Self-renewal of a purified Tie2+ hematopoietic stem cell population relies on mitochondrial clearance

See allHide authors and affiliations

Science  02 Dec 2016:
Vol. 354, Issue 6316, pp. 1156-1160
DOI: 10.1126/science.aaf5530

How to maintain hematopoietic stem cells

Hematopoiesis provides the body with a continuous supply of blood cells (see the Perspective by Sommerkamp and Trumpp). Taya et al. report that amino acid content is important for hematopoietic stem cell (HSC) maintenance in vitro and in vivo. Dietary valine restriction seems to “empty” the mouse bone marrow niche. Ito et al. used single-cell approaches and cell transplantation to identify a subset of HSCs at the top of the HSC hierarchy. Self-renewal relied on the induction of mitophagy, a quality-control process linked to a cell's metabolic state. Both studies may be helpful in improving clinical bone marrow transplantation.

Science, this issue p. 1103, p. 1152; see also p. 1156


A single hematopoietic stem cell (HSC) is capable of reconstituting hematopoiesis and maintaining homeostasis by balancing self-renewal and cell differentiation. The mechanisms of HSC division balance, however, are not yet defined. Here we demonstrate, by characterizing at the single-cell level a purified and minimally heterogeneous murine Tie2+ HSC population, that these top hierarchical HSCs preferentially undergo symmetric divisions. The induction of mitophagy, a quality control process in mitochondria, plays an essential role in self-renewing expansion of Tie2+ HSCs. Activation of the PPAR (peroxisome proliferator–activated receptor)–fatty acid oxidation pathway promotes expansion of Tie2+ HSCs through enhanced Parkin recruitment in mitochondria. These metabolic pathways are conserved in human TIE2+ HSCs. Our data thus identify mitophagy as a key mechanism of HSC expansion and suggest potential methods of cell-fate manipulation through metabolic pathways.

Precise mechanisms enable efficient and regulated hematopoietic stem cell (HSC) renewal versus differentiation, and these processes are coordinated. HSCs must withstand various stresses to maintain hematopoiesis for the duration of their life cycles. The repair or clearance of damage is critical to precisely controlling their cell fates and maintaining stemness upon division, especially during HSC expansion through symmetric self-renewing division. HSC exhaustion can result from defective cellular metabolism and/or impaired autophagy (16), a lysosomal degradation pathway that breaks down damaged or unwanted proteins and/or organelles (7). The FOXO3A (forkhead box O3a)–driven proautophagy program, for example, protects HSCs from metabolic stress (5). In a similar process in mammary epithelial stemlike cells, older mitochondria are pushed into daughter cells fated to differentiate by asymmetric division (AD), thus maintaining high-quality stem cell homeostasis (8). Despite extensive study of HSC renewal and differentiation, however, much remains unclear regarding the mechanisms of division balance due to the heterogeneity of HSC-enriched fractions.

In searching for a potential HSC marker, we observed high Tie2 levels in the CD34 HSC-enriched compartment (fig. S1A). We therefore established a Tie2–green fluorescent protein (GFP) reporter line (Tie2-reporter mice or Tie2 Tg) to validate Tie2-expression as an HSC marker (fig. S1, B and C). Tie2-GFP was specifically enriched in the HSC fraction but was reduced during differentiation, in parallel with mRNA levels (fig. S1, D to G). Tie2-GFP+ and -GFP fractions in CD34CD150+CD48low/–CD135KSL cells were purified from Tie2 reporter mice (hereafter, Tie2-GFP+ or Tie2-GFP HSCs), and Tie2 levels were detected in Tie2-GFP+ HSCs but not in Tie2-GFP HSCs (fig. S1H). The cell cycle kinetics (9) revealed that cell cycling was slightly slower in Tie2-GFP+ versus Tie2-GFP HSCs (fig. S1I). A higher proportion of phenotypic Tie2-GFP+ HSCs than Tie2-GFP HSCs was closely associated with arteriolar structure (fig. S1J) and expressed endothelial and HSC markers (10, 11) (fig. S1, K to M).

In addition, the Tie2-GFP+ fraction displayed a highly significant (5.75-fold) increase in long-term culture–initiating cell (LTC-IC) frequency relative to the Tie2-GFP fraction (Fig. 1A). Single-cell transplantation showed that a high percentage of single Tie2-GFP+ cells (68.0%) exhibited reconstitution capacity without lineage bias and maintained high donor chimerism upon secondary transplantation. Donor contribution was observed in all recipients transplanted with three Tie2-GFP+ HSCs, whereas 40% showed reconstitution capacity with three Tie2-GFP HSCs (Fig. 1B and fig. S2, A to D), demonstrating that Tie2-GFP identifies an HSC fraction with minimal heterogeneity. To accurately supply the bone marrow with individual HSCs, we employed multiphoton microscopy guidance (local transplantation) (fig. S2E) to deliver single Tie2-GFP+ HSCs near an opening to the marrow of live mice, and we then tracked these cells during subsequent homing to locations some distance from the delivery site (Fig. 1C and movies S1 and S2). Recipients showed reconstitution from locally transplanted single (100%) and multiple Tie2-GFP+ HSCs (fig. S2F). A single Tie2-GFP+ HSC can exhibit reconstitution capacity in a recipient mouse for more than 6 months, once it is located in the bone marrow.

Fig. 1 Tie2-GFP marks a distinct HSC subset.

(A) LTC-IC frequencies of Tie2-GFP+ and Tie2-GFP HSCs were determined in a limiting-dilution assay (LDA). (B) Percentages of successful long-term reconstitution. (C) In vivo imaging of single Tie2-GFP+ HSCs immediately after local delivery in the calvarium. (Left) Top (upper) and side (lower) views of the delivery site. The arrow indicates the opening to the bone marrow. (Right) In vivo imaging of the same cell, 24 hours after local delivery (en face maximum intensity projection). (D) Single-cell gene expression analysis using the BioMark System array (Fluidigm) (see materials and methods).

We also tested, in vivo and in vitro, whether Tie2 acts as an HSC marker under stress conditions. Of the polyinosinic:polycytidylic acid (pIpC, inducing the interferon response)–treated single Tie2-GFP+ HSCs, 65.0% showed repopulation capacity at levels comparable to those seen in control mice (fig. S3, A to D). But whereas the Tie2-GFP+/Tie2-Ab+ fraction exhibited high reconstitution capacity, neither the Tie2-GFP/Tie2-Ab+ fraction nor any of the 12 single Tie2-GFP+ KSL cells showed high reconstitution. The majority of the re-sorted Tie2-GFP+ cells (70.6% after 3-day culture; 85.7% after 7-day culture) exhibited multilineage reconstitution, though Tie2-GFP cells did not. None of 90 Tie2-GFP HSCs provided donor-derived Tie2-GFP+ HSCs in recipient mice. These data suggest that Tie2 serves as a useful marker that can prospectively identify HSCs (figs. S2G and S3, E to L).

Single-cell gene expression assays showed that a substantial number of genes—including HSC factors and markers (e.g., Tie2, Hif1a, Mpl, and Foxo3a), as well as Pml (promyelocytic leukemia) and its downstream signaling pathways, which regulate fatty acid oxidation (FAO) via Cpt1a (carnitine palmitoyltransferase 1a), Ppard (peroxisome proliferator-activated receptor-delta), and Acox1 (acyl–coenzyme A oxidase1)—were markedly higher in the Tie2-GFP+ fraction than the Tie2-GFP fraction (Fig. 1D). Hierarchical clustering, HSC signature, and signatures for both PPAR signaling and fatty acid metabolism demonstrate that Tie2-GFP+ HSCs are molecularly farther away from both KSL cells and Tie2-GFP HSCs (fig. S4). The functions of the PPAR-FAO pathway in Tie2-GFP+ HSCs were then assessed. Impaired reconstitution ability of single Pml−/− Tie2-GFP+ HSCs could be rescued by GW501516, a PPARδ-specific agonist (fig. S5, A to C). Thus the PPAR-FAO pathway plays critical roles in directly regulating the self-renewal of the top hierarchical HSCs.

Mitochondria are central metabolic organelles responsible for the final steps of FAO, which cleaves two carbons every cycle to form acetyl–coenzyme A. Once the PPAR-FAO pathway is activated, fatty acids have been imported into the mitochondrial matrix (12). Our gene expression assays revealed higher expression of mitochondrial autophagy (mitophagy)–related genes, including Parkin (Park2), Pink1 (PTEN-induced putative kinase 1), Optineurin, Tom 7, Map1lc3a (Lc3), and p62/Sqstm1, in Tie2-GFP+ HSCs at steady-state conditions. Most of these genes were further up-regulated by activation of the PPAR-FAO pathway (Fig. 2A and fig. S5, D to F). To assess whether mitophagy-related genes are regulated by the PPAR-FAO pathway at the transcriptional level, Pink1-luciferase reporters were used. A PPARδ agonist transactivated Pink1 reporters, which were reduced by FAO inhibition or mutation of FOXO-response elements. Foxo3a was up-regulated in HSCs upon treatment with a PPARδ agonist, and this process was attenuated by the ablation of Ppard or Cpt2, which are FAO-related genes. These data suggest that the PPAR-FAO pathway transcriptionally regulates PINK1, in part through FOXO3a (fig. S6).

Fig. 2 The PPAR-FAO pathway activates mitochondrial autophagy in HSCs.

(A) Parkin and Pink1 levels in GW501516-treated (GW) Tie2-GFP+ HSCs. (B and C) PPARδ activators, GW501516 or L-165041 (L165), enhance colocalization of Tom20 with LAMP1. Treatment with carbonyl cyanide m-chlorophenylhydrazone (CCCP) was used as a control for mitophagy induction [(B), left]. Mitochondrial net flux was calculated by Tom20-accumulation in the presence of leupeptin (Leup) [(B), right]. Representative immunofluorescence (IF) images of colocalized Tom20 with LAMP1 [4′,6-diamidino-2-phenylindole (DAPI), blue; Tom20, red; LAMP1, green; Merge, yellow] in PPARδ-agonist–treated HSCs. White dashed boxes represent the magnified images shown at right. Scale bars, 2 μm (C). (D) Representative images of GW501516-treated HSCs immunostained to label mtDNA (left) and quantified for mitophagy [12 hours oligomycin and antimycin A] (right top) and net mtDNA flux (right bottom). In (A), (B), and (D), error bars represent mean ± SEM (at least 10 cells analyzed per each condition per each experiment).

We then sought to determine whether PPARδ-induced mitophagy correlates with a specific metabolic state or profile. Cellular FAO activity was specifically enhanced after PPARδ agonist treatment and was accompanied by Pink1, whereas glucose uptake or levels of pyruvate and tricarboxylic acid cycle metabolites were not altered (fig. S7, A to D). PPARδ agonists also enhanced mitophagy, with the increased colocalization of Tom20 (translocase of outer membrane 20) with LAMP1 (lysosomal-associated membrane protein 1). Flux analyses were then performed to quantify mitochondrial turnover in lysosomes (the organelles where mitophagy occurs) by evaluating the accumulation of Tom20 in the presence of leupeptin, an inhibitor of lysosomal proteolysis. Pharmacological PPAR-FAO activation enhanced the net mitochondrial flux toward lysosomal degradation. These data strongly suggest that the PPAR-FAO pathway activates mitophagy (Fig. 2, B and C).

Other findings supported the enhancement of mitophagy by a PPAR agonist, including increased colocalization of Parkin with pyruvate dehydrogenase (PDH), a matrix mitochondrial protein. Mitophagy enhancement was again confirmed in GW501516-treated HSCs by the increased colocalization of Parkin with both Tom20 and the autophagosome marker LC3 (figs. S7E and S8, A to D). To explore another mitophagy indicator, mitochondrial DNA (mtDNA) nucleoids were quantified after mitochondrial damage with oligomycin and antimycin A. GW501516-treated HSCs were nearly devoid of mtDNA and exhibited increased net mtDNA flux, whereas mtDNA was still partially retained in control HSCs (Fig. 2D). Thus, damaged mitochondria are cleared more quickly due to enhanced mitophagy activation in Tie2-GFP+ HSCs treated with a PPAR agonist.

Hematopoiesis-specific Ppard conditional knockout mice were then generated. Ppard-deleted HSCs exhibited neither enhanced Tom20 and Parkin colocalization nor increased mitochondrial net flux after PPARδ agonist treatment, although these agonists stimulated mitophagy in wild-type HSCs, and a mitochondrial uncoupler, carbonyl cyanide m-chlorophenylhydrazone (CCCP), induced mitophagy in both genotyped HSCs (fig. S8E). Conditional heterozygous deletion of Cpt2 partially but significantly attenuated the enhanced Parkin recruitment in mitochondria induced by a PPAR agonist (fig. S8F). Taken together, these direct genetic approaches reveal that PPARδ agonists stimulate mitophagy in a PPAR-FAO–dependent manner.

To test whether mitophagy is critical for Tie2-GFP+ HSC expansion, Parkin and Pink1 were first silenced by RNA interference–mediated knockdown. siPark2 decreased net mitochondrial flux (a hallmark of mitophagy) in GW501516-treated HSCs, as indicated by decreased colocalization of Tom20 and PDH with LAMP1, whereas cell death, proliferation rate, FAO activity, and Cdk6 levels were not altered (Fig. 3A and figs. S7A and S9). Parkin-knockdown in GW501516-treated HSCs led to accumulating signals of Tom20 and PDH, as well as reduced colocalization of LAMP1 with LC3 (fig. S10, A to C). mtDNA was retained in Parkin-silenced HSCs, and these cells were refractory to PPAR agonist treatment (Fig. 3B and fig. S10D). These data suggest that Parkin has an essential role in PPAR-FAO–induced mitophagy.

Fig. 3 Parkin is important for PPAR-FAO–induced mitophagy in HSCs.

(A) Relative Tom20 colocalization (left) and mitochondria net flux (right) in GW501516-treated HSCs in the presence of siPark2. (B) Quantification of mitophagy based on the remaining mtDNA staining 12 hours after treatment with oligomycin and antimycin A. (C) Pink1-silencing reduces PPARδ-enhanced mitophagy. Relative Tom20 colocalization (top left), net mitochondria flux (top right), and representative IF images depicting LAMP1 and Tom20 colocalization in GW501516-treated HSCs in the presence of siPink1 (bottom). (D) GW501516 enhances colocalization of TOM20 with PARKIN (left) and increases mitochondrial net flux (right). Scale bars in (C) and (D), 2 μm.

Because Pink1 is known to be essential for mitophagy and Pink1 levels were increased by PPAR-FAO activation, we next assessed the effect of Pink1 knockdown on the process. Pink1 silencing reduced Tom20 colocalization with both LAMP1 and Parkin and decreased net flux of mitochondria in GW501516-treated HSCs, accompanied by accumulating Tom20 signals. mtDNA was also retained in Pink1-silenced HSCs (Fig. 3, B and C, and fig. S10, D to G). Consequently, Parkin and Pink1 play important roles in PPAR-FAO–induced mitophagy in HSCs.

We then explored potential clinical applications of pharmacological FAO modulation in human hematopoiesis. TIE2 expression is enriched in the HSC fraction, and TIE2-positivity enhances LTC-IC frequency and in vivo repopulation capacity in human bone marrow (fig. S11, A to G). GW501516 enhances LTC-IC frequency with mitophagy activation, whereas low doses of an FAO inhibitor reduce its capacity. Fatty acid metabolism can therefore control self-renewal capacity and the cell fates of human HSCs (Fig. 3D and fig. S11, H to J).

Finally, we explored the effects of FAO-induced mitophagy on Tie2-GFP+ HSC expansion. In vivo paired daughter cell (PDC) assays can determine the division patterns of HSCs retrospectively, by assessing the reconstitution capacity of each daughter cell (3, 13). In vivo PDC assays revealed that Tie2-GFP+ HSCs preferentially undergo symmetric division (SD) (in 20 out of 27 divisions, or 74.1%, both daughter cells showed a reconstitution capacity) rather than AD or symmetric commitment [(SC), both daughter cells exhibited no reconstitution capacity], whereas Tie2-GFP HSCs undergo either AD or SC (Fig. 4, A to C, and fig. S12). To further test the restoration capacity of hematopoiesis by single Tie2-GFP+ HSCs, we performed transplantation assays under high-stress conditions, as are found in single-cell transplantation without supporting cells. The survival ratio after single Tie2-GFP+ HSC transplantation was 85.7%, whereas no recipients of single Tie2-GFP HSCs survived (Fig. 4D and fig. S13, A and B). Functional HSCs are present in the Tie2-GFP fraction, but they fail to adequately respond to the demand for hematopoietic recovery under high-stress conditions without the enhanced symmetric expansion provided by Tie2-GFP+ HSCs.

Fig. 4 Self-renewing expansion of Tie2+ HSC through mitophagy.

(A) Representative pictures of Tie2-GFP positivity, with bright-field images (BF) collected during symmetric division. (B) Representative repopulation kinetics of SD during the first and second bone marrow transplantation (BMT). PB-MNCs, peripheral blood mononuclear cells. (C) Division patterns of Tie2-GFP+ HSCs (left) and Tie2-GFP HSCs (right), in terms of reconstitution capacity. SD, symmetric division; AD, asymmetric division; SC, symmetric commitment. (D) Single Tie2-GFP+ HSC transplantation without supporting cells. Survival of recipient mice was examined by plotting Kaplan-Meier survival curves (also see fig. S13B). (E) PPARδ activation induces in vivo expansion of Tie2-GFP+ HSCs. The number of transplanted Tie2-GFP+ HSCs (donor cells) is denoted by the dashed line marked “Original” (also see fig. S13H). (F) Frequency of successful hematopoietic reconstitution of shParkin-silenced Tie2-GFP+ HSCs (also see fig. S13J).

As expected, PPAR-FAO activation increased the number of Tie2-GFP+ cells in vitro with an increased rate of SD (fig. S13, C to G). Tie2-GFP+ HSC numbers were increased in comparison with the original number of transplanted donor cells in recipients and were further increased by PPARδ agonist treatment (about a fourfold increase). Furthermore, PPARδ agonist treatment expanded Tie2-GFP+ HSCs under physiological stress, as induced by pIpC (Fig. 4E and figs. S3B and S13H).

However, silencing Park2 or Pink1 not only abrogated the expansion of Tie2-GFP+ cells in vitro but also inhibited their maintenance (fig. S13I). Most single Parkin-silenced Tie2-GFP+ cells failed to show reconstitution capacity (92.3%), whereas the majority of control Tie2-GFP+ cells (66.7%) exhibited this capacity (Fig. 4F and fig. S13J). Pink1−/− HSCs failed to adequately respond to the demand for hematopoietic recovery under high-stress conditions, even though a PPARδ agonist was administered (fig. S13K). In conclusion, Tie2-GFP+ HSCs can maintain stem cell potential during cell cycling due to high mitochondrial clearance through mitophagy and up-regulation of Parkin and Pink1 (fig. S14). Our data from this highly purified HSC fraction identifies mitochondrial clearance by induction of mitophagosome formation as a key mechanism in maintaining stemness.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S14

References (1459)

Movies S1 and S2

References and Notes

Acknowledgments: We thank all members of the Ito lab and Einstein Stem Cell Institute for their comments on HSC self-renewal and M. Wolfgang, A. Carracedo, H. You, and the Einstein Flow Cytometry and Analytical Imaging core facilities (grant P30CA013330) for help and materials. This work was supported by NIH (grants R01DK98263 and R01DK100689 to Ke.I.) and NYSTEM (New York State Stem Cell Science single-cell-core, grant C029154 to Ke.I.), Harvard Stem Cell Institute (to C.P.L.), NIH and Ellison Medical Foundation (to R.S.), NIH and Leukemia Lymphoma Society (to P.S.F.), and Japan Society for the Promotion of Science (to T.S.). We declare no competing financial interests.
View Abstract

Navigate This Article