Integrin αV is necessary for gastrulation movements that regulate vertebrate body asymmetry

Integrins mediate cell-cell and cell-extracellular matrix (ECM) interactions. The eighteen α and eight β integrin subunits in mammals can assemble into 24 different αβ heterodimers (Hynes, 2002). Specific integrins recognize a restricted range of ECM ligands (Plow et al., 2000; Wendel et al., 1998), and integrin affinity for these ligands is controlled by `inside-out' signaling pathways that converge upon the integrin cytoplasmic and transmembrane domains to activate the extracellular domains. Ligand binding to integrin triggers `outside-in' signals that mediate anchorage-dependent events, including cell migration, proliferation, differentiation and survival (Hynes, 2002; Luo et al., 2007).

In mammals, the αV integrin subunit can associate with any of five β subunits (β1, β3, β5, β6 or β8) and αV integrins typically recognize Arg-Gly-Asp-containing ligands, such as fibronectin, vitronectin, osteopontin and latency-associated peptide-TGF1β (Hynes, 2002; Takada et al., 2007). Postnatally, αV integrins have been implicated in cellular responses to injury, immunity, angiogenesis and aspects of tumor progression (Nemeth et al., 2007; Takada et al., 2007). During vertebrate development, αV integrins exhibit a wide distribution of overlapping expression domains in mammalian, avian and zebrafish embryos (Ablooglu et al., 2007; Delannet et al., 1994; Neugebauer et al., 1991; Testaz et al., 1999; Yamada et al., 1995). The developmental importance of αV is illustrated by the phenotype of αV (itgav – Zebrafish Information Network) knockout mice, which demonstrate improper formation of embryonic cerebral blood vessels and defective axon and glia interactions in the postnatal central nervous system (Bader et al., 1998; McCarty et al., 2005; McCarty et al., 2002). However, as αV is maternally deposited in mice (Sutherland et al., 1993) and up to 20% of αV-null mice survive to birth (Bader et al., 1998), the opportunity to uncover potential roles for αV in very early mouse development has been limited.

Here, we used antisense morpholino oligonucleotides (MOs) to transiently knockdown integrin αV in zebrafish (Eisen and Smith, 2008; Nasevicius and Ekker, 2000). We provide the first evidence that depletion of αV, along with depletion of one of its potential β subunit partners, β1b, leads to defective dorsal forerunner cell (DFC) migration during gastrulation. Recent reports have shown that DFC migration is important for the formation of Kupffer's vesicle (KV), a ciliated organ involved in left-right body axis specification in zebrafish (Amack et al., 2007; Amack and Yost, 2004; Essner et al., 2005; Essner et al., 2002; Oishi et al., 2006; Schneider et al., 2008). Indeed, we find that KV is abnormally formed in αV and β1b morphants and that both exhibit body asymmetry defects later in development.

Wild-type Danio rerio and Tg[sox17:gGFP] embryos were raised at 28.5°C. Embryos from natural matings were kept in 1-phenyl-2-thiourea (PTU; 0.003%) to inhibit pigmentation and staged according to Kimmel et al. (Kimmel et al., 1995). Zebrafish were housed in the UCSD animal facility and experiments were performed in accordance with the guidelines of UCSD Institutional Animal Care and Use Committee.

MOs used (see Figs S1 and S5 in the supplementary material) were: αV1, 5′-AGTGTTTGCCCATGTTTTGAGTCTC-3′; αV2, 5′-AGTAGATGGAGATCGCGCTGTTTGT-3′; αVEI10, 5′-GTCAGTGCAAATCATTACTCACCCA-3′; αV1miss (mutated residues in lower case), 5′-AcTcTTTcCCgATcTTTTcAGTgTC-3′; standard control MO, 5′-CCTCTTACCTCAGTTACAATTTATA-3′; β1b1, 5′-GGAGCAGCCTTACGTCCATCTTAAC-3′; β1bEI10, 5′-GCCAGTTTGAGTGAATAACTCACCT-3′. All MOs were obtained from Gene Tools (Philomath, OR, USA). MOs were injected at the 1- to 4-cell-stage blastulae except where noted. The impact of exon-intron-specific MOs on splicing was determined by RT-PCR with RNAs extracted from 3-6 somite stage (SS) AB embryos using the following primers (forward, backward, targeted exons, PCR fragment size): αV, 5′-GTTATTTGGGTTACTCTGTGGCTGTT-3′, 5′-GTTTGATGACACTGTTGAAGGTGAAGC-3′, exons 7 and 11, 336 bp; β1b, 5′-GCTCCAACATCTCCATTGGGGACGA-3′, and 5′-CAGATGTCAGTGCCATTATCCATAC-3′, exons 9 and 11, 334 bp. The altered splicing events were identified by size and DNA sequencing: αVEI10 MO resulted in a deletion of exon 10, yielding a 283 bp fragment; β1bEI10 MO caused an insertion of intron 10, yielding a 409 bp fragment.

Zebrafish αV mRNA (mαV RNA) was synthesized with mMESSAGE mMACHINE T7 Ultra Kit (Ambion, Austin, TX, USA). The 5′ UTR of αV cDNA was altered and missense and silent mutations were introduced to prevent and reduce possible αV1 interactions at its recognition site (mutated residues in lower case, start codon in parenthesis): 5′-agtggcggccgC(ATG)GGgAAgCAtT-3′. The resulting cDNA did not retain the αV2 segment (see Fig. S1 in the supplementary material). For MO rescue experiments, 200 pg of capped RNA was co-injected with 1.25 ng αV1 or 1.75 ng αV2 at the 1- to 4-cell-stage blastulae.

An N-terminal His-tagged 813 bp fragment of zebrafish αV cDNA, resulting a 34.2 kDa recombinant protein, coding for amino acid residues His106 to Leu376, was bacterially expressed and purified with Ni-NTA agarose beads (QIAGEN, Valencia, CA, USA). Approximately 1 mg of recombinant protein was used for immunization of rabbits (ab18001, Millipore, Temecula, CA, USA).

General methods for cell culture, protein analysis and western blotting were similar to Ablooglu et al. (Ablooglu et al., 2007). The antigen specificity of rabbit serum raised against recombinant zebrafish αV was tested in cell culture assays. A total of 20 μg CHO cell lysates or total zebrafish embryonic lysates from 5-8 SS embryos were resolved on 7.5% SDS-PAGE and analyzed by western blotting with the following antibodies: rabbit anti-zebrafish αV (ab18001, Millipore, Temecula, CA, USA) and β-actin antibody (ab6276, Abcam, Cambridge, MA, USA). Immunoreactive signals were detected and quantified by infrared emission spectrometry (Odyssey, Li-Cor Biosciences, Lincoln, NE, USA).

General methods for WISH were similar to Ablooglu et al. (Ablooglu et al., 2007). Standard molecular cloning techniques were used to prepare antisense riboprobes and their GenBank accession numbers are as follows: cas, AF362749; ntl, NM_131162; sox17, NM_131287; spaw, NM_180967; vtn, NM_001139461; pdx1, NM_131443.

Paraformaldehyde-fixed (4%) embryos were kept in methanol and rehydrated in incubation buffer containing 1× PBS, 0.5% Triton X-100, 5% BSA and 2% goat serum. The primary antibodies were used at 1:200 dilution: mouse anti-acetylated tubulin (T-6793, Sigma, St Louis, MO, USA); rabbit anti-aPKC-ζ (sc-216, Santa Cruz Biotechnology Inc., CA, USA); mouse anti-ZO-1 (339100, Invitrogen, Carlsbad, CA, USA). The secondary antibodies were used at 1:500 dilution: goat anti-mouse IgG (H+L) Alexa-Fluor 488 (A11017, Invitrogen, Carlsbad, CA, USA); goat anti-mouse IgG (H+L) Alexa-Fluor (A11031, Invitrogen); goat anti-mouse IgG (H+L) Alexa-Fluor 647 (A21237, Invitrogen); goat anti-rabbit IgG (H+L) Alexa-Fluor 568 (A11011, Invitrogen). Deyolked and Hoechst-stained (1:5000, Invitrogen) embryos were transferred into 50% glycerol.

Methods to determine individual DFC orientation and length-width ratio measurements were similar to previous studies (Davidson et al., 2006; Ezin et al., 2006). Half-rose diagrams were generated in six sectors with 30 degree (θ) intervals, representing anterior (a) with 0 θ, mediolateral axis (ml) with 90 θ and posterior (p) with 180 θ. Individual DFC length (L), the longest axis, and width (W), the widest distance across the DFC that is perpendicular to the length, were used to establish L-W ratios. Confocal images were captured with a Nikon Eclipse 80i microscope with a 40× objective, or with an Olympus FV1000 with a 20× objective. Multiple focal-plane confocal images of KV were acquired at 0.3 μm step-size z-series and of DFCs at 1 μm step-size z-series. 3D or 4D immunofluorescence images were assembled using Volocity 5.2.0 (Improvision, MA, USA), ImageJ (NIH) and Photoshop (Adobe) software. All images were examined in blind and analyzed by both ANOVA as a group and by Student's t-test. From the confocal images of Tg(sox17:GFP) embryos, we were able to identify 49-60% of individual DFC boundaries per morphant per embryo, percentages significant enough (P<0.05) to assume they are representative of the entire DFC population (Cochran, 1977).

In order to visualize DFC migration and individual DFC shape in vivo, 1- to 4-cell-stage blastomeres of Tg(sox17:GFP) embryos were injected either with MOs alone or with BODIPY TR ceramide (B-34400, Invitrogen) to outline cellular boundaries. Embryos were manually dechorionated and used for time-lapse, multiple focal plane 2 μm step-size z-series (4D) microscopy at 24°C. A method to detect the cell-cluster edge was adapted from Machacek and Danuser (Machacek and Danuser, 2006) and, instead of individual cell edge detection, we identified uniform multicellular DFC cluster edges. Unlike αV1miss control morphants, DFCs in αV1 and β1bEI10 morphants were not clustered homogeneously and had fewer cell-cell contacts, with visible gaps, and often formed multiple DFC clusters (compare Movies 1 and 2 versus 3-6 in the supplementary material). In these instances we identified smaller GFP-positive DFC clusters that remained as clustered during the length of imaging (e.g. αV1 in Fig. 5B). Individual cells that did not form a DFC cluster and migrated in all three axes were excluded from migration analyses. Velocity maps were calculated by finite differences of positions in consecutive frame triplets at T–1, T and T+1. Large protrusions were analyzed as previously described (Davidson et al., 2006), where each large and distinct protrusion was manually tracked and its appearance-disappearance cycle confirmed during the recorded time-lapse sequence.

Gene knockdown in zebrafish can be accomplished with translation-blocking or splice-inhibiting antisense MOs (Eisen and Smith, 2008; Nasevicius and Ekker, 2000). The former interfere with translation of both maternal and zygotic transcripts, whereas the latter interfere primarily with unspliced zygotic transcripts (Abrams and Mullins, 2009; Eisen and Smith, 2008; Nasevicius and Ekker, 2000). Translation-blocking MOs (αV1, αV2) were designed to target the AUG start site within the αV 5′UTR (see Fig. S1 in the supplementary material). The efficacy of αV1 in knocking down αV was determined on western blots of embryo lysates probed with an αV-specific antibody. In addition, we also used a splice-inhibiting MO (αVEI10) designed to yield a nonfunctional protein, and its efficacy was confirmed by RT-PCR (see Fig. S1 in the supplementary material).

As αV knockout mice exhibit myocardial abnormalities (Bader et al., 1998) and one of the earliest gross phenotypic abnormalities in αV1 morphants was abnormal heart development and pericardial edema (see below), we focused first on heart tube formation (Fig. 1A-D). As expected, the heart tube was on the left side of the body in the vast majority (83.8±3.8%, n=567) of embryos injected with the standard control MO (Fig. 1E; see Table S1 in the supplementary material). By contrast, this pattern was reversed in αV1 morphants, with the heart tube on the left in only 47.6±4.0% of embryos (n=569). When embryos were co-injected with αV1 and morpholino-resistant mαV RNA, the αV1 morphant phenotype was partially reversed (63.2±1.9% left, n=224), suggesting that the effects of αV1 were specific. This heart tube phenotype was also induced with a second translation-blocking MO (αV2), and it too was partially reversed by co-injection of mαV RNA (Fig. 1E; see Table S1 in the supplementary material).

To further assess the consequences of αV loss-of-function on left-right patterning, we examined expression of the earliest known asymmetric marker spaw in the anterior lateral plate mesoderm (LPM) (Long et al., 2003), lft2 in the left heart primordia (Amack and Yost, 2004), vtn in the liver (Thisse et al., 2004) and pdx1 in the pancreatic bud (Ober et al., 2003) (see Fig. S2 in the supplementary material). Compared with embryos injected with control MOs, αV morphants displayed randomized left-right gene expression profiles, with partial rescue of each defect by co-injection of mαV mRNA (see Fig. S2 and Table S2 in the supplementary material).

αV transcripts are widely present in gastrulating embryos (Fig. 2A-C) and are selectively localized in developing notochord at the beginning of segmentation (Ablooglu et al., 2007). As defective midline structures are known to cause laterality defects (Bisgrove et al., 2000), one potential explanation for our results is a compromise in the integrity of midline structures. However, αV1 morphants exhibited intact, ntl-positive notochords (see Fig. S3 in the supplementary material), arguing against this possibility.

Given the central role of KV in specifying laterality in zebrafish (Essner et al., 2005), our attention then focused on DFCs, which migrate at the leading edge of the dorsal blastoderm margin (Cooper and D'Amico, 1996; Oteiza et al., 2008) and are precursors of KV. Specifically, we wondered if αV might be present in DFCs and required for its function during gastrulation. Although maternal (Fig. 2A) and zygotic (Fig. 2B,C) expression profiles suggested that αV mRNA, and possibly αV protein, is present in DFCs, we chose a more direct and functional assay to generate DFC-selective morphants (DFC^αV1), which were created by injecting αV1 MO into the yolk cell at mid-blastula stage (Amack and Yost, 2004).

Unlike standard control MO and αV1miss-injected morphants (Fig. 2D,E), when αV1, αV2 or αVEI10 MOs were delivered at the 1-4 cell stage, αV morphants exhibited abnormal cerebellum, hydrocephaly involving the fourth ventricle, and pericardial edema (Fig. 2F-H). Although DFC^αV1 morphants were grossly similar to wild-type embryos (Fig. 2I-L), their heart tube location was randomized compared with controls (DFC^αV1, 48.2±5.9% left, n=223; DFC^control, 87.0±4.5% left, n=251) (Fig. 2M; see Table S3 in the supplementary material). However, when αV1 was delivered into the yolk cell at an even later stage [dome stage to 30% epiboly (30% E)], when DFC connections with yolk cells are considered to be closed (D'Amico and Cooper, 1997; Essner et al., 2005), these yolk^αV1 morphants displayed normal heart tube asymmetry (Fig. 2M; see Table S3 in the supplementary material). Thus, selective knockdown of αV supported its presence in DFCs by phenocopying the heart laterality defect.

Given the prominent role of αV integrins in mammalian cell migration (Hynes, 2002), we asked whether αV might be required for DFC migration. DFCs were identified in gastrulating embryos at 80% E by utilizing cas (Kikuchi et al., 2001), sox17 (Alexander and Stainier, 1999) or ntl (Amack and Yost, 2004) as markers (Fig. 3; see Fig. S4 in the supplementary material). Over 76% of uninjected embryos or embryos injected with αV1miss showed ovoid DFC clustering forming 5-6 tiers of cells from the margin (Fig. 3A,B). By contrast, when embryos were injected with αV1, αV2 or αVEI10, 53-69% of αV morphant DFCs were confined to a linear domain that had occasional gaps (Fig. 3C-E). This mutant DFC phenotype was independent of the markers used to identify the cells (see Fig. S4 in the supplementary material) and it could be rescued by co-injection of mαV mRNA. For example, although only 17.6±6.3% (n=69) of αV1 morphants showed an ovoidal wild-type DFC clustering pattern, 56.3±1.1% (n=23) of morphants co-injected with mαV mRNA demonstrated the wild-type pattern (see Table S4 in the supplementary material). Rescue by mαV mRNA was also observed in αV2 morphants (Fig. 3F). Finally, DFC-selective, DFC^αV1 morphants also exhibited DFC phenotypes similar to that of the αV1, αV2 or αVEI10 morphants (see Table S3 in the supplementary material). Thus, the loss of αV function during gastrulation appears to impair DFC migration but not specification.

Of the several β subunits that can pair with αV (Bouvard et al., 2001), zebrafish express β5, β6, β8 and multiple forms of β1 (β1a, β1b, β1b.1, β1b.2) and β3 (β3a, β3b) (Ablooglu et al., 2007; Julich et al., 2005; Julich et al., 2009; Mould et al., 2006; Thisse et al., 2001). Based on the reported spatial and temporal expression patterns of these β subunits and on overlapping expression patterns with αV during gastrulation, β1b (itgb1b – Zebrafish Information Network) and β5 (itgb5 - Zebrafish Information Network) appeared to be the only potential partners for αV in DFCs. Consequently, we examined their localization patterns at gastrulation to establish the identity of the potential αV partner in DFCs. Both β integrins showed distinct localization patterns at mid-gastrulation stages (Fig. 4A,B). For example, although β1b transcripts were mainly present in the embryonic axis at 80% E (Fig. 4A), β5 transcripts were only present in the marginal cells and there was a gap in its expression field (Fig. 4B). When embryos were examined at 80% E for cas expression in DFCs and for integrin β subunit expression by double wholemount in situ hybridization (WISH), the gap in the β5 expression field in the marginal cells overlapped the embryonic axis and only β1b was present in DFCs (Fig. 4C,D). Consequently, the binding partner for αV in DFCs might be β1b.

To study β1b function, a translation-blocking MO (β1b1) and a splice-inhibiting MO (β1bE10) were employed (see Fig. S5 in the supplementary material). After injection of either MO at the 1-4 cell stage, DFC markers were still expressed but there was a DFC mutant phenotype similar to that observed in αV morphants (Fig. 4E-H). Later in development at 32 hours post-fertilization (hpf), β1b1 morphants had pericardial edema and showed shorter and undulated midlines associated with U-shaped somites (Fig. 4I,J). When 0.7 ng β1b1 or 5 ng β1bEI10 were delivered at the 1-4 cell stage, these morphants showed reduced left-sided location of the liver (see Table S2 in the supplementary material). At these MO doses, considerable numbers of β1b1 morphants exhibited an absence of liver primordium and some β1bEI10 morphants had situs inversus totalis. When both β1b1 and β1bEI10 were co-injected, no liver primordium was evident (see Table S2 in the supplementary material). These results suggested that pleiotropic phenotypes might be caused by midline defects (Biemar et al., 2001) or lack of endoderm (Alexander and Stainier, 1999; Kikuchi et al., 2000; Komada and Soriano, 1999). As β1b is deposited maternally (Ablooglu et al., 2007; Mould et al., 2006) and its zygotic expression is maintained exclusively in the developing midline during gastrulation (Fig. 4A) (Julich et al., 2005), knocking down of maternal β1b with β1b1 MO could contribute pleiotropic defects. Consequently, we used a lower dose of β1b1 (0.5 ng) to minimize possible pleiotropic defects. Under these conditions, β1b1 alone did not significantly alter the asymmetric spaw expression pattern (Fig. 4L; see Table S2 in the supplementary material). However, when 0.5 ng β1b1 and 5 ng β1bEI10 were each co-injected, the severity of randomized spaw expression increased (Fig. 4K,L). In order to obtain more direct evidence for β1b function in DFCs, DFC-selective morphants (DFC^β1bEI10) were generated by injecting β1bEI10 into the yolk cell at the mid-blastula stage (Fig. 4M,N). The resulting DFC^β1bEI10 morphants looked grossly similar to wild-type embryos (Fig. 4M versus Fig. 4J), but heart tube location was still randomized as with DFC^αV1 morphants (Fig. 2M). By contrast, when β1bEI10 was delivered to the yolk cell at dome stage to 30% E, these yolk ^β1bEI10 embryos showed normal heart tube asymmetry (Fig. 4O).

Expression profiles and early tissue-specific knockdown phenotypes of αV and β1b indicated that development of proper body asymmetry requires their presence in DFCs. In order to examine possible αV and β1b genetic interaction, we used substantially lower doses of αV1 (0.41 ng) or β1bEI10 (1.1 ng) injections, to a level at which less then 15% embryos had an unclustered mutant DFC phenotype (see Fig. S6 and Table S5 in the supplementary material). However, co-injection of αV1 and β1bEI10 at these doses caused a dramatic increase in the frequency of the DFC clustering defect in embryos. Conversely, no such effect was observed when control αV1miss was co-injected with β1bEI10. These results are consistent with the idea that αV and β1b might interact in DFCs.

As DFCs are migratory cells and WISH analysis of integrin morphants showed abnormal DFC clustering (Figs 3 and 4; see Fig. S4 in the supplementary material), we sought to identify possible migratory defects in DFCs by time-lapse imaging in Tg(sox17:GFP) transgenic fish (Mizoguchi et al., 2008) (Fig. 5). Migratory DFC progenitors in this transgenic line have been shown to intercalate mediolaterally and form a compact and oval-shaped DFC cluster at mid-gastrulation stages (Oteiza et al., 2008). We reasoned that either proliferation of DFC progenitors and/or motile properties and directionality of DFCs might be affected in αV and β1b integrin morphants. When examined at mid-gastrulation, both integrin morphants appeared to have numbers of DFCs comparable with control morphants: αV1, 38.9±9.5 cells/embryo (±s.d., n=19); β1bEI10, 39.2±8.6 (n=18); αV1miss, 37.9±9.5 (n=15). However, live imaging of DFCs in Tg(sox17:GFP) embryos revealed that αV or β1b morphants failed to form a single DFC cluster (Fig. 5B,C; see Movies 3-6 in the supplementary material), unlike αV1miss-injected embryos where DFCs formed an oval-shaped cluster that migrated towards the vegetal pole (Fig. 5A; see Movies 1 and 2 in the supplementary material). Furthermore, examination of DFC cluster edge protrusions and protrusion velocity maps (Fig. 5) revealed that both morphant DFCs exhibited disarrayed cluster edge protrusions (Fig. 5D-F).

To study the effect of integrin knockdown in individual DFCs in more detail, we evaluated DFC orientation, shape and the number of large and discrete protrusions per DFC in Tg(sox17:GFP) embryos. Although the majority (64.4±2.0%) of control DFCs aligned strongly to the mediolateral axis (e.g. between 60 and 120 degrees), only 46.6±3.7% of αV1 and 41.4±4.5% of β1bEI10 morphant DFCs did so (Fig. 5G-I), a significant difference compared with controls (P≤0.0002). Expressed another way, αV1 and β1b morphant DFCs were oriented more randomly than control DFCs (ANOVA for anterior orientation, P<0.014; mediolateral orientation, P<0.00015; posterior orientation, P<0.016). Furthermore, individual control DFCs tended to be elongated, with a mean length-width ratio of 1.92±0.15, whereas integrin morphant DFCs tended to be rounder, with a mean length-width ratio of 1.33±0.09 (P<0.0001) (Fig. 5J). When large and discrete new protrusions in clustered or unclustered individual DFCs were tracked in time-lapse sequences, the average lifespan of protrusions was similar in controls and morphants (∼4.6 minutes; ANOVA P>0.10). However, the number of protrusions formed by morphant DFCs was significantly reduced (P≤0.02; αV1, 34.7±8.7 protrusions per embryo per hour; β1bEI10, 55.0±10.8; αV1miss control, 89.6±8.2; Fig. 5K). Thus, knockdown of αV1 or β1b in DFCs appears to affect their orientation, shape and protrusive activity.

Although DFCs in αV1 and β1b morphants showed morphologic and migratory defects, their vegetal migration did not seem to be affected, suggesting that αV and β1b might not be essential for this latter process. As earlier studies indicated that DFC and enveloping layer (EVL) attachments couple epiboly movements of both tissues towards the vegetal pole (D'Amico and Cooper, 1997; Oteiza et al., 2008; Solnica-Krezel et al., 1996), we examined DFC-EVL interactions. Confocal images of GFP(+) DFCs of Tg(sox17:GFP) morphants were examined by aPKC-ζ and ZO-1 immunolabeling. At 80% E, the DFC-EVL interface was enriched for tight junction components aPKC-ζ and ZO-1 in all three morphants (αV1miss, αV1, β1bEI10; Fig. 6). Thus, DFC-EVL connections in αV and β1b morphants are still maintained.

As migratory DFCs are precursors of KV (Cooper and D'Amico, 1996; D'Amico and Cooper, 1997; Essner et al., 2005; Melby et al., 1996), we reasoned that KV organogenesis might be affected by αV and β1b loss-of-function. Therefore, the physical dimensions of KV and the number of cilia per KV were determined (Fig. 7A-G). There was a strong positive correlation between KV volume and cilia number per KV, both in controls and in embryos in which αV and β1b had been knocked down (r=0.858; Fig. 7H). Compared with controls, both KV volume and cilia number per KV were significantly reduced in αV and β1b morphants (Fig. 7H). In uninjected embryos or those injected with control MO αV1miss, the mean number of cilia per KV was 38.5±3.4 and 36.8±4.9, respectively. However, in αV morphants (αV1, αVEI10, αV2 MOs), cilia number per KV (±s.e.m.) was 17.7±4.1, 19.9±5.8 and 25.4±5.3, respectively, whereas in β1b morphants (β1b1 or β1bEI10 MOs) it was 16.2±4.4 and 22.1±5.2, respectively (Fig. 7I). Furthermore, because KV organogenesis could be uncoupled from KV ciliogenesis by disruption of non-canonical Wnt signaling to yield short cilia (Oishi et al., 2006), we also examined cilia length and determined that it was similar in all morphants (Fig. 7I, inset).

As both integrin mutants have fewer cilia in their KVs and apparent KV volume is reduced (see Fig. 8), we examined the DFC-to-KV organization process in more detail. In a recent study it was reported that KV ciliogenesis and lumen formation are directly coupled (Oteiza et al., 2008). When DFCs of control αV1miss morphants were examined at 6-8 somite stage (SS), GFP(+) cells in Tg(sox17:GFP) were connected with tight junctions that were localized in a single cell layer that formed the lining of a single fluid-filled KV lumen (Fig. 8A). However, neither αV1 nor β1bEI10 morphants developed such uniform tight junctions demarking a fluid-filled lumen within the GFP(+) cell cluster (Fig. 8B,C). Furthermore, although 3D renderings of ZO-1-labeled control embryos had a uniform tight junction lattice (Fig. 8D), DFCs in αV1 and β1bEI10 morphants failed to aggregate properly and yielded a dysmorphic ZO-1 lattice (Fig. 8E,F). Unlike in control embryos, GFP(+) DFCs were partially polarized in the integrin morphants and they displayed discontinuous or absent aPKC-ζ labeling and fewer cilia within DFC clusters (Fig. 8G-I). Overall, these observations suggest that body asymmetry defects in αV and β1b morphants originate from disrupted DFC organization and consequent KV malformation.

Deletion of αV in mice causes early lethality at mid-gestation owing to placental defects; however, 20% of pups survive gestation with cerebral and intestinal blood vessel defects that contribute to their early demise (Bader et al., 1998). Lack of very early developmental defects in these animals could be attributed, in part, to maternally deposited αV protein (Sutherland et al., 1993). As depletion of both maternal and zygotic αV is difficult to achieve in mammalian models, the functional contribution of maternal αV cannot be inferred. By contrast, transient gene knockdown in zebrafish by injection of antisense MOs at the 1- to 4-cell stage could deplete zygotic transcripts and maternal transcripts not yet translated in the egg (Eisen and Smith, 2008; Nasevicius and Ekker, 2000). As zebrafish αV is deposited maternally like its murine counterpart and is expressed throughout gastrulation (Ablooglu et al., 2007), we used a series of MOs to determine the role of αV during early zebrafish development. Several new conclusions can be drawn from these studies. First, knockdown of αV in zebrafish leads to left-right asymmetry defects affecting multiple visceral organs, in addition to vascular defects previously described in αV knockout mice. Second, randomized visceral organ asymmetry in αV morphants could be explained by defective DFC migration and/or organization, leading to a malformed KV laterality organ. Third, the integrin β1b subunit is expressed in DFCs and its knockdown also causes a laterality phenotype. These results demonstrate a novel contribution of αV to early vertebrate development and suggest a previously unrecognized role for integrin αVβ1b within DFCs to form a normal KV, which in turn is essential for the establishment of proper left-right body asymmetry.

In preliminary work with αV MOs, we noticed altered heart tube asymmetry that could be partially rescued by injection of mαV mRNA (Fig. 1). As expression of left-sided genes (spaw and lft2) was randomized in αV morphants, the observed body asymmetry defects most likely did not originate from defective midline structures (Bisgrove et al., 2000). This was further supported by the observations that αV morphants had intact ntl-positive notochords and showed randomization of the normal left-right asymmetric locations of the liver and pancreas (see Fig. S2 in the supplementary material), features not characteristic of anterior midline defects (Bisgrove et al., 2000).

Given the generalized left-right asymmetry defects in αV morphants, we focused our attention on DFCs, which arise at the onset of gastrulation (Cooper and D'Amico, 1996; D'Amico and Cooper, 1997; Melby et al., 1996; Oteiza et al., 2008) and are precursors of KV (Amack et al., 2007; Amack and Yost, 2004; Essner et al., 2005; Essner et al., 2002; Oishi et al., 2006; Schneider et al., 2008). The robust early embryonic αV expression profile suggested that αV mRNA might be present in DFCs, and when we delivered αV1 MO selectively into these cells by injecting into the yolk cell at mid-blastula stage, heart tube location was randomized (Fig. 2). Consequently, these data support an essential role for αV in DFCs and laterality specification. To identify a potential β partner for αV in DFCs, mRNA expression domains for several integrin β subunits were examined. Of these, only β1b transcripts were present in DFCs (Fig. 4). Moreover, when the locations of migratory DFCs were identified with multiple DFC markers (cas, ntl and sox17), the selective knockdown of either αV or β1b resulted in the same DFC phenotype: a linear domain with occasional gaps, in contrast to ovoid DFC clusters in controls (Figs 3 and 4; see Fig. S4 in the supplementary material). Similar DFC mutant phenotypes have been reported in embryos with defective Ca²⁺ fluxes in DFCs, and in G_α12 or G_α13a morphants (Lin et al., 2009; Schneider et al., 2008). As αV and β1b morphant DFCs still express forerunner-specific markers, we posit that αV and β1b are not required for DFC specification. Migratory DFC progenitors normally intercalate mediolaterally and form an oval-shaped DFC cluster by mid-gastrulation (Oteiza et al., 2008). The unclustered and disoriented DFCs in integrin morphants suggest that αVβ1b is required for DFC migration and clustering.

It is interesting to note that the migratory defect seems to be specific for mediolateral DFC clustering but not for epiboly movements towards the vegetal pole because DFC-EVL connections that couple epiboly movements of both tissues towards the vegetal pole are still maintained in both morphants (Fig. 6). As the DFC-EVL connections allow a subset of DFCs to be pre-polarized, by having their apical contact with the overlying EVL (Oteiza et al., 2008), the maintenance of DFC-EVL connections in αV and β1b morphants allows us to assume the integrin αVβ1b is not essential for DFC polarization. However, because KVs in αV and β1b morphants formed a dysmorphic tight junction lattice that normally demarks the fluid-filled KV lumen (Fig. 8), integrin αVβ1b appears essential for DFC-to-KV organization. In addition, our studies do not exclude the possibility of further αVβ1b involvement in events later in KV development.

αVβ1 has not been studied extensively in mammals, in part because it is usually found in cells that express many different integrins, including other αV integrins, and because there is a dearth of αVβ1-specific antibodies and inhibitors. In addition to serving as a receptor for several RGD-containing matrix proteins (Hynes, 2002; Takada et al., 2007), αVβ1 serves as a receptor or co-receptor for several viruses, including adenovirus (Davison et al., 2001; Li et al., 2001), foot-and-mouth disease virus (Jackson et al., 2002), parechovirus (Triantafilou et al., 2000) and metapneumovirus (Cseke et al., 2009). Although many integrins function as both adhesion and signaling receptors (Hynes, 2002; Takada et al., 2007), the role of αVβ1 in cell signaling remains to be fully explored. Several in vitro studies have hinted at a role for αVβ1 in developmental processes. For example, αVβ1 has been implicated in adhesion and spreading of mouse embryonic cells on fibronectin (Yang and Hynes, 1996), migration and/or differentiation of rodent embryonic astrocytes and oligodendrocyte precursors (Milner et al., 1996; Milner et al., 2001), binding of chicken myotubes to agrin (Martin and Sanes, 1997) and cell binding to neural cell adhesion molecule L1 (Felding-Habermann et al., 1997). By taking advantage of the zebrafish system, the present study has uncovered a necessary role for αV and β1 during vertebrate gastrulation in regulating DFC migration, and DFC-to-KV organization that is essential for proper development of left-right asymmetry. Additional studies will be required to prove that αVβ1b is a heterodimer in DFCs, to identify the relevant matrix ligand(s) for αVβ1b during gastrulation and to determine if the signaling as well as the adhesive functions of this integrin are required for development of a proper KV.

We thank Drs N. Chi and D. Y. R. Stainier for providing Tg(sox17:GFP) fish, Oleg Tsivkovsk and Emerald Butko for technical assistance and the UCSD Fish Facility for maintaining zebrafish stocks. Confocal image analyses were done at the UCSD Neuroscience Microscopy Shared Facility (NIH P30 NS047101) and at the UCSD Cancer Center (NIH P30 CA23100). This work was supported by NIH grants F32 HL094012-01 to E.T. and HL78784 and HL56595 to S.J.S. Deposited in PMC for release after 12 months.