Human definitive hematopoietic specification from pluripotent stem cells is regulated by mesodermal expression of CDX4

The generation of hematopoietic stem cells (HSCs) from human pluripotent stem cells (hPSCs) is a major goal for regenerative medicine. To reproducibly achieve this goal, we must first understand human hematopoietic ontogeny. Embryonic hematopoiesis is classically defined by the spatiotemporal emergence of at least 2 distinct programs.¹  The first, primitive hematopoiesis, does not give rise to HSCs,²  but instead transiently gives rise to a limited subset of lineages, including HBE-expressing erythroblasts or primitive erythroid colony-forming cells (EryP-CFCs), which can be reliably used as an indicator of the primitive hematopoietic program.^3,4  Shortly thereafter, the definitive program emerges and gives rise to all lineages found in the adult, including the HSC.^5,6  When differentiating hPSCs, the development of the definitive program can be distinguished from the primitive program by using in vitro assays for HBG-expressing erythroblasts and T lymphocytes.³  However, the genetic regulation of human definitive hematopoietic specification, both in vivo and in vitro, remains unclear.

We recently developed a method to independently specify progenitors of the primitive or definitive hematopoietic programs from hPSCs via stage-specific manipulation of canonical WNT signaling.⁴  This approach generates a WNT-dependent KDR⁺CD235a⁻ mesodermal population that gives rise to CD34⁺ definitive hematopoietic progenitors, as well as a WNT-independent KDR⁺CD235a⁺ mesodermal population that gives rise to CD43⁺ primitive hematopoietic progenitors. Using this tractable system, we interrogated the transcriptional regulation of definitive hematopoietic specification at its earliest identifiable mesodermal progenitor, and identified CDX4 as a critical regulator of human definitive hematopoietic progenitor specification.

The hPSC line H1 (WA01; WiCell) was grown and differentiated as described previously.⁴  CD34⁺CD43⁻ hemogenic endothelium (HE) was isolated by fluorescence-activated cell sorting (FACS) and allowed to undergo the endothelial-to-hematopoietic transition as described in detail previously.⁷  Analysis of hematopoietic colony potential was performed as described previously.^3,7  Full experimental details are found in the supplemental Methods, available on the Blood Web site.

Given that our hPSC differentiation system gives rise to populations of mesoderm harboring exclusively primitive or exclusively definitive hematopoietic progenitors (Figure 1A),⁴  we asked whether these populations could identify which transcription factor(s) regulate definitive hematopoietic specification within early mesoderm. We isolated by FACS KDR⁺CD235a⁻ and KDR⁺CD235a⁺ mesoderm, generated by CHIR99021 or IWP2 treatment, respectively (Figure 1B),⁴  and performed whole-transcriptome expression analyses. Differential gene expression analysis⁸  revealed significant enrichment of the CDX and HOX genes within definitive hematopoietic mesoderm (Figure 1C; supplemental Figure 1; supplemental Table 1). CDX1, CDX2, and CDX4 were all highly expressed in definitive, but not primitive, hematopoietic mesoderm, and have been previously identified as being expressed during hPSC-derived definitive hematopoietic specification.⁹ 

Interestingly, qRT-PCR analyses of each gene over the first 6 days of differentiation revealed that CDX1 and CDX2 are expressed within 24 hours of differentiation, whereas CDX4 was instead upregulated twofold at the time of CHIR99021 treatment (Figure 1D). CDX expression immediately decreased following CHIR99021 removal. This suggested that CDX1 or CDX2 may not specifically regulate definitive hematopoietic progenitor specification, but instead regulate mesendoderm specification, and thus may affect the emergence of both programs, whereas CDX4 expression correlates with definitive hematopoietic progenitor specification by WNT signaling.

In addition to WNT signaling, differentiation cultures at this stage employ the use of BMP4 and basic fibroblast growth factor (FGF),^4,7  both of which contribute to the expression of CDX genes in early mesoderm.^10,11  Although inhibition of BMP4 signaling with recombinant NOGGIN caused a complete block in mesoderm formation (not shown), the inhibition of basic FGF signaling with PD173074 at the same time as CHIR99021 treatment led to the emergence of a CD34⁺CD43⁻ population that lacked a distinct CD73⁻CD184⁻ HE (Figure 1E).^7,12  Interestingly, FGF receptor inhibition (FGFRi) had no effect on the expression of CDX1 or CDX2, but did repress CDX4 expression (Figure 1F). Collectively, these results suggested that CDX4 is a relevant transcriptional target during human definitive hematopoietic specification.

In both mouse and zebrafish models, cdx4 deficiency has been implicated in the regulation of embryonic hematopoiesis.^11,13-17  Although Cdx4^−/− mice showed no significant definitive hematopoietic defects,¹⁸  zebrafish exhibit a moderate decrease in definitive hematopoietic potential that was significantly enhanced when combined with cdx1 knockdown,^13,14,16  suggesting species-specific usage of different Cdx genes during definitive hematopoietic specification. As we observed definitive hematopoietic mesoderm-specific expression of CDX4 (Figure 1C), we reasoned that the role of CDX4 in human embryonic hematopoiesis may be elucidated using our hPSC model system, by monitoring the ontogeny of primitive and definitive hematopoietic progenitors from early mesoderm.⁴ 

We first generated an inducible expression system using the AAVS1 locus¹⁹  to allow for CDX4 expression at any stage of differentiation. Exogenous CDX4 expression from days 2 to 3 of differentiation (supplemental Figure 2A) repressed the specification of the primitive hematopoietic progenitors, as we observed a dramatic decrease in the emergence of CD43⁺ primitive hematopoietic progenitors²⁰  (Figure 2A), and a 10-fold decrease in detectable EryP-CFCs (Figure 2B). Because this resembled CHIR99021 treatment of differentiation cultures,⁴  which causes CDX4 expression (Figure 1C-D), we asked whether induction of CDX4 expression during the same window of time would yield a similar effect as CHIR99021 treatment, resulting in an enrichment of definitive hematopoietic specification.⁴  Therefore, differentiation cultures did not have WNT signaling manipulated, so as to allow for the specification of both programs (Figure 1A). Under these conditions, the definitive hematopoietic progenitors were still specified with doxycycline treatment, because functional CD34⁺CD43⁻CD73⁻CD184⁻ HE^7,12  was still specified with similar efficiency to control conditions (Figure 2C-D; supplemental Figure 2B). However, CDX4 overexpression caused a functional increase within this population, as indicated by a fivefold increase in definitive erythroid progenitors, similar to that observed following CHIR99021 treatment (Figure 2E). Collectively, these observations indicate that mesodermal CDX4 expression recapitulates WNT activation during mesoderm specification,⁴  resulting in definitive hematopoietic specification and primitive hematopoietic repression.

We next used a CDX4 knockout (CDX4^−/Y) hPSC line, which completely lacked WT CDX4 expression, whereas CDX1 and CDX2 were still expressed (Figure 2F). Under all differentiation conditions (Figure 1A), CDX4^−/Y cultures gave rise to approximately threefold more CD43⁺ cells (Figure 2G). CDX4^−/Y hPSCs gave rise to threefold more EryP-CFCs in the absence of WNT signaling (Figure 2H), likely due to the complete absence of CDX4 expression in comparison with control IWP2-treated cultures (Figure 2F). CHIR99021 treatment still repressed primitive hematopoiesis, suggesting other β-catenin transcriptional targets also repress primitive hematopoietic specification. In contrast, when CHIR99021-derived definitive CD34⁺CD43⁻ cells were assessed for hematopoietic potential, the CDX4^−/Y cells exhibited an approximately sevenfold reduction in HBG-expressing BFU-E potential, and a fourfold decrease in myeloid potential (Figure 2I; supplemental Figure 2C). Further, limiting dilution analyses of T-lymphoid potential revealed a fivefold decrease in lymphoid potential in CDX4^−/Y CD34⁺ cells (Figure 2J), indicating a broad decrease in multilineage definitive hematopoietic potential.

To determine the cause of this, we asked whether CDX4^−/Y hPSCs have either defective HE specification or an impaired endothelial-to-hematopoietic transition. CDX4^−/Y CD34⁺CD43⁻ cells exhibited an average fivefold reduction in the specification of CD73⁻CD184⁻ HE (Figure 2K-L), indicating they have impaired definitive hematopoietic specification. However, this HE population remained functional, giving rise to definitive BFU-E with similar efficiency to control hPSCs (Figure 2M). Because both CDX1 and CDX2 were still expressed under these conditions (Figure 2F), it is possible that either may be redundantly²¹  contributing to definitive hematopoietic specification in the absence of CDX4, similar to that observed in zebrafish.¹⁸  Collectively, these results establish that CDX4 is a key regulator of specifying human definitive hematopoietic progenitors within mesoderm, by regulating the specification of HE, and thus provides a mechanistic basis for WNT-mediated definitive hematopoietic progenitor specification.⁴ 

CDX/HOX expression has been suggested to be essential for definitive hematopoietic specification from hPSCs.^9,22  Together, our data demonstrate the importance of CDX4 expression for human definitive, but not primitive, hematopoietic progenitor specification within mesoderm, prior to HE emergence. With this insight, it will be possible to interrogate the intrinsic and extrinsic regulators of human definitive hematopoietic specification, so as to ultimately increase hPSC-derived definitive hematopoiesis for regenerative medicine applications.

This work was supported by T32 training grants from the National Institutes of Health, National Heart, Lung, and Blood Institute (HL007088-41) (J.P.C. and C.D.) and (HG000045) (M.C.V.); T.E.D. is supported by a grant from the Children’s Discovery Institute of Washington University and St. Louis Children’s Hospital; and C.M.S. is supported by an American Society of Hematology Scholar Award.

Contribution: J.P.C., Q.R., and C.D. designed and performed experiments, and analyzed the data; J.T.K.H. and M.C.V. analyzed expression data; T.E.D. supervised expression analyses; and C.M.S. designed and supervised the research project, performed experiments, analyzed the data, and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Christopher M. Sturgeon, Washington University in St. Louis, 660 South Euclid Ave, Box 8125, St. Louis, MO 63110; e-mail: csturgeon@wustl.edu.