Introduction

MFAP4 (Microfibril-associated glycoprotein 4) encodes an extracellular matrix (ECM) protein that belongs to the fibrinogen-related protein superfamily. MFAP4 was first associated with Smith-Magenis syndrome (SMS, OMIM #182290), a rare sporadic disorder characterized by skeletal anomalies and intellectual deficit, which is predominantly caused by a 3.7 megabase (Mb) interstitial 17p11.2 deletion encompassing the MFAP4 locus1. Subsequently, copy number variant (CNV) analysis in 464 individuals identified MFAP4 as a candidate gene for the severe cardiac malformations associated with left-sided congenital heart disease (LS-CHD)2. More recently, numerous studies have implicated MFAP4 as a prognostic marker (or biomarker) for assorted cancers, including ovarian3,4, mammary5 and prostate6, as well as hepatic and pulmonary fibrosis7 and chronic obstructive pulmonary disease (COPD)8. Despite the growing list of diseases that implicate MFAP4 in their aetiology, the exact function of MFAP4 in vivo remains largely unknown.

Biochemically, MFAP4 homo-oligomerizes into trimers and hexamers, and interacts with tropoelastin and fibrillins during elastic fiber assembly9 via its highly conserved fibrinogen-related (FReD) domain10. Immunohistochemistry revealed that MFAP4 is highly enriched in vascular smooth muscle cells (VSMCs) in the blood vessel wall of many human tissues including heart, lung, kidney, liver, testis and dermis11,12. In addition to elastic fiber assembly, MFAP4 is thought to play a role in cell adhesion and/or intercellular interactions through its amino (N-) terminal Arg-Gly-Asp (RGD) cell-binding motif10,13. Indeed, MFAP4 binds integrin αvβ3 in vitro and promotes the migration and proliferation of VSMCs14.

We previously exploited the directed differentiation of human embryonic stem cells (hESCs) into the definitive endoderm (DE) lineage as a platform to identify novel genes upregulated after treatment with the TGFβ-related ligands Activin and BMP415. MFAP4 emerged from this screen alongside cardinal markers of the mouse primitive streak, where the mesoderm and DE emerge during gastrulation, including T/BRACHYURY, CER1, LHX1 MIXL1 and SOX17. We subsequently demonstrated by chromatin immunoprecipitation coupled with massively parallel deep sequencing (ChIP-Seq) that MFAP4 is directly bound by the Activin/Nodal effector proteins SMAD2/3, as well as the T-box transcription factor EOMESODERMIN (EOMES) in differentiating hESCs16,17. These findings strongly suggest that MFAP4 is a direct target of SMAD2/3/EOMES transcriptional regulation during the earliest events of germ layer specification in the developing mammalian embryo. Given these results, we further reasoned that loss (or diminished levels) of MFAP4 during early human development could contribute to the pathologies of SMS and LS-CHD.

Anticipating strong evolutionary conservation among vertebrate MFAP4 orthologs, we embarked upon a disease modeling approach using zebrafish (Danio rerio). We first demonstrated that mfap4 is indeed expressed throughout zebrafish embryonic development, but is entirely dispensable as adult mfap4 null mutants generated by CRISPR/Cas9 gene editing are viable and fertile. To date, zebrafish mfap4 has been most widely used as a macrophage marker18,19,20. Coupling our larval and adult mfap4 loss-of-function zebrafish with tail fin injury models, we investigated the role of Mfap4 in the zebrafish innate immune system and revealed an unexpected role for Mfap4 in regulating lineage restriction during haematopoiesis.

Results

mfap4 is expressed during zebrafish embryogenesis

We first established the evolutionary relationship between human, mouse and zebrafish MFAP family members. This phylogenetic analysis demonstrated that all MFAP proteins derive from a common ancestor, and that the human and mouse MFAP proteins are more closely related to each other than to their respective zebrafish orthologs (Fig. 1a). The latter is most likely due to rapid gene duplication and species divergence in response to environmental factors10. Interestingly, zebrafish Mfap4 is closely related to its paralog Mfap3l, as they are part of the same clade. However, there are no reports indicating any functional relationships between these two proteins.

Figure 1
figure 1

Alignment of vertebrate Mfap4 proteins and mfap4 expression during zebrafish development. (a) Phylogenetic analysis of human (Homo sapiens, Hs) MFAP family members and their murine (Mus musculus, Mm) and piscine (Danio rerio, Dr) orthologs. Scale bar represents 1 amino acid substitution per site. (b) Amino acid sequence alignment of human, murine and zebrafish MFAP4 proteins displayed in ClustalX color scheme. Red arrows demarcate the start and end of the 16 amino acid signal peptide (www.uniprot.org); black box highlights the RGD sequence; black arrow indicates the start of the fibrinogen-related domain (FReD). (c–e) Zebrafish mfap4 expression analysis by (c,d) quantitative RT-PCR (mean expression in twenty pooled embryos/larvae per time point) relative to the expression of the housekeeping gene actin, beta 1 (actb1) and by (e) whole-mount RNA in situ hybridization (purple staining)—ventral (13–17 somite stage, ss), dorsal (15–17 ss and 24 h post fertilization, hpf) and lateral (24 hpf) views. Phylogenetic analysis in panel (a) was performed with Phylogeny.fr online software (available at www.phylogeny.fr)63,64. Alignment in panel (b) was performed with the online PRALINE Multiple Sequence Alignment software (Vrije Universiteit Amsterdam, Centre for Integrative Bioinformatics VU, available at www.ibi.vu.nl/programs/pralinewww/). Graphs in panels (c) and (d) were generated with Microsoft Excel (for Mac 2011, version 14.7.7). This figure was created with Inkscape software version 0.92 (available at https://www.inkscape.org/).

Full size image

Although many genes were duplicated during D. rerio evolution, zebrafish only have a single ortholog for human MFAP4 (https://www.zfin.org)9,21. Protein alignment revealed that MFAP4 is well-conserved from humans to zebrafish, with 66% sequence identity and a similar domain structure, including an N-terminal signal peptide and a FReD domain9,22. Significantly, zebrafish Mfap4 lacks an N-terminal RGD sequence (Fig. 1b; see also10).

We next analyzed mfap4 expression by quantitative RT-PCR (qPCR) and whole-mount in situ hybridization (WISH). qPCR revealed that mfap4 is expressed in zebrafish embryos as early as the 50% epiboly stage, which is just prior to the onset of gastrulation (Fig. 1c). mfap4 transcripts were detected throughout the segmentation period (7 to 20 somites)—a result consistent with our previous observation that Mfap4 mRNA levels steadily increase in the developing mouse embryo during the early somite stages (Fig. 1c)15. From the 20 somite stage (ss), zebrafish mfap4 levels slowly increase to around 6 days post fertilization (dpf) (Fig. 1d). These expression dynamics are supported by mfap4 WISH of embryos at mid-segmentation stages (13–17 somites) and at 24 h post fertilization (hpf) and beyond (Fig. 1e and Supplementary Fig. S1). Increased mfap4 expression at 24 hpf correlates with the emergence of the bloodstream (Fig. 1d)23. Indeed, zebrafish mfap4 was first described by Zakrzewska et al.19 as a transcriptional target of Spi1b (Spi-1 proto-oncogene b), which is more commonly known as Pu.119. Consistent with this finding and as previously reported, mfap4 expression conspicuously labels macrophages (Fig. 1e and Supplementary Fig. S1).

The generation of mfap4-deficient zebrafish and their phenotypic characterization

To investigate the role of Mfap4 in zebrafish development, we used CRISPR/Cas9 to generate a 150 base pair (bp) deletion in mfap4 (hereafter designated mfap4∆) using two guide RNAs that target the 5′ and 3′ ends of exon 2 (Fig. 2a,b). The resulting deletion is predicted to encode a truncated protein product of 74 amino acids (p.S17Rfs58X), leaving only the first 7% of Mfap4 intact (Fig. 2a). mfap4+/∆ heterozygous fish were intercrossed and adult mfap4∆/∆ homozygous mutants were recovered that were indistinguishable from their wild-type (WT) siblings (Fig. 2a) and were moreover fertile with a normal lifespan. Very low levels of mfap4 mRNA were detected by qPCR in WT zygotes (Fig. 1c), suggesting minimal maternal contribution. Indeed, maternal-zygotic mfap4∆/∆ mutants are also viable and fertile.

Figure 2
figure 2

Generation of a loss-of-function mutation in zebrafish mfap4. (a) Guide RNA target sites (T1 and T2, in blue) in exon 2 for CRISPR/Cas9 gene editing of the zebrafish mfap4 genomic locus. The resulting 150 base pair (bp) deletion is predicted to generate a truncated 74 amino acid (a.a.) protein product, p.S17Rfs58X. Abbreviations: FReD, fibrinogen-related domain; C, C-terminal; Chr, Chromosome; N, N-terminal; S, signal peptide. (b) Genotyping analysis of wild-type (WT, +/+) and mfap4 heterozygous (+ /∆) and homozygous (∆/∆) mutants. The WT amplicon length is 946 bp and the mutant amplicon length is 795 bp (see Materials and Methods for additional details). The full gel image can be found in Fig. S5. (c) Representative images of 4-month-old WT and mfap4∆/∆ fish. (d) Quantitative RT-PCR analysis with primers indicated in panel (a) by red arrows relative to actb1 reveals that the mutant mfap4 transcript is barely detectable (significant at 1 dpf (P = 0.029612) and 3 dpf (P = 0.041617), but not at 2 dpf (P = 0.084486); assessed by Student’s t-test), which suggests the mfap4p.S17Rfs58X mutant transcript is degraded. Error bars represent the standard error of the mean (s.e.m.) of technical triplicates of 20 pooled embryos per genotype per time point. (e) Whole-mount in situ hybridization of mfap4 in WT and mfap4∆/∆ larvae at 3 days post fertilization. The graph in panel (d) was generated with Microsoft Excel (for Mac 2011, version 14.7.7). This figure was created with Inkscape software version 0.92 (available at https://www.inkscape.org/).

Full size image

qPCR with primers downstream of the 150 bp deletion in exon 2 failed to amplify mutant mfap4 RNA (Fig. 2d)24. WISH analysis on 3 dpf mfap4∆/∆ mutants further demonstrated the loss of mfap4 mRNA (Fig. 2e). Taken together, these results indicate that the mfap4∆ mutation is a complete loss-of-function. Interestingly, we observed increased expression of mfap2 and mfap3l at various larval stages in mfap4∆/∆ mutants, suggesting that Mfap2 and Mfap31 might compensate for the lack of Mfap4, perhaps by transcriptional adaptation, and thus prevent the manifestation of an overt loss-of-function phenotype (Supplementary Fig. S2)25.

mfap4 mutants have reduced numbers of macrophages and elevated production of neutrophils

Upon tissue damage, circulating monocytes extravasate at the site of injury and differentiate into macrophages26. Resident macrophages then produce mediators such as cytokines to promote inflammation and elimination of apoptotic cells26. Recent studies have indicated a direct relationship between macrophages and tissue repair and regeneration27,28. Accordingly, amputated zebrafish tail fins do not fully regenerate in the absence of macrophages28. Given that mfap4 expression specifically labels macrophages, we examined the response of macrophages to injury in the absence of mfap4. mfap4∆/∆ mutant fish were crossed into the Tg(mpeg1:EGFP) reporter background in which macrophages are fluorescently labelled29. Next, we partially amputated the tail fin primordium and assessed the effect of loss of mfap4 on macrophage number (Supplementary Fig. S3). Flow cytometry analysis performed on 3 dpf larvae pre-amputation and immediately after amputation demonstrated considerably fewer macrophages in mfap4∆/∆ mutants as compared to WT larvae, which was statistically significant at multiple analyzed time points. However, macrophage recruitment to the amputation site was not impaired by mfap4 loss. These findings suggest that Mfap4 plays a role in macrophage differentiation during embryonic, primitive haematopoiesis that gives rise to primitive monocytes and macrophages (see below). Around 32–36 hpf, definitive haematopoiesis commences, which gives rise to differentiated monocytes and macrophages from 5 dpf onwards30,31,32,33. As such, we analyzed the macrophage response in partial adult tail fin amputations (Fig. 3a,b). Similar to 3 dpf larvae, although six-month-old mfap4∆/∆ mutants had significantly fewer macrophages as compared to age-matched WT fish (Fig. 3c), macrophage recruitment to the amputation site was not impaired (Fig. 3d). Regenerated fins in both adult WT and mfap4∆/∆ mutant fish following amputation were moreover indistinguishable (Fig. 3e). These findings suggest that loss of mfap4 quantitatively impacts the differentiation of macrophages from circulating monocytes, but those mfap4-deficient macrophages that do emerge show no defects in their recruitment to the site of injury.

Figure 3
figure 3

mfap4 mutant fish have reduced numbers of macrophages. (a) The adult tail fins of wild-type (WT, mfap4+/+) and mfap4∆/∆ mutant fish were amputated (red dotted line). Cells in the remaining part of the tail fin (blue dotted box) were dissociated and macrophage numbers quantitated by FACS using the Tg(mpeg1:EGFP) fluorescent reporter. (b) FACS analysis depicting side scatter (SSC) on the y-axis and Tg(mpeg1:EGFP) fluorescence on the x-axis. (c) Graphical representation of mean macrophage numbers of fifteen biological replicates per genotype [determined as in panel (b)] for various time points post amputation. At most analyzed time points, there are significantly fewer macrophages in mfap4∆/∆ tail fins as compared to the WT fins (0 days post amputation (dpa), P = 0.0043; 1 dpa, P = 0.0264; 4 dpa, P = 0.0049; 7 dpa, P = 0.0892; 11 dpa, P = 0.0235; 14 dpa, P = 0.0145; assessed by Mann–Whitney-U test). (d) Quantification of macrophage numbers in tail fins at the time points shown in (c) relative to uncut fins. There is no delay in the recruitment of macrophages. Error bars in (c) and (d) represent the standard error of the mean (s.e.m.) of biological replicates. (e) Representative images of regenerated WT and mfap4∆/∆ tail fins two months post amputation. Images in panel (b) were generated with FlowJo V10 software (FlowJo LLC). Graphs in panels (c) and (d) were generated with Microsoft Excel (for Mac 2011, version 14.7.7). This figure was created with Inkscape software version 0.92 (available at https://www.inkscape.org/).

Full size image

Previous studies in mice uncovered important interactions between macrophages and neutrophils during wound healing34,35. Neutrophils are the first leukocytes recruited to the site of injury where they migrate from blood vessels to promote inflammation and to destroy foreign material by phagocytosis36. We therefore assessed neutrophil count in mfap4∆/∆ mutant larvae after tail fin amputation using whole-mount immunofluorescence for the neutrophil-specific marker Mpx (Myeloid-specific peroxidase). Neutrophil numbers were four-fold higher in mfap4∆/∆ mutant larvae as compared to WT larvae prior to amputation (Fig. 4b,c, 0 hpa). Upon tail fin primordium amputation, a significantly larger number of neutrophils was recruited to the site of injury in mfap4∆/∆ larvae as compared to WT (Fig. 4b,c, later time points). However, the recruitment of neutrophils to the amputation site follows similar dynamics in WT and mutant larvae (Fig. 4d). In conclusion, mfap4∆/∆ mutant zebrafish have significantly reduced numbers of macrophages with a concomitant increase in the number of neutrophils.

Figure 4
figure 4

Neutrophil recruitment to the amputation site is increased in mfap4 mutant fish. (a) The tail fin primordium of 3 dpf wild-type (WT, mfap4+/+) or mfap4∆/∆ mutant larvae was amputated (red line) and subsequently analyzed by whole-mount immunofluorescence with the neutrophil marker mpx (black box). (b) Representative images of various time points post amputation of WT and mfap4∆/∆ fins. White dotted lines outline the amputation site. (c) Graphical representation of mean neutrophil numbers in the imaged area near the amputation site [as shown in (b)] of ten biological replicates per genotype for various time points post amputation. Quantification of neutrophil numbers reveals significantly more neutrophils are recruited to the amputation site in mfap4∆/∆ mutants as compared to WT larvae at most analyzed time points (0 hours post amputation (hpa), P = 0.0021; 0.5 hpa, P = 0.0002; 1 hpa, P < 0.0001; 2 hpa, P = 0.1051; 3 hpa, P < 0.0001; 6 hpa; P = 0.0002; assessed by Mann–Whitney-U test). (d) Quantification of neutrophil numbers in the larval tail near the amputation site at the time points shown in (c) relative to uncut tails. The recruitment of neutrophils is not accelerated in mfap4∆/∆ mutants. Error bars in (b) and (c) represent the standard error of the mean (s.e.m.) of ten biological replicates. Graphs in panels (c) and (d) were generated with Microsoft Excel (for Mac 2011, version 14.7.7). This figure was created with Inkscape software version 0.92 (available at https://www.inkscape.org/).

Full size image

Loss of mfap4 alters haematopoiesis in zebrafish

The altered macrophage and neutrophil counts in mfap4∆/∆ mutant fish suggest that mfap4 plays a role during haematopoiesis. In zebrafish, haematopoiesis consists of two distinct phases—the primitive and definitive waves30. During early somitogenesis, ventral mesoderm differentiates into hemangioblasts31. The primitive wave begins around the 3 ss when hemangioblasts undergo differentiation and specification events to generate common myeloid progenitors (CMPs, Fig. 5a), which give rise to primitive erythrocytes and granulocytes/monocytes32. Primitive erythrocytes promote tissue oxygenation and further embryonic development, while primitive macrophages protect the embryo against invading microorganisms in the presence of injury30. Commitment to either the erythroid or the myeloid lineage is regulated by the transcription factors Gata1a (GATA binding protein 1a) and Spi1b/Pu.1, respectively37. Previous reports indicated that Gata1a and Spi1b/Pu.1 physically interact, resulting in reciprocal inhibition of their respective transcriptional targets30. During the primitive wave at the 10–14 ss, gata1a is highly enriched in erythroid progenitors, and consequently spi1b/pu.1 expression in myeloid progenitors is repressed (Fig. 5a). However, in mfap4∆/∆ mutants, the expression of gata1a is downregulated, while spi1b/pu.1 is upregulated three-fold (Fig. 5b). Consistent with this observation, the expression of the early macrophage commitment marker l-plastin/lpc1 (lymphocyte cytosolic protein 1) showed a 14-fold increase in mfap4∆/∆ mutants (Fig. 5c)38. However, there is no statistically significant change in the expression of the neutrophil-specific marker mpx (Fig. 5d). Next, we analyzed the expression of a suite of hemangioblast marker genes including gata2a, lmo2 (LIM domain only 2) and tal1/scl (stem cell leukemia T-cell acute lymphocytic leukemia 1), all of which were unaltered in mfap4∆/∆ mutant larvae at the 10 to 14 ss when compared to WT (Fig. 5e)39. Taken together, these findings suggest that Mfap4 plays a role in development of the myeloid lineage during the primitive wave of haematopoiesis.

Figure 5
figure 5

Loss of Mfap4 alters myeloid haematopoiesis in zebrafish. (a) Schematic of the primitive wave of zebrafish haematopoiesis (reviewed in47,48). (b–e) Expression of markers specific to the primitive wave in 10–14 somite stage WT (mfap4+/+) or mutant (mfap4∆/∆) embryos relative to the average expression of housekeeping genes (actb1 and eukaryotic translation elongation factor 1 alpha 1, like 1 (eef1a1)). In mutants, gata1a expression is significantly reduced (P = 0.000157; Student’s t-test), while expression of spi1b and lpc1 (P = 0.000147 and P = 0.000515, respectively; Student’s t-test) are increased. Expression of mpx, gata2a (P = 0.7 for both; Mann–Whitney-U test), lmo2 and tal1 (P = 0.855612 and P = 0.254594, respectively; Student’s t-test) are not significantly affected by loss of Mfap4. (f) Schematic of the definitive wave of zebrafish haematopoiesis (reviewed in47,48). (g–j) Expression of markers specific to the definitive wave in 7-day-old WT or mutant larvae, relative to the average expression of housekeeping genes (actb1 and eef1a1). At 7 dpf, expression of gata2a, runx1 and spi1b (P = 0.001146, P = 0.02048, and P = 0.002625, respectively; Student’s t-test) is significantly reduced. Expression of gata1a at 7 dpf is unaffected by Mfap4 loss (P = 0.136085; Student’s t-test). Expression of mpeg1.1 is considerably, yet not significantly, reduced in mutants (P = 0.1; Mann–Whitney-U test). Expression of mpx is significantly increased in mutants (P = 0.022413; t-test), while ikaros and cpa5 expression is significantly reduced (P = 0.001643 and P = 0.000204, respectively; Student’s t-test). Error bars represent the standard error of the mean (s.e.m.) of technical triplicates of twenty pooled embryos/larvae per genotype per time point. Abbreviations: HSC, hematopoietic stem cell; CMP, common myeloid progenitor; MEP, megakaryocyte/erythroid progenitor; GMP, granulocyte/macrophage progenitor. Graphs in panels (b–e) and (g–j) were generated with Microsoft Excel (for Mac 2011, version 14.7.7). This figure was created with Inkscape software version 0.92 (available at https://www.inkscape.org/).

Full size image

The transition from primitive to definitive haematopoiesis begins shortly after the start of blood circulation at 24 hpf33. The definitive wave generates hematopoietic stem cells (HSCs) that reside in the kidney marrow where they undergo self-renewal and differentiate into erythroid, myeloid and lymphoid lineages throughout the life of the fish (Fig. 5f)32. Runx1 (Runt-related transcription factor 1) is essential for the specification of HSCs from the hemogenic endothelium40, and has been reported to initiate HSC commitment through Notch signaling in a Gata2a-dependent manner41. At 7 dpf, we observed a significant reduction in the expression of runx1 and gata2a in mfap4∆/∆ mutant larvae (Fig. 5g). In addition to its role in HSC differentiation, Runx1 forms a negative feedback loop with Spi1b/Pu.1, controlling commitment to the macrophage versus neutrophil lineages42. High levels of Spi1b/Pu.1 promote macrophage development at the expense of neutrophil development42. Unlike its expression during the primitive wave, spi1b/pu.1 expression is reduced during the definitive wave in mfap4∆/∆ mutant larvae (Fig. 5h). Furthermore, Gata2a and Spi1b/Pu.1 are essential for early mast cell lineage commitment43. Consistent with this, the expression of the mast cell progenitor marker cpa5 (carboxypeptidase A5) is reduced in mfap4∆/∆ mutant larvae (Fig. 5j). Lymphoid progenitors that differentiate from HSC are labeled by ikaros (IKAROS family zinc finger 1), whose expression levels are also reduced in mfap4∆/∆ mutant larvae (Fig. 5i)44. As for myeloid cells, the neutrophil marker mpx has significantly higher expression in mfap4∆/∆ as compared to WT. mpeg1.1 mRNA levels trended lower in mutant larvae (20 sampled) but did not however achieve statistical significance (Fig. 5i). When taken together, our data suggest that the loss of mfap4 alters the balance between myeloid and lymphoid lineages during definitive haematopoiesis. Although gata1a expression is reduced during the primitive wave (Fig. 5b), gata1a expression is not significantly altered in mfap4∆/∆ mutant larvae (Fig. 5h). As the number of early erythroid cells, polychromatophils, is not altered significantly in adult mfap4∆/∆ mutants (Supplementary Fig. S4), the effect of mfap4 loss on erythropoiesis remains unclear.

Discussion

Apart from their crucial role in the innate immune response, macrophages and neutrophilic granulocytes are involved in wound healing and tissue regeneration28,45. For instance, previous studies have demonstrated that depletion of macrophages during different phases of wound healing impairs wound closure and tissue regeneration26. In this study, we show that in zebrafish loss of mfap4, which is conspicuously expressed in macrophages, results in an average of 1.5 times reduction of the total number of macrophages (Fig. 3) and a concomitant increase in the number of neutrophils (Fig. 4). Using a tail fin amputation model to investigate tissue repair, we further demonstrate that the recruitment of these two cell types to the injury site in mfap4∆/∆ mutant fish is unaffected and amputated fins regenerate normally (Fig. 3). Similar to our mfap4∆/∆ mutant zebrafish, other in vivo models in which macrophage differentiation is affected, such as interferon regulatory factor 8 (irf8) mutant fish, are viable and do not exhibit an overt phenotype46.

Our analysis of the expression of key marker genes (gata1a, gata2a, runx1 and spi1b/pu.1) suggests that Mfap4 plays a regulatory role during zebrafish haematopoiesis47,48. Spi1b/Pu.1 and Gata1a physically interact and function antagonistically to specify erythroid and myeloid cell fates30. Upon binding GATA1, human SPI1/PU.1 was previously shown to displace the transcriptional activator CREB-binding protein (CBP) to repress GATA1 target genes49. Our results show that during the primitive wave, spi1b/pu.1 expression increases in mfap4∆/∆ mutants, which we predict inhibits the expression of Gata1a target genes and tips the balance toward the macrophage lineage, as evidenced by increased levels of l-plastin/lpc1 mRNA (Fig. 5b,c). Jin et al.42 reported that high levels of Spi1b/Pu.1 promote macrophage lineage fate, while low levels of Spi1b/Pu.1 support neutrophil production42. Consistent with these findings, we observe that reduced levels of spi1b/pu.1 during definitive haematopoiesis are associated with an upregulation of neutrophil but downregulation of macrophage marker genes (Fig. 5g–i).

Gata2a and Runx1 play significant roles in the specification and differentiation of HSC upstream of the CMP (Fig. 5f)41. During definitive haematopoiesis, gata2a and runx1 expression was reduced in mfap4∆/∆ mutant larvae (Fig. 5g). Due to its sequence similarity with angiopoietin (Angptl), previous in vitro studies incorporated purified MFAP4 as a growth factor for the ex vivo expansion of mouse HSCs50. Like Mfap4, Angptl contains a C-terminal fibrinogen-like domain, and zebrafish Angptl2 interacts with Notch to in turn activate target genes such as gata2a41,51. These findings raise the intriguing possibility that these two extracellular proteins function similarly in vivo, playing a role in the development of definitive HSCs through the regulation of gata2a and runx1. In support of this, the Mfap4 relative Mfap2 induces the shedding of the Notch1 extracellular domain and activates Notch signaling in vitro52.

Alternatively, given the physical association between the MFAP family of proteins and fibrillins, who have been extensively described for their regulation of the bio-availability of assorted growth factors (GFs) (reviewed in53,54,55), it is possible that Mfap4 is similarly involved in the extracellular regulation of growth factor signaling. MFAP2 has indeed been shown to bind to active, but not latent, forms of members of the Transforming Growth Factor β (TGFβ) superfamily of signaling cytokines in vitro, including TGFβ1 and Bone Morphogenetic Protein 7 (BMP7)56,57. A connection between Mfap4 and the TGFβ signaling axis is intriguing given that we first identified MFAP4 as a potential Activin/Nodal target gene in differentiating hESC15,16. It is well established that in the early mouse embryo, Nodal signaling gradients are shaped not only by the auto-regulation of Nodal transcription itself but also by the direct activation of genes encoding secreted Nodal antagonists and binding proteins, such as Lefty1/2 and Cerberus58. Perhaps zebrafish Mfap4 also serves to refine TGFβ signaling by ligand sequestration or concentration. Additional molecular studies are nevertheless needed to elucidate how Mfap4 in the ECM impacts the regulation of hematopoietic genes in the nucleus.

The human MFAP4 protein localizes principally to elastic fibers located in adult blood vessels11. This is in striking contrast to zebrafish mfap4 where expression resolves to macrophages by 24 hpf. To date, there are no reports describing the spatiotemporal distribution of Mfap4 transcripts by WISH in the developing mouse embryo or enrichment of expression in mammalian macrophages. Thus, there appear to be prominent species-specific differences between the piscine and mammalian Mfap4 orthologs. These differences are further underscored by the lack of the RGD sequence in zebrafish Mfap4. Nevertheless, both zebrafish mfap4 and mouse Mfap4 loss-of-function mutations result in homozygous viability14,59. Interestingly, zebrafish mfap2, like mfap4, is expressed in the somites during segmentation60, which raises the possibility that the absence of a more pronounced mfap4 null mutant phenotype is due to compensation by mfap2 during early development. By 1 dpf, however, mfap2 transcripts label the caudal somites and the trunk mesenchyme as well as the connective tissue surrounding the dorsal aorta and posterior cardinal vein60, whereas mfap4 is expressed specifically in macrophages. Given the expression in these distinct tissue compartments, it is difficult to envision the rescue of mfap4∆/∆ mutants by Mfap2 at later stages of development. It is important to note however that in the mouse Mfap2 is widely expressed in mesodermally derived connective tissue61 and that inactivation of Mfap2 results in monocytopenia, with fewer circulating monocytes55. The molecular mechanism underlying this phenotype is as yet unknown. Lastly, the expression of mfap1, mfap3l and mfap5 remains poorly described (https://www.zfin.org) and invites additional experiments to understand the interrelationships of the remaining zebrafish Mfap family members.

Methods

Zebrafish breeding and maintenance

Zebrafish were bred and reared at the Institute of Molecular and Cellular Biology (IMCB) zebrafish facility in compliance with approved IACUC #161172 (“Breeding and maintenance of fish stocks for research use by IMB investigators and providing microinjection capabilities”) administered by the Agency for Science, Technology and Research (A*STAR) Biological Resource Centre (BRC) (https://www.a-star.edu.sg/brc/a-star-iacuc). Larvae were kept until 14 days post fertilization (dpf) in E3 medium (0.1 mM NaCl, 3.4 µM HCl, 6.6 µM CaCl2 and 6.6 µM MgSO4, calibrated to pH 7.4) at 28.5 °C before transfer into a closed water system. All experimental protocols were carried out in accordance with relevant guidelines and regulations.

Genotype analysis

Dermal scales were isolated from adult zebrafish and lyzed overnight at 56 °C in lysis buffer (100 mM Tris–HCl pH 7.5, 50 mM EDTA pH 8.0, 0.5% SDS and 0.5 mg/mL Proteinase K). Genomic DNA was extracted using phenol:choloform:isoamyl alcohol (Merck-Calbiochem, 516,726) and dissolved in nuclease-free water. GoTaq DNA Polymerase (Promega, M5123) was used for PCR amplification of extracted DNA. Primers used to amplify WT and mfap4∆150 genomic DNA (indicated in Fig. 2a) are listed in Supplementary Table S1.

Whole-mount in situ hybridization

Whole-mount in situ hybridization (WISH) was performed as previously described62. In brief, the mfap4 antisense riboprobe sequence was amplified from WT 1 dpf larval cDNA and subsequently in vitro transcribed using T7 polymerase. Primers used in Figs. 1 and 2 are listed in Supplementary Table S1. Images were taken with a Leica M80 stereomicroscope with Plan 1.0X lens and DFC425 5 megapixel camera, operated with Leica Application Suite (LAS) software version 4.9.

RNA extraction and quantitative PCR

At desired embryonic stages, 20 dechorionated embryos were pooled per time point, homogenized in TRIzol (Ambion, 15,596,018) using a 23G needle, chloroform/isopropanol precipitated and resuspended in RNase-free water. Purified RNA samples were DNase I (New England Biolabs, M0303S) treated prior to cDNA synthesis. The High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, 4,368,814) was used to synthesize 2 µg of RNA into cDNA. Quantitative PCR (qPCR) reactions were carried out using SYBR Select Master Mix (Applied Biosystems, 4,472,919). A two-sided Student’s t-test was used to determine statistical significance. Primers used are listed in Supplementary Table S1.

Tissue dissociation and flow cytometry

The Tg(mpeg1:EGFP) fluorescent macrophage reporter was previously described (https://zfin.org/ZDB-TGCONSTRCT-120117-1)29 and was crossed into the mfap4∆/∆ background. Subsequent generations were identified by Tg(mpeg1:EGFP) fluorescence and genotyped for mfap4∆ as described above. Adult zebrafish were anaesthetized in Tricaine, and tail fins carefully cut within 3–5 mm from the fin cleft under a dissecting microscope. Amputated adult tail fins or entire larvae were dissociated in Hank’s solution containing 100 mg/mL collagenase type IV and 30 mg/mL protease type XIV for 1 h at 30 °C with agitation (600 rpm). Samples were gently triturated with an 18G needle every 15 min. Cells were further dissociated with 0.05% trypsin–EDTA (Gibco, 25,200,056) for 10 min under the same conditions. Prior to flow cytometry, dissociated cells were rinsed and quenched with 2% fetal bovine serum (FBS) and 2 mM EDTA in 1X PBS. For both adult tail fins and entire larvae, the number of events was set at 100,000. Data analysis was performed with FlowJo V10 software (FlowJo LLC).

Whole-mount immunofluorescence

Whole-mount immunofluorescence was performed with an anti-Mpx antibody (GeneTex, GTX128379) at a 1:250 dilution and a goat anti-rabbit Alexa488-conjugated secondary antibody (Thermo Fisher Scientific, A21212) at a 1:500 dilution. Fin images were captured at 10X with an Olympus IX81 confocal microscope with FV1000 scanner head. GFP-positive cells were counted with Fiji software (ImageJ version 2.0.0-rc-15/1.49 k). ImageJ was used for quantification.

Statistics

Mean values of technical triplicates of RT-PCR ΔΔ Ct values, and of biological replicates of flow cytometry counts and fluorescence microscopy neutrophil count were assessed per genotype for normal distribution by Shapiro–Wilk test and subsequently assessed for statistically significant differences between genotypes (per time point, where applicable) by either two-sampled, non-pooled, two-tailed Student’s t-test following t-distribution, or two-tailed Mann–Whitney-U test (Statistics Study version 4.31; Statext LLC, Carlstadt, New Jersey, USA). For all tests, a P-value < 0.05 was used as cut-off point to reject H0.