells enter mitosis, chromosomes condense (Caravaca et al. 2005) and the genome becomes silent (Prescott and Bender 1962; Spencer et al. 2000). Only a fraction of transcription factors are retained on mitotic chromosomes (Martinez-Balbas et al. 1995; Michelotti et al. 1997; Burke et al. 2005; Yan et al. 2006; Egli et al. 2008), and a subset of these facilitate rapid gene reactivation post-mitosis (Young et al. 2007; Blobel et al. 2009; Dey et al. 2009; Kadauke et al. 2012). The basis for the marked differential in transcription factors' binding to mitotic chromatin and how it reflects the factors' roles in genome reactivation is not clear.

In liver development, binding sites for FoxA and GATA factors are occupied on the silent liver gene alb1 in undifferentiated embryonic endoderm (Gualdi et al. 1996; Bossard and Zaret 1998). Upon hepatic induction, nearby binding sites for NF-1, C/EBP, and other factors become occupied, and the liver gene is activated. Among the factors that promote liver development, only FoxA proteins can bind their target sites on nucleosomes and enable the other factors to bind (Cirillo and Zaret 1999; Cirillo et al. 2002); hence, FoxA factors have been called “pioneer factors” (Cirillo et al. 2002; Zaret and Carroll 2011). While GATA4 is dependent on FoxA for binding nucleosomes (Cirillo and Zaret 1999), it can bind to compacted chromatin that is inaccessible to the other factors (Cirillo et al. 2002) and hence can be considered a subordinate pioneer factor. The structure of the DNA-binding domain (DBD) of FoxA resembles that of linker histone (Clark et al. 1993; Ramakrishnan et al. 1993), and the FoxA C-terminal domain, which is unlike that of linker histone, interacts with core histones and promotes local chromatin opening (Cirillo et al. 2002). The highly related FoxA1 and FoxA2 are encoded by unlinked genes and both are necessary for the activation of the hepatic program (Lee et al. 2005); FoxA1 has been shown to remain bound to mitotic chromosomes in adult liver cells (Zaret et al. 2011). We therefore sought to investigate the mechanism and role of FoxA binding to the mitotic genome in hepatic cells.

Results

Pioneer factors bind more strongly than other factors to mitotic chromatin

We previously assessed the interphase chromatin binding and mobility of GFP-tagged versions of FoxA1, GATA4, C/EBPα, NF-1, and other proteins that are expressed in the liver and contain different DBD structures (HMGB1, c-Myc, and linker histone H1o). Notably, FoxA1 moves much more slowly in chromatin than the other factors, correlating with its high nucleosomal binding ability, although not as slow as H1o (Sekiya et al. 2009). Here, we expressed the constructs in HUH7 adult human hepatoma cells that had been blocked in mitosis with nocodazole and visualized GFP fluorescence in live cells by high-resolution deconvolution microscopy (Agard 1984). GFP-FoxA1 was seen almost exclusively bound to chromosomes in the metaphase-arrested cells as well as in drug-free control cells passing through mitosis, mimicking the pattern of GFP-H1o (Fig. 1A). We estimate that the GFP transfected cells expressed ∼10-fold more of the respective amounts of the transcription factor than the endogenous protein (data not shown). When we used 20-fold lower amounts of transfected GFP-FoxA1 plasmid DNA, we observed much fainter signals but still marked binding to mitotic chromosomes (Supplemental Fig. 1a). GFP-GATA4 and GFP-HMGB1 fluorescence was seen both on the mitotic chromosomes and in the cytoplasm (Fig. 1B), while GFP-C/EBPα gave fainter signals on mitotic chromosomes than the other factors. Western blotting of endogenous C/EBPα showed it to be several-fold less stable in mitotic hepatoma cells, whereas FoxA1 was equal in abundance in mitotic and asynchronous cells (Supplemental Fig. 1b). GFP-c-Myc and GFP-NF1 exhibited background fluorescence on the mitotic chromosomes, reflecting factor exclusion (Fig. 1B). Cells released from the metaphase mitotic block and fixed at anaphase and telophase showed that GFP-FoxA1 remained bound to chromosomes throughout mitosis, while a GFP protein fused to a nuclear localization sequence was excluded (Supplemental Fig. 2a). Endogenous FoxA1 exhibited similar properties but with a more diffuse signal that is typical of fixed cells, compared with that seen when live cells are imaged (Supplemental Fig. 2b). GFP fused to the FoxA1 DBD was sufficient to bind mitotic chromosomes (Fig. 1A).

Figure 1.

Diverse modes of hepatic transcription factor binding to mitotic chromosomes. (A,B) GFP fluorescence in live HUH7 hepatoma cells visualized by deconvolution microscopy with or without nocodazole for mitotic arrest. (A) GFP and GFP/bright-field (BF) views showing that GFP-FoxA1, GFP-FoxA1 DBD, and GFP-H1o remain quantitatively bound to mitotic chromosomes. (B) A portion of the cellular GATA4 and HMGB1 is released from mitotic chromosomes in nocodazole-treated cells; C/EBPα becomes unstable, yet a portion remains bound; and c-Myc and NF1 are excluded from the mitotic chromosomes.

Thus, within this group of regulatory factors for the liver lineage, the earliest pioneer factor in development is the most strongly bound to mitotic chromatin; the subordinate pioneer factor GATA4 is partially bound, similar to that seen for GATA1 elsewhere (Kadauke et al. 2012); and later developmental factors are either partially bound, bound yet unstable in mitosis, or excluded from the mitotic chromosomes altogether.

Highly transcribed gene targets of FoxA1 remain specifically bound in mitosis

To assess where endogenous FoxA1 binds the genome in mitotic cells compared with asynchronously cycling cells, we performed chromatin immunoprecipitation (ChIP) coupled with deep sequencing (ChIP-seq) on triplicate cell populations, with and without nocodazole treatment (Supplemental Fig. 3a,b). More than 94% of the nocodazole-arrested cells were in mitosis, as assessed by cell morphology and H3Ser10 phosphorylation, while <2% of the asynchronously cycling wells were in mitosis (Supplemental Fig. 3a). Using model-based analysis of ChIP-seq (MACS) (Zhang et al. 2008) to assign peaks, we discovered 546 FoxA1-bound sites in mitotic cells and 3509 sites in asynchronous cells (Fig. 2A). Eighty-seven of the called FoxA1 peaks were specific to mitotic cells, but visual inspection of the unique sequence reads and quantitative PCR (qPCR) analysis of independent chromatin preparations indicated that the FoxA1-bound sites in mitotic cells are also bound in asynchronous cells (e.g., Fig. 2B [intergenic site on ch. 5 and MIPEP site at +92.7 kb, red arrowheads], B [ChIP-qPCR validations are of mitotic and asynchronous cell chromatin]). Certain weak FoxA1 peaks in mitosis that were not called by MACS were significant by ChIP-qPCR (Fig. 2B, AFP −4.1- and −2-kb sites and TTR promoter), similar to the weaker mitotic site at the HNF4a −7-kb site and unlike a negative control site on ch. 13 (Fig. 2B). ChIP for histone H3 at these sites showed comparable signals between the asynchronous and the mitotic chromatin (Supplemental Fig. 3c), demonstrating that differences seen for FoxA1 were specific to the factor and not the preparations of chromatin.

Figure 2.

FoxA1 in mitosis occupies the most strongly expressed and strongly bound genes in asynchronous HUH7 cells. (A) Peaks were pooled from three replicate ChIP-seq samples, revealing 3509 asynchronous FoxA1-binding sites, of which 546 are also bound in mitosis. (B) FoxA1 ChIP-seq data tracks and ChIP-qPCR confirmation in independent chromatin samples with signals normalized to input and per million sequence reads. Red arrowheads depict sites of mitotic binding, and those that were verified by qPCR are shown on the right. The circle indicates the negative control site. The break in the HNF4a and TTR bar graphs is to accommodate different scales for the asynchronous (as.) and mitotic (mi.) data. (C) FoxA1 ChIP-seq signals over all sites bound in mitotic and asynchronous cells compared with all sites bound only in asynchronous cells; the signal is normalized to input DNA quantity and the total number of aligned sequence reads. FoxA1 binding is much stronger to sites in mitotic and asynchronous cells than to sites bound in asynchronous cells only. (D) Common FoxA-binding motif at sites bound in mitosis versus asynchronous only. (E) Nearly all of the FoxA1-bound sites in mitosis are associated with genes either within the transcribed region or <20 kb upstream. (F) Box and whisker plots showing that genes bound by FoxA1 in mitotic and asynchronous cells correspond to those more highly expressed genes in hepatoma cells; (***) P < 10⁻¹⁵. (G) FoxA1 remains bound to hepatic transcription factor genes in mitosis.

FoxA1 peaks that occurred in both mitotic and asynchronous cells were, on average, markedly stronger in asynchronous cells than the FoxA1 peaks that occurred only in asynchronous cells (Fig. 2C, where overlapping peaks were merged). De novo motif analysis of FoxA1 peaks in mitosis revealed a Fox consensus sequence that was essentially the same as that seen at asynchronous-only sites (Fig. 2D). We found 601 genes bound by FoxA1 in both asynchronous and mitotic cells and 2722 genes bound by FoxA1 only in asynchronous cells (Fig. 2E). Triplicate microarray analyses of HUH7 cells showed that mitotic and interphase FoxA1 target genes represent those that are among the most highly expressed in interphase (Fig. 2F). Mitotic FoxA1 targets include transcription factor genes that are important for hepatic differentiation, such as HNF4a and FOXA1 itself (Fig. 2G), and genes for kinase signaling pathways (Supplemental Fig. 3d). In summary, we estimate that ∼15% of the FoxA1 sites that are bound in interphase cells are also bound in mitosis, corresponding to the strongest bound sites at the more highly expressed genes in interphase.

FoxA1 target sites that remain bound in mitosis have a higher intrinsic nucleosome occupancy score (INOS) than sites bound only in asynchronous cells

To further characterize potential differences between target sites for FoxA1 that remain bound in mitosis compared with sites bound only in asynchronous cells, we first compared the binding events with available histone modification data in the human ENCODE database. No substantial differences were observed to be centered over the FoxA1-binding sites (data not shown). Given the extensive prior data documenting the ability of FoxA1 to bind nucleosomal DNA (Cirillo et al. 1998; Cirillo and Zaret 1999; Chaya et al. 2001), we compared the FoxA1 mitotic and asynchronous binding data with a computational model that predicts how well a nucleosome could form at underlying 147-base-pair (bp) lengths of DNA (Kaplan et al. 2009; Tillo and Hughes 2009; Tillo et al. 2010). Figure 3 shows the average INOS for FoxA1 ChIP-seq peaks in mitotic and asynchronous cells across 1500 bp spanning each binding event. In both cases, there is a peak in INOS near the center of DNA binding. Peaks in INOS of ∼300 bp have been observed at binding events for other transcription factors (Tillo et al. 2010). A difference here is that the absolute value of INOS for FoxA1 is low, indicating that the protein does not generally bind to CG-rich regions of the genome. Indeed, the target sequence for FoxA1 is relatively AT-rich (Fig. 2D).

Figure 3.

FoxA1-bound sites in mitotic cells have higher predicted nucleosome occupancy compared with sites bound only in asynchronous cells. Average INOS for FoxA1 3509 ChIP-seq peaks seen only in asynchronous HuH7 cells and 544 peaks seen in mitotic cells within ±750 bp from the center of the peak; also shown are 100,000 sequences selected at random from human genome (hg18). The average INOS profiling of FoxA1 in mitotic HuH7 cells is higher than seen in asynchronous cells (t-test, P = 9.5 × 10⁻⁶).

Interestingly, the INOS for FoxA1 in mitotic cells is significantly higher than that in asynchronous cells (P = 9.5 × 10⁻⁶). The higher INOS for mitotic peaks suggests that the underlying nucleosome is typically more stable at chromosomal sites where FoxA1 remains bound in mitotic cells compared with that seen at sites where FoxA1 binds only in asynchronous cells. This possibility is consistent with the aforementioned features of FoxA1 binding its target site on nucleosomes in vitro and in vivo and suggests that more stable nucleosomal targets help predict mitotic chromosome binding.

Substantial nonspecific mitotic chromosome binding by FoxA1

We next addressed the apparent conundrum that FoxA1 is stable in mitosis (Supplemental Fig. 1b) and retained globally on the chromosomes (Fig. 1A) yet dissociates from many specific interphase binding sites (Fig. 2A). To investigate changes in the chromatin-binding properties of FoxA1, we performed fluorescence recovery after photobleaching (FRAP) assays in unfixed cells on GFP-FoxA1, with GFP-H1o as a control. GFP-H1o in HUH7 cells exhibited a threefold increase in FRAP half-time in metaphase compared with interphase (Fig. 4A); a similar increase was seen for H1c-GFP in other cells (Chen et al. 2005). Thus, linker histone moves more slowly in metaphase chromatin, consistent with the chromosomes' high degree of compaction. In marked contrast, GFP-FoxA1 exhibited a more than twofold decrease in FRAP half-time in metaphase (Fig. 4A). Thus, the dramatic compaction of chromatin in mitosis is associated with an increase in the mobility of FoxA1 compared with interphase.

Figure 4.

Increased mobility and nonspecific binding of FoxA1 in mitotic cells. (A) Relative fluorescence intensity (RFI) analysis showing that while GFP-H1o moves threefold more slowly in mitotic chromatin compared with interphase nuclei, GFP-FoxA1 moves 2.5-fold more quickly. Error bars denote standard error of the mean (SEM). Primary FRAP data for RFI analysis. White circles indicate the bleached area. (B) Unique FoxA1 ChIP-seq signals from two to 10 reads per million per 25-bp interval mapped to the left arm of human ch. 7 depicting higher nonspecific, background binding in mitotic cells (red arrows) and many more peaks in asynchronous cells. Input DNA is shown at two to 20 reads per 25-bp interval.

We questioned whether the increased mobility of FoxA1 in mitosis could be visualized in the genomic ChIP-seq data as an increase in nonspecific DNA-binding signals. Indeed, when viewed at the chromosomal level, there was a greater background of nonpeak FoxA1 ChIP-seq reads in mitotic cells compared with asynchronous cells (Fig. 4B, blue lines below red arrowheads; Supplemental Fig. 4). Thus, the increased mobility of FoxA1 in mitosis is associated with significant amounts of FoxA redistributing from specific sites to the flanking chromosomal domains.

We questioned whether the nonspecific binding at the chromosomal level in mitotic cells was dependent on FoxA1 interactions with DNA. To address this, we transfected HUH7 cells with GFP-tagged variants of FoxA1 that perturb nonspecific DNA binding (FoxA1-RR) or specific DNA binding (FoxA1-NH) (Sekiya et al. 2009). FoxA1-RR has alanine substitutions at two residues that make phosphate contacts with the DNA backbone (Fig. 5A). The FoxA1-RR mutant exhibits a marked decrease in overall and nonspecific DNA binding but still recognizes FoxA target sites, albeit more weakly than wild type (Sekiya et al. 2009). FoxA1-NH has alanine substitutions at two residues that make base contacts with DNA (Fig. 5A). The FoxA1-NH mutant exhibits normal nonspecific DNA binding but no longer recognizes FoxA target sites (Sekiya et al. 2009). Strikingly, when observed in live mitotic cells, much of the GFP-FoxA1-RR mutant was cytoplasmic (Fig. 5B). In contrast, the GFP-FoxA1-NH mutant was largely retained on the mitotic chromosomes (Fig. 5B), demonstrating that specific binding is not necessary for bulk mitotic retention. From these data, we conclude that much of the FoxA1 in mitotic chromatin is bound nonspecifically to the DNA.

Figure 5.

FoxA1 mutants that perturb specific or nonspecific DNA binding reveal significant nonspecific binding to mitotic chromatin. (A) Crystal structure view of the FoxA DBD depicting residues mutated to perturb specific (NH) and nonspecific (RR) DNA binding (Sekiya et al. 2009). (B) The FoxA1 mutant with impaired nonspecific DNA binding is partially released from mitotic chromosomes, while the FoxA1 mutant with impaired specific DNA binding is mostly retained. (C) ChIP-qPCR assays on transfected HUH7 cells in mitosis showing that the FoxA1-RR nonspecific binding mutant can still recognize the Afp −4.1-kb and Ttr promoter target sites that contain FoxA-binding motifs (shown in the top row, red arrowheads), whereas the FoxA1-NH specific binding mutant cannot. Thus, even when loss of nonspecific binding results in most of the FoxA1 being lost from mitotic chromosomes (B), FoxA1 can bind specifically to target sites in mitotic cell chromatin (C).

To assess the contributions of specific and nonspecific binding to target site binding in mitosis, we performed ChIP for the GFP tag in cells transfected with GFP-FoxA1 and the variants (Fig. 5C; Supplemental Fig. 5a). As expected, wild-type GFP-FoxA1 bound to endogenous FoxA1 targets in both asynchronous and mitotic cells (Supplemental Fig. 5b). Of three target sites that possess a FoxA-binding motif, binding by the GFP-FoxA1-RR mutant at the AFP −4.1-kb site and the TTR promoter was diminished, but still significant, in comparison with the IgG ChIP control (Fig. 5C, red arrowheads). In contrast, the GFP-FoxA-NH mutant exhibited a loss of binding to all of these sites (Fig. 5C). Therefore, despite a loss in nonspecific DNA binding and significant release from chromosomes, the FoxA-RR mutant can still bind in mitosis to sites with FoxA1 motifs. We conclude that binding by FoxA1 in mitotic chromosomes can take two forms: nonspecific binding that is not reflected in specific peaks in the ChIP-seq data and specific binding at a subset of asynchronous cell FoxA1 target sites.

Specific and nonspecific mitotic binding promotes target gene reactivation post-mitosis

Previous knockdown studies of Brd4 (Dey et al. 2009), MLL (Blobel et al. 2009), and GATA1 (Kadauke et al. 2012) found that these mitotic bookmarking factors are required for the timely reactivation of genes to which they bind in mitosis and are not required for the initial reactivation of genes to which they are bound only in interphase. To assess FoxA1 in this context, we compared the target gene reactivation after mitotic release in cells transfected for 2 d with a FOXA1 knockdown siRNA versus a control siRNA (Supplemental Fig. 5c,d). The relatively short time period is sufficient for determining effects on nascent transcript induction (see below) but, in our hands, not long enough for marked effects on steady-state levels of mRNA. By immunofluorescence, RNA polymerase II was absent from metaphase chromosomes in blocked HUH7 cells and then rebound chromatin 80 min after release in some cells (Supplemental Fig. 5e). Incorporation of ethynyl uridine (EU) (Jao and Salic 2008) into nascent RNA was undetectable in arrested cells; sparse staining appeared in some cells by 80 min after release and was more uniform across the nucleus by 90–100 min post-release in late telophase (Supplemental Fig. 5e), similar to other cell types (Prasanth et al. 2003). Based on these findings, we collected triplicate RNA time points before and after transcriptional reactivation and analyzed genes by RT-qPCR with primer sets that span intron–exon junctions to exclusively detect nascent mRNA. Many genes exhibited a continual increase in nascent transcripts for hours after mitotic release, while others exhibited an initial burst of expression (Fig. 6A; Supplemental Fig. 6).

Figure 6.

FoxA1 is necessary for timely reactivation of target genes as cells exit mitosis. (A) RT-qPCR analysis of primary transcripts in HUH7 cells treated with siRNAs for FoxA1 (si6689) or a control siRNA blocked in mitosis and released for various time points. Data are normalized to Gapdh and to the “block” time point prior to mitotic release. The first two genes in each row exhibited a net increase in synthesis over time, while the others exhibited an initial burst of synthesis. The top row depicts genes that are bound by FoxA1 in mitosis; these are dependent on FoxA1 for late telophase reactivation. The middle row depicts genes that are bound by FoxA1 only in asynchronous cells, many of which depend on FoxA1 for its initial activation. The bottom row depicts genes that are not bound by FoxA1 and are independent of FoxA1 for early reactivation. See Supplemental Figure 5 for more genes of each type. Error bars denote SEM; asterisks indicate significance by a one-tailed Student's t-test: (*) P < 0.05; (**) P < 0.01. (B) The average nascent transcript induction at 105 min post-mitotic block release is shown as a ratio of that in the presence of the FoxA1 siRNA over that for the control siRNA for genes in the three categories in A and Supplemental Figure 5. Error bars denote SEM, and asterisks indicate significance by a one-tailed Student's t-test: (*) P < 0.05; (**) P < 0.01. A separate Mann-Whitney test, not shown, revealed that the FoxA siRNA selectively perturbed mitotic FoxA1-bound versus unbound to P < 0.00013 and asynchronous-only bound versus unbound to P < 0.022. The data show that FoxA1 facilitates the early reactivation of genes bound in mitosis as well as genes bound solely in interphase cells and does not indirectly enhance the reactivation of genes to which it does not bind.

Regardless of the initial re-expression pattern, most of the genes that were bound by FoxA1 in mitosis exhibited a statistically significant dependence on FoxA1 for their initial activation in late telophase (Fig. 6A; Supplemental Fig. 6, top rows), demonstrating that FoxA1 functions as a bookmarking transcription factor. None of the genes that were not bound by FoxA1 in mitosis or interphase were dependent on FoxA1 for early reactivation, demonstrating selectivity for FoxA1 target genes (Fig. 6A; Supplemental Fig. 6, bottom rows). Interestingly, many of the genes that were bound by FoxA1 in interphase but not in mitosis were also dependent on FoxA1 for their initial activation post-mitosis, although with a wider variation in response (Fig. 6A; Supplemental Fig. 6, middle rows). When we grouped all genes assayed in each category at the earliest time point and assessed the average fold induction in the presence of FoxA1 siRNA over that with the control siRNA, both the mitotic-bound and asynchronous-only-bound FoxA1 target genes exhibited a significant difference from the genes not bound by FoxA (Fig. 6B). Similar results were obtained with a less extensive study using a different siRNA to knock down FoxA1 (Supplemental Fig. 7). We conclude that FoxA1's increased mobility in mitotic chromatin (Fig. 4A) yet high nonspecific chromatin-binding capacity (Figs. 5, 6) keep the factor on chromatin and facilitate rapid reactivation during mitotic exit. The most highly expressed FoxA1 targets retain FoxA1 binding in mitosis, whereas most other FoxA1 targets appear to be dependent on nonspecific binding to mitotic chromatin. Importantly, FoxA1 does not indirectly enhance the reactivation of genes to which it is not bound in interphase. We conclude that the pioneer factor FoxA1 has mitotic chromatin-binding features that distinguish its activity from other bookmarking factors that have been characterized.

Discussion

In this study, we demonstrated that FoxA1, a pioneer factor involved in early steps of liver development, also bookmarks targets genes during mitosis. FoxA1 employs two different modes of mitotic binding that contributes to gene reactivation during exit from mitosis. The specific binding mode is responsible for mitotic marking of genes that are highly expressed in interphase, including important liver genes, and the nonspecific binding mode keeps FoxA1 in the vicinity of other target genes by random site retention on mitotic chromosomes. The nonspecific chromatin binding occurs via the high intrinsic affinity of FoxA1 for nucleosomal DNA (Cirillo and Zaret 1999; Sekiya et al. 2009) and the factor's increased chromatin mobility in mitosis (Fig. 4). By retaining FoxA1 on the DNA in mitosis, albeit nonspecifically, the factor has initial access to its specific target sites that would precede what would be observed for factors that do not bind mitotic chromatin and thus could facilitate early gene reactivation during mitotic exit (Fig. 7).

Figure 7.

Specific and nonspecific modes of FoxA1 binding to the mitotic genome allow rapid reactivation of FoxA target genes during mitotic exit.

What determines whether a transcription factor will remain bound to a particular site in mitosis? The answer to this question has eluded the field, and there may be different mechanisms employed by different factors. As seen by other factors (Kadauke et al. 2012), the binding sequence motif for FoxA1 is essentially the same at mitotic-bound sites compared with sites bound only in interphase (Fig. 2D). We failed to discover histone modifications at FoxA1 target sites in asynchronous cells that would predict mitotic binding. However, this negative result could be due to a limitation of the available chromatin modifications currently in the online databases or to a lack of correspondence between the cell lines that have been assayed and HUH7 cells in which we performed our experiments. Importantly, we were able to discern that predicted INOSs (Kaplan et al. 2009) are sufficient to distinguish the subset of sites that remain bound by FoxA1 in mitosis; i.e., the sites that retain FoxA1 binding in mitosis had significantly higher INOSs. While nucleosome mapping in vivo will be required to assess the validity of these predictions, they are striking in light of FoxA1 having a markedly lower off rate for its target sites on nucleosomes compared with free DNA (Cirillo and Zaret 1999). Furthermore, sequential ChIP studies showed that FoxA1 binds a nucleosomal target site in vivo (Chaya et al. 2001). Because many, but not all, other transcription factors lack the high intrinsic affinity for nucleosomal DNA, it seems unlikely that the INOS will be a general predictor of mitotic chromosome binding. However, for this class of protein, it could be useful.

How is nonspecific DNA binding by FoxA1 more prominent in mitosis than in interphase? Our FRAP experiments revealed that despite linker histone moving more slowly in mitotic chromatin, in agreement with prior studies (Chen et al. 2005), the mobility of FoxA1 in chromatin is increased compared with interphase. This observation, coupled with the high intrinsic nucleosome and chromatin-binding capacity of FoxA1, could be sufficient to explain a decrement in specific target residence time with a concomitant increase in nonspecific chromatin binding. As for what causes an increased mobility of FoxA1 in mitotic chromatin, we speculate that a mitosis-specific modification of the protein could play a role, the overall compaction of chromatin could help exclude many specific sites, and/or an altered modification of the chromatin itself could contribute.

By assaying in parallel diverse types of factors critical for a single cell lineage, we discovered a diversity of mitotic chromatin-binding types, including total occupancy, specific and nonspecific binding, partial or total chromosome exclusion, and differential mitotic stability. We note that the earliest transcription factors in liver development, including FoxA1 and HNF1β (Lokmane et al. 2008), are necessary in the endoderm for hepatic induction and exhibit high-level binding to mitotic chromosomes (Verdeguer et al. 2010). In contrast, the other hepatic transcription factors tested exhibit successively less mitotic binding capacity. This raises the possibility that the diverse modes of transcription factor binding in mitosis may mimic binding hierarchies in development. The mitotic binding hierarchy could be a way to ensure the initial exclusion of factors that would otherwise promote an alternate cell fate when a given cell type exits mitosis.

Materials and methods

Mitotic arrest of HUH7 cells

HUH7 cells were plated in fresh DMEM High Glucose with L-glutamine (Invitrogen, 11965), 10% FBS (HyClone), and 1% penicillin–streptomycin (Invitrogen) and grown overnight to 75% confluence. Cells were blocked in S phase by addition of fresh medium containing 2 mM thymidine (Sigma, T1895). After 24 h, cells were washed three times with PBS and released into medium containing 0.06 μg/mL fresh nocodazole (Sigma, M1404). After 18 h, ∼94% of the cells were blocked in metaphase (Supplemental Fig. 3a) and used for imaging or ChIP. Drug washout experiments showed that the cells remained viable and proliferative. For mitotic block release studies, arrested cells were either (1) harvested by gentle shaking and replated into fresh medium or (2) washed three times with PBS, and fresh medium was added to the cells. Cells were allowed to proceed for the periods indicated.

Fluorescence imaging of HUH7 cells

HUH7 cells were plated in glass-bottomed microwell dishes (Mat Tek Corporation) and transfected on the next day with GFP-tagged fusion proteins (Sekiya et al. 2009) with FuGENE 6 (Roche). After 48 h of mitotic arrest, asynchronous and nocodazole-blocked cells were photographed at 100× using a Deltavision Core Deconvolution microscope (Applied Precision) from an Olympus IX70 microscope equipped with a Photometrics CoolSNAP HQ 12-bit monochrome cooled CCD camera. We used a 100× 1.4 NA oil immersion PlanSApo lens (Olympus, #UPLSAPO 100XO) objective with epi-illumination provided by a 300 W xenon arc lamp. The resulting images were deconvoluted using the constrained iterative algorithm with softWoRx (Applied Precision) acquisition software.

For conventional fluorescence microscopy, cells were fixed with 3.7% formaldehyde (Fisher Scientific) for 20 min, washed twice with PBS, permeabilized for 10 min in methanol (Fisher Scientific) at −20°C, and then rinsed three times with PBS. Cells were stained with 4,6-diamidino-2-phenylindole (DAPI) for 1 min. After three washes with PBS, the samples were imaged at 40× using a Nikon Optiphot microscope with an Optronics CCD camera. Alternatively, after the first PBS washes, cells were incubated with a monoclonal antibody against H3phospho-S10 (Abcam, ab14955) diluted 1/375 in 3% FBS, 5 mM KCl, 100 mM NaCl, 0.2% Triton X-100%, 1% BSA, and 10 mM Tris-HCl (pH 7.5) for 2 h at 4°C, then rocked 3× with PBS for 10 min at room temperature and incubated with a secondary antibody conjugated to Alexafluor 488 (Invitrogen; A21202) at 1/500 in dilution buffer for 1 h at room temperature. After rinsing twice with PBS, cells were stained with DAPI and imaged as above. For RNA polymerase II, cells were fixed for 30 min in PBS containing 4% paraformaldehyde and permeabilized for 20 min in PBS containing 0.5% Triton X-100. Blocking was in PBS, 0.1% Triton X-100 (PBS-T), and 10% goat serum (Sigma, G9023) for 1 h. 8WG16 (Abcam, ab817) was diluted 1:1000 in PBS-T and applied overnight at 4°C. Cells were washed three times with PBS for 10 min each and then treated with Alexa Fluor 488 goat anti-rabbit IgG (1:10,000 in PBS-T; Invitrogen, A11001) for 1 h at room temperature. Cells were washed three times with PBS. DAPI (Sigma, D9542) at 0.5 μg/mL PBS was applied to the cells for 2 min. Cells were washed with PBS and imaged at 60× using an Eclipse TE2000-U inverted microscope (Nikon) and CoolSNAP EZ camera (Photometrics).

For immunostaining of endogenous FoxA1, HuH7 cells were washed twice with PBS before fixation in PBS with 4% paraformaldehyde for 12 min at room temperature and then rehydrated in cold PBS, 0.5% Triton X-100 for 10 min. After rinsing twice with PBS, cells were treated with blocking solution containing PBS-T, 4% FBS, 10% glycerol, and 0.1 M glycine for 1 h at room temperature. Primary incubation was done with a goat antibody against FoxA1 (Abcam, ab5089) diluted 1/200 in blocking solution for 1 h at room temperature. Cells then were rocked three timeswith PBS-T for 5 min and incubated with Alexa Fluor 488 anti-goat (1/1000 in blocking solution; Invitrogen, A11055) for 1 h at room temperature. Finally, after rocking the cells three times with PBS-T for 10 min, cells were stained with DAPI and imaged as for H3phospho-S10.

To visualize new RNA synthesis after release from a mitotic block, HUH7 cells were synchronized by thymidine–nocodazole block and released into fresh medium. One hour before the desired time point for fixation, EU was added to 0.5 mM (Jao and Salic 2008). After this 1-h pulse, the cells were fixed and permeabilized as for RNA polymerase II. Cells were washed once with PBS, and the 30-min “click” reaction to Alexa Fluor 594-azide was initiated and later quenched as specified (Click-It RNA Alexa Fluor 594 imaging kit; Invitrogen, C10330). Cells were then washed three times with PBS while protected from light and counterstained with DAPI, washed with PBS, and imaged at 60×.

ChIP

Asynchronous and nocodazole/thymidine-blocked cells were cross-linked with 1% formaldehyde for 20 min at room temperature followed by addition of glycine to 125 mM. After two washes with PBS, cells were collected and frozen. Cell pellets were lysed, and genomic chromatin was sonicated to 200–500 bp (Diagenode Bioruptor). For ChIP-qPCR, 25 μg (for GFP-ChIP) or 35 μg (for FoxA1-ChIP) were precleared with salmon sperm DNA and protein A agarose (Millipore, #16-157). Samples were split, the first aliquot was incubated with 2 μg of rabbit IgG (Abcam, ab46540), and the second aliquot was incubated with 2 μg of rabbit polyclonal to GFP (Abcam, ab290), FoxA1 (Abcam, ab23738), or Histone H3 (Abcam, ab1791). Samples were rotated overnight at 4°C and immunoprecipitated with salmon sperm DNA–protein A agarose and low- and high-salt wash steps. Cross-linked products were reversed, RNase-treated, and DNA-purified. Samples were analyzed on an iCycler iQ multicolor real-time PCR (Bio-Rad) using SYBR Green (Bio-Rad). PCR primer sets are listed below. We used two biological replicates for the GFP-ChIP and three independent replicates for the FoxA1 and H3-ChIP.

ChIP-seq

ChIP was performed on triplicate chromatin samples of 335 μg divided for FoxA1 and IgG immunoprecipitations, and products were used to generate libraries (Illumina, IP-102-1001) of 100–200 bp (Bioanalyzer, Agilent) on an Illumina Sequencer. Additionally, one DNA input library each was made from mitotic and asynchronous chromatin. Sequence reads were aligned to the human genome (NCBI Build 36) with ELAND using default parameters. We uniquely mapped 40,002,286 and 43,176,234 reads for mitotic and asynchronous cells, respectively. Peaks were called for each lane separately with MACS (Mfold parameter 16; false discovery rate [FDR] 5% for mitotic peaks and 0.05% for asynchronous peaks) using the input lanes as background. A more lenient FDR was used to assess mitotic peaks because the majority of reads fell into regions of nonspecific binding (see Fig. 3C), making peak identification for this ChIP more difficult. The trends observed in Figure 2, C and F, are also observed when both peak sets are filtered at 5% FDR (data not shown). Peak sets from each replicate were intersected, and peaks that share ≥50% sequence were merged to define distinct binding sets. RefSeq transcripts were classified as targets if a FOXA1 peak was present in the gene body or within 20 kb upstream of the transcription start site. Data track images of peaks were constructed by pooling sequence reads from all replicate lanes, assessing tag counts at each position, normalizing the count per million aligned reads, and subtracting input. These counts were written into a wiggle file and uploaded to the University of California at Santa Cruz Genome Browser. Whole-chromosome views were created similarly, except that sequence tags were binned at 25-bp intervals prior to normalization. Motifs were discovered using MEME and TOMTOM. The presence or absence of the human FoxA1-binding sequence was assessed using the available site matrix from JASPAR (MA0148.1). ChIP-seq and gene expression array data have been uploaded to Gene Expression Omnibus.

Intrinsic nucleosome occupancy calculation

The INOS based on the Lasso algorithm (Kaplan et al. 2009; Tillo and Hughes 2009; Tillo et al. 2010) was calculated for evaluation of intrinsic nucleosome occupancy. For each 1500-bp sequence, we calculated the INOSs for each slide window of 147 bp and moved the window base by base to get the profiling of INOSs. The control set was 100,000 sequences randomly selected from hg18.

Gene expression microarrays

Total RNA was collected from three different plates of asynchronous Huh7 cells using the RNeasy minikit (Qiagen). Expression microarrays were performed with a Human Gene 1.OST array (Affymetrix) at the University of Pennsylvania Microarray Core Facility and were evaluated using Partek.

Western blot analysis

Primary antibodies used were as follows: FoxA1 (0.001 mg/mL; Abcam, ab23738), C/EBPa (1:400; Santa Cruz Biotechnology, sc-61), and vinculin antibodies (1:5000; Sigma-Aldrich, V-9131). Secondary antibodies used were as follows: goat anti-rabbit IgG (H+L)-HRP (1:5000; Bio-Rad, 1706515) and goat anti-mouse IgG-HRP (1:5000; Santa Cruz Biotechnology, sc-2005).

FRAP assays

Cells cultured in glass-bottomed microwell dishes were transfected with GFP-tagged proteins and arrested in mitosis. The dishes were mounted onto a spinning-disk confocal microscope with a Yokogawa CSU X1 scan head and an Olympus IX 81 microscope. The cells were kept at 37°C using an Okolab Uno incubator. Acquisition and hardware were controlled by MetaMorph, version 7.7 (Molecular Devices). An Andor iXon3 897 EMCCD camera (Andor Technology) was used for image capture. Solid-state lasers for excitation (488 nm for GFP) were housed in a launch constructed by Spectral Applied Research. An Olympus 100×, 1.4 NA UPlanSApo oil immersion objective was used for all experiments.

FRAP was performed using the iLas² system (Roper Scientific), using a 50-mW diode-pumped crystal laser at 405 nm (CrystaLaser, model DL405-050-O) controlled by MetaMorph. Laser power for bleaching was attenuated to 7.5%. For each experiment, four to six single-prebleach images were acquired before an area of 1.5 μm² was bleached. Images were collected over 150 sec (FoxA1) or 227 sec (H1) every 0.3 sec. At least nine cells were analyzed for each time.

To calculate the relative fluorescence intensity (RFI) in the bleached area at time t, we used the equation RFI (t) = [I_B (t) − I_BG (t)/I_B (t₀) − I_BG (t₀)] [I_U (t₀) − I_BG (t₀)/I_U (t) − I_BG (t)]. I_B (t) is the fluorescence intensity in the bleached region at time t, I_BG (t) is the fluorescence intensity in an area containing no cells (background) at time t, I_U (t) is the fluorescence intensity in a nonbleached region in the same cell at time t, I_B (t₀) is the fluorescence intensity of the bleached region before bleaching, I_BG (t₀) is the fluorescence intensity in the region containing no cells before bleaching, and I_U (t₀) is the fluorescence intensity in the nonbleached area before bleaching. Curve fitting was performed using Prism 5 (Graphpad Software). Error bars from individual time points represent standard error of the mean (SEM).

FoxA1 knockdown

Huh7 cells at 30% confluence were reverse-transfected with 3 nM siRNA (FoxA1 ID: si6689) (Fig. 6; Supplemental Figs. 5, 6) or s6687 (Supplemental Fig. 7) and negative control #1 Silencer Select siRNAs (Ambion) using RNAiMAX (Invitogen, 13778-075). After 58 h, mitotic arrested cells were harvested by shake-off and plated with fresh medium. Total RNA was collected at the time points indicated using the RNaEasy microkit (Qiagen). Nascent mRNA of target genes were analyzed by real-time PCR using SYBR Green (Applied Biosystems). Primer sets (see below) were designed to detect primary transcript, nonspliced mRNA by spanning intron–exon junctions and the adjacent exon. Expression levels were normalized to the levels of spliced GAPDH. The data are represented as fold induction over time 0 h: 2^ΔΔCt = 2^{Ctgene_t − CtGapdh_t}/2^{Ctgene_t0 − CtGapdh_t0}. Error bars indicate SEM. One asterisk denotes P < 0.05 and two asterisks represent P < 0.01, calculated using a one-tailed Student's t-test.