Pluripotent stem cell model of early hematopoiesis in Down syndrome reveals quantitative effects of short-form GATA1 protein on lineage specification

Children with Down syndrome (DS) are susceptible to two blood disorders, transient abnormal myelopoiesis (TAM) and Down syndrome-associated acute megakaryocytic leukemia (DS-AMKL). Mutations in GATA binding protein 1 (GATA1) have been identified as the cause of these diseases, and the expression levels of the resulting protein, short-form GATA1 (GATA1s), are known to correlate with the severity of TAM. On the other hand, despite the presence of GATA1 mutations in almost all cases of DS-AMKL, the incidence of DS-AMKL in TAM patients is inversely correlated with the expression of GATA1s. This discovery has required the need to clarify the role of GATA1s in generating the cells of origin linked to the risk of both diseases. Focusing on this point, we examined the characteristics of GATA1 mutant trisomy-21 pluripotent stem cells transfected with a doxycycline (Dox)-inducible GATA1s expression cassette in a stepwise hematopoietic differentiation protocol. We found that higher GATA1s expression significantly reduced commitment into the megakaryocytic lineage at the early hematopoietic progenitor cell (HPC) stage, but once committed, the effect was reversed in progenitor cells and acted to maintain the progenitors. These differentiation stage-dependent reversal effects were in contrast to the results of myeloid lineage, where GATA1s simply sustained and increased the number of immature myeloid cells. These results suggest that although GATA1 mutant cells cause the increase in myeloid and megakaryocytic progenitors regardless of the intensity of GATA1s expression, the pathways vary with the expression level. This study provides experimental support for the paradoxical clinical features of GATA1 mutations in the two diseases.

GATA1 is a representative hematopoietic transcription factor involved in early hematopoiesis and erythro-megakaryocytic cell development [17][18][19][20][21][22][23][24][25][26][27]. Various mutations in exons 2 to 3 of GATA1 result in the loss of the full-length protein (GATA1fl) and the production of only the short-form protein (GATA1s) translated from the second ATG site, which lacks the amino-terminal activation domain [10,28]. This means that, regardless of the pattern of the mutation, the resulting protein is always a single alternative form produced even without the mutation, albeit in small amounts. This distinguishes this mutation from other oncogenic mutations.
Despite the obvious necessity for GATA1 mutations in trisomy-21 cells, the quantitative impact of GATA1s protein produced as a result of the mutations has not been fully elucidated. Indeed, although some meta-clinical analyses have shown a significant association between the GATA1s expression levels predicted from the variants and the severity of TAM and the frequency of AMKL [29], the early stage pathogenesis is not fully understood. In particular, it remains unclear whether there is a direct causal relationship beyond correlation between the amount of GATA1s protein, rather than its presence per se, and early hematopoietic cell fate associated with disease-specific blood findings. An in vitro model using PSCs was reported to be useful for analyzing diseases of early hematopoiesis [30][31][32]. Of course, it is hard to precisely address if the level of gene expressions in PSC-derived hematopoietic cells be the same in cells of comparable stages in primary disease development during fetal hematopoiesis, but several PSC models of TAM have been already reported to recapitulate a differentiation preference for myelocytes due to GATA1 mutations and an increase in CD34 + immature megakaryoblasts associated with expression level of GATA1s [33][34][35], which correspond to the features observed in patients. Furthermore, recent study using trisomy-21 PSCs identified an CD34 + CD43 + CD11b -CD71 + CD41 + CD235amegakaryocytic progenitor population largely responsible for the myeloid proliferation in the absence of GATA1fl [36]. Interestingly, despite being an erythro-megakaryocytic progenitor population, cells in this fraction possessed an expression profile that showed a tendency for myeloid differentiation, which suggested the need for a more detailed analysis of the effect of GATA1s on the nature of progenitors in earlier developmental stages. Current study therefore examined the effects of higher or lower amount of GATA1s protein levels on each lineage cell by additionally induce GATA1s expression in early-stage hematopoietic cells derived from GATA1 mutant PSCs.

Ethical statement
To establish and use induced pluripotent stem cells (iPSCs), written informed consent was obtained from the guardians of the DS patient (ID: CiRA12345 at Kyoto University and 778 at

Cells and cell culture
The cell line Ts21-ES-GATA1-WT, in which a human chromosome 21 was transferred into the human ESC line, KhES-1-derived subline, and Ts21-ES-GATA1s, in which the GATA1 mutation was introduced into the KhES-1-derived subline and then a human chromosome 21 was transferred into the GATA1s-ES, were previously established [33]. TAM-iPS-GATA1s, which was generated from the blasts of TAM patients with DS, and TAM-iPS-GATA1-WT, in which the GATA1 mutation of TAM-iPS-GATA1s was repaired, were established as described previously [36]. All PSCs were cultured on 0.25 μg/cm 2 Laminin511-E8 fragment iMatrix-511 silk (Nippi, Tokyo, Japan)-coated culture plates with StemFit AK02 medium (Ajinomoto, Tokyo, Japan). For passage, the cells were dissociated into single cells with 0.5×TrypLE Select (Thermo Fisher Scientific, Waltham, MA, USA) and plated at 265 cells/cm 2 . 10 μM Rock inhibitor Y-27632 (Nacalai Tesque, Kyoto, Japan) was used at the time of the plating, and the medium was exchanged with fresh AK02 medium without Y-27632 the next day.

Cell sorting and flow cytometric analyses
The isolation of HPCs on day 6 and subsequent flow cytometric analysis were performed by using a FACS Aria Ⅱ (BD Biosciences). The antibodies used are described in Table 1. Collected cells were counted using C-chip (NanoEnTek, Seoul, Korea) or Countess 1 Ⅱ FL automated cell counter (Thermo Fisher Scientific) and stained in PBS containing 2% FBS for 20 minutes on ice. Samples were analyzed using FlowJo software (FlowJo LLC, Ashland, OR, USA).

Immunoblotting
To confirm the expression of Dox-inducible GATA1 protein, protein was extracted from human PSCs treated with or without 1 μg/mL Dox for 24 hours with RIPA buffer (Wako) supplemented with 2% protease inhibitor cocktail (Nacalai, Kyoto, Japan). Each sample was separated by 10% sodium dodecyl sulfate polyacrylamide gel electrophoresis and transferred to PVDF membranes (Merck Millipore). The membrane was blocked with 5% dry milk and incubated with an anti-GATA1 primary antibody (CST #4589, 1/1,000, Danvers, MA, USA) overnight at 4˚C. The membrane was then incubated with anti-rabbit IgG, HRP-linked secondary antibody (CST #7074, 1/5,000) for 1 hour at room temperature. To confirm the amount of loaded protein, the membrane was stripped with WB stripping solution strong (Nacalai) and probed with -actin (13E5) rabbit mAb (CST #4970, 1/2,000). Signals were detected with Chemi-Lumi One Super (Nacalai) and scanned with ImageQuant LAS 4000 (GE Healthcare, Chicago, IL, USA).

Statistical analyses
Statistical analyses were performed with GraphPad Prism 6 (GraphPad Soft, La Jolla, CA, USA). Results are shown as the mean ± SD and compared with the unpaired Student's t-test.

CD235a -CD34 + CD43 + early-phase multipotent progenitors recapitulate the hematopoietic features of TAM
In order to precisely analyze the effect of GATA1 genotype on the hematopoietic differentiation process, we prepared two sets of isogenic PSC pairs with trisomy of chromosome 21. One pair was human ESCs transferred chromosome 21 (Ts21-ES-GATA1-WT) and the same line with GATA1 mutation introduced (Ts21-ES-GATA1s) [33]. The other pair was iPSCs (TAM-iPS-GATA1s) established from the blasts of a TAM patient with DS and with the GATA1 mutation that repaired (TAM-iPS-GATA1-WT) [36] (Fig 1A). To compare these isogenic pairs, we conducted hematopoietic differentiation (Fig 1B).
Compared to the GATA1-WT strains, early HPCs in GATA1s strains produced few erythroid lineage cells and much more myeloid lineage cells (Fig 1C and 1E). Of note, while immature myeloid cells derived from the GATA1-WT strains continued to decrease with time, those from the GATA1s strains increased until day12 of the culture and were maintained significantly longer than in the GATA1-WT strains thereafter (Fig 1D and 1F). Both strains gave rise to megakaryocytic lineage cells (Fig 1C and 1E), which is consistent with previous studies that showed GATA1fl is not essential for specification into megakaryocytes, unlike erythrocytes [22, 33-35, 41, 42]. Taken together, these data indicated that early HPCs can recapitulate the hematopoietic features of TAM [1].

Establishment of Doxycycline-inducible GATA1s-or GATA1fl-expressing clones
Previous studies have reported that GATA1s is not just the cause of increased myelocytes in TAM, but also that higher expression levels correlate with severe disease groups [29,41]. On the other hand, the incidence of DS-AMKL, which is an oncogenic blast proliferation derived from megakaryocytic progenitors, correlates with a lower expression of GATA1s, suggesting that GATA1s has different effects on the myeloid and megakaryocytic lineages in the absence of GATA1fl [7,29]. To clarify this spatiotemporal quantitative effect of GATA1s protein on the nature of multipotent progenitors and each lineage cell type, we next analyzed the differentiation properties of GATA1s strains introduced with Dox-inducible GATA1s expression cassettes (Fig 2A and S2A and S2B Fig). Additionally, we generated GATA1-WT strains with Dox-inducible GATA1s expression cassettes and GATA1s strains in which we added the Doxinducible GATA1fl expression cassettes to evaluate the emergence and rescue of disease phenotypes, respectively (S2C Fig and Fig 2B). The insertion of the GATA1s expression cassette was confirmed by genomic PCR (S2B Fig), and protein expressions induced by Dox treatment were confirmed by western blotting analyses (Fig 2C). Karyotypes of each clones was confirmed by Q-banding analysis (S3A-S3E Fig). To confirm whether there is reproducibility beyond the clones, we also generated corresponding subclones in TAM-iPS clones (S4A Fig), and confirmed karyotypes and Dox-inducible expression of GATA1 protein (S4B-S4G Fig).

GATA1s protein acts to quantitatively sustain immature myeloid cells in competition with GATA1fl
Using the series of modified cells, we examined the quantitative effects of GATA1s by lineage. GATA1s overexpression in early HPCs on day 6 significantly increased commitment into myeloid lineage (Fig 3A and 3B). Moreover, overexpression from day 9 of the differentiation, when immature myeloid progenitors had already appeared in culture (Fig 1D), also significantly increased the number of immature myeloid progenitors (Fig 3C and 3D). Considering that GATA1fl deficiency itself led to an increase in myeloid cells even without exogenous GATA1s expression (Fig 1D), these results suggested that GATA1s leads to a further proliferation of the myeloid lineage brought about by the loss of GATA1fl by sustaining committed progenitors. Consistent with this result, we observed that overexpression of GATA1s tended to increase the number of colonies containing non-megakaryocytic (non-Mk) cells in colonyforming unit assay of megakaryocytic progenitors (CFU-Mk) (S5A, S5B and S5D Fig) and larger non-Mk colonies was seen in GATA1s overexpressed samples (S5E Fig) as previously reported [30]. In TAM-iPS-GATA1s derived clones, due to differences in the differentiation properties, it was not possible to detect increase myeloid commitment by quantitative increase of GATA1s (S6A and S6B Fig), but there was tendency toward enhanced maintenance of immature myeloid cells (S6C and S6D Fig). These results are consistent with the exacerbation of myeloproliferation in patients with a higher expression of GATA1s. Similar results were obtained in GATA1-WT strains introduced with GATA1s (S7A and S7B Fig) and similar result was obtained for TAM-iPS-GATA1-WT derived clone (S8A and S8B Fig). Whereas, the opposite was observed in GATA1s strains that overexpressed GATA1fl (Fig 3C and 3D), demonstrating that GATA1s and GATA1fl competitively increase and decrease myeloid lineages.

GATA1s protein has conflicting effects on megakaryocyte commitment and persistence in the absence of GATA1fl
Contrary to the correlation with myeloproliferation seen in TAM, meta-clinical analyses on the impact of GATA1 mutation in DS-AMKL are somewhat paradoxical. Although almost all DS-AMKL patients have a GATA1 mutation, some studies have shown that an increased expression of GATA1s is inversely associated with the risk of DS-AMKL [29]. We therefore evaluated the spatiotemporal effects of GATA1s on megakaryocytic lineage, a potential origin of DS-AMKL, following differentiation. GATA1s overexpression in early HPCs significantly reduced megakaryocytic commitment in GATA1s strains (Fig 4A and 4B). Similar results was obtained with TAM-iPS-GATA1s derived clone (S9A and S9B Fig). Consistent with this result, we observed that the overexpression of GATA1s significantly reduced the total number of CFU-Mk (S5A-S5C Fig). Furthermore, an effect of GATA1s overexpression was observed in GATA1s strains but not in GATA1-WT strains (S7C and S7D Fig) and in TAM-iPS-GATA1-WT derived clone (S8C and S8D Fig), suggesting that the effects on megakaryocytic lineage are counteracted by endogenous GATA1fl, even at high concentrations of GATA1s. On the other hand, unexpectedly, GATA1fl overexpression did not restore the megakaryocytic differentiation of GATA1s strains, but rather reduced it as in the case of GATA1s overexpression (S10A, S10B, S11 and S11B Figs). Because the predominant restoration of erythroid differentiation was observed at this time (S10C-S10E and S11C-S11E Figs), these results indicated that GATA1fl at the endogenous expression level is important for the commitment to both erythroid and megakaryocytic lineages, but a higher expression at this stage leads to a significant bias towards erythroid commitment due to its essential role in erythropoiesis, which consequently suppresses megakaryocyte commitment.
The inhibitory effects of GATA1s on megakaryocytic commitment could explain the lower risk of DS-AMKL progression in cases of high GATA1s expression among TAM patients [29]. Nevertheless, it is still clinically evident that GATA1 mutations are by far the most important risk factor for developing DS-AMKL, even in patients with a high expression of GATA1s [1,2]. These facts led us to examine if there is another cause of the accumulation of immature megakaryocytes that could be responsible for DS-AMKL even in GATA1s high-expressing cells with suppressed commitment. Indeed, we found the overexpression of either GATA1s and GATA1fl significantly increased the percentage of total megakaryocytes in GATA1s strains after day 12 of the differentiation (Fig 4C and 4D). However, when focusing on immature megakaryocytic progenitor cells, GATA1s overexpression had a significantly increased CD34 + CD41 + subpopulation, but GATA1fl overexpression did not. (Fig 4E and 4F). In TAM-iPS-GATA1s derived clone, although there was no significant difference in total megakaryocytes, there was a trend toward an increase (S9C  and S9D Fig). Whereas, when we focused on immature megakaryocytic cells, we found that the overexpression of GATA1s in megakaryocytic progenitors on later stage significantly increased the persistence of immature megakaryocytic cells, but GATA1fl overexpression did not (S9E and S9F Fig). These results indicated that GATA1s works to maintain immature cells in megakaryocytic lineage as well as myeloid lineage, but unlike the myeloid lineage, the overexpression of GATA1s in the GATA1-WT strain did not have any effect on immature megakaryocytic cells (S7E, S7F, S8E and S8F Figs). Therefore, the effects of higher GATA1s expression on the maintenance of mutant strain-derived megakaryocytic progenitors are dependent on differences in the responsiveness of the target cells to GATA1s protein, which are conferred by the mutation itself.

Discussion
The exclusive expression of GATA1s protein as a result of GATA1 mutations is an essential process for the onset of both TAM and DS-AMKL. Even though blasts in patients in most cases have been found to be a heterogeneous population with a variety of GATA1 mutations at different expression levels, no study has experimentally examined how the intensity of the gene expression contributes to the pathologies of both diseases. Focusing on this point, we clarified how the spatiotemporal shift of GATA1s protein expression affects the progenitor cells from which both diseases originate by using a PSC model and stepwise hematopoietic differentiation. We successfully observed the quantitative impact of the GATA1s expression level on each stage of each lineage by utilizing a Dox-inducible expression system.
PSC-based studies can reveal new effects of mutant genes that cannot be elucidated by studies using patient primary cells after the disease onset or cell lines that are already addicted to the mutations themselves. Moreover, with respect to DS, there is no suitable mouse model that replicates the phenotypes of human trisomy-21. While previous studies including the overexpression of GATA1s in fetal liver progenitor cells of Gata1 ΔN mice and cord blood CD34 + hematopoietic progenitor cells have reported the GATA1s-dependent expansion of GATA1 mutant cells in myeloid and megakaryocytic lineages [41,43], our study distinguished the effects of GATA1s on the commitment and proliferation of the myeloid and megakaryocytic lineages in the absence of GATA1fl by focusing on the progenitor cells which correspond to common myeloid progenitors, originally defined as an origin of both granulocyte/macrophage progenitors and megakaryocyte/erythrocyte progenitors. Specifically, we found that commitment to megakaryocytes at the early HPC stage were significantly reduced by elevated GATA1s expression, and only in the absence of GATA1fl were the megakaryocyte progenitors maintained in response to GATA1s expression levels. These mutation-and differentiation stage-specific reversal effects contrasted the results regarding myeloid lineage, where GATA1s simply sustained and increased progenitor cells in competition with GATA1fl.
Two hypotheses may explain why once committed megakaryocytic progenitors acquire the ability to proliferate in response to GATA1s like myeloid progenitors only under conditions without GATA1fl. First, some additional genetic or epigenetic modifications that occur during tumorigenesis might confer GATA1s-responsive growth characteristics. Alternatively, GATA1fl deficiency itself might provide intracellular signaling for the perturbation. Indeed, a previous study using trisomy -21 PSCs revealed that the expression profile of a GATA1fl-deficient megakaryocytic progenitor subpopulation responsible for myeloproliferation was biased toward the myeloid lineage [36]. Therefore, GATA1s could hijack the myeloid mechanism to promote the proliferation of megakaryocytic progenitors. Further study of this hypothesis using methods that directly examine access of the GATA1 protein to genomic DNA, such as electrophoretic mobility shift assays and chromatin immunoprecipitation, are needed. Such studies could also reveal new molecular mechanisms, by which the higher expression of GATA1s suppresses megakaryocytic commitment in early HPCs.
Collectively, our results suggested that although GATA1 mutant cells cause the increase in myeloid and megakaryocytic progenitors regardless of the intensity of GATA1s expression, the pathways vary with their expression levels (Fig 5). This model provides an explanation for the paradoxical clinical features in which higher and lower GATA1s expressions are inversely correlated with the severity of TAM and development of DS-AMKL among patients with TAM even though GATA1 mutations are the definitive etiology of both diseases. Future in vitro and in vivo studies are expected to provide more definitive evidence for this model.