Genome-Wide Maps of Histone Modifications Unwind In Vivo Chromatin States of the Hair Follicle Lineage

Purifying Quiescent and Activated SCs and Their TA Progeny from Adult HFs

To ascertain the possible physiological relevance of chromatin modifications to the program of gene expression in adult SCs and their lineages, our aim was to map mRNA expression profiles and global chromatin occupancies of key histone modifications in HFs in their native context. The technological demands of ChIP-seq were met by the synchrony of HFs, enabling us to purify 10⁷ cells of each of three desired populations for our study: quiescent HF-SCs (qHF-SCs), activated HF-SCs (aHF-SCs) and TA-matrix cells (HF-TACs) (Figure 1A).

For each experiment, we used fluorescence activated cell sorting (FACS) and ~10–15 adult female CD-1 mice transgenic for K14-GFP, active in skin epithelium (Rendl et al., 2005). qHF-SCs were isolated from telogen HFs in the midst [postnatal day P52–58] of a 3 week resting phase. While the aHF-SC population will always contain quiescent SCs, appreciable SC cycling exists in the earlymid growth phase (anagen) (Hsu et al., 2011). Hence we chose P28–30 for isolating aHF-SCs and HF-TACs. Populations were sorted based on GFP fluorescence and Abs against differentially expressed surface proteins (Figure 1B). In addition to established markers of bulge HF-SCs (CD34, α6 integrin; Blanpain et al., 2004), Sca1-Abs were used to remove/discard basal epidermal and upper ORS cells, and ephrin B1-Abs enriched for HF-TACs. Immunofluorescence of cytospin samples confirmed >95% purity for each population.

Molecular Signatures of HF Lineage States

Total RNAs from each population were analyzed in duplicate by microarray for expression of ~30,000 mouse genes. Comparisons provided molecular signatures of mRNAs ≥2X increased in one population relative to the others (p<0.02). The signatures are presented as a heatmap (Figure 1C) and were confirmed by reverse transcription (RT) and quantitative (q)-PCR (Figure S1).

As expected from previous studies (Lowry et al., 2005), mRNAs expressed by aHF-SCs and qHF-SCs overlapped significantly (n=672) and included key HF-SC TFs (Figure S1A). In addition, genes involved in cell signaling appeared on this list (Figure 1D; Table S1). The qHF-SC-specific signature (n= 651) was enriched for balancers of cell survival/apoptosis and negative regulators of gene expression (p<0.001). The aHF-SC signature (n=231, anagen bulge alone) favored mRNAs associated with cell migration/proliferation (p<0.001) (Figure 1D; Table S1). The HF-TAC signature was the most highly enriched for cell cycle-enhancing as well as HF lineage/differentiation mRNAs, e.g. Lef1, Msx1 and Cux1 (Figure S1B; Table S1; p<0.001). These new comparisons provided a more comprehensive and clearer picture than previously of the changes in mRNA expression that occur as HF-SCs transition from a quiescent to activated and finally TA state. Importantly, they also established steady-state expression profiles for each population to subsequently compare to their chromatin states.

Genome-wide Profiles of H3K4me3 and H3K27me3 Unveil Differences Between qHFSCs and ESCs

In cultured ESCs, genes encoding key master regulators of lineages are bivalent/poised, i.e. marked by both H3K4me3 and by H3K27me3; in embryonic fibroblasts (MEFs), these genes are partially resolved; in neural progenitor cells (NPCs), they are largely resolved to either a fully-repressed or fully-activated state (Figure 2A; Lee et al., 2006; Mikkelsen et al., 2007). Given these in vitro findings and the potential importance of PcG modifications to adult SCs in vivo, we used ChIP-seq to determine the global progenitor H3K4me3 and H3K27me3 modification profiles from our purified skin populations.

90% of H3K4me3 peaks identified in qHF-SCs were located within 2kb of promoter regions of known genes/transcripts (Table S2). Moreover, similar to in vitro lineage-committed progenitors and in striking contrast to ESCs, qHF-SCs in vivo displayed very few genes (89/15,284) in a bivalent state (Figure 2A). These results added in vivo relevance to the prior notion that the bivalent, “poised” state may be uniquely tailored to enable ESCs to maintain the flexibility to choose among a myriad of lineage options (Bernstein et al., 2006; Boyer et al., 2006; Lee et al., 2006; Mikkelsen et al., 2007).

The majority of promoters bivalent in ESCs were resolved to either H3K4me3 alone (39%) or H3K27me3 (38%) in qHF-SCs (Figure 2B). Interestingly, genes bivalent in ESCs but PcG-repressed in qHF-SCs included not only key genes in non-skin tissue-specific lineages, but also hair lineage differentiation (Figure 2C; Table S3). In contrast to bivalent genes, most ESC genes marked only by H3K4me3 were similarly marked in qHF-SCs (Figure 2B). Predictably, ChIP-seq showed that RNA polymerase II also marked most of these genes (data not shown). Since the majority of these singly marked genes were also activated in MEFs and NPCs, this list seemed to be enriched for features (including housekeeping functions) common to progenitor states and irrespective of cycling status.

Only 2.3% of genes marked by only H3K4me3 in ESCs were transcriptionally repressed by H3K27me3 in qHF-SCs (Figure 2B). That said, this short-list was informative, as exemplified by its inclusion of Sox2 (Figure 2D), which along with Oct4, Nanog and Tcf3, establishes the genomic state that maintains ESC pluripotency (Bernstein et al., 2006; Boyer et al., 2005; Boyer et al., 2006; Lee et al., 2006). Only Tcf3, a transcriptional effector of Wnt signaling that intersects with the ES core circuitry (Cole et al., 2008), was also actively transcribed in HF-SCs and also critical for their maintenance (Figure 2D; Nguyen et al., 2009).

Surprisingly, despite the silenced state of pluripotency genes in HF-SCs, only a small subset of the genes bound by Sox2/Oct4/Nanog and H3K4me3-marked in ESCs (Cole et al., 2008; Marson et al., 2008) were PcG-repressed in qHF-SCs (Figure 2E). Most, including Tcf3 and many other chromatin regulators, were active in both SC types. Conversely, of the HF-SC signature genes, >60% also scored as active in ESCs, including cell cycle regulation and apoptosis/survival genes (not shown). Together, these data suggest that a common set of chromatin, transcriptional and cell cycle/survival regulators may govern both pluripotent, rapidly-proliferating ESCs and lineage-restricted, slow-cycling HF-SCs. Moreover, since Sox2, Oct4 and Nanog are not expressed by HF-SCs, other TFs must contribute to the activation of these core stemness target genes in adult SCs.

Identifying Key HF Regulators from Aspects of Histone Modification

Since bivalent ESC genes include developmental regulators required for lineage specification, it seemed likely that key regulators of HF-SC specification and maintenance might be bivalent in ESC and H3K4me3-marked genes in qHF-SCs. Consistent with this notion, TF family genes that were bivalent in ESCs and PcG-derepressed in qHF-SCs were often featured in the HF-SC signature (Figure 3A; Table S4A). This included Sox9, Lhx2, Gata3, Runx1, Dlx3 and Hoxc13, whose functional importance to the hair lineage is known (Fuchs, 2007) (Figure 3B; Table S4B). The significant enrichment of established regulators led us to wonder whether less explored members of this list, e.g. TF genes Foxc1, Klf4, Irx and Tbx1, or signaling factor genes e.g. Fgf18, Grem1, Fst and Inhbb, might also play critical roles in HF-SC behavior.

In ESCs, key pluripotency genes such as Oct4 (Pouf5f1) display expanded H3K4me3 peaks over their promoters (Mikkelsen et al., 2007), and in hematopoietic progenitors, broad H3K4me3 domains coincide with known hematopoietic regulators (Adli et al., 2010). To address whether similar features might be characteristic of key HF-SC regulators, we analyzed H3K4me3 peak sizes in the promoter regions of qHF-SC genes. 180 qHF-SC genes displayed expanded peaks (>4kb) and were highly enriched for HF-SC signature genes (47% vs 10%; p<0.001). Genes encoding HF-SC marker proteins, e.g. CD34 and K15, were included in this group, as were essential HF-SC TF genes Lhx2, NFATc1 and Sox9 (Figure 3C). The sheer number of TFs (32; 38%) in this group was notable, given that only 10% of all mouse genes encode TFs. Curiously, many of the same signaling factors noted above also appeared on this list.

Transitioning to HF-TACs Involves PcG-Mediated Repression of HF-SC Genes

To determine whether and how epigenetic changes might be involved in the HF lineage, we developed genome-wide maps of histone modifications for quiescent vs activated HF-SCs and for HF-TACs. For this study, we included H3K79me2, which distinguishes initiated (H3K4me3 only) from actively elongating (H3K4me3+H3K79me2) genes (Vakoc et al., 2006). As expected, promoters of genes marked by H3K79me2 in our populations were always modified by H3K4me3 as well (Figure 4A). Conversely, ~20–30% displayed only the H3K4me3 mark and not the H3K79me2, even when analyses were restricted to genes preferentially transcribed in these populations (Figure 4B).

By comparing chromatin and mRNA expression data, we learned that genes displaying the H3K79me2 mark were on average transcribed at higher levels than those with only the H3K4me3 mark (Figure 4B, S2A). This analogy seemed to be meaningful, since 31% of H3K4me3-marked genes in HF-SCs gained the H3K79me2 mark in HF-TACs (Figure 4C), and these genes were more highly expressed than those retaining solely the H3K4me3-mark in HF-TACs (p<0.001; Figure S2B). These data suggested that the acquisition of an H3K79me2 mark can provide an epigenetic readout to gauge transcriptional levels within a lineage.

Among qHF-SC signature genes marked by H3K4me3 ± H3K79me2, only 12.5% became PcG-silenced in HF-TACs (Figure 4B). Intriguingly, 66.5% of this subset exists in a bivalent repressed state in ESCs. Underscoring the potential importance of this group was the inclusion of established HF-SC genes e.g. CD34, Sox9, Nfatc1 and Fgf18. Intriguingly, the putative cell signaling regulators, Grem1, Fst and Gdf10 surfaced again on this list (Figure 4D; Table S5).

The importance of PcG-mediated chromatin modifications in repressing SC regulators upon HF-TAC commitment was further highlighted by comparing histone modification peak intensities on key PcG-regulated HF-SC genes versus housekeeping genes expressed in both SCs and TACs. As shown in Figure 4E, key HF-SC signature genes were enriched for H3K4me3 and H3K79me2 modifications in HF-SCs, but showed H3K27me3 repression in HF-TACs.

HF-TAC Commitment Involves Derepression of PcG-Silenced Key Matrix Regulators

Before terminally differentiating along one of 7 different HF lineages, matrix cells undergo a brief spurt of proliferation as undifferentiated cells. During this time, HF-SC regulators are repressed while a unique set of matrix regulators are induced (Blanpain and Fuchs, 2009). To determine whether the matrix induction switch is also governed by PcG, we compared the chromatin state of HFTAC signature genes in matrix, where they are active to the bulge, where they are silenced.

77% of HF-TAC signature genes displayed H3K4me3 and H3K79me2 modifications, reflective of a highly transcribed state (Figure 5A). Many of these genes were “induced” (acquired H3K79me2 mark; n=312) during the qHF-SC to TAC switch, while only a small subset were “derepressed” (lost H3K27me3 mark; n=18) from a PcG-repressed state (Tables S6). Bioinformatics analyses revealed that the “induced” TAC subset was composed mostly of cell cycle-related genes, whereas the “derepressed” PcG group encompassed known regulators of matrix biology (e.g. Lef1,Cux1, Sp6, Runx2, Msx1) (Figure 5B).

Peak intensities of histone modifications further highlighted the dynamic chromatin states of HF-TACs (Figure 5C). Consistent with the markedly enhanced proliferation in HF-TACs, cell cycle genes were transcribed at much higher levels than in qHF-SCs (compare intensities of H3K79me2 marks). However, the most striking chromatin differences occurring as HF-SCs transitioned to HF-TACs were in H3K27me3 marks, where key matrix genes were PcG-derepressed, and key HF-SC genes were repressed.

Maintaining Stemness during the Transition from a Quiescent to Proliferative State: Inducing Cell Cycle Regulators without Affecting Key PcG-regulated Matrix Genes

In striking contrast to HF-TACs, genes expressed by aHF-SCs retained many of the same modifications and displayed similar mRNA patterns as their quiescent counterparts (Figures 6A and 1). Notably, aHF-SCs showed no signs of the PcG-silencing of key HF-SC genes that occurred in HF-TACs (e.g. Sox9 in Figure 6B). Interestingly, while only 3.3% of HF-SC signature genes acquired PcG-repressive marks in aHF-SCs (Figure 6A), this small group included genes that encode known regulators of HF-SC quiescence (e.g. Nfatc1 and Fgf18) (Figures 6B’, S3A; Table S7A). Although the overall level of H3K27me3 modification over these genes was weak, this is likely reflective of the fact that only a fraction of the HF-SC population becomes activated during anagen (Hsu et al., 2011).

Interestingly, more HF-TAC signature genes gained the H3K79me2 mark in aHF-SCs than in qHF-SCs (42% vs 29%; Figure 6A’; Table S7B). Many of these appeared to be non-PcG-regulated cell cycle genes (Figure S3A’). Comparing all three populations, a hierarchy in cell cycle gene transcription and in mRNAs emerged: HF-TACs > aHF-SCs > qHF-SCs (Figures 6C, S3B). In contrast to cell cycle genes, H3K27me3-marked key HF-TAC regulators remained PcG-repressed in aHF-SCs (Figure 6C’, S3B). These results suggest that the qHF-SC to aHF-SC transition is controlled largely by genes changing from initiating to elongating states.

Since Wnt signaling/β-catenin stabilization is associated with anagen onset and transgenic mice which express elevated stabilized β-catenin (ΔN-βcat) display a shortened telogen, we asked whether elevated β-catenin signaling in qHF-SCs is sufficient to de-repress PcG-marked key HF-TAC genes which are essential for the cell fact switch. To test this possibility, we performed ChIP-qPCR on key matrix regulator genes (Figure S4A). In agreement with their mRNA expression pattern (Figure S4B), key matrix regulator genes remained repressed while cell cycle genes, frequently upregulated during SC activation, remained un-marked by H3K27me3 in ΔN-βcat(+) qHF-SCs (Figure S4A). These data suggest that stabilization of β-catenin promotes transcription of a number of non-PcG-regulated cell cycle genes involved in the process of HF-SC activation. The bar for switching PcG-regulated fate determiners may require more robust signaling and/or additional pathways. The model in Figure 6D summarizes our findings on histone modification aspects in HF-SC activation and the TAC fate switch.

Loss of H3K27 Trimethylation Elicits Early Transcriptional Changes Which Precede Phenotypic Signs of Flipping the PcG-Mediated HF Fate Switch

Previously, we conditionally targeted mice Ezh1/2 resulting in a lack of H3K27me3 in the skin epithelium of P0 mice (Ezhkova et al., 2011). Although HFs had developed in these mice and no phenotypic signs of a fate-switch defect had been observed, this issue merited revisiting now that we had identified the key fate-switch genes that are PcG-regulated. To do so, we conducted mRNA profiling and qPCR analyses of purified populations of Ezh1/2 null vs WT ORS/bulge and matrix. Intriguingly, in the absence of Ezh1/2, mRNAs expressed by key matrix genes were significantly upregulated in ORS/bulge, and conversely key HF-SC regulators failed to completely shut off in matrix (Figure S5). That said, aberrant expression of these key PcG-regulated genes occurred at markedly lower levels than in the natural HF-SC to HF-TAC switch (shown). These results are reminiscent of those obtained previously with PcG-regulated non-skin genes (Ezhkova et al., 2011). Most importantly, however, they show that like Wnt activation, H3K27me3 loss is necessary but not sufficient to flip the powerful PcG switch that sets the high threshold in HF-SCs for TAC fate commitment.

Exploiting Histone Modifications to Uncover New Insights into the Balance between Quiescence and Activation in the HF Stem Cell Niche

Our findings led us to wonder whether our knowledge of histone modification patterns might be used to facilitate the identification of key HF-SC regulators. We posited that these genes should be enriched in the subset that are a) bivalent in ESCs; b) actively transcribed in HF-SCs; c) PcG-repressed in HF-TACs; and in some cases, d) marked by a broad H3K4me3 peak. Interestingly, this increasingly shorter list still contained several TFs (Sox9, Nfatc1), transmembrane proteins (Cd34, Lgr5) and regulatory growth factors (Fgf18), already linked to HF-SC behavior (Figure 7A). Most notable was the persistence of several signaling genes, which although functionally untested in HFSCs, are predicted to function in BMP and Activin pathways.

A scan of histone modification and mRNA profiles indicated that the surface receptor genes for FGF, BMP and Activin pathways are all actively transcribed in both quiescent and activated HF-SCs (not shown). Figure 7B summarizes their corresponding ligands identified from our short-list, and indicates how we posit they might be involved in HF-SCs based upon what is known (Blanpain and Fuchs, 2009; Greco et al., 2009; Hsu et al., 2011).

We first focused on Grem1 and Gdf10, encoding putative BMP inhibitory factors. HF-SCs in vitro are growth-restricted when exposed to 400 ng/ml BMP6 (Blanpain et al., 2004). Interestingly, BMP6-mediated HF-SC growth inhibition was significantly ameliorated by co-exposure to either growth factor, with Gremlin having the most potent effect (Figure 7C). The counteracting effects of these factors were reflected in the restoration of HF-SC cycling, as measured by 1 hr exposure to the nucleotide analogue EdU prior to harvesting. Moreover, the antagonizing effects of Gremlin on BMP6 signaling were dose-dependent (shown). Since both Grem1 and Gdf10 genes are PcG-repressed after HF-SCs transition to the activated state (Figure S6A), we speculate that these cell autonomous, BMP-inhibitory factors might function with DP-expressed Noggin to eventually override the high BMP6 levels imposed by the K6 inner bulge cells (Hsu et al., 2011) and activate HF-SCs (see Figure 7C).

Previously it was reported that Activin-antagonist Follistatin are expressed by human HF-SCs (Ohyama et al., 2006), and Activin in vivo function in hair cycling (Qiu et al., 2011). Intriguingly, Inhbb encoding Activin B (activinβB-activinβB), and Fst encoding Follistatin (Figure 7B), were both marked by H3K79me2 and also displayed broad H3K4me3 peaks throughout the transition from HF-SC quiescence to activation (Figure S6B). Thus, we next focused on the possible relevance of this pathway.

Recombinant Activin B had a potent inhibitory effect, significantly reducing HF-SC proliferation in vitro in a dose-dependent fashion (Figures 7D, S7A). At 500 ng/ml, 90% of HF-SC growth was restricted, an effect that was significantly counter-balanced by Follistatin in a dose-dependent fashion (Figures 7D, S7B). Interestingly, Follistatin is also known to interfere with BMP signaling, albeit with lower affinity/potency. These data provide new insights into why Fst null embryos show retarded HF development (Nakamura et al., 2003) and suggest that its stimulatory effects may be required in adult mice to overcome the intrinsic inhibitory effects of Activin B signaling in HF-SCs (Figure 7D).

Cell Type vs Lineage Fate Switches

By monitoring chromatin marks in an adult SC population through stages of quiescence, activation and lineage commitment, we discovered that in contrast to ESCs, only a small number of genes are bivalently marked by both H3K27me3 and H3K4me3. Based upon similar in vitro findings on cultured neural cells differentiated from ESCs, our data suggest that bivalency may no longer be needed once tissue lineages have been selected and cell type options have been restricted.

Despite minimal signs of bivalency, PcG-mediated repression and derepression featured prominently in HF-SC and lineage genomes. As expected from prior studies made on individual genes, our global analyses showed that non-skin lineage master transcriptional regulators are uniformly silenced by H3K27me3 in the HF lineage. More interesting and unexpected were the select group of regulators involved in HF-SC lineage progression which were marked by H3K27me3. Some of these genes--the matrix regulators--remained PcG-repressed in both quiescent and active HF-SCs, and were only derepressed in HF-TACs. Others--the HF-SC maintenance genes--remained active in HF-SCs and were only repressed in HF-TACs. Taken together, these findings unveil PcG-mediated repression/derepression as a double-edged sword in the lineage commitment switch from stemness to differentiation. Moreover, the congruent timing of these events suggests that the mechanism for repression and derepression is integrally linked.

It was also intriguing that many of the PcG-regulated genes in the HF lineage existed in a bivalent state in ESCs. Based upon these findings, it could be that the foundations for the PcG switch are established during early embryogenesis, and then modified during development to suit the tissue-specific needs of adult SCs. This finding is interesting in light of recent studies on epigenetic reprogramming of induced pluripotent SCs, where H3K27me3 marks remained largely unchanged in the face of thousands of changes in H3K4me2 (Koche et al., 2011).

From Histone Methylation to Gene Expression to Functional Significance

In recent years, it has become increasingly clear that RNA polymerase II occupancy of the transcription initiation site and the accompanying H3K4me3 chromatin modification is necessary but not sufficient to dictate transcription, and that “mediator” complexes associate with pausing factors to control elongation (Conaway and Conaway, 2011; Kagey et al., 2010). Although reliable measures of transcriptional elongation, such as H3K79me2 and H3K36me3, are now available (Vakoc et al., 2006), chromatin modifiers have been less optimal in gauging a gene’s relative level of expression, even with the recent addition of so-called “chromatin readers” which fine-tune chromatin surrounding modified histone marks (Vermeulen et al., 2010).

While the use of chromatin modifications as a blueprint for determining transcript levels is still in its early stages, we gained several new insights by comparing relative mRNA levels with associated chromatin modifications as HF-SCs transition through dormant and proliferative states and finally lineage commitment. Most notably, a gene more highly transcribed at one time within the lineage often displayed a broader H3K4me3 peak accompanied by stronger H3K79me2 marks throughout the body of the gene than when at a stage where its expression was lower. Possibly relevant to our findings are those on Arabidopsis, where highly dense and narrow distributions of H3K4me3 near transcriptional start sites were associated with constitutive expression of genes involved in translation, while broad distributions toward coding regions correlated with expression of genes involved in key plant processes e.g. photosynthesis (Ha et al., 2011).

Indeed our data revealed that by screening the genes in our HF-SC signature first for bivalency in ESCs, then for H3K79me2-marked transcriptional elongation, often with a broad H3K4me3 peak, and finally for PcG repression in TACs, a short-list of candidates sharing these histone modification criteria could be generated that was highly enriched for established HF-SC regulators. We exploited these data by functionally testing several signaling pathway genes on this short-list. We focused on candidates whose corresponding surface receptor genes also showed H3K4me3 and H3K79me2 marks, suggesting that they could be cell-autonomous regulators of HF-SCs. Although further studies will be necessary to fully understand the roles of these signaling pathway members in HF-SC biology, our functional assays were sufficient to illustrate the power of using histone modification patterns as a blueprint for honing in on the key gene expression changes that occur as HF-SCs embark upon lineage commitment. Placed in context with the current literature, our results suggest a newfound complexity in the checks and balances imposed by intrinsic and extrinsic factors and signaling pathways on the HF-SC niche (Figure 7E).

The Role of PcG-mediated Repression/Derepression in Lineage Commitment and Cell Fate Determination

The relative paucity of PcG-mediated changes in transitioning from quiescent to activated HFSCs was as dramatic as the switch in PcG-repressed/relieved genes upon transition from HF-SCs to TA-progeny. Additionally remarkable was the considerable threshold to PcG-mediated change that persisted in the face of enhanced stabilized β-catenin in HF-SCs. Indeed, while telogen HF-SCs exhibited transcriptional changes and increased proliferation in response to elevated β-catenin, the affected cell cycle genes did not seem to be ones marked by H3K27me3. By contrast, PcG-regulated cell fate commitment genes remained largely silent.

Our findings suggest the following mechanistic framework: By utilizing transient and/or low Wnt/β-catenin to activate cell-cycle genes that are predominantly free of PcG-mediated control, HF-SCs can transition between quiescence and self-renewal while maintaining their undifferentiated state. By requiring PcG modifications to regulate critical cell fate commitment genes, activated HF-SCs then must undergo a second step to overcome this threshold and transition to a point of no return. Finally, the PcG gatekeeper in this key decision is a two-way switch, reinforcing the likelihood that it will be irreversible.

How is this switch in PcG-target genes achieved during lineage commitment? Elevating the levels of stabilized β-catenin does not seem to be sufficient, nor does relieving PcG modification by loss of the H3K27 methylases Ezh1 and Ezh2. However, we know that during the normal hair cycle, the switch happens at the base of the mature HF, when prolonged interactions occur between DP and matrix. Taken together, additional and/or elevated signals arising from this sustained crosstalk are likely to contribute with Wnts to elicit the SC→TAC transition. By placing a strong PcG-regulatory switch at cellular signaling crossroads, the commitment step involving key regulatory TFs becomes distinguished from the events that merely control HF-SC proliferation and don’t change fate.

The requirements for multiple signaling inputs and transcriptional effectors in the switch mechanism provides us with a clearer understanding of why loss of PcG H3K27me3 modifications in the HF did not result in massive rapid activation of PcG-regulated genes, nor did it appear to cause a phenotypic fate switch in the HF-SC or matrix compartments (Ezhkova et al., 2011). Moreover, although HF-SC and matrix TAC proliferation were severely crippled by a marked deregulation of the PcG-regulated Ink4a/Ink4b/Arf locus upon Ezh1/2 loss-of-function (Ezhkova et al., 2011), we found that changes in the Ink4a/Ink4b/Arf locus do not come into play during either normal HF-SC activation or lineage progression. Moreover, most cell cycle genes transcribed upon Wnt-mediated activation of HF-SCs and further upregulated during the transition to HF-TACs, did not appear to be under PcG control.

In closing, our findings show that adult SCs in the HF become active and self-renew by mechanisms that appear to involve mostly non-PcG regulated genes, while a bimodal switch involving PcG-regulation is critical to control fate determination and lineage progression. In normal homeostasis, this rheostat does not trigger until multiple signaling inputs converge on the activated HF-SCs, unveiling a requirement for both transcriptional effectors and histone modifications in governing the complex processes of tissue homeostasis and regeneration.

Mice

Mice were housed in the AAALAC accredited Comparative Bioscience Center (CBC) at The Rockefeller University in accordance with NIH guidelines.

ChIP-seq Assay

All materials and methods for ChIP-seq have been described (Lee et al., 2006; Marson et al., 2008). Independent immunoprecipitations were performed on FACS-sorted populations from female mice. 3×10⁶ cells were used for each ChIP-seq run. Sorted cells were cross-linked by addition of 1/10th volume of fresh 11% formaldehyde solution for 10 minutes (min.) at room temperature (r.t.) and then rinsed twice with 1X phospho-buffered saline (PBS) prior to freezing in liquid nitrogen and storing at – 80°C. Before ChIP, cells were resuspended, lysed and sonicated to solubilize and shear cross-linked DNAs. For sonication, lysates were treated with Triton X-100 to 1% and then subjected to a Bioruptor Sonicator (Diagenode, UCD-200) according to a 30X regiment of 30 seconds (sec.) sonication followed by 60 sec. rest. The resulting whole cell extract was incubated overnight at 4°C with 10 µl of Dynal Protein G magnetic beads (Invitrogen) which had been pre-incubated with ~5 µg of the appropriate Ab. After ChIP, samples were washed with low salt, high salt, LiCl and Tris-EDTA buffer for 5 min. Bound complexes were eluted and crosslinking was reversed by overnight (o/n) incubation at 65°C. Whole cell extract DNA was also treated for crosslink reversal.

ChIP-Seq Sample Preparation and Analysis

ChIP DNA was prepared for sequencing (Marson et al., 2008) with a few modifications. Briefly, a single adenine nucleotide overhang was added to the fragmented DNAs to allow for directional ligation. A 1:500 dilution of the Adaptor Oligo Mix (Illumina) was used in the ligation step. A subsequent PCR step with 25 amplification cycles added the additional Solexa linker sequence to the fragments to prepare them for annealing to the Genome Analyzer flow-cell. After amplification, a narrow range of fragment sizes was selected by separation on a 2% agarose gel and excision of a band between 150–300 bp. The DNA was purified from the agarose and diluted to 10 nM for loading on the flow cell. Sequencing was performed on an Illumina/Solexa Genome Analyzer 2 following manufacturer protocols.

ChIP-Seq reads were aligned to the mouse genome (mm9, build 37) using the Bowtie program within the Illumina Analyzer Pipeline (Langmead et al., 2009). Unique reads mapped to a single genomic location (allowing two mismatches) were kept for peak identification using the software MACS (Zhang et al., 2008). Sequencing reads from input DNA were used as control for MACS. Peak heights were subsequently computed with in-house software and used to filter out false positive peaks of low signal intensities. Annotated mouse RefSeq genes with a peak at their promoters (−2kb to +2kb of transcription start sites) were then considered as targets of K4, K27 or K79 modifications. Genes without a modification mark in any of our cell types were excluded in our statistics. To compute the level of a histone modification, we summed the total of ChIP-Seq reads mapped to individual promoters and this total was then normalized with sequencing depth to obtain comparable data across cell types. To identify genes with large H3K4me3 peaks in qHF-SCs, we utilized the peak sizes from MACS and selected for peaks that were >4 kb and whose sizes were >3X larger in qHF-SCs than in matrix cells (or no peak identified in matrix cells). Data have been submitted to http://www.ncbi.nlm.nih.gov/geo.