A KRAB/KAP1-microRNA cascade regulates erythropoiesis through the stage-specific control of mitophagy

Abstract

Hematopoiesis is orchestrated by a succession of lineage- and stage-specific transcription factors working in concert with chromatin modifiers. Here, we explored the role of KRAB-containing zinc finger proteins (KRAB-ZFPs) and their cofactor KAP1 in this process. The hematopoietic-restricted deletion of Kap1 in the mouse resulted in severe hypoproliferative anemia, with Kap1-deleted erythroblasts failing to induce mitophagy-associated genes, hence to eliminate mitochondria. This was due to persistent expression of microRNAs targeting mitophagy transcripts, itself secondary to a lack of repression by stage-specific KRAB-ZFPs. This KRAB/KAP1-microRNA regulatory cascade is evolutionary conserved, as it also controls mitophagy during human erythropoiesis. A multilayered transcription regulatory system is thus unveiled, where protein- and RNA-based repressors are super-imposed in combinatorial fashion to govern the timely triggering of an essential differentiation event.

Erythropoiesis releases about one hundred billion new red cells every day from the human adult bone marrow. This process is initiated by the differentiation of hematopoietic stem cells (HSC) into the earliest erythroid progenitor identified ex vivo, the slowly growing burst-forming unit-erythroid (BFU-E). This cell morphs into the rapidly dividing CFU-E (colony-forming unit-erythroid), the proliferation of which is stimulated by the hypoxia-induced hormone erythropoietin. Further differentiation occurs through a highly sophisticated program orchestrated by lineage- and stage-specific combinations of protein- and RNA-based transcription regulators (1–3). It culminates in the elimination of all intracellular organelles including mitochondria and the nucleus to yield the fully mature erythrocyte, with on board some 250 million molecules of hemoglobin as almost sole cargo. Much is still to be learned about the molecular mechanisms of these events, not only to understand more fully the cause of red cell disorders, but also to envision the in vitro manufacturing of the large supplies of oxygen-carrying cells needed for transfusion.

Higher vertebrates encode hundreds of KRAB-ZFPs that can bind DNA in a sequence-specific fashion through a C-terminal array of C2H2 zinc fingers and recruit the corepressor KAP1 via their N-terminal KRAB domain (4–7). KAP1, also known as TRIM28 (tripartite motif protein 28), TIF1β (transcription intermediary factor 1 beta) or KRIP-1 (KRAB-interacting protein 1), acts as a scaffold for a multi-molecular complex that silences transcription through the formation of heterochromatin (8–11). The KRAB/KAP1 system probably evolved initially to minimize retroelement-induced genome perturbations (12–14), but recent data indicate that it also regulates multiple aspects of mammalian physiology (15–24). Accordingly, the present study was undertaken to explore its role in hematopoiesis.

The hemato-specific knockout of Kap1 in the mouse (Suppl. Fig. 1) resulted in a series of hematological abnormalities (Suppl. Table 1). The clinically most prominent was an ultimately fatal hypo-regenerative anemia, characterized by the accumulation of transferrin receptor/CD71+ glycophorin-A-associated/Ter119- early erythroblasts and an almost complete absence of mature CD71-Ter119+ cells in the bone marrow (Fig. 1A). Electron microscopy and Mitotracker staining further revealed that KO erythroblasts contained markedly more mitochondria than their wild type counterparts (Fig. 1B), and this correlated with decreased expression of mitophagy genes such as Nix/Bnip3L, Ulk1, GABARAP, sh3glb1, Beclin1 and Bcl2l1 (Fig. 2A). Since the KRAB/KAP1 pathway is mostly known to induce transcriptional repression, it seemed likely that this effect was indirect. An examination of the miRNA expression profile of control and Kap1 KO CD71+Ter119+ cells indeed revealed that, amongst 455 miRNAs tested, 5 were downregulated and 11 upregulated more than two-fold in KO cells. A recently described in silico approach (25, 26) suggested that several of these miRNAs had mitophagy-associated deregulated transcripts as their targets, notably miR-351, predicted to act on Bnip3L (Fig. 2A). Consistent with this hypothesis, levels of miR-351 abruptly dropped in CD71+Ter119+ cells, compared to their CD71+Ter119- precursors, mirroring Bnip3L induction (Fig. 2B). Furthermore, transduction of mouse erythroleukemia (MEL) cells with a GFP-expressing lentiviral vector harboring 3’ of GFP the Bnip3L 3’UTR sequence predicted to be targeted by miR-351 resulted in miR-351-dependent downregulation of the reporter (Fig. 2C). Finally, similar to their KAP1-depleted counterparts, miR-351-overexpressing MEL cells were blocked in differentiation and accumulated mitochondria, and this phenotype was reversed by expression of a Bnip3L transcript devoid of this 3’UTR sequence (Suppl. Fig. 2).

Fig. 1.

Blocked maturation and accumulation of mitochondria in Kap1-deleted erythroblasts. (A) FACS analysis of CD71 and Ter119 in bone marrow from control (Ctrl) and Kap1 KO mice 7 weeks after pIC injection. (B) Electron microscopy (left, stars indicate mitochondria; middle, average number of mitochondria visualized per cell; n=10, *p <0.05) and Mitotracker staining (right, n=4, *p<0.05). Decreased nuclear density was frequent in Kap1 KO cells, perhaps reflecting altered chromatin condensation.

Fig. 2.

A KAP1-miRNA cascade controls red cell mitophagy. (A) Top, mitophagy-related transcripts in erythroblasts from control (Ctrl) and Kap1 KO mice (n=4, *p<0.05, **p<0.01, ***p<0.001). Bottom, indicated miRNAs expression in same samples; predicted miRNA-target pairs are indicated by X. (B) miR351 and Bnip3L expression in megakaryocyte/erythroid progenitors (MEP: Lin−Sca1−CD117+ CD34−CD16.32−, in which expression was set at 1) and indicated erythroblast subsets. (C) MiR-351 targets the Bnip3L 3’UTR. Ctrl, for which the normalized value was set at 1, was a combination of MEL cells not overexpressing miR-351 and transduced with a GFP-expressing lentiviral vector with the Bnip3L 3’UTR, and cells overexpressing miR-351 but transduced with the same vector without this sequence (n=3, *p<0.05).

MiR-503 and miR-322*, located next to miR-351 on chromosome X, were also upregulated (2.46 and 2.17 fold, respectively) in Kap1 KO erythroblasts. Consistent with a role for KRAB/KAP1 in regulating this miRNA gene cluster, chromatin immunoprecipitation coupled to DNA sequencing (ChIPSeq) detected a strong KAP1 peak less than 4kb away (Fig. 3A). Because KAP1 is not a DNA binding protein, we postulated that it might be tethered to this and other relevant loci by stage-specific KRAB-ZFPs. Nine KRAB-ZFP genes were identified, which both had a human orthologue and were expressed exclusively in CD71+Ter119- and/or CD71+Ter119+ erythroblasts, but not in other hematopoietic cells. Six could be efficiently knocked down in MEL cells by lentivector-mediated RNA interference, and two of them, ZFP689 and ZFP13, emerged as potential Bnip3L regulators (Suppl. Fig. 3). Interestingly, ZFP689 is expressed in CD71+Ter119- erythroblasts, whereas ZFP13 is expressed only in their CD71−Ter119+ counterparts (Fig. 3B). Both could repress reporter expression in MEL cells transduced with a lentiviral vector harboring the miR-351-close KAP1-binding site upstream of a human phosphoglycerate kinase promoter murine secreted alkaline phosphatase (mSEAP) cassette (Fig. 3C). We then validated these two candidates in vivo by transplanting CD45.2 hematopoietic stem cells (lineage-, Sca1+ and cKit+, or LSK) transduced with lentiviral vectors producing GFP and shRNAs against Zfp689, Zfp13, or Kap1 as a control, into irradiated CD45.1 mice, allowing the dual discrimination of donor vs. recipient and transduced vs. untransduced cells. Analyses of the red cell compartment in bone marrow harvested eight weeks after the graft revealed that knockdown of either Zfp689 or Zfp13 led to a decrease in CD71+Ter119- cells as pronounced as that observed with the Kap1 knockdown (Fig. 3D). Furthermore, RNA analyses of sorted transduced CD71+Ter119+ cells demonstrated that ZFP689-, ZFP13- and KAP1-depleted cells all exhibited an upregulation of miR-351 (Fig. 3E) and a marked downregulation of Bnip3L (Fig. 3F).

Fig. 3.

Erythroblast-specific KRAB-ZFPs control the miR-351/Bnip3L/mitophagy axis. (A) Screen shots from the UCSC Genome Browser, with results of a KAP1 ChIPSeq analysis performed on CD71+Ter119+ bone marrow cells. (B) Zfp689 and Zfp13 are induced during erythroid differentiation. (C) ZFP689 and ZFP13 repress a lentiviral vector carrying a miR-351-close KAP1-binding site in transduced MEL cells (Ctrl is a combination of ZFP-overexpressing cells transduced with a vector without the KAP1-binding site and cells LacZ-overexpressing cells transduced with a vector carrying the KAP1-binding site; n=3, *p<0.05). (D) CD45.2+ LSK cells were transduced with GFP-expressing, empty or scramble (Ctrl), Kap1-, Zfp689- or Zfp13-directed shRNA lentiviral vectors, engrafted into irradiated CD45.1+ mice, and erythroid differentiation was evaluated by FACS 8 wks later. The CD71+Ter119+, CD45.2+, eGFP+ population was then sorted and analyzed by RT-QPCR for miR351 (E) and Bnip3L (F) expression (n=6, *p<0.05).

In a last series of experiments, we asked whether this erythropoiesis-regulating system has its equivalent in humans. We first found that Kap1 knockdown impaired the differentiation of human erythroleukemia (HEL) cells and increased their mitochondrial content (Fig. 4A-C), blocking several mitophagy effectors including Nix/Bnip3L (Fig. 4D). We further verified that KAP1-depleted HEL cells had increased levels of hsa-miR-125a-5p, which has the same seed as murine miR-351, and that overexpressing this miRNA triggered a downregulation of Nix and a rise in the mitochondrial content of these cells (Fig. 4E). Finally, when we knocked down Kap1 in human cord blood CD34+ cells, it resulted in decreasing their ability to undergo cytokine-induced ex vivo erythroid differentiation, which correlated with reduced Nix expression and elevated mitochondrial content (Fig. 4F), a phenotype that could be reproduced by hsa-miR-125a overexpression (Fig. 4G).

Fig. 4.

KAP1-regulated RNA interference controls human red cell mitophagy. HEL transduced with scramble or Kap1-specific shRNA-expressing lentiviral vectors and induced or not to differentiate were evaluated for Kap1 mRNA expression (A), and by benzidine (B) (n=3, counting 100 cells for each condition) or Mitotracker (C) (n=3) staining (*p <0.05). (D) hsa-miR-125a-5p (miR125a) and Nix expression measured respectively by NanoString nCounter direct RNA quantification and RNA sequencing in HEL cells transduced with empty or Kap1 knockdown vectors. (E) Nix expression in Ctrl (setting normalized value at 1) or hsa-miR-125a-5p-overexpressing HEL cells, measuring their mitochondrial content by Mitotracker staining (n=4, *p<0.05). (F) Decreased erythroid differentiation of Kap1 knockdown human cord blood CD34+ cells, assessed by CD235a surface expression at seven (D7) and eleven (D11) days. At D7, sorted CD235a+eGFP+ cells were analyzed by RT-QPCR for Nix and hsa-miR125a expression, and for mitochondrial content by Mitotracker staining (n=3, *p<0.05). (G) Percentage of CD235a-expressing cells 7 days after inducing the differentiation of CD34+ cells transduced with empty or unrelated-miRNA- (Ctrl) or hsa-miR-125a-5p-overexpressing lentiviral vectors (n=3, *p<0.05).

Altogether, these results unveil a multilayered transcription regulatory system, where protein- and RNA-based repressors are super-imposed in combinatorial fashion to govern the timely triggering of an essential step of erythropoiesis. Besides miR-351, several other microRNAs with predicted targets in the mitophagy pathway were upregulated in Kap1-deleted murine erythroblasts (Fig. 2). This apparent redundancy, or rather addition of parallel effects aimed at a same physiological process, is commonly observed with RNA interference. Our discovery that it can be further modulated by KRAB-ZFP-mediated repression, and that the latter can itself be multifactorial, adds a remarkable level of modularity to this type of regulation. In human erythroblasts, while KAP1 represses the Nix-targeting hsa-miR-125a-5p, downregulation of several other miRNAs, including hsa-miR-24, -221, -222 and -223, was previously found important for erythroid differentiation, which on the opposite requires the upregulation of hsa-miR-144/451 cluster (2, 3). Whether stage-specific KRAB-ZFPs are involved in controlling some of these other miRNAs remains to be determined. While it is likely that KAP1 influences erythropoiesis by more than just allowing mitophagy, it is interesting to note that Znf205 and Znf689, the respective human orthologues of murine Zfp13 and Zfp689, are expressed in HEL cells and induced upon erythroid differentiation of CD34+ cells (Suppl. Fig. 4). Therefore, polymorphism or mutations in any genetic component of the pathway unveiled here, whether Znf205, Znf689, the genomic binding sites of their products, hsa-miR-125a-5p and other KAP1-regulated miRNA genes, or the sequences targeted by these RNA regulators, could underlie red cell-related pathologies such as anemia, polycythemia or erythroleukemia.

Supplementary Materials

Materials and Methods

Figures S1-S4

Tables S1

One Sentence Summary.

A multilayered protein- and RNA-based transcriptional regulation system governs the timely removal of mitochondria during red cell differentiation