Developmental regulation of myeloerythroid progenitor function by the Lin28b–let-7–Hmga2 axis

We examined the WT mouse myeloerythroid compartment at various stages of definitive hematopoiesis ranging from the E14.5 fetal liver (FL) to the mature BM, first assessing CMP, MEP, and GMP populations (Fig. 1 A). Compared with adult BM, numbers of MEPs in E14.5 FL were sixfold higher than GMPs, a ratio that switched with time to GMP predominance in adult mice (Fig. 1, A and B). By querying granulocyte populations at different stages (Sakamoto et al., 2000), we observed a progressive increase in mature granulocytes with time, paralleling the increase in GMPs (Fig. 1, C and D). We also delineated stages of erythroid development over time (Socolovsky et al., 2001). Compared with adult BM, we found more Ter119⁺-maturing erythroid cells in E14.5 FL, as well as a larger CD71⁺/Ter119⁻ population containing early proerythroblasts (Fig. 1, E and F). Next, we cultured FL and BM mononuclear cells in methylcellulose with myeloerythropoietic growth factors and monitored hematopoietic colony formation. We observed a larger proportion of erythroid colonies in the FL compared with the BM, with more granulocytic colonies arising from BM cells (Fig. 1 G). Finally, as the CMP population is thought to differentiate to either MEPs or GMPs, we analyzed CMP transcriptome profiles by whole RNA sequencing (RNAseq; n = 3 biological replicates for each condition). A comparison with published datasets (Wong et al., 2013; Wu et al., 2014) revealed that transcripts specific to BM CMPs are enriched in myeloid lineage–specific transcripts (relative to erythroblasts), whereas transcripts specific to FL CMPs are enriched in erythroblast-specific transcripts (Fig. 2 A). Consistent with this finding, CMPs from BM contained a significantly higher expression of myeloid-specific genes, including Elane, Mmp8, Ctsg, Mmp9, Ngp, and Ltf, whereas FL CMPs expressed higher levels of the platelet/erythroid factors Rcor2, Ptsg1/Cox1, and P2ry12 (Fig. 2 B). Collectively, these data demonstrate that MEPs and their differentiated progeny predominate in the E14.5 FL compared with BM, in which GMPs and granulocytes predominate, possibly because of transcriptional priming of lineage commitment of CMPs, and describe normal developmental maturation within the myeloerythroid compartment.

Using RNAseq, we found that a total of 1,081 genes were higher and 1,636 genes were lower in E14.5 FL CMPs compared with BM CMPs at a significance cutoff of 0.001 (Table S1). Lin28b was the second most significantly enriched transcript in the FL, which we confirmed by quantitative PCR (qPCR; Fig. 3 A). We noted enrichment of known let-7 target mRNAs in FL CMPs, suggesting that FL CMPs contain lower overall let-7 levels, possibly as a consequence of relatively higher levels of Lin28b compared with BM CMPs (Fig. 3 B). Therefore, we examined developmental distribution of mature let7 forms and observed significantly higher levels of several mature let-7 microRNAs in adult BM CMPs compared with FL CMPs (Fig. 3 C). Because both transcriptional and posttranscriptional mechanisms govern let-7 levels, we measured levels of pre– and pri–let-7 in FL and BM CMPs, as well as levels of a let-7 host gene. We observed significantly higher levels of several let-7 precursors as well as the host gene for pre–let-7b and pre–let-7c2 in adult BM compared with FL CMPs (Fig. 3, D and E). Together, these data support the notion that let-7 accumulates in CMPs with hematopoietic maturation.

Because several mature let-7 species are present at higher levels in adult CMPs relative to FL CMPs, we endeavored to inhibit let-7 activity in the adult and assess effects on myeloerythropoiesis. To this end, we used three complementary tools: (1) ectopic overexpression of LIN28B to impair maturation of let-7 precursors, (2) expression of a pan–let-7 sponge in BM CMPs to functionally sequester mature let-7 species from their endogenous mRNA targets, and (3) use of a mouse model deficient for the polycistronic transcript encoding let-7b (the let-7 form most enriched in BM CMPs compared with FL CMPs) and let-7c2.

We first used a mouse model of doxycycline-induced human LIN28B expression (iLIN28B mice; Shyh-Chang et al., 2013). 14 d of doxycycline exposure achieved a robust induction of LIN28B in BM cells (Fig. 4 A), suppression of several mature let-7 species (Fig. 4 B), as well as a decrease in GMPs and an FL-like increase in MEPs in the BM (Fig. 4, C and D) compared with unexposed iLIN28B mice. LIN28B-expressing adult hematopoietic tissues were generally more proliferative than control tissues by Ki67 staining (not depicted), but we did not observe obvious differences in overall tissue cellularity or differences in the rates of CMP proliferation or apoptosis as measured by in vivo bromodeoxyuridine incorporation and annexin V staining between iLIN28B and control mice (control 21.9 ± 5.5% compared with iLIN28B 25.9 ± 5.7% bromodeoxyuridine-positive CMPs after a 4-h in vivo pulse [P = 0.63]; control 20.5 ± 3.4% compared with iLIN28B 27.0 ± 4% annexin V–positive CMPs in vivo [P = 0.25]; n = 5 control and iLIN28B mice across three independent experiments for both assays; both analyses by Student’s t test). LIN28B induction resulted in increased numbers of Ter119-positive committed erythroid cells (as well as CD71⁺/Ter119⁻ early erythroblasts, similar to E14.5 FL) and decreased numbers of mature BM granulocytes compared with controls (Fig. 4, E–H). By morphology, we observed more erythroblasts with fewer mature neutrophils in the BM (Fig. 4 I) and expansion of the splenic red pulp (Fig. 4 J) in doxycycline-exposed TRE-LIN28B mice compared with unexposed mice. We cultured iLIN28B BM cells in methylcellulose with myeloerythropoietic growth factors and observed increased erythroid colony forming activity in LIN28B-induced cells, analogous to WT FL-derived cultures (compare Fig. 4 K with Fig. 1 G). Competitive transplantation of iLIN28B HSCs with WT cells (Fig. 5, A and B) as well as a hematopoietic-specific LIN28B ectopic expression model showed similar effects on myeloid progenitors (Fig. 5, C–I).

Our second method of interfering with let-7 function in adult myeloerythropoiesis was introduction of a let-7 sponge construct in sorted BM CMPs (Kumar et al., 2008), followed by culture of the sponge- or empty vector–transduced BM CMPs in methylcellulose in the presence of growth factors. Sequestration of let-7 resulted in significantly increased proportions of burst-forming unit erythroid (BFU-E) colonies and mRNA levels of mouse globin genes compared with control cells in culture (Fig. 6, A and B). Third, we analyzed the myeloerythroid progenitor compartment of a strain of mice bearing a deletion of the let-7b and let-7c2 polycistronic host gene (let-7 KO) and observed a trend toward decreased GMPs and increased MEPs (Fig. 6, C and D), as well as fewer BM granulocytes compared with WT isogenic controls (Fig. 6, E and F), but not a significant effect on erythroid maturation or distribution of myeloerythroid colony formation in vitro (not depicted), potentially because of compensatory effects of other let-7 species in this model.

If let-7 function specifies adult BM myeloerythropoiesis, we hypothesized that mistimed ectopic activation of let-7 in the FL would result in an aberrant adult-like myeloerythroid phenotype. To this end, we used a mouse strain wherein expression of a Lin28-resistant chimeric form of let-7g (let-7S21L) is induced by doxycycline exposure (iLet-7 mouse; Zhu et al., 2011) and assessed myeloerythroid colony formation in E14.5 FL-derived cells from iLet-7 embryos and controls after 2–4 d of in utero induction before harvesting of embryos. Induction of let-7S21L resulted in an increased proportion of CFU granulocyte colonies at the expense of BFU-E and CFU-GEMM (granulocyte, erythrocyte, monocyte, and megakaryocyte) colonies (Fig. 7 A). To assess effects of ectopic let-7 in vivo, we induced females bearing mixed litters of iLet-7 embryos and littermate controls at E10.5 with doxycycline in the drinking water, which resulted in robust let-7g expression in the FL when harvested at E14.5 (Fig. 7 B). Although many iLet-7 embryos died, analysis of myeloid progenitors in viable embryos at E14.5 revealed depletion of MEPs (Fig. 7, C and D), a trend toward increased FL granulocytes (Fig. 7, E and F), and diminished numbers of CD71⁺/Ter119⁻ early erythroid progenitors with an adult-like pattern of erythroid differentiation (Fig. 7, G and H).

Because Lin28b decreases during myeloerythroid maturation and ectopic LIN28B drives an FL-like myeloerythroid phenotype in BM CMPs, we examined the progenitor pool in Lin28b-deficient E14.5 FLs and observed an increased number of GMPs in Lin28b-deficient embryos compared with littermate controls (Fig. 8, A and B). Although we did not observe a significant phenotype related to erythroid maturation (not depicted), we did find increased numbers of granulocytes in the FLs of Lin28b-deficient embryos compared with WT littermates (Fig. 8, C and D). Culture of FL mononuclear cells in methylcellulose with growth factors did not demonstrate a significant difference in colony formation between genotypes (not depicted). Importantly, we did not observe a compensatory up-regulation of Lin28a in Lin28b-deficient FLs (not depicted). Consistent with a fetal-specific role for Lin28b in myelopoiesis, complete blood counts in adult Lin28b-deficient mice were similar to control mice, with the exception of a mild anemia in the null mice (not depicted). We hypothesized that the incomplete induction of an adult BM-like myelopoietic phenotype in Lin28b-deficient embryos might be caused by a relative paucity of precursor let-7 species in FL CMPs compared with adult BM CMPs (Fig. 3 C). Therefore, we assessed the ability of Lin28b-deficient E14.5 CMPs to form mature let-7 microRNAs. In the absence of Lin28b, we observed an increase in a subset of mature let-7 forms in FL CMPs (let-7d, let-7g, and let-7i); however, levels of let-7 were still lower than those observed during physiological myeloerythroid maturation (compare Fig. 8 E with Fig. 3 C). Thus, it is likely that the consequences of loss of Lin28b in the FL are mitigated by an inability to generate adult BM-like levels of let-7 microRNAs, possibly caused by low levels of basal transcription of let-7 precursors.

We performed RNAseq on CMPs isolated from WT FL, WT adult BM, Lin28b-deficient FL, and iLIN28B adult BM. The numbers of differentially expressed genes between relevant comparisons are shown in Fig. 9 A. We observed significant enrichment of let-7 target genes in three comparisons (Fig. 9 B), in patterns consistent with the relative let-7 abundance across development described above (Fig. 3 C). Gene set enrichment analysis (GSEA) revealed that mRNAs affected by LIN28B induction in adult BM CMPs had a highly significant coordinate pattern of regulation with those altered during normal development from FL to BM myeloerythropoiesis (Fig. 9 C). As such, we compared the sets of mRNAs altered in FL CMPs and iLIN28B BM CMPs relative to WT BM CMPs and noted a significant degree of overlap (Fig. 9 D). On the list of commonly regulated mRNAs, we found several markers of mature neutrophils (Ctsg and Slpi) and monocytes (Mafb and Spic) that were more abundant in WT BM compared with iLIN28B BM and WT FLs (Fig. 9 E). Among the commonly enriched transcripts in FL CMPs and iLIN28B BM CMPs, 16/61 (26%) had predicted let-7 binding sites, whereas among the transcripts that were commonly lower, only 5/134 (4%) were predicted let-7 targets. Interestingly, LIN28B induction resulted in a 4.58-fold increased expression of embryonic globin Hbb-y relative to WT BM CMPs.

As Lin28/let-7 appears to regulate myeloerythropoiesis at the level of the CMP, possibly via transcriptional priming, we assessed the myeloerythroid differentiation potential of MPPs—which give rise to CMPs in hematopoietic ontogeny—in iLIN28B and control mice. Within the LSK (Lineage⁻, Sca-1⁺, c-kit⁺) compartment, we resolved MPP and long-term and short-term HSC populations, defining MPPs as CD150⁻ and CD48^−/low (Fig. 10 A; Oguro et al., 2013). We did not observe differences in MPP frequency between iLIN28B and control BM (Fig. 10 B). MPPs were sorted and cultured in methylcellulose in the presence of myelopoietic cytokines, and colonies were scored by morphology after 10–14 d. We did not observe differences in colony distribution, save for restraint of full CFU granulocyte potential by ectopic LIN28B expression, and no effect on erythroid differentiation (Fig. 10 C), consistent with an effect on myeloerythroid differentiation mediated predominantly at the CMP level.

By RNAseq, the let-7 target gene Hmga2 (Copley et al., 2013) was one of the most significantly increased mRNAs in both iLIN28B CMPs and FL CMPs compared with adult BM CMPs (Fig. 9 E). We confirmed this differential expression by qPCR (Fig. 10 D). Moreover, we found lower levels of Hmga2 in Lin28b-deficient FL CMPs (Fig. 10 E). Because Hmga2 is a well-known hematopoietic regulator (Ikeda et al., 2011; Lam et al., 2014), we isolated adult BM CMPs and ectopically expressed a mouse Hmga2 transgene or a control vector and cultured the CMPs in methylcellulose with hematopoietic growth factors for 14 d. Expression of ectopic Hmga2 resulted in an increased proportion of BFU-E colonies relative to control cultures after 14 d (Fig. 10 F), paralleling an FL-like phenotype and indicating that Hmga2 is a mediator of the effect of the Lin28–let-7 axis on hematopoietic maturation.

Hematopoiesis is a tightly regulated but adaptable process in which production of terminally differentiated cells responds to pathophysiologic stimuli such as hypoxia, hemorrhage, infection, or inflammation (Orkin and Zon, 2008; Paulson et al., 2011; Takizawa et al., 2012). Basal myeloerythroid output is poised to meet specific developmental stage–specific demands, with erythroid production predominant in the FL to support growth under hypoxic conditions and granulocyte production favored in the adult BM to provide innate immunity and tissue repair capability. Candidate regulators of hematopoietic developmental maturation have begun to emerge (He et al., 2011; Mochizuki-Kashio et al., 2011; Xie et al., 2014). In this study, we document developmental maturation of myeloerythroid progenitors and link this mechanistically to regulation by the Lin28–let-7–Hmga2 axis, which is known to control developmental timing from worms to humans (Ambros and Horvitz, 1984; Yuan et al., 2012; Copley et al., 2013; Shyh-Chang and Daley, 2013).

Lin28 genes function in vertebrates to specify timing of kidney and germ cell development (Ambros and Horvitz, 1984; Moss et al., 1997; West et al., 2009; Urbach et al., 2014), as well as in tissue regeneration (Shyh-Chang et al., 2013) and muscle and nerve cell differentiation (Polesskaya et al., 2007; Yang et al., 2015). Recently, two studies have implicated Lin28 in the timing of hematopoietic phenotypes during development (Yuan et al., 2012; Copley et al., 2013). Specifically, ectopic expression of Lin28 in adult BM cells confers fetal-like HSC self-renewal properties and lymphopoietic phenotypes, indicating that ectopic Lin28 expression may be sufficient to recapitulate prenatal hematopoietic programs (Yuan et al., 2012; Copley et al., 2013). Both of these studies were based on transcriptional analyses showing predominance of let-7 in adult and Lin28 in fetal hematopoietic cells (Yuan et al., 2012; Copley et al., 2013). Our study extends this paradigm to the myeloerythroid system. However, in contrast to these prior studies based exclusively on ectopic overexpression of Lin28, we have dissected the functionality of the Lin28–let-7 axis in myeloerythroid differentiation by manipulating the timing of this developmental switch through both gain-of-function and loss-of-function studies of endogenous alleles in both adult and fetal scenarios, thereby confirming the role of endogenous Lin28 in regulating normal hematopoietic development. Furthermore, our findings provide a paradigm in which the function of this axis or other candidate regulators of developmental timing can be interrogated in other systems.

We observed strong reduction of Lin28b levels and reciprocal recruitment of several mature let-7 species during developmental maturation of CMPs from fetus to adult, as well as coordinate modulation of several let-7–regulated mRNAs with developmental changes in endogenous Lin28b and let-7 expression. Mature let-7 activation in the adult seemed to occur as a result of both the down-regulation of Lin28b—which would be expected to permit full processing of let-7 precursors to mature forms—as well as transcriptional activation of precursor let-7, indicative of two levels of let-7 regulation. Therefore, to investigate the roles of Lin28 and let-7 in early myeloerythroid differentiation, we used the complementary strategies of inhibition of let-7 function in adult tissues and inappropriate let-7 activation in the fetal state.

In the FL, wherein erythroid output predominates, mistimed activation of let-7 activity inappropriately drives granulocyte accumulation at the expense of erythroid activity. This observation agrees with previous studies that let-7 expression increases during terminal granulocyte differentiation (Wong et al., 2014) and that disruption of Lin28 expression can affect granulopoiesis (Wang et al., 2015). These results also parallel studies from the leukemia literature wherein let-7 activity has been shown to promote myeloblast differentiation (Viswanathan et al., 2009; Emmrich et al., 2013; Pelosi et al., 2013). Consistent with both transcriptional and posttranscriptional mechanisms of let-7 regulation, Lin28b deficiency alone did not fully recapitulate an adult-like myeloerythroid phenotype in the FL. In the setting of low levels of endogenous precursor let-7 transcription, there appears to be scant let-7 precursors present for Lin28b to impair. Therefore, the blunted phenotype of Lin28b-deficient FLs is likely caused by a low degree of mature let-7 accumulation, compared with that which occurs during normal myeloerythroid maturation as a consequence of the dual effects of Lin28b depletion and pre/pri–let-7 transactivation. Very little is known regarding regulators of precursor let-7 transcription, with one study implicating HOXA10 as a transactivator of a human let-7–containing tricistron (Emmrich et al., 2014). It is likely that enhanced understanding of the regulation of coordinated Lin28b transcriptional repression and let-7 precursor transcriptional activation will provide further insight into developmental hematopoiesis. Given the number of mRNAs deregulated by Lin28b deficiency in the FL, it could also be speculated that Lin28b may play a significant role in regulating mRNA stability, as has been previously reported (Polesskaya et al., 2007).

We used three methods of let-7 inhibition in adult BM myeloerythroid progenitors: ectopic/mistimed expression of LIN28B, a novel mouse model of precursor let-7b1/c2 deficiency, and a let-7 sponge construct. Although prior studies have suggested that ectopic Lin28 expression in adult hematopoiesis can drive fetal-like hematopoietic output (Yuan et al., 2012; Copley et al., 2013), our findings extend this observation to the myeloerythroid system. Furthermore, our results mechanistically attribute this effect to Lin28 inhibition of let-7 maturation, given the similar effects on myeloerythropoiesis provoked by modulating let-7 activity directly using our let-7 KO mice and let-7 sponge construct. Importantly, we observed that inhibition of let-7 directly in FACS-sorted adult BM CMPs impacted production of differentiated progeny. Together, these findings advance a mechanism by which Lin28b works at least in part via a linear axis involving control of let-7 microRNA maturation and let-7 target genes to specify CMP fate and cell output in the FL and adult BM. Conversely, we observed a very subtle effect of ectopic activation of LIN28B in FACS-sorted MPPs. Together with our RNAseq data, these findings suggest that transcriptional priming at the level of the CMP specifies myeloerythroid differentiation across development. Indeed, human and mouse myeloid progenitor populations possess intrinsic priming in transcriptional programs and ultimate lineage commitment at the single-cell level (Paul et al., 2015; Notta et al., 2016). Given the changes in progenitor lineage priming that occur across development (Notta et al., 2016), we propose that as heterochronic regulators of myeloerythropoiesis, Lin28b/let-7 may play a central role in this process, a hypothesis that deserves further investigation. Finally, our genetic and transplant strategies for hematopoietic-specific LIN28B activation demonstrate that this effect is autonomous within the hematopoietic system.

In WT FL and adult BM, we observed varying patterns of erythroid cell differentiation from the earliest stages of MEP fate commitment through reticulocyte formation. We observed that mistimed LIN28B expression in the adult or let-7 expression in the FL could recapitulate fetal or adult erythroid states, respectively, at inappropriate developmental times, effectively generating erythroid heterochronic mutants. It has been shown that let-7 microRNAs are enriched in adult human erythroid cells compared with umbilical cord cells (Noh et al., 2009), implicating the Lin28–let-7 axis in human developmental erythropoiesis as well. Indeed, these studies have been extended to demonstrate that the Lin28­–let-7 axis acts as a regulator of hemoglobin switching (Lee et al., 2013). Further study is necessary to delineate potential roles for let-7 microRNAs in sequential phases of erythroid precursor differentiation.

Our RNAseq data show that CMPs from FL and adult BM differentially express transcripts characteristic of erythroid and granulocytic differentiation, in agreement with priming for production of appropriately differentiated progeny, as described previously (Pronk et al., 2007; Notta et al., 2016). Our data suggest that this priming may be regulated at least in part by Lin28 and let-7 given the coordinate patterns of gene expression we observed between iLIN28B adult BM and normal FL CMPs. We identified Hmga2 as one of the most highly up-regulated mRNAs in FLs and iLIN28B adult BM CMPs compared with WT adult BM CMPs, consistent with a role as a candidate effector of Lin28b/let-7–driven regulation of CMP output. Hmga2 has been previously implicated as a regulator of myeloid progenitor cell function (Lam et al., 2014), and consistent with our results, its ectopic expression can drive adult BM and extramedullary erythropoiesis (Ikeda et al., 2011). Interestingly, this effect parallels a study wherein insertional activation of HMGA2 by a gene therapy lentivirus in a human β-thalassemia patient resulted in autonomous erythropoiesis and transfusional independence (Cavazzana-Calvo et al., 2010). Further supportive of an intersection with Lin28–let-7 function, Hmga2 previously has been shown to regulate developmental timing in neural stem cells (Nishino et al., 2008).

Our data posit that the Lin28–let-7 axis acts as a heterochronic switch within the hematopoietic system. The roles of Lin28/let-7 as heterochronic regulators appear to be relevant in pathological contexts as well. In cancer, Lin28 acts as an oncofetal factor, whereby developmentally mistimed Lin28 expression provokes aberrant induction of fetal-like cell dedifferentiation and proliferation in certain tumors (Jiang et al., 2012; Rao et al., 2012; Nguyen et al., 2014; Urbach et al., 2014; Wang et al., 2015).

In sum, we propose that the highly conserved Lin28–let-7–Hmga2 signaling axis serves as a heterochronic genetic regulatory system that matches the myeloerythroid phenotype to the appropriate developmental state to meet physiological needs. We speculate that this paradigm of heterochronic regulation is applicable not only to the hematopoietic system, but also to other mammalian developmental processes.

iLIN28B mice, Lin28b-deficient mice, and iLet-7 mice that carry a Lin28-resistant chimeric form of let-7g consisting of the stem loop structure from let-7g fused to the loop of the LIN28-resistant miR-21 have been described previously (Zhu et al., 2011; Shinoda et al., 2013). For doxycycline-inducible expression of LIN28B or let-7g, transgenic mice were crossed to the B6.Cg-Gt(ROSA)26Sor^{tm1(rtTA*M2)Jae}/J strain (006965; The Jackson Laboratory; Hochedlinger et al., 2005). Experiments using WT adult mice and embryos used the C57BL/6J strain. All animal procedures were performed according to animal care guidelines approved by the Institutional Animal Care and Use Committee. To generate let-7 KO mice, a mouse genomic clone containing the let-7b and let-7c-2 isoforms was obtained from a genomic P1-derived artificial chromosome (PAC) library (RPCI-21 129s6/SvEvTAC). A gene replacement vector was constructed by recombineering essentially as previously described (Rodriguez et al., 2007). The final vector contained the MC1-DTA–negative selection cassette and the loxP-PGK-EM7-NeobpA-loxP–positive selection cassette. Targeted AB1 embryonic stem cell clones were identified by Southern blots and then used for blastocyst injections. Male chimeras were mated to 129S5 isogenic-CMV-Cre mice to remove the selection cassette. PCR sequencing was then used to confirm the presence of the 801-bp genomic deletion of let-7b and let-7c-2. The deletion breakpoints are as follows: 5′-GACCTCAAGAAGCCACAACATC-(810-bp deletion)-TACAACCTACTGCCTTCCCTGA-3′. Heterozygous mice were intercrossed to obtain homozygous mutant animals. Mutant animals used in these studies were backcrossed at least six generations into a C57BL/6 genetic background. This strain was also analyzed in separate experiments on the 129-strain background.

Cells were isolated from BM by flushing marrow cavities. Cells were isolated from FL from timed pregnancies. Cells were analyzed on a Fortessa flow cytometer or sorted on a FACS Aria (BD). The following anti–mouse fluorescence-conjugated antibodies were used: anti-B220, CD3, Ter119, Mac-1, Gr-1 conjugated to Pacific blue as a lineage cocktail, anti–c-kit–APC-Cy7, anti-CD34–FITC, and anti-CD16/32–PerCp-Cy5.5 (eBioscience and Affymetrix); anti–Sca-1–PE-Cy7; anti-Ter119–APC and anti–Gr-1–PE; anti–Mac-1–PE-Cy5 and anti-CD150–APC (BioLegend); and anti–Mac-1–AF700, anti-CD48–PE, and anti-CD71–FITC (BD). Cells were counterstained for viability with Sytox blue or propidium iodide (Thermo Fisher Scientific). Viability for cell quantification was determined using a parallel tube stained with Sytox blue.

BM mononuclear cells were isolated and immunostained with surface markers per the flow cytometry methods in the previous paragraph to delineate progenitor populations. They were then bound with annexin V APCs (eBioscience and Affymetrix) according to the manufacturer’s instructions.

Adult mice were injected intraperitoneally with 2 mg bromodeoxyuridine. 4 h later, BM cells were isolated and immunostained with surface markers to delineate progenitor populations, after which they were fixed and permeabilized for intracellular staining of incorporated bromodeoxyuridine with a kit from BD, according to the manufacturer’s instructions.

The let-7 sponge construct was provided in pMSCV-puro and was a gift from P. Sharp (Massachusetts Institute of Technology, Boston, MA; plasmid 29766; Addgene; Kumar et al., 2008). pcDNA 3.1 WT mouse Hmga2 was a gift from D. Bartel (Massachusetts Institute of Technology, Boston, MA; plasmid 14789; Addgene; Mayr et al., 2007). The WT mouse Hmga2 insert was subcloned into pBabe-puro for retrovirus packaging.

The following antibodies were used for Western blotting: rabbit anti-LIN28B (4196; Cell Signaling Technology) and goat anti-actin (sc-1616; Santa Cruz Biotechnology, Inc.).

RNA was isolated using TRIzol reagent (Thermo Fisher Scientific), and cDNA was synthesized using the miScript II kit (QIAGEN). Target cDNAs were detected using a SYBR green detection kit (Thermo Fisher Scientific) and gene-specific primers.

FL and BM mononuclear cells were stained with Wright-Giemsa and May-Grunwald stains (Sigma-Aldrich), and images were photographed.

Myeloid progenitors were isolated by FACS. Ecotropic retrovirus was generated by cotransfection of the appropriate retroviral vector with an ecotropic packaging plasmid. Myeloid progenitors were exposed to retroviral supernatant for 48 h in the presence of 100 ng/ml stem cell factor, 10 ng/ml IL-3, and 10 ng/ml IL-6 (all from R&D Systems), as well as 5 µg/ml polybrene (Sigma-Aldrich).

BM or FL mononuclear cells were resuspended and plated in methylcellulose medium according to the manufacturer’s protocol (M3434; STEMCELL Technologies). In some experiments, cells were cultured in the presence of 1 µg/ml doxycycline (Sigma-Aldrich) or 2 µg/ml puromycin (Thermo Fisher Scientific).

For reverse transcription and library preparation, we used a SMARTer Ultra Low RNA kit (V3) for sequencing and a Low Input Library Prep kit (Takara Bio Inc.) according to the manufacturer’s protocol. The RNAseq libraries were sequenced on the HiSeq 2500 platform (Illumina) yielding 82 million raw reads per sample on average (Table S1). From the raw reads, adapter sequences were removed using a Cutadapt tool (Martin, 2011) allowing some mismatches (the specific parameters we used were −n 2 −e 0.02 −o 15). The cleaned reads were aligned to the mouse genome/transcriptome (10 mm and corresponding University of California, Santa Cruz, gene model) using TopHat2 software (Kim et al., 2013). The mean mapping rate was 81%, and this yields 67 million aligned reads per sample on average (Table S1). The RNAseq raw data and analyzed data have been deposited in NCBI’s Gene Expression Omnibus (GEO) database (Edgar et al., 2002) and are accessible through GEO Series accession no. GSE72521.

To estimate expression levels for each gene, aligned reads per gene were counted using an htseq-count tool (Anders et al., 2015). Counts were then normalized and converted to counts per million and reads per kilobase per million values for further analysis using an edgeR package (Robinson et al., 2010). Differentially expressed genes between any two groups in this study were identified according to the edgeR’s guidelines. Low-level expressed genes, which did not have counts per million values over 1 in at least three samples, were filtered out. Dispersions were estimated, and an exact test was performed. Most computational experiments were conducted on the Orchestra High Performance Compute Cluster at Harvard Medical School. For GSEA analysis, publicly available datasets for normal erythroblasts (GEO accession no. GSM995533), promyelocytes (GEO accession no. GSM1174469), myelocytes (GEO accession no. GSM1174470), and granulocytes (GEO accession no. GSM1174471) were used (Wong et al., 2013; Wu et al., 2014).

Under isoflurane anesthesia, mice were retroorbitally bled, and complete blood counts were obtained on a hematology system (Hemavet 950FS; Drew Scientific).

This was done at the Dana-Farber/Harvard Cancer Center Specialized Histopathology Core.

BM mononuclear cells from age- and gender-matched male or female mice were directly retroorbitally injected (3 × 10⁶ CD45.1 cells/recipient with 150 long-term HSCs [lineage⁻, c-kit⁺, Sca-1⁺, CD48⁻, and CD150⁺]) into anesthetized, lethally irradiated (950 rads) congenic CD45.1 recipients. Recipients received trimethoprim-sulfamethoxazole in their drinking water for 4 wk after transplantation to prevent infection.

Statistics were done with Student’s t test unless otherwise indicated.

Table S1 shows RNAseq data comparing transcript expression between E14.5 FL and adult BM CMPs.

We thank Joanna Han, Anna Zhao, and the Harvard Medical School Research Computing Team for RNAseq data management.

This work is supported by grants from the US National Institutes of Health: National Institute for Diabetes and Digestive and Kidney Diseases (R24-DK092760 and R24-DK49216), the National Heart, Lung, and Blood Institute (NHBLI Progenitor Cell Biology Consortium grant UO1-HL100001), and the National Institute of General Medical Sciences (RO1GM107536; PN2 EY018244 and P30DK036836 to A.J. Wagers). G.Q. Daley is an associate member of the Broad Institute and an investigator of the Howard Hughes Medical Institute and the Manton Center for Orphan Disease Research. R.G. Rowe was supported by NHLBI (5T32HL007574-34). L.D. Wang was supported by a Damon Runyon-Sohn Foundation Pediatric Cancer Research fellowship (DSRG 02-12), an Alex’s Lemonade Stand Foundation for Childhood Cancer Young Investigator award (31582), and a St. Baldrick’s Foundation PALS Scholar award (243625).

The authors declare no competing financial interests.

Author contributions: R.G. Rowe and L.D. Wang designed the research, performed experiments, analyzed data, and wrote the manuscript. S. Coma performed experiments, analyzed data, and performed mouse husbandry. A. Rodriguez generated let-7 KO mice. P.T. Nguyen, S. Ross, and P. Sousa performed experiments and mouse husbandry. R. Mathieu and D.S. Pearson performed experiments. A. Han analyzed RNAseq data. A.J. Wagers and G.Q. Daley supervised experiments, designed the research, and wrote the manuscript.