Live imaging of Fibronectin 1a-mNeonGreen and Fibronectin 1b-mCherry knock-in alleles during early zebrafish development

Abstract

Within the developing embryo, cells assemble and remodel their surrounding extracellular matrix during morphogenesis. Fibronectin is an extracellular matrix glycoprotein and is a ligand for several members of Integrin adhesion receptor family. Here, we compare the expression pattern and loss of function phenotypes of the two zebrafish fibronectin paralogs fn1a and fn1b. We engineered two fluorescently tagged knock-in alleles to facilitate live in vivo imaging of the Fibronectin matrix. Genetic complementation experiments indicate that the knock-in alleles are fully functional. Fn1a-mNeonGreen and Fn1b-mCherry are co-localized in ECM fibers on the surface of the paraxial mesoderm and myotendinous junction. In 5-days old zebrafish larvae, Fn1a-mNeonGreen predominantly localizes to the branchial arches, heart ventricle, olfactory placode and within the otic capsule while Fn1b-mCherry is deposited at the pericardium, proximal convoluted tubule, posterior hindgut and at the ventral mesoderm/cardinal vein. We examined Fn1a-mNeonGreen and Fn1b-mCherry in maternal zygotic integrin α5 mutants and integrin β1a; β1b double mutants and find distinct requirements for these Integrins in assembling the two Fibronectins into ECM fibers in different tissues. Rescue experiments via mRNA injection indicate that the two fibronectins are not fully inter-changeable. Lastly, we examined cross-regulation between the two Fibronectins and find fn1a is necessary for normal Fn1b fibrillogenesis in the presomitic mesoderm, but fn1b is dispensable for the normal pattern of Fn1a deposition.

1. Introduction

Cells and extracellular matrix (ECM), along with water, are the primary components of animal tissues. Genetic studies demonstrate the importance of the ECM and their primary receptors, Integrins, for normal vertebrate development. The large multidomain glycoprotein Fibronectin is secreted as a soluble dimer and is assembled into an insoluble ECM by Integrins on the cell surface (Schwarzbauer and DeSimone, 2011). Fibronectin mutant mice exhibit neural tube and somite defects, and these embryos die around E8.0–8.5 (George et al., 1993; Georges-Labouesse et al., 1996). Integrins are heterodimeric transmembrane proteins with one α and one β subunit. Integrin α5β1 is one of the most important Fibronectin receptors and binds the amino acid sequence RGD as well as at a second “synergy site” in Fibronectin (Hynes, 2002). Targeted disruption of integrin α5 in mice results in pronounced defects in the posterior trunk (Yang et al., 1993) while loss of function of mouse integrin β1 causes an early developmental arrest at day E5.5 (Bouvard et al., 2001; Fassler and Meyer, 1995; Stephens et al., 1995). Zebrafish have two fibronectin genes, fn1a and fn1b (Howe et al., 2013; Sun et al., 2005). fn1a mutants have defects in heart development as well as abnormal somites in the anterior trunk, yet some mutants survive to adulthood (Koshida et al., 2005; Trinh and Stainier, 2004). fn1b mutants are largely normal and viable, but fn1; fn1b double mutant embryos have defects in the morphogenesis of all somites (Guillon et al., 2020) and do not survive past larval stages. integrin α5 mutant zebrafish have defects in anterior somites and in craniofacial development and die by 10 days post fertilization (dpf) (Crump et al., 2004; Julich et al., 2005; Koshida et al., 2005). Zebrafish integrin β1a; β1b mutants exhibit severe morphological defects (Yamaguchi et al., 2022) and are not viable past 5 dpf.

The dynamics of Fibronectin fibril assembly and fibril remodeling under live conditions have been visualized using murine cell-lines that express Fibronectin endogenously tagged with green fluorescent protein (GFP) either within the FNIII repeats in the center of the protein (Ohashi et al., 1999; Ohashi et al., 2002) or at the N-terminus (Tomer et al., 2022). Fibronectin matrix dynamics have been visualized in living Xenopus and avian embryos using fluorochrome-labeled anti-Fibronectin antibodies (Davidson et al., 2008; Filla et al., 2004). In zebrafish, we previously created a Fibronectin-GFP using the tagging strategy of Ohashi et al. and visualized Fibronectin matrix dynamics in the paraxial mesoderm in transgenics under control the tbx6l enhancer/promoter as well as ubiquitous expression via mRNA injection (Guillon et al., 2020; Julich et al., 2015). We used a transgenic zebrafish in which Fibronectin was tagged with a photoconvertible protein and under control of a heat shock promoter to visualize the pattern of matrix remodeling at the interface between the neural tube and paraxial mesoderm (Guillon et al., 2020).

In this study, we report the engineering of zebrafish knock-in alleles labeling endogenous Fn1a and Fn1b with mNeonGreen and mCherry, respectively. Double homozygotes for either allele behave as wild-type alleles indicating that the knock-ins are fully functional. fn1a and fn1b have both overlapping and distinct patterns of transcription in the developing zebrafish, and the most commonly used anti-Fibronectin polyclonal antibodies recognize both zebrafish Fibronectins. Thus, we use the Fn1a-mNeonGreen and Fn1b-mCherry knock-in alleles to define the paralog-specific ECM fibril patterns during zebrafish embryogenesis and early organogenesis. We characterize the ECM patterns in maternal zygotic integrin α5 mutants (MZitgα5−/−) and integrin β1a; integrin β1b double mutants (itgβ1a−/−; β1b−/−) and find distinct requirements for each Integrin in assembling the two Fibronectins into ECM fibers in different tissues. Compared to human FN1, Fn1a has a polymorphism in the amino acid sequence near the “synergy site” which is bound by Integrin α5β1. Zebrafish cells do not adhere well to human FN1, and this polymorphism has been proposed to underlie this difference (Mould et al., 2009). Fn1b has the same amino acid as human FN1 at this residue, thus we tested whether this polymorphism is responsible for the functional difference between fn1a and fn1b. However, rescue experiments using chimeric fn1a/fn1b constructs indicate that this polymorphism is neither necessary nor sufficient to account for the functional differences between Fn1a and Fn1b. Lastly, we examined cross-regulation between the two Fibronectins and find fn1a is necessary for normal Fn1b fibrillogenesis in the presomitic mesoderm, but fn1b is dispensable for the normal pattern of Fn1a deposition.

2. Methods

2.1. Animal husbandry

Zebrafish, Danio rerio, were raised, maintained and handled according to standard protocols (Nüsslein-Volhard and Dahm, 2002) approved by Yale University’s Institutional Animal Care and Use Committee.

2.2. Zebrafish strains

Tü and TL were used as wild-type lines. Mutant alleles include itgα5^thl030, fn1a^ya13Tg, fn1b^ya14Tg (Guillon et al., 2020; Julich et al., 2005). The itgβ1a; itgβ1b mutant line was a generous gift from the Holger Knaut laboratory (Yamaguchi et al., 2022). itgβ1a^−/−; itgβ1b^−/− embryos were derived from incrosses of itgβ1a^−/−; itgβ1b^−/+ adults and were sorted based on phenotype. Itgα5^−/− embryos can be rescued to adulthood by mRNA injection of wild-type itgα5 mRNA at the one cell stage (Julich et al., 2009). Therefore, itgα5 homozygous mutants, as well as fn1a and fn1b homozygous mutants, were derived from homozygous mutant adults. fn1a^−/−; fn1b^−/− embryos were collected from incrosses of fn1a^−/+; fn1b^−/− adults. To generate fn1a^−/−; fn1b^−/− mutant adults, we incrossed heterozygous fn1a; homozygous fn1b mutant adults that carried the Tg(hsp70: fn1a-mKikGR13.2) cassette, heat-shocked the embryos at 8–10 somites for 40 minutes at 38°C and raised them. The adults were then genotyped for fn1a.

The fn1a: fn1a-mNeonGreen knock-in line (fn1a^ya15Tg) was generated in the Tü background and outcrossed to TL and Tü; the fn1b: fn1b-mCherry knock-in line (fn1b^ya16Tg) was generated and maintained in the TL background.

Embryos imaged at the 10–12 somites stage and 18 somites stage were raised at 28.6°C for 4 hours, then cooled to 23.6°C for 3 hours and thereafter cooled to 22°C to slow development. All others were raised at 28.6°C.

2.3. Generation of fibronectin knock-in lines

fn1a: fn1a-mNeonGreen (fn1a^ya15Tg): Based on the recommendation by Tomoo Ohashi, we chose the same insertion site for the fluorophore into the fn1a locus as in Tg(tbx6l: fn1a-GFP) (Julich et al., 2015; Ohashi et al., 1999). gRNA target sites were identified using ChopChop (https://chopchop.cbu.uib.no/) and CRISPOR (http://crispor.tefor.net/), and two gRNA target sequences closest to the insertion site tested for cutting efficiency, ultimately opting for the target site 5’-GGACTCCAGGTTCAGATCAGtgg-3’. The guide DNA template was generated as described in (Gagnon et al., 2014) and gRNA synthesized with HiScribe T7 according to the manufacturer’s instruction (NEB E2040S).

Approximately 800 bp of genomic sequence flanking the insertion site within exon 23 of fn1a were amplified from Tü genomic DNA via PCR and verified by sequencing. These homology arms were then ligated together with mNeonGreen into the pKHR4 plasmid (a gift from David Grunwald (Hoshijima et al., 2016)). The gRNA target sequence was added to both ends of the cassette. Silent mutations were introduced into the gRNA target sequence within the cassette to prevent cutting: 5’-GctCTCCAGaTTCAGgTCtGTtG-3’.

fn1b: fn1b-mCherry (fn1b^ya16Tg): To identify the locus for insertion of the fluorophore, the protein sequence of Fn1b was aligned with Fn1a (DNASTAR MegAlign) and, based on sequence homology, determined to be within exon 23; plasmid construction and gRNA generation were in principle the same as for fn1a. The gRNA target sequence 5’-GGCTCCACATTTAGATCAGTTGG-3’ was selected, with the silent mutations 5’-GGtTCtACgTTTAagTCcGTgGG-3’ introduced to prevent cleavage by the gRNA/Cas9 complex.

Prior to injections, 335 ng/μl Cas9 protein (PNA Bio CP01) was gently mixed with ~150 ng/μl gRNA and incubated at 37°C for 10 minutes to allow complex formation, then placed on ice and 50 ng/μl plasmid and phenol red added. The mixture was injected at the one-cell stage within 20 minutes of fertilization. Injected embryos were incubated at 28°C and sorted for fluorescence at the 20 somites stage. Embryos with fluorescence resembling the published FN immunostaining were raised to adulthood. Founder fish were identified by crossing to wild-type and fluorescent F1 embryos were raised to establish stable lines.

2.4. pCS2+ fn1a-sfGFP and pCS2+ fn1b-mCherry plasmid construction

Due to the size of fn (>7 kbps) we amplified the coding sequences for each gene in two fragments from Tü cDNA for fn1a and TLF cDNA for fn1b. Primer sequences: fn1a N-terminal fragment fwd 5’-atgtttggtggccctttg-3’, rev 5’-tggagaaagaggagtgac-3’, C-terminal fragment fwd 5’-cctctttctccacccactgatctgaacctg-3’, rev 5’-aattagtactcccgatacccg-3’; fn1b N-terminal fragment fwd 5’-atgacccgtgagtcagtaaagag-3’, rev 5’-tggggataaggatgtgacaac-3’, C-terminal fragment fwd 5’-tccttatccccaccaactgatc-3’, rev 5’-tcactcttgggttttgggattg-3’. Please note that the fwd primer for the C-terminal fragments includes the duplication of the last 4 amino acids of the N-terminal fragment. Together with PCR-amplified sfGFP, the two fn1a PCR fragments were ligated into the EcoRI and XbaI sites of pCS2+ to create an EcoRI-fn1a (bps 1–3525)-SalI-sfGFP-EcoRV-PLSP (duplication)-fn1a (bps 3526–7443)-XbaI insertion. The sequence deviates from fn1a NM_131520_2 as follows: aa 58 K to Q, aa 1800 S to L, aa 2276 Q to E and aa 2475 H to Q. The two fn1b fragments were, together with mCherry, ligated into the BstBI and XhoI sites of pCS2+ via Gibson Assembly (BstBI-NheI-fn1b (bps 1–3531)-SalI-mCherry-SacII-SLSP (duplication)-fn1b (3532–7218)-XhoI.

The sequence for fn1b used here differs from fn1b NM_001013261_1 in the following: aa 294 R to H, aa 417 A to T, aa 616 G to V, aa 762 R to Q, aa 860 G to E, aa 913 P to L, aa 938 C to R, aa 1060 G to D, aa 1229 E to D (this results in a second RGD domain upstream of aa 1527–1529 RGD), aa 1267 D to deletion, aa 1624 K to E, aa 1638 Y to N, aa 1647 L to F, aa 1653 N to T, aa 1661 G to V, aa 1682 K to R, aa 1786 V to I, aa 2153 V to F, aa 2157 P to S, aa 2160 P to L, aa 2161 A to V, aa 2163 G to C, aa 2322 W to C, aa 2331 M to I and aa 2341 R to G.

2.5. mRNA synthesis

10 μg each, pCS2+ fn1a-sfGFP and pCS2+ fn1b-mCherry, were linearized with NotI-HF, extracted with phenol/chloroform and isoamyl alcohol (Sigma P2069), and precipitated with 1/10 volume 3M sodium acetate and 2.5 volumes 100% ethanol. After a 2 hours incubation at −20°C, the template was pelleted by centrifugation for 10 minutes at 4°C, washed with 70% ethanol, dried and resuspended in 8 μl RNase free water. The transcription reaction with the mMESSAGE mMACHINE SP6 (Ambion AM1340) was supplied with 1 μl GTP (for a 20 μl reaction) provided in the kit and incubated for 2 hours at 37°C. Enzyme, buffer and unincorporated nucleotides were removed from the mRNA via SpinColumns (BioRad 732–6250). Approximately 350 ng/μl mRNA was injected into one-cell stage embryos.

2.6. in situ hybridization

DNA templates for fn1a and fn1b digoxygenin-labeled antisense RNA probes were amplified from cDNA with primers fn1a fwd 5’-ctgggactgtacttgcattgg-3’, rev 5’-TAATACGACTCACTATAGaaacgcaggatactgatgagc-3’; fn1b fwd 5’-tcagccatctgaaattagtcg-3’, rev 5’-TAATACGACTCACTATAGcgttgaatcatgaccagtagg-3’ (T7 binding site in capital letters). The hybridization procedure followed established protocols (Thisse and Thisse, 2014).

2.7. Immunohistochemistry

Embryos were fixed in ice-cold 4% PFA at 26 hpf and incubated overnight at 4°C. After two rinses in PBS plus 1% DMSO and 0.1% Triton-X100 (‘PBSDT’) embryos were permeabilized by incubation in PBSDT plus 5 μg/ml proteinase K (Roche/Millipore Sigma 3115836001) for 8 minutes, rinsed twice in PBSDT and post-fixed in 4% PFA for 20 minutes at room temperature followed by two rinses in PBSDT and blocking for 2 hours with 1% blocking solution (Roche 11096176001) in PBSDT. The primary antibody (rabbit anti-FN Sigma-Aldrich F3648) was diluted 1:100 in PBSDT/block and incubated with the embryos overnight at 4°C. Next, the embryos were rinsed twice with PBSDT/block followed by two rinses in PBSDT, 15 minutes each. The secondary antibody Alexa Fluor 555 donkey anti-rabbit (Thermo Fisher Scientific A-31572) was diluted 1:200 in PBSDT and incubated with the embryos overnight at 4°C. The embryos were then rinsed 3 times in PBSDT, 10 minutes each and post-fixed with 4% PFA for 20 minutes, rinsed twice in PBSDT and taken though a 25%, 50% and 75% glycerol series.

2.8. Imaging

Brightfield images of embryos embedded in 3% methylcellulose and embryos from the in situ hybridization stains as well as immunohistochemistry mounted in 75% glycerol/PBST were captured on an Olympus MVX10 using its cellSens Standard software.

Confocal images were acquired on a Zeiss LSM 880 Airyscan NLO equipped with EC PlnN 10x/0.3, PlnApo 20x/0.8 and C Apo 40x/1.2W objectives. Embryos were embedded in 1.2 % low-melt agarose (American Bioanalytical AB00981, dissolved in E2) on a 24X50–1.5 microscope cover glass (Fisher Scientific 12544E), covered with a 22X22–1.5 cover glass (Fisher Scientific 12541B) with putty on all four corners as spacer and surrounded by E2. Airyscan images were processed in ZEN 2.3 SP1 with standard parameters provided by Zeiss. Embryos older than 24 hpf were immobilized with 200 μg/ml Tricaine-S (Syndel/Western) added to the E2.

Images were adjusted for brightness and contrast and cropped in ImageJ/Fiji. 3D confocal projections were processed in IMARIS (Bitplane). 3–5 embryos were imaged for all genetic backgrounds at the indicated stages.

3. Results and Discussion

3.1. Comparison of loss of function, mRNA expression and immunolocalization of zebrafish Fibronectins

Most amniote/tetrapod genomes contain a single gene for Fn. However, the sequencing of the zebrafish genome identified two genes subsequently named fn1a and fn1b (Howe et al., 2013; Sun et al., 2005). These two paralogs likely arose from a single common ancestor during the teleost-specific whole genome duplication 320–350 million years ago (Postlethwait et al., 1998).

To initiate a comparison of these two paralogs, we first evaluated the loss of function phenotypes of MZfn1a^−/−, MZfn1b^−/− and double homozygous mutant embryos at 8–10 somites stage, ~18 somites stage, ~26 hours post-fertilization (hpf) and 5-days post-fertilization (dpf). The mutant lines were previously generated via a Stop-Codon cassette insertion into exon 4 (fn1a) or exon 5 (fn1b) using CRISPR/Cas9 (Gagnon et al., 2014; Guillon et al., 2020). MZfn1a^−/− embryos fail to maintain somite boundaries in the anterior trunk and exhibit u-shaped myotome boundaries (Julich et al., 2005; Koshida et al., 2005; Snow et al., 2008) (see also Fig. 5C). By 5 dpf we observed three main groups of phenotypes. One group of embryos developed severe edema, failed to inflate the swim bladder, exhibited extensive tissue necrosis and died by 10 dpf (54%). A second group failed to inflate the swim bladder and developed less severe edemas (29%), and a third group of embryos that exhibited no gross morphological defects and were viable (18%) (3 experiments, n=180) (Fig. 1 B–B”’). In contrast, MZfn1b^−/− appeared wild-type at all stages examined (Fig. 1 C–C”’). Double homozygous mutants derived from fn1a^+/−; fn1b^−/− adults developed slower than their single mutant counterparts and displayed a border maintenance defect in all somites. fn1a^−/−; MZfn1b^−/− embryos developed severe edema by 5 dpf, accompanied by extensive tissue necrosis (Fig. 1 D–D”’). Simultaneous loss of function of both genes results in 100% embryonic lethality. We observed a similar phenotype for double homozygous mutants derived from fn1a^−/−; fn1b^−/− adults (data not shown). The phenotype of MZfn1a^−/−; MZfn1b^−/− embryos at the 10–12 somites stage is similar to fn1a^−/−; MZfn1b^−/−, and both phenotypes are less severe than the morpholino-mediated double knockdown of fn1a and fn1b (Julich et al., 2005). This difference may stem from transcriptional adaptation, though there are no other fibronectin paralogs in the zebrafish genome, or may be due to off-target effects of the morpholinos (El-Brolosy et al., 2019; Kok et al., 2015).

Fig. 5.

Effects of loss of function of one fibronectin paralog on the localization of the ECM of the other paralog.

Endogenous localization of (A-G) Fn1b-mCherry in MZfn1a^−/− mutant embryos. (A, B) 10–12 somites stage, (C) 26 hpf and (D-G) 5 dpf. (A) Lateral view. (B) Dorsal anterior PSM. (B) White arrow: gaps in the fibril meshwork. Orange arrow: nascent somite boundary. (C) Lateral view of myotome boundaries. White arrow: u-shaped myotome boundary. (D) Ventral view of the head. (E) Lateral view of the branchial arches and otic vesicle. White arrow: proximal convoluted tubule. Orange arrow: epidermis. (F) Lateral view of the mid-trunk. Yellow arrow: epidermis. White arrow: ventral mesoderm/posterior cardinal vein. (G) Digitally reconstructed transverse section through the trunk anterior to the urogenital opening. White arrow: axial vein. Yellow arrow: epidermis. (H-N) Endogenous localization of Fn1a-mNeonGreen in MZfn1b^−/− mutant embryos at three developmental stages. (H, I) 10–12 somites stage, (J) 26 hpf and (K-N) 5 dpf. (H) Lateral view. (I) Dorsal anterior PSM. White arrow: nascent somite boundary. White circle: radial fibril alignment. (J) Lateral view of myotome boundaries. (K) Ventral view of the head. White arrow: heart ventricle. Pink arrow: olfactory organ. Yellow arrow: retina. (L) Lateral view of the branchial arches and otic vesicle. White arrow: matrix spanning the lumen of the ear. Cyan arrow: pectoral fin. Magenta arrow: branchial arches. (M) Lateral view of the mid-trunk. White arrow: dorsal aorta. Purple arrow: urogenital opening. (N) Digitally reconstructed transverse section through the trunk anterior to the urogenital opening. (A-N) 3D projections of confocal z-stacks. (A-F) (H-M) Anterior to the right. Scale bars: (H, K-N) 100 μm, (I, J) 20 μm.

Fig. 1.

Loss of function phenotype, expression and immunolocalization of zebrafish fn1a and fn1b.

(A-D”’) Brightfield images of wild type, MZfn1a^−/−, MZfn1b^−/− and fn1a^−/−; MZfn1b^−/− embryos at (A-D) 6–8 somites stage, (A’-D’) 18 somites stage, (A”-D”) 26 hpf and (A”’-D”’) 5 dpf. (B”’) Insert: most severe phenotype. (E-H) fn1a expression at the 10–12 somites stage in (E) wild type, (F) MZfn1a^−/−, (G) MZfn1b^−/− and (H) fn1a^−/−; MZfn1b^−/−. The expression is greatly reduced in the double mutant. (E) Inserts: flat-mounted embryos showing the expression in the anterior lateral plate mesoderm, the notochord and paraxial mesoderm. (I-L) fn1b expression at the 10–12 somites stage in (I) wild type, (J) MZfn1a^−/−, (K) MZfn1b^−/− and (L) fn1a^−/−; MZfn1b^−/−. (I) Inserts: flat-mounted embryos showing the strong expression throughout the paraxial mesoderm. (M-P) FN immunolocalization at 26 hpf. Head and yolk manually removed. (M) wild type, (N) MZfn1a^−/−, (O) MZfn1b^−/− and (P) fn1a^−/−; MZfn1b^−/− embryos. Fn1a and Fn1b are both recognized by these polyclonal antibodies. Immunofluorescence is not observed in the double mutant. (A-P) Lateral view with anterior to the right. Inserts in E and I: dorsal view with anterior to the top.

Next, we evaluated the expression pattern of fn1a and fn1b at the 10–12 somites stage. fn1a is expressed in the notochord along the anterior-posterior axis, in a receding fashion within the paraxial mesoderm with the strongest expression at the posterior, in the anterior lateral plate mesoderm/endocardial progenitors and hatching gland (Fig. 1E inserts) and in the epidermis covering the yolk (see also (Koshida et al., 2005; Sakaguchi et al., 2006; Thisse et al., 2004; Trinh and Stainier, 2004)). fn1b is transcribed throughout the paraxial mesoderm and tailbud, but all other tissues are devoid of detectable transcripts (Fig. 1I inserts) (Baxendale et al., 2009). Despite the strong expression of fn1b throughout the paraxial mesoderm, loss of function of fn1b alone does not perturb somite or myotome boundaries (Fig. 1C’, C”; see also Fig. 5J).

We next assayed the effect of the loss of function of either fn gene on their respective mRNA expression patterns. fn1a transcripts can be observed in the anterior lateral plate and posterior paraxial mesoderm of MZfn1a^−/− albeit at a much lower level (Fig. 1F). The expression in the notochord appears absent. fn1a mRNA expression levels in MZfn1b^−/− appear similar to wild type (Fig. 1G). fn1a^−/−; MZfn1b^−/− double homozygous mutant embryos exhibit low-level expression of fn1a within the anterior lateral plate mesoderm and posterior paraxial mesoderm (Fig. 1H). fn1b expression appears unaffected in MZfn1a^−/− (Fig. 1J), reduced in MZfn1b^−/− (Fig. 1K) and further reduced in fn1a^−/−; MZfn1b^−/− double mutant embryos (Fig. 1L).

To examine endogenous protein localization, we subjected wild type, MZfn1a^−/−, MZfn1b^−/− and fn1a^−/−; MZfn1b^−/− embryos at 24 hpf to immunohistochemistry using a rabbit-polyclonal antibody against human FN (Fig. 1M–P) (Chou et al., 2013; Crawford et al., 2003; Dray et al., 2013; Hayes et al., 2012). These antibodies recognize both Fibronectin paralogs, and thus preclude the ability to attribute protein localization specifically to one or the other paralog in wild-type embryos.

3.2. Endogenous Fn1a-mNeonGreen and Fn1b-mCherry localization

FN consists of several structural repeats termed FN type I, type II and type III (Schwarzbauer and DeSimone, 2011). We previously demonstrated that fluorescent proteins can be inserted between FN type III repeats 6 and 7 of the zebrafish Fn1a protein sequence without perturbing Fn1a function (Guillon et al., 2020; Julich et al., 2015; Ohashi et al., 1999; Ohashi et al., 2002).

We employed the same design principle to knock-in mNeonGreen into the fn1a genomic locus via CRISPR/Cas9 (Fig. 2A). To construct the donor plasmid, we amplified ~800 bps of genomic DNA adjoining the desired target site to serve as homology arms, ligated mNeonGreen in between the homology arms, and flanked the resulting cassette with the selected RNA guide target sequence. We chose a guide target sequence no more than 10 bps away from the target site. To avoid recutting of integrated DNA, we introduced silent mutations into the guide target sequences within the donor cassette. Cas9 protein complexed with the gRNA plus the donor plasmid were injected into one-cell stage embryos within 20 minutes of egg collection and, 24 hours after injection, visually screened for fluorescence similar to the published protein localization pattern of Fn. We implemented a similar methodology to insert mCherry into the fn1b genomic locus (Fig. 2A).

Fig. 2.

Endogenous localization of Fn1a-mNeonGreen and Fn1b-mCherry.

(A) Schematic of the domain structure of Fn1a and insertion site of the fluorescent protein mNeonGreen in Fn1a and mCherry in Fn1b. The Integrin-binding sequence RGD is located in Fn1a FN type III repeat 11, downstream of the synergy site (amino acid sequence indicated). Protein sequence detailed at the insertion site: FN type III-6 ends in VTPLS, followed by one residue of FN type III-7 (P), restriction sites BamHI and XmaI (GSPG), mNeonGreen MVSKG….DELYK, EcoRI restriction site (EF) and a duplication of the FN type III-6/7 interface (PLSP). mCherry in Fn1b is flanked by restriction sites BglII (RS) on its 5’ and XmaI (PG) on its 3’ end. (B-H”) 3D projections of confocal z-stacks showing localization of (B-H) Fn1a-mNeonGreen and (B’-H’) Fn1b-mCherry at three developmental stages in live embryos. (B”-H”) Overlay. (B-C”) 10–12 somites stage, (D-D”) 26 hpf and (E-H”) 5 dpf. (B-B”) Lateral view. (B) White arrow: deposition of Fn1a-mNeonGreen between the extending tail and the head. (C-C”) Dorsal view of the anterior PSM. (C-C’) White arrow indicates the nascent somite boundary. White circle indicates radial fiber alignment toward a central focal point. (D-D”) Lateral view of myotome boundaries atop the yolk extension. (E-E”) Ventral view of the head. (E) White arrow: heart ventricle. Pink arrow: olfactory placode. Yellow arrow: retina. (E’) White arrow: pericardium. Yellow arrow: jaw joint. (F-F”) Lateral view of the branchial arches and otic vesicle. Overlying epidermis has been partly digitally removed. (F) White arrow: matrix spanning the lumen of the ear. Cyan arrow: branchial arches. (F’) White arrow: proximal convoluted tubule (G-G”) Lateral view of the mid-trunk. (G) White arrow: urogenital opening. Orange arrow: floorplate. (G’) Cyan arrow: posterior cardinal vein region. White arrow: gut (H-H”) Digital transverse section through the trunk anterior to the urogenital opening. (H’) White arrow: gut. Cyan arrow: posterior cardinal vein/pronephric duct. (B-G”) Anterior to the right. Scale bars: (B”, E”, F”, G”/H”) 100 μm, (C”, D”) 20 μm

Embryos derived from incrosses of transgenic adults appear phenotypically wild type. However, given the variability of the fn1a^−/− mutant phenotype and the absence of a mutant phenotype in fn1b^−/− embryos, we performed complementation experiments in which we tested the ability of the knock-in alleles to rescue the fn1a^−/−; fn1b^−/− phenotype. fn1a^mNG/mNG; fn1b^−/− embryos are phenotypically wild type indicating that the fn1a-mNeonGreen allele is functional. Similarly, fn1a^−/−; fn1b^{mCherry/mCherry} embryos resemble fn1a^−/− embryos rather than fn1a^−/−; fn1b^−/− embryos indicated that the Fn1b-mCherry allele is functional. Hereafter, we refer to the knock-in lines as Fn1a-mNG and Fn1b-mCherry.

We next examined the localization of Fn1a-mNG and Fn1b-mCherry at the 10–12 somites stage, ~26 hpf and 5 dpf via live confocal imaging (Fig. 2B–H”). At the 10–12 somites stage, Fn1a-mNG is observed at the dorsal surface along the entire anterior-posterior axis of the embryo (Fig. 2B) and enriched over the yolk between the head and the extending tail (Fig. 2B white arrow). In contrast, Fn1b-mCherry localizes solely to the developing trunk (Fig. 2B’). Both proteins colocalize into the same ECM fibers and form a dense matrix on the surface of the paraxial mesoderm (Fig. 2C–C”) with some individual fibrils radially aligned to focal points (Fig. 2C, 2C’ white circle), a pattern previously reported in Xenopus (Davidson et al., 2008). This colocalization persists at the myotendinous junctions at 26 hpf (Fig. 2D–D”). By 5 dpf, differences in localization are more apparent. Fn1a-mNG is observed in the olfactory placode (Fig. 2E pink arrow), retina (Fig. 2E yellow arrow), the heart ventricle (Fig. 2E white arrow), branchial arches (Fig. 2F blue arrow) and the otic placode (Fig. 2F white arrow) (Haddon and Lewis, 1996). The localization is consistent with reports demonstrating a requirement of fn1a for heart, craniofacial and eye development (Crump et al., 2004; Gao et al., 2022; Hayes et al., 2012; Trinh and Stainier, 2004). The urogenital opening (Fig. 2G white arrow) and the floor plate (Fig. 2G orange arrow) are positive for Fn1a-mNG. Fn1b-mCherry is prevalent at the pericardium (Fig. 2E’ white arrow) and proximal convoluted tubule (Fig. 2F’ white arrow) (Gerlach and Wingert, 2013). We observe significant deposition of Fn1b-mCherry within the ventral mesoderm along the posterior cardinal vein/pronephric duct (blue arrow in Fig. 2G’, 2H’) and around the gut/intestine within the trunk (white arrow in Fig. 2G’, 2H’). It underlies the epidermis throughout the embryo (Kimelman et al., 2017), ensheathes the neurocoele (not shown) and localizes to the branchial arches albeit at a low level. Both proteins are enriched in the pectoral fins (not shown).

3.3. Fn1a-mNG and Fn1b-mCherry localization in MZitgα5^−/− and itgβ1a; itgβ1b^−/−mutant embryos

Fibronectin matrix assembly is mediated by the transmembrane Integrin adhesion receptor family, and the Integrin α5β1 heterodimer is considered the principal receptor for FN (Schwarzbauer and DeSimone, 2011). We extended our live imaging of Fn1a-mNG and Fn1b-mCherry to embryos deficient for Integrin α5 (MZitga5^−/−) and Integrin β1a; Integrin β1b (itgβ1a^−/−; β1b^−/−). MZitgα5^−/− mutants do not maintain somite boundaries (Fig. 3A, 3A’) (Julich et al., 2005; Koshida et al., 2005), form abnormal pharyngeal arch derivatives (Crump et al., 2004), do not inflate their swim bladder and develop edemas around the heart. Although the latter phenotype is variable (Fig. 3A”’), the mutation causes 100% embryonic lethality. In contrast, somite boundary formation and maintenance appear to occur normally in itgβ1a^−/−; β1b^−/− double mutant embryos at the 8–10 somites stage (Fig. 3I) but starts to become aberrant within the anterior somites by the 18 somites stage (Fig. 3I’). These double mutant embryos develop eyes, parts of the brain, a string-like heart, pigmentation and an anterior trunk by day 3 pf. However, the posterior body never develops (Fig. 3I”’) and 100% of the embryos die by day 5 (Yamaguchi et al., 2022).

Fig. 3.

Localization of Fn1a-mNeonGreen and Fn1b-mCherry in MZitgα5^−/− and MZitgβ1a^−/−; itgβ1b^−/− mutants.

Brightfield images of (A-A”’) MZitgα5^−/− homozygous mutants at (A) 8–10 somites stage, (A’) 18 somites stage, (A”) 26 hpf and (A”’) 5 dpf. (A”’) Insert: most severe phenotype. (B-H”) Endogenous localization of Fn1a-mNeonGreen and Fn1b-mCherry in MZitgα5^−/− mutant embryos at three developmental stages. (B-H) Fn1a-mNeonGreen. (B’-H’) Fn1b-mCherry. (B”-H”) Overlay. (B-C”) 10–12 somite stage, (D-D”) 26 hpf and (E-H”) 5 dpf. (B-B”) Lateral view. (C-C”) Dorsal anterior PSM at the nascent somite boundary. (C-C’) White arrow: thick cable-like fibril. (D-D”) Lateral view of myotome boundaries. White arrow: myotome boundary. (E-E”) Ventral view of the head. (E) Pink arrow: olfactory placode, white arrow: heart ventricle. (E’) White arrow: pericardium. (F-F”) Lateral view of the branchial arches and otic vesicle. (F) White arrow: matrix spanning the lumen of the otic placode. (F’) White arrow: epidermis. Green arrow: proximal convoluted tubule. (G-G”) Lateral view of the mid-trunk. (G’) White arrow: trunk cardinal vein region. (H-H”) Digital transverse section through the trunk anterior to the urogenital opening. (H’) White arrow: posterior cardinal vein. (B-H”) 3D projections of confocal z-stacks. (A-G”) Anterior to the right. Scale bars: (B”, E”, F”, G”/H”) 100 μm, (C”, D) 20 μm.

(I-I”’) Brightfield images of MZitgβ1a^−/−; itgβ1b^−/− double homozygous mutants at (I) 8–10 somites stage, (I’) 18 somites stage, (I”) 26 hpf and (I”’) 3 dpf. (J-P”) Endogenous localization of Fn1a-mNeonGreen and Fn1b-mCherry in MZitgβ1a^−/−; itgβ1b^−/− mutant embryos at three developmental stages. (J-P) Fn1a-mNeonGreen. (J’-P’) Fn1b-mCherry. (J”-P”) Overlay. (J-K”) 10–12 somite stage, (L-L”) 26 hpf and (M-P”) 3 dpf. (J-J”) Lateral view. (K-K”) Dorsal anterior PSM at the nascent somite boundary. (K-K’) White circle: radial fibril alignment. (L-L”) Lateral view of myotome boundaries. White arrow: myotome boundary. (M-M”) Ventral view of the head. (M’) Green arrow: epidermis. (N-N”) Lateral view of the branchial arches and otic vesicle. (O-O”) Lateral view of the mid-trunk. (O, O’) White arrow: myotome boundary. Green arrow: epidermis. (O’) Insert: magnified section of the epidermis overlying the yolk extension. (P-P”) Digital transverse section through the trunk anterior to the urogenital opening. (P’) Green arrow: epidermis overlying the yolk extension. (J-P”) 3D projections of confocal z-stacks. (I-O”) Anterior to the right. Scale bars: (J”, M”, N”, O”, P”) 100 μm, (K”, L”) 20 μm.

Instead of a dense mesh-like ECM, Fn1a-mNG and Fn1b-mCherry fibrils appear cable-like and sparsely span the anterior paraxial mesoderm medio-laterally in 10–12 somite old MZitgα5^−/− mutant embryos (Fig. 3C, C’ white arrow). Fn1a-mNG and Fn1b-mCherry are observed at the irregular and u-shaped myotome boundaries typical for MZitgα5^−/−(Fig. 3D, D’ white arrow). At 5 dpf Fn1a-mNG appears reduced within the olfactory placode (Fig. 3E pink arrow) while its localization to the heart ventricle (Fig. 3E white arrow) is unperturbed. Although the morphology of the branchial arches is abnormal in these mutants, Fn1a-mNG is present as it is within the otic placode (Fig. 3F white arrow). Similar to wild type, Fn1b-mCherry is prevalent in the proximal convoluted tubule (Fig. 3F’ green arrow), and it exhibits a punctate pattern underneath the epidermis throughout the embryo (white arrow in Fig. 3E’ and 3F’). It also localizes around the posterior cardinal vein (white arrow in Fig. 3G’ and 3H’).

At the 10–12 somites stage, Fn1a-mNG and Fn1b-mCherry colocalize and cover the paraxial mesoderm of MZitgβ1a^−/−; β1b^−/− mutant embryos with an architecture similar to wild type: a dense meshwork with some individual fibrils radially aligned to focal points (white circle in Fig. 3K and 3K’). Fn1a-mNG and Fn1b-mCherry concentrate at the myotome boundaries dorsal to the yolk extension at 26 hpf and highlight their abnormal, neither chevron- nor u-shaped, morphology in this double mutant background (white arrow in Fig. 3L and 3L’). On day 3 pf, many structures within the head such as branchial arches, heart atrium and ventricle and the otic placode are essentially absent in MZitgβ1a^−/−; β1b^−/−, and we observe minimal Fn1a-mNG localization (Fig. 3M, 3N). Some Fn1a-mNG and Fn1b-mCherry can be seen in the anterior trunk outlining myotome boundaries (white arrow in Fig. 3O, 3O’). The urogenital opening, posterior intestine, trunk artery and cardinal vein appear absent (Fig. 3O, 3O’). Fn1b-mCherry is noticeably enhanced at the epidermis throughout these embryos (green arrow in Fig. 3M’, 3O’, 3P’).

The differences in the MZitgα5^−/− and MZitgβ1a^−/−; β1b^−/− phenotypes and Fn1amNG/Fn1b-mCherry localization at the 10–12 somite was surprising as Itgα5 heterodimerizes with Itgβ1 to form the principal FN receptor. Co-IP and mass spectrometry also indicate that Itgβ1a and Itgβ1b are the predominant heterodimer partner for Itgα5 (Sun et al., 2021). However, zebrafish has several itgβ1 homologs (itgβ1a, itgβ1b, itgβ1b.1 and itgβ1b.2) which differ in their expression and it is possible that the other two paralogs partly compensate for itgβ1a and itgβ1b in the double mutants (Mould et al., 2006; Wang et al., 2014).

3.4. Functional differences between Fn1a and Fn1b

Aside from the different expression patterns, the fn1a and fn1b phenotypes suggest that there are functional differences between the two proteins. It has been reported that human Integrin α5β1 can bind zebrafish Fn1 while zebrafish Integrin α5β1 does not interact with human FN (Mould et al., 2009). The authors hypothesized that a highly conserved asparagine residue in the Integrin-binding region of human FN, which is replaced by a less bulky glycine in zebrafish Fn1a, causes a steric hindrance thus preventing human FN from binding to zebrafish Integrin α5β1. This amino acid polymorphism lies just downstream of the synergy site PHSRN (human).

We aligned the protein sequence of the Integrin-binding region for human FN with zebrafish Fn1a and Fn1b (Fig. 4A) and discovered that the glycine residue in Fn1a is an asparagine residue in Fn1b, the same residue as in the human protein (Fig. 4A underlain in blue). To test whether this polymorphism is responsible for the functional difference between Fn1a and Fn1b, we replaced the glycine residue in Fn1a with asparagine and replaced the asparagine in Fn1b with glycine. We injected in vitro synthesized mRNA encoding these chimeric Fibronectins into MZfn1a^−/− embryos to test for complementation. fn1a-sfGFP mRNA (Fig. 4B) and fn1a G1509N-sfGFP mRNA (Fig. 4C) both rescue myotome boundary morphology while neither fn1b-mCherry (Fig. 4D) nor fn1b N1420G-mCherry mRNA (Fig. 4E) did. We note that despite the failure to rescue myotome boundary morphology, the injection of both wild type and chimeric fn1b mRNA increased the viability of MZfn1a^−/− embryos (n=536; percentage of injected embryos that developed normally and inflated the swim bladder at 5 dpf: fn1a-sfGFP mRNA 76%; fn1a G1509N-sfGFP mRNA 86%; fn1b-mCherry mRNA 79%; fn1b N1420G-mCherry mRNA 64%; non-injected control 18%).

Fig. 4.

An amino acid variation near the synergy site is neither necessary nor sufficient to account for the functional difference between Fn1a and Fn1b.

(A) Alignment of human FN, zebrafish Fn1a and Fn1b protein sequence centered on the synergy site (yellow). Modified residues are highlighted in blue. Homologous residues are red. (B-E) Lateral view of myotome boundaries in the anterior trunk at 26 hpf in in vitro synthesized mRNA injected MZfn1a^−/− mutant embryos. (B) Fn1a-sfGFP deposition along chevron-shaped myotome boundaries. (C) fn1a G1509N-sfGFP mRNA also rescues myotome morphology. (D) Fn1b-mCherry localization to u-shaped myotome boundary characteristic for fn1a^−/−. (E) fn1b N1420G-mCherry mRNA does not rescue myotome morphology. (B-E) 3D projections of confocal z-stacks. Anterior to the right. Scale bars: 30 μm.

3.5. Examination of the inter-dependence of the Fn1b matrix on Fn1a and vice versa

We next examined whether fn1a is required for normal Fn1b matrix formation and whether fn1b is required for normal Fn1a matrix formation. MZfn1a^−/−; fn1b-mCherry adults are viable and fertile. In MZfn1a^−/− mutant embryos at the 10–12 somites stage, Fn1b-mCherry localizes to the surface of the paraxial mesoderm (Fig. 5B). However, the matrix is less dense compared to wild type with areas devoid of Fn1b-mCherry deposition (Fig. 5B white arrow) and individual fibrils do not radially align to a focal point. Fn1b-mCherry is assembled at the u-shaped myotome boundaries typical for MZfn1a^−/− mutants (Fig. 5C white arrow) and localizes to the proximal convoluted tubule (Fig. 5E white arrow) and the ventral mesoderm/posterior cardinal vein (Fig. 5F white arrow) at 5 dpf. Loss of function of fn1b has no obvious effect on the assembly and distribution of Fn1a-mNG at either 10–12 somites stage, 26 hpf or 5 dpf (Fig. 5H–N).

4. Conclusion

We have performed an overview of the function, regulation and ECM morphology of the two zebrafish fibronectin paralogs. Genetic experiments indicate that fn1a and fn1b are partially redundant in that the double mutant phenotype is much more severe than either single mutant phenotype. However, chimeric over-expression experiments suggest functional differences as fn1b cannot fully rescue fn1a mutants. Analysis of Fn1a-mNG and Fn1b-mCherry knock-in alleles reveal the two proteins colocalize in ECM fibers of some tissues such as the paraxial mesoderm. Other tissues, such as the branchial arches and the gut, have ECM containing only one of the two paralogs. Within the paraxial mesoderm, there appears to be a hierarchy within the ECM as fn1a is required for normal Fn1b fiber morphology, but fn1b appears dispensable for Fn1a fibrillogenesis. Genetic analysis of subunits of the principal receptor for Fibronectin, Integrin α5β1, indicates distinct requirements in different tissues. These differences likely stem from several factors including varied expression patterns of itgα5, itgβ1a and itgβ1b, the existence of two additional itgβ1 paralogs in zebrafish, and potential redundancy with five other Integrin heterodimers that are capable of binding to RGD peptides (Hynes, 2002; Mould et al., 2006; Wang et al., 2014). Ultimately, the most significant contribution of this study is the creation and validation of two knock-in lines that fluorescently label Fn1a and Fn1b. These lines should prove valuable for live imaging of the Fibronectin matrix in the analyses of its functions in embryogenesis, organogenesis and homeostasis.

Highlights.

• Fully functional Fn1a-mNeonGreen and Fn1b-mCherry knock-in alleles in zebrafish

• Fn1a-mNeonGreen and Fn1b-mCherry colocalize in fibers in the paraxial mesoderm

• Fn1a-mNeonGreen and Fn1b-mCherry have non-overlapping patterns during organogenesis

• integrin α5 mutants and integrin β1a; β1b double mutants have distinct phenotypes

• fn1a is required for normal Fn1b fibers in the presomitic mesoderm