Single Cell Transcriptome Profiling of Highly Purified Human Megakaryocyte-Erythroid Progenitors (MEP) Reveals New Insights into the MEP Fate Decision

Megakaryocyte-Erythroid Progenitors (MEPs) are bipotent cells capable of generating megakaryocytic (Mk) or erythroid (E) progeny. However, neither the cell fate-determining componentry nor the initial molecular consequences of lineage specification have been defined. To elucidate this, it is critical to rigorously purify MEP from primary cell sources. Unfortunately, existing purification strategies to do this fail to yield pure, bipotent cells. To improve upon existing approaches for the enrichment of primary human MEPs from G-CSF mobilized peripheral blood (MPB) and BM, we used the cell surface markers CD36 and CD110 in order to further enrich MEP from CD34+CD38+Lin-Flt3-CD45RA- cells. We then quantitated the Mk and E potential of those cells using single cell colony assays. Using this approach demonstrated that CD36/CD110 selection led to an increase of biphenotypic MEP (assessed as CFU-Mk/E) from ~15% to ~40% of colonies that grew. However, it was unclear from colony assay data alone whether or not the heterogeneity of the underlying population was accurately reflected. To address this, we subjected the FACS-sorted MEP-enriched population to single cell mRNA deep sequencing using the Fluidigm C1 platform. For comparison to MEP, we also performed single cell deep sequencing of CD34+CD38+CD41+Flt3- and CD34+CD38+Flt3-CD36+ cells, which are highly enriched for megakaryocyte progenitors (MkP) and erythroid progenitors (ErP), respectively. A total of 150 single cells were captured and sequenced with an average of 3 million reads per cell (1x100bp sequencing). The mRNA deep sequencing data was analyzed by a combination of gene and cell bi-clustering approach to identify both transcripts and cells that exhibited shared or differential patterning. Initial expression patterns and cell groups were identified using stringent expression filtering for transcripts that exhibited >10 FPKM in at least one cell, and iteratively defined and refined based on known E, Mk, and other hematopoietic genes, and then extended for all strongly expressed transcripts. For the MkP and ErP groups, the resulting clusters of cells expressed genes indicative of commitment to E or Mk differentiation. In contrast, within the MEP-enriched population, while a few cells clustered with MkP and ErP, the vast majority of cells fell into distinct subsets of uncommitted cells, supporting the idea that the MEP-enriched population was unique and distinct from MkP or ErP. Analysis of the gene expression patterns from the MEP, ErP and MkP revealed two remarkable trends.

First, the transcription factors GATA1 and GATA2 showed distinct expression patterns in the different clusters of cells; there was a subset of MEP that had high GATA2 expression with little to no GATA1 expression (GATA2 cluster), and an opposite cluster containing high GATA1 expression and low or absent GATA2 expression (GATA1 cluster). The genes most positively correlated with GATA2 expression were also low or absent in the GATA 1 cluster. Closer analysis revealed that the GATA 1 cluster cells were predominantly erythroid and megakaryocyte committed, while the GATA2 cluster appeared uncommitted. A third cluster was present, containing intermediate expression of both GATA1 and GATA2. This cluster is as yet undefined, but appears to contain both MkP and MEP, suggesting a possible link between these two cell types. The second pattern we noted was that the genes in the GATA1 cluster correlated very strongly with cell cycle activity and cell cycle progression while the GATA2 cluster geneset had very low cell cycle activity. This suggested that the commitment of the MEP to E or Mk fates could not be unlinked from their cell cycling status. Such a finding could only be ascertained using single cell sequencing. Using single cell sequencing also provided us with a gene expression signature for primary human MkP, something which was not possible before because there is no reliable way to sort pure human MkP. Regarding GATA1 and GATA2 clusters, real time RT-PCR analysis of primary human ErP, MkP, and MEP point to a scenario where the ratio of GATA2/GATA1 is critical to determining the E vs. Mk fate decision. These findings will be further addressed in future studies aiming to understand the link between cell cycle and the MEP fate decision. Our new findings will help clarify genetic events critical for the E/Mk fate decision.