Csf1rb regulates definitive hematopoiesis in zebrafish

Hematopoietic stem/progenitor cells (HSPCs) are responsible for generating multiple blood cell lineages, including erythrocytes and leukocytes, during early development and adulthood (Jagannathan-Bogdan and Zon, 2013). With erythrocytes providing oxygen and leukocytes providing immune protection for the body, disruption of hematopoiesis leads to severe blood diseases such as leukemia (Hatzimichael and Tuthill, 2010). Transplantation of HSPCs to a recipient in order to restore the normal hematopoietic system of the patient is a promising potential therapy for hematological disorders (Kumar and Geiger, 2017). To date, various strategies have been applied to increase the expansion of HSPCs in vivo or ex vivo. These strategies include stimulation with cytokines or growth factors (Matsunaga et al., 1998; Wohrer et al., 2014), ectopic expression of transcription factors for HSPC maintenance and expansion (Amsellem et al., 2003), modulation of the metabolic state of HSPCs by chemical inhibitors (Huang et al., 2012; Perry et al., 2011), or generation of HSPCs by reprogramming endothelium cells (Lis et al., 2017). However, all of these strategies resulted in the limited expansion of HSPCs (Kumar and Geiger, 2017). Therefore, a more comprehensive understanding of the mechanisms underlying HSPC maintenance and expansion in vivo is required to improve therapeutic applications.

In mammals, HSPCs originate from endothelial cells in the aorta, gonads and mesonephros region (Boisset et al., 2010; Cumano et al., 2001), and they migrate to the fetal liver for expansion and maturation before finally colonizing the bone marrow and sustaining adult hematopoiesis (Kieusseian et al., 2012). To fulfill the intense demand of short-lived mature blood cells to support the normal function of different organs, HSPCs need to be highly active to replenish mature blood cells and at the same time preserve their own cell pool (Bowie et al., 2006). The maintenance and cell fates of HSPCs are tightly regulated by a complicated interplay of intrinsic and extrinsic signals (Wei and Frenette, 2018). Among these signals, cytokine/receptor pairs are essential for modulating HSPC behaviors related to multiple aspects, including proliferation, differentiation, survival and mobilization in their niche. For example, SCF (also known as KITLG)/KIT, TPO/MPL and CXCL4 (also known as PF4)/CXCR2 signaling are necessary for the survival and maintenance of HSPCs (Deshpande et al., 2013; Kimura et al., 1998; Matsunaga et al., 1998; Sinclair et al., 2016; Tan et al., 1990). CXCL12/CXCR4 signaling is known for its important role in retaining HSPCs in the bone marrow niche in adult mice (Broxmeyer et al., 2005). Other cytokines, such as insulin-like factor 2 and angiopoietin 1, have been reported to enhance the expansion of HSPCs (Arai et al., 2004; Zhang and Lodish, 2004). Despite the fact that several cytokines and receptors have been demonstrated to be pivotal in directing HSPCs behaviors and fates, the whole picture of cytokines/receptors in regulating HSPC maintenance and expansion remains obscure.

CSF1R, also known as macrophage colony-stimulating factor receptor (M-CSFR) or Fms, belongs to the class III tyrosine kinase receptor family (Heldin and Lennartsson, 2013). The role of CSF1R has been well studied in mammals with a major focus on its functions in myeloid cells, especially the differentiation, proliferation and survival of tissue-resident macrophages (Stanley and Chitu, 2014). The function of CSF1R in the macrophage lineage depends on its ligands CSF1 or IL34 (Zelante and Ricciardi-Castagnoli, 2012). IL34 is essential for the development of Langerhans cells in the epidermis and microglia in the brain (Wang and Colonna, 2014; Wang et al., 2012; Wu et al., 2018), whereas CSF1 is required for other tissue-resident macrophages (Wiktor-Jedrzejczak et al., 1982). Apart from myelopoiesis, CSF1R also participates in the regulation of embryonic lymphopoiesis, as Csf1r marks myeloid-primed B-cell progenitors (Zriwil et al., 2016) or lymph-myeloid progenitors (Böiers et al., 2013). Intriguingly, Csf1r is also expressed in HSPCs (Miyamoto et al., 2002), and CSF1/CSF1R signaling instructs myeloid fate commitment from hematopoietic stem cells (HSCs) at the single-cell level (Mossadegh-Keller et al., 2016). PLX5622, a chemical inhibitor that is presumed to target CSF1R, displays long-term suppressive effects on CD117 (KIT)⁺ hematopoietic progenitors and CD34⁺ HSCs in mice (Lei et al., 2020). However, a recent study has indicated that the Csf1r expression in mouse HSPCs may be attributed to an artefact caused by association of macrophage remnants (Millard et al., 2021). Accordingly, CSF1R signal is undetectable in HSPCs in the bone marrow of CSF1R reporter mice (Grabert et al., 2020). The kinase dead mutation or hypomorphic mutation of Csf1r causes no defects on HSPCs (Rojo et al., 2019; Stables et al., 2022). These inconsistent findings suggest that further study is required to clarify the role of CSF1R in HSPC regulation.

Zebrafish is emerging as a powerful model for studying hematopoiesis with the advantages of external development, transparent embryos and easy genetic manipulation. Like mammals, zebrafish contains all mature blood cell types (Davidson and Zon, 2004). Initiating at around 30 h post-fertilization (hpf), HSPCs are generated via the endothelial-to-hematopoietic transition (EHT) process in the ventral wall of the dorsal aorta (VDA) and posterior blood island (Kissa and Herbomel, 2010; Tian et al., 2017). From 2 days post-fertilization (dpf) onwards, VDA-derived HSPCs migrate to caudal hematopoietic tissue (CHT), the equivalent hematopoietic tissue of mammalian fetal liver, for expansion and differentiation. Then, they populate the thymus and kidney after 3 dpf and 4 dpf, respectively. Similar to the bone marrow in mammals, zebrafish kidneys function as the major adult hematopoietic organ (Jin et al., 2007; Murayama et al., 2006). Apart from the similar ontogeny and cellular development of hematopoietic cells, molecular signals that regulate hematopoiesis are largely conserved between zebrafish and mammals (Perlin et al., 2017; Zhang et al., 2013). As a result of genome duplication in teleost fish, Csf1r has two orthologs in zebrafish: csf1ra and csf1rb (Braasch et al., 2006). Csf1ra is essential for microglial colonization in the brain and the development of xanthophore cells (Parichy et al., 2000; Wu et al., 2018; Herbomel et al., 2001). Csf1ra and Csf1rb work together as the functional homolog of mammalian CSF1R in regulating microglia development and distribution in the brain (Oosterhof et al., 2019, 2018). In addition, these two receptors also participate in the regulation of osteoclast cell activity during skeleton formation (Caetano-Lopes et al., 2020). Recently, Ferrero et al. found that csf1rb mutant zebrafish with a splice site mutation predominantly showed myeloid and B-lineage defects (Ferrero et al., 2020). However, the role of Csf1ra/Csf1rb in regulating hematopoiesis at a higher hierarchy has not yet been identified.

In the present study, we propose an unexpected role of Csf1rb in regulating HSPCs in zebrafish. csf1rb, but not csf1ra, was detected in HSPCs. Loss of function of Csf1rb, but not of Csf1ra, led to impaired maintenance of HSPCs during both early developmental stages and adult hematopoiesis. Consequently, this caused multiple mature blood-lineage defects. Surprisingly, the hematopoietic effects of Csf1rb did not rely on any of the dedicated ligands, including Csf1a, Csf1b and Il34, suggesting the existence of additional, as-yet-unknown signals for HSPC development.

To explore the role of Csf1r during early hematopoiesis in zebrafish, we determined the expression pattern of two zebrafish CSF1R orthologs, csf1ra and csf1rb, at the embryonic stage. Consistent with previous reports, whole-mount in situ hybridization (WISH) showed that csf1ra was exclusively detected in mpeg1⁺ macrophages and neural crest-derived xanthophores (Fig. S1D-F) (Herbomel et al., 2001; Ellett et al., 2011; Wu et al., 2018). Likewise, csf1rb was also detected in macrophages, but the number of csf1rb-positive cells was lower than that of csf1ra-positive cells (Fig. S1A-C). Surprisingly, csf1rb was also found to be expressed in non-macrophages in the VDA region at 2 dpf (Fig. S1B), and csf1rb but not csf1ra transcripts were highly enriched in the VDA region at 30-36 hpf (Fig. S1G), when definitive HSPCs began to emerge from hemogenic endothelium (HE) in the VDA (Kissa and Herbomel, 2010). This observation coincided with that reported by Ferrero et al. (2020). We suspected that csf1rb was expressed in the HE and newly formed HSPCs. Indeed, the co-staining analysis showed that csf1rb was co-expressed with the endothelium marker flk1 (kdrl) and HSPC marker cmyb (myb) at 30-36 hpf (Fig. 1A,B) (Choi et al., 2007; Ellett et al., 2011; North et al., 2007). Consistent with this notion, a substantial portion of the csf1rb-positive cells in the VDA of 2 dpf embryos co-expressed the HSPC markers cmyb and runx1 (Fig. 1C,D) (Bertrand et al., 2008; Gering and Patient, 2005; Lam et al., 2010; Zhang et al., 2011), but not mpeg1 (Fig. S1B), although not all cmyb⁺ or runx1⁺ cells displayed csf1rb expression. Collectively, these data demonstrate that csf1rb is enriched in HE and HSPCs at the embryonic stages of zebrafish development. This conclusion was further supported by the observation that csf1rb expression was completely abolished in the VDA and CHT in runx1-deficient mutant fish (Fig. S1H), in which the formation of HSPCs was largely abrogated (Jin et al., 2009).

The enriched expression of csf1rb in HE and HSPCs implies that it may play an important role in the top hierarchy of definitive hematopoiesis. To address this question, we generated a csf1rb-deficient zebrafish mutant, csf1rb^hkz15 (called csf1rb mutant in the following text; siblings referred to wild-type and heterozygous fish), with the CRISPR-Cas9 system. The mutant csf1rb genome harbored a 2333 bp fragment deletion from intron 9 to exon 13 (Fig. S2A), and the relative expression level of csf1rb in csf1rb mutants was much lower than that in siblings (Fig. S2B). To test whether loss of Csf1rb function would perturb HSPC formation, we first performed time-lapse imaging to track the process of EHT, during which the HE sprout from the aorta and turn into HSPCs (Kissa and Herbomel, 2010; Tian et al., 2017). To monitor and quantify the events of EHT, we utilized Tg(flk1:NLS-Eos) fish (Fukuhara et al., 2014), in which endothelial cells express the photo-convertible nuclear localization signal (NLS)-tagged Eos fluorescence protein (NLS-Eos). Upon UV light irradiation, Eos switches from green fluorescence to red fluorescence (Fig. 2A). At 28-30 hpf, we converted nine endothelial cells in the VDA (one per somite) in each embryo and followed their behaviors for approximately 26 h with time-lapse imaging (Fig. 2A, lower panels, Movies 1-2). The results showed that the budding number of converted endothelial cells was comparable between csf1rb mutants and siblings (3.3 cells versus 3.8 cells on average per fish) (Fig. 2B). Consistent with the results of time-lapse imaging, the expression of cmyb, a well-known HSPC marker (Bertrand et al., 2008; Xia et al., 2021; Xue et al., 2017; Zhang et al., 2011), in csf1rb mutants was similar to that in siblings from 30 hpf to 2 dpf (Fig. 2C-E). These data indicate that Csf1rb is not required for the specification and budding of HSPCs during early development. However, as the mutant fish developed to 3 dpf, when most of the HSPCs accumulate in the CHT (Jin et al., 2007), the number of cmyb⁺ cells in the CHT region was severely reduced (Fig. 2F,G). Consistent with this, another HSPC-enriched marker, runx1 (Gering and Patient, 2005; Tamplin et al., 2015; Xue et al., 2017), also displayed a similar reduction in csf1rb mutants (Fig. 2H,I). To confirm the HSPC defects in csf1rb mutants, we further checked cmyb⁺ pu.1 (spi1b)⁻ cells (non-myeloid HSPCs) in csf1rb siblings and mutants within the Tg(cmyb:GFP) line (North et al., 2007), to exclude the possibility that the reduction of cmyb⁺ cells in csf1rb mutants could be attributed to myeloid cell defects (Ferrero et al., 2020). Our results showed that a significant portion of cmyb-GFP positive cells were pu.1 negative, and not only pu.1⁺ cells (myeloid cells) but also cmyb⁺pu.1⁻ cells (non-myeloid HSPCs) showed a significant decrease in csf1rb mutants (Fig. S3A-D). We also examined downstream blood cell markers in csf1rb mutants at 5-6 dpf. Unlike primitive erythrocytes, which only retain AE1-globin protein but decrease αe1-globin (hbae1.1) RNA in the cell, and mainly exist in the circulation, definitive erythrocytes maintain αe1-globin RNA expression and locate in the CHT at 5 dpf (Fig. S4) (Jin et al., 2009). We found that not only αe1-globin but also βe2-globin (hbbe2) (Ganis et al., 2012) and Gata1 (Monteiro et al., 2011) were reduced in csf1rb mutants at 5-6 dpf (Fig. 2J,K, Fig. S3E-H), suggesting defects in definitive HSPC-derived erythroid cells. We also observed a substantial reduction of Lcp1⁺ myeloid cells (Meijer et al., 2008) and lck⁺/rag1⁺ T cells (Langenau et al., 2004) in csf1rb mutants at 5 dpf (Fig. 2L-O, Fig. S3I-J). Considering T cells at early larval stages are mixed origins from both non-HSC progenitors and HSCs (Tian et al., 2017; Ulloa et al., 2021), we further checked T cells after 40 dpf, when T cells in the thymus are originated purely from HSCs (Tian et al., 2017). It has been demonstrated that a coro1a-driven reporter can mark T cells in the thymus region (He et al., 2020; Xue et al., 2021), and we found that the coro1a⁺ thymus area in csf1rb mutants at 45 dpf was significantly smaller than that in siblings (Fig. S3K-M). These results suggest definitive hematopoiesis failure in csf1rb mutants. The HSPC defect was confirmed in the compound heterozygous mutants of csf1rb by crossing our csf1rb mutant with the published csf1rb^re01 mutant (Fig. S5) (Oosterhof et al., 2018), demonstrating that the observed phenotypes were a consequence of csf1rb deficiency. Interestingly, the development of primitive myeloid and erythroid lineages was largely unaffected in csf1rb mutants (Fig. S6A-C), and primitive microglia only displayed a slight reduction (Fig. S6D and E), compared with a near-complete depletion of primitive microglia in csf1ra mutants (Oosterhof et al., 2018; Wu et al., 2018). These results suggest that, in contrast to the important role of Csf1ra in primitive myeloid cell development (Wu et al., 2018), Csf1rb appears to play a crucial role in definitive HSPC development. This conclusion was further supported by the findings that csf1ra deficiency had no obvious effect on HSPC development and csf1ra/b double deficiency did not cause a more severe HSPC phenotype (Fig. S7). Taken together, these data show that Csf1rb is essential for HSPC development at the early stages of zebrafish development.

To examine the cellular basis underlying the defect of HSPCs in csf1rb mutants, a terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay was performed to monitor the death of HSPCs in the Tg(cmyb:GFP);csf1rb transgenic mutant line. TUNEL staining showed that there was no excessive death of cmyb-GFP⁺ cells in csf1rb mutants (Fig. S8), suggesting that the reduction of HSPCs was not caused by apoptotic cell death. To support this argument further, we performed time-lapse imaging of the Tg(cmyb:GFP);csf1rb transgenic mutant line to trace the fate of cmyb-GFP⁺ HSPCs in the VDA and CHT from 2.5 dpf to 4 dpf, when cmyb⁺ HSPCs were significantly lower in the mutant embryos (Fig. 2F,G). Consistent with the TUNEL assay results, time-lapse imaging revealed no excessive death of cmyb-GFP⁺ HSPCs in csf1rb mutants (Movies 3,4). These data indicate that the reduction of HSPCs in csf1rb mutants cannot be attributed to excessive cell death.

We therefore speculated that the reduction of HSPCs in csf1rb mutants could be due to a defect in cell proliferation. We performed a bromodeoxyuridine (BrdU) incorporation assay to examine the proliferation of HSPCs in the Tg(cmyb:GFP);csf1rb transgenic mutant line. The results showed that not only the cmyb-GFP⁺ HSPC number but also the percentage of BrdU⁺ proliferating HSPCs were significantly reduced in the CHT region of csf1rb mutants compared with that in siblings (Fig. 3, Fig. S9), demonstrating the important role of Csf1rb in regulating the expansion of HSPCs. Reporter lines driven by the coro1a promoter have been demonstrated to faithfully label HSPCs/hematopoietic cells in the VDA and CHT region at embryonic stages (Huang et al., 2019; Li et al., 2012); thus, we also utilized the Tg(coro1a:kaede) line to monitor the proliferation of coro1a⁺ HSPCs/hematopoietic cells in the VDA region of csf1rb mutants, in which these cells could be converted from green to red by 405 nm UV light (Mizuno et al., 2003; Xu et al., 2016). Twenty to thirty coro1a⁺ cells in the VDA were converted at about 2.5 dpf and their proliferation was traced for 20 h by time-lapse imaging (Fig. S10A,B). The results showed that the percentage of dividing coro1a⁺ HSPCs/hematopoietic cells in the VDA of csf1rb mutants (35%) was significantly lower than that in siblings (52%) (Fig. S10C). Taken together, Csf1rb is essential for the expansion of HSPCs, and the absence of Csf1rb results in a severe reduction of HSPCs and their downstream blood lineages.

As csf1rb-deficient mutants can survive to adulthood, we next investigated whether Csf1rb also plays a similar role in adult hematopoiesis. Blood cells were collected from the peripheral circulation and whole kidney marrow (WKM), the adult hematopoietic organ in zebrafish (Traver et al., 2003), from adult (3-8 months) csf1rb mutants and siblings and subjected to flow cytometry. We examined the blood lineages utilizing Tg(globin:GFP) and Tg(coro1a:DsRedx), in which erythroid cells and leukocytes/progenitors, respectively, are labeled (Huang et al., 2019; Li et al., 2012; Tian et al., 2017). We found that the coro1a⁺ cells in the WKM were significantly reduced in adults with csf1rb deficiency (12.7% of the WKM cells in mutants versus 34.8% in siblings) (Fig. 4A,B, Fig. S11A). By contrast, the percentage of the globin⁺ erythroid cells in the WKM were comparable between csf1rb siblings and mutant (Fig. 4C,D, Fig. S11B). To confirm the normal erythrocyte population in adult csf1rb mutants, we took the same volume of peripheral blood from csf1rb-deficient adult fish and siblings to check the absolute cell number of globin-GFP⁺ erythroid cells using flow cytometry (Figs S11C and S12A). As expected, the number of erythroid cells in circulation was normal in csf1rb mutant fish (Fig. S12B). We also examined the morphology of erythrocytes and found no difference between csf1rb mutants and siblings (Fig. S12C). All of these results suggested that the shortage of erythroid population at embryonic stages stemming from csf1rb deficiency can be compensated for and recover with the growth of zebrafish. To test this possibility further, we examined the blood lineages at earlier stages utilizing Tg(globin:GFP) and Tg(coro1a:DsRedx). The results showed that in csf1rb mutant fish, the percentage and absolute cell number of globin⁺ erythroid cells, but not coro1a⁺ cells, recovered to the normal level at as early as 11 dpf (Figs S11D, S12D-G). lck⁺ T lymphocytes (Tian et al., 2017) in the thymus remained defective in csf1rb mutants at 11 dpf (Fig. S12H).

To investigate further the HSPC defects in csf1rb mutant adult fish, we injected pu.1 morpholino (MO) into runx1 mutant embryos to deplete all hematopoiesis-derived immune cells and obtain immune-deficient embryos (Kissa and Herbomel, 2010; Rhodes et al., 2005; Fraint et al., 2020). We transplanted the same dosage of adult WKM cells (on average 400 cells per fish) from csf1rb mutants or siblings with Tg(globin:GFP) and Tg(coro1a:DsRedx) into immunodeficient embryos at 2 dpf and raised these fish into adults to determine the donor contribution of leukocytes and erythrocytes with the globin-GFP and coro1a-DsRedx reporters (Fig. 5A). We found that four out of 13 fish were reconstituted by csf1rb sibling donor-derived hematopoietic cells, as shown by the DsRedx⁺ cells in the skin/kidney and GFP⁺ erythroid cells in the circulation (Fig. 5B-E, Fig. S11A). However, none of the 24 fish transplanted with csf1rb mutant WKM cells was grafted by donor-derived blood cells (Fig. 5B-E, Fig. S11A), indicating the impaired repopulating ability of adult HSPCs in csf1rb mutant fish. These data suggest that Csf1rb deficiency also attenuates the hematopoietic capacity of adult HSPCs.

In mammals, CSF1 and IL34 serve as the ligands for CSF1R during the regulation of myeloid cell development (Stanley and Chitu, 2014). In zebrafish, the Il34-Csf1ra axis has been demonstrated to mediate microglia seeding in the brain (Kuil et al., 2019; Wu et al., 2018), and Csf1a/b-Csf1ra regulate xanthophore cell development and pattern melanocytes to direct stripe formation (Parichy and Turner, 2003; Patterson and Parichy, 2013; Wu et al., 2018). Recently, Il34 was reported to regulate granulopoiesis via Csf1rb (Hason et al., 2022). We wondered which ligand functioned via Csf1rb in regulating HSPC maintenance, so we examined the cmyb⁺ HSPCs in csf1a, csf1b and il34 mutant fish (Wu et al., 2018). However, no defects in HSPCs were observed in these ligand mutants (Fig. 6A,B). Considering possible compensation effects between these three ligands, the phenotypes of double mutants and triple mutants were examined. Similar to the single mutants, HSPCs were normal in all of the mutants (Fig. 6A,B). However, the csf1a^−/− csf1b^−/− fish fully recapitulated the pigmentation defects of csf1ra^−/− mutant fish (Fig. S13A) (Parichy and Turner, 2003), indicating total loss of function of both ligand forms. Il34 mutant fish have also been demonstrated to have defective microglia (Wu et al., 2018). To test whether these ligands work as the ligand of Csf1rb, we overexpressed each ligand utilizing the Tg(hsp70:csf1a), Tg(hsp70:csf1b) and Tg(hsp70:il34), in which the expression of the ligand was activated after heat shock. The embryos were heat shocked at 2, 2.5 and 3 dpf for 1 h, respectively, and cmyb⁺ HSPCs were evaluated at 3.5 or 4 dpf. The results showed that ectopic expression of csf1a or csf1b increased the population of xanthophore cells on the surface of the embryos (Fig. S13B,C), and overexpression of il34 augmented macrophages in the embryos (Fig. S13D,E), demonstrating gain of function of these ligands after heat shock. However, we found that overexpression of Csf1a or Csf1b or Il34 was insufficient to boost HSPCs, although they could expand xanthophores and macrophages (Fig. 6C-E). In conclusion, the above data demonstrate that Csf1a, Csf1b and Il34, the canonical ligands for Csf1r in zebrafish, are not likely to be the ligands of Csf1rb in regulation of hematopoiesis.

In the present study, we provide the first genetic evidence to suggest that Csf1r can function at a higher hematopoietic hierarchy. We reveal a crucial role of Csf1rb in preserving the HSPC pool by maintaining the expansion of HSPCs at embryonic stages. We also found that Csf1rb regulates the reconstitution ability of adult HSPCs, which was attenuated by csf1rb deficiency. Interestingly, the impaired proliferation of HSPCs in csf1rb-deficient embryonic fish led to defects in multiple lineages, including erythrocytes, myelocytes and lymphocytes (Figs 2 and 3, Fig. S3). In the adult fish kidney, hematopoietic defects remained severe, but erythroid cells recovered (Fig. 4, Fig. S12A-C).

Csf1r is regarded as a major regulator of myeloid cell development in vertebrates. Here, our data reveal the function of Csf1rb in the higher hierarchy during hematopoiesis. Some reports have suggested that CSF1R may function in HSCs or progenitors in mammals. For instance, CSF1R has been detected in HSPCs (Miyamoto et al., 2002). In Csf1^op/op or Csf1r mutant mice or rats, both blood lymphocytes and monocytes were decreased (Dai et al., 2002; Keshvari et al., 2021; Pridans et al., 2018). However, no detailed analysis of the populating ability of HSPCs, such as a transplantation assay with HSPCs from CSF1R signaling-deficient mice, has been performed, and most of the data is based on the colony formation assay that may not be able to reveal the function of CSF1R in HSPCs. Recently, it has been reported that PLX5622, a chemical inhibitor presumably targeting CSF1R, can reduce HSC and hematopoietic progenitors in adult mouse bone marrow (Lei et al., 2020). On the contrary, abolishing Csf1r expression in the bone marrow via deletion of a csf1r enhancer shows no effects on hematopoietic progenitors (Rojo et al., 2019). Therefore, whether CSF1R in mouse or human is similar to Csf1rb in zebrafish in regulating HSPC maintenance remains unclear and needs further exploration.

The expression of csf1rb in HSPCs (Fig. 1) and HSPC defects in csf1rb mutants (Fig. 2F-I, Fig. S3A-C) strongly suggest that T-cell defects in csf1rb mutants are related to the cell-autonomous effect of csf1rb in HSPCs. Yu et al. reported that in pu.1/spi1b double-deficiency zebrafish larvae, macrophages were severely disrupted before 8 dpf, but the rag1⁺ T cells at 5 dpf were not affected (Yu et al., 2017). Faiza et al. showed that csf3r mutation caused a significant reduction of neutrophils, but the rag1⁺ T cells at 5 dpf were normal (Basheer et al., 2019). These studies suggest that myeloid cell defects have little impact on T-cell development in early zebrafish larvae. Therefore, the T-cell defects in the csf1rb mutant were unlikely an indirect consequence of the myeloid cell defects.

The BrdU assay is a well-established method to detect cell proliferation. BrdU incorporation occurs when the synthesis of DNA is ongoing. Usually, a short duration of BrdU application labels S-phase cells, whereas a longer duration labels both proliferating and proliferated cells. We performed BrdU labeling over a short time (within 1 h), which is not long enough for normal cells to complete a whole cell cycle. Therefore, the 0.5 ratio may not mean that half of all cmyb:GFP⁺ cells proliferated within 1 h of BrdU labeling. Instead, it suggests that about half of all cmyb:GFP⁺ cells were undergoing DNA synthesis within this short incubation time. This HSPC proliferation rate was similar to the BrdU⁺ incorporation ratio of HSPCs reported in other studies at around 2-3 dpf, when the BrdU incorporation was performed within 1-3 h (Xia et al., 2021; Xue et al., 2017; Zhang et al., 2011).

A recent publication by Ferrero et al. about zebrafish csf1rb mutants concluded that csf1rb deficiency predominantly affected myeloid and B-lineage development (Ferrero et al., 2020). These results are within the scope of the previous understanding of Csf1r functions in downstream hematopoietic lineages such as myeloid cells. In contrast, our findings clearly showed that HSPCs were impaired in csf1rb mutants. This conclusion is supported by three different experiments: the decrease of definitive hematopoietic cells in embryonic/adult csf1rb mutants, the impaired definitive hematopoiesis in a complementary test with two different csf1rb mutant alleles, and the failure of hematopoietic reconstitution after transplantation with the kidney marrow cells from csf1rb mutants. The difference between Ferrero's and our findings may be attributed to the use of different csf1rb mutant alleles. Ferrero's csf1rb mutant allele led to a splicing defect and an additional isoform of csf1rb was observed in Ferrero's study. Therefore, Ferrero's csf1rb mutant allele may not fully abolish csf1rb functions. In contrast, four exons were deleted in our csf1rb mutant allele to ensure the full loss of csf1rb functions.

Surprisingly, the phenotype for erythroid lineages in csf1rb mutant fish was not consistent between embryonic stages and adulthood, as the defects of mature erythroid cells in csf1rb mutant fish recovered in the early juvenile to adult stage (Fig. 4C,D, Fig. S12). This phenotype is in accordance with normal red blood cells in Csf1r mutant rats aged between 6 and 12 weeks (Dai et al., 2002; Pridans et al., 2018), but the erythroid lineage phenotype in embryonic mutant mice or rats was not reported in the previous study. We examined the expression of csf1rb in the Atlas for Zebrafish Development database from the UCSC cell browser (https://cells.ucsc.edu), which contains the single-cell RNA sequencing data of 1, 2 and 5 dpf zebrafish embryos. It turned out that csf1rb was also detected in mature myeloid cell and lymphoid cell clusters, but not in mature erythroid cells (Fig. S14; Speir et al., 2021). In addition, we have shown that csf1rb depletion did not interfere with the development of primitive erythroid cells, indicating that Csf1rb may not be required for the maturation of erythroid cells from gata1⁺ committed precursors, but is required for the downstream development of myeloid and lymphoid cells. Some studies have suggested that under certain conditions, such as bone marrow failure, tissue inflammation, abnormal systemic chemokine production, and especially severe hematologic disorders, compensatory reactivation occurs to fulfill the need for blood cell production by increasing the maturation of downstream lineages (Johns and Christopher, 2012). Thus, we speculate that the deficiency of erythroid lineages at early stages may lead to hypoxia. As the maturation of myeloid and lymphoid cells depends on Csf1rb, but erythroid cell maturation does not rely on Csf1rb, the compensatory hematopoiesis compromises erythroid but not myeloid and lymphoid defects over time.

Zebrafish orthologs of dedicated ligands for CSF1R in mammals, including Csf1a/Csf1b and Il34, have been demonstrated to work through Csf1ra to regulate xanthophore cell development or microglial colonization in the brain, respectively (Kuil et al., 2019; Parichy and Turner, 2003; Patterson et al., 2014; Wu et al., 2018). Surprisingly, none of these ligands is likely to be the ligand for Csf1rb in regulating HSPCs, as suggested by the loss-of-function and gain-of-function studies (Fig. 6). Cytokines, such as IL3, IL5, CSF2, and the ligand of Flt3 are involved in hematopoiesis in mammals. However, they have not been identified in zebrafish (Bartunek et al., 2018). Interestingly, the whole gene cluster for IL3, IL5 and CSF2 is depleted in the zebrafish genome, and the receptor genes were also not identified. The function of these genes may be compensated for by other regulators (Liongue and Ward, 2007). Recently, the Flt3 receptor was identified in zebrafish and was shown to be important for the development of definitive HSPCs (He et al., 2014). However, the Flt3 ligand in zebrafish is still unknown. The poor sequence homology of cytokines between zebrafish and mammals has made it challenging to identify new cytokine orthologs in zebrafish. Therefore, the ligand for Csf1rb may be an unknown gene. Alternatively, other known cytokines may work through Csf1rb to regulate HSPCs in zebrafish. Different cytokines may have the ability to activate the same receptor to exert similar functions (Ozaki and Leonard, 2002). For instance, CXCL9, CXCL10 and CXCL11 share the receptor CXCR3 and mediate overlapping effects in T cells and intestinal myofibroblasts (Campanella et al., 2008; Kouroumalis et al., 2005). CSF2, IL3 and IL5 regulate the differentiation and function of eosinophils via a common β receptor subunit during airway inflammation (Asquith et al., 2008). The redundancy of cytokines reminds us that multiple cytokines may function through Csf1rb to regulate HSPC maintenance, and they are not limited to Csf1/Il34, which may explain why both gain of function and loss of function in Csf1/Il34 resulted in no obvious HSPC phenotype. Other cytokines that have been reported in the hematopoietic system may participate in the regulation of HSPCs via Csf1rb. Further study is required to address the ligand for Csf1rb in zebrafish.

All zebrafish lines were raised and maintained according to standard protocols (Westerfield, 2000). In this study, ABSR wild type, Tg(flk1:mCherry) and Tg(flk1:GFP) (Choi et al., 2007), Tg(cmyb:GFP) (North et al., 2007), Tg(runx1:en-GFP) (Zhang et al., 2015), Tg(mpeg1:GFP), Tg(flk1:NLS-Eos) (Fukuhara et al., 2014), Tg(coro1a:DsRedx) (Xu et al., 2016), Tg(coro1a:Kaede) (Zhan et al., 2018), Tg(globin:GFP), Tg (globin:loxP-DsRedx-loxP-GFP), Tg (lck:loxP-DsRedx-loxP-GFP) (Tian et al., 2017), Tg(hsp70:csf1a), Tg(hsp70:csf1b), Tg(hsp70:il34), Runx1^w84x (Jin et al., 2009), csf1ra^j4e1(Herbomel et al., 2001), csf1rb^re01 (Oosterhof et al., 2018), csf1a^hkz9, csf1b^hkz10, il34^hkz11 (Wu et al., 2018) and csf1rb^hkz15 were used.

All animal experiments were performed under the approval of the Animal Ethics Committees of the Hong Kong University of Science and Technology and South China University of Technology.

The csf1rb mutant fish line was generated by the CRISPR-Cas9 method following previous reports (Chang et al., 2013; Wu et al., 2018). The csf1rb mutant was identified with primers 5′-GTTCACCATACCAATGTGGC-3′ and 5′-AGCATTCACAGACTGGCTGA-3′ for the mutant allele, 5′-ATTGACCCCAGCCAACTTCC-3′ and 5′-TAGCGTAGCCTAGAGGCCTA-3′ for the wild-type allele.

To generate Tg(hsp70:csf1a), Tg(hsp70:csf1b) and Tg(hsp70:il34), the promoter of zebrafish heat shock protein 70 (hsp70) was constructed into the pBSK vector with minimal Tol2 transposon sequence and an SV40 polyA sequence, and the cDNA of csf1a, csf1b or il34 was cloned under the control of the promoter. Plasmid injection and stable line screening were performed as previously described (Wu et al., 2018).

WISH of csf1ra, csf1rb, cmyb, runx1, pu.1, mfap4, lck, rag1, gata1, αe1-globin and βe2-globin was performed as previously described (Wu et al., 2018).

Immunofluorescent antibody staining was conducted following previous studies (Dai et al., 2016). The primary antibodies utilized in this study were: rabbit anti-DsRedx antibody (632496; Clontech; 1:400), goat anti-GFP antibody (ab6658; Abcam; 1:400), rabbit anti-Gata1 (ab210539, Abcam; 1:100), mouse anti-BrdU (1117037600, Roche; 1:50), rabbit anti-Lcp1 (Jin et al., 2009; 1:400) and rabbit anti-AE1-globin (Du et al., 2011; 1:400). Secondary antibodies used in this study were: Alexa 488 anti-goat antibody (A11055, Invitrogen), Alexa 555 anti-rabbit antibody (A31572, Invitrogen), Alexa 488 anti-rabbit antibody (A21206, Invitrogen), Alexa 555 anti-mouse antibody (A31570, Invitrogen). All of the images were taken with Leica SP8 or Zeiss LSM800 confocal microscope and the quantification was conducted manually.

The BrdU-based proliferation assay was performed as previously reported (Dai et al., 2016). In brief, embryos were incubated with BrdU for 30 min followed by 30 min recovery before fixation. Before 2 N HCl treatment, the head region was cut for genome typing. Embryos were separated according to genome types and stained with anti-BrdU and anti-GFP antibodies simultaneously.

A cell death assay based on TUNEL was performed as previously described (Dai et al., 2016). Neutral Red staining was conducted as previously described (Dai et al., 2016).

Photo-conversion was conducted with a 405-nm UV laser as previously described (Tian et al., 2017). For the HSPC budding assay, nine endothelia in the VDA for each fish were converted with one per somite at 28-30 hpf. Live imaging was performed immediately for about 26 h with a time interval of 10.8 min. For cell conversion in the Tg(coro1a: Kaede) line, kaede⁺ cells from the VDA region were converted at 2.5 or 3 dpf, respectively. Live imaging of the VDA region was conducted right after conversion for overnight. Live imaging of HSPCs in the Tg(cmyb:GFP) line was captured from 2.5 dpf for about 26 h with a time interval of 4 or 6 min. All time-lapse imaging was captured using a Leica SP8 confocal microscope.

Heat shock was performed in a 37°C water bath for 1 h at 2, 2.5 and 3 dpf.

Flow cytometry was performed with BD FACS Aria III, and data were analyzed on BD FACS Aria III or FlowJo. Adult fish kidney marrow cell suspension was prepared according to a previous study (Tian et al., 2017). The concentration of adult whole kidney marrow cell suspension was determined using a hemocytometer, and 2.4×10⁵ cells were prepared in 2 μl injection medium containing 1.8 U heparin (Sigma-Aldrich) and 0.4 U DNaseI (QIAGEN) in 5% fetal bovine serum (FBS) in PBS; 3.5 nl of the cell solution was injected into the circulation of 2 dpf runx1 mutant embryos injected with pu.1 morpholino (Gene Tools).

For 11 dpf juvenile fish, fish were anesthetized and killed on ice, then two fish were grouped in 5% FBS in PBS and homogenized with a 24 or 27 gauge needle. The cell suspension was filtered through a 40-μm nylon mesh before flow cytometry analysis (FACS Aria III; BD Biosciences). In addition, 40 μl adult peripheral blood was collected from each fish into 500 μl 5% FBS/PBS for FACS analysis and 10 μl peripheral blood was collected in 100 μl 3 U/μl heparin in 5% FBS/PBS to perform May–Grunwald and Giemsa staining as previously described (Lin et al., 2019).

Tg(coro1a:DsRedx) fish at 45 dpf were anesthetized and body length measured. The skin and bone covering the thymus were removed and imaged under a Zeiss LSM800 confocal microscope. The thymus area indicated by rag1/lck expression at 5 dpf or by coro1a-DsRedx expression at 45 dpf was manually marked according to the signal and defined automatically using ZEISS ZEN microscope software.

After heat-shock treatment, wild-type control fish and Tg(hsp70:csf1a) or Tg(hsp70:csf1b) fish were collected and anesthetized, and imaged using a Zeiss AXIO Zoom.V16 microscope. The surface region of each individual fish was marked with same size and location, and the mean intensity value of the bright channel within the marked region was determined automatically using ZEISS ZEN microscope software.

The relative expression level of csf1rb in csf1rb siblings and mutants were determined by QPCR. In brief, 4 dpf individual embryos were subjected to lysis, and, after genotyping, sibling and mutant embryos were pooled together for total RNA extraction with the RNeasy mini kit (QIAGEN, 74104), and cDNA was obtained using the SuperScript II reverse transcription kit (Thermo Fisher Scientific, 18064022). The expression of csf1rb was normalized to ef1a (eef1a1l1). The primers for csf1rb were 5′-GATGCTCTGTGTCGCTCTCAT-3′ and 5′-TCTGCGCCTGGTCTTCCATA-3′; primers for ef1a were 5′-CTTCTCAGGCTGACTGTGC-3′ and 5′-CCGCTAGCATTACCCTCC-3′.

All the statistical analysis was performed using GraphPad Prism version 8 with unpaired Student's t-tests, and two-tailed P-values were used for all t-tests.

We would like to thank Prof. Tjakko J. van Ham from the University Medical Center for sharing the csf1rb^re01 zebrafish mutant line, and Prof. Feng Liu from Chinese Academy of Science for sharing the Tg(runx1:en-GFP) fish. We thank International Science Editing (https://www.internationalscienceediting.com/) for editing this manuscript.