Glutamylation of deubiquitinase BAP1 controls self-renewal of hematopoietic stem cells and hematopoiesis

Many blood components, including erythrocytes, neutrophils, and megakaryocytes, are short-lived and are constantly regenerated. All hematopoietic lineages are derived from a limited pool of hematopoietic stem cells (HSCs), which represents one of the most canonical adult stem cells (Orkin and Zon, 2008). At the very top of hematopoietic hierarchy is the long-term HSC (LT-HSC). LT-HSCs possess pluripotency to generate all blood cells throughout their lifetime, and they keep predominantly quiescent at G₀ phase (Wilson et al., 2008). Oxidative stress (Tothova et al., 2007), infection (Sato et al., 2009), and aging (Flach et al., 2014; Takubo et al., 2010) can activate HSCs to enter the cell cycle and replenish blood cells, but excessive mobilization burdens HSCs and renders them exhausted. LT-HSCs give rise to daughter stem cells through self-renewal, as well as downstream short-term HSCs and multipotent progenitors (MPPs), along with progressively bereaving self-renewal ability (Rossi et al., 2012). Thus, HSCs are essential to guarantee lifelong hematopoiesis.

Protein posttranslational modifications, such as methylation, phosphorylation, ubiquitination, and sumoylation, have been reported to participate in regulating hematopoiesis (Cimmino et al., 2017; Liu et al., 2014; Nakagawa et al., 2015; Zhu et al., 2011). Glutamylation is another posttranslational modification that was initially identified on tubulins (Eddé et al., 1990). By adding glutamate side chains onto the γ-carboxyl groups of glutamic acid residues of the target proteins, glutamylation alters charge characteristics, protein–protein interaction, stability, and activity of the modified targets. As a reversible process of glutamylation, a group of tubulin tyrosine ligase-like (TTLL) enzymes add the glutamate side chains (Janke et al., 2005), while members of the cytosolic carboxypeptidase (CCP) family of enzymes remove them (Rogowski et al., 2010). Given that TTLLs and CCPs are reversible-modification enzyme members with unique distributions, they might harbor nonredundant roles in the regulation of cellular processes by orchestrating glutamylation and deglutamylation of target proteins (Janke, 2014). Besides glutamylation of tubulin, several other target proteins have been recently identified to be glutamylated (van Dijk et al., 2008). We previously reported that Mad2 can be glutamylated to modulate megakaryocyte maturation (Ye et al., 2014). IL-7Rα is glutamylated to regulate the development of group 3 innate lymphoid cells (ILC3s; Liu et al., 2017a). Cyclic GMP-AMP synthase (cGAS) can be glutamylated to inhibit its synthase activity during DNA virus infections (Xia et al., 2016). However, how glutamylation regulates HSC self-renewal is still unclear.

BAP1 is a deubiquitinase that is involved in many cellular processes, including transcription regulation, cell cycle, proliferation, DNA damage, and cell death (Bononi et al., 2017). BAP1 mutations have been reported to be implicated in oncogenesis of several malignancies (Pilarski et al., 2014). Moreover, cancer-derived BAP1 mutations that abolish autodeubiquitination and promote its cytoplasmic sequestration abolish its function as a tumor suppressor (Mashtalir et al., 2014). Cytoplasmic BAP1 can deubiquitylate and stabilize IP3R3, which modulates calcium release from endoplasmic reticulum and enhances apoptosis (Bononi et al., 2017). Deletion of BAP1 in the hematopoietic system causes myelodysplastic syndrome (Dey et al., 2012). Here, we show that CCP3 deficiency impairs HSC self-renewal and hematopoiesis. CCP3 can deglutamylate BAP1 to promote its stability, which enhances Hoxa1 expression, leading to HSC self-renewal. BAP1 glutamylation at Glu651 is catalyzed by TTLL5 and TTLL7. Moreover, TTLL5 and TTLL7 deficiencies sustain BAP1 stability to promote HSC self-renewal and hematopoiesis.

We previously demonstrated that CCP6 deficiency in mice causes underdeveloped megakaryocytes and dysfunctional platelets (Ye et al., 2014). To further explore how glutamylation regulated hematopoiesis, we tested expression levels of Ccps in mouse bone marrow (BM). We found that only Ccp3 was most highly expressed in BM cells (Fig. 1 A). Of note, Ccp3-deleted (Ccp3^−/−) mice displayed splenomegaly and increased spleen weight (Fig. 1 B). Moreover, Ccp3^−/− mice showed disordered structure of splenic white pulp, suggesting extramedullary hematopoiesis (Fig. S1 A). In addition, Ccp3^−/− mice decreased cell counts of erythrocytes, myeloid cells, and lymphocytes in peripheral blood (Fig. 1 C, Fig. S1 B, and Table S1). By contrast, CCP4-deficient mice displayed normal spleen and cell numbers of blood lineages we tested (Fig. 1 B and Fig. S1, A and B). Consequently, Ccp3^−/− mice showed reduced BM cellularity compared with Ccp3^+/+ mice (Fig. 1 D and Fig. S1 C).

Blood cells can be replenished a short time after being damaged. To determine this regeneration capacity in Ccp3^−/− mice, we used 5-fluorouracil (5-FU) to eliminate proliferating cells and enforce hematopoietic regeneration. We observed that peripheral white blood cell (WBC) numbers in Ccp3^−/− mice were dramatically reduced after 5-FU treatment, and they failed to replenish blood cells and BM cells (Fig. S1, D and E). Furthermore, Ccp3^−/− mice could not survive three rounds of continuous 5-FU treatment, while WT mice could (Fig. S1 F). We used carboxypeptidase inhibitor phenanthroline (Rogowski et al., 2010) to suppress CCP3 activity in WT mice during 5-FU treatment. We noticed that phenanthroline treatment showed phenotypes similar to Ccp3^−/− mice (Fig. S1, G and H), suggesting CCP3 enzymatic activity is indispensable for hematopoietic regeneration capacity.

HSCs dominate the ultimate source of blood cell regeneration. We found that Ccp3 was most highly expressed in HSCs compared with other lineages (Fig. S1 I). Of note, Ccp3^−/− mice showed decreased total numbers and total frequencies of LSKs (Lin^–Sca-1⁺c-Kit⁺), MPPs (Lin^–Sca-1⁺c-Kit⁺CD48⁺CD150^–), and LT-HSCs (Lin^–Sca-1⁺c-Kit⁺CD48^–CD150⁺; Fig. 1 E and Fig. S1 J). HSCs give rise to common myeloid progenitors (CMPs) and common lymphoid progenitors (CLPs) and further differentiate into mature blood components hierarchically (Xia et al., 2015). Consequently, Ccp3^−/− mice showed decreased numbers of CMPs and CLPs as well as blood cell counts (Fig. 1 F, Fig. S1 K, and Table S1). In addition, we performed a CFU assay to test the role of CCP3 in the activity of hematopoietic stem and progenitor cells (HSPCs) in vitro. We observed that Ccp3^−/− BM cells produced fewer colonies compared with those of Ccp3^+/+ littermate control mice (Fig. 1 G). Moreover, Ccp3^−/− BM cells generated much lower numbers of colonies in secondary replating formation (Fig. 1 H). To determine in vitro differentiation and maintenance ability of Ccp3-deficient LT-HSCs, we performed serial plating CFU assays using Ccp3^+/+ and Ccp3^−/− LT-HSCs. CFU formations of primary plating were comparable between Ccp3^+/+ and Ccp3^−/− LT-HSCs (Fig. 1 I). However, Ccp3^−/− LT-HSCs produced much fewer CFUs in successive plating (Fig. 1 I), suggesting Ccp3^−/− LT-HSCs disrupt maintenance and self-renewal ability. A single LT-HSC can undergo population expansion for a certain time after being cultured with cytokine cocktail (Himburg et al., 2010). We found that Ccp3^−/− LT-HSCs lost their expanding and regenerating ability through single-cell culture assay (Fig. 1 J). In addition, phenanthroline treatment impeded WT LT-HSC expansion, and the CCP enzyme agonist CoCl₂ (Berezniuk et al., 2012) remarkably increased their expansion. However, with phenanthroline or CoCl₂ treatment, expansion abilities of Ccp3^−/− LT-HSCs were unchanged, suggesting a specific role of CCP3 in the regulation of LT-HSC expansion (Fig. 1 J). Of note, Ccp3^−/− LT-HSCs failed to be recovered after 5-FU treatment (Fig. S1 L). Finally, we observed that Ccp3^−/− mice had fewer quiescent LT-HSCs and greater number of cycling HSCs (Fig. 1 K). Of note, Ccp3^−/− LT-HSCs did not show apparent cell death (Fig. 1 L). Collectively, CCP3 deficiency impairs hematopoiesis and HSC self-renewal.

We next sought to determine whether CCP3 deficiency–mediated impairment of hematopoiesis was intrinsic or extrinsic. We transplanted CD45.2⁺ Ccp3^+/+ or Ccp3^−/− BM cells into lethally irradiated CD45.1⁺ recipients (Fig. 2 A). Engraftment of Ccp3^−/− BM cells decreased numbers of peripheral blood cells and BM cells (Fig. 2, B and C). Then we performed LT-HSC transplantation. We found that homing efficiency of Ccp3^−/− LT-HSCs was comparable to Ccp3^+/+ LT-HSCs (Fig. S2 A). However, Ccp3^−/− LT-HSC transplantation generated fewer peripheral blood cells and BM cells and failed to maintain LT-HSC self-renewal and hematopoiesis, whereas transplantation of Ccp3^+/+ LT-HSCs could produce normal numbers of peripheral blood cells and BM cells and maintain normal long-term hematopoiesis as well (Fig. 2, D and E; and Fig. S2, B and C).

We also performed competitive BM transplantation assays. We transplanted a 1:1 mixture of CD45.1⁺ WT and CD45.2⁺ Ccp3^−/− or Ccp3^+/+ BM cells into lethally irradiated recipient mice (Fig. 2 F). Engraftment of Ccp3^+/+ BM cells kept long-term balance with their competitors, whereas transplantation of Ccp3^−/− BM cells was gradually reduced during competition (Fig. 2 G). 16 wk after transplantation, transplanted Ccp3^−/− BM cells had reduced counts of peripheral blood cells and LT-HSCs compared with their WT competitors (Fig. 2, G and H). These results indicate that CCP3 is an intrinsic factor in regulating HSC self-renewal and hematopoiesis.

It has been reported that Glu540-to-Ala mutation abolishes deglutamylase activity of CCP3 (Tort et al., 2014). To examine whether CCP3 deficiency–mediated impairment of hematopoiesis was dependent on CCP3 enzymatic activity, we generated the Glu540-to-Ala mutation construct (CCP3-mut) and overexpressed it into Ccp3^−/− LT-HSCs, followed by reconstitution assay (Fig. S2 D). Through BM transplantation assay, we found that engraftment of LT-HSCs with CCP3-wt overexpression could restore hematopoiesis, whereas transplantation of LT-HSCs with CCP3-mut overexpression still abrogated hematopoiesis (Fig. 2 I). These data suggest that enzymatic activity of CCP3 is required for the regulation of hematopoiesis.

To further determine the molecular mechanism of CCP3-mediated hematopoiesis, we analyzed lysates of Ccp3^−/− or Ccp3^+/+ LSK cells by immunoblotting with a glutamylation-specific antibody: GT335. The antibody GT335 specifically recognizes the branch points of glutamate side chains and detects all glutamylation forms of target proteins (Ye et al., 2018). After Western blot analysis, one band around 95 kD appeared in the lane of CCP3-deficient LSK lysates (Fig. 3 A). This band was undetectable in the corresponding lane location from the littermate control LSK lysates. Therefore, this band could be a potential candidate substrate for CCP3. To identify the candidate substrates of CCP3, we generated CCP3-mut through Glu540-to-Ala mutation. CCP3-wt and CCP3-mut were immobilized in Affi-Gel 10 resin to go through mouse BM lysates for affinity chromatography. The eluted fractions were resolved by SDS-PAGE, followed by silver staining. This band was present in the gel analyzing CCP3-mut and was cut for mass spectrometry, where the band was identified as BAP1 (Fig. 3 B and Fig. S3 A). BAP1 is a nuclear-localized deubiquitinating enzyme that participates in the regulation of development, cell proliferation, and tumor transformation (Bononi et al., 2017; Dey et al., 2012; Yu et al., 2010). However, how glutamylation of BAP1 regulates hematopoiesis is still unknown.

We next overexpressed Myc-tagged BAP1 and Flag-tagged CCP3-wt or CCP3-mut in HEK293T cells for coimmunoprecipitation assay. We observed that the Flag-tagged CCP3-mut could pull down the Myc-tagged BAP1. Furthermore, CCP3-mut was able to immunoprecipitate endogenous BAP1, whereas CCP3-wt could not (Fig. 3 D). Since CCP3-wt could remove glutamylation of BAP1, degutamylated BAP1 could not bind to active CCP3-wt. In addition, endogenous BAP1 was highly glutamylated in Ccp3^−/− LSK lysates, while it was less glutamylated in Ccp3^+/+ LSK lysates (Fig. 3 E). Finally, phenanthroline treatment remarkably increased BAP1 glutamylation, whereas CoCl₂ treatment almost abolished the glutamylation of BAP1 (Fig. 3 F). Collectively, BAP1 is indeed glutamylated in LSK cells.

Nine glutamylase members of the TTLL family have been recently identified (Garnham et al., 2015). We wanted to explore which TTLL members catalyzed BAP1 in HSCs. We tested expression levels of these nine TTLL members in HSCs and mature lineage cells. We observed that Ttll5 and Ttll7 were most highly expressed in LSKs, especially in LT-HSCs (Fig. 4 A). We then expressed mouse TTLL5 and TTLL7 in HEK293T cells and incubated them with recombinant mouse BAP1 for in vitro glutamylation assay. We noticed that BAP1 was highly glutamylated by TTLL5 and TTLL7 (Fig. 4, B and C). By contrast, CCP3 could remove the glutamylation of BAP1 (Fig. 4, B and C). These results indicate that BAP1 is glutamylated by TTLL5 and TTLL7, whose glutamylation is removed by CCP3.

Glutamylation has been found to be modified on the glutamate-rich stretches and acidic environment at the acceptor sites (van Dijk et al., 2008). We analyzed conservative amino acid sequences of BAP1 and screened out conserved identical glutamic acid residues for mutations (Fig. S3 B). Through in vitro glutamylation assay, we identified that only the Glu651-to-Ala mutation of BAP1 (BAP1-E651A) abolished its glutamylation, indicating that BAP1 is catalyzed by TTLL5 and TTLL7 at Glu651 (Fig. 4 D and Fig. S3, C and D). We next generated Ttll5- and Ttll7-deficient mice, as well as Bap1^E651A knock-in mutant mice via CRISPR-Cas9 technology (Fig. S3, E–G). As expected, we found that glutamylation of BAP1 was dramatically reduced in TTLL5- and TTLL7-deficient LSKs, and BAP1 failed to undergo glutamylation in Bap1^E651A knock-in mutant mice (Fig. 4 E). These data confirm that BAP1 is glutamylated by TTLL5 and TTLL7 at Glu651.

Of note, we observed that Bap1^E651A mice displayed an increased percentage and number of LT-HSCs and consequently increased numbers of peripheral blood cells and BM cells (Fig. 4, F and G). As expected, more LT-HSCs were kept in G₀ phase in Bap1^E651A mice (Fig. 4 H). Through serial LT-HSC transplantation, we noticed that engraftment of Bap1^E651A LT-HSCs could produce more LT-HSCs and peripheral blood cells (Fig. 4, I and J). We conclude that glutamylation negatively regulates BAP1 function in the regulation of HSC maintenance and hematopoietic reconstruction ability.

It has been reported that the ubiquitin-conjugating enzyme UBE2O interacts with BAP1 to catalyze its ubiquitination (Mashtalir et al., 2014). We found that BAP1 protein level in Ccp3^−/− LSKs was markedly lower than that in Ccp3^+/+ cells (Fig. 5 A). However, Bap1 mRNA levels were comparable in Ccp3^−/− and Ccp3^+/+ LSKs (Fig. 5 B). These data suggest that BAP1 glutamylation might facilitate its degradation. With cycloheximide (CHX) treatment, BAP1 in Ccp3^−/− LSKs was rapidly degraded, whereas BAP1 in Ccp3^+/+ LSKs was more stable (Fig. 5 C). Of note, the proteasome inhibitor MG132 could impede BAP1 degradation (Fig. 5 C). These data suggest that BAP1 glutamylation facilitates its degradation. In addition, the interaction between BAP1 and UBE2O was confirmed by coimmunoprecipitation assay (Fig. 5 D). Moreover, BAP1 glutamylation enhanced their interaction (Fig. 5 D). In parallel, BAP1 glutamylation surely promoted its UBE2O-mediated ubiquitination (Fig. 5 E). Consistently, BAP1 glutamylation only enhanced K48-linked ubiquitination, but not K63-linked ubiquitination (Fig. 5 F). More importantly, K48-linked ubiquitination of BAP1 was much more accumulated in Ccp3^−/− LSKs than in Ccp3^+/+ BM LSKs (Fig. 5 G). However, BAP1 in Bap1^E651A LSKs was still stable following CHX treatment (Fig. 5 H). Altogether, BAP1 glutamylation promotes its interaction with UBE2O and accelerates K48-linked ubiquitination of BAP1 for degradation.

We next conducted transcriptome analysis between Ccp3^+/+ and Ccp3^−/− LT-HSCs through RNA sequencing. Of differential gene changes of Ccp3^+/+ and Ccp3^−/− LT-HSCs, 601 genes were downregulated and 532 genes were upregulated (Fig. S4 A). Hematopoiesis and lymphocyte differentiation-related genes were downregulated (Fig. S4 B). Through gene set enrichment analysis (GSEA), HSPC-related gene sets were enriched in Ccp3^+/+ LT-HSCs, whereas cell cycle–related gene sets were enriched in CCP3-deficient LT-HSCs (Fig. S4 C). Since Ccp3^−/− cells showed reduced BAP1 protein levels, we supposed that some differential genes might overlap between Ccp3^−/− HSCs and BAP1-deficient HSCs. In fact, Ccp3^−/− HSCs displayed overlapping differential genes with BAP1-deficient LSKs (Fig. S4 C). Transcription factors (TFs) were essential to HSC function, and we found that many TFs were differentially regulated in Ccp3^−/− versus Ccp3^+/+ LT-HSCs (Fig. 6 A). The top 10 downregulated TFs in Ccp3^−/− HSCs were further verified via quantitative PCR (qPCR; Fig. S4 D). We then silenced these 10 TFs in HSCs by shRNAs and followed by single-cell culture assay (Fig. S4 E). We observed that Hoxa1 depletion most significantly impaired HSC expansion (Fig. 6 B). Hoxa1 is a member of the homeobox (Hox) gene encoding TFs, which plays a critical role in development and tumor progression (Bach et al., 2010; Makki and Capecchi, 2010; Wang et al., 2015). Of note, we observed that only Hoxa1 was remarkably downregulated in Ccp3^−/− LT-HSCs, while other HoxA genes were unchanged between Ccp3^+/+ and Ccp3^−/− LT-HSCs (Fig. S4 F). We next wanted to determine how BAP1 regulated Hoxa1 expression. Through chromatin immunoprecipitation (ChIP), we found that BAP1 accumulated onto Hoxa1 promoter (Fig. 6 C). BAP1 overexpression in HSCs facilitated Hoxa1 transcription, whereas overexpression of BAP1 with Hoxa1 promoter binding region deletion had no such effect (Fig. 6 D and Fig. S4 G). By contrast, BAP1 depletion decreased Hoxa1 expression. Additionally, BAP1 overexpression in Ccp3^−/− HSCs rescued Hoxa1 expression (Fig. 6 E and Fig. S4 H). Consistently, CCP3 deficiency or BAP1 depletion caused the Hoxa1 promoter to be more resistant to DNase I digestion, and BAP1 overexpression in Ccp3^−/− HSCs made the Hoxa1 promoter more accessible (Fig. 6 F). These data suggest that BAP1 accumulates on the Hoxa1 promoter and facilitates Hoxa1 expression.

As a nuclear deubiquitylase, BAP1 can remove monoubiquitin from histone H2A lysine 119 (H2AK119), playing a critical role in regulating gene expression (Balasubramani et al., 2015). We then measured H2AK119Ub and H3K27me3 on Hoxa1 promoter in Ccp3^−/− LSKs. We found elevated enrichment of H2AK119Ub and H3K27me3 on the Hoxa1 promoter in Ccp3^−/− LSKs (Fig. S4, I and J). These data suggest that BAP1 could indirectly regulate Hoxa1 expression through effects on H2AK119Ub and other chromatin marks in HSCs.

We next generated Ttll5 and Ttll7 double knockout (DKO) mice (Ttll5^−/−;Ttll7^−/−). We found that BAP1-wt overexpression in Ttll5^−/−;Ttll7^−/− HSCs could promote Hoxa1 expression (Fig. 6 G and Fig. S4 K). However, co-overexpression of BAP1-wt with TTLL5 or TTLL7 in Ttll5^−/−;Ttll7^−/− HSCs suppressed this transcription-promoting effect. By contrast, overexpression of BAP1^E651A plus TTLL5 or TTLL7 in Ttll5^−/−;Ttll7^−/− HSCs restored Hoxa1 transcription (Fig. 6 G and Fig. S4 K). These data suggest that BAP1 glutamylation suppresses Hoxa1 transcription. In addition, Hoxa1 was dramatically decreased in Ccp3^−/− HSCs, whereas it was increased in Bap1^E651A HSCs (Fig. 6 H). Parallelly, BAP1 was not deposited on Hoxa1 promoter in Ccp3^−/− HSCs, while it was enriched on the promoter region in Bap1^E651A HSCs (Fig. 6 I). Accordingly, BAP1-E651A caused Hoxa1 promoter to be more accessible to DNase I digestion, indicating the open state of the Hoxa1 promoter (Fig. 6 J). These data indicate that BAP1 glutamylation causes its degradation to suppress Hoxa1 expression.

We next generated Hoxa1^−/− LT-HSCs via Cas9 knock-in mice as previously described (Platt et al., 2014; Fig. S4 L), followed by BM transplantation assay. We observed that engraftment of Hoxa1^−/− LT-HSCs generated reduced numbers of LT-HSCs and peripheral blood cells (Fig. 6, K and L). Of note, Hoxa1^−/− LT-HSCs could not sustain their quiescence, similar to Ccp3^−/− LT-HSCs (Fig. S4 M). However, Hoxa1^−/− LT-HSCs did not undergo apparent apoptosis (Fig. S4 N). Through serial LT-HSC transplantation, Hoxa1^−/− LT-HSCs failed to reconstitute peripheral blood cells and could not maintain their self-renewal (Fig. 6, M and N). Transplantation of Hoxa1-overexpressing Ccp3^−/− LT-HSCs could restore the normal numbers of HSCs and peripheral blood cells compared with engraftment of WT LT-HSCs (Fig. 6 O and Fig. S4 O). These results indicate that Hoxa1 is required for the maintenance of HSC self-renewal.

Given that BAP1 glutamylation was catalyzed by TTLL5 and TTLL7, we next tested how Ttll5 and Ttll7 deficiencies affected hematopoietic phenotypes. We observed that Ttll5- and Ttll7-deficient mice displayed higher numbers and percentages of LSKs and LT-HSCs compared with their WT littermate control mice (Fig. 7 A and Fig. S5 A). Moreover, Ttll5 and Ttll7 DKO mice showed much higher numbers of LSKs and LT-HSCs compared with Ttll5- or Ttll7-deficient mice alone (Fig. 7 A). As expected, Ttll5^−/− or Ttll7^−/− LSKs showed much less glutamylation of BAP1 and much higher protein levels of BAP1 compared with WT control mice (Fig. 7 B). In addition, Hoxa1 was highly expressed in Ttll5^−/− and Ttll7^−/− HSCs (Fig. 7, B and C). Consequently, Ttll5- and Ttll7-deficient mice had higher numbers of BM and peripheral blood cells than WT mice (Fig. 7 D and Fig. S5 B). In addition, through in vitro replating clone formation assay, Ttll5^−/− and Ttll7^−/− LT-HSCs showed stronger expansion capacities, and DKO LT-HSCs showed much stronger expansion capacity than Ttll5^−/− or Ttll7^−/− mice alone (Fig. 7 E). Through BM transplantation, engraftment of Ttll5^−/− and Ttll7^−/− LT-HSCs displayed increased numbers of LT-HSCs and peripheral blood cells (Fig. 7, F and G). Finally, Ttll5^−/− and Ttll7^−/− LT-HSCs kept quiescent similar to Bap1^E561A LT-HSCs (Fig. 7 H). Collectively, TTLL5 and TTLL7 deficiencies promote HSC self-renewal and hematopoiesis.

HSCs maintain self-renewal with a sophisticated mechanism to provide lifelong hematopoiesis. However, the mechanism underlying HSC self-renewal is still elusive. In this study, we showed that CCP3 is most highly expressed in BM cells among CCP members. CCP3 deficiency impairs HSC self-renewal and hematopoiesis. BAP1 is a substrate for CCP3 in LT-HSCs. BAP1 is catalyzed at Glu651 by TTLL5 and TTLL7, and BAP1-E651A mutation abrogates BAP1 glutamylation. BAP1 glutamylation accelerates its ubiquitination to trigger its degradation. CCP3 can remove glutamylation of BAP1 to promote its stability, which enhances Hoxa1 expression leading to HSC self-renewal. Bap1^E651A mice produce higher numbers of LT-HSCs and peripheral blood cells, and Bap1^E651A LT-HSCs were more quiescent. Moreover, TTLL5 and TTLL7 deficiencies sustain BAP1 stability to promote HSC self-renewal and hematopoiesis (Fig. S5 C).

Protein glutamylation is catalyzed by a family of polyglutamylases, also called TTLLs (Janke et al., 2005). The well-known substrates of glutamylation are tubulins. Tubulins are glutamylated at their acidic, glutamate-rich C termini, which are the binding sites for most microtubule (MT)-associated proteins (Janke and Bulinski, 2011). Thus, tubulin glutamylation generates functionally divergent MTs by regulating the affinity between MT-associated proteins and MTs (van Dijk et al., 2008). Through regulating MT character, glutamylation is therefore proposed to be involved in MT-related cellular processes, including stability of centrosomes, motility of cilia and flagella, and neurite outgrowth, as well as neurodegeneration (Bosch Grau et al., 2013; Rogowski et al., 2010). Given that TTLLs have different expression patterns in diverse tissues and show nonredundant functions, more novel substrates of glutamylation need to be defined. We recently reported that TTLL4 and TTLL6 are highly expressed in megakaryocytes and catalyze glutamylation of Mad2 to modulate megakaryocyte maturation (Ye et al., 2014). IL-7Rα can be catalyzed by TTLL4 and TTLL13, and IL-7Rα glutamylation initiates TF Sall3 expression in common helper-like innate lymphoid progenitors, leading to development of ILC3 cells (Liu et al., 2017a). In this study, we demonstrated that TTLL5 and TTLL7 are most highly expressed in HSCs and catalyze BAP1 glutamylation to facilitate its ubiquitylation for degradation, which regulates HSC self-renewal and hematopoiesis.

Protein glutamylation is also a reversible modification, whose glutamylation is removed by a family of carboxypeptidases, also called CCPs (Janke et al., 2005). It has been reported that CCP family members catalyze deglutamylation of tubulins and display enzymatic specificities (Rogowski et al., 2010). We previously showed that CCP6 hydrolyzes Mad2 glutamylation to modulate megakaryocyte maturation (Ye et al., 2014). We also reported that CCP2 is highly expressed in common helper-like innate lymphoid progenitors, the progenitor of ILCs, and catalyzes IL-7Rα deglutamylation, leading to ILC3 development (Liu et al., 2017a). We recently demonstrated that cGAS can be glutamylated by TTLL4 and TTLL6, whose glutamylation can be hydrolyzed by CCP5 and CCP6 (Xia et al., 2016). Glutamylation and deglutamylation of cGAS tightly regulate immune responses to DNA virus infections. Herein, we showed that CCP3 is most highly expressed in BM cells and hydrolyzes BAP1 glutamylation to modulate HSC self-renewal and hematopoiesis. These findings suggest that different tissue and cell type distributions of TTLLs and CCPs exert unique roles in the regulation of different physiological processes.

BAP1 is a member of the ubiquitin C-terminal hydrolase subfamily of deubiquitylating enzymes, which is associated with multiprotein complexes to regulate various cellular processes (Carbone et al., 2013). BAP1 mutations are implicated in various malignancies, including myeloid transformation (Abdel-Wahab and Dey, 2013). BAP1 catalytic mutation has been identified in myelodysplastic syndrome patients (Carbone et al., 2013). In addition, BAP1-deficient mice also manifest myeloid transformation (Dey et al., 2012). It has been reported that BAP1 locates at the nucleus and interacts with TFs or multiprotein complexes to regulate transcription initiation and elongation (Carbone et al., 2013; Yu et al., 2010). However, how BAP1 regulates HSC self-renewal and hematopoiesis is still unknown. In actuality, we found elevated enrichment of H2AK119Ub and H3K27me3 on the Hoxa1 promoter in Ccp3^−/− LSKs, suggesting that BAP1 could indirectly regulate Hoxa1 expression through effects on H2AK119Ub and other chromatin marks in HSCs. We are still exploring the molecular mechanism by which BAP1 glutamylation modulates Hoxa1 expression in a direct or indirect manner in HSCs. TTLL5- and TTLL7-mediated BAP1 glutamylation promotes its interaction with ubiquitin-conjugating enzyme UBE2O to facilitate K48-linked ubiquitination for its degradation. TTLL5 and TTLL7 deficiencies maintain BAP1 stability, and stable BAP1 further promotes Hoxa1 expression, leading to enhancement of HSC self-renewal and hematopoiesis. As expected, Hoxa1 deletion causes reduced numbers of LT-HSCs and peripheral blood cells. In summary, glutamylation and deglutamylation of BAP1 play a critical role in the regulation of HSC self-renewal and hematopoiesis.

The following commercial antibodies were used: mouse hematopoietic lineage eFlour 450 cocktail (eBioscience; 22–7775), PerCP-Cy5.5–anti-CD45.1 (eBioscience; 45–0453), FITC–anti-CD45.2 (eBioscience; 11–0454), FITC–anti–IL-7Rα (eBioscience; 11–1271), APC–anti-Ly6A/E (Sca-1, eBioscience; 17–5981), PE–anti-CD117 (c-Kit, eBioscience; 12–1171), PE-Cy7–anti-CD16/32 (eBioscience; 25–0161), APC-eFluor 780–anti-CD48 (eBioscience; 47–0481), PE-CY7–anti-CD150 (eBioscience; 25–1502), eFluor 450–anti-CD3 (eBioscience; 48–0031), PE–anti-CD19 (eBioscience; 12–0193), FITC–anti–Gr-1 (eBioscience; 11–5931), Alexa Fluor 700–anti-CD34 (eBioscience; 56–0341), FITC–anti–Ki-67 (eBioscience; 7B11), and FITC–anti–Annexin V (eBioscience; VAA-33). Anti-CCP3 (16990–1-AP) antibody was purchased from Proteintech. Anti-CCP4 (T-17), anti-GST (6G9C6), anti-Myc (9E10), anti-Flag (M1), anti–β-actin (SP124), and anti-His (6AT18) antibodies were bought from Sigma-Aldrich. GT335 antibody was obtained from AdipoGen. BAP1 (D1W9B), K48 linkage–specific polyubiquitin (D9D5), ubiquityl-histone H2A (Lys119; D27C4), and tri-methyl-histone H3 (Lys27; C36B11) antibodies were purchased from Cell Signaling Technology. TTLL5 (ARP75440_P050) antibody was obtained from Aviva Systems Biology. TTLL7 (N1N2) antibody was bought from GeneTex. Hoxa1 (BA3730-2) antibody was bought from Boster Biological Technology. Paraformaldehyde, CHX, MG132, phenanthroline, 5-FU, and DAPI were purchased from Sigma-Aldrich. CoCl₂ was from Sinopharm Chemical Reagent. EDTA-free protease inhibitor cocktail and DNase I were purchased from Roche Molecular Biochemicals. West Pico and West Femto plus chemiluminescent substrate were purchased from SageBrightness.

Ccp1- and Ccp6-deficient mice were described previously (Ye et al., 2014). Ccp2-, Ccp3-, Ccp4-, and Ccp5-deficient mice were generated using CRISPR-Cas9 approaches as described (Xia et al., 2016). For generation of Ttll5- and Ttll7-deficient mice, vector pST1374-NLS-flag-linker-Cas9 (Addgene plasmid #44758) expressing Cas9 and pUC57-sgRNA (Addgene plasmid #51132) expressing small guide RNAs (sgRNAs) targeting Ttll5 and Ttll7 genes were constructed (Table S2). Mixtures of Cas9 mRNA (100 ng/µl) and sgRNA (50 ng/µl) were microinjected into the cytoplasm of C57BL/6 fertilized eggs, followed by transfer to the uterus of pseudo-pregnant ICR-background female mice, from which viable founder mice were obtained. Genotyping of KO mice with indicated primers (Table S3) was performed as previously described (Zhu et al., 2014). We chose frameshift mutation by PCR screening and TA clone for sequencing and then confirmed knockout efficiency by Western blot. For generation of Bap1^E651A mice, the genome locus of Bap1 gene was knocked in with BAP1-E651A mutation via a CRISPR-Cas9 approach (Ye et al., 2018). A mixture of Cas9 mRNA, sgRNA, and BAP1-E651A donor templates was microinjected into the cytoplasm of C57BL/6 fertilized eggs and transferred into the uterus of pseudo-pregnant ICR females. BAP1-E651A mutants were identified by PCR screening and TA clone for DNA sequencing. The gRNA sequence for Bap1^E651A mice was up, 5′-GCC​CCT​AAG​GTA​TAC​AAT​GT-3′; down, 5′-CAG​CTG​TCC​TTG​GGC​AGT​AG-3′. Gt(ROSA)26Sor^{tm1(CAG-xstpx-cas9,-EGFP)Fezh} mice were purchased from The Jackson Laboratory (Platt et al., 2014). For deletion of Hoxa1, LT-HSCs from Gt(ROSA)26Sor^{tm1(CAG-xstpx-cas9,-EGFP)Fezh} knock-in mice were sorted and infected with lentivirus containing the indicated sgRNA and Cre recombinase expression. Then infected LT-HSCs were mixed with 5 × 10⁵ helper cells and transplanted into lethally irradiated recipient mice (CD45.1). 1 mo later, GFP⁺ BM cells were sorted from the recipient mice to confirm Hoxa1 deficiency with Western blot. GFP⁺ LT-HSCs were sorted for further transplantion. All the mice we used were C57BL/6 background and ~3 mo old. We used littermates with the same age and gender for each group. We performed three independent experiments of each mouse from at least three mice for each group. Animal use and protocols were approved by the Institutional Animal Care and Use Committees at the Institute of Biophysics, Chinese Academy of Sciences.

Mouse spleens were fixed in 4% paraformaldehyde for 12 h at room temperature. Mouse femurs were fixed in PBS buffer containing 10% formaldehyde for 12 h and then decalcified in decalcifying buffer (10% EDTA in PBS, pH 7.4) for 48 h, changing new decalcifying buffer every 24 h. Fixed tissues were washed twice using 70% ethanol and embedded in paraffin, followed by sectioning and staining with H&E according to standard laboratory procedures.

BM cells were flushed out from femurs with precooled PBS buffer containing 2% FBS and sifted through 50-µm cell strainers. RBCs were removed by suspending cells in RBC lysis buffer, followed by washing twice with PBS. BM cells were counted with a blood counting chamber at least three times and then checked again through flow cytometry with a FACSAria III instrument (BD Biosciences). For hematopoietic lineage analysis, 10⁷ BM cells were incubated with fluorophore-conjugated antibodies at 4°C for 1 h and then washed twice. LSKs (Lin^–Sca-1⁺c-Kit⁺), LT-HSCs (Lin^–Sca-1⁺c-Kit⁺CD48^–CD150⁺), short-term HSCs (Lin^–Sca-1⁺c-Kit⁺CD48^–CD150^–), MPPs (Lin^–Sca-1⁺c-Kit⁺CD48⁺CD150^–), HSPCs (Lin^–Sca-1^–c-Kit⁺), CMPs (Lin^–Sca-1^–c-Kit⁺CD34⁺CD16/32^–), and CLPs (Lin^–CD127⁺Sca-1^lowc-Kit^low) were analyzed or sorted with a FACSAria III instrument. Lineage cocktail antibodies contained anti-B220, anti-CD3, anti-Ter119, anti–Gr-1, anti-CD11b, anti-CD19, and anti-NK1.1. For peripheral WBC flow cytometric analysis, blood samples were collected through tail veins, and 10 µl of blood was incubated with 200 µl RBC lysis buffer at room temperature for 2 min, followed by washing with PBS twice. T cells (CD3⁺), B cells (CD19⁺), and myeloid cells (CD11b⁺Gr-1⁺) were analyzed with a FACSAria III instrument. Data were analyzed using the FlowJo 7.6.1 software.

4 × 10⁴ BM cells from WT or KO mice were mixed with cytokine-supplemented methylcellulose medium (Methocult, M3434; STEMCELL Technologies) and plated in 35-mm tissue-culture dishes. After 10 d of culture at 37°C in 5% CO₂, granulocyte colonies, macrophage colonies, granulocyte-macrophage colonies, granulocyte, erythroid, macrophage, and megakaryocyte colonies, erythroid colonies, and megakaryocyte colonies were observed with an inverted microscope and assigned scores (Hou et al., 2015). For continuous replating colony-forming assay, 1 × 10⁴ BM cells were initially plated in M3434 medium. Colonies were scored after 7-d cultures, and cells were resuspended and washed with PBS. Then, 1 × 10⁴ cells were cultured again for a second or third replating (Moran-Crusio et al., 2011).

BM cells from the indicated mice (CD45.2) were flushed out from femurs with precooled PBS buffer and sifted through 70-µm cell strainers. 1 × 10⁶ BM cells were then transplanted into lethally irradiated (10 Gy) recipient mice (CD45.1) through tail veins. The indicated parameters were detected 16 wk after BM transplantation. For LT-HSC transplantation, 1 × 10² LT-HSCs were sorted and mixed with 5 × 10⁵ BM helpers (CD45.1) and then transplanted into lethally irradiated recipient mice (CD45.1), followed by examination at the times mentioned. For competitive BM transplantation, 1 × 10⁶ BM (CD45.2) cells were mixed with 1 × 10⁶ BM competitors (CD45.1) to be transplanted into lethally irradiated recipient mice (CD45.1⁺CD45.2⁺). Peripheral WBCs were detected every 4 wk. LT-HSCs from different donors were identified with anti-CD45.1 and anti-CD45.2 antibodies through flow cytometry 16 wk after transplantation.

In vitro single LT-HSC culture assay was performed as described previously (Rathinam et al., 2011). In brief, we isolated LT-HSCs (Lin^–Sca-1⁺c-Kit⁺CD48^–CD150⁺) by flow cytometry and sorted single cells into 96-well plates. Single LT-HSCs were cultured in vitro in the presence of the following recombinant cytokines: mouse stem cell factor (50 ng/ml), mouse thrombopoietin (10 ng/ml), mouse IL-3 (10 ng/ml), mouse IL-6 (10 ng/ml), and human Flt3L (50 ng/ml; all from Peprotech). Cells were cultured in IMDM supplemented with 10% (volume/volume) FCS, 2 mM L-glutamine, 1% (volume/volume) penicillin-streptomycin, and 1 mM nonessential amino acids. 3 d later, cells were transferred into 24-well plates with 1 ml fresh medium containing cytokines for further culture. After 7-d cultures, cells were counted with a blood counting chamber or analyzed by flow cytometry.

10⁶ sorted LSKs or cultured HSCs were lysed with RIPA buffer (150 mM NaCl, 0.5% sodium deoxycholate, 0.1% SDS, 1% NP-40, 1 mM EDTA, and 50 mM Tris, pH 8.0, containing protease inhibitor cocktail) at 4°C for 30 min, followed by separation with SDS-PAGE. Samples were then transferred onto nitrocellulose membranes and incubated with primary antibodies in 5% BSA at room temperature for 2 h. After washing with Tris-buffered saline containing 0.1% Tween-20 three times, membranes were incubated with HRP-conjugated secondary antibodies at room temperature for 1 h. Chemiluminescent signals were generated using West Pico or West Femto plus enhanced chemiluminescent substrate.

HEK293T cells were transfected with the indicated plasmids and cultured for 48 h. For endogenic immunoprecipitation, 1 × 10⁶ LSKs were sorted or LT-HSCs were cultured for expansion. Cells were lysed with radioimmunoprecipitation assay buffer at 4°C for 1 h. Lysates were incubated with the indicated antibodies for 2 h and immunoprecipitated with protein A/G agarose beads for 1 h, followed by SDS-PAGE separation and immunoblotting.

Cell populations were isolated by flow cytometry. Total RNAs were extracted with RNA Miniprep Kit (Tiangen Biotech) according to the manufacturer’s protocol. Then, cDNA was synthesized with M-MLV reverse transcription (Promega). mRNA transcripts were analyzed with the ABI 7300 qPCR system using specific primer pairs as listed in Table S3. Relative expressions were calculated and normalized to endogenous Actb expression.

cDNAs were cloned from a BM cDNA library. CCP3-wt, CCP3-mut, BAP1-wt, and indicated BAP1 mutants were cloned into pGEX6p-1 plasmid for GST-tagged protein expression. Plasmids were transformed into Escherichia coli strain BL21 (DE3), followed by induction with 0.1 mM isopropyl β-D-1-thiogalactopyranoside at 16°C for 24 h. Cells were collected and lysed by an ultrasonic cell disruption system (Branson), followed by purification with GST resins.

ChIP assay was described previously (Liu et al., 2017b). In brief, 1 × 10⁶ LSKs or cultured HSCs were cross-linked with 1% formaldehyde at 37°C for 10 min. Then, cells were washed twice with PBS, lysed, and sonicated to get 300–500-bp DNA fragments. Lysates were incubated with 4 µg anti-BAP1 antibody rolling overnight at 4°C. Salmon sperm DNA/protein agarose beads were added for DNA immunoprecipitation. After washing, DNA was eluted from beads and purified. DNA fragments were extracted and analyzed with qPCR. Primers used for ChIP are listed in Table S4.

Target sequences for RNA interference were designed according to MSCV-LTRmiR30-PIG (LMP) system instructions. LMP vectors containing target sequences were constructed. shRNA sequences are listed in Table S5. For gene overexpression, the indicated genes were cloned into pMYs-IRES-GFP (pMYs) retrovirus vectors. LMP or pMYs vectors were cotransfected with packaging plasmid PCL122 into HEK293T cells for 48 h. Media containing virus particles were collected and ultracentrifugated at 25,000 rpm (82,700 g) for 2 h for viral concentration. Pellets were resuspended in IMDM, and viral titers were determined by infecting HEK293T cells with diluted viruses. LT-HSCs were sorted and incubated with viruses in the presence of 8 µg/ml polybrene, followed by centrifuging at 500 g for 2 h. After 36-h culture to allow gene expression, GFP⁺ cells were sorted for transplantation or further cultured for another 2 d to get enough cells for analyzing gene expression or other experiments as mentioned.

DNase I digestion assay was performed as described previously (Liu et al., 2017a). In brief, nuclei were purified from 1 × 10⁶ LSKs or cultured HSCs, according to the manufacturer’s protocol, with the Nuclei Isolation Kit (Sigma-Aldrich). Then, nuclei were resuspended with DNase I digestion buffer and treated with indicated units of DNase I (Sigma-Aldrich) at 37°C for 5 min. 2× DNase I stop buffer (20 mM Tris, pH 8.0, 4 mM EDTA, and 2 mM EGTA) was added to stop reactions. DNA was extracted and examined by qPCR.

Detailed protocol for in vitro glutamylation assay was previously described (Ye et al., 2014). In brief, CCP3, TTLL5, and TTLL7 were transfected into HEK293T cells for 48 h. Cells were harvested and lysed. Supernatants were incubated with GST-BAP1 or indicated mutants at 37°C for 2 h. GST-BAP1 was precipitated, followed by immunoblotting to detect glutamylation with GT335 antibody.

10⁴ LT-HSCs (Lin^–Sca-1⁺c-Kit⁺CD48^–CD150⁺) from Ccp3^+/+ and Ccp3^−/− mice were sorted by flow cytometry. Total RNA was extracted with the RNA Miniprep Kit and was qualified using Agilent 2100 for the construction of sequencing libraries. Libraries were sequenced on BGISEQ-500 using 50-bp single-end reads. Heatmap.2, ggplot2, and clusterProfiler in Bioconductor were used for generating heatmap, volcano plot, and gene ontology analyzing. GSEA v4.0.1 was used. RNA-seq data have been deposited under GEO accession no. GSE138298.

For statistical analysis, data were analyzed by Sigma Plot or GraphPad Prism 5.0. Two-tailed unpaired Student’s t test was used in this study. P values < 0.05 were considered significant (*, P < 0.05; **, P < 0.01; ***, P < 0.001); P values > 0.05 were considered nonsignificant. All flow cytometry data were analyzed with FlowJo (Treestar).

Fig. S1 shows that Ccp3 deletion impairs HSC expansion. Fig. S2 shows that CCP3 plays an intrinsic role in HSC self-renewal. Fig. S3 shows that BAP1 is glutamylated at E651 by TTLL5 and TTLL7. Fig. S4 shows that BAP1 targets the Hoxa1 promoter to enhance Hoxa1 expression. Fig. S5 shows that Ttll5 and Ttll7 deficiencies promote HSC self-renewal. Table S1 shows hematopoietic cell counts in peripheral blood of Ccp3^+/+ and Ccp3^−/− mice. Table S2 shows sgRNA sequences used in this study. Table S3 shows sequences of primers used for genotyping and qPCR. Table S4 shows sequences of primers used in ChIP assays. Table S5 shows sequences for shRNAs used in this study.

We thank Peng Xue, Jianhua Wang, Di Liu, Yan Teng, Junying Jia, Shu Meng, Jing Cheng, Yihui Xu, Xudong Zhao, and Jianhui Li for technical support. We thank Xiang Shi and Xing Gao for animal procedures. We thank Zhimin Wang and the HPC Service Station for data analysis.

This work was supported by the National Natural Science Foundation of China (31930036, 81921003, 31530093, 91640203, 31871494, 31671531, 31670886, 81601361, 31570872, 31728006, 81572433, 81772646, 31601189); the Strategic Priority Research Programs of the Chinese Academy of Sciences (XDB19030203, XDA12020219); and the Beijing Natural Science Foundation (7181006).

Author contributions: Z. Xiong designed and performed experiments, analyzed data, and wrote the paper. P. Xia performed experiments and analyzed data. X. Zhu generated genetic mice. J. Geng, S. Wang, B. Ye, and X. Qin performed some experiments. Y. Qu, L. He, and D. Fan crossed and genotyped some mice. Y. Du analyzed some data. Y. Tian initiated the study and built up animal models; Z. Fan initiated the study, organized, designed, and wrote the paper.

The authors declare no competing financial interests.