STAT5-regulated microRNA-193b controls haematopoietic stem and progenitor cell expansion by modulating cytokine receptor signalling

Haematopoietic stem cells (HSCs) require the right composition of microRNAs (miR) for proper life-long balanced blood regeneration. Here we show a regulatory circuit that prevents excessive HSC self-renewal by upregulation of miR-193b upon self-renewal promoting thrombopoietin (TPO)-MPL-STAT5 signalling. In turn, miR-193b restricts cytokine signalling, by targeting the receptor tyrosine kinase c-KIT. We generated a miR-193b knockout mouse model to unravel the physiological function of miR-193b in haematopoiesis. MiR-193b−/− mice show a selective gradual enrichment of functional HSCs, which are fully competent in multilineage blood reconstitution upon transplantation. The absence of miR-193b causes an accelerated expansion of HSCs, without altering cell cycle or survival, but by decelerating differentiation. Conversely, ectopic miR-193b expression restricts long-term repopulating HSC expansion and blood reconstitution. MiR-193b-deficient haematopoietic stem and progenitor cells exhibit increased basal and cytokine-induced STAT5 and AKT signalling. This STAT5-induced microRNA provides a negative feedback for excessive signalling to restrict uncontrolled HSC expansion.

M icroRNAs (miRs) are small non-coding RNAs, which regulate gene expression by either degrading mRNAs or by inhibiting protein translation 1 . They simultaneously target various mRNAs and thereby fine-tune entire gene expression networks 1 . The necessity of miRs for normal longterm repopulating haematopoietic stem cell (LT-HSC) function became apparent by the haematopoietic-specific deletion of Dicer, an essential nuclease for the generation of mature miRs, leading to the loss of their life-long self-renewal ability 2 . MiR-125a and miR-125b support HSC maintenance, function and renewal 2,3 , and their upregulation may even drive leukaemogenesis 4 . Conversely, miR-126 impedes cell-cycle progression in LT-HSCs, and downregulation of miR-126 causes LT-HSC expansion 5 . Various miRs act at different haematopoietic developmental stages and lineages 6 , and altered miR expression may play a pivotal role in leukaemia onset and progression 1,7,8 .
We aimed to identify miRs induced by the self-renewalpromoting cytokine thrombopoietin (TPO)-mediated signalling in LT-HSCs. In particular, we assessed miRs downstream of TPOinduced activation of signal transducer and activator of transcription (STAT) 5A/B 9 . The STAT5A/B signalling pathway is activated by cytokine receptors, such as myeloproliferative leukemia (MPL), c-KIT and the receptors of interleukin 3, granulocyte-macrophage colony-stimulating factor and erythropoietin 10 . LT-HSCs require STAT5A/B activity for self-renewal and maintenance 11 . Constitutively active STAT5A/B promotes marked LT-HSC expansion and the subsequent development of a myeloproliferative syndrome 12 . Therefore, STAT5A/B signalling needs fine-tuning for normal LT-HSC self-renewal.
In this study, we identify miR-193b as a regulatory feedback molecule restricting excessive HSC self-renewal upon the activation of the self-renewal-promoting TPO-MPL-STAT5 signalling. To execute this function, miR-193b restricts cytokine signalling by targeting the tyrosine kinase c-KIT. In turn, miR-193b-deficient haematopoietic stem and progenitor cells (HSPCs) from miR-193b knockout mice exhibited increased basal and cytokine-induced STAT5 and AKT signalling, thereby promoting the consecutive expansion of HSCs. This STAT5-regulated miR balances cytokine signalling via the STAT5 and AKT pathways, providing a negative feedback for excessive signalling to restrict uncontrolled HSC expansion.

Results
In vivo expansion of LT-HSCs in the absence of miR-193b. To identify miRs that are extrinsically regulated by the selfrenewal-promoting signalling axis comprising TPO, its receptor MPL and the transcription factors STAT5A/B, we compared miR expression patterns in LT-HSCs of STAT5A/B-deficient and wild-type (WT) control mice 13 that were stimulated with TPO or kept unstimulated, by quantitative PCR (qPCR; Fig. 1a). The differential miR pattern revealed five miRs that were 42-fold upregulated by TPO only in the presence of STAT5A/B: miR-193b, miR-132, miR-125a, miR-331-5p and miR-669a ( Fig. 1a and Supplementary Data 1). We focused on the function of the intergenic miR-193b in haematopoiesis, because miR-193b is selectively expressed in LT-HSCs and to a lesser extend in multipotent progenitors (MPPs), but not in lineagecommitted progenitors and mature blood cells, as shown by us (Supplementary Fig. 1a) and others 3,6 . Furthermore, haematopoietic stress induced by the cytokine storm 10 days after 5-fluorouracil (5-FU) treatment upregulated miR-193b expression in LT-HSCs (about 2.5-fold in comparison to steady-state). Although the induction of miR-193b expression was even more pronounced in lineage-committed progenitors and mature blood cells than in LT-HSCs caused by 5-FU treatment, the expression level in these committed cells was still 1,000 times lower than in LT-HSCs ( Supplementary Fig. 1b). Recently, we demonstrated that STAT5A/B binds to the miR-193b promoter in the murine mammary gland 14 . Here we could show that STAT5A/B is required for the cytokine-induced miR-193b transcription in LT-HSCs.
To unravel the function of miR-193b in haematopoiesis, we generated miR-193b knock-out mice, which were viable without visible abnormalities 15 . First we investigated the steady-state haematopoiesis of 2-to 3-month-old miR-193b À / À mice. Compared with WT mice, no significant differences (according to t-tests) were observed in the mature blood cell lineages in peripheral blood, bone marrow (BM) or spleen of miR-193b À / À mice ( Supplementary Fig. 2a-c). The percentage and number of defined BM progenitor cells were also unchanged ( Fig. 1b and Supplementary Fig. 2d,e). However, miR-193b À / À mice over 6 months of age displayed an unexpected increase in LT-HSCs in the LSK (Lineage À Sca1 þ c-KIT þ ) compartment (Fig. 1b), whereas total LSK cell numbers were not altered ( Supplementary  Fig. 2e). The accumulation of LT-HSCs increased with age, as 1-year-old mice showed a 1:1 ratio of LT-HSCs and MPPs (Fig. 1b). Yet, we only determined the LT-HSC frequency by their well-established marker phenotype, but we needed to confirm their true identity by their long-term blood reconstitution ability. To corroborate that miR-193b À / À LT-HSCs were fully functional, we performed a competitive transplantation of LT-HSCs from 1-year-old miR-193b-deficient or WT mice into recipients and then monitored donor blood reconstitution (Fig. 1c). The miR-193b-deficient LT-HSCs reconstituted equally well as WT LT-HSCs (Fig. 1d) and exhibited normal production of T, B and myeloid cells ( Supplementary Fig. 2g). Strikingly, when we analysed the distribution of LT-HSC and progenitor cells in primary recipient BM, we determined a more than twofold increase in phenotypic LT-HSC numbers in the absence of miR-193b in comparison to the WT controls (Fig. 1e). Although donor cell engraftment in the BM was only slightly enhanced in the absence of miR-193b ( Supplementary Fig. 2f), overall BM donor cellularity was markedly increased, thereby suggesting that miR-193b À / À LT-HSCs self-renew extensively after transplantation stress to repopulate the recipient ( Supplementary Fig. 2h). We further challenged the self-renewal ability of miR-193b-deficient LT-HSCs by transplanting unfractionated BM cells from primary recipients into secondary recipient mice. Again, both the WT and knockout group reconstituted the secondary recipients almost equally well, which clearly indicated that miR-193b-deficient LT-HSCs were fully functional (Fig. 1f). When we gated for LT-HSCs in the BM of secondary recipients, we again measured an increased proportion and number of LT-HSCs in those recipients that received miR-193b À / À cells (Fig. 1g). Assessing the donorderived HSPC distribution in primary and secondary recipient BM, we observed a consistent increase in miR-193b À / À LT-HSCs and subsequent increase in LSK cell numbers (Fig. 1h,i). Although no difference was found in the primary recipients of either miR-193b À / À or miR-193b þ / þ LT-HSCs (Fig. 1h), secondary recipient BM accumulated miR-193b À / À LT-HSCs and LSK cells (Fig. 1i). This proves that the self-renewal of LT-HSCs leading to an enlarged compartment of competent LT-HSCs is intrinsically promoted in the absence of miR-193b. Of note, the recipients that received the miR-193b À / À donor cells displayed a slightly lower donor cell chimerism in the BM ( Supplementary Fig. 2f,h), indicating that although they received more LT-HSCs from primary recipients and again showed an enhanced self-renewal and expansion of LT-HSCs, the ability of these LT-HSCs to produce the same output on mature cells seemed altered. Whether this is due to a delay in differentiation of LT-HSCs or an altered clonal fitness remains to be clarified.
More miR-193b-deficient LT-HSCs are in active cell cycle. Most LT-HSCs are quiescent (G0 phase) in homeostasis. The expansion of the LT-HSC population may indicate that an increased proportion of LT-HSCs are actively cycling in the absence of miR-193b. Therefore, we examined the cell-cycle phases in 2-to 3-month-old mice (Fig. 2a). Indeed, there were a significant higher proportion of LT-HSCs and MPPs in cell-cycle (Ki67 þ ) in miR-193b À / À mice (Fig. 2a). The continuous observation of individual LT-HSCs by time-lapse microscopy-based cell tracking 16,17 allowed us to determine the time of entry into the first division once cells were exposed to in vitro culture. The miR-193b À / À LT-HSCs displayed an earlier time point of division compared with WT controls under minimal (stem cell factor (SCF) only) and self-renewal-promoting (TPO and SCF)   ARTICLE cytokine conditions (Fig. 2b). Next, we functionally confirmed these findings by repetitively treating mice with 5-FU, which eliminates proliferating cells. Quiescent LT-HSCs only become susceptible to 5-FU during haematopoietic stress, driving them into cycle. We injected 5-FU into miR-193b À / À and miR-193b þ / þ mice once a week, until the mice suffered from fatal haematopoietic failure (Fig. 2c). Although most of the WT mice died after the third round of 5-FU injections, all miR-193bdeficient mice died already after two subsequent injections, most likely because the hyperproliferative LT-HSCs were quickly extinguished (Fig. 2c). These results show at the phenotypic, molecular and functional level that the absence of miR-193b leads to an intrinsically controlled expansion of fully functional LT-HSCs, as they exhibit reduced quiescence and increased self-renewal.

MiR-193b controls HSPC expansion by altering differentiation.
We then assessed the immediate consequences of the absence of miR-193b on the function and fate of LT-HSCs. First, we measured the expansion of LT-HSCs in liquid culture (Fig. 3a) and found that LT-HSCs from miR-193b À / À mice gave rise to a twofold increase in cell numbers over a 7-day-period compared with WT LT-HSCs (Fig. 3a). Conversely, lentiviral expression of miR-193b in LT-HSCs led to a dramatic reduction of miR-193bexpressing progeny over time, whereas the percentages of control vector transduced cells or of cells lentivirally expressing the unrelated miRs-132/212 remained largely constant (Fig. 3b,c, and Supplementary Figs 3a and 4a). Of note, ectopic expression of miR-193b in GMPs did not influence their expansion, and the percentage of transduced cells did not change over time ( Supplementary Fig. 4b), indicating a distinct function of miR-193b in LT-HSCs. The increased expansion of LT-HSCs and their progeny leading to elevated levels of mature blood cells may be explained by increased proliferation, reduced apoptosis or delayed differentiation. To dissect these various cell fates at a single-cell resolution, we monitored the individual LT-HSCs and their progeny of miR-193b-deficient and WT mice using videomicroscopy-based cell tracking ( Fig. 3d and Supplementary  Fig. 5). Interestingly, the absence of miR-193b had no influence on cell-cycle duration in dividing LT-HSCs and their progeny for many generations (Fig. 3e). Moreover, there was no decrease of cell death events in the absence of miR-193b (Fig. 3f). To verify that the absence of miR-193b does not influence the cell-cycle duration of HSPCs, we determined HSPC proliferation in vivo by pulsing miR-193b À / À and miR-193b þ / þ mice with 5-bromodeoxyuridine (BrdU) for 4 h. Accordingly, there was no difference in the cell-cycle distribution in LT-HSCs, MPPs, GMPs or MEPs, which further confirmed that the absence of miR-193b does not shorten the cell cycle of HSPCs (Fig. 3g). These results show that the absence of miR-193b increases the number of LT-HSCs in active cell cycle (Fig. 2), but does not alter the cell-cycle progression of proliferating LT-HSCs or their progeny during steady-state haematopoiesis (Fig. 3). We asked whether miR-193b À / À LT-HSCs show a delayed differentiation that causes an accumulation of highly proliferative progenitors. Therefore, we determined distinct differentiation stages of LT-HSCs that were cultured for several days via fluorescence-activated cell sorting (FACS; Fig. 3h). After 5 days of culture, 25% of WT cells expressed markers similar to GMP-like cells (CD117 þ , CD16/32 þ ). Meanwhile, most miR-193b-deficient cells remained more immature (CD117 þ , CD16/32 À ; Fig. 3i,j). Overall, the increased expansion of cells from miR-193b À / À LT-HSCs was due to a delay in differentiation and not a change in proliferation or survival.  Increased cytokine signalling in miR-193b À / À LT-HSCs. Next, we aimed to assess the molecular mechanism by which miR-193b restrains HSPC expansion and LT-HSC self-renewal. We performed RNA sequencing of LSKs from miR-193b À / À and miR-193b þ / þ mice to elucidate differences in the gene expression profile. Using a regulation threshold of 1.5-fold (Po0.05), we identified 41 upregulated and 117 downregulated genes in the absence of miR-193b (Supplementary Fig. 6 and Supplementary Data 2). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of these regulated genes using DAVID 18 and Ingenuity Pathway Analysis suggested the involvement of altered cell signalling ( Fig. 4a and Supplementary Data 3 and 4). This result guided us to investigate major signalling pathways that are activated by cytokines, such as Janus kinase/STAT, PI3K/AKT (phosphatidylinositol 3-kinase/thymoma viral proto-oncogene) and MAPK/ERK (mitogen-activated protein kinase). Phosphoflow cytometry of phosphorylated STAT3, STAT5, AKT and ERK revealed an increase in the basal signalling levels of STAT5 and AKT in miR-193b À / À BM cells (Fig. 4b). The hyperactivation of signal transduction, particularly of STAT5 (ref. 12), may lead to the observed expansion of LT-HSCs in miR-193b-deficient mice. Therefore, we assessed tonic and cytokineactivated signalling in LT-HSCs using phosphoflow cytometry. Although tonic signalling levels after cytokine starvation were similar (Fig. 4c), STAT5 and AKT phosphorylation were 2-and 1.5-fold stronger after cytokine stimulation, respectively, in miR-193b-deficient LT-HSCs compared with WT counterparts (Fig. 4d). Next, we determined the kinetics of STAT5 and AKT signalling in temporal relation to miR-193b expression in LT-HSCs upon cytokine stimulation, and investigated changes in STAT5 and AKT signalling in the absence of miR-193b (Fig. 4e). After a rapid induction already at 5 min of stimulation, the levels of pSTAT5 and pAKT further increased to a maximum at 20 min in LT-HSCs from miR-193b þ / þ and miR-193b À / À mice. pSTAT5 and pAKT levels remained constant for 60 min before declining again at 120 min. Intriguingly, miR-193b À / À LT-HSCs showed an overshooting activation of pSTAT5 at 20 min, that did not decline as much as the pSTAT5 levels in miR-193b þ / þ LT-HSCs at 120 min (Fig. 4e). MiR-193b À / À LT-HSCs showed a higher induction of pAKT at 20 min, which remained above the pAKT levels of miR-193b þ / þ LT-HSCs until 120 min (Fig. 4e). The expression of miR-193b was instantly induced by cytokine stimulation and showed the strongest expression at 60 min after stimulation before it already declined at 120 min. Importantly, already 20 min after stimulation there was a twofold increase in miR-193b expression in comparison to unstimulated LT-HSCs (Fig. 4e).
These results underline that the cytokine-induced activation of STAT5 leads to the rapid upregulation of the miR-193b as a STAT5-dependent miR, and the time shift between maximum pSTAT5 (at 20 min) and maximum miR-193b expression (at 60 min) would be expected from a transcribed target gene of STAT5. However, in the absence of miR-193b there is an MiR-193b targets c-KIT and thereby modulates signalling. The increased signalling output may be caused by kinase hyperactivation. Therefore, we applied PamGene array technology to quantitatively compare the activity of hundreds of tyrosine and serine/threonine kinases in the BM cells of miR-193b À / À and miR-193b þ / þ mice in an unbiased way. Overall, we did not observe global changes in tyrosine or serine/threonine phosphorylation patterns in miR-193b-deficient cells. However, we clearly determined specific changes in a small group of peptides ( Fig. 5a and Supplementary Data 5). One peptide found to be hyperphosphorylated was a target of c-KIT, the receptor of the cytokine SCF and an essential receptor tyrosine kinase in haematopoiesis. Altered c-KIT signalling has detrimental effects on LT-HSC biology [19][20][21] . Another hyperphosphorylated peptide was a target of the serine/threonine kinase mTOR (mechanistic target of rapamycin), which acts downstream of PI3K/AKT signalling. Interestingly, PI3K/AKT signalling was enhanced in miR-193b À / À HSPCs at basal levels and after cytokine stimulation.
As the c-Kit mRNA contains a predicted conserved miR-193b target sequence (Fig. 5b), and c-KIT activity was elevated in the absence of miR-193b, we predicted that miR-193b can modulate c-KIT expression in HSPCs. Indeed, there was a 30% reduction of c-Kit mRNA expression in LT-HSCs ectopically expressing miR-193b (Fig. 5c). More importantly, we determined a 40% reduction of c-KIT surface expression on HSPCs that expressed miR-193b in comparison to control vector transduced cells, measured by FACS (Fig. 5d). Because LT-HSCs with diminished c-KIT function are severely impaired in recipient repopulation 19,20 , we assessed the blood reconstitution of miR-193b-expressing LT-HSCs after transplantation. As hypothesized, no donor cell reconstitution was detected in mice transplanted with LT-HSCs ectopically expressing miR-193b (Fig. 5e).
Reduced cell expansion and lack of blood reconstitution following transplantation with LT-HSCs overexpressing miR-193b resembles the phenotype of LT-HSCs harbouring dysfunctional c-KIT 19,20 . Next, we assessed the c-KIT protein expression in miR-193b À / À BM cells and showed via FACS that miR-193b-deficient cells expressed 30% more c-KIT protein in comparison to their respective miR-193b þ / þ counterparts (Fig. 6a), indicating that the absence of miR-193b leads to higher c-KIT protein levels.
To consolidate that the observed phenotype in miR-193bexpressing LT-HSCs was at least partly due to the reduced c-KIT expression, we lentivirally expressed both c-Kit lacking the miR-193b target site (Supplementary Fig. 3b) and miR-193b in LT-HSCs (Fig. 6b). We then assessed the expansion of doubletransduced cells in culture. Although cells expressing miR-193b and a control vector nearly disappeared over the course 7 days, ARTICLE miR-193b-expressing cells co-transduced with c-KIT expanded in culture, thereby rescuing the miR-193b-mediated effect (Fig. 6c).
These results show that STAT5-activated miR-193b regulates c-KIT expression, probably among other important factors, and thereby subsequently influences signalling networks to control the fate of LT-HSCs.

Discussion
Two prominent signalling pathways guiding LT-HSC self-renewal and proliferation, STAT5A/B and PI3K/AKT, are hyperactivated in the absence of miR-193b leading to a more active LT-HSC population that expands over time ( Supplementary Fig. 7). Constitutive active STAT5 signalling induces LT-HSC expansion and finally results in a myeloproliferative disease in mice 12 . Furthermore, many leukaemogenic alterations (for example, mutated c-Kit, Flt3-ITD and BCR-ABL) require chronically high STAT5 activation for disease development [22][23][24] . PI3K/AKT signalling promotes cell proliferation and survival 25 , and hyperactivation of this pathway results in LT-HSC expansion and exhaustion, and in the development of leukaemia 26 . Furthermore, SCF is essential for the survival and proliferation of all HSPCs, and it transmits its signal after binding to c-KIT leading to STAT5, PI3K/AKT and ERK signalling 21 . Activating c-KIT mutations, which are common in acute leukaemias, cause SCF-independent constitutive, dysregulated signalling and eventual uncontrolled expansion of leukaemic blasts 27 . Conversely, in mouse models, hypomorphic c-KIT mutations compromise LT-HSCs, which can easily be outcompeted by WT LT-HSCs in transplantations without harsh conditioning 28 . STAT5 and AKT signalling need to be in tight balance 26 , and are controlled by negative regulators such as suppressor of cytokine signaling (SOCS) and phosphatase and tensin homolog (PTEN) 10,[29][30][31][32] . PTEN-deficient LT-HSCs display increased PI3K/AKT signalling and a hyperproliferative phenotype with long-term exhaustion 29,30,32 . In this study, we found that miR-193b is an important negative regulator of basal and cytokine-stimulated signalling and hyperactivation in LT-HSCs. Therefore, as miR-193b expression is triggered by STAT5 signalling, this axis represents a classical negative feedback mechanism. Strikingly, although miR-193b À / À LT-HSCs were more actively cycling and expanding, they did not exhaust and remained fully functional. This aspect distinguishes miR-193b function from other haematopoietic negative regulators, such as PTEN, thereby suggesting that miR-193b is able to fine-tune HSC numbers. The tolerated expansion of LT-HSCs over time might also be supported by our findings that the absence of miR-193b does not change the cell-cycle length in proliferating HSPCs, but rather regulates the decision of active cell-cycle entry versus quiescence.
It is not surprising that miR-193b is downregulated in leukaemia and other cancer entities 7,33-36 . Although we did not observed increased cancer incidence in 1-year-old miR-193b À / À mice, it would be intriguing to test whether the absence of miR-193b cooperates with known oncogenes. Especially, the initiation of pre-cancerous (stem) cells at an early disease stage might be supported by enhanced self-renewal in the absence of the tumour-suppressing miR-193b, as STAT5 activation plays a key role in establishing pre-cancerous clonal dominance in stem cells 37,38 . We recently reported the expansion of mammary epithelial stem cells in the absence of miR-193b 14 , which suggests a general miR-193b function in restricting adult stem cell proliferation. Further evaluation is warranted to determine whether miR-193b downregulation is an early event in tumourigenesis.
FACS analysis and sorting of HSPCs. BM cells isolated from femurs, tibias, coxae and sternum were either crushed or flushed (excluding sternum) followed by a depletion of lineage marker-positive cells (EasySep Biotin Selection Kit, Stemcell Technologies) using the following biotin-labelled antibodies (CD3, CD45R, CD19, CD11b, CD41, Ter119 and Gr1). Alternatively, BM mononuclear cells were enriched using Ficoll and then stained using the same lineage markers. Streptavidin was used to stain for the remaining lineage marker-positive cells. The cells were stained with fluorochrome-labelled antibodies and sorted using a FACS Aria I and III (BD) or analysed using a LSR Fortessa (BD). The following surface marker combinations were used to identify various HSPC populations: KL (Lin À CD117 þ Sca1 À ), LSK (Lin À CD117 þ Sca1 þ ), (LT)-HSCs (Lin À CD117 þ Sca1 þ CD150 þ CD34 low/ À CD48 À ), MPPs (Lin À CD117 þ Sca1 þ CD150 À CD34 þ ), GMPs (Lin À CD117 þ Sca1 À CD150 À CD34 þ ) and MEPs (Lin À CD117 þ Sca1 À CD150 þ ). The antibodies used are listed in Supplementary Table 1. The gating was performed as previously described 16 . Viable sorted cells were counted with trypan blue exclusion. FACS data analysis was performed with DIVA 7.0 software (BD) or FlowJo software (FlowJo). using a miRNeasy mini kit (Qiagen). cDNA was synthesized from total RNA using a miR gene-specific RT-primer pool according to the MicroRNA Megaplex Assay protocol (Megaplex Pools-Applied Biosystems). The preamplification reaction was performed according to the manufacture's protocol. The TaqMan Array MicroRNA Rodent Cards were analysed using the 384-well TaqMan Low Density Array default thermal cycling conditions (ABI PRISM 7900HT). The CT (threshold cycle) values were determined using default threshold settings. Three different housekeeping genes were used for data normalization. To identify TPO-induced miRs that are dependent on STAT5, we measured the fold expression change of miRs upregulated following TPO stimulation in the presence (STAT5 fl/fl ) and absence (STAT5 fl/fl x Mx1:Cre) of STAT5. All statistical analyses were performed using a two-tailed unpaired t-test. N ¼ 2 sets of 16-20 mice per experiment and group. The miR expression array data can be accessed from Figshare at: http:// dx.doi.org/10.6084/m9.figshare.1554878.
Quantitative RT-PCR of c-Kit expression. To quantify the ability of miR-193b to target c-Kit mRNA expression, FACS-sorted LT-HSCs and MPPs were lentivirally transduced for ectopic miR-193b expression or control at multiplicity of infection (MOI) ¼ 100 and cultured for 4 days. Fluorescent cells were harvested and directly lysed and reverse transcribed using the Cells-to-Ct-Kit (Life Technologies) according to the manufacturer's protocol. The pre-amplification for 14 cycles was performed according to the manual of TaqMan-PreAmp-Master-Mix (Life Technologies) using TaqMan assays for c-Kit (ID: MM00445212) and Gapdh (ID: MM099999915_g1), before qPCR was performed according to the manufacturer's protocol using the respective TaqMan assays and the Gene Expression Master Mix (Life Technologies). The data were normalized to Gapdh in each sample and displayed as Dct.
Cells were lysed according to the manufacturer's protocol of the Cells-to-Ct-Kit (Life Technologies) the reverse transcription was performed with the TaqMan MicroRNA RT Kit (Life Technologies) according to the manufacturer's instructions using the mmu-miR-193b-3p (ID002467) and snoRNA202 (ID001232) assays. LT-HSCs and MPPs were pre-amplified for 12 cycles using the TaqMan PreAmp Master Mix (Life Technologies) according to the manufacturer's protocol. Reverse transcribed samples from KL and Lin þ cells were directly used for qPCR according to the manufacturer's instructions of the Cells-to-Ct-Kit. The TaqMan Universal Master Mix II with UNG (Life Technologies) was used for the qPCR performed on a StepOne instrument (Applied Biosystems). SnoRNA202 was used for normalization.
Time-lapse imaging and cell tracking. Microscopy and tracking of LT-HSCs and their progeny was performed using a self-written computer programme (TTT) as previously described 16,17 until the fate of all progeny in the third cell generation was determined. The generation time of an individual cell was defined as the time span from cytokinesis of its mother cell division to its own division. Dead cells were easily depicted by their shrunken, non-refracting and immobile appearance. Cell tracking was carried out by scientists; the current analysis did not rely on data generated by an unsupervised computer algorithm for automated tracking.
Peripheral blood cell counts. Peripheral blood cell counts were determined from tail vein blood using a ScilVet animal blood cell counter (Scil Animal Care Company).
In vitro cell proliferation assay. FACS-sorted LT-HSCs (100 cells per well) were cultured for 7 days in SFEM supplemented with 100 ng ml À 1 murine SCF and TPO, 20 ng ml À 1 murine IL3 and IL6 (Peprotech) and 5 U ml À 1 human EPO (PromoKine). Viable cells were assessed using the ViaLight Plus Cell Proliferation and Cytotoxicity BioAssay Kit (Lonza) at days 3, 5 and 7 according to the manufacturer's instructions. Luminescence was measured using a Mithras LB940 luminometer (Berthold Technologies). For ectopic miR-193b expression experiments, FACS-sorted LT-HSCs (100 cells per well in 96-well format) were lentivirally transduced (MOI ¼ 100) and cultured for up to 9 days in SFEM supplemented with 100 ng ml À 1 SCF and TPO. Viable cells were counted with Trypan blue exclusion. The percentage of transduced cells (eBFP þ ) was analysed via FACS (BD CantoII). Ectopic expression of the unrelated miRs-132/212 in LT-HSCs using the same lentiviral expression strategy served as an additional control. FACS-sorted GMPs (300 cells per well in 96-well format) were lentivirally transduced (MOI ¼ 20) and cultured for up to 7 days in SFEM supplemented with 100 ng ml À 1 SCF and 20 ng ml À 1 murine IL3 and IL6 (Peprotech). The percentage of transduced cells (eBFP þ ) was analysed via FACS.
Rescue experiment with ectopic c-KIT. Freshly sorted LT-HSCs (100 cells per well) were double transduced with four different combinations of either a vector coding for murine c-KIT (without 3 0 -untranslated region and miR-193b target site, Supplementary Fig. 3b) or a corresponding empty vector (both co-expressing VENUS), and a miR-193b expression vector or a corresponding empty vector (both co-expressing eBFP2). A MOI 50 was used for each vector. The cells were cultured for up to 9 days in SFEM (Stemcell Technologies) supplemented with 100 ng ml À 1 murine SCF and TPO. Reporter fluorescence was measured over time via FACS (Canto II, BD). The ratio of double transduced cells was calculated by dividing the absolute number of double transduced cells at day 9 by the initial absolute number of double transduced cells at day 3.
Virus production. Vesicular Stomatitis Virus-G-pseudotyped lentiviral particles were produced using a split genome approach via calcium-phosphate-mediated transient transfection of human embryonic kidney HEK293T producer cells as recently described 16 . After 48 h, supernatant was collected, filtered (45 mm) and enriched via ultracentrifugation (50,000 g, 2 h). Viral titres were determined via transduction of NIH3T3 cells with various concentrations of virus supernatant and FACS analysis.
RNA sequencing. A total 10,000 LSK cells (CD117 þ Sca1 þ Lineage À ) from four miR-193b À / À and miR-193b þ / þ mice were isolated via FACS. The cells of one genotype were then pooled. The RNA of two independent sets of pooled LSK cells was sequenced. The quality and concentration of the libraries were determined using an Agilent 2,100 Bioanalyser and RiboGreen fluorescence on QuBit (Life Technologies). The libraries were sequenced using a HiSeq2000 system (Illumina). Sample preparation and data analysis were performed as previously described 15 . The RNA-Seq statistical analysis was performed using Partek genomics suit 6.6 software.
DAVID Bioinformatics Resources functional cluster analysis. The list of significantly differentially regulated genes (1.5-fold upregulated or downregulated in miR-193b À / À LSK, Po0.05) was submitted as a list with official gene symbols for DAVID analysis and aligned with murine genetic background. The count threshold and the 'EASE' value were set to 2 and 0.1, respectively. Functional annotation was performed via KEGG pathway analysis.

Ingenuity pathway analysis.
To obtain information about the miR-193bdependent biological processes from the RNAseq data, gene expression data were analysed using Ingenuity Pathways Analysis with default settings according to the manufacturer's instructions (IPA) v.5.0 (Ingenuity Systems Inc). This tool provides information about diseases, molecular function and biological process categories, as well as biological pathways related to the genes obtained from the RNAseq analysis (Supplementary Data 2). In addition, IPA maps each gene within a global molecular network developed from information contained in the Ingenuity Pathways Knowledge Base. Gene networks are generated algorithmically based on their connectivity in terms of expression, activation, transcription and/or inhibition. We used IPA to identify canonical pathways that are affected by miR-193b and upstream regulators predicted to be responsible for the observed mRNA expression changes.
Phosphoflow cytometry. To assess basal signalling activity, the BM cells were flushed in ice-cold PBS and directly fixed in Fix Buffer I after centrifugation. To examine cytokine activation, FACS-sorted LT-HSCs were starved for 1 h in SFEM at 37°C/5% CO 2 . Next, the cells were either left untreated (control) or stimulated for 20 min with 100 ng ml À 1 SCF and TPO, 20 ng ml À 1 IL3 and IL6 and 5 U ml À 1 EPO. For the stimulation kinetics, LT-HSCs were stimulated for 0, 5, 20, 60 and 120 min with the aforementioned cytokine cocktail. The cells were then fixed with Fix Buffer I (BD). After washing the cells in PBS þ 1% FCS þ 0.09% NaN 3 , they were permeabilized with ice-cold Perm Buffer III (BD) for 30 min on ice according to the manufacturer's instructions. Permeabilized cells were subsequently stained with antibodies against p-STAT5-PE, p-ERK1/2-FITC, p-AKT-BV421 and p-STAT3-AF647 (BD) to analyse the phosphorylation status of the cells via FACS. Freshly isolated BM cells stimulated with 100 ng ml À 1 SCF and TPO, 20 ng ml À 1 IL3 and IL6 and 5 U ml À 1 EPO for 20 min before fixation served as a positive control.
PamGene serine/threonine and tyrosine array. The kinase activity assessment of total BM cells was accomplished using PamChip Tyrosine and Serine/Threonine Kinase Array Chips and a PamStation 12 system according to the manufacturer's instructions. Briefly, 2 Â 10 6 total BM cells isolated from four individual male mice of each genotype (miR-193b À / À and miR-193b þ / þ ) were lysed in M-PER Mammalian Extraction Buffer (Pierce).
For the tyrosine kinase array, 2 mg of cleared cellular lysate was mixed with 4 ml of 10 Â protein tyrosine kinase reaction buffer (PK), 0.4 ml of 1 M dithiothreitol, 0.4 ml 100 Â BSA, 1 ml of 4 mM ATP and 0.3 ml 1 mg ml À 1 monoclonal anti-phosphotyrosine FITC conjugate (clone PY20), which was adjusted to 40 ml with distilled H 2 O. All chemicals were provided by PamGene International BV. Each array was blocked with 0.2% bovine serum albumin and washed with PK solution. A kinase reaction was then carried out at 30°C. The reaction mix was pulsed back and forth through the porous material of the PamChip for 60 cycles. A picture was taken with a built-in CCD camera every fifth cycle.
For the Serine/Threonine Kinase Array, 2 mg of cleared cellular lysate was mixed with 4 ml of 10 Â PK buffer, 0.4 ml of 100 Â BSA and 1 ml of 4 mM ATP adjusted to 40 ml with distilled H 2 O. Each array was blocked with 0.2% bovine serum albumin and washed with PK solution. A kinase reaction was subsequently carried out at 30°C. The reaction mix was pulsed back and forth through the porous material of the PamChip for 60 cycles. Next, the Detection Mix consisting of 3 ml of 10 Â Antibody buffer, 0.34 ml of STK antibody mix and 0.4 ml of STK antibody FITC-labelled, adjusted to 30 ml with distilled H 2 O was added to the chip. The reaction mix was pulsed back and forth through the porous material of the PamChip for an additional 30 cycles. A picture was taken with a built-in CCD camera every fifth cycle.
Spot intensities were normalized to the local background signal by subtracting the median background signal from the median spot intensity using BioNavigator (PamGene) software. A constant background of 40 was subtracted from each raw median signal. The value of 40 was assessed in a preliminary experiment from an empty array, which was not prepared with any sample. To make arrays comparable and to remove a typical mean-variance-dependency, we normalized the data using the 'vsn' method. This method also allows for negative values and therefore uses the arcsinh instead of the logarithm for variance stabilization. All analyses were performed with R software (version 3.0.1). Comparing miR-193b À / À and miR-193b þ / þ samples revealed differentially expressed peptides using the linear models proposed by Smyth 40 , which are implemented in the R-package 'limma'. To reduce the number of false positive results, raw P-values were adjusted using the method of Benjamini and Hochberg. The log2 fold change (log2 FC) was assessed to determine whether genes were upregulated or downregulated.