PGE2 alters chromatin through H2A.Z-variant enhancer nucleosome modification to promote hematopoietic stem cell fate

Significance Hematopoietic stem and progenitor cells (HSPCs) require tight transcriptional regulation. In-depth knowledge on how epigenetic mechanisms regulate chromatin landscape in order to control gene expression, particularly in response to external stimuli, will be key to understand the pathogenesis of hematologic disorders and improve blood stem cell-based therapies. Prostaglandin E2 (PGE2) and its metabolically resistant analog 16,16-dimethyl-PGE2 (dmPGE2) were previously identified as a potent external stimulus to enhance HSPC engraftment. PGE2 has been evaluated in various clinical trials to improve engraftment of stem cells. We identified that dmPGE2 mediates chromatin flexibility at HSPC-specific enhancers through histone-variant H2A.Z acetylation to promote master transcription factor (TF) binding and reinforce expression of genes involved in stem cell fate and engraftment.


Significance
Hematopoietic stem and progenitor cells (HSPCs) require tight transcriptional regulation. In-depth knowledge on how epigenetic mechanisms regulate chromatin landscape in order to control gene expression, particularly in response to external stimuli, will be key to understand the pathogenesis of hematologic disorders and improve blood stem cell-based therapies. Prostaglandin E2 (PGE 2 ) and its metabolically resistant analog 16,16-dimethyl-PGE2 (dmPGE 2 ) were previously identified as a potent external stimulus to enhance HSPC engraftment. PGE 2 has been evaluated in various clinical trials to improve engraftment of stem cells. We identified that dmPGE 2 mediates chromatin flexibility at HSPC-specific enhancers through histone-variant H2A.Z acetylation to promote master transcription factor (TF) binding and reinforce expression of genes involved in stem cell fate and engraftment. Here, we report that PGE 2 -induced changes in chromatin at enhancer regions through histone-variant H2A.Z permit acute inflammatory gene induction to promote HSPC fate. We found that dmPGE 2 -inducible enhancers retain MNase-accessible, H2A.Z-variant nucleosomes permissive of CREB transcription factor (TF) binding. CREB binding to enhancer nucleosomes following dmPGE 2 stimulation is concomitant with deposition of histone acetyltransferases p300 and Tip60 on chromatin. Subsequent H2A.Z acetylation improves chromatin accessibility at stimuli-responsive enhancers. Our findings support a model where histone-variant nucleosomes retained within inducible enhancers facilitate TF binding. Histone-variant acetylation by TF-associated nucleosome remodelers creates the accessible nucleosome landscape required for immediate enhancer activation and gene induction. Our work provides a mechanism through which inflammatory mediators, such as dmPGE 2 , lead to acute transcriptional changes and modify HSPC behavior to improve stem cell transplantation.
hematopoietic stem cell | prostaglandin | chromatin Hematopoietic stem cells are characterized by their unique ability to self-renew and differentiate into all mature blood cell lineages. During normal homeostasis and in conditions of stress such as injury or inflammation, HSPCs maintain an appropriate balance of the hematopoietic system. Hematopoietic stem and progenitor cells (HSPCs) sense and respond to a variety of extrinsic signals that regulate their quiescence, proliferation, and differentiation (1)(2)(3).
Prostaglandins are physiologically active lipids produced in response to mechanical, chemical, or immunological stimuli. They sustain a variety of homeostatic and pathogenic functions. This includes roles in the inflammatory response. Prostaglandin E2 (PGE 2 ) is one of the most abundant prostaglandins produced in the body (4). PGE 2 and its stable derivative dmPGE 2 act as important regulators of vertebrate HSPC development and homeostasis (1,5). We previously demonstrated that ex vivo pulse exposure of HSPCs to dmPGE 2 enhances engraftment and self-renewal in mice and clinical studies indicate benefits for hematopoietic stem cell transplantation (HSCT) outcomes in humans (6,7). DmPGE 2 predominantly exerts its effects by binding to the PGE 2 receptor (EP) subtypes EP2 and EP4 on HSPCs 2 . Interaction with these G-coupled protein receptors enhances intracellular cAMP levels, which activates signaling cascades and downstream effectors, for instance Wnt and β-catenin (8). Improved engraftment presumably results from upregulation of genes implicated in HSPC homing, such as CXCR4 (9). As enhancement of HSPC function by external stimuli supports a strategy to improve HSCTs, understanding the mechanism of gene regulation in response to inductive signals can provide a significant clinical opportunity.
A main mechanism of adaptation involves the activation of TFs that are downstream of signal transduction pathways to ensure appropriate transcriptional responses upon stimulation (10)(11)(12). In higher eukaryotes, gene expression is regulated by the coordinated action of enhancers and promoters (13). Stimuli-responsive TFs (STFs) tend to operate within the cis-regulatory repertoire that is established during cell fate specification and maintained by constitutive binding of lineage-specific master TFs (MTFs) (14). The access of STFs to these regulatory elements and their interaction with cofactors, such as transcriptional activators and chromatin remodeling complexes, depend largely on the local chromatin OPEN ACCESS architecture (15). Generally accepted features of active regulatory regions include open chromatin conformation, histone modifications, and TF binding (16). While promoters consist of a nucleosome-depleted region that is established by chromatin remodelers, general TFs, and the basal transcription machinery, the typical chromatin organization and nucleosome configuration at enhancers remain unclear (17).
Here, we sought to address how transcriptional induction is regulated during the HSPC response to dmPGE 2 . We exploited inducible TF binding to chromatin and identified, and then mechanistically dissected, enhancers controlling inflammatory gene expression changes in HSPCs. We assessed the chromatin accessibility and nucleosome organization of regulatory regions responsive to dmPGE 2 and studied how the higher-order chromatin structures changed following induction. We found that stimuli-induced enhancers retained MNase-accessible nucleosomes. These enhancer nucleosomes were enriched with the noncanonical histone-variant H2A.Z and remodeled but not evicted during acute stimulation. Rather than prohibiting TF binding, we observed enrichment of the dmPGE 2 -responsive TF CREB at accessible nucleosomes within inducible enhancers. CREB binding is concomitant with deposition of the chromatin remodelers p300 and Tip60 that acetylate histone-variant H2A.Z following dmPGE 2 stimulation. This may further improve nucleosome accessibility at stimuli-responsive enhancers and allow for binding of additional TFs and coactivator complexes. We show that the nucleosome organization at enhancers is not exclusively repressive to gene regulation but favors STF binding, which enables rapid enhancer activation and inflammatory gene induction. Our study uncovers a mechanism at the chromatin level that supports acute changes in gene expression, underlying the biologic effect of PGE 2 to induce changes in stem cell engraftment potential.

CREB Regulates Gene Expression Changes through Binding
at Enhancer Elements. We sought to define the molecular mechanisms that underlie the transcriptional response of HSPCs to dmPGE 2 . dmPGE 2 is a synthetic stable derivative of PGE 2 . While this metabolically stable PGE 2 analog is not naturally generated in vivo, it has previously been shown to act as an important regulator of vertebrate HSPC development in ways similar to PGE 2 (1, 8). We exposed human mobilized peripheral blood CD34 + HSPCs to 10μM dmPGE 2 for 2 h and performed extensive gene expression and chromatin profiling (Fig. 1A). Using RNA-sequencing (RNA-Seq), we identified a total of 687 consistent differentially expressed genes (DEGs) after 2 h of dmPGE 2 treatment, when compared to vehicle (DMSO)-treated control cells (Fig. 1B). The effect of dmPGE 2 on gene expression was overwhelmingly stimulatory. More specifically, 535 genes were ≥1.5-fold up-regulated. This includes a significant number of genes involved in cell migration and cell cycle regulation (SI Appendix, Fig. S1 A and C). Among the up-regulated set were genes representative of dmPGE 2 /cAMP/PKA signaling, including PDE4B and PTGS2 (18); several chemokines and cytokines, such as CXCL2 and CXCL8 (19); and genes known to restrict HSPC proliferation and differentiation, for instance NR4A1 and JUNB (20,21) (Fig. 1C and SI Appendix, Table S1). We validated the gene expression signature observed by RNA-Seq through RT-qPCR in HSPCs for representative genes (SI Appendix, Fig. S1D). A total of 152 genes showed a ≤0.67-fold decrease in expression (SI Appendix, Fig. S1B and Table S1). Among the repressed genes, we found enrichment of genes that regulate cell division, such as HOXB4 and CCNF (22). To evaluate the functional impact of gene expression changes in human HSPCs, we assessed cell migration after dmPGE 2 treatment in vitro. By transwell migration assay, which served as a proxy for the previously observed engraftment phenotype (1), we observed greater migration of HSPCs after exposure to dmPGE 2 (SI Appendix, Fig. S1E). These data showed that enhanced engraftment in vivo after dmPGE 2 stimulation is, in part, driven by transcriptional induction of migration genes.
The TF CREB has previously been associated with dmPGE 2 signaling (23). Indeed, our transcriptomic analysis revealed that 30% (158/535) of up-regulated genes are known targets of CREB (SI Appendix, Fig. S2A). CREB binds to cyclic-AMP response elements (CREs) near its target genes where it, upon phosphorylation at serine 133 (S133) by protein kinases, promotes the recruitment of coactivator proteins (23). This increases transcription of CREB-dependent genes (24). One of the protein kinases that phosphorylates CREB at S133 is protein kinase A (PKA). We assessed S133 phosphorylation of CREB in HSPCs after dmPGE 2 treatment. Western blot analysis revealed increased abundance of S133 phosphorylated CREB (pCREB), while total TF protein levels remain unaltered (SI Appendix, Fig. S2B). To correlate pCREB with gene induction, we performed chromatin immunoprecipitation sequencing (ChIP-Seq) with an antibody against pCREB. We found 31,198 binding sites (False Discovery Rate < 0.1%) in CD34 + HSPCs treated for 2 h with dmPGE 2 compared to 8,332 sites in DMSO (SI Appendix, Fig. S2C and Table S2). Correlation of pCREB occupancy to DEGs showed enrichment of pCREB at genes induced by dmPGE 2 (Fig. 1D). In fact, 79% of the up-regulated genes contained at least one pCREB peak in the proximity, that is a window from −5 kb upstream of the transcription start site (TSS) to +5 kb downstream of the TTS, after dmPGE 2 stimulation compared to 47% prior to treatment (Fig. 1D). This value increased to 85% by extending the window up to −100 kb from the TSS to +25 kb from TTS (SI Appendix, Fig. S2C). Besides a higher percentage of up-regulated genes containing pCREB, the TF density was also >2 times higher at up-regulated genes compared to noninduced genes ( Fig. 1E and SI Appendix, Fig. S2D). We observed no clear correlation between the magnitude of the transcriptional response and the density of pCREB (SI Appendix, Fig. S2F). Genes down-regulated by dmPGE 2 showed no additional enrichment in pCREB compared to noninduced genes. Since both the phosphorylation status and TF occupancy of CREB may change in response to dmPGE 2 , we performed ChIP-Seq using a CREB antibody that is nondiscriminating against its phosphorylation status and compared the total CREB occupancy against pCREB-specific occupancy across the genome. These data demonstrated that, relative to total CREB levels, dmPGE 2 preferentially induced pCREB binding near dmPGE 2 -responsive genes (SI Appendix, Fig. S2G). Overall, these data suggested that dmPGE 2 regulates pCREB activity and genomic binding near transcriptionally induced genes.
We next assessed the genomic location of pCREB-bound regions. A total of 23,386 (75%) sites displayed unique pCREB enrichment ("inducible" pCREB regions) in dmPGE 2 -treated HSPCs compared to control-treated cells. The other 7,812 (25%) pCREB sites were present in both control and stimulated HSPCs ("ubiquitous" pCREB regions). The majority of inducible pCREB sites were located distal to the TSS of genes, with a strong representation in intronic and intergenic sequences (Fig. 1 F and G). Ubiquitous pCREB sites were predominantly enriched in promoter regions. This indicated that pCREB binding at putative distal regulatory elements plays a pivotal role in up-regulating dmPGE 2 response genes.

Inducible Enhancers Gain Master and Signaling TF Binding.
To understand the epigenetic impact of dmPGE 2 on distal regulatory elements, we determined the effects of a 2-h stimulation on the state of enhancers in HSPCs. H3K27ac distinguishes active enhancers from primed and poised enhancer elements that are marked by H3K4me1 alone or with H3K27me3, respectively (25). We performed ChIP-Seq for these histone modifications. We assessed H3K27ac enrichment in control and dmPGE 2 -stimulated HSPCs and identified a total of 25,998 active distal regulatory elements (SI Appendix, Table S3). Enhancers that regulate stimulus-responsive gene programs can be distinguished from other active enhancers by their specific increase in H3K27ac upon receipt of the stimulus (26)(27)(28). Comparison of H3K27ac enrichment between control and dmPGE 2 -treated HSPCs allowed us to identify putative distal enhancers involved in the response to dmPGE 2 . We identified a total of 954 (3.7%) enhancers that gained significant enrichment in H3K27ac following dmPGE 2 treatment (SI Appendix, Fig. S3 A and B and Materials and Methods). A subset of these stimuli-inducible enhancers (498/954, 52%) were only detected as active regulatory regions after dmPGE 2 stimulation. While these "de novo" enhancers were depleted of H3K27ac prior to stimulation, their H3K4me1 high /H3K27me3 low state indicated that de novo enhancers reside in a primed state prior to activation ( Fig. 2 A and B). The other fraction of dmPGE 2 -inducible enhancers (456/954, 48%) displayed significant enhancement in H3K27ac enrichment after dmPGE 2 treatment ("enhanced", Fig. 2A, and SI Appendix, Fig. S3C). Since the number of noninducible enhancers (25,044 regions, 96.3%) is much larger than the set of inducible enhancers, we randomly sampled a comparable number of noninducible enhancers ("background", 486) to ensure that observed differences are not due to variations in the size of defined enhancer categories. To assess epigenetic transitions at inducible enhancers, we profiled genome accessibility by assay for transposase-accessible   (E) Correlation between pCREB binding and gene expression in response to dmPGE 2 . pCREB density was calculated by dividing the total number of pCREB peaks associated to each gene category (up−, down−, and nonregulated genes) by the total amount of base pairs that this category occupies in the genome. pCREB peaks were assigned to a gene when located from +5 kb upstream of the TSS to +5 kb downstream of the TTS. Peak density in the genome was calculated by considering random distribution of pCREB sites in the whole genome.
(G) Enrichment of pCREB binding at four representative dmPGE 2 response genes: CXCL2 (promoter, intergenic), CXCL8 (promoter, intergenic), PDE4B (intronic), and FOSL2 (promoter, intronic). Gray bars indicate intronic and intergenic pCREB peaks. Genomic location of presented window is indicated at the bottom of the panels. chromatin sequencing (ATAC-Seq). We found a profound increase in DNA accessibility at inducible enhancers compared to background enhancers after dmPGE 2 stimulation (Fig. 2 A and B), which is a sign of active chromatin reorganization (29). In addition, the mean ATAC-Seq signal prior to activation of de novo enhancers suggested preexisting, yet minimal, accessibility before stimulation. This revealed that dmPGE 2 stimulation results in rapid activation of stimuli-responsive enhancers that is concomitant with increased chromatin accessibility.
It is well known that TFs act as anchors to recruit chromatin remodelers to regulate gene expression (12). Most STFs do not possess pioneering activity and therefore preferentially bind DNA elements located within nucleosome-depleted regions. Moreover, STFs often land at regulatory elements predefined by lineage-specific MTF that have pioneer functions (27,30). Given these insights, we assessed both pCREB occupancy and binding of the HSPC MTFs GATA2 and PU.1 at enhancers prior to and after dmPGE 2 treatment. We found that pCREB colocalized with GATA2 and PU.1 at stimuli-responsive enhancers (Fig. 2 A and C). We moreover observed that inducible enhancers not only gain pCREB but also GATA2 and PU.1 enrichment after dmPGE 2 stimulation (Fig. 2 B and C). No enrichment in MTF occupancy was observed in background enhancers ( Fig. 2 A and B and SI Appendix, Fig. S3E). These data showed that dmPGE 2 -inducible enhancers gain both STF and additional MTF deposition following stimulation.
The contribution of distal regulatory regions for transcriptional activation of dmPGE 2 target genes is illustrated by the changes in the expression of genes regulated by stimuli-responsive enhancers compared to those regulated by background enhancers (Fig. 3A and SI Appendix, Fig. S4A). We tentatively assigned enhancers to individual nearest genes. To reasonably limit arbitrariness in gene assignment, we only considered genes with a mapped TSS within 15 kb of an enhancer. The genes nearest to stimuli-inducible enhancers showed a greater transcriptional response than genes annotated to background enhancers ( Fig. 3A and SI Appendix, Fig. S4A). This indicates that differential enhancer activity is directly reflected in gene expression levels. Importantly, genes associated with inducible enhancers belonged to several pathways, including cell signaling and blood cell migration such as SGK1, CALCRL, CXCL2, CXCL5, and ITGA4 ( Fig. 2C and SI Appendix, Table S4). We found a clear correlation between the number of genes showing differential expression and the changes in enhancer activity. Sixteen percent of genes nearest to de novo enhancers display ≥1.5-fold induction after dmPGE 2 treatment, compared to 9% of enhanced and 3% of background enhancers ( Fig. 3B and SI Appendix, Fig. S4B). In total, 83 (16%) of the 535 up-regulated genes were regulated by at least one stimuli-responsive enhancer ( Fig. 3C and SI Appendix, Fig. S4C). This demonstrated that dmPGE 2 -mediated activation of stimuli-responsive enhancers correlates with modulation of gene expression.
To gain a better understanding of the role of pCREB at inducible enhancers, we further segmented enhancers based on the presence or absence of pCREB after dmPGE 2 treatment. We found that 208/498 (42%) de novo, 210/456 (46%) enhanced, and 125/486 (26%) background enhancers to contain at least 1 pCREB ChIP-Seq peak after dmPGE 2 treatment. Chromatin accessibility and MTF binding increased more profoundly, but not exclusively, at pCREB + stimuli-inducible enhancers (Fig. 3D). The results support a prominent, but not restricted, role for pCREB in regulating gene expression through binding at inducible enhancers. Additionally, pCREB + stimuli-responsive enhancer showed greater transcriptional changes than those without pCREB ( Fig. 3E and SI Appendix, Fig. S4D). Together, our data suggested that increased enhancer chromatin accessibility and the cooperative binding of MTFs and STFs at stimuli-responsive regulatory regions drive transcriptional induction.

Inducible Enhancers Retain Accessible Nucleosomes after
Inflammatory Stimulation. Various studies showed that TFs preferentially bind to sites of open, accessible chromatin (31). However, recent work indicated that accessible chromatin may not necessarily represent nucleosome-depleted regions (32,33). To test whether greater chromatin accessibility at stimuli-responsive enhancers is due to nucleosome depletion, we performed micrococcal nuclease sequencing (MNase-Seq). This allowed us to map nucleosome positions and nucleosome occupancy changes after dmPGE 2 stimulation. Because nucleosomes have varying sensitivities to enzymatic digestion, MNase titrations were performed to obtain a comprehensive map of the nucleosome landscape within a genome (32)(33)(34). While comparing occupancy profiles between individual titration points provides insights on nucleosomal accessibility, the combinatorial analysis of all titration points within a given condition generates a complete view of the nucleosome organization (32) (SI Appendix, Fig. S5A). We prepared native nuclei from HSPCs treated with either DMSO or dmPGE 2 and exposed them to increasing units of MNase. We selected four digestion points that generated increasingly larger fractions of mononucleosomalsize DNA. Mononucleosomal fractions comprised around 10%, 25%, 50%, and 75% of the input chromatin, respectively (SI Appendix, Fig. S5B). Individual MNase titration profiles within a condition, as well as pooled average nucleosome occupancy profiles, revealed a TSS pattern similar to those previously reported (32) (SI Appendix, Fig. S5 C and D). We observed lower average nucleosome occupancy at TSS-proximal regions of dmPGE 2 -responsive genes after treatment (SI Appendix, Fig. S5E). This inverse correlation aligns with former studies which identified high levels of transcription to be concomitant with nucleosome eviction at the TSS, as elongation by RNA Polymerase II is thought to disrupt the nucleosomal organization (35,36).
When assessing the nucleosome organization at inducible enhancers, we found that these regions presented in high nucleosome occupancy states when compared to active promoters (Fig. 4A). Although high occupancy regions are traditionally thought of as "closed", recent work indicated that accessible regulatory regions can retain nucleosomes (32)(33)(34). We determined whether nucleosomes were repositioned or evicted at dmPGE 2 -responsive enhancers following stimulation. No decrease in average nucleosome occupancy was observed at stimuli-responsive enhancers. Inducible enhancers demonstrated higher nucleosome occupancy after dmPGE 2 treatment (Fig. 4A). This was not observed at background enhancers, indicating specificity of the phenomena to dmPGE 2 -inducible enhancers. Our data suggest that stimuli-responsive enhancers retained nucleosomes upon activation, rather than remodeled to a nucleosome-free organization via nucleosome eviction.
Nucleosome profiles from individual MNase titration points can be leveraged to determine how MNase sensitivity changes in specific regions (SI Appendix, Fig. S5A). Light MNase conditions preferentially release accessible and unstable nucleosomes, whereas stable nucleosomes are more resistant to enzyme digestion and only released at higher MNase concentrations (32)(33)(34). We found greater low MNase-Seq signal at inducible enhancers after dmPGE 2 stimulation (Fig. 4 B, upper panel), whereas high MNase digestion degraded the nucleosomal DNA fragments (Fig. 4 B, bottom panel). No changes in low MNase sensitivity were observed in background enhancers. Greater low MNase sensitivity after dmPGE 2 treatment indicates a higher presence of MNase accessible nucleosomes. This shows that nucleosomes retained within inducible enhancers gained low MNase sensitivity upon stimulation.
To ensure that the low MNase-Seq signal represents nucleosomes, we performed ChIP-Seq for the core histones H2B and H4 ( Fig. 4C and SI Appendix, Fig. S6A). Inducible enhancers were not depleted of core histones prior to or after dmPGE 2 stimulation. This is in line with our MNase-Seq data and suggested that nucleosomes are indeed present, and retained at, dmPGE 2 -responsive enhancers. In contrast to our expectations that increased chromatin accessibility at inducible enhancers (Fig. 2B) resulted from nucleosome displacement or eviction, our MNase-Seq and ChIP-Seq data revealed that stimuli-responsive enhancers retained accessible nucleosomes upon activation.
Based on the accessible nucleosomes at enhancers, we next evaluated the relationships between nucleosomes and binding of pCREB at inducible enhancers. A total of 10,169/23,386 (43%) of pCREB binding sites uniquely present after dmPGE 2 treatment are located within the 25,998 active enhancers identified in HSPCs. We observed enrichment of phased nucleosomes at the summit of dmPGE 2 -unique pCREB peaks located within enhancers, both before and after     Fig. S7A). To exclude the possibility that the MNase-Seq fragments observed at pCREB sites within enhancers represent nonhistone proteins protecting from MNase digestion, we assessed fragment size distribution (SI Appendix, Fig. S6B). We found a fragment periodicity that is characteristic for MNase-digested nucleosomes (37,38). The majority of the MNase-Seq fragments that coincide with pCREB binding was 148 base pairs (bp) in length. This corresponds to precisely trimmed nucleosome core particles. Subnucleosomal peaks showed a clear ~10 bp periodicity that reflects the accessibility of DNA as it is wound along the surface of the histone octamer (39). This analysis suggested that MNase-Seq fragments mapping to pCREB sites within enhancers represent nucleosomes. Looking specifically at pCREB + stimuli-responsive enhancers, we observed that the effects of dmPGE 2   (47) De Novo + Enhanced (13) Enhanced (20) De Novo + Background (1) on nucleosome occupancy and low MNase sensitivity described earlier are further amplified at pCREB + regions (SI Appendix, Fig. S7  B and C). Assessment of individual loci confirmed pCREB binding at stimuli-responsive enhancers that show greater low MNase signal and increased nucleosome occupancies after dmPGE 2 treatment (Fig. 4C). Together, the data demonstrated that accessible nucleosomes are retained at dmPGE 2 -responsive enhancers and that these nucleosomes are not prohibitive of pCREB binding.

Modification of H2A.Z-Variant-Accessible Nucleosomes at
Stimuli-Responsive Enhancers. We hypothesized that nucleosome at enhancers may exert important roles in pCREB binding.
Nucleosome-driven TF binding has been observed for several other stress-responsive TFs (40). To understand the conformational changes that underlie increased low MNase sensitivity of enhancer nucleosomes after dmPGE 2 stimulation, we investigated other mechanisms that influence accessibility to nucleosomal DNA. Weakening of internucleosomal interactions by covalent modifications of histones as well as the introduction of histone variants increases DNA accessibility (41). The histone variants H2A.Z and H3.3 can be incorporated in replication-independent manners and are associated with enhancers (42)(43)(44). Once incorporated, these noncanonical histones make for destabilized "fragile" nucleosomes (45). We assessed histone-variant abundance at regulatory regions in the presence and absence of dmPGE 2 stimulation through ChIP-Seq analysis for H2A.Z and H3.3. We observed a positive correlation between histone-variant deposition and enhancer activity, with more profound presence of H2A.Z and H3.3 at higher H3K27ac levels ( Fig. 5 A and B). The histone variants were also found to occupy slightly different sites within enhancers. H2A.Z localized more central to enhancers, the region where TFs bind, whereas H3.3-variant nucleosomes followed a more dispersed pattern and localized to the flanks of enhancers in a profile similar to H3K27ac (Fig. 5 A and B and SI Appendix,  Fig. S6C). We noted minimal changes in H2A.Z enrichment following dmPGE 2 stimulation. We did observe incorporation of the histone-variant H3.3 in the nucleosomes of enhancers (Fig. 5 A and B). When specifically assessing histone variants at pCREB + stimuli-responsive enhancers, we found that pCREB binding directly overlaps with H2A.Z-variant nucleosomes (Fig. 6A). To confirm interaction of pCREB with H2A.Z-variant nucleosomes, we performed complex immunoprecipitation (Co-IP) for pCREB and found association between the TF and H2A.Z histones (Fig. 6B). These results indicated that pCREB overlaps with H2A.Z and suggest that DNA binding of the TF is not prohibited by H2A.Z-variant nucleosomes. H2A.Z is associated with both gene repression and activation (46). This dual function is attributed to posttranslational regulation of the histone variant. Different H2A.Z histone tail modifications recruit distinct interactors that mediate varying transcriptional outputs. H2A.Z acetylation is associated with active transcription and dynamically regulated in response to environmental signals (47,48). Acetylation of H2A.Z occurs at active regulatory regions where it promotes nucleosome destabilization and an open chromatin conformation. The histone acetyltransferases (HATs) p300 and Tip60 that acetylate H2A.Z are known interactors of pCREB (45,(49)(50)(51)(52)(53). Both chromatin remodelers hold important roles in hematopoietic stem cell fate (45,(51)(52)(53). Whereas Tip60 alone is not sufficient to acetylate H2A.Z, p300 can rapidly and effectively acetylate H2A.Z-containing nucleosomes on its own (50). Hence, we assessed whether acetylation of H2A.Z-variant nucleosomes at stimuliresponsive enhancers underlies increased MNase sensitivity and nucleosome accessibility after dmPGE 2 treatment. We found specific H2A. Zac enrichment at stimuli-inducible enhancers, whereas this was not observed in background enhancers (Fig. 5 A-C). We furthermore noted that dmPGE 2 increased the abundance of both p300 and Tip60 at enhancers (Fig. 5 A-C). pCREB interacts with a variety of chromatin remodelers, including p300 and Tip60 (50,54). Binding patterns of two HATs at enhancers were highly similar to pCREB after dmPGE 2 treatment (Fig. 5B) and are suggestive of complex interaction at enhancers. Together, our data revealed that pCREB binds to H2A.Z-variant nucleosomes within enhancers following dmPGE 2 stimulation. pCREB binding at stimuli-responsive enhancers is accompanied by nucleosome remodeling through H2A.Z acetylation, likely mediated through interaction of CREB with p300 and Tip60. H2A.Z acetylation at enhancers may underlie increased enhancer accessibility, allowing additional chromatin factors to engage and ultimately drive acute gene expression changes.
There has been a significant body of literature that suggests the importance of p300 in mediating H2A.Z acetylation (13,49,50,55). CBP is highly homologous to p300 and demonstrates over 60% overlap at the amino acid level (56). Several studies suggest both comparable and distinct contributions of p300 and CBP with regard to protein acetylation (57). Here, we do not discriminate against potential contribution of CBP and believe both p300 and CBP could acetylate H2A.Z to transcriptionally activate dmPGE 2 target genes in HSPCs. Only recently have a small number of studies implicated a role for Tip60 in blood development (51,58).
As such, the exact role of Tip60 in regulating blood cells, specifically during transcriptional activation of stress-induced hematopoietic regeneration, remains elusive. To document the role of Tip60 in modulating the transcriptional effects of dmPGE 2 -pCREB signaling, we followed two separate loss-of-function approaches to knock down TIP60 expression and assess downstream transcriptional effects of dmPGE 2 as compared to DMSO. We reduced TIP60 expression by siRNA and with the Tip60 inhibitor NU9056 (51,58). We observed a moderate yet statistically significant inhibition of dmPGE 2 induction of target genes under siRNA-mediated knockdown of TIP60 in CD34 + HSPCs (SI Appendix, Fig. S8 A and B). Additionally, NU9056 treatment of U937 cells demonstrated a substantial decrease of H2A.Z acetylation and under such conditions, dmPGE 2 target genes appeared to show a reduction in gene induction compared to their respective control condition (SI Appendix, Fig. S8 C and D). While reduction in PGE2-mediated gene expression upon loss of Tip60 is moderate, we did not anticipate huge drop given very high fold-change observed under normal conditions. However, taken together, inhibition of Tip60 demonstrated a statistically significant trend toward compromised transcriptional responsiveness during dmPGE 2 signaling.

Discussion
PGE2 is an important regulator of HSPC homeostasis. The distinct molecular mechanisms through which PGE 2 and its stable derivative dmPGE 2 affect HSC function are critical to understand yet remain elusive. Here, we find that the TF CREB is a key player in the acute transcriptional response to dmPGE 2 by binding to, and activating, distal regulatory elements. Specifically, we find that pCREB binds to H2A.Z-variant nucleosomes that are retained within stimuli-induced, active enhancers and is concomitant with acetylation of these histones. Upon dmPGE 2 treatment, intensity of both H2A.Z and H2A.Zac increases specifically at pCREB-bound, dmPGE 2 -inducible enhancers, suggesting that acetylation of H2A.Z may likely increase on-chromatin interaction of pCREB and H2A.Z. Such interactions between pCREB and H2A.Z may further increase H2A.Z acetylation of enhancer nucleosomes to synergistically promote local chromatin accessibility, which may subsequently help recruit and/or stabilize other HSPC-specific TFs such as GATA2 to stimulate gene transcription and strengthen blood stem cell fate.
CREB is a ubiquitously expressed nuclear, basic leucine zipper (bZIP) TF that regulates over 5,000 genes in the mammalian genome. This includes genes controlling proliferation, differentiation, and cell survival (59). The TF is activated by a wide variety of environmental stimuli and is an important regulator of cellular responses to stress. Most studies focused on promoter-proximal effects of CREB binding. We showed that inducible gene expression is characterized by binding of pCREB at TSS-distal enhancer regions. Regulation of transcriptional responses through binding of CREB at enhancers has also been observed in pancreatic beta cells (60). Our data show that CREB employs distinct mechanisms to regulate steady-state versus inducible gene expression.
Genome-wide assessment of the epigenetic landscape revealed that dmPGE 2 works within the predetermined enhancer repertoire of HSPCs. dmPGE 2 stimulation activates a set of preexisting H3K4me1 + enhancers through chromatin reorganization. dmPGE 2 -responsive enhancers rapidly gain accessibility and TF binding. Our work complements the studies that showed that STFs localize to binding sites adjacent to master regulators (30,61). Although we did not identify the surfacing of latent enhancers, i.e., genomic regulatory elements devoid of TFs and enhancer marks in unstimulated cells (27), a 2-h pulse of dmPGE 2 may be too short to allow for partial reprogramming of the available cis-regulatory landscape. Latent enhancer activation may also be more associated with differentiated cells rather than stem and progenitor populations (27).
We found stimuli-driven enrichment of the HSPC-specific MTFs GATA2 and PU.1 at inducible enhancers, especially those that gained pCREB binding after 2 h of treatment. Cofactor driven binding is common among nonpioneer TFs. It was only recently implied that interaction between TFs can enhance pioneer factor binding at previously sampled target sites (62)(63)(64). Many studies have described the vital role of GATA2 in establishing the regulatory landscape for STFs in HSPCs (30,65). GATA2 facilitates enhancer-promoter loop formation (66), yet few have proposed signaling factors to play a role in GATA2 binding and recruitment to chromatin. We suggest that GATA2 occupancy may be directed and stabilized through cooperativity with STFs at stimuli-responsive enhancers. Our work supports a model where pioneer factor occupancy at specific subsets of enhancers is partially determined by engagement with signaling specific cofactors.
An open chromatin structure surrounding TF binding sites constitutes a prerequisite for transcriptional regulation. Our results show that inducible enhancers rapidly gain accessibility and affinity for STFs and MTFs. While open chromatin regions are presumed nucleosome-depleted regions, we find that dmPGE 2 -responsive enhancers exhibit a high nucleosome occupancy both prior to and after stimulation of HSPCs. Moreover, retained enhancer nucleosomes are not prohibitive of STF binding. pCREB occupancy at enhancers overlaps with well-positioned nucleosomes, suggesting that pCREB can initiate chromatin engagement and access binding sites organized within a positioned nucleosome. Our work indicates that accessible nucleosomes at enhancers may facilitate cooperativity between STFs and MTFs to ensure rapid transcriptional induction. Retention of MNase-accessible nucleosomes at regulatory elements was proposed to play a crucial role in hormone signaling and tissue-specific gene activation (33,40). Retained nucleosomes likely stabilize the interaction of TFs with DNA by facilitating interactions between TF-associated factors, such as chromatin remodeling complexes and histone tails (40).
Except for pioneer factors, most TFs are thought to be unable to bind nucleosomal DNA. Although CREB is not traditionally described as a pioneer factor, novel studies revealed the ability of CREB to open chromatin (67). CREB was identified as a TF that displays an orientated, asymmetric, binding preference near the dyad axis of the nucleosome when engaging nucleosomal DNA (64,65). We hypothesize that MNase-accessible nucleosomes within stimuli-responsive enhancers enable cooperative TF binding for rapid gene activation.
Nucleosomes retained within stimuli-inducible enhancers were epigenetically premarked by the histone-variant H2A.Z. Although the precise function of H2A.Z at enhancers at remains unclear, H2A.Z is an important regulator of enhancer activity in response to stimuli. H2A.Z-rich enhancers display higher chromatin accessibility and gene induction by promoting RNA polymerase II recruitment (42,46,68). In contrast to hormone stimulation, which was found to increase H2A.Z incorporation at enhancer nucleosomes (69), we found limited changes in H2A.Z distribution after dmPGE 2 treatment. Instead, we noted that H2A.Z-variant nucleosomes undergo histone tail acetylation following dmPGE 2 stimulation. Acetylated forms of H2A.Z are associated with an open chromatin conformation and directly regulate transcription of enhancer RNAs (50,54,55,58,70,71). Our work implies that dmPGE 2 -inducible H2A.Z acetylation underlies increased low MNase sensitivity and enhanced nucleosome accessibility at stimuli-responsive enhancers following dmPGE 2 treatment. We found that changes in posttranslational acetylation of H2A.Z at stimuli-responsive enhancers correlate with gene expression changes, indicating that H2A.Zac is a prerequisite for appropriate transcriptional induction following dmPGE 2 stimulation.
Labile, H2A.Z-marked nucleosomes do not present an obstacle for pCREB binding but may facilitate TF binding and enhancer activity. We suggest that a critical feature of pCREB is the recruitment of remodelers that opens the local nucleosome structure through H2A.Z acetylation in enhancers. pCREB interacts with a variety of chromatin remodelers. This includes the HATs p300 and Tip60, both known to interact with H2A.Z and catalyze its acetylation (50,54,58). The recruitment of p300 to chromatin in a stimulus-dependent manner observed here is consistent with interactions between chromatin remodelers and other STFs (72,73). Localization of p300 is concomitant to pCREB binding, suggesting pCREB-p300 complex formation at enhancers upon dmPGE 2 stimulation. pCREB-facilitated recruitment of Tip60 to enhancers complements previously described observations of Tip60 binding to a subset of enhancers and acetylating H2A.Z to promote expression of HSC genes (58,74). The acetylation of H2A.Z-variant enhancer nucleosomes may create a chromatin environment permissive of enhancer activity and transcription.
This study reveals how specific genomic reorganization at a stimuli-responsive group of enhancers is directly translated into regulatory element activation and transcriptional induction. Our findings support a model where STFs and MTFs cooperate with nucleosomes to regulate the activity of cis-regulatory elements that mediate adequate responses to environment signals. While the combination of cooperative lineage-specific MTF and inducible STF binding provides context and responsiveness to external stimuli, histone-variant nucleosomes retained within inducible enhancers may facilitate TF binding by stabilizing chromatin complexes. Subsequent acetylation of histone-variant nucleosomes by TF-associated nucleosome remodelers creates the accessible nucleosome landscape required at active transcriptional enhancers to ensure strong gene activation.
Our research previously identified PGE 2 as a potent regulator of HSPC fate, which became the first compound to move from a zebrafish screen to the clinic (6). Starting with the discovery of PGE 2 in 2007, we have since reported its scientific journey from the bench to the bedside, describing the potential of PGE 2 to improve the success of blood stem cell transplantation in clinical trials (5,6,8). However, the exact mechanisms through which PGE 2 regulates HCPC fate have remained elusive. This study demonstrates the mechanism by which inflammatory lipids such as PGE 2 alter chromatin and HSPC engraftment to improve transplantation. PGE 2 and its stable derivate dmPGE 2 mediate chromatin accessibility at stimuli-inducible enhancers through histone-variant H2A.Z acetylation to promote master TF binding. The acute changes at enhancer elements drive reinforcement in the expression of genes involved in stem cell fate and engraftment. Higher expression of such genes may drive selection of stem cells with greater potency and long-term activity in the natural bone marrow niche after transplantation. As PGE 2 is essential for various types of stem cells (75), this mechanism may improve stem cell function in organ systems beyond those of the blood. We propose that other stress-responsive signals besides PGE 2 , such as estrogen signaling, may utilize similar mechanisms through acetylation of histone variants (48,68). Plasticity of chromatin architecture achieved through modification of labile nucleosomes provides a rapid mechanism to promote stem cell fate through a variety of stress-activated signaling cues over a large extent of tissue types.

Materials and Methods
Expansion of CD34 + Cells. Human CD34 + (HSPCs), isolated from granulocyte-colony-stimulating factor (G-CSF) mobilized peripheral blood of healthy volunteers, were purchased from the Fred Hutchinson Cancer Research Center. The cells were maintained in suspension culture as previously described by Trompouki et al. (30).
Cell Culture. U937 cells were maintained in suspension culture in RPMI-1640 supplemented with 10% (v/v) heat-inactivated fetal bovine serum, 1× GlutaMax, and 1% penicillin-streptomycin at 37° in a humidified atmosphere of 5% CO 2 . dmPGE 2 Treatment. 16,16-dimethyl PGE2 was purchased, reconstituted in DMSO from Cayman Chemicals (cat. #14750), aliquoted, and stored in −80 °C until use. Cells were counted, collected, and resuspended in StemSpan medium with 2% PenStrep (CD34+ Cells) or RPMI with 1% penicillin-streptomycin, but in the absence of additional cytokines or growth factors. The cells were treated with 10 μM dmPGE 2 (Cayman chemicals) or DMSO (vehicle control) for 2 h. qPCR Analysis. RNA was extracted using the RNeasy plus mini kit (Qiagen). cDNA synthesis was performed using the Superscript VILO (Invitrogen) and using equal amounts of starting RNA. The cDNA was analyzed with the Light Cycler 480 II SYBR green master mix (Applied Biosystems), and the QuantStudio 12K Flex (Applied Biosystems). All samples were prepared in triplicate. The PCR cycle conditions used are: Data, Materials, and Software Availability. Next Gen Sequencing Data data have been deposited in GEO DataSets (76).