Distinct regulatory networks control the development of macrophages of different origins in zebrafish

Macrophages are a subclass of innate immune cells that reside in almost all tissues of vertebrates.^1-4  The importance of this population of cells is exemplified by their broad involvement in organogenesis, tissue regeneration and remodeling, and immune defense, and perturbation of their development and functions often leads to detrimental consequences, resulting in occurrence of various disorders such as leukemia, infectious diseases, and autoimmune diseases. Therefore, a comprehensive understanding of how tissue-resident macrophages are formed and how they function will be clinically beneficial to cure macrophage-associated disorders.

In mammals, macrophages are predominantly derived from the hematopoietic precursors born in the yolk sac (YS) and the aorta-gonad-mesonephros (AGM), where embryonic and adult hematopoiesis occurs.^5-9  Recent cell fate–mapping studies with the conventional promoter-directed CreER-loxP system have indicated that macrophages derived from the YS and AGM manifest distinct colonization potential and tissue preferences.^10-12  Likewise, macrophages originated from the rostral blood island (RBI) and ventral wall of dorsal aorta (VDA), the zebrafish hematopoietic tissue equivalent to the mammalian YS and AGM for myelopoiesis, respectively,^13-16  colonize the central nervous system with distinct colonization kinetics and potential.¹⁷  However, the genetic programs governing the development of these different macrophage populations and their cellular functions remain unclear.

It is well known that many key steps of macrophage development are dictated by tissue-specific transcription factors.¹⁸  PU.1, the founding member of the spleen focus forming virus proviral integration oncogene (SPI) subfamily of E-twenty-six (ETS) transcription factors, is one of the best-characterized tissue-specific transcription factors involved in macrophage development.^19,20  PU.1-null mice die either in late gestation or shortly after birth and display profound defects in myeloid cell formation during early embryogenesis.^19,20  Subsequent studies in mice and zebrafish have revealed that PU.1 regulates specification and maturation of macrophages and neutrophils in a dosage-dependent manner.^21,22  Because PU.1-null mice cannot survive to adulthood and spatial-specific lineage tracing in PU.1 knockout mice is still lacking, it remains unclear whether PU.1 is differentially required for different waves of macrophage development. In addition to PU.1, the mammalian SPI subfamily of ETS transcription factors includes 2 other members, SPI-B and SPI-C. Interestingly, although SPI-B shares the highest degree of similarity with PU.1, knockout studies in mice have shown that SPI-B is dispensable for the development of the macrophage lineage.²³  By contrast, SPI-C is required for the development of red pulp macrophage in spleen.²⁴  Similar to mammals, 3 members of the Spi subfamily (Pu.1, Spi1-like and Spi-c) have been annotated in zebrafish. Genetic ablation and morpholino (MO) knockdown studies have shown that Pu.1 is the functional ortholog of mammalian PU.1,^21,25  whereas Spi1-like has been shown to play a role in embryonic myelopoiesis.²⁶  Intriguingly, our recent study showed that while RBI-born (embryonic) macrophages are depleted in hypomorphic pu.1^G242D mutant zebrafish, VDA-derived (adult) macrophages remain largely unaffected,¹⁷  suggesting that the development of RBI- and VDA-born macrophages is differentially regulated by Pu.1 and implying that additional factors are involved in the regulation of VDA-born macrophages.

In this study, we used reverse genetics and fate-mapping analysis to dissect the genetic programs governing the development of RBI- and VDA-born macrophages. We showed that while pu.1-null mutation is sufficient to block the formation of RBI-derived macrophages, simultaneous inactivation of Pu.1 and Spi-b is necessary to abolish the development of VDA-derived macrophages. Epistatic analysis revealed that Pu.1 and Spi-b act in parallel and cooperatively to regulate VDA-derived macrophage development, whereas Pu.1 acts upstream of Spi-b during the development of RBI-derived macrophages. We further documented that Irf8 is a common key downstream target of these distinct Pu.1−Spi-b genetic loops for both RBI- and VDA-derived macrophage development.

Zebrafish were maintained according to standard protocol.²⁷  AB wild-type, pu.1^G242D,²¹ pu.1^Δ839, pu.1^Δ371, spi-b^Δ232, irf8^Δ171, Tg(zpu.1:eGFP)^df5,²⁸ Tg(hsp70:mCherry-T2aCreERT2)^#12,²⁹ Tg(coro1a:eGFP)hkz05t,³⁰ Tg(mpeg1:loxP-DsRedx-loxP-eGFP)hkz015t,³¹  and Tg(mpeg1:loxP-eGFP) lines were used in this study.

Transcription activator-like effector nuclease (TALEN) constructs were generated using the “unit assembly” method.³²  Synthesis of capped TALEN messenger RNA (mRNA) was performed with the Sp6 mMESSAGE mMACHINE Kit (Ambion) according to the manufacturer’s instructions.

pu.1^G242D genotyping was performed by BsmFI digestion as described previously.¹⁷  The other mutants (pu.1^Δ839, pu.1^Δ371, spi-b^Δ232, and irf8^Δ171) were identified by DNA electrophoresis of polymerase chain reaction (PCR)–amplified genomic DNA. The following primers were used for PCR amplification: pu.1^Δ839 or pu.1^Δ371, 5′-CCTTGTGTAGACTTCGATGAACTG-3′/5′-CGTTCGCTTGTGTGTTAGTGTCAC-3′; Spi-b^Δ232, 5′-GGTCATGGCCAAGACTTGATC-3′/5′-CTAAACTG TGCAGGGCTTCGG-3′; and irf8^Δ171, 5′-CTGCTTGGATGCCGTGAGTATG-3′/5′-CAT GTACACAGGGATCTATGATC-3′.

IR-LEGO-CreER-loxP labeling was performed as previously described.¹⁷  In brief, embryos were mounted in methylcellulose and subjected to heat shock with a 1345-nm infrared laser in the RBI and VDA, respectively. The power of the 1340-nm laser was ∼180 to 200 mW, and the duration of each heat shock was 30 seconds. Three spots in one side of the RBI of 12– to 18–hours postfertilization (hpf) embryos and 4 spots in the VDA of 28- to 30-hpf embryos were irradiated. Immediately after the irradiation, the embryos were treated with 4-OHT for 17 to 22 hours, and the labeled cells were marked by GFP expression.

The 9-kb promoter of pu.1²⁸  and flag-myc double-tagged zebrafish pu.1 or spi-b were cloned into the modified pBluescript II SK(+) vector containing 2 arms of Tol2 sequence. The resulting -9kbpu.1:flag-myc-pu.1 or -9kbpu.1:flag-myc-Spi-b construct, together with transposase mRNA,³³  was injected into 1-cell-stage embryos. The injected embryos were raised to adulthood for further screening of germline transmission.

Antisense RNA probes and whole-mount in situ hybridization (WISH) were performed according to standard protocol.²⁷ 

Two-color fluorescence in situ hybridization (FISH) and costaining of 1-color FISH with antibody were performed as previously described.³⁴  Primary antibodies used in this study were goat anti-GFP (Abcam), rabbit anti-DsRed (Clontech), and rabbit anti-Lcp1.¹⁵ 

Neutral red staining was performed as previously described.¹⁴ 

Acridine orange and terminal deoxynucleotidyltransferase-mediated dUTP nice end labeling (TUNEL) staining was performed according to manufacturer’s instruction or as previously reported.³⁵ 

All the samples were fixed with 4% paraformaldehyde for 1 or 2 days and washed with PBS for another day. For 15- and 30–days postfertilization (30-dpf) juveniles, the fixed fish were directly subjected to cryosection at 30-µm intervals and subsequently costained with GFP/DsRed and Lcp1 antibodies. For adult fish, the brains of the fixed fish were dissected for cryosection and antibody staining.

Fluorescent signals from live reporter lines, FISH, or immunostaining were imaged under a Leica SP8 confocal microscope. The HC PL APO 20×/0.70 DRY objective was used in this study.

To dissociate cells from zebrafish embryos or juveniles, 22G or 24G needles were used to break up the tissues, followed by 37°C digestion with dispase (5% final, Sigma) for 20 min. The dissociated cells were re-suspended with 400 to 600 µL cold Hank’s balanced salt solution, passed through 40-µm nylon filters, and finally subjected to fluorescence-activated cell sorting (FACS) with a Becton Dickinson FACSAria IIIu.

Total RNA were extracted from FACS-sorted cells with NucleoSpin RNA XS kit (Macherey-Nagel) and subjected to reverse transcription (RT) and amplification with the REPLI-g WTA Single Cell Kit (QIAGEN) according to the manufacturer’s instructions. For large amount of samples, total RNA was extracted with TRIzol reagent or the RNeasy Mini Kit (QIAGEN), and complementary (cDNA) was synthesized with SuperScript III reverse transcriptase (Thermo Fisher).

Quantitative PCR was performed with SYBR Green Supermix (Roche) on a 7500 Fast Real-time PCR system (ABI) and analyzed using the ΔΔCt method. The following primers were used for quantitative PCR: elf1a, 5′-CTTCTCAGGCTGACTGTGC-3′/5′-CCGCTAGCATTACCCTCC-3′; lyz, 5-GCCTGTTCAGACTTGCTTAACG-3′/5′-CAGGCTCGGAGGCTTTGTTTG-3′; mpo, 5′-GCTATACCAGGTTATAATGCATG-3′/5′-CCACAACCTATCGCCATCTCG-3′; mfap4, 5′-GGATGGACGGTGATTCAGAG-3′/5′-CAGATAAAGAGTCGCCTGCT-3′; mpeg1, 5′-CTCCACAGAAAACCAGCGCA-3′/5′-CGTCAGCGATTTCTTCTGCC-3′; pu.1, 5′-GACAGTCAGAACGATCACTCTT-3′/5′-GGAGAGGAGATGGCTGGACG-3′; Spi-b, 5′-GACTCATCCAGAGAGCGGATG-3′/5′-ACAGCGGATGGGAGATGTAAG-3′; and irf8, 5′-CTGCTTGGATGCCGTGAGTATG-3′/5′-ACCTAACTTTTGTTCCTCCTCTG-3′.

Myc-tagged fusion mRNA for each gene were coinjected with eGFP mRNA into 1-cell-stage zebrafish embryos. At 6 hpf, ∼100 embryos were washed with 500 µL PBS and passed through a 100-µL pipette to remove the YS. After centrifugation at 200g for 5 minutes, the pellets were resuspended with 100 µL SDS lysis buffer (63 mM Tris-HCl, pH 6.8, 10% glycerol, 5% β-mercaptoethanol, and 0.002% bromophenol blue). After centrifugation at 13 000 rpm for 10 minutes, 5 µL supernatant was loaded for immunoblotting.³⁶  Anti-Myc antibody (Santa Cruz Biotechnology) and anti-GFP antibody (a kind gift from Robert Qi in the Hong Kong University of Science and Technology) were used as primary antibodies.

Statistics were analyzed using the 2-tailed Student t test. A result was considered significant if P < .05. Values represent mean ± standard error of the mean.

Previous studies have revealed that reducing Pu.1 activity in hypomorphic pu.1^G242D mutant zebrafish blocks the development of RBI-derived macrophages but has little effect on the generation of VDA-derived macrophages.¹⁷  This observation suggests that perhaps Pu.1 is dispensable for the VDA-derived macrophage development. To confirm this notion, we generated an amorphic pu.1 allele pu.1^∆839 (supplemental Figure 1A, available on the Blood Web site), in which an 839-bp genomic fragment of the pu.1 gene was deleted, leading to a 194-bp deletion in exon 4 (supplemental Figure 1B) and synthesis of an unstable truncated Pu.1^∆839 protein lacking the Q-rich transcriptional activation, PEST, and Ets DNA-binding domains (supplemental Figure 1C-F). Thus, pu.1^∆839 likely represents a null allele. Phenotypic characterization revealed loss of macrophages in pu.1^∆839 mutants during early development as indicated by the lack of mfap4 staining in the periphery at 30 hpf and the loss of neutral red staining in the brain at 3 dpf (Figure 1A-D). In addition, neutrophils (lyz⁺ cells) and early myeloid progenitors (cebpα⁺ cells) were also dramatically reduced in pu.1^∆839 mutants (Figure 1E-H). Since these early macrophages and neutrophils are derived from the RBI,^13,14,16,37  these data suggest that loss of Pu.1 blocks the development of RBI-born macrophages and neutrophils. Consistent with the null mutation, the phenotypes in pu.1^∆839 mutants are significantly more severe than the myeloid defects observed in a hypomorphic pu.1^G242D mutant, where neutrophils and early myeloid progenitors are intact.²¹  However, macrophages (as indicated by macrophages on the skin and microglia in the brain) in pu.1^∆839 mutants recovered from 5 dpf onwards (supplemental Figure 2A-D). The recovered cells were likely to be able to differentiate into mature macrophages as they displayed typical branched morphology (supplemental Figure 2A′-B′) and could phagocytose bacteria (supplemental Figure 2E-F). Likewise, neutrophils also became evident from 4 dpf onwards (Figure 1L-M). The recovery of macrophages in pu.1^∆839 mutants is similar to that observed in hypomorphic pu.1^G242D mutants,¹⁷  suggesting that complete loss of Pu.1 function fails to block VDA-born macrophage development. To prove this is indeed the case, we outcrossed pu.1^∆839 with Tg(hsp70:mCherry-T2a-CreER^T2;mpeg1:loxP-DsRedx-loxP-eGFP) (we herein refer to loxP-DsRedx-loxP-eGFP as LRLG) reporter line, in which CreER recombinase is induced by spatial-restricted infrared light–mediated heat shock and subsequently activated by tamoxifen, resulting in the conversion of mpeg1-DsRed into mpeg1-GFP.¹⁷  The Tg(hsp70:mCherry-T2a-CreER^T2;mpeg1:LRLG);pu.1^∆839 mutants were then used to determine the source of the recovered macrophages in pu.1^∆839 mutants using infrared light–induced IR-LEGO-CreER-loxP system.¹⁷  Results showed that the recovered macrophages (GFP⁺ cells) in pu.1^∆839 mutants were exclusively derived from the VDA (Figure 1I-K ). From these observations, we conclude that Pu.1 is essential for the development of RBI-derived macrophages but is dispensable for those derived from the VDA.

The normal development of VDA-born macrophages and neutrophils in pu.1^∆839 mutants suggests that additional factors, perhaps other Ets transcription members, may compensate for the loss of Pu.1 function in the development of VDA-derived myelopoiesis. We therefore turned our attention to Spi1-like, which is thought to be a co-ortholog of mammalian PU.1 and has been shown to be involved in early myeloid development in zebrafish.²⁶  Intriguingly, our phylogenic tree analysis and protein sequence alignment showed that Spi1-like shares the highest similarity with the mammalian SPI-B among Ets family transcription factors in zebrafish (Figure 2A-B). Furthermore, critical amino acids in the acidic subdomain and Q-rich region of the PU.1 transactivation domain³⁸  were absent in Spi1-like protein (Figure 2C, underlined black letters). Instead, a unique P/S/T-rich region known to be essential for mammalian SPI-B transactivation activity³⁹  is present in Spi1-like (Figure 2C, dashed black box). Thus, Spi1-like is likely to represent the zebrafish ortholog of mammalian SPI-B and is herein designated as Spi-b. To dissect the role of Spi-b, we generated a null allele spi-b^∆232, which carried a 232-bp deletion and a 4-bp insertion in exon 5, resulting in the production of a truncated Spi-b^∆232 protein lacking 78 amino acids in the Ets domain (supplemental Figure 3). Surprisingly, spi-b^∆232 mutant fish developed normally to adulthood without obvious myeloid phenotypes (data not shown). To further confirm that loss of Spi-b function indeed had no effect on the VDA-derived myeloid development, we outcrossed spi-b^∆232 mutants with the Tg(hsp70:mCherry-T2a-CreER^T2;mpeg1:LRLG) reporter line and examined the generation of VDA-derived macrophages using IR-LEGO-CreER-loxP labeling. Results showed that spi-b^∆232 mutants produced comparable numbers of VDA-derived macrophages to siblings (Figure 2D-F). Likewise, spi-b^∆232 mutants contained comparable VDA-derived neutrophils (Figure 2G-H). These results demonstrate that inactivating Spi-b has no effect on the VDA-born myeloid cell development.

The above observations suggest that Pu.1 and Spi-b may play complementary roles in VDA-born myeloid cell development. We therefore examined VDA-born macrophages and neutrophils in pu.1^Δ839spi-b^∆232 double mutants. We found that, in contrast to pu.1^Δ839 and spi-b^∆232 single mutants, macrophages derived from both origins were severely impaired in pu.1^Δ839spi-b^∆232 double mutants as indicated by the lack of mpeg1⁺ cells from 3 dpf onwards (Figure 3B-H), and by 15 dpf, only a few mpeg1⁺ macrophages were observed in pu.1^Δ839spi-b^∆232 double mutants (Figure 3F-H). The depletion of macrophages in pu.1^Δ839spi-b^∆232 double mutants was attributed to excessive apoptosis of macrophages as indicated by the increase of TUNEL⁺/coro1a-GFP⁺ cells in the VDA and caudal hematopoietic tissue (CHT) of 4-dpf double mutants (Figure 3I-J) and AO⁺mpeg1⁺ macrophages in the trunk of 15-dpf double mutants (Figure 3K-L). The residual mpeg1⁺ cells found in pu.1^Δ839spi-b^∆232 double mutants displayed a small-rounded cell shape with very few branches (Figure 3F′-3G′), suggesting a severe impairment of macrophage differentiation in pu.1^Δ839spi-b^∆232 double mutants. Likewise, microglia, the central nervous system–resident macrophages were completely absent in pu.1^Δ839spi-b^∆232 double mutants (Figure 3M-N). However, VDA-born hematopoietic stem/progenitor cells (cmyb⁺ cells), neutrophils (SB⁺ cells), and T lymphocytes (rag1⁺ cells) developed normally (supplemental Figure 4). These results demonstrate that Pu.1 and Spi-b are essential for both the RBI- and VDA-born macrophage development. Notably, the double mutants could not survive to adulthood and died at ∼25 to 30 dpf, perhaps due to the lack of macrophages (data not shown).

To dissect the genetic hierarchy of pu.1 and spi-b in the VDA-born macrophage development, we first performed spi-b RNA WISH and anti-GFP antibody double staining to examine the expression of pu.1 and spi-b in Tg(pu.1:GFP) transgenic fish²⁸  at 4 dpf, when VDA-born hematopoietic cells are known to colonize the CHT.^40,41  Results showed that majority of the pu.1-GFP cells in the CHT were positive for spi-b mRNA (Figure 4B-D), showing that pu.1 and spi-b are co-expressed in the same VDA-born myeloid cells. Thus, Pu.1 and Spi-b appear to act in parallel to regulate the development of VDA-derived macrophages. To support this notion, we examined the expression of spi-b and pu.1 by monitoring spi-b transcripts and GFP signal in the CHT region of pu.1^Δ839 mutants and spi-b^∆232;Tg(-9.0kbpu.1:eGFP) mutants. As indicated in Figure 4E-H, spi-b and pu.1 transcripts were clearly not downregulated in pu.1^Δ839 and spi-b^∆232;Tg(-9.0kbpu.1:eGFP) mutants, respectively (Figure 4E-H). In fact, real-time RT-PCR analysis with VDA-derived macrophages isolated from pu.1 MO-injected wild-type Tg(mpeg1:LRLG) (pu.1 MO blocks the development of RBI-, but not VDA-derived, macrophages; supplemental Figure 5), pu.1^Δ839;Tg(mpeg1:LRLG), and spi-b^∆232;Tg(mpeg1:LRLG) mutants revealed an upregulation of spi-b and pu.1 transcripts in pu.1^Δ839 and spi-b^∆232 mutants, respectively, perhaps due to the compensation effect (Figure 4I-J). Consistent with the genetic results that pu.1 and spi-b are capable of fully compensating each other, pu.1 and spi-b were expressed at comparable levels in the VDA-derived macrophages (Figure 4K). Taken together, we conclude that pu.1 and spi-b act in parallel and cooperatively to regulate the VDA-derived macrophage development (Figure 4L).

A previous study reported by Bukrinsky et al showed that the suppression of Spi-b function by spi-b MO severely blocks embryonic (RBI-born) myeloid cell development.²⁶  We were therefore keen to find out whether RBI-derived myeloid cells are impaired in spi-b^∆232 mutants. Surprisingly, we found that RBI-born embryonic neutrophils and myeloid progenitors, peripheral macrophages, and microglia in the brain developed normally in spi-b^∆232 mutants (Figure 5A-H). This result is somewhat unexpected, as spi-b is highly expressed in RBI-derived myeloid cells during early myelopoiesis²⁶  (Figure 6C). We hypothesized that because of the compensation of Pu.1, the role of Spi-b in the RBI-born myeloid cell development could not be revealed in spi-b–deficient mutants. To test this hypothesis, we used a weak hypomorphic pu.1 allele, pu.1^∆371, which harbors an in-frame 42-amino-acid deletion in the transactivation domain, resulting in the synthesis of a truncated Pu.1 protein with reduced transactivation activity (supplemental Figure 6A-D). In these weak hypomorphic mutants, RBI-derived macrophages and microglia were only partially reduced (supplemental Figure 6E-H). We then outcrossed pu.1^Δ371 mutants with spi-b^∆232 and asked if inactivating Spi-b in this pu.1^Δ371 hypomorphic mutant background would cause a more severe phenotype. Indeed, the number of RBI-derived macrophages (Figure 5I-J,M), but not neutrophils (Figure 5K-M), was further reduced in pu.1^Δ371;spi-b^∆232 double mutants. These results demonstrate that Spi-b participates in RBI-derived macrophage development, although its role is limited.

To delineate the molecular basis underlying the differential requirements of Pu.1 and Spi-b for RBI-derived macrophage development, we examined the expression of spi-b and pu.1 in RBI-derived myeloid cells in hypomorphic pu.1^G242D mutants and spi-b^∆232 mutants. WISH showed that the expression of spi-b was completely abolished in pu.1^G242D mutants (Figure 6C-D), despite the presence of RBI-derived myeloid progenitors (Figure 6A-B). In contrast, the pu.1 transcripts were intact in spi-b^∆232 mutants (Figure 6E-F). These results indicate that spi-b is genetically downstream of Pu.1 during embryonic myeloid cell development. Interestingly, real-time RT-PCR analysis revealed that the expression level of spi-b in embryonic myeloid cells was significantly lower than that of pu.1 (Figure 6G), indicating that the limited role of Spi-b in RBI-derived macrophage development is due to its low expression. Indeed, when expressed at a similar level using a previously characterized pu.1 promoter,²⁸  the ability of Spi-b or Pu.1 to rescue RBI-derived macrophages was indistinguishable in pu.1^Δ839 mutants (Figure 6H-N). Taken together, we conclude that during RBI-derived macrophage development, Pu.1 is a master regulator that acts upstream of Spi-b, which, upon induction by Pu.1, can partially compensate for the function of Pu.1 (Figure 6O).

The above genetic analysis has shown that pu.1 and spi-b form 2 distinct genetic networks, a vertical and a parallel cascade, to regulate RBI- and VDA-born macrophages, respectively. To identify the key targets downstream of these distinct Pu.1−Spi-b regulatory networks, we focused on Irf8, a transcription factor known to be a Pu.1 downstream target and essential for macrophage development.^34,42,43  Indeed, WISH showed that the irf8 expression in RBI-born myeloid cells was completely absent in pu.1-deficient mutants (both hypomorphic pu.1^G242D and null pu.1^Δ839 alleles) (Figure 7A-B). Likewise, real-time RT-PCR analysis with the residual VDA-derived macrophages isolated from 20-dpf Tg(mpeg1:LRLG);pu.1^Δ839;spi-b^Δ232 double mutants showed that the irf8 expression was also abolished in the VDA-born macrophages in pu.1^Δ839;spi-b^Δ232 mutants (Figure 7C). To confirm that irf8 is a key downstream target of both the vertical and the parallel Pu.1−Spi-b regulatory cascades, we generated an irf8-null allele, irf8^Δ171, which harbors a 171-bp deletion and a 3-bp insertion in exon 1, resulting in the generation of a truncated protein lacking the DNA-binding domain due to the usage of the downstream ATG for translation initiation (supplemental Figure 7). Similar to pu.1^Δ839;spi-b^Δ232 double mutants, RBI- and VDA-born macrophages (peripheral macrophages and microglia) in irf8^Δ171 mutants were drastically reduced or absent (Figure 7D-I,P-Q). As expected, the expression of pu.1 and spi-b in RBI- and VDA-derived myeloid cells remained either unchanged or upregulated in irf8^Δ171 mutants (Figure 7K-O), perhaps due to a feedback mechanism. These data demonstrate that the vertical and parallel Pu.1−Spi-b regulatory networks orchestrate the development of RBI- and VDA-derived macrophages, at least in part, by regulating a common downstream regulator, Irf8. Notably, in contrast to the complete absence of microglia and early lethality of pu.1^Δ839;spi-b^Δ232 double mutants, irf8^Δ171 mutant fish are viable and have a partial recovery of mpeg1⁺ cells in the brain in adulthood (Figure 7R-V). The recovered mpeg1⁺ cells in adult irf8^Δ171 mutants appeared to be macrophages, as they lacked Mpx activity, a typical feature of neutrophils (supplemental Figure 8C-D), and were able to engulf bacteria (supplemental Figure 8A-B). This observation suggests that the development of a small portion of VDA-born macrophages is independent of Irf8.

In this report, by combining genetics and a temporal-spatial–resolution fate-mapping method, we reveal that the development of macrophages of different origins is orchestrated by distinct genetic networks comprising the transcription factors Pu.1, Spi-b, and Irf8.

Our study has shown that genetic inactivation of Spi-b does not affect the development of both RBI- and VDA-born myeloid cells. This result is clearly different from those of a previous study, where Bukrinsky et al showed that transient knockdown of spi-b expression by the spi-b MO severely impairs the development of embryonic or RBI-born myeloid cells.²⁶  Three possible mechanisms could explain this discrepancy. One explanation is that the discrepancy is caused by the toxic and off-target effects of MO⁴⁴  or, alternatively, the difference is attributed to the complementary effect that occurs only in the genetic spi-b mutants.⁴⁵  Finally, although the majority of the Spi-b Ets domain was removed in spi-b^Δ232 mutants, we cannot exclude the possibility that weak Spi-b activity was retained, which might be sufficient to sustain the development of RBI-derived macrophages.

Our findings challenge the current view that zebrafish Pu.1 and Spi1-like are co-orthologs of mammalian PU.1. Two lines of evidence support the notion that Spi1-like is the ortholog of mammalian SPI-B.⁴⁶  The first is provided by protein sequence alignment and phylogenic tree analysis, which reveals that Spi1-like shares the highest similarity with mammalian SPI-B among Ets family transcription factors and the transactivation domain of Spi1-like mimics that of mammalian SPI-B, but not PU.1. The second line of evidence is provided by functional studies showing that similar to spi-b^Δ232 mutant zebrafish, SPI-B knockout mice are viable, with no obvious hematopoietic lineage developmental defects,²³  and SPI-B acts cooperatively with PU.1 and can compensate for the function of PU.1 in hematopoietic cell development.^47-49 

Our study demonstrates that RBI- and VDA-derived macrophages are regulated by distinct genetic networks formed by Pu.1 and Spi-b. Although the establishment of these distinct genetic networks remains unknown, the parallel Pu.1−Spi-b cascade does give a competitive advantage as a single copy of either gene is sufficient to support the development of macrophages and ensure the survival of the organism. Unexpectedly, both vertical and parallel Pu.1−Spi-b cascades govern the macrophage development through, at least in part, regulating a common key downstream target irf8. Notably, a recent study in mice has reported that during dendritic cell development PU.1 directly binds the distal enhancers (−50 kb) of the IRF8 locus, leading to the activation of IRF8 transcription.⁵⁰  Remarkably, promoter analysis reveals 2 discrete clusters of potential Pu.1 binding sites at the −16-kb to −14-kb and −10.8-kb to −8.8-kb regions of the zebrafish irf8 locus (supplemental Figure 9), suggesting a conserved regulatory mechanism across species. However, forced expression of irf8 using the hsp70 heat shock promoter is insufficient to restore the development of either RBI-derived macrophages in pu.1^Δ839 single³⁴  or VDA-derived macrophages in pu.1^Δ839;spi-b^Δ232 double mutants (data not shown), suggesting that other transcription factors that act in parallel with Irf8 are involved in macrophage development. This notion is supported by the finding that macrophage defects in irf8-null mutants are less severe than those seen in pu.1^Δ839;spi-b^Δ232 double mutants and by the finding that a small proportion of macrophages continue to develop into microglia in an Irf8-independent fashion. Further studies will be required to uncover these factors.

The authors thank Michael Brand for sharing the Tg(hsp70:mCherry-T2a-CreERT2) transgenic line, Thomas Look for providing the Tg(zpu.1:eGFP)^df5 transgenic line, and Robert Qi for providing anti-GFP antibody.

This work was supported by grants from the National Natural Science Foundation of China (31229003), the Research Grants Council of the Hong Kong Special Administrative region (16102414, 16103515, HKUST5/CRF/12R, AoE/M-09/12, and T13-607/12R), and the Innovation and Technology Commission of the Hong Kong Special Administrative region (ITCPD/17-9)

Contribution: T.Y., W.G., Y.T., J.X., and J.C. designed the research, performed experiments, and analyzed data; L.L. generated irf8^Δ171 mutants; and Z.W. designed the research and analyzed data.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Zilong Wen, Division of Life Science, State Key Laboratory of Molecular Neuroscience and Center of Systems Biology and Human Health, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong, People’s Republic of China; e-mail: zilong@ust.hk.