Dhx38 is required for the maintenance and differentiation of erythro-myeloid progenitors and hematopoietic stem cells by alternative splicing

Alternative splicing is a fundamental regulatory mechanism of gene expression in higher organisms, requiring accurate regulation of RNA-splicing machinery (Baralle and Giudice, 2017). Mutations in components of the RNA-splicing machinery are present in >50% of myelodysplastic syndrome (MDS) patients, and nearly 60% of genes were predicted to undergo alternative splicing during human hematopoiesis (Chen et al., 2014). Emerging data show that splicing-factor mutations, to some degree, define distinct clinical phenotypes and often have different prognostic impacts (Pellagatti and Boultwood, 2020). Understanding the diversity and complexity of alternative splicing in hematopoiesis remains a challenge.

Hematopoiesis is a hierarchical system, in which hematopoietic stem and progenitor cells (HSPCs) differentiate into progressively committed progenitors and mature cells (Zhang et al., 2018). HSPCs maintain self-renewal and multilineage blood differentiation with a sophisticated mechanism to provide lifelong hematopoiesis. Dysplasia of HSPCs is frequently associated with an increased risk of hematopoiesis-related diseases, such as MDS and acute myeloid leukemia (Baeten and de Jong, 2018). Before the emergence of HSPCs, lineage-restricted progenitors, such as erythro-myeloid progenitors (EMPs), appear and overlap with the latter half of primitive hematopoiesis and the emergence of HSPCs from the dorsal aorta at 36 to 48 h post fertilization (hpf) (Da'as et al., 2012). EMPs give rise to fetal erythrocytes, myeloid lineages, natural killer cells and adult tissue-resident macrophages, and can be distinguished via the expression of lmo2 and gata1 (Dege et al., 2020; Li et al., 2018b; Bertrand et al., 2007, 2008; Forrester et al., 2012; Xia et al., 2021). Compromised EMPs impair the formation of fetal erythrocytes and adult macrophages, resulting in fetal hemoglobinopathy, late-onset neurodegeneration or chronic inflammatory diseases (Hoeffel et al., 2015; Mass et al., 2017; McGrath et al., 2011). Although several studies have been carried out to demonstrate the role of splicing factors in hematopoietic development, few studies have found the role of splicing factors in EMP defects.

EMPs and HSPCs both emerge from the hemogenic endothelium and differentiate into multiple blood cells. In this process, they share common and also distinct mechanisms of development (Yokomizo et al., 2018). For example, canonical Wnt signaling is a common mechanism regulating the emergence of EMPs and HSPCs (Frame et al., 2016). However, in a mechanism distinct from that of HSPCs, EMPs do not require Notch signaling during their appearance (Bertrand et al., 2010). HSPC dysfunction has been observed in splicing-factor knockout models, such as sf3b1, ddx41 and u2af1 (Danilova et al., 2010; De La Garza et al., 2016; Weinreb et al., 2021). However, the mechanism by which splicing factors regulate both the differentiation and maintenance of EMPs and HSPCs is still required to be fully determined.

dhx38 encodes the RNA helicase PRP16 and is required to destabilize the U2-U6 helix I between the first and second catalytic steps in the splicing pathway (Fica et al., 2017). Dhx38 is the only RNA helicase that binds to mitotic noncoding RNA, and interference with DHX38 function in human cells affects mitotic function and leads to disrupted chromatin arrangement during the M phase (Nishimura et al., 2019). In fission yeast, mutations of Prp16 affect a small subset of introns with weak 5′SS-U6 snRNA interactions (Vijayakumari et al., 2019). DHX38 is an endogenous inhibitor of the protein phosphatase PP4, which regulates DNA damage repair and microtubule development (Han et al., 2015). Missense mutations in dhx38 affect the splicing of cell cycle-related genes and regulate heterochromatinization of the centromere. Latif et al. (2018) identified that the missense variant c.971G>A of dhx38 is involved in the early-onset etiology of retinitis pigmentosa. However, there are no theoretical or experimental reports so far regarding the regulation of DHX38 in the hematopoietic system. We found that high expression of DHX38 was restricted to the lymphoid neoplasm diffuse large B-cell lymphoma and thymoma in the Gene Expression Profiling Interactive Analysis (GEPIA) database, and in 2006, DHX38 was reported to be amplified in acute myeloid leukemia (Ma et al., 2006). Whether and how DHX38 acts in the hematopoietic system needs to be further explored.

Zebrafish is an ideal model for investigating embryonic hematopoiesis, as it can survive for several days independently of the cardiovascular system (Li et al., 2018a). Here, we constructed a dhx38 knockout zebrafish line, and observed its hematopoietic development during embryogenesis. Our study highlights the physiological function and splicing mechanisms of Dhx38 during hematopoiesis, and also demonstrates its role in regulating both EMPs and HSPCs.

To establish the expression pattern of dhx38 in zebrafish embryos, we performed whole-mount in situ hybridization (WISH) using a dhx38 probe. We observed that dhx38 was ubiquitously expressed throughout the zebrafish embryos. Outside of the head and tail, we also identified high expression of dhx38 in the posterior blood island (PBI) at 36 hpf and in caudal hematopoietic tissue (CHT) at 48 hpf, where EMPs and HSPCs are known to colonize (Fig. S1A). Moreover, the dhx38 promoter drove mCherry expression in the PBI at 36 hpf and colocalized with cmyb (or myb), a marker of EMPs and HSPCs (Bertrand et al., 2008; Hoeffel et al., 2015) (Fig. 1A,B). These data imply that the expression of dhx38 is potentially required for the development of EMPs and HSPCs. We generated a dhx38^−/− zebrafish line using CRISPR/Cas9 technology (Fig. 1C), with a 10 bp deletion in dhx38 cDNA (c.11_21delTTCCTCCCTG), which was predicted to generate a truncated protein (p.Ser5_ThrfsTer7) (Fig. 1D). The mRNA level of dhx38 was undetectable in the dhx38^−/− embryo at 36 hpf (Fig. 1E).

Primitive hematopoiesis occurs as an early but transitory wave, generating primitive erythrocytes and macrophages at 12-26 hpf. The myeloid progenitor marker pu.1 (also known as spi1b), erythroid progenitor marker gata1 and hemangioblast marker scl (or tal1) are typically expressed in the lateral plate mesoderm of dhx38 mutants at 13 hpf (Fig. S1B). The expression of myelocyte markers mpx and lyz in the rostral blood island and the erythrocyte marker hbae1 (or hbae1.1) in the intermediate cell mass showed no difference between wild-type siblings and dhx38 mutants at 24 hpf (Fig. S1C). Primitive hematopoiesis was unaffected in dhx38^−/− zebrafish.

HSPCs originate from the arterial endothelium. The arterial marker dll4 and venous marker dab2 at 36 hpf showed no significant difference between dhx38^−/− embryos and their wild-type siblings (Fig. S1D). In addition, the patterns of the dorsal aorta, posterior cardinal vein, caudal artery, caudal venous plexus and intersegmental vessels indicated well-developed angiogenesis in dhx38^−/−;Tg (flk1:mCherry) embryos at 36 hpf and 48 hpf (Fig. S1E). These data suggest that the development of the vasculature appears to be unaffected in the dhx38 mutants.

Definitive hematopoiesis initiates through committed EMPs in the PBI that arise independently from HSPCs at 30-36 hpf (Bertrand et al., 2007). EMPs can differentiate into multiple lineages of blood cells, including erythrocytes, macrophages, granulocytes and mast cells, and are labeled by gata1⁺/lmo2⁺ double-positive cells (Bertrand et al., 2010). The expression of lmo2⁺ and gata1⁺ cells (EMPs) was normal at 30 hpf (Fig. S2A), but dramatically increased at 33 hpf in the PBI of dhx38 mutants (Fig. 1F,J). We further examined the expression of genes downstream of EMP differentiation, including the promyelocyte markers pu.1 and cebpa, which are highly expressed in immature cells and within undifferentiated myeloid cells in some hematopoietic cancers (Dai et al., 2016). Both cell types are expressed normally in the PBI of the dhx38 mutant at 30 hpf (Fig. S2B), with expression increasing at 33 hpf (Fig. 1G,J), suggesting a crucial role of Dhx38 in mediating EMP development.

We investigated the expression of the mature granulocyte marker mpx, the mature macrophage marker mfap4 (or mfap4.1), the macrophage marker mpeg1 (or mpeg1.1) and the mature erythrocyte marker hbae1 in the dhx38 mutants and their wild-type siblings. The mRNA profiles were almost identical in their PBIs at 30 hpf (Fig. S2C), but decreased at 33 hpf and 36 hpf in the mutants (Fig. 1H-J; Fig. S3B). These results were confirmed by quantitative real-time PCR (qRT-PCR) (Fig. S3C).

Bertrand et al. isolated EMPs using lmo2⁺/gata1⁺ zebrafish at 30 hpf, and validated the tracing of EMP cells by using cmyb⁺ and gata1⁺ cells (Bertrand et al., 2010; Forrester et al., 2012). Consistent with our WISH results, the number of gata1⁺/cmyb⁺ cells and pu.1⁺ cells in the PBI region of dhx38 mutants was significantly increased compared with wild-type siblings at 36 hpf and 48 hpf, suggesting the accumulation of immature EMPs and promyelocytes in the dhx38 mutants (Fig. 2A,B,G,H; Fig. S3A).

We then employed a zebrafish line, Tg(mpeg1⁺:eGFP), to trace developing macrophages. Mature mpeg1⁺ cells displayed a typical branched morphology in the PBI of wild-type siblings, but fewer branched mature macrophages in the PBI of dhx38 mutants at 48 hpf (Fig. 2C,I). Granulocytes identified by Tg(mpx:eGFP) were also significantly reduced in the mutants (Fig. 2D,J). Taken together, these results suggest that EMP maturation is impaired in the dhx38^−/− embryo. Embryos stained by Sudan Black showed a lower number of and smaller granulocytes, representing immature granulocytes, in dhx38^−/− embryos than in wild-type siblings (Fig. 2E,K). Staining of phagocytic macrophages with Neutral Red showed that the number of functional macrophages in the PBI of dhx38 mutant was significantly reduced (Fig. 2F,L). The increase in the number of immature blood cells and the reduction in the number of mature blood cells clearly show that EMP differentiation is blocked in the dhx38^−/− embryo.

Similar to mammals, definitive hematopoiesis in zebrafish is divided into two independent stages: EMP development and HSPC development. WISH and fluorescent microscopy were employed to investigate whether loss of dhx38 affects the development of definitive HSPCs in embryos. HSPCs labeled with a cmyb probe were correctly identified and localized in the ventral wall of dorsal aorta (VDA) of dhx38 mutants at 36 hpf. Moreover, HSPCs in the dhx38 mutants were found to be higher in number compared with their wild-type counterparts at 48 hpf, but this increase disappeared completely by 56 hpf (Fig. S3D). Fluorescence microscopy of dhx38^−/−;Tg(flk:mCherry;cmyb:eGFP) fish revealed that the formation of the hematogenic endothelium (flk⁺;cmyb⁺) in the VDA in dhx38 mutants was comparable with that in the wild-type siblings at 36 hpf. HSPCs accumulated in the CHT at 48 hpf, but then subsequently decreased by 56 hpf in the CHT (Fig. 3A). The emergence of HSPCs was normal, whereas HSPC maintenance was perturbed in the dhx38 mutants.

Given the impaired maintenance of HSPCs in dhx38 mutants at 56 hpf, we examined the differentiation of HSPCs. Unsurprisingly, dhx38^−/− embryos lacked the expression of the myeloid marker lyz and the erythroid marker hbae1 in the CHT at 3 dpf, and the lymphoid mark rag1 in the thymus at 4 dpf (Fig. S3E). Defects in HSPC maintenance in the dhx38 mutants further resulted in a complete loss of lifelong hematopoiesis.

EMPs appear to overlap with the HSPCs in the PBI region from 30 hpf to 40 hpf, but can be distinguished by visualizing cd41⁺/gata1⁺ cells (EMPs) and cd41⁺/gata1⁻ cells (HSPCs), as previously reported (Bertrand et al., 2010, 2008; Forrester et al., 2012; Xia et al., 2021). We then distinguished HSPCs from EMPs by visualizing the cmyb and gata1 markers in the PBI region. The numbers of both cmyb⁺/gata1⁺ and cmyb⁺/gata1⁻ cells were increased at 48 hpf, but decreased rapidly at 56 hpf in the dhx38^−/− embryos (Fig. 3B-E), indicating that EMPs and HSPCs share similar dynamic characteristics, accumulating at first and then decreasing at later stages. Using an F0 strain, dhx38 del7 (c.14_21delCTCCCTG), we also confirmed the role of dhx38 in hematopoiesis, as this mutant phenocopied that of the dhx38 del10 mutant (Fig. S8).

To further interrogate the biological basis by which EMPs and HSPCs accumulate in the dhx38 mutants, we examined their proliferation. Experiments with 5-ethynyl-2'-deoxyuridine (EdU) revealed that the number of cells undergoing S phase was significantly reduced in the PBI of dhx38^−/− embryo at 36 hpf (Fig. 4A-D). In contrast, immunofluorescence of phosphorylated histone H3 (pH3) showed that both cell types were arrested in the M phase in the PBI of dhx38 mutants at 36 hpf (Fig. 4E-H). Meanwhile, pH3⁺ cells exhibited an abnormal chromosome karyotype in dhx38^−/− embryos (Fig. S4A). Small interfering RNA (siRNA)-mediated interference of DHX38 in K562 leukemia cells phenocopied the G2/M arrest observed in the dhx38 mutants (Fig. 4I; Fig. S4B). We hypothesized that the cell cycles of EMPs and HSPCs were arrested in the M phase in dhx38^−/− embryos. Moreover, an enhanced DNA damage signal was observed in dhx38 mutants at 36 hpf and 56 hpf, indicating an impaired cell cycle (Fig. 4J).

The numbers of EMPs and HSPCs were decreased in dhx38^−/− embryos at 56 hpf. We therefore evaluated whether EMPs and HSPCs underwent apoptosis in dhx38^−/− embryos, via the terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay. Although there was no significant difference between dhx38 mutants and wild-type siblings with regards to the proportion of apoptotic EMPs at 36 hpf in the PBI (Fig. S4C,D), apoptosis increased significantly in the PBI of dhx38^−/− embryos at 56 hpf (Fig. 4K-N). Also, inhibiting P53 (encoded by p53 or tp53) activity in dhx38 mutants failed to rescue the phenotype of granulocyte (mpx) and erythrocyte (hbae3) reduction at 36 hpf (Fig. 8C; Fig. S4E). Therefore, we eliminated the possibility that the changes in granulocyte and erythrocyte numbers in dhx38 mutants at 36 hpf were due to defects in the viability of differentiated cells. But our findings strongly suggest that the observed reduced numbers of EMPs and HSPCs at 56 hpf are the result of increased apoptosis.

We then performed a more detailed investigation of mitosis by monitoring the state of the chromosomes and the spindle during mitosis in dhx38^−/− embryos. A wide range of EMPs and HSPCs exhibited karyotypic abnormalities in the PBI of dhx38^−/− zebrafish at 36 hpf and 48 hpf (Fig. 5A,B). In metaphase, EMPs and HSPCs in wild-type siblings exhibited a typical chromosome alignment on the equatorial plane, with spindles attached to the centromeres, whereas EMPs and HSPCs in the dhx38 mutants displayed an increase in the impaired alignment of chromosomes. Intriguingly, a previous study found a similar chromatin phenotype, called the ‘grape’ phenotype, when DHX38 was depleted from HeLa cells (Nishimura et al., 2019). In anaphase, the dhx38 mutants exhibited an abnormal chromatin phenotype accompanied by progressively abnormal spindle morphology (Fig. 5C-F; Fig. S5).

We profiled the transcriptomes of the PBI region of wild-type siblings and mutants by RNA sequencing (RNA-seq) at 36 hpf. This analysis revealed that 164 downregulated genes and 386 upregulated genes were mainly enriched in centromere-related genes and the P53 signaling pathway (Fig. S6A-C; Tables S1 and S2). These findings are consistent with the observation of abnormal mitosis and increased apoptosis in dhx38 mutants. As Dhx38 mainly mediates the splicing process of pre-mRNA, we speculate that abnormal splicing events may be the cause of hematopoietic abnormalities. RNA-seq analysis revealed 150 genes with retained introns (RI), 685 genes with skipped exons (SE) and 131 alternatively spliced genes (i.e. genes containing an alternative 3′ splice site or A3SS, an alternative 5′ splice site or A5SS, or a mutually exclusive exon, MXE) were differentially expressed in dhx38^−/− embryos (Fig. 6A; Table S3). The most significant events in differential splicing events are displayed in percent spliced in index (PSI) DOWN of SE events (600/966) and PSI UP of RI events (145/966), which implies an increase in exon skipping and intron retention (Fig. 6A). Compared with the unaffected splice sites, the PSI DOWN group of SE events have significantly longer 3′ and 5′ introns and shorter exons. Meanwhile, the PSI UP group of RI events showed shorter introns. The PSI DOWN group of SE events displayed that the exons have lower GC content (Fig. 6C). The shorter length of affected exons/introns and lower GC content may require higher accuracy for splicing factors, and therefore may be more sensitive to dhx38 deletion.

The biological processes of the cell cycle, double-strand break repair, the microtubule cytoskeleton and chromosome segregation were mainly enriched among all the differentially spliced events (Fig. 6B,D). They were verified by semi-quantitative PCR (semi-qPCR) and qRT-PCR in wild-type sibling and dhx38 mutants (Fig. 7A,B; Fig. S7). Significant splicing changes and decreased expression were observed in genes such as those involved in chromatin segregation (mis18a, cenpk, kdm8, smc5 and knl1) and those related to cell cycle kinases (ccng2, ccnb2 and aaas), DNA damage (dtl, trip12, rad9a, eme1, tonsl and recql) and the microtubule cytoskeleton [scrib, ift20, cep131, lzts2a, sept6 (also known as septin6), mak and pard3ab]. Many of these genes are components of centrioles or the spindle apparatus. Previous studies have shown that cenpk, mis18a and knl1 mediate centromeric functions (Cheeseman et al., 2008; Medina-Pritchard et al., 2020). The smc5 and kdm8 genes are involved in chromatin segregation during mitosis (Hsia et al., 2010; Wehrkamp-Richter et al., 2012). Loss of function of ccng2 and ccnb2 has been shown to result in a defect in chromatin condensation (Gong and Ferrell, 2010). aaas (also known as aladin) is a spatial regulator of Aurora A, which plays a critical role in multiple steps of mitotic progression (Carvalhal et al., 2015). The scrib, ift20, cep131, lzts2a, sept6, mak and pard3ab gene products participate in microtubule motor activity and microtubule polymerization (Carvalho et al., 2015; Chen et al., 2009; Holdgaard et al., 2019; Hong et al., 2010; Sudo and Maru, 2008; Wang and Kung, 2012; Zhu et al., 2017). The dtl, trip12, rad9a, eme1, tonsl and recql genes are essential for genome integrity and chromosomal stability (Calzetta et al., 2020; Gatti et al., 2020; Huang et al., 2019; Piwko et al., 2016; Sansam et al., 2006). Of the genes assessed, cenpk, smc5, knl1, kdm8, eme1, tonsl, ccng2, ccnb2 and aaas are predicted to undergo premature termination codon-nonsense-mediated mRNA decay (PTC-NMD) events, whereas trip12, scrib, pard3ab, sept6, ift20 and lzts2a undergo differential isoform transition, with mis18a, dtl, rad9a, mak and cep131 generating disordered isoforms (Fig. 7C).

Given the essential role of these alternatively spliced genes in mitosis, these observations likely explain the abnormal mitosis events observed in the dhx38 mutants. Western blotting of CCNB2 in si-DHX38-641-treated cells also displayed decreased protein expression, compared with that in cells treated with negative control siRNA (si-NC) or si-DHX38-325 (Fig. 7D). All the isoform transition or NMD events of differentially spliced genes are provided in Tables S3 and S4.

Accumulation of EMPs and HSPCs was followed by induction of apoptosis. The expression of p53 and its target genes p21 (also known as cdkn1a), mdm2 and puma (or bbc3) is upregulated in dhx38^−/− embryos (Fig. 8A). The levels of the P53 protein also appeared to be overly increased in the mutants at 56 hpf (Fig. 8B). WISH assays revealed increased expression of p53 in the CHT of the mutants at 36 hpf and in the aorta-gonad-mesonephros and CHT of the mutants at 56 hpf (Fig. 8C,G). We rescued the embryos by inhibiting P53 activity. As expected, the number of HSPCs at 56 hpf was significantly increased after incubation with 1 μM of the P53 inhibitor PFT-β in the dhx38 mutants (Fig. 8D,E,H-J). Both treatment with PFT-β and p53 knockout extended the survival ratio of the dhx38 mutants (Fig. 8F). Taken together, these data suggest that the induction of apoptosis in dhx38 mutant is P53 dependent.

In this study, our findings demonstrate that Dhx38 is essential for the maintenance and differentiation of EMPs and HSPCs during zebrafish embryogenesis. The importance of EMPs in fetal hematopoiesis has not been fully elucidated to date. Herein, we emphasize the key role of the splicing factor Dhx38 in EMP development. In dhx38^−/− zebrafish, EMPs accumulated, but failed to differentiate into mature myeloid cells. This result suggests that loss of dhx38 may lead to fetal hematopoiesis defects, which may prevent the embryo from surviving. Moreover, EMPs give rise to macrophage precursors that distribute within embryonic tissues and differentiate into adult tissue-resident macrophage subsets, such as the Kupffer cells in the liver, Langerhans cells in the epidermis and alveolar macrophages in the lungs. Impaired EMP differentiation caused by dhx38 deletion may result in the loss of the tissue-resident macrophage pool in adults, leading to related diseases.

Previously, we demonstrated that the splicing factor Bcas2 is vital for HSPC maintenance by regulating P53 signaling to inhibit apoptosis (Yu et al., 2019). Here, we report that the deletion of the splicing factor Dhx38 inhibits cell cycle progression and differentiation of EMPs and HSPCs, and increases apoptosis in a P53-dependent-manner. These defective phenotypes observed in EMPs overlapped with those of HSPCs, pointing to common mechanisms underlying their development and maintenance.

As both EMPs and HSPCs exhibit hematopoietic progenitor cell characteristics, they are likely to have similar pathogenic mechanisms but also have distinct regulatory preferences. Of note, EMPs with fetal characteristics are developmentally restricted, but can be used for long-term multilineage reconstitution upon transplantation into adult recipients (Chen et al., 2011). This hematological model can serve as a useful model for understanding the temporal and spatial control of EMPs and HSPCs and the etiology of hematopoietic malignancies, and for facilitating the development and testing of novel therapeutic interventions.

Mitotic disorders affected by splicing factor dysfunction could be caused by the direct regulation of mitosis by these splicing factors or an indirect consequence of compromised splicing in cells. Previous studies have revealed that mutation of DHX38 affects the function of Aurora B and centromeric heterochromatin, which results in defective chromosome segregation in the mitotic phase (Nishimura et al., 2019; Vijayakumari et al., 2019). However, only a limited number of genes have been characterized as the targets of DHX38 in these studies. Our study is the first to elucidate the role of the splicing factor Dhx38 in regulating mitosis and differentiation by alternative splicing during hematopoiesis. As RNA splicing is often tissue specific, it is plausible that the Dhx38-regulated program in EMPs and HSPCs differs from that of other tissues. Here, we found that deletion of dhx38 causes mitosis arrest in EMPs and HSPCs, accompanied by a ‘grape’ karyotype due to the defects in chromosome alignment. RNA-seq analysis identified abnormally spliced genes regulated by Dhx38, which are related to chromosome segregation, the microtubule cytoskeleton, cell cycle kinases and DNA damage (Fig. 6C,D). Although molecular mechanisms of how each process affects the mitosis of EMPs and HSPCs are currently unknown, it is likely that multiple regulatory mechanisms of Dhx38 co-exist in cells and work together to regulate the cell cycle.

In addition to regulating mitosis, deletion of dhx38 leads to a shorter S phase and increased DNA damage in EMPs and HSPCs, implying an important role for Dhx38 in the maintenance of genomic integrity during DNA replication. Splicing-factor depletion could affect DNA replication during S phase, resulting in R-loop formation and associated genomic instability, which activates the DNA damage response (Chakraborty et al., 2018). Mitotic errors, such as lagging chromosomes, and mitotic delay can also cause DNA damage, which can trigger P53-dependent apoptosis (Edwards et al., 2021; Ganem and Pellman, 2012; Hayashi and Karlseder, 2013; Jelluma and Kops, 2014). A defective DNA damage response can lead to genomic instability underlying many diseases, including hematological disorders and cancer (Bugai et al., 2019; Yoshioka et al., 2021). The interplay between cell cycle inhibition, DNA damage and apoptosis is complex, and how these processes affect each other in dhx38 mutants is unknown; thus, further investigation is required.

In summary, the most striking result to emerge from the data is the new insight into the role of Dhx38 in maintaining differentiation and survival of EMPs and HSPCs, by regulating pre-mRNA splicing of genes involved in mitosis. We hypothesize that mutation of splicing factors in cells with high proliferative requirements, including hematopoietic progenitors, neural progenitors or retinal progenitors, may share similar phenotypes. Mutations of splicing factors in acute myeloid leukemia and related myeloid malignancies have been confirmed; thus, this study contributes to our understanding of the potential role of altered pre-mRNA splicing in the pathogenesis of clonal hematopoietic malignancies.

Zebrafish embryos were reared and kept at 28.5°C under standard aquaculture conditions, as described previously (Hu et al., 2019). Female and male zebrafish of the AB strain under 1 year of age were used. The transgenic zebrafish used in this study included Tg(cmyb:eGFP) (CZ273), Tg(gata1:dsred) (CZ64), Tg(flk:mCherry), Tg(pu.1-gal4-vp16-uas-GFP) (CZ83) and Tg(mpeg1:eGFP) (CZ98), and were purchased from the China Zebrafish Resource Center. Tg(mpx:eGFP) was obtained from Prof. Yiyue Zhang, South China University of Technology (Lian et al., 2018). All animal experimental procedures were reviewed and approved by the Ethics Committee of Huazhong University of Science and Technology.

dhx38^−/− zebrafish were generated by CRISPR/Cas9 technology as described previously (Li et al., 2021). The guide RNAs (gRNAs) were designed by CHOPCHOP (http://chopchop.cbu.uib.no/). mMESSAGE mMACHINE T7 Transcription Kit (Invitrogen) and TranscriptAid T7 High Yield Transcription Kit (Thermo Fisher Scientific) were used to generate Cas9 mRNA and gRNAs, respectively. Zebrafish carrying dhx38 mutations were identified by Sanger sequencing. The dhx38 knockout zebrafish lines generated in this study have been deposited in the China Zebrafish Resource Center (CZRC).

Total RNA from the trunk region of siblings and dhx38^−/− embryos at 36 hpf was extracted using Trizol (Invitrogen). Using the Nanodrop and the Bioanalyzer 2100 (Agilent), RNA sample quality and quantity were assessed. Beijing Novogene and Biomarker used an Illumina HiSeq2000 platform to carry out RNA-seq. RNA-seq quality control and filtering were conducted using fastp software (https://github.com/OpenGene/fastp). Input files following quality control were aligned to the Zebrafish genome (Ensembl GRCz11) by STAR software using the two-pass strategy (Dobin and Gingeras, 2015). Gene expression data were extracted using featureCounts software (Liao et al., 2014) and differential analysis was performed using edgeR (Robinson et al., 2010) (Tables S1 and S2). Alternative splicing analysis was implemented by the software rMATS (Shen et al., 2014) and visualized using rmats2sashimiplot (https://github.com/Xinglab/rmats2sashimiplot) for significant differential splicing events (Tables S3 and S4; Fig. S7) (Li et al., 2021). Metascape (http://metascape.org) was used to identify Gene Ontology terms enriched in the differentially expressed genes and differentially spliced genes (Tables S1 and S2). RNA-seq data in this study have been uploaded to the Gene Expression Omnibus under the accession number GSE165203.

WISH for zebrafish embryos was performed as described previously (Yu et al., 2019). More than 40 zebrafish embryos were used to ensure that the proportion of homozygous embryos was above 10. The genes examined using RNA in situ hybridization included scl, gata1, pu.1, cmyb, lmo2, dll4, dab2, hbae1, lcp1, lyz, cebpa, mpeg1, mpx, hbbe1.1, dhx38, mfap4 and rag1. The primers used to synthesize the probes are listed in Table S5.

Semi-qPCR was performed to confirm abnormal alternative splicing. Primers were designed for differentially spliced genes with aberrant splice sites. The PCR products were analyzed by gel electrophoresis and visualized using the gel ultraviolet detector ZF-401 (Shanghai Guanghao Analytical Instrument). PSI values from 0 to 100 reflected the ratio of correct splicing from the total junction reads (Li et al., 2021). Exons/introns that are constitutively retained in all transcripts and are never skipped have a PSI of 100.

qRT-PCR was conducted with AceQ qPCR SYBR Green Master Mix (Vazyme Biotech) in a StepOnePlus real-time PCR machine (Applied Biosystems). Gene expression was normalized to gapdh. The primers for semi-qPCR and qRT-PCR are summarized in Table S5.

To prepare the staining solution, 0.3 g of Sudan Black (4197-25-5, Sigma-Aldrich) was added to 100 ml of anhydrous ethanol (64-17-5, Sinopharm) and heated or ground for 1-2 days to dissolve completely. To prepare the buffer solution, 16 g of crystalline phenol (108-95-2, Sigma-Aldrich) dissolved in 30 ml of anhydrous ethanol was added to 100 ml of water containing 0.3 g of sodium dihydrogen phosphate (7558-80-7, Sinopharm). These solutions were stored at 4°C and used immediately after mixing the staining solution and buffer solution in a 3:2 ratio. The fixed embryos were incubated in 1 ml Sudan Black staining solution (mixed solution) for 20 min, then washed with 70% ethanol and photographed under the Nikon Eclipse 80i Advanced Research Microscope (Nikon, RRID:SCR_015572).

Neutral Red can be phagocytosed by macrophages and concentrated in their lysosomes, thus causing live macrophages to be labeled red. To prepare the Neutral Red solution, 0.01 g of Neutral Red powder was dissolved in 1 ml of pure water, and the completely dissolved Neutral Red (10 mg/ml) was stored at 4°C and protected from light. Live embryos at 36 hpf were incubated in 1 ml of Neutral Red staining solution (2.5 μg/ml) at 28°C for 8 h and protected from light. Embryos were washed three times in PBS with 0.1% Triton X-100 (PBST) for 5 min each, fixed with 4% paraformaldehyde (PFA) and photographed by Eclipse 80i (Nikon).

Zebrafish embryos were incubated with 1 μM PFT-β (0477-34-1, MedChemExpress) from 12 hpf. The number of zebrafish was counted and genotyped daily to supervise the survival rate.

Embryos were collected and fixed with 4% PFA overnight at 4°C. After washing with PBST, embryos were soaked in acetone for 15 min. Embryos were then blocked with PBST containing 10% goat serum at room temperature for 1 h. After blocking, embryos were incubated with the primary antibody overnight at 4°C. The fluorescent dye-conjugated secondary antibody was used to visualize the signals. The primary and secondary antibodies used are listed in Table S6. The immunofluorescence signals were observed and imaged using an FV1000 confocal microscope (Olympus, RRID:SCR_020337).

TUNEL staining was performed using the TUNEL BrightRed/Green Apoptosis Detection Kit (A113/A112, Vazyme). For the EdU assay, live embryos at 34 hpf were incubated in EdU (2 mM) for 30 min at 4°C and then switched to fresh water for 2 h. Embryos were then fixed with 4% PFA overnight at 4°C, and apoptotic cells were detected by the Cell-Light EdU Apollo567/488 in vitro Kit (C10310/C10338, RiboBio, Guangzhou, China). Images were taken with a Nikon Eclipse 80i Advanced Research Microscope (Nikon, RRID:SCR_015572).

The K562 cell line was obtained from Prof. Yong You's lab (Huazhong University of Science and Technology) and tested for contamination (Zhou et al., 2021). siRNAs targeting different coding regions of human DHX38 were designed by RiboBio . Two independent siRNAs were used for the DHX38 gene to avoid off-target effects. The siDHX38-641 RNA is at the c.641 position of the human DHX38 gene, and siDHX38-328 is at c.328. A nonspecific siRNA duplex was used as an siRNA negative control (si-NC). The silencing effects of these two siRNAs were confirmed by western blotting. The sequences of the siRNAs used in this study are listed in Table S7. K562 cells were transfected with DHX38 siRNA using Lipofectamine 3000 (Invitrogen) according to the manufacturer's protocol. After 72 h, suspension cells were collected and washed with cold PBS. Cells were then resuspended in 700 μl of PBS with 300 μl of 70% ice-cold ethanol slowly added on a gentle shaker. Cells were then fixed overnight in the freezer at −20°C. After washing the cells with PBS, 1 ml of propidium iodide solution (at a final concentration of 20 μg/ml propidium iodide, containing 0.1% Triton X-100 and 0.2 mg/ml RNase A) was added, the cells were evenly dispersed and stained for at least 30 min protected from light. Analysis was performed on Cytoflex S (Beckman Coulter, RRID:SCR_019627) and cell cycle calculations were performed with FlowJo software (https://www.flowjo.com/) using the Dean–Jett–Fox model (RRID:SCR_008520).

For statistical analysis, data are represented as mean±s.d., and a statistically significant difference between control and experimental groups was determined by an unpaired, two-tailed Student's t-test. The level of significance was set to P<0.05.

The authors would like to thank Dr Feng Liu, University of Chinese Academy of Sciences, for advising on this paper, Dr Yiyue Zhang, South China University of Technology, for providing transgene fluorescence zebrafish and commenting on this paper, and Dr Yong You, Huazhong University of Science and Technology, for providing K562 cells.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.200450.reviewer-comments.pdf