Genetic and epigenetic orchestration of Gfi1aa-Lsd1-cebpa in zebrafish neutrophil development

Neutrophils are a subset of granulocytes that play crucial roles in host defense and homeostasis (Newburger, 2006; Amulic et al., 2012). Dysregulation of neutrophil development is known to be associated with a variety of human diseases, such as severe congenital neutropenia (SCN), myelodysplasia syndrome (MDS) and leukemia (Moretti et al., 1994; Newburger, 2006). In mammals, the earliest neutrophil progenitors are granulocyte-monocyte progenitors (GMPs), which are specified from common myeloid progenitors (CMPs) (Manz et al., 2002). Recent studies of adult hematopoiesis revealed that GMPs differentiate into committed neutrophil progenitors, which are unipotent and capable of further differentiation into mature neutrophils (Ng et al., 2019; Kwok et al., 2020). Neutrophil lineage specification, proliferation and maturation are governed by the sequential activation of a series of key transcription factors (Zhang et al., 1997; Gombart and Koeffler, 2002; Li et al., 2011; Jin et al., 2012; Yáñez et al., 2015).

Growth factor independence 1 (GFI1) is a zinc-finger DNA-binding protein that acts as a transcriptional repressor predominantly involved in the formation of various blood lineages (van der Meer et al., 2010; Möröy et al., 2015). GFI1 has been shown to repress Meis1-Pbx1-HoxA9, Csf1 and Egr2 to regulate neutrophil progenitor proliferation and differentiation (Laslo et al., 2006; Zarebski et al., 2008; Horman et al., 2009). Recent sequence data has identified a cluster of GFI1 target candidates (Khandanpour et al., 2012; Olsson et al., 2016; Muench et al., 2020), although the specific downstream targets have not been fully elucidated. In addition to genetic regulation, GFI1 is also an epigenetic regulator that recruits chromatin remodelers, such as G9a (EHMT2), LSD1 (KDM1A) and histone deacetylases (HDACs), known to modify histones (Saleque et al., 2007). The role of GFI1 in adult granulopoiesis has been studied extensively in mammals, with genetic and epigenetic regulation of embryonic granulopoiesis being less well characterized.

Dysregulation of GFI1 function often leads to the initiation and progression of various blood diseases (Karsunky et al., 2002; Person et al., 2003; Yücel et al., 2003; Möröy and Khandanpour, 2019). Humans bearing GFI1 mutations and Gfi1-null mice develop SCN (Karsunky et al., 2002; Hock et al., 2003; Person et al., 2003). Furthermore, GFI1 mutations have been associated with acute myeloid leukemia (AML) (Khandanpour et al., 2010) and MDS (Botezatu et al., 2016), and are a possible risk factor for myeloproliferative disorder (Horman et al., 2009; Khandanpour et al., 2012), with progression to AML. Interestingly, the association of low levels of GFI1 with inferior progress to AML has been suggested (Hönes et al., 2016). Loss of GFI1 negatively influences RUNX1-ETO (RUNX1T1)⁺ leukemia development (Marneth et al., 2018), suggesting that GFI1 plays a role in leukemogenesis in a context-based manner. Thus, clarification of a role for GFI1 in disease pathogenesis is essential in order to improve prognosis and treatment. A suitable animal model would be especially useful for not only analysis of the role of GFI1 in pathogenesis but also the development of potential therapeutic approaches.

The zebrafish is an ideal organism for the study of hematopoiesis because it has a transparent body and external development that permits suitable tracing of blood cell development, allowing for study of disease pathogenesis from embryonic stages to adulthood. Zebrafish gfi1aa (Wei et al., 2008), the ortholog of mammalian GFI1, is known to be involved in primitive erythropoiesis (Wei et al., 2008; Moore et al., 2018). Its role in neutrophil development and myeloid disorder pathogenesis is unclear. In this study, we generated a gfi1aa zebrafish mutant and found that gfi1aa deficiency induces the expansion of neutrophil progenitors as early as the embryonic stage, with progression at different levels to MDS-like phenotypes with age. We demonstrate that Gfi1aa, with the help of its co-factor Lsd1 (Kdm1a), inhibits neutrophil progenitor proliferation by targeting cebpa, both genetically and epigenetically.

To dissect the role of Gfi1aa in granulopoiesis, we generated a gfi1aa-deficient zebrafish mutant, gfi1aa^smu10, in which a 23 bp fragment in exon 3 was deleted, resulting in the synthesis of a premature Gfi1aa protein lacking the DNA-binding domain (Fig. 1A-C). To monitor myelopoiesis in gfi1aa-deficient zebrafish mutants, we performed whole-mount in situ hybridization (WISH) to examine the expression of neutrophil and macrophage markers in gfi1aa^smu10 embryos. Results showed that the expression of the neutrophil markers mpx and lyz was markedly increased in gfi1aa^smu10 mutants, whereas the expression of the macrophage marker mfap4 remained unchanged (Fig. 1D-I), indicating that gfi1aa deficiency causes neutrophil lineage expansion without affecting macrophages. We further showed that the neutrophil lineage expansion of gfi1aa^smu10 mutants could be recovered by the reconstitution of gfi1aa (Fig. S1A-C). Taken together, these data indicate that gfi1aa inhibits neutrophil lineage expansion.

Based on the current murine neutrophil development model, neutrophils are classified into two populations: proliferative mitotic neutrophil progenitors and sequential non-proliferative maturation cells (Xie et al., 2020). To identify which phase was affected in gfi1aa^smu10 mutants, mpx:GFP⁺ neutrophil lineage cells from 3 days post-fertilization (dpf) Tg(mpx:GFP);gfi1aa^smu10 mutants and their siblings were sorted by fluorescence-activated cell sorting (FACS) and May-Grünwald-Giemsa stained. First, consistent with the neutrophil lineage expansion indicated by WISH, FACS showed an increased percentage of mpx:GFP⁺ neutrophil lineage cells in gfi1aa^smu10 mutants (Fig. S2A-C). Then, zebrafish developing neutrophils were separated into three stages: N1, N2 and N3 (Fig. 2A). The results showed that the cell numbers of N1, which represents neutrophil progenitors with a round-shape morphology and a large nucleus (Fig. 2A), were markedly increased in gfi1aa^smu10 mutants compared with their siblings (Table S1). Upon examination of each population ratio, we found that the proportion of N1 cells was doubled in gfi1aa^smu10 mutants (∼14%) compared with their siblings (∼7%), whereas the proportion of well-differentiated banded and segmented neutrophils (N3) was decreased in the mutants (Fig. 2B). These results indicate that gfi1aa deficiency leads to excessive neutrophil progenitors and a decreased mature neutrophil population.

Given that the expansion of neutrophil progenitors is the major phenotype of gfi1aa^smu10 mutants, we speculated that gfi1aa deficiency may lead to excessive proliferation of mitotic neutrophil progenitors. To assess this speculation, we performed a bromodeoxyuridine (BrdU) incorporation assay to monitor the proliferation rate of the neutrophil lineage. The results showed that the percentage of mpx:GFP and BrdU double-positive cells was significantly higher in gfi1aa^smu10 mutants (∼6%) compared with their siblings (∼3%) (Fig. 2C,D), demonstrating that gfi1aa deficiency leads to excessive proliferation of neutrophil progenitors.

To identify the key downstream targets of Gfi1 during granulopoiesis, we focused on the Ets and Cebp families of transcription factors, both of which are well-known key regulators involved in neutrophil development (Smith et al., 1996; Zhang et al., 1997). Notably, a recent single-cell-based study in mice revealed that Cebpa is highly expressed in proliferating neutrophil progenitor populations but rapidly declines during maturation (Xie et al., 2020). This observation suggests that CEBPA may play a crucial role in mitotic neutrophil progenitor proliferation. We therefore hypothesized that loss of gfi1aa may cause an upregulation of cebpa, resulting in the accumulation of neutrophil progenitors. To test this hypothesis, we monitored the expression of cebpa in gfi1aa^smu10 mutants. As we anticipated, cebpa transcripts detected by qPCR were elevated in gfi1aa-deficient neutrophil lineage cells (Fig. 3A), and cebpa expression detected by WISH was also increased in gfi1aa^smu10 mutants (Fig. 3B,C). Overexpression of Gfi1aa could reduce the high cebpa level in gfi1aa^smu10 mutants (Fig. S3A,B). Moreover, overexpression of cebpa in zebrafish led to neutrophil lineage expansion (Fig. S4A-C), mimicking the neutrophil lineage accumulation in the gfi1aa^smu10 mutant. These observations support the hypothesis that the neutrophil lineage expansion in gfi1aa-deficient mutants can be attributed to the upregulation of cebpa.

To confirm this observation, we outcrossed gfi1aa^smu10 mutants with cebpa^hkz7 mutant fish (Dai et al., 2016) and tested whether reducing cebpa levels reversed the gfi1aa^smu10 phenotype. Consistent with a previous report (Dai et al., 2016), the neutrophil lineage was completely absent in cebpa^hkz7/hkz7 null mutants but remained normal in cebpa^hkz7/+ heterozygous mutants (Fig. 4). Interestingly, the neutrophil lineage in gfi1aa^smu10/smu10;cebpa^hkz7/+ double mutants (gfi1aa homozygous and cebpa heterozygous embryos) was comparable to that of wild-type (WT; gfi1aa^+/+;cebpa^+/+) embryos (Fig. 4), indicating that the increased number of neutrophil lineage cells in gfi1aa^smu10/smu10 single homozygous mutants was restored by the loss of one allele of cebpa. These results suggest that Gfi1aa prevents inappropriate expansion of the neutrophil lineage by repressing cebpa during neutrophil development. We also found that gfi1aa expression was comparable in cebpa^hkz7 mutants and their siblings (Fig. S4D,E), suggesting that cebpa is genetically downstream of Gfi1aa in embryonic granulopoiesis.

To prove whether Gfi1aa represses cebpa, we detected the transcriptional activity of cebpa using a luciferase reporter assay. The results showed that cebpa transcriptional activity was significantly repressed by overexpression of Gfi1aa (Fig. 5A). To explore further whether Gfi1aa could directly target cebpa in zebrafish, we performed Gfi1aa chromatin immunoprecipitation with sequencing (ChIP-seq) by injecting the pTol2-hsp-gfi1aa-eGFP construct into zebrafish embryos to overexpress a Gfi1aa-eGFP fusion protein (Fig. 5B). To assess the Gfi1aa ChIP-seq data, we compared the peaks between Gfi1aa and input at the transcription start site (TSS) (±2 kb) and found that GFI1 binding peaks were enriched near the TSS (Fig. S5A). Then, we analyzed the genomic distributions of the ChIP-seq peaks and found that most of the peaks located at the introns, proximal promoter regions and distal intergenic regions (Fig. S5B), indicating a crucial role of Gfi1aa in transcriptional regulation. To investigate further the conservation of Gfi1aa and mammalian GFI1, we identified the Gfi1aa binding motif through the ChIP-seq peaks by Homer de novo motif analysis and found that the Rank 1 motif matched the known human GFI1 motif (Fig. S5C). Furthermore, we compared the Gfi1aa ChIP-seq target genes with mouse GFI1 ChIP-seq target genes (Olsson et al., 2016) and found that 4523 homologous genes overlapped (Fig. S5D), including previously known GFI1 target genes [e.g. csf1 (Zarebski et al., 2008) and Egr1 (Zhang et al., 2019)], suggesting high functional conservation between zebrafish Gfi1aa and mammalian GFI1. To verify the ChIP-seq data, we monitored the transcription factor irf8, a known GFI1 target in mice (Muench et al., 2020), which has Gfi1aa binding peaks in its promoter region (Fig. 5C). We found three obvious Gfi1aa binding peaks (−90∼+200, up1 and up2) in the proximal promoter region of cebpa (Fig. 5C), whereas no obvious binding peaks were found in the regulatory region of pu.1 (spi1b) (Fig. 5C), suggesting that Gfi1aa directly targets cebpa but not pu.1. Notably, by overlapping the target genes from zebrafish Gfi1aa ChIP-seq and previous reported mouse GFI1 ChIP-seq data (Olsson et al., 2016), we found that cebpa was also included in the overlapping 4523 targets (Fig. S5D, Table S9). By re-analyzing other GFI1 ChIP-seq data (Khandanpour et al., 2012), we showed that GFI1 was indeed bound to the Cebpa proximal promoter region in mouse (Fig. S5E), suggesting conservation of GFI1 binding to the Cebpa regulatory region between zebrafish and mouse during neutrophil development. Consistent with this, by JASPAR website (http://jaspar.genereg.net/) prediction, we selected the top five GFI1-binding sites and found four out of five predicted Gfi1aa-binding sites located in the obvious ChIP-seq peaks of the cebpa proximal promoter region (sites a and b were within the up1 peak, sites d and e were within the up2 peak, and site c did not overlap with any obvious peaks) (Fig. S5F). These results suggest that Gfi1aa may repress cebpa expression by directly binding to its promoter region.

Gfi1aa has been shown to regulate downstream targets epigenetically (Möröy et al., 2015). Because Gfi1aa genetically represses cebpa during granulopoiesis, we assessed whether Gfi1aa could regulate associated histone methylation status. As H3K4 methylation is a marker of transcriptional activation (Barski et al., 2007), it is possible that alteration of H3K4 methylation may be one means by which Gfi1aa suppresses cebpa. To examine methylations at the regulatory regions of cebpa, we performed H3K4me1, H3K4me2 and H3K4me3 ChIP-PCR to detect H3K4 methylation in gfi1aa^smu10 mutant embryos. The results showed that loss of gfi1aa was associated with a significantly increased level of H3K4me2 at the up1 but not up2 promoter region of cebpa compared with WT embryos (Fig. 6A-D, Fig. S6A-C). By performing a luciferase reporter assay, we found that mutation of cebpa up1 (which contains GFI1-binding sites a and b) resulted in the recovery of cebpa transcriptional activity (Fig. S6D). These data suggest that cebpa is epigenetically regulated by Gfi1aa in zebrafish embryos.

The demethylase LSD1 is a co-factor that could be recruited by GFI1 to repress downstream transcriptional control of hematopoietic cell fate (Velinder et al., 2016). LSD1 knockout mice have an accumulation of granulomonocytic progenitors (Sprüssel et al., 2012). Furthermore, LSD1 inhibitors can activate PU.1 and CEBPα in AML patients (Cusan et al., 2018), but whether LSD1 regulates CEBPα during neutrophil development is unclear. Zebrafish studies suggest that Lsd1 and Gfi1aa may have some collaborative role in erythropoiesis, as lsd1 morphants and mutants have primitive erythroid defects (Takeuchi et al., 2015), and gfi1aa deficiency produces an identical phenotype (Takeuchi et al., 2015; Moore et al., 2018). Whether Gfi1aa and Lsd1 cooperate in neutrophils is unknown.

Recent RNA-seq data suggested an inverse correlation between GFI1 and Cebpa in mouse neutrophil development (Zhu et al., 2018). By further analyzing the RNA-seq data (Zhu et al., 2018), we found similar expression changes of Gfi1 and Lsd1 in neutrophil development, which were opposite to that of Cebpa (Fig. S7A). We speculated that Gfi1aa may epigenetically repress its myeloid target, cebpa, with the help of Lsd1. Hence, we assessed whether knockdown of lsd1 produced neutrophil lineage expansion similar to that observed in gfi1aa-deficient zebrafish. As expected, lsd1 zebrafish morphants exhibited increased lyz⁺ neutrophils compared with control embryos (Fig. 6E), which mimics the neutrophil lineage expansion phenotype of gfi1aa^smu10 mutant embryos. Furthermore, we found cebpa expression to be increased by lsd1 knockdown by morpholino (MO) (Fig. 6F). The level of H3K4me2 at cebpa up1, but not up2, was increased in lsd1 MO compared with WT embryos (Fig. S8A-G). These results imply that Lsd1 may be involved in the Gfi1aa-cebpa regulatory pathway. In order to assess whether Lsd1 could interact with Gfi1aa in zebrafish, co-immunoprecipitation (co-IP) experiments were carried out by co-expressing Myc-Lsd1 and Gfi1aa-eGFP in embryos. The results showed that Gfi1aa-eGFP could co-immunoprecipitate with Myc-Lsd1 (Fig. 6G), suggesting that Gfi1aa physically interacts with Lsd1 in zebrafish embryos. We further assessed whether Lsd1 regulation of cebpa depended on Gfi1aa by Lsd1 ChIP-PCR of the cebpa up1 regulatory region in WT and gfi1aa mutants. The results showed that the binding of Lsd1 to cebpa was markedly decreased in the gfi1aa^smu10 mutant compared with WT (Fig. 6H), suggesting that the regulation of Lsd1 targeting of cebpa is dependent on Gfi1aa. Taken together, these results suggest that Gfi1aa, collaborating with its co-factor Lsd1, epigenetically regulate cebpa H3K4 methylation during granulopoiesis.

As the neutrophil lineage proportion was abnormal during the embryonic stage, we monitored the neutrophil lineage proportion in adult Tg(mpx:GFP);gfi1aa^smu10 mutants. Consistent with the embryonic stage, the mpx:GFP⁺ cell proportion in 10-month-old gfi1aa^smu10 mutant fishes was higher than in their siblings (Fig. S9A-C). The N1 proportion in gfi1aa^smu10 mutants (∼12%) was significantly higher than WT (∼7%), and the N3 proportion in gfi1aa^smu10 mutants was slightly lower than WT (Fig. 7A, Table S2). We further monitored the neutrophil lineage proportion in older Tg(mpx:GFP);gfi1aa^smu10 adult fish (14 months). The N1 proportion was much higher and the N3 proportion was much lower in gfi1aa^smu10 mutants than WT (N1: 16% versus 3%; N3: 8% versus 22%) (Fig. 7B, Table S2), suggesting that the abnormal neutrophil lineage ratio was more severe in older adult gfi1aa^smu10 mutants. These data suggest that the abnormal neutrophil lineage ratio in gfi1aa^smu10 mutants was sustained to adulthood.

As GFI1 mutations are associated with SCN (Karsunky et al., 2002; Person et al., 2003), AML (Khandanpour et al., 2010) and MDS (Botezatu et al., 2016), we further analyzed the adult blood phenotypes of gfi1aa^smu10 mutants to determine whether adult gfi1aa^smu10 mutants exhibited any myeloid disorders. For blood cell counts of 10-month-old fish, we found a myeloid cell (including neutrophil lineage and monocyte lineage) expansion in the kidney marrow (KM) of gfi1aa^smu10 mutants compared with WT (Fig. 7C, Fig. S10A, Table S3). However, peripheral blood (PB) cell counts revealed no apparent change in gfi1aa^smu10 mutants (Fig. 7D, Fig. S10B, Table S3). To trace the pathological phenotypic progression, we further dissected the blood phenotypes of 18-month-old gfi1aa^smu10 mutants and WT fish. The KM blood cell counts demonstrated that precursors were significantly increased in the mutant compared with WT, whereas other blood populations were not significantly altered (Fig. 7E, Fig. S10C, Table S3). The PB exhibited neutrophil and monocyte lineage cell expansions along with relatively decreased lymphocyte proportion in gfi1aa^smu10 mutants compared with WT (Fig. 7F, Fig. S10D, Table S3). Notably, in the gfi1aa^smu10 PB neutrophil lineage, the N1 proportion was significantly increased (Fig. S11A, Table S4). These results suggest that myelodysplasia was aggravated in older gfi1aa^smu10 mutants, with KM blood blast accumulation and PB content imbalance, resembling the ineffective hematopoiesis and the disease progression with age observed in some MDS patients.

We further investigated the survival of gfi1aa^smu10 mutants. The survival rates were similar between WT and gfi1aa^smu10 mutants up to 15 months, but the survival rate of gfi1aa^smu10 mutants was significantly reduced from 16 months onward (Fig. S11B). Notably, the mortality rate of gfi1aa^smu10 mutants at 18 months approached 50% (Fig. S11B), which resembles the shorter survival time of progressed MDS patients.

Collectively, the results described above suggest that the gfi1aa^smu10 zebrafish mutant displays some abnormal blood phenotypes resembling those of MDS patients, with myeloid expansion, late-stage blast-crisis, and consequent lower survival rate.

In this study, by utilizing gfi1aa^smu10 and cebpa^hkz7 zebrafish mutants as well as an Lsd1 zebrafish morphant, we demonstrated that Gfi1aa, Lsd1 and cebpa form a regulatory network controlling neutrophil development, providing new mechanistic insights into granulopoiesis. Moreover, we show that a gfi1aa mutant develops an MDS-like phenotype from embryogenesis to adulthood. As such, this zebrafish model can serve as a valuable tool by which to investigate disease pathogenesis.

The neutrophil lineage can be separated into a mitotic neutrophil progenitor population and a non-mitotic maturation neutrophil population (Ng et al., 2019). Neutrophil development is currently thought to develop from unipotent neutrophil progenitors (Zhu et al., 2018; Kwok et al., 2020), derived from multipotent GMPs. Several single-cell-based studies indicate that during neutrophil differentiation GFI1 is specifically upregulated in the preNeu stages (Evrard et al., 2018), with peak expression during the early neutrophil activation program (Giladi et al., 2018; Xie et al., 2020), consistent with a role for gfi1aa in proliferative neutrophil progenitor development. Thus, we speculate that committed neutrophil progenitors require appropriate Gfi1aa for proper progenitor expansion, which prevents over-proliferation and benefits subsequent differentiation and maturation.

Using transcription activator-like effector nuclease (TALEN) technology, we generated a gfi1aa^smu10 zebrafish mutant and demonstrated a role for Gfi1aa in zebrafish embryonic granulopoiesis. Deregulation of gfi1aa results in the expansion of neutrophil progenitors, suggesting that GFI1-caused dysplasia of the neutrophil lineage begins at an early embryonic stage. Furthermore, Gfi1aa has an inhibitory role in neutrophil progenitor proliferation, indicating that the full function of Gfi1aa prevents the uncontrolled proliferation of neutrophil progenitors. This result suggests that Gfi1aa is a risk factor for myelodysplasia disease progression.

Cebpα is thought to be a key transcription factor for GMPs, as a Cebpa null mutation results in neutrophil loss in both mice and zebrafish (Zhang et al., 1997; Dai et al., 2016). Overexpression of Cebpα results in expansion of the neutrophil lineage cell but not monocytic cells in mice (Quintana-Bustamante et al., 2012; Ma et al., 2014), suggesting a selective role in neutrophil development. Cebpα expression is greatest in proliferating neutrophil progenitors, and declines during maturation (Xie et al., 2020), which suggests that proper levels are tightly controlled. Xie et al. have also demonstrated an overlap in the expression of Cebpa and Gfi1 in unipotent neutrophil progenitors in mouse bone marrow (Xie et al., 2020). Furthermore, Zhu et al. reported that the expression changes of Gfi1 and Cebpa during neutrophil development are inversely correlated (Zhu et al., 2018). By re-analyzing mammalian GFI1 ChIP-seq data (Khandanpour et al., 2012; Olsson et al., 2016), we found that GFI1 could also bind to Cebpα regulatory regions (Fig. S5E). Here, in zebrafish, we demonstrated that the gfi1aa mutation results in a high level of cebpa expression that leads to expansion of the neutrophil progenitor population. Downregulation of cebpa by a cebpa^hkz7/+heterozygous mutant in a gfi1aa mutant prevents this neutrophil lineage overproduction. Biochemical and genetic analyses suggest that Gfi1aa controls cebpa levels and the neutrophil lineage expansion process, revealing a new mechanism that is crucial for the proliferation of committed neutrophil progenitors.

Mammalian GFI1 is also considered to be an epigenetic regulator, interacting with the histone demethylase LSD1 to cooperatively repress downstream genes (Saleque et al., 2007). For hematopoietic stem cell emergence, GFI1 recruits LSD1 to regulate target genes (Thambyrajah et al., 2016). Here, we found that zebrafish lsd1 morphants exhibit a neutrophil lineage deficiency similar to gfi1aa mutants. Gfi1aa and Lsd1 may act on cebpa to alter H3K4 methylation status such that Gfi1aa inhibits cebpa. Further studies are required to determine whether Gfi1aa interacts with other co-factors (e.g. G9A and HDACs) to regulate early granulopoiesis.

In humans, GFI1 mutations are commonly associated with SCN (Person et al., 2003), AML (Khandanpour et al., 2010) and MDS (Botezatu et al., 2016). The gfi1aa^smu10 zebrafish model more closely resembles MDS than does the SCN phenotype of Gfi1 knockout mice (Karsunky et al., 2002; Hock et al., 2003). The zebrafish model exhibits neutrophil lineage expansion from the embryonic stages, adult KM or PB myeloid expansion, late-stage blast-crisis, and a reduced survival rate with age. Similarly, MDS patients exhibit ineffective hematopoiesis resulting in BM dysplasia, PB cytopenia, and disease progression with age (Arber et al., 2016). The gfi1aa^smu10 zebrafish closely mimics the disease phenotype, in that neutrophil lineage expansion developed into myeloid expansion in young adults with blast increment in older adults. MDS patients often develop secondary AML with shortened survival time (Greenberg et al., 1997; Tanaka and Bejar, 2019). The gfi1aa^smu10 zebrafish exhibited a similar blast-crisis phenotype with poor survival in late-stage adults, resembling MDS progression. Furthermore, older mutant fish exhibit increased PB monocytes, which might be similar to the syndrome of chronic myelomonocytic leukemia.

In summary, these results provide valuable mechanistic insights into neutrophil development by revealing the Gfi1aa-Lsd1-cebpa regulatory pathway involved in granulopoiesis. Furthermore, a zebrafish model of MDS progression is described that contributes to an understanding of the precise regulation of neutrophil development. Application of these insights and use of the zebrafish model will provide new avenues of investigation that could shed light on our understanding and treatment of GFI1-associated myeloid disorders.

Zebrafish were maintained in standard conditions as described (Westerfield, 2000). The following strains were used: the AB strain, the Tg(mpx:GFP)ⁱ¹¹⁴ transgenic line (Renshaw et al., 2006), the cebpa^hkz7mutant (Dai et al., 2016) and the gfi1aa^smu10 mutant.

The gfi1aa TALEN pair was designed to target exon 3 as described (Huang et al., 2011), and paired TALEN mRNAs were synthesized by the in vitro SP6 mMESSAGE mMACHINE Kit (Invitrogen, AM1340). Then, the mRNA pair was injected into one-cell-stage AB embryos and F1 generation individuals were screened for mutations by sequencing. PCR primers are listed in Table S5.

For WISH, staged embryos were collected and fixed with 4% paraformaldehyde (PFA) as previously described (Thisse and Thisse, 2008). Probes for mpx, cebpa, lyz, mfap4 and gfi1aa were prepared as previously described (Lieschke et al., 2001; Lyons et al., 2001; Liu and Wen, 2002; Zakrzewska et al., 2010; Wei et al., 2008). For BrdU labeling, Tg(mpx:GFP);gfi1aa^smu10/+ intercross embryos at 3 dpf were incubated with a 10 mM BrdU solution for 2 h, fixed with 4% PFA, and labeled as previously described (Jin et al., 2012). Antibodies used were: goat anti-GFP (Abcam, ab6658; 1:400), mouse anti-BrdU (Roche, 11170376001; 1:50), Alexa Fluor 488-conjugated anti-goat (Invitrogen, A32814; 1:400) and Alexa Fluor 555-conjugated anti-mouse (Invitrogen, A31570; 1:400). Images were captured with a Zeiss Axio Zoom.V16, Zeiss Axio Imager.A2 and Zeiss LSM880 confocal microscope system.

Three hundred Tg(mpx:GFP);gfi1aa^smu10/+ and Tg(mpx:GFP);gfi1aa^smu10 larvae were collected and then dissociated into a cell suspension as described (Manoli and Driever, 2012). For FACS of adult KM, fish were dissected as described (Traver et al., 2003), and single-cell suspensions analyzed or sorted with a MoFlo Astrios^EQ sorter (Beckman Coulter). The sorted GFP⁺ cells were collected in PBS with 5% fetal bovine serum. For May-Grünwald-Giemsa staining, sorted cells were centrifuged by cytospin onto slides and stained with May-Grünwald-Giemsa (Sigma-Aldrich, May-Grünwald solution, 63950 and Giemsa solution, 32884) following the manufacturer's instructions. For qPCR, sorted cells were lysed with Trizol (Invitrogen, 15596026), and RNA was extracted and reverse transcribed. qPCR was performed with a LightCycler 96 system (Roche).

Total RNA was extracted with TRIzol reagent according to the manufacturer's instructions. cDNA was transcribed with M-MLV Reverse Transcriptase (Promega, M1701). Myc-tagged lsd1 mRNA was synthesized with an in vitro sp6 mMESSAGE mMACHINE Kit (Invitrogen, AM1340).

For the pTol2-hsp70:gfi1aa-eGFP construct, gfi1aa cDNA containing the coding region but not the stop codon was cloned into the pTol2 vector under the control of the hsp70 promoter, with eGFP fused after gfi1aa cDNA. For overexpression of Gfi1aa-eGFP, 75 ng/μl of DNA construct and 50 ng/μl transposase mRNA were co-injected into one-cell-stage WT embryos, then subjected to a heat-shock assay.

For Gfi1aa-eGFP ChIP, ∼250 embryos were injected with the pTol2-hsp70:gfi1aa-eGFP construct, raised to 12 h post-fertilization (hpf), heat-shocked for 2 h at 39°C, recovered for 1 h, and then fixed with 1.1% formaldehyde at 15 hpf. For H3K4me1, H3K4me2 and H3K4me3 ChIP, ∼100 embryos for each group were deyolked, and fixed at 36 hpf for ChIP assay. For Myc-Lsd1 ChIP, ∼250 embryos were injected with Myc-lsd1 mRNA and collected for fixation at 15 hpf. The following antibodies were used for the ChIP assay: anti-GFP (Abcam, ab6658; 1:200), anti-monoMeK4H3 (Abcam, ab8895; 1:200), anti-diMeK4H3 (Sigma-Aldrich, 07-030; 1:200), anti-triMeK4H3 (Abcam, ab8580; 1:200) and anti-Myc (MBL, M192-3; 1:200).

The ChIP assay was performed as previously described (Leichsenring et al., 2013). Briefly, cross-linked embryos were lysed and sonicated with a Covaris M220 sonicator (duty factor 20%, peak incident power 75, cycles of burst 200, time=10 min) to a size range of 200-700 base pairs. Then sonicated chromatin was incubated together with Dynal Protein G magnetic beads and 5 µg of antibody. After washing and reversal of cross-linking, DNA was purified with phenol-chloroform-isoamyl alcohol. For ChIP-seq, libraries were prepared using the NEBNext Ultra II DNA Library Prep Kit for Illumina and sequenced with PE150 strategy on the Illumina NovaSeq6000 sequencer. For ChIP-PCR, analysis was performed with the SYBR Select Master Mix (Applied Biosystems, 4472908). qPCR primers are listed in Table S5.

The data analysis was carried out as previously described (Furey, 2012). Using Burrows-Wheeler Aligner software (v0.1.17) (Li and Durbin, 2009), we mapped the sequencing reads to Ensembl Zebrafish reference genome (GRCZ11). After SAMtools (Li et al., 2009) to align the reads, we used the bamCoverage in deepTools (v3.5.1) (Ramirez et al., 2016) to visualize the reads in Integrative genomics viewer (IGV). The Gfi1aa binding peaks were called using Macs2 (v2.2.6) (Zhang et al., 2008). Gene annotation and genomic distribution of these peaks were performed using ChIPseeker (v1.24.0) (Yu et al., 2015). We analyzed the Gfi1aa binding motif using Homer Software (v4.11) (Heinz et al., 2010). The annotated nearest genes that were 5000 bp upstream to 5000 bp downstream from the TSS were obtained (the filtered genes are listed in Table S8). For overlapped gene analysis, using the homologous gene file provided by MGI (www.informatics.jax.org/downloads/reports/HOM_AllOrganism.rpt), we mapped the filtered genes to mouse homologous genes (the overlapped genes are listed in Table S9).

HEK 293T cells were transfected with the pGL4.2-cebpa luciferase reporter (150 ng), or the pGL4.2-cebpa-mut luciferase reporter, the pRL-CMV Renilla luciferase reporter (5 ng), the pCS2-gfi1aa expression vector (750 ng), or the pCS2 empty vector (750 ng) using the Lipofectamine 3000 reagent (Invitrogen, L3000015). Then cebpa luciferase activity was measured with a Dual Luciferase Reporter Assay kit (Promega, E1910).

MO for lsd1 (5′-GTTATTCACACCTTGTTGAGATTTC-3′) was obtained from Gene Tools as described (Takeuchi et al., 2015). The MO oligo was diluted to a concentration of 1 mM and injected into one-cell-stage embryos.

For the Co-IP assay, one-cell-stage embryos were injected with Myc-tagged lsd1 mRNA, pTol2-hsp-gfi1aa-eGFP plasmid, or both; then, for each group, ∼200 embryos were subjected to heat shock for 2 h at 20 hpf and collected at 23 hpf. Embryos were deyolked with cold deyolking buffer (1× PBS with 1 mM EDTA, 0.3 mM PMSF) and lysed with cell lysis buffer (Cell Signaling Technology, 9803S) and cocktail protease inhibitors (Roche, 11697498001). Immunoprecipitation for GFP-fusion proteins were performed according to the GFP-Trap_MA manufacturer's instructions (Chromotek, gtma-20). For input control, 20% of the immunoprecipitated extract was preserved. Then, Co-IP western blot was performed. For western blot, after all samples were boiled for 5 min at 95°C, they were loaded onto a 10% SDS-polyacrylamide gel for electrophoresis and the protein was then transfered onto the nitrocellulose membranes. The membranes were analyzed with antibodies. Primary antibodies anti-Myc (MBL, M192-3; 1:2000) and anti-GFP (Abcam, ab6658; 1:2000) were used. Secondary antibodies anti-mouse IgG-HRP (KPL, 074-1806, 1:5000) and anti-goat IgG-HRP (Santa Cruz, sc-2345; 1:1000) were used.

All data were statistically analyzed by GraphPad Prism 7.0. Student's t-test was used to compare two groups that followed normal distribution, otherwise, the Mann–Whitney test was used. One-way ANOVA was used to compare multiple groups. Fisher's exact test was used to compare two categorical variables and the log-rank test was used to analyze the survival rate. P<0.05 was considered statistically significant.

We thank Dr Bo Zhang for providing TALEN reagents and protocol and Kemin Chen for providing cebpa overexpression plasmid.

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.199516