Polycomb repressive complex 1.1 coordinates homeostatic and emergency myelopoiesis

Polycomb repressive complex (PRC) 1 regulates stem cell fate by mediating mono-ubiquitination of histone H2A at lysine 119. While canonical PRC1 is critical for hematopoietic stem and progenitor cell (HSPC) maintenance, the role of non-canonical PRC1 in hematopoiesis remains elusive. PRC1.1, a non-canonical PRC1, consists of PCGF1, RING1B, KDM2B, and BCOR. We recently showed that PRC1.1 insufficiency induced by the loss of PCGF1 or BCOR causes myeloid-biased hematopoiesis and promotes transformation of hematopoietic cells in mice. Here we show that PRC1.1 serves as an epigenetic switch that coordinates homeostatic and emergency hematopoiesis. PRC1.1 maintains balanced output of steady-state hematopoiesis by restricting C/EBPα-dependent precocious myeloid differentiation of HSPCs and the HOXA9- and β-catenin-driven self-renewing network in myeloid progenitors. Upon regeneration, PRC1.1 is transiently inhibited to facilitate formation of granulocyte-macrophage progenitor (GMP) clusters, thereby promoting emergency myelopoiesis. Moreover, constitutive inactivation of PRC1.1 results in unchecked expansion of GMPs and eventual transformation. Collectively, our results define PRC1.1 as a novel critical regulator of emergency myelopoiesis, dysregulation of which leads to myeloid transformation.


Introduction
While lifelong hematopoiesis is considered driven by hematopoietic stem cells (HSCs) (Sawai et al., 2016;Säwen et al., 2018), recent evidence pointed out a major role for multipotent progenitors (MPPs) and lineage-committed progenitors in hematopoiesis (Busch et al., 2015;Sun et al., 2014). Hematopoietic stem and progenitor cells (HSPCs) are highly responsive to various stresses such as infection, inflammation, and myeloablation (Trumpp et al., 2010;Zhao and Baltimore, 2015), which facilitate myelopoiesis by activating HSPCs to undergo precocious myeloid differentiation and transiently amplifying myeloid progenitors that rapidly differentiate into mature myeloid cells (Hérault et al., 2017;Pietras et al., 2016). This reprogramming of HSPCs, termed 'emergency myelopoiesis,' serves to immediately replenish mature myeloid cells to control infection and regeneration (Manz and Boettcher, 2014). Recent evidence further suggested that uncontrolled activation of the myeloid regeneration programs results in the development of chronic inflammatory diseases and hematological malignancies (Chiba et al., 2018;Zhao and Baltimore, 2015). Emergency myelopoiesis is driven via activation of key myeloid transcriptional networks at the HSPC and myeloid progenitor cell levels (Rosenbauer and Tenen, 2007). However, the epigenetic regulatory mechanisms governing emergency myelopoiesis remained largely unknown.
PCGF4/BMI1-containing canonical PRC1 (PRC1.4) has been characterized for its role in maintaining self-renewal capacity and multipotency of HSCs (Sashida and Iwama, 2012). We and others have reported that BMI1 transcriptionally represses the loci for CDKN2A and developmental regulator genes (e.g., B cell regulators) to maintain self-renewal capacity and multipotency of HSPCs (Iwama et al., 2004;Oguro et al., 2010;Park et al., 2003). We also reported that PCGF5-containing PRC1.5 regulates global levels of H2AK119ub, but is dispensable for HSPC function (Si et al., 2016). On the other hand, we and others recently showed that PRC1.1 components, PCGF1, KDM2B, and BCOR, maintain normal hematopoiesis and suppress malignant transformation of hematopoietic cells (Andricovich et al., 2016;Isshiki et al., 2019;Tara et al., 2018). PRC1.1 consists of PCGF1, RING1A/B, KDM2B, and BCOR or BCLRL1. KDM2B binds to non-methylated CpG islands through its DNA-binding domain, thereby recruiting other PRC1.1 components (Farcas et al., 2012;He et al., 2013). PCGF1 was found to restrict the proliferative capacity of myeloid progenitor cells by downregulating Hoxa family genes in in vitro knockdown experiments (Ross et al., 2012). Correspondingly, we demonstrated that Pcgf1 loss induces myeloid-biased hematopoiesis and promotes JAK2V617F-induced myelofibrosis in mice (Shinoda et al., 2022). Bcor loss also showed myeloid-biased hematopoiesis and promoted the initiation and progression of myelodysplastic syndrome in collaboration with Tet2 loss . However, detailed analysis of the role for PRC1.1 in hematopoiesis, especially in the context of hematopoietic regeneration and emergency myelopoiesis, has not yet been reported.
Here, we analyzed the murine hematopoiesis in the absence of PCGF1 and uncovered critical roles of PCGF1-containing PRC1.1 in homeostatic, emergency, and malignant hematopoiesis.
Pcgf1 Δ/Δ mice exhibited mild anemia and leukopenia, which was mainly attributed to the reduction in B cell numbers ( Figure 1B, Figure 1-figure supplement 2A and B). Myeloid cell numbers in PB were relatively maintained and their proportion increased over time in Pcgf1 Δ/Δ mice ( Figure 1B and Figure 1-figure supplement 2B). While BM cellularity was comparable between control and Pcgf1 Δ/Δ mice, Mac1 + mature myeloid cells were increased at the expense of B cells in Pcgf1 Δ/Δ BM and spleen ( Figure 1C and  3A). The reduction in the HSPC pool size was accompanied by decreased cells in cycling phases in Pcgf1 Δ/Δ HSPCs (CD34and CD34 + LSK cells) ( Figure 1D). Despite the reduction in phenotypic HSPCs, limiting dilution assays with competitive BM transplantation revealed that the numbers of HSPCs that established long-term repopulation of myeloid and T cells (B cells were excluded due to the low contribution of Pcgf1 Δ/Δ HSPCs to B cells, see below) were comparable between control and Pcgf1 Δ/Δ BM ( Figure 1E). Extramedullary hematopoiesis was evident in the Pcgf1 Δ/Δ spleen, as judged by markedly increased numbers of LSK HSPCs, GMPs, and mature myeloid cells including monocytes and neutrophils ( To further evaluate the role of PCGF1 in hematopoiesis, we transplanted BM cells with the same number of CD45.1 wild-type (WT) competitor cells ( Figure 1F). In this competitive setting, only a mild decrease was detected in the overall chimerism of CD45.2 + Pcgf1 Δ/Δ cells in PB ( Figure 1G). In contrast, the chimerism of Pcgf1 Δ/Δ cells in myeloid cells (Mac-1 + and/or Gr1 + ) was markedly increased while that in B cell lineage (B220 + ) was decreased ( Figure 1G). In BM, Pcgf1 Δ/Δ cells outcompeted the competitor cells in the myeloid lineage compartments from the common myeloid progenitor (CMP) stage ( Figure 1H). Since Pcgf1 Δ/Δ cell showed reductions in the numbers of HSCs and MPPs in a non-competitive setting ( Figure 1C), we examined the absolute numbers of test and competitor cells in this competitive repopulation. Of interest, the competitive Pcgf1 Δ/Δ recipients also exhibit similar changes in BM hematopoietic cell numbers. Both CD45.2 + Pcgf1 Δ/Δ and CD45.1 + WT cells were depleted in HSPC fractions in Pcgf1 Δ/Δ recipients, while the total numbers of myeloid progenitors and mature myeloid cells were maintained or rather increased ( Figure 1I). These findings suggest that Pcgf1 Δ/Δ hematopoietic cells suppress expansion of both Pcgf1 Δ/Δ and co-existing WT HSPCs through non-autonomous mechanisms. It is possible that accumulating Pcgf1 Δ/Δ mature myeloid cells produce inflammatory cytokines that are suppressive to HSPCs. The underlying mechanism is an intriguing issue to be analyzed. To evaluate the impact of PCGF1 loss on long-term hematopoiesis, we harvested BM cells from primary recipient mice 4 mo after tamoxifen injection and transplanted them  into secondary recipients. Pcgf1 Δ/Δ cells reproduced the myeloid-biased hematopoiesis in secondary recipients ( Figure 1J). We next evaluated the capacity of Pcgf1 Δ/Δ HSCs to differentiate to myeloid and lymphoid cells in culture. Pcgf1 Δ/Δ HSCs displayed slower population doubling under HSPC-expanding culture conditions ( Figure 1-figure supplement 4A), which is in good agreement with fewer cycling Pcgf1 Δ/Δ HSPCs ( Figure 1D). Nevertheless, Pcgf1 Δ/Δ HSCs showed better population doubling than control cells under myeloid culture conditions ( Figure 1-figure supplement 4A). On the other hand, limiting dilution assays using a co-culture system with TSt-4 stromal cells (Masuda et al., 2005) revealed that the capacity of Pcgf1 Δ/Δ HSCs to produce B and T cells was declined by two-and fivefold, respectively, compared to the control (Figure 1-figure supplement 4B). The discrepancy in T cell production between in vitro and in vivo may be due to the compensatory expansion of T cells in the thymus. These results indicate that PCGF1 loss enhances myelopoiesis at the expense of lymphopoiesis. These phenotypes were similar to those of mice expressing a carboxyl-terminal truncated BCOR that cannot interact with PCGF1 . PCGF1 inhibits precocious myeloid commitment of HSPCs through repression of C/EBPα, which is critical for balanced output of HSPCs To clarify the molecular mechanisms underlying myeloid-biased differentiation of Pcgf1 Δ/Δ HSPCs, we performed RNA sequence (RNA-seq) analysis of HSPCs from mice 4 wk after the tamoxifen injection. Principal component analysis (PCA) showed shifts of the transcriptomic profiles of Pcgf1 Δ/Δ MPP2 and MPP3 toward pre-GMs ( Figure 2A). Gene set enrichment analysis (GSEA) revealed upregulation of genes involved in myeloid cell development (Brown et al., 2006), CEBP targets (Gery et al., 2005), and genes upregulated upon C/EBPα overexpression (Loke et al., 2018) in Pcgf1 Δ/Δ HSPCs compared to controls ( Figure 2B and Supplementary file 1). C/EBP family are master transcription factors for myeloid differentiation (Avellino and Delwel, 2017;Rosenbauer and Tenen, 2007). RT-PCR analysis (n = 9) and Pcgf1 Δ/Δ (n = 14) mice. (C) Absolute numbers of total BM cells, hematopoietic stem cells (HSCs), multipotent progenitors (MPPs), myeloid progenitors, and CLPs (ALP and BLP) in a unilateral pair of femur and tibia 4 wk after the tamoxifen injection (n = 4-5). (D) Cell cycle status of CD34 -LSK HSCs and CD34 + LSK MPPs assessed by Ki67 and 7-AAD staining 4 wk after the tamoxifen injection. (E) In vivo limiting dilution assay. Limiting numbers of BM cells (1 × 10 4 , 4 × 10 4 , 8 × 10 4 , and 2 × 10 5 ) isolated from BM of primary recipients (control and Pcgf1 Δ/Δ mice after transplantation) were transplanted into sublethally irradiated secondary recipient mice with 2 × 10 5 of competitor CD45.1 BM cells (n = 5 each). Due to the low contribution of Pcgf1 Δ/Δ HSPCs to B cells, mice with chimerism of donor myeloid and T cells more than 1% in the PB at 16 wk after transplantation were considered to be engrafted successfully, and the others were defined as non-engrafted mice. The frequencies of HSPCs that contributed to both myeloid and T cells are indicated. (F) Strategy for analyzing Pcgf1 Δ/Δ hematopoietic cells. Total BM cells (2 × 10 6 ) from Rosa26 CreERT and Rosa26 CreERT ;Pcgf1 fl/fl CD45.2 mice were transplanted into lethally irradiated CD45.1 recipient mice with the same number of competitor CD45.1 BM cells. Pcgf1 was deleted by intraperitoneal injections of tamoxifen at 4 wk post-transplantation. Secondary transplantation was performed using 5 × 10 6 total BM cells from primary recipients at 4 mo post-intraperitoneal injections of tamoxifen. (G) The chimerism of CD45.2 donor cells in PB CD45 + leukocytes, Mac-1 + and/or Gr1 + myeloid cells, B220 + B cells, and CD4 + or CD8 + T cells in control and Pcgf1 Δ/Δ mice (n = 6 each) after the tamoxifen injection. (H) The chimerism of CD45.2 donor-derived cells in BM 4 wk after the tamoxifen injection (n = 5). (I) Absolute numbers of CD45.1 and CD45.2 total BM cells, HSCs, MPPs, myeloid progenitors, and Mac-1 + mature myeloid cells in a unilateral pair of femur and tibia 4 wk after the tamoxifen injection (n = 5). Statistical significance is based on the overall number of cells. (J) The chimerism of CD45.2 donor-derived cells in PB in primary (n = 3 each) and secondary (n = 4-5) transplantation. Data are shown as the mean ± SEM. *p<0.05, **p<0.01, ***p<0.001 by the Student's t-test. Each symbol is derived from an individual mouse. A representative of more than two independent experiments is shown.

Deletion of Pcgf1 affects levels of H2AK119ub1
To understand how PCGF1 loss affects H2AK119ub1 status in HSPCs, we performed chromatin immunoprecipitation followed by sequencing (ChIP-seq) analysis using control and Pcgf1 Δ/Δ HSPCs. Since none of the anti-PCGF1 antibodies were suitable for ChIP analysis, we used 3×Flag-PCGF1-expressing BM LK cells obtained by retrovirally transducing LSK cells and transplanting them to lethally irradiated mice. We defined 'PRC1 targets' and 'PRC2 targets' as genes with H2AK119ub1 and H3K27me3 enrichment greater than twofold over the input signals in control LSK cells at promoter regions (transcription start site [TSS] ± 2.0 kb), respectively (Supplementary file 2). GSEA revealed that both PRC1 the Student's t-test. (D) Effects of Cebpa overexpression on HSPC differentiation in vitro. LSK cells were transduced with either control (GFP) or Cebpa retrovirus, then cultured on TSt-4 stromal cells in the presence of IL-7 (upper). The proportions of myeloid (Mac1 + and /or Gr-1 + ), B cells (CD19 + ), and others (Mac1 -Gr-1 -CD19 -) among CD45.2 + GFP + hematopoietic cells on day 17 of culture are indicated (lower left; n = 4 each). RT-qPCR analysis of Cebpa in LSK cells transduced with control or Cebpa retrovirus on day 14 of culture (n = 3). Hprt1 was used to normalize the amount of input RNA (lower right). Each symbol is derived from an individual culture. Data are shown as the mean ± SEM. **p<0.01, ***p<0.001 by the Student's t-test. (E) Impact of Cebpa haploinsufficiency on myeloid-biased differentiation of Pcgf1 Δ/Δ HSPCs. LSK cells from Rosa26 CreERT , Rosa26 CreERT ;Pcgf1 fl/fl and Rosa26 CreERT ;Pcgf1 fl/fl ;Cebpa fl/+ mice were treated with 4-OHT (200 nM) for 2 d in culture to delete Pcgf1 and Cebpa. Cells were further cultured on TSt-4 stromal cells in the presence of IL-7 (upper). The proportions of myeloid (Mac1 + and /or Gr-1 + ), B cells (CD19 + ), and others (Mac1 -Gr-1 -CD19 -) among CD45.2 + GFP + hematopoietic cells on day 17 of culture are indicated (lower; n = 4 each；* versus control). Each symbol is derived from an individual culture (lower left). RT-qPCR data of Cebpa in LSK cells on day 14 of culture (n = 3). Hprt1 was used to normalize the amount of input RNA (lower right). Data are shown as the mean ± SEM. *p<0.05, **p<0.01, ***p<0.001 by the Student's t-test (lower left) or the one-way ANOVA (lower right).
The online version of this article includes the following source data and figure supplement(s) for figure 2: Source data 1. Raw data for Figure 2.    and PRC2 targets were upregulated in Pcgf1 Δ/Δ HSPCs and GMPs ( Figure 2B). K-means clustering divided PRC1 targets into two clusters depending on the levels of H2AK119ub1 and H3K27me3. Cluster 1 genes (1835 RefSeq ID genes) were marked with high levels of H2AK119ub1 and H3K27me3 at promoter regions, while cluster 2 genes (2691 RefSeq ID genes) showed moderate levels of H2AK119ub1 and H3K27me3 ( Figure 3A). PCGF1 showed stronger binding to the promoters of cluster 2 genes than cluster 1 genes ( Figure 3A). Interestingly, only cluster 2 genes showed moderate but significant reductions in H2AK119ub1 and H3K27me3 levels in Pcgf1 Δ/Δ HSPCs ( Figure 3B). The loss of PCGF1 was also significantly associated with de-repression of PRC1 target genes in clusters 1 and 2 ( Figure 3B). These results suggest that cluster 2 genes are the major targets of PCGF1containing PRC1.1 and Cebpa was included in cluster 2 genes. Bivalent genes, which are enriched for developmental regulator genes marked with both active and repressive histone marks (H3K4me3 and H3K27me3, respectively, mostly with H2AK119ub1), are classical targets of canonical PRC1 and PRC2 (Bernstein et al., 2006;Ku et al., 2008) and are implicated in multipotency of HSPCs (Oguro et al., 2010). Of note, bivalent genes defined by our previous ChIP-seq data of HSPCs ; Supplementary file 2) were more enriched in cluster 1 genes than cluster 2 genes ( Figure 3C).
We then defined 'PCGF1 targets' whose H2AK119ub1 levels were decreased by Pcgf1 deletion greater than twofold at promoter regions in HSPCs (997 RefSeq ID genes; Supplementary file 2). We found that Cebpa and Cebpe were included in PCGF1 targets, showed reductions in H2AK119ub1 levels ( Figure 3D and E), and were de-repressed in expression in Pcgf1 Δ/Δ LSK cells ( Figure 2C and Figure 2-figure supplement 1C). ChIP-qPCR confirmed a significant reduction in H2AK119ub1 levels at the promoter region of Cebpa in Pcgf1 Δ/Δ HSPCs ( Figure 3F). These results indicate that the deletion of Pcgf1 compromises PRC1.1 function and causes precocious activation of key myeloid regulator genes in HSPCs.
To clarify whether the Pcgf1 deletion has any impact on the chromatin accessibility in HSPCs, we performed an assay for transposase accessible chromatin with high-throughput sequencing (ATAC-seq) analysis using CD135 -LSK HSPCs, which include HSCs and MPP1-3 but not lymphoid-primed MPP4. ATAC-seq profiles open chromatin regions enriched for transcriptional regulatory regions, such as promoters and enhancers. ATAC peaks were significantly enriched at the promoter regions of PCGF1 target genes, but not of PCGF1 non-target genes, in Pcgf1 Δ/Δ HSPCs compared to the corresponding controls ( Figure 3G), further validating de-repression of PCGF1 targets upon the deletion of Pcgf1. Motif analysis of ATAC peaks in the proximal promoter regions (TSS ± 2 kb) revealed that the CTCF motif, which has been reported to be associated with differentiation of HSCs (Buenrostro et al., 2018;Takayama et al., 2021), was highly enriched in Pcgf1 Δ/Δ HSPCs ( Figure 3H). Interestingly, the other top-ranked motifs were related to stress response transcription factors, such as Bach family (Bach1 and 2), CNC family (Nrf2 and NF-E2), AP1 family, and Atf3 ( Figure 3H and Supplementary file 3). In contrast, the binding motifs for transcription factors essential for T and B cell commitment, including HEB and E2A (de Pooter and Kee, 2010), were negatively enriched in Pcgf1 Δ/Δ HSPCs ( Figure 3H). Together with the significant upregulation of Cebpa and Cebpe in Pcgf1 Δ/Δ HSPCs ( Figure 2C and

PCGF1 inhibition facilitates emergency myelopoiesis
Myeloid-biased hematopoiesis in Pcgf1-deficient mice reminded us of the myeloproliferative reactions caused by emergencies such as regeneration (Manz and Boettcher, 2014). Individual GMPs scatter throughout the BM in the steady state, while expanding GMPs evolve into GMP clusters during regeneration, which, in turn, differentiate into granulocytes. Inducible activation of β-catenin and Irf8 controls the formation and differentiation of GMP clusters, respectively (Hérault et al., 2017). Immunofluorescence analyses of BM sections readily detected GMP clusters in steady-state Pcgf1 Δ/Δ BM ( Figure 4A), which was reminiscent of those observed during regeneration after 5-fluorouracil (5-FU) treatment ( Figure 4-figure supplement 1A). To address whether PCGF1 also regulates myelopoiesis at the GMP level during regeneration, we challenged control and Pcgf1 Δ/Δ mice with a single dose of 5-FU ( Figure 4B). Consistent with the previous report (Hérault et al., 2017), control mice showed transient expansion of BM HSPCs and GMPs around day 14 and subsequent burst of circulating PB myeloid cells around day 21. In sharp contrast, Pcgf1 Δ/Δ mice displayed sustained GMP expansion until day 28 without efficient production of PB myeloid cells, leading to the accumulation of excess GMPs ( Figure 4C). GMPs can be divided into steady-state GMPs (ssGMP) and self-renewing GMPs (srGMP), the latter of which transiently increase during regeneration (Hérault et al., 2017). We performed RNA-seq analysis of GMPs isolated from 5-FU-treated WT mice at various time points ( Hérault et al., 2017). Of note, a significant portion of upregulated DEGs in day 8 5-FU-treated GMPs were also upregulated in Pcgf1 Δ/Δ GMPs ( Figure 4D and Supplementary file 4). These overlapping genes included Hoxa9, a PRC1.1 target (Figure 4-figure supplement 1D; Ross et al., 2012;Shinoda et al., 2022;Tara et al., 2018), which is highly expressed in srGMPs (Hérault et al., 2017). RT-qPCR confirmed significantly higher expression of Hoxa9 in Pcgf1 Δ/Δ GMPs than control GMPs ( Figure 4E). Correspondingly, 5-FU treatment transiently decreased H2AK119ub1 levels at Hoxa9 locus around day 10 ( Figure 4F). ChIP-seq analysis also revealed significant reductions in H2AK119ub1 levels at promoters of upregulated DEGs in day 8 5-FU-treated GMPs in Pcgf1 Δ/Δ GMPs ( Figure 4G). These results suggest that transient inhibition of PRC1.1 de-represses genes critical to expand srGMP during myeloid regeneration, although the expression of PRC1.1 genes remained largely unchanged during regeneration (Figure 4-figure supplement 1E).
To better understand the role of PRC1.1 in myeloid progenitors, we performed single-cell RNA-seq (scRNA-seq) of Lin -Sca-1c-Kit + myeloid progenitors from control and Pcgf1 Δ/Δ mice at steady state. We used data from 6171 control and 6198 Pcgf1 Δ/Δ single cells and identified 10 major clusters based on dimension reduction by UMAP ( Figure 4H). Functional annotation of respective UMAP clusters using previously reported myeloid progenitor cell gene expression profiles (Nestorowa et al., 2016) assigned clusters 0, 2, and 4 to GMPs ( Figure 4H). PCA subdivided GMPs into two major groups ( Figure 4I). These two groups exhibited distinct expression profiles of Hoxa9, Irf8, Csf1r, and Il6ra, key genes differentially expressed between ssGMPs (Hoxa9 lo , Irf8 hi , Csf1r hi , Il6ra hi ) and srGMPs (Hoxa9 hi , Irf8 lo , Csf1r lo , Il6ra lo ) (Hérault et al., 2017), and we classified clusters 2 and 4 as ssGMPs and cluster 0 as srGMPs ( Figure 4J). Gene Ontology and pathway enrichment analyses using DEGs (cluster 2 or 4 versus cluster 0) revealed that clusters 2 and 4 represented more mature myeloid cell populations than cluster 0 ( Figure 4K). As expected, Pcgf1 Δ/Δ myeloid progenitors had a greater proportion of total GMPs including srGMPs than controls ( Figure 4L). Of note, the frequency of srGMPs was also increased in Pcgf1 Δ/Δ myeloid progenitors ( Figure 4L). These results indicate that PRC1.1 restricts expansion of self-renewing GMPs and suggest that transient PRC1.1 inhibition allows for temporal amplification of GMPs and their subsequent differentiation to mature myeloid cells.

PCGF1 restricts GMP self-renewal network
To further investigate the mechanism by which PCGF1 regulates GMPs, we took advantage of in vitro culture experiments. Remarkably, Pcgf1 Δ/Δ GMPs displayed better proliferation than control GMPs under myeloid-expanding culture conditions ( Figure 5A). Moreover, while comparable numbers of cells expressing immunophenotypic GMP markers (CD34 + FcγR + c-Kit + Sca-1 -Lineage -) (immunophenotypic GMPs) were produced by control and Pcgf1 Δ/Δ HSCs on day 7 of culture, control HSCs showed a rapid decline in immunophenotypic GMP production afterward but Pcgf1 Δ/Δ HSCs persistently produced immunophenotypic GMPs until day 23 ( Figure 5B). Of interest, differentiation of the against the input signals (ChIP/Input) (TSS ± 2 kb) of H2AK119ub1 between control and Pcgf1 Δ/Δ LSK cells. PRC1 targets are indicated in a green box. (E) Snapshots of Flag-PCGF1 and H2AK119ub1 ChIP signals at the Cebpa and Cebpe gene loci. (F) ChIP qPCR assays for H2AK119ub1 at the Cebpa promoter in control and Pcgf1 Δ/Δ LSK cells. The relative amounts of immunoprecipitated DNA are depicted as a percentage of input DNA. Data are shown as the mean ± SEM (n = 3). ***p<0.001 by the Student's t-test. (G) Box-and-whisker plots showing ATAC signal levels at proximal promoters (TSS ± 2 kb) in Pcgf1 Δ/Δ CD135 -LSK cells relative to those in control cells. The data of all ATAC peaks (n = 18,417), PCGF1 target genes (TG) (n = 670), and non-PCGF1 TG (n = 17,747) are shown. Boxes represent 25-75 percentile ranges. The whiskers represent 10-90 percentile ranges. Horizontal bars represent medians. Mean values are indicated by red crosses. ***p<0.001 by the one-way ANOVA. (H) Top DNA motifs identified in ATAC peaks at proximal promoters (TSS ± 2 kb) positively or negatively enriched in Pcgf1 Δ/Δ CD135 -LSK cells compared to corresponding controls.
The online version of this article includes the following source data for figure 3: Source data 1. Raw data for Figure 3.  expanded Pcgf1 Δ/Δ GMPs was largely blocked, as indicated by reduced Mac1 + differentiated myeloid cells/GMP ratios until day 19 ( Figure 5-figure supplement 1A), suggesting that Pcgf1 Δ/Δ GMPs underwent enhanced self-renewal rather than differentiation. However, Pcgf1 Δ/Δ GMPs did not lose its differentiation potential to mature myeloid cells and did not show evident differentiation block, with more Mac1 + cells ultimately generated from HSC cultures than control GMPs ( Figure 5-figure  supplement 1A). This was also consistent that the number of mature myeloid cells expressing myeloid differentiation markers increased in Pcgf1 Δ/Δ BM and spleen (Figure 1-figure supplement 2C and  D). Furthermore, Pcgf1 Δ/Δ HSPCs showed remarkably sustained colony formation activity upon serial replating with myeloid cytokines, which is in line with the elevated self-renewing activity of Pcgf1 Δ/Δ GMPs ( Figure 5C).
srGMPs have increased levels of nuclear β-catenin, which is known to confer aberrant self-renewal features to leukemic GMPs (Wang et al., 2010) and directly suppresses Irf8 expression (Hérault et al., 2017). The proportion of nuclear β-catenin was significantly increased in Pcgf1 Δ/Δ GMPs at later time points of culture when GMPs in control culture shrunk but GMPs in Pcgf1 Δ/Δ culture kept expanding ( Figure 5B and D). Pcgf1 Δ/Δ GMPs in culture possessed a transcriptional profile typical to srGMPs; upregulation of Hoxa9 and downregulation of Irf8, Csf1r, and Il6ra ( Figure 5E). GSEA revealed activation of the Wnt/β-catenin pathway (Shooshtarizadeh et al., 2019) and downregulation of the Irf8 targets (Kubosaki et al., 2010) in Pcgf1 Δ/Δ GMPs ( Figure 5F). We hypothesized that Hoxa9, a direct target of PCGF1, could have a role in the GMP self-renewal network. Overexpression of Hoxa9 in HSPCs significantly enhanced their growth and induced persistent production of GMPs for a long period ( Figure 5G and Figure 5-figure supplement 1B). Most Hoxa9-overexpressing GMPs had nuclear β-catenin ( Figure 5H). These results indicate that HoxaA9 can reinforce activation of β-catenin, thus placing Hoxa9 as a component of the GMP self-renewal network and PCGF1-PRC1.1 as a negative regulator of this network ( Figure 5I). The promoter region of the Ctnnb1 gene contains at least one predicted sequence for Hoxa9 binding (CTTA TAAA TCG) (data not shown). We performed qPCR analysis using Hoxa9 overexpression samples at day 12 of culture, in which nuclear localization of β-catenin was observed ( Figure 5H). Ctnnb1 gene expression was slightly but significantly upregulated in Hoxa9 overexpression samples compared to mock control ( Figure 5-figure supplement  1C). These results raise the possibility that Hoxa9 is a direct transcriptional activator of the Ctnnb1 gene, but this warrants further analysis.

Constitutive PCGF1 loss promotes malignant transformation
In leukemia, GMP clusters are constantly produced owing to persistent activation of the myeloid self-renewal network and a lack of termination cytokines that normally restore HSC quiescence (Hérault et al., 2017). A significant portion of Pcgf1 Δ/Δ mice, which exhibit constant production of GMP clusters, developed lethal myeloproliferative neoplasms (MPN) with severe anemia and massive     Figure 6A-E). Morphological analysis of BM and spleen sections revealed accumulation of mature myeloid cells but no obvious differentiation block like acute myeloid leukemia (AML) was observed ( Figure 6F). Indeed, AML was not observed in long-term observation of Pcgf1 Δ/Δ mice. It is assumed that the epigenetic status that allows GMP selfrenewal was enforced over time in the absence of Pcgf1, leading to enhanced production of mature myeloid cells that mimic MPN. A part of Pcgf1 Δ/Δ mice also developed lethal T-cell acute lymphoblastic leukemia (T-ALL) ( Figure 6A and G) like Bcor mutant mice . These results indicate that constitutive activation of the GMP self-renewal network in Pcgf1 Δ/Δ mice serves to promote malignant transformation. Taken together, these results highlight the importance of PRC1.1-dependent suppression of the myeloid self-renewing network to prevent malignant transformation.

Discussion
In this study, we demonstrated that PCGF1 contributes to balanced hematopoiesis by restricting precocious myeloid commitment of HSPCs and expansion of myeloid progenitors while its inhibition promotes emergency myelopoiesis and myeloid transformation. These findings present a sharp contrast with PCGF4/BMI1 essential for self-renewal of HSCs (Iwama et al., 2004;Oguro et al., 2006;Park et al., 2003) and underscore distinct functions between canonical PRC1 and non-canonical PRC1.1 in hematopoiesis ( Figure 6H). PcG and trithorax group proteins mark developmental regulator gene promoters with bivalent histone domains to keep them poised for activation in ES cells (Bernstein et al., 2006). We previously reported that canonical PRC1 reinforces bivalent domains at the B cell regulator genes, Ebf1 and Pax5, to maintain B cell lineage commitment poised for activation in HSPCs (Oguro et al., 2010). In contrast, PCGF1 appeared to target non-bivalent PRC1 target genes marked with moderate levels of H2AK119ub1 and H3K27me3. Among these, PCGF1 targets myeloid regulator genes, such as Cebpa, thereby negatively regulating myeloid commitment. Our findings indicate that canonical and noncanonical PRC1 restrict the lymphoid and myeloid commitment of HSPCs, respectively, by targeting different transcriptional programs of differentiation, thereby fine-tuning the balance of HSPC commitment ( Figure 6H). Although there might be considerable functional redundancy between canonical and non-canonical PRC1 complexes, our results uncovered a unique function of PRC1.1 in the lineage commitment of HSPCs. The reduction in the levels of Cebpa expression by introducing a Cebpa fl/+ allele in Pcgf1 Δ/Δ LSK cells was sufficient to restore balanced production of myeloid and B cells by Pcgf1 Δ/Δ HSPCs in culture ( Figure 2E). However, we could not confirm these data in mice probably due to inefficient deletion of Cebpa using a Cre-ERT system (data not shown). The real impact of de-repressed Cebpa in myeloid-biased differentiation of Pcgf1 Δ/Δ HSPCs requires further validation in vivo.
Myeloid-biased output from HSPCs is one of the hallmarks of emergency hematopoiesis (Trumpp et al., 2010;Zhao and Baltimore, 2015). In mouse models of regeneration, myeloid-biased MPP2 and MPP3 are transiently overproduced, suggesting that HSCs produce functionally distinct lineage-biased MPPs to adapt blood production to hematopoietic demands (Pietras et al., 2015). In the present study, we found that Pcgf1-deficient hematopoiesis recapitulates sustained emergency myelopoiesis, immunophenotypic GMPs are shown on the right panel. Data are shown as the mean ± SEM (n = 3). (E) Quantitative RT-PCR analysis of Hoxa9, Irf8, Csf1r, and Il-6ra in sorted control and Pcgf1 Δ/Δ immunophenotypic GMPs in HSC culture in (B) at the indicated time points. Hprt1 was used to normalize the amount of input RNA. Data are shown as the mean ± SEM (n = 3). (F) Gene set enrichment analysis (GSEA) using RNA-seq data. The gene sets used are indicated in Supplementary file 1. (G) Growth of mock control and Hoxa9-expressing LSK cells. LSK cells transduced with a Hoxa9 retrovirus harboring mCherry marker gene were cultured in triplicate under myeloid culture condition-2 (25 ng/mL SCF, TPO, Flt3L, and IL-11 and 10 ng/mL IL-3 and GM-CSF). The proportion of GMPs in culture is depicted on the right panel. Data are shown as the mean ± SD (n = 4). (H) Proportion of GMPs with nuclear β-catenin in mock control and Hoxa9-expressing GMPs in LSK culture on day 12 in (G). Data are shown as the mean ± SEM (n = 5-6). (I) Model of the molecular network controlling GMP self-renewal and differentiation. *p<0.05; **p<0.01; ***p<0.001 by the Student's t-test (A-D, H, and G) or the one-way ANOVA (E). Each symbol is derived from an individual culture.
The online version of this article includes the following source data and figure supplement(s) for figure 5: Source data 1. Raw data for Figure 5.   The online version of this article includes the following source data for figure 6: Source data 1. Raw data for Figure 6. although the production of circulating myeloid cells was not enhanced. Expanding GMPs, GMP clusters during regeneration, which, in turn, differentiate into granulocytes (Hérault et al., 2017). Of note, PCGF1 loss induced constitutive GMP cluster formation at steady state and sustained GMP expansion in mice after myeloablation and in culture. Correspondingly, Pcgf1-deficient mice had a greater number of self-renewing GMPs than control mice. This unique phenotype may implicate the importance of transient but not constitutive PCGF1 repression for proper myeloid regeneration. β-catenin and Irf8 constitute an inducible self-renewal progenitor network controlling GMP cluster formation, with β-catenin directly suppressing Irf8 expression while restoration of Irf8 expression terminating the self-renewal network and inducing GMP differentiation (Hérault et al., 2017; Figure 5I). Hoxa9, which is upregulated in srGMPs, is one of the PRC1.1 targets in myeloid progenitors (Figure 4figure supplement 1D; Ross et al., 2012;Shinoda et al., 2022;Tara et al., 2018). We demonstrated that Hoxa9 expression activates β-catenin and promotes GMP self-renewal, identifying Hoxa9 as a component of the GMP self-renewal network. Of note, PRC1.1 is transiently inhibited to de-repress such GMP self-renewal network genes. This transient nature of PRC1.1 inhibition allows for srGMP expansion and GMP cluster formation followed by proper differentiation of expanded GMPs. As expression levels of PRC1.1 components remained unchanged during hematopoietic regeneration, non-canonical PRC1.1 activity could be modulated by post-translational modifications in response to extracellular stimuli like canonical PRC1 (Banerjee Mustafi et al., 2017;Liu et al., 2012;Nacerddine et al., 2012;Voncken et al., 2005). How extrinsic signals modulate PRC1.1 functions to regulate myelopoiesis remains an important question.
The molecular machineries that drive emergency myelopoiesis are often hijacked by transformed cells (Hérault et al., 2017). A significant portion of Pcgf1-deficient mice eventually developed lethal MPN after a sustained myeloproliferative state. These findings indicate that PRC1.1 functions as a critical negative regulator of myeloid transformation. Among the components of PRC1.1, BCOR and BCLRL1, but not PCGF1 are targeted by somatic gene mutations in various hematological malignancies, including myelodysplastic syndrome (MDS), chronic myelomonocytic leukemia (CMML), and acute myeloid leukemia (AML) . We reported that mice expressing a carboxyl-terminal truncated BCOR, which cannot interact with PCGF1, showed myeloid-biased hematopoiesis like Pcgf1-deficient mice. Importantly, HSPCs in these mice showed a growth advantage in the myeloid compartment, which was further enhanced by the concurrent deletion of Tet2, leading to the development of lethal MDS . De-repression of myeloid regulator genes, such as Cebp family and Hoxa cluster genes, were also detected in Bcor mutant progenitor cells . These findings also support the idea that PRC1.1 restricts myeloid transformation by transcriptionally repressing aberrant activation of myeloid regeneration programs.
Collectively, our findings highlight a critical role of PRC1.1 in coordinating steady-state and emergency hematopoiesis and preventing malignant transformation. They also suggest that transient inhibition of PRC1.1 would be a novel approach to temporarily induce emergency myelopoiesis and enhance myeloid cell supply while avoiding the potential risk for malignant transformation.

Mice
Wild-type mice (C57BL/6) and Rosa::Cre-ERT2 mice were purchased from the Japan SLC and Taconi-cArtemis GmbH, respectively. Pcgf1 fl and Cebpa fl mice were kindly provided by Haruhiko Koseki and Daniel G. Tenen, respectively, and previously reported (Almeida et al., 2017;Zhang et al., 2004). All experiments using mice were performed in accordance with our institutional guidelines for the use of laboratory animals and approved by the Review Board for Animal Experiments of Chiba University (approval ID: 30-56) and the University of Tokyo (approval ID: PA18-03).

Quantitative RT-PCR
Total RNA was extracted using a RNeasy Micro Plus Kit (QIAGEN) or TRIZOL LS solution (MOR) and reverse transcribed by the SuperScript IV First-Strand Synthesis System (Invitrogen) or the ReverTra Ace α-(TOYOBO) with an oligo-dT primer. Real-time quantitative PCR was performed with a StepO-nePlus Real-Time PCR System (Life Technologies) using FastStart Universal Probe Master (Roche) and the indicated combinations of the Universal Probe Library (Roche), or TB Green Premix Ex Taq II (TaKaRa Bio). All data are presented as relative expression levels normalized to Hprt expression. The primer sequences used are shown in Supplementary file 5 (Murakami et al., 2021;Sonntag et al., 2018).
Cell cycle assay BM cells were stained with antibodies against cell-surface markers. After washing, cells were fixed and permeabilized with a BD Phosflow Lyse/Fix Buffer and a BD Phosflow Perm Buffer II (BD Biosciences) according to the manufacturer's instructions. Cells were stained with FITC-Ki67 antibody (#11-5698-82; Thermo Fisher Scientific) at room temperature for 30 min and then with 1 μg/mL 7-AAD (Sigma-Aldrich). Flow cytometric analyses were performed on FACSAria IIIu (BD Biosciences).

Retroviral vector and virus production
Full-length Pcgf1 and Bmi1 cDNA tagged with a 3×Flag at the N-terminus was subcloned into the retroviral vector pGCDNsam-IRES-EGFP. Full-length Hoxa9 cDNA was subcloned into the retroviral vector pMYs-IRES-mCherry. A recombinant retrovirus was generated by a 293gpg packaging cell line. The virus in supernatants of 293gpg cells was concentrated by centrifugation at 6000 × g for 16 hr.

Immunofluorescence imaging of bone marrow and spleen sections
Isolated mouse femurs were immediately placed in ice-cold 2% paraformaldehyde solution (PFA/PBS) and fixed under gentle agitation for 16 hr. The samples were then incubated in 15 and 30% sucrose for cryoprotection overnight. Samples were embedded in O.C.T. (Sakura) and frozen in cooled hexane. The 7 μm frozen sections were generated with a cryostat (Cryostar NX70, Thermo Scientific) using Kawamoto's tape method (Kawamoto, 2003). Sections on slide glasses were blocked with staining buffer (10% normal donkey serum in TBS) and an Avidin/Biotin Blocking Kit (VECTOR), then stained with biotinylated anti-lineage antibody cocktail and anti-c-Kit antibody (#AF 1356; R&D Systems), or anti-FcγR-Alexa Fluor 647 (#101314; BioLegend) in staining buffer overnight at 4℃. For secondary staining, sections were incubated with streptavidin-Alexa Fluor 488 (#S11223; Invitrogen) and donkey anti-goat Alexa Fluor 555 (#A21432; Invitrogen) antibody for 3 hr at room temperature. Finally, sections were incubated with 1 μg/mL DAPI/TBS for 10 min and mounted with ProLong Glass Antifade Mountant (Thermo Scientific). Images of sections were captured on a confocal microscope (Dragonfly, Andor, or A1Rsi, Nikon) and processed using Fiji. ImageJ was used for image quantification.

Immunofluorescence imaging of purified GMPs
GMPs were sorted directly onto glass slides using BD AriaIIIu. The cells were washed three times with PBS for 5 min between each staining step. Cells were fixed with 4% PFA for 15 min, permeabilized with 0.1% Triton X-100 for 10 min, and then blocked with 3% BSA for 1 hr at room temperature. The cells were then incubated with rabbit anti-mouse β-catenin (#9582S; Cell Signaling) primary antibody at 4°C overnight. The cells were then stained with anti-rabbit AF488A (#20015; Biotium) secondary antibody for 2 hr at room temperature. After staining with 1 μg/mL DAPI/PBS for 5 min, the cells were mounted with ProLong Glass Antifade Mountant (Thermo Fisher). DragonFly (Andor, ×40 objective) was used for image acquisition.
Bulk RNA-seq and data processing Total RNAs were extracted from 1000 to 5000 cells using an RNeasy Plus Micro Kit (QIAGEN) and cDNAs were synthesized using a SMART-Seq v4 Ultra Low Input RNA Kit for Sequencing (Clontech) according to the manufacturer's instructions. The ds-cDNAs were fragmented using S220 or M220 Focused-ultrasonicator (Covaris), then cDNA libraries were generated using a NEBNext Ultra DNA Library Prep Kit (New England BioLabs) according to the manufacturer's instructions. Sequencing was performed using HiSeq1500 or HiSeq2500 (Illumina) with a single-read sequencing length of 60 bp. TopHat2 (version 2.0.13; with default parameters) was used to map the reads to the reference genome (UCSC/mm10) with annotation data from iGenomes (Illumina). Levels of gene expression were quantified using Cuffdiff (Cufflinks version 2.2.1; with default parameters). Significant expression differences were detected edgeR (version 3.14; with default parameters), with raw counts generated from String Tie. The super-computing resource was provided by the Human Genome Center, the Institute of Medical Science, the University of Tokyo (http://sc.hgc.jp/shirokane.html). The enrichment analyses were performed using g:Profiler tool.
In 3×Flag-Pcgf1ChIP assay, BM LK cells were fixed with 1% FA at 25°C for 10 min, lysed in RIPA buffer, and sonicated for 11 s ×15 times by a homogenizer (NR-50M; Micro-tec Co.). After centrifugation, supernatants were immunoprecipitated at 4°C overnight with Dynabeads Sheep Anti-Mouse IgG (Invitrogen) conjugated with an anti-FLAG antibody (Sigma-Aldrich). Then, the samples were treated in the same way as ChIP assays for histone modifications.
In ChIP-qPCR assay, quantitative real-time PCR was performed with a StepOnePlus Thermal Cycler (Thermo Fisher Scientific) using SYBR Premix Ex Taq II or TB Green Premix Ex Taq II (Takara Bio). The primer sequences used are shown in Supplementary file 5.
ChIP-seq libraries were prepared using a ThruPLEX DNA-seq Kit (Clontech) according to the manufacturer's instructions. Bowtie2 (version 2.2.6; with default parameters) was used to map the reads to the reference genome (UCSC/mm10). The RPM (reads per million mapped reads) values of the sequenced reads were calculated every 1000 bp bin with a shifting size of 100 bp using bedtools. In order to visualize with Integrative Genomics Viewer (IGV) (http://www.broadinstitute.org/igv), the RPM values of the immunoprecipitated samples were normalized by subtracting the RPM values of the input samples in each bin and converted to a bigwig file using wigToBigWig tool. The supercomputing resource was provided by the Human Genome Center, the Institute of Medical Science, the University of Tokyo (http://sc.hgc.jp/shirokane.html).
Assay for transposase accessible chromatin with high-throughput (ATAC)-sequencing BM CD135 -LSK cells (1.6-3.0 × 10 4 ) and GMPs (3.0 × 10 4 ) were lysed in cold lysis buffer (10 mM Tris-HCl pH 7.4, 10 mM NaCl, 3 mM MgCl 2 , 0.1% IGEPAL CA-630) on ice for 10 min. After centrifugation, nuclei pellets were resuspended with 50 μL of transposase reaction mix (25 μL Tagment DNA buffer (illumine), 2.5 μL Tagment DNA enzyme (illumine) and 22.5 μL water), incubated at 37°C for 35 min and were purified with a MinElute PCR Purification Kit (QIAGEN). After the optimization of PCR cycle number using SYBER Green I Nucleic Acid gel Stain (Takara Bio), transposed fragments were amplified using NEBNext High Fidelity 2×PCR Master mix and index primers, and were purified with a MinElute PCR Purification Kit (QIAGEN). Library DNA was sized selected (240-360 bps) with BluePippin (Sage Science). Sequencing was performed using HiSeq1500 or HiSeq2500 (Illumina) with a single-read sequencing length of 60 bp. Bowtie2 (version 2.2.6; with default parameters) was used to map reads to the reference genome (UCSC/mm10) with annotation data from iGenomes (Illumina). Reads mapped to mitochondria were removed. To ensure even processing, reads were randomly downsampled from each sample to adjust to the smallest read number of samples. MACS (version 2.1.1; with default parameters) was used to call peaks in downsampled reads. The catalogue of all peaks called in any samples was produced by merging all called peaks that overlapped by at least one base pair using bedtools. The MACS bdgcmp function was used to compute the fold enrichment over the background for all populations, and the bedtools map function was used to count fragments in the catalogue in each population. Fragment counts at each site in the catalogue were quantile normalized between samples using the PreprocessCore package in R (3.3.2). We used the Homer package with command annotatePeaks. pl using default parameters to annotate regions with promoter and distal labels and the nearest gene, and with command findMotifsGenome. pl using default parameters to identify enriched motifs, and the catalogue of all called peaks as a background.

Quantification and statistical analysis
Statistical tests were performed using Prism version 9 (GraphPad). The significance of difference was measured by the Student's t-test or one-way ANOVA. Data are shown as the mean ± SEM. Significance was taken at values of *p<0.05, **p<0.01, and ***p<0.001. and Scientific Research on Innovative Area 'Replication of Non-Genomic Codes' (#19H05746) from Japanese Society for the Promotion of Science, and Moonshot (#21zf0127003h0001) and UTOPIA (JP223fa627001) projects from the Japan Agency for Medical Research and Development. • Supplementary file 4. List of differentially expressed genes (DEGs) in GMPs and those overlapped between "5-FU day 8 vs day 0 GMPs" and "Control vs Pcgf1 Δ/Δ GMPs.

Additional information
• Supplementary file 5. List of oligonucleptide and TaqMan gene expression assay materials.
Data availability RNA sequence, ChIP sequence and ATAC sequence data were deposited in the DDBJ (accession numbers DRA008518 and DRA013523).
The following datasets were generated: Author (