Trim33 conditions the lifespan of primitive macrophages and onset of definitive macrophage production

Wandering within most tissues from early on in ontogeny, tissue-resident macrophages are often pictured as sentinels, always on alert, with the mission to continually detect and adequately react to any fluctuations within the surrounding tissues, from the small and natural consequences of proper organ functioning to extreme pathological/inflammatory aggressions. These roles require a high level of transcriptional versatility, to provide macrophages with a wide range of possible reactions and quick reactivity. As it is thought that, at least in mammals, tissue-resident macrophage populations maintain themselves throughout life, they also need robust self-renewal capacities and a long lifespan. The transcriptional toolbox necessary for macrophages to face such diverse tasks for such an extended period of time is still poorly known.

Macrophages arise in successive waves in vertebrate ontogeny – first in the yolk sac and later within the developing larva or fetus as one of the differentiation outcomes of the hematopoietic stem/progenitor cells (HSPCs), in the successive niches where so-called ‘definitive’ hematopoiesis takes place. In zebrafish development, the initial – or ‘primitive’ – wave of macrophages originates from the anterior-most lateral mesoderm. By mid-somitogenesis, myeloid progenitors arise directly from this mesoderm and migrate into the adjacent yolk sac to differentiate into primitive macrophages and neutrophil precursors, which then quickly invade the embryo's mesenchyme and then the epidermis. Only the macrophages further colonize the brain and retina, to become the primitive microglia (Herbomel et al., 1999, 2001; Le Guyader et al., 2008).

Definitive hematopoiesis begins in the embryo with the emergence of HSPCs from the endothelium of the ventral wall of the aorta in the trunk (Kissa and Herbomel, 2010). These HSPCs then home to a first niche in the tail, the caudal hematopoietic tissue (CHT), where they expand and undergo multi-lineage differentiation, including in lymphoid cells, and later home to the kidney, the niche of lifelong hematopoiesis (Murayama et al., 2006; Kissa et al., 2008). In fish, the CHT then the ‘kidney marrow’ are thus the hematopoietic niche homologs of the fetal liver then bone marrow in mammals.

The transcription co-factor Trim33/Tif1γ has been known for years to be an important actor of hematopoiesis. It regulates transcription by binding to various DNA-binding transcription factors, such as SMADs, PU.1 and Scl/Tal1 (Ferri et al., 2015; Kusy et al., 2011; Xi et al., 2011). In zebrafish, it was first identified as essential cell-autonomously for the differentiation of the primitive erythrocytes from the embryo mesoderm (Ransom et al., 2004). moonshine (trim33) mutant embryos, which harbor a null mutation in Tif1γ/Trim33, appear ‘bloodless’, as the primitive erythroid progenitors undergo early apoptosis by mid-somitogenesis. Monteiro et al. (Monteiro et al., 2011) later found that the definitive, HSPC-derived erythropoiesis normally occurring in the CHT is also defective in moonshine swimming larvae, while granulopoiesis is enhanced. In mammals, the hematopoietic role of Trim33/Tif1-γ has only been studied in postnatal/adult mice. Trim33-deficient HSPCs showed signs of premature aging (Quéré et al., 2014) and an increased capacity to generate myeloid progenitors, at the expense of other lineages (Aucagne et al., 2011). The generated myeloid progenitors showed a diminished capacity to differentiate in macrophages (Chrétien et al., 2016; Gallouet et al., 2017).

More recently, Trim33 has been shown to be necessary for several essential functions of differentiated macrophages. In mice, Romeo and co-workers found that Trim33-deficient macrophages were unable to resolve inflammation (Ferri et al., 2015; Gallouet et al., 2017; Petit et al., 2019). In the zebrafish, we found that in moonshine embryos, primitive macrophages and neutrophils were produced, and then dispersed throughout the interstitium of the embryo, but they displayed reduced basal mobility and were unable to respond to developmental or inflammatory recruitment signals. This notably led to an absence of primitive microglia (Demy et al., 2017).

We now report that over the next 2 days of development of moonshine larvae, the primitive macrophages (but not the neutrophils) prematurely disappear. In addition, the first wave of definitive macrophage production from aorta-derived HSPCs is missing, while the production of HSPC-derived neutrophils and thrombocytes is enhanced.

When proceeding with the characterization of the myeloid phenotype of our own moonshine (Trim33-null mutant) allele, mon^NQ039 (Demy et al., 2017), we discovered that after the strong navigation defect displayed by both primitive macrophages and neutrophils by 2-3 days post-fertilization (dpf), the macrophages (but not the neutrophils) had mostly disappeared altogether from the mutant larvae by 4 dpf (Fig. 1A; zebrafish embryos become ‘swimming larvae’ after hatching, by 2.5 dpf). We confirmed this previously unreported phenotype of moonshine by counting mCherry-positive (mCherry⁺) macrophages at 4 dpf in mon^NQ039 Tg(mpeg1:mCherry-F) larvae, both manually on live individual larvae (Fig. 1B) and by fluorescence-associated cell sorting (FACS) of pools of over 50 mutant larvae and their siblings (Fig. 1C). We had previously shown that moonshine mutant embryos have a normal number of macrophages and neutrophils at 48 h post-fertilization (hpf) (Demy et al., 2017). So we looked more closely at the kinetics of this macrophage disappearance by revealing the macrophage population of mon^NQ039 mutant embryos and their siblings by whole-mount in situ hybridization for csf1ra from 2 to 4 dpf (Fig. S1A), and we found that the disappearance of the mutant macrophages was very gradual, starting around 54 hpf. We obtained the same results via in vivo imaging of mon^TB222 Tg(mfap4:mCherry-F) larvae (Fig. 1D). Following up individual larvae over time and counting their macrophages at 2, 3 and 4 dpf confirmed the gradual disappearance of their macrophages (Fig. 1E).

To check whether this disappearance of primitive macrophages in moonshine mutants is premature, we used UV exposure to completely photoconvert the Kaede proteins of Tg(mpeg1:Gal4/UAS:Kaede) embryos from green to red at 2 dpf. Upon UV-driven photoconversion, all primitive macrophages become red (Fig. S1B). Over the next few days, new (green) Kaede protein was synthesized in these photoconverted macrophages, which now appeared yellow as they contained both green and red Kaede protein (Fig. S1B). Numerous such cells were detected up to 9 days post-photoconversion (Fig. 1F), meaning that the lifespan of wild-type primitive macrophages is at least 11 days and that the disappearance of the whole primitive macrophage population that we witness in moonshine mutants between 3 and 4 dpf is indeed very premature.

Further in vivo observations and time-lapse imaging between 54 hpf and 4 dpf revealed that several cellular characteristics of mon primitive macrophages are lost upon their progressive disappearance. By 3 dpf, although most differentiated interstitial macrophages in wild-type larvae have acquired an elongated and ramified morphology, macrophages at similar locations in mutant larvae appear more round and less ramified (Fig. 2A-C). It seems that they also tend to loose adhesion capacity, as many are detected circulating in the heart and blood vessels, carried away by the blood flow (Fig. 2D). High-magnification in vivo observations of macrophages in Tg(mfap4:mCherry-F) mon^TB222 mutants and their siblings at 3 dpf also revealed a reduction in fluorescence in these cells prior to their disappearance (Fig. 2E), both in cell area (Fig. 2F) and total intensity (Fig. 2G). Quantification of the fluorescence of mon^TB222 macrophages expressing both mfap4:mCherry-F (membrane-associated) and mfap4:turquoise (whole-cell reporter) transgenes in vivo (Fig. 2H,I; Fig. S2A,B) confirmed the global decrease of mCherry-F fluorescence in mutant macrophages, but also highlighted that this farnesylated fluorescent protein had mostly disappeared from the plasma membrane (Fig. 2I, black arrowheads) and persisted only at the center of the cell (Fig. 2I, purple arrowhead). These mCherry⁺ hot spots within the mutant macrophages colocalized with numerous vesicles that were highly refractile upon Nomarski imaging (Fig. 2J) and also acidic (Fig. 2K), suggesting that they are of phagocytic origin, unlike most of the mCherry⁺ intracellular hotspots of macrophages in sibling larvae (Fig. 2K). In vivo and time-lapse imaging of macrophages in moonshine^TB222 Tg(mfap4:turquoise) mutants between 3 and 4 dpf then allowed us to witness dead cells still expressing macrophage-driven fluorescence (Fig. 2L), and macrophages dying by bursting into several turquoise⁺ smaller pieces reminiscent of apoptotic bodies (Fig. 2M). Of note, macrophage death in vivo cannot be reliably assessed by the usual markers of apoptosis, for many of these professional phagocytes contain engulfed dead cell remnants that stain positive for these markers – hence our use of in vivo time-lapse imaging.

Observation of moonshine larvae later in development revealed that macrophages progressively reappeared in the mutants starting around 6 dpf (Fig. 3A,C; Fig. S3A), and that a full macrophage population was usually restored by 8.5 dpf (Fig. 3B). These macrophages seemed able to invade most larval tissues, including the brain where they settled to become microglia (Fig. 3D). To assess whether this later wave of macrophage production similarly occurred in wild-type larvae, we performed several experiments aimed at removing all primitive macrophages from the embryos and monitoring the (re)appearance of these cells in mon mutants and their siblings (Fig. S3B; Fig. 3E). First of all, we used two antisense morpholinos (Mo), targeting the transcription factors Pu.1 and Irf8, respectively, that have been shown to suppress primitive macrophage production, but not definitive (HSPC-derived) macrophages (Rhodes et al., 2005; Yu et al., 2017). Pu.1 Mo prevents all primitive myeloid cell formation (both primitive macrophages and neutrophils), whereas the Irf8 Mo disrupts the balance between these two myeloid cell fates, so that all progenitors become neutrophils. Both morpholinos resulted in a complete absence of macrophages at 48 hpf (Fig. S3B). We also took advantage of two techniques already used successfully in zebrafish to kill all macrophages at 48 hpf: clodronate-filled liposome injection in the blood flow (Bernut et al., 2014), and metronidazole (Mtz) treatment of Tg(mpeg1:Gal4; UAS:nfsB-mCherry) embryos, wherein macrophages specifically express a nitroreductase that converts Mtz into toxic metabolites (Davison et al., 2007; Palha et al., 2013). We then monitored macrophage reappearance over time in these different conditions of primitive macrophage depletion, both in mon^NQ039 mutants (Fig. 3E, left graph) and wild-type larvae (Fig. 3E, right graph). Macrophage reappearance in moonshine mutants was pretty homogeneous between conditions and consistent with our previous observations, starting by 6 dpf (Fig. 3E, left graph). In contrast, macrophage reappearance occurred much earlier in wild-type larvae, placing the beginning of larval (definitive) macrophage production at 4 dpf (Fig. 3E, right graph).

We then used a different approach, in which we did not manipulate primitive macrophage production or survival. We took advantage of the Tg(gata2b:Gal4; UAS:LifeAct-eGFP) line, in which eGFP specifically highlights aorta-derived HSPCs over the next few days, albeit somewhat mosaically (Butko et al., 2015). We crossed it with Tg(mpeg1:mCherry-F) and monitored in the resulting embryos the first appearance of double-positive (GFP⁺, mCherry-F⁺) cells, which would represent the first definitive (HSPC-derived) macrophages. At 4 dpf (Fig. 3F, upper panel), and as early as 3.5 dpf (Fig. S3C), definitive (mCherry⁺,GFP⁺) macrophages were detected in the CHT of wild-type larvae, whereas mon mutants were devoid of any macrophages at this stage (Fig. 3F, lower panel). At 4.5 dpf, single- and double-positive macrophages could be spotted in various tissues of wild-type larvae, including the thymic region (Fig. 3G, upper panel); macrophages were still absent from mon mutants at this stage (Fig. 3G, lower panel). Finally, starting at 6.5 dpf, numerous macrophages were detected in the thymic and kidney regions of mon mutants, and many of them were double labelled, evidencing their definitive (HSPC-derived) origin (Fig. 3H, lower panel; Movie 1). So whether or not we experimentally perturbed primitive macrophage production or survival, in either case the first definitive (HSPC-derived) macrophages arose by 3.5-4 dpf in wild type, and only by 6-6.5 dpf in Trim33-deficient mutants.

We have thus far established that in Trim33-deficient mutants, primitive macrophages die prematurely – between 3 and 5 dpf – and that, independently, the production of definitive macrophages is delayed by at least 2 days. To assess whether these effects of Trim33 deficiency on macrophages are cell-autonomous, we performed an embryonic parabiosis experiment (Demy et al., 2013), in which we fused wild-type Tg(mpeg1:eGFP-F) blastulas with mon^TB222 Tg(gata1:DsRed; mfap4:turquoise) blastulas (the gata1:DsRed transgene allowed us to determine whether or not the parabiote obtained from mon^tb222 heterozygous carrier parents was a moonshine homozygous mutant, i.e. producing no circulating Dsred⁺ primitive erythrocytes). Primitive macrophages originating from both parabiotes/genetic backgrounds were able to colonize, survive and co-exist in both parabiotes in wild-type and sibling (Fig. 4A) as well as in wild-type and mutant embryo combinations (Fig. 4B). We monitored mutant mon^TB222/mfap4:turquoise⁺ macrophages in the wild-type embryo (Fig. 4C, left panel) and wild-type mpeg1:eGFP-F⁺ macrophages in the mutant embryo (Fig. 4C, right panel) over time. At 2 dpf, both macrophage populations had been able to transcolonize and settle in the other parabiote. By 4 and 5 dpf, wild-type macrophages had survived and greatly expanded in the mutant parabiote, whereas mutant macrophages had gradually decayed in the wild-type parabiote. These observations demonstrate a cell-autonomous role for Trim33 in both the survival of primitive macrophages and the onset of definitive macrophage production.

In moonshine mutants, we further found that the delay in definitive macrophage production correlates with a higher level of definitive neutrophil production, as documented in vivo (Fig. 5A,B). Whole-mount Sudan Black staining confirmed that these overproduced neutrophils differentiate properly, as they develop fully stained mature granules, including in the kidney marrow niche (Fig. 5C).

In addition, from 3 dpf onwards, moonshine larvae recovered gata1:Dsred⁺ circulating cells, but it appeared that these cells were mostly not erythrocytes, but thrombocytes, as they were also CD41:eGFP⁺ (Fig. 5D,E). moonshine mutants produced a lot more thrombocytes than their siblings (Fig. 5D,E and Movie 2), from both the CHT and kidney marrow (Fig. 5F,G). This enhanced production of definitive neutrophils and thrombocytes was not due to the lack of primitive macrophages, as it did not occur in PU.1 morphant larvae, which lack primitive macrophages and neutrophils (Fig. S4). It also did not result from an inflammatory condition, as we detected no induction of the main inflammatory cytokines – TNFα, IL1β and Cxcl8 – by qPCR in moonshine mutants (Fig. S5). Altogether, these new observations suggest a disruption in the balance of HSPC-derived macrophage versus neutrophil and of erythrocyte versus thrombocyte production in moonshine mutants, thus uncovering a new perspective regarding the central role of Trim33 as a hematopoietic regulator.

In this study, we have found that in Trim33-deficient zebrafish, macrophages of the primitive lineage disappear prematurely, and that the production of new macrophages from the definitive hematopoiesis (DH) is delayed. The combination of these two features leads to a complete depletion of macrophages throughout the swimming larva by 4 and 5 dpf. Recent lineage-tracing studies have shown that primitive microglia, derived from primitive macrophages, still predominate in the zebrafish brain by 25 dpf and are then progressively replaced by DH-derived microglia over the following weeks (Ferrero et al., 2018; Xu et al., 2015). Similarly, the primitive macrophages that settle in the epidermis still predominate by 6 dpf, and are progressively replaced by DH-derived macrophages over the next week (He et al., 2018). Here, we have found that interstitial macrophages wandering in the mesenchyme are still present by 11 dpf.

We previously showed that in moonshine embryos, primitive macrophages and neutrophils disperse through the embryo, but are unable to navigate according to developmental or inflammatory signals. Notably, the macrophages do not colonize the brain and retina, a chemokine-driven process (Herbomel et al., 2001; Wu et al., 2018). Thus, both macrophages and neutrophils essentially remain in the interstitium between organs/tissues (Demy et al., 2017). In contrast, the later demise of primitive macrophages that we document in the present study does not apply to neutrophils. Early signs of this demise appeared to be a less elongated and ramified morphology, and a loss of expression of macrophage-specific genes – the endogenous mcsfr1 gene, and the mpeg1 and mfap4 promoter-driven reporter transgenes. Indeed, the fluorescence of farnesylated (hence membrane-targeted) mCherry expressed from the mpeg1 or mfap4 promoter faded rapidly, first disappearing from the plasma membrane while it was still detected in intracytoplasmic membrane compartments (where the mCherry-F signal is highest, both in wild-type and mutant macrophages). At the same time, the turquoise fluorescent reporter, expressed throughout the cell from the same promoter, was still detected uniformly within the mutant macrophages; this allowed us to document the subsequent apoptotic death using this reporter via time-lapse in vivo imaging. We suspect that the membrane-bound mCherry-F protein has a shorter half-life in macrophages due to the huge membrane turnover that characterizes these cells. Even though the mCherry-F signal is highest in intracellular membrane compartments in macrophages of both mutant and wild-type larvae, these compartments seem to have distinctive features in the mutant macrophages; they appear to mostly coincide with numerous highly refringent and acidic particles, suggesting they could possibly result from the earlier extensive efferocytosis of the dead primitive erythroid progenitors. Indeed, we know that before their demise, the Trim33-deficient primitive macrophages are phagocytically competent and can handle the elimination of the primitive erythroid progenitors that undergo apoptosis during somitogenesis (Demy et al., 2017). Yet this possibly challenging task early in macrophage development cannot be the only cause of their later premature death. Indeed, in our parabiosis experiment in which a moonshine embryo is fused with a wild-type sibling, both the wild-type and mutant primitive macrophages were exposed to the apoptotic bodies from the mutant primitive erythroid progenitors; yet only the mutant macrophages gradually decayed between 2 and 5 dpf. This demonstrates that Trim33 is required cell-autonomously for the lifespan of the yolk sac-derived primitive macrophages in zebrafish. In mammals, yolk sac-derived macrophages have been shown to constitute the majority of resident macrophages in many adult tissues throughout life (Kierdorf et al., 2015). Our results suggest that Trim33 could well be one of the key transcriptional regulators fostering their remarkably long lifespan and self-renewal capacity.

We then found that, in moonshine mutants, the first production of definitive macrophages (from aorta-derived HSPCs) is delayed by at least 2 days relative to wild type. To establish this, we first had to determine when the first definitive macrophages normally arise, as this is not known. One reason is that the first macrophages (and neutrophils) to be found in the CHT niche are actually of primitive origin. Indeed, when the primitive blood circulation starts by 25 hpf, in the yolk sac, it is not yet enclosed in a vessel; hence, it bathes the primitive myeloid cells that are still maturing there (Herbomel and Levraud, 2005; Herbomel et al., 1999); part of them are taken by the blood flow and most of these home to the tail, within the forming CHT vasculature (Le Guyader et al., 2008; Murayama et al., 2006). Therefore, to discern the beginning of definitive macrophage and neutrophil production in the CHT, it was necessary either to remove the interference of the primitive myeloid cells or to use a cell labelling method that is able to discriminate the two origins. We used both approaches, and they led to the very same results – the first definitive monocytes/macrophages arise in the CHT by 3.5 dpf. This consistency notably tells us that the absence of primitive macrophages did not trigger, for example, some acceleration of macrophage production from definitive hematopoiesis. As aorta-derived HSPCs begin to home to the CHT by 1.5 dpf (Kissa et al., 2008), this delay could mean that the differentiation process from freshly born HSPCs to monocytes/macrophages takes a minimum of 2 days (of note, although in mammals macrophage production from definitive hematopoiesis is known to occur through a monocyte intermediate, in the zebrafish model we do not yet have markers to discriminate monocytes from macrophages within hematopoietic organs).

In contrast, in moonshine larvae, the first definitive macrophages arose no earlier than 6 dpf, not specifically in the CHT but instead in the anterior part of the larvae, notably the thymus and anterior kidney areas. As the definitive HSPCs have begun to colonize the final hematopoietic organ, the anterior kidney, by 4 dpf, these observations could suggest that, in Trim33-deficient larvae, monocyte/macrophage production from the definitive HSPCs may not occur at all in the CHT, but only later in the kidney marrow.

In addition to the delayed (and possible absence of) production of definitive macrophages in the CHT of Trim33-deficient larvae, we found that their CHT overproduces not only neutrophils, but also thrombocytes, relative to wild-type larvae. Based on the defective erythropoiesis and enhanced granulopoiesis in that niche, Monteiro et al. (Monteiro et al., 2011) proposed that Trim33 deficiency caused an imbalance between the myeloid and erythroid fates for the HSPCs in the CHT. The more complete view proposed by our results, i.e. the concomitant lack of macrophage production and excess of thrombocytes, is an alternative hypothesis: an imbalance between the erythroid and thrombocytic fates on one side, and between the macrophage and neutrophil fates on the other side. However, the impact of Trim33 deficiency on HSPC differentiation output is still different in the final niche, the kidney marrow (where erythropoiesis is still compromised; Ransom et al., 2004), neutrophils and thrombocytes are still overproduced, but monocytes/macrophages are also now produced. It will be interesting to determine whether the differential impact of Trim33 deficiency in the two niches also occurs in the homologous hematopoietic niches of mammals – the fetal liver then the bone marrow.

Wild-type, transgenic and mutant zebrafish embryos were raised at 28°C in Embryo Water [Volvic water containing 0.28 mg/ml Methylene Blue (M-4159; Sigma) and 0.03 mg/ml 1-phenyl-2-thiourea (P-7629; Sigma)], and staged according to Westerfield (Westerfield, 2007). The moonshine^{NQ039 or t30813} (Demy et al., 2017) and moonshine^TB222 (Ransom et al., 2004) mutations were maintained by outcross of heterozygous carriers with wild-type and/or Tg fish from the following lines: Tübingen wild-type fish (ZIRC), Tg(mpeg1:mCherry-F)^ump2 (Nguyen-Chi et al., 2014), Tg(lyz:DsRed2)^nz50 (Hall et al., 2007), Tg(mfap4:mCherry-F)^ump6 (Phan et al., 2018), Tg(mpeg1:Gal4FF)^gl25 (Palha et al., 2013), Tg(UAS:Kaede)^rk8 (Hatta et al., 2006), Tg(mfap4:turquoise)^xt27 (Walton et al., 2015), Tg(UAS-E1b:Eco.NfsB-mCherry)^c264 (Davison et al., 2007), Tg(gata2b:Gal4;UAS:lifeAct-eGFP) (Butko et al., 2015), Tg(mpeg:GFP-F)^sh425 (Keatinge et al., 2015), Tg(mpx:GFP)ⁱ¹¹⁴ (Renshaw et al., 2006), Tg(gata1:dsRed)^sd2 (Traver et al., 2003) and Tg(CD41:GFP)^la2 (Lin et al., 2005).

Pools of ∼100 zebrafish larvae from four replicate moonshine^NQ039 couples were anesthetized with 0.16 mg/ml Tricaine (A-5040; Sigma), screened for the mon^NQ039 phenotype (based on absence of circulating erythrocytes) and their total RNA was extracted using TRIzol (Invitrogen), following the manufacturer's protocol. cDNA was obtained using MMLV H Minus Reverse Transcriptase (Promega) with a dT17 primer. Trim33 PCR was performed using LA Taq polymerase (Takara) and the following primer pair: 5′-GCTCCTACTGCTCCTCCGTCCA-3′ and 5′-GATGGACGAGCTGGAGTGTG-3′.

Sequencing of the amplicon was performed by Eurofins using the forward primer, and led to the identification of a C>T change (Gln>Stop) at the 1745 bp position, resulting in a truncated protein retaining 581 out of the 1176 amino acids of the wild-type protein. The other moonshine mutant used in this study, mon^tb222 (Ransom et al., 2004), harbors a nonsense mutation resulting in a Trim33 protein retaining the first 433 amino acids.

Zebrafish Tg(mpeg1:mCherry-F) larvae obtained from mon^NQ039 heterozygous parents were anesthetized at 4 dpf with 0.16 mg/ml Tricaine (A-5040; Sigma), screened for the mon^NQ039 phenotype (based on absence of circulating erythrocytes) then dissociated into single-cell suspensions as previously described (Covassin et al., 2006). Fluorescence-activated cell sorting (FACS) of red fluorescent macrophages was performed using a FACSCALIBUR and a FACSAria cytometers (BD Biosciences). Data analysis was performed using FlowJo software (TreeStar). The experiment was replicated three times, using pools of over 50 larvae per condition.

Embryos and larvae were anesthetized with Tricaine at the stage of interest, then fixed overnight at 4°C in 4% methanol-free formaldehyde (Polysciences, 040181). Whole-mount in situ hybridization was performed according to Thisse and Thisse (2014). Whole-mount immunohistochemistry was performed as described previously (Murayama et al., 2006), omitting the acetone treatment. The primary antibody used was a rabbit anti- zebrafish L-plastin polyclonal antibody (at 1:5000) (Le Guyader et al., 2008) and the secondary antibody was a Cy3-coupled anti-rabbit antibody (111-166-003; Jackson Immunoresearch) used at 1:800 dilution.

Sudan Black staining of neutrophils in formaldehyde-fixed embryos was carried out as previously described (Le Guyader et al., 2008).

Embryos from the Tg(mpeg1:gal4; UAS:Kaede) line were dechorionated at 2 dpf, then screened under a Leica M165 FC fluorescence stereomicroscope. Embryos were exposed in groups of 10 under UV light to photoconvert Kaede from green to red using a HBO lamp, and their macrophages were then followed up every day from 2 to 11 dpf. All larvae were kept in the dark and observed with a UV filter during the whole duration of the study.

Staining of acidic compartments was achieved by bathing live zebrafish embryos for over 30 min in the dark in a LysoID solution (Enzo, 510005) diluted at 1/1000. Larvae were then rinsed three times, and quickly imaged. For macrophages ablation with Metronidazole (Mtz), embryos were dechorionated and screened for red fluorescence using a fluorescence stereomicroscope. They were incubated in the dark at 28°C overnight with 10 mM Mtz freshly diluted from powder (Sigma M3761) in embryo water+1% DMSO. Embryos were washed three times before observation/imaging. To arrest the heartbeat in order to image the heart content, 6 dpf zebrafish larvae were placed for a few hours in water containing 10 mM BDM (2.3-butanedione monoxime: Sigma, B-0753) as previously described (Miyasaka et al., 2011).

The two antisense morpholinos (MOs) used in this study were synthesized by Gene Tools: Anti-Pu.1 MO (Rhodes et al., 2005), 5′-GATATACTGATACTCCATTGGTGGT-3′; anti-Irf8 MO (Yu et al., 2017), 5′-AATGTTTCGCTTACTTTGAAAATGG-3. 1-5 nl of 0.6 mM MO solution was microinjected in one- to two-cell stage embryos.

Clodronate liposomes (clodronateliposomes.com) were stained with a DiO solution (2.5 mg DiO in 1 ml ethanol). 10 μl of DiO solution were added per ml of liposomes solution. The sample was incubated for 10 min at room temperature, then centrifuged twice at 20,000 g for 10 min. The supernatant was removed and replaced with sterile PBS. A few nanoliters of the final solution was injected in the Duct of Cuvier of 48 hpf zebrafish embryos, and proper injection in the blood flow was subsequently verified under a fluorescence dissecting scope.

Parabiotic embryos were generated as previously described (Demy et al., 2013) by fusing wild-type Tg(mpeg1:GFP-F) embryos at the late blastula stage to stage-matched mon^NQ039 Tg(gata1a:DsRed; mfap4:turquoise) embryos; mutant embryos were identified by the absence of circulating DsRed-positive erythroid cells.

Low-magnification bright-field images were acquired using video cameras mounted on a Leica Macrofluo driven by the Metavue (Metamorph) software or a Zeiss Macrofluo driven by the Zen software (Zeiss). Wide-field video-enhanced (VE) Nomarski/differential interference contrast (DIC) and fluorescence microscopy were performed as described previously (Herbomel and Levraud, 2005; Murayama et al., 2006) through the 40×/1.00 water-immersion objective of a Nikon 90i microscope or the 40×/1.00 oil-immersion objective of a Reichert Polyvar 2 microscope. Images were obtained from a HV-D20 3-CCD camera (Hitachi) and digitized through a GVD-1000 DV tape recorder (Sony). Still images were then collected using the BTVpro software (Bensoftware). For fluorescence confocal imaging, embryos and larvae were mounted as previously described (Demy et al., 2017). Images were then captured at selected times on an inverted Leica SP8 set-up allowing multiple point acquisition, in order to image mutants and their siblings in parallel. Images were treated and analyzed using the Fiji (ImageJ) software.

Cells were manually counted on Fiji (ImageJ) from maximum projections of total z-stacks of whole live zebrafish larvae, obtained with the SP8 Leica confocal set-up.

At the stages of interest, live mutant and wild-type larvae were observed under a Leica M165FC fluorescence stereomicroscope and binary scored as having a ‘normal’ macrophage population or not, according to the phenotype presented in Fig. 1A.

Fluorescence intensity was automatically measured on Fiji (ImageJ) after manual thresholding of images obtained with the SP8 Leica confocal set-up. It was then expressed in Area (number of pixels above threshold) and Pixel Intensity. Macrophage roundness was automatically measured on Fiji (ImageJ) from images obtained with the SP8 Leica confocal set-up, after manual determination of cells. Macrophage ramifications were manually counted on Fiji (ImageJ) from maximum projections of z stacks of live zebrafish larvae, obtained with the SP8 Leica confocal set-up.

To evaluate differences between means of non-Gaussian data, they were analyzed with a Mann-Whitney-Wilcoxon test. When appropriate, distributions were normalized by log, and an analysis of variance (two-way ANOVA) was performed (for experiments with two variables). P<0.05 was considered statistically significant (****P<0.0001; ***P<0.001; **P<0.01; *P<0.05). Statistical analyses and graphic representations were carried out using Prism software.

We thank our fish facility team for their excellent care of the fish, and Léa Torcq and Valerio Laghi for their help in distant work procedures.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.200835.reviewer-comments.pdf