Caveolin-1 is required for lateral line neuromast and notochord development.

Caveolae have been linked to diverse cellular functions and to many disease states. In this study we have used zebrafish to examine the role of caveolin-1 and caveolae during early embryonic development. During development, expression is apparent in a number of tissues including Kupffer's vesicle, tailbud, intersomite boundaries, heart, branchial arches, pronephric ducts and periderm. Particularly strong expression is observed in the sensory organs of the lateral line, the neuromasts and in the notochord where it overlaps with expression of caveolin-3. Morpholino-mediated downregulation of Cav1α caused a dramatic inhibition of neuromast formation. Detailed ultrastructural analysis, including electron tomography of the notochord, revealed that the central regions of the notochord has the highest density of caveolae of any embryonic tissue comparable to the highest density observed in any vertebrate tissue. In addition, Cav1α downregulation caused disruption of the notochord, an effect that was enhanced further by Cav3 knockdown. These results indicate an essential role for caveolin and caveolae in this vital structural and signalling component of the embryo.


Introduction
Caveolae are specialised plasma membrane microdomains enriched in glycosphingolipids, cholesterol and integral membrane proteins termed caveolins. These small 50-80 nm 'cave-like' invaginations of the plasma membrane are found in many cell types but are particularly abundant in adipocytes, endothelial cells and muscle cells (Parton and Simons, 2007;Rothberg et al., 1992;Scherer et al., 1996;Tang et al., 1996;Way and Parton, 1995). Determining the role of caveolins and caveolae in normal tissues is fundamental to understanding their role in diseases such as cancer, where caveolin is a candidate tumour suppressor gene and in muscle disease, where mutations in caveolin have been associated with a number of different clinical symptoms and conditions (Li et al., 2006;Razani and Lisanti, 2001;Williams and Lisanti, 2005). Caveolins are a family of small integral membrane proteins that bind cholesterol and fatty acids (Murata et al., 1995;Trigatti et al., 1999). Three caveolin genes exist in higher eukaryotic genomes. Caveolin-1 (Cav1) and Caveolin-2 (Cav2) are expressed in numerous tissues in mammalian cells with particularly high expression in tissues with abundant caveolae, whereas caveolin-3 (Cav3) is the major caveolin isoform in skeletal and cardiac muscle (Tang et al., 1996;Way and Parton, 1995). The Cav1 and Cav2 genes each generate two isoforms. In the case of Cav1, alternative transcription initiation sites produce two separate mRNA species (Kogo and Fujimoto, 2000).
Cav1 has been implicated in an increasing number of cellular processes by the study of Cav1-null mice and by downregulation of Cav1 expression in cultured cells. Cav1 is sufficient to drive the formation of caveolae in lymphocytes, which do not normally express caveolins (Fra et al., 1995), and ablation of Cav1 in mice results in a lack of caveolae in all cells where Cav1 is normally expressed (Drab et al., 2001;Parton, 2001;. Although Cav1-null mice are relatively healthy and fertile with a relatively mild phenotype, in-depth characterisation of these mice has revealed a wide range of effects (Drab et al., 2001;. These studies implicate caveolae in a broad range of cellular activities including lipid regulation, cell signalling and mechanosensation (Parton and Simons, 2007). For example, Cav1-null mice show resistance to diet-induced obesity indicating changes in lipid metabolism or adipocyte function (Drab et al., 2001;Razani et al., 2002), consistent with in vitro studies (Cohen et al., 2004;Frank et al., 2001;Le Lay et al., 2001). Loss of Cav1 also leads to reduced survival of mice after partial hepatectomy (Fernandez et al., 2006) associated with a severe reduction in lipid droplet formation and a block in the cell cycle, suggesting an important role for Cav1 in the coordination of lipid metabolism with cell cycle progression (Fernandez et al., 2006). Cav1-null mice also show increased sensitivity to carcinogens (Capozza et al., 2003) and changes in mammary development (Williams et al., 2006), consistent with decreased expression of Cav1 in some tumours (Williams and Lisanti, 2005) and a role for Cav1 in regulating specific signalling pathways. Cav1 has also been strongly linked to mechanosensation (Boyd et al., 2003;Parton and Simons, 2007;Radel and Rizzo, 2005), particularly in smooth muscle cells (Sedding et al., 2005) and endothelia (Rizzo et al., 2003;Yu et al., 2006). Recent studies have shown a crucial role for Caveolae have been linked to diverse cellular functions and to many disease states. In this study we have used zebrafish to examine the role of caveolin-1 and caveolae during early embryonic development. During development, expression is apparent in a number of tissues including Kupffer's vesicle, tailbud, intersomite boundaries, heart, branchial arches, pronephric ducts and periderm. Particularly strong expression is observed in the sensory organs of the lateral line, the neuromasts and in the notochord where it overlaps with expression of caveolin-3. Morpholino-mediated downregulation of Cav1␣ caused a dramatic inhibition of neuromast formation. Detailed ultrastructural analysis, including electron tomography of the notochord, revealed that the central regions of the notochord has the highest density of caveolae of any embryonic tissue comparable to the highest density observed in any vertebrate tissue. In addition, Cav1␣ downregulation caused disruption of the notochord, an effect that was enhanced further by Cav3 knockdown. These results indicate an essential role for caveolin and caveolae in this vital structural and signalling component of the embryo.
caveolae in linking shear stress generated by laminar flow on the surface of the endothelium to NO generation (Yu et al., 2006).
Despite these numerous studies, a universal role for caveolae linking these apparently diverse functions, which presumably require the unique and conserved architecture of caveolae, is still lacking. We and others have thus started to investigate the function of caveolins in a wider range of eukaryotic systems including the nematode, Caenorhabditis elegans, the frog, Xenopus laevis and the zebrafish, Danio rerio. The C. elegans genome contains two caveolin family members (Tang et al., 1997). Cav1 is expressed in most (or all) cells throughout embryonic and larval development. Interference with Cav1 specifically advances meiotic progression in the worm, a process that is Ras/MAP kinase dependent (Scheel et al., 1999). Interestingly, morphological caveolae have not been described in C. elegans and recent studies suggest that C. elegans caveolin may not generate caveolae (Sato et al., 2006) (M. Kirkham, S.N., D. Wakeham et al., unpublished results).
Here we used zebrafish to study the role of caveolae. Previous Cav1 zebrafish knockdown experiments showed early developmental abnormalities with defects in axis elongation and somite patterning. In these studies, embryos did not survive to 5 days post fertilisation (Fang et al., 2006;Smart et al., 2004). We examined the expression of Cav1 at the mRNA and protein level and the formation of caveolae in various tissues at the ultrastructural level. We also carried out detailed characterisation of Cav1␣ function. We show high levels of cav1 expression in neuromasts of the lateral line, the organs that aid in sensing water movement (Dijkgraaff, 1989). Knockdown of Cav1␣ expression disrupts neuromast maturation. We also observe strong expression in the notochord and an incredibly high density of caveolae within specific regions of notochord cells, particularly the cell-cell junctional regions within the centre of the notochord. We also demonstrate the evolutionary conservation of caveolin notochord expression in chick and mouse. We propose a novel structural role for caveolae in the notochord and a hitherto unexpected role in neuromast development.

Results
Putative signalling tissues express Cav1 during early development To further our understanding of Cav1 and its role in development we examined the developmental expression pattern of cav1 by mRNA in situ hybridisation. cav1 mRNA first appears in a spatially restricted pattern in the notochord just prior to the six-somite stage (12 hours), and this expression persists throughout most of the segmentation period (Fig.  1A,E). At the six-somite stage, the periderm, a tightly sealed epithelial monolayer of specialised impermeable cells that covers the entire embryo after gastrula stages (Kimmel et al., 1995), also expresses cav1 (Fig. 1A). Peridermal labelling persists throughout subsequent embryonic and early larval stages ( Fig. 1A-G,I-L,N). Later, the epithelial lining (demonstrated by the ring, Fig. 1B) of Kupffer's vesicle (KV) expresses cav1. The KV is thought to be important in many signalling processes (Cooper and D'Amico, 1996;Melby et al., 1996). At the 14-20-somite stage (16-19 hours), KV labelling disappears (consistent with loss of this structure) and cav1 expression is then detected in a cluster of cells in the tailbud, probably originating from the KV (Fig. 1C,D). The expression in the tailbud persists until the 20-somite stage (Fig. 1E). Thus, during early development, major signalling centres express cav1.
By the long-pec stage (48 hours), anterior and posterior lateral line neuromasts express high levels of cav1 ( Fig. 1N; example of labelling at 72 hours). The neuromasts are deposited along the lateral line at regular intervals, from a migrating primordium; however, cav1 expression is undetectable in the migrating primordium (Fig. 1F). This migration is completed by about 42 hours when the last neuromast is deposited, and neuromasts mature approximately 6 hours later (Gompel et al., 2001), suggesting an association of Cav1 with maturing neuromasts.
At the protruding-mouth stage (72 hours), we observe cav1 labelling in the heart, branchial arches, gut, pronephric ducts and the intersomite boundaries, presumably representing expression by the vasculature (Fig. 1H-N). By 6 days, cav1 expression in the heart is restricted to the ventricle (Fig. 1M). The correlation of this mRNA expression pattern with protein levels was confirmed by labelling with a pan-caveolin antibody as used previously (Nixon et al., 2005)

Notochord cells develop a high density of intermediatefilament-associated caveolae
The small size of the zebrafish embryo provides the opportunity to analyse the formation of caveolae in various tissues during development and to therefore gain new insights into the role of these structures at specific developmental stages. We analysed the formation of caveolae in tissues that express Cav1 by examining ultrathin sections cut transversely across the zebrafish trunk at various stages of development. We detected a low density of caveolae on the basolateral surface of peridermal cells from 48 hours, consistent with the strong expression of Cav1 in the peridermal layer.
The most dramatic concentration of caveolae forms in the large vacuolating cells of the notochord consistent with caveolin expression (Fig. 2F). After segmentation stages (24 hours), caveolae are present in all cells of the notochord. However, the highest concentration of caveolae is located within the cell-cell junctional regions that cross the centre of the notochord ( Fig. 2A-E). This is even more dramatic by 48 hours (Fig. 2B,C,E) and 72 hours; even though the cells are extremely vacuolated and the cytosol largely devoid of other recognisable structures, the central cell-cell contact regions (forming 'struts') across the centre of the notochord show an incredible density of caveolae, greater than the highest density of caveolae observed in any tissue to our knowledge (estimated as 37 caveolae/m 2 in mouse adipocytes; 38 caveolae/m 2 in 30-hour notochord; 88 caveolae/m 2 in 48-hour notochord; see Materials and Methods for details).
In addition to the incredible density of caveolae in the junctional regions of the notochord, we noted a striking organisation of filaments in areas enriched in caveolae. The filaments are largely excluded from the areas enriched in caveolae (Fig. 3A,c) but form a distinct filamentous network to the cytoplasmic side of the caveolae (Fig. 3A,f). To identify these filaments we measured their diameters as 10±2 nm. This size corresponds to the size of intermediate filaments.
Labelling with a pan-cytokeratin antibody on sections confirmed labelling within the notochord (Fig. 3D,E) and revealed a pattern within the 'struts' similar to the Cav1 labelling ( Fig. 3C). Immunogold labelling of the zebrafish notochord revealed caveolin within caveolae and cytokeratin labelling either side of the caveolae consistent with the intermediate filaments seen in Fig. 3A. By electron microscopy, apparent connections between the band of intermediate filaments and the 'bulb' of the caveolae are frequently observed (Fig. 3A, arrowheads). In addition to the intermediate filaments, putative desmosomes are seen (Fig. 2E, arrowheads). To gain further insight into this organisation in 3D at high resolution, we analysed thick sections of 48-hour embryos by electron tomography and generated a dual-axis-tilt series of a 300 nm section of the notochord region of a 3-day post fertilisation embryo (see Movie 1 in supplementary material and single images in Fig. 3B). Connections between intermediate filaments running parallel to the cell surface and the caveolae but not to the junctions are apparent in the tomograms. This unusual organisation suggests a strong link between caveolae, intermediate filaments, and the cell-cell contacts within the notochord.

Morpholinos downregulate expression of Cav1␣ protein
To explore the role of Cav1␣ during development we first optimised conditions to knockdown Cav1 expression. One-to two-cell-stage embryos were injected with two independent morpholino antisense oligonucleotides (MO), one targeted to the region flanking the start codon of cav1␣ (cav1-MO1) and a second MO directed to a sequence upstream of the start codon in the 5ЈUTR (cav1-MO2). These MOs should inhibit expression of Cav1␣ but not Cav1␤. Cav1 expression was examined by labelling with a pan-caveolin antibody (Luetterforst et al., 1999), which we have used previously in zebrafish (Nixon et al., 2005). Caveolin labelling at the junctions between peridermal cells (Fig. 4I) and in intersomite boundaries is lost after MO injection (Fig. 4J), but, importantly, labelling of muscle cells known to express Cav3 (Nixon et al., 2005) persists ( Fig. 4J) showing the specificity and efficacy of Cav1 knockdown. We also performed western blot analysis on embryos injected with cav1-MO1 and cav1-MO2 at 48 hours post injection. A band of ~24 kDa was detected by western blot analysis using a caveolin polyclonal antibody in the control MO-injected embryos but embryos injected with either MO targeted to Cav1 show complete

Knockdown of Cav1␣ affects development of heart and notochord
We then examined the effects of cav1-MO1 and cav1-MO2 morpholinos, injected at various concentrations, on embryonic development. The cav1-MO1-injected embryos consistently show more-severe defects than the cav1-MO2 embryos but, importantly, the two morpholinos cause similar effects, which are summarised in Table 1. Injected embryos are characteristically curved and show notochord defects ( Fig. 4B,E,F and see below). cav1-MO2 injected embryos also display a shortened body axis and blebbing around the blood island ( Fig. 4B). At 72 hours, the embryos injected with cav1-MO1 at higher concentrations (3 and 4.5 ng/embryo) exhibit enlarged pericardial sacs and abnormal heart chambers with incorrect heart looping. Embryos injected with this MO at higher concentrations also showed identical effects to those shown (Fang et al., 2006) including retinal pigmented epithelial defects, disruption to neural tissues and disorganised somites (not shown). All embryos injected with a control MO up to 4.5 ng/embryo develop normal morphology including hearts. The cav1 MO phenotype was dose-dependent; embryos injected with 1.5 ng/embryo of cav1-MO1, which only caused a partial ablation of Cav1␣ expression measured by western blotting (not shown) were generally normal but had slightly curved tails (not shown). Thus, the phenotype observed with two different MOs, but not a control MO even at higher concentrations, correlates with the level of Cav1 reduction and demonstrates the specificity of the effect. A number of additional control experiments were carried out, firstly the cav1-MOs were co-injected with a MO to p53 to suppress non-specific offtarget defects through the activation of p53 (Langheinrich et al., 2002) at concentrations from 0.3-3 ng/embryo. The same phenotype was observed (results not shown). In addition, embryos coinjected with the two cav1 MOs at concentrations that do not cause any defects when injected alone, produced identical phenotypes to those described above, demonstrating the specificity of the MOs and confirming that the effects seen are not due to MO toxicity (results not shown).
To further dissect the role of Cav1, we designed a spliceblocking MO directed to the exon2-intron2 splice junction. The MO caused a reduction, but not loss, of full-length cav1 mRNA as well as a second, smaller band corresponding to mis-spliced Journal of Cell Science 120 (13) RNA as judged by RT-PCR. This is consistent with contribution of maternal mRNA to caveolin expression. We also observed no significant phenotype upon injection of this MO (results not shown). Taken together with the results of Fang et al., this strongly suggests a role of maternally expressed Cav1 at early stages of development (Fang et al., 2006).

Caveolin functions in notochord
In view of the incredible abundance of caveolae in the zebrafish notochord, we further examined the effect of Cav1␣ MOs on notochord development compared with embryos injected with a control MO. A shortened embryo length is characteristic of disrupted notochord development (Coutinho et al., 2004;Stemple et al., 1996). Consistent with a role for Cav1␣ in the notochord, knockdown of Cav1␣ using cav1-MO2 causes significantly shorter embryos (Fig. 4L). This was also seen with cav1-MO1 but to a lesser extent. At 72 hours, embryos injected with cav1-MO2 had a length equivalent to that of control 48hour embryos (3.06±0.33 mm, n=38; control MO injected embryos 3.59±0.18 mm, n=37, from three independent experiments). This reduction in length is an underestimate because only straight embryos could be measured accurately and these embryos showed the mildest phenotype. In addition, loss of Cav1␣ results in an undulating notochord and the appearance of rounded vacuoles within the notochord (Fig. 4E). Ultrastructural examination of the notochord of cav1-MO2injected embryos at 30 hours shows a dramatic, although somewhat variable, reduction in caveolae density within the cell-cell contact regions (40-75% decrease compared with the control MO-injected embryos from two independent experiments). This incomplete loss might reflect an overlap in function with Cav3 (Nixon et al., 2005) or Cav1␤ in the notochord, rather than incomplete knockdown of Cav1␣, because caveolae are undetectable in the peridermal cells of the cav1-MO2 injected embryos (results not shown). The notochord also shows smaller rounder vacuoles in the control MO-injected embryos and more cellularity than the 'empty' vacuoles of the control embryos at the same stage consistent with the light microscopic observations (Fig. 5B). The medial layer of the perinotochordal basement membrane, or sheath, is also thicker and more disorganised; fibre bundles were observed at various orientations to the notochord in contrast to the uniform orientation of control MO-injected embryos (Fig. 6D). We also noted that in control MO-injected embryos the caveolae-rich plasma membranes of neighbouring cells were closely apposed (Fig. 5E). By contrast, the plasma membranes of the equivalent regions in cav1-MO2-injected embryos were Journal of Cell Science 120 (13) In cav1-MO2 injected embryos the notochord appears 'bubbly' and undulating (E) and the undulating notochord is still apparent at 48 hours (F). When embryos are labelled for caveolin with the pan-caveolin antibody at 2 days post fertilisation, control MO injected embryos have normal Cav1 labelling at the cell surface (G) with normal Cav3 labelling in muscle (H), but when embryos are injected with cav1-MO1 at 1.5 ng/embryo, the labelling at the junction of these cells is diminished, although some weak intracellular labelling remains (I). Staining in the muscle is unaffected by cav1-MO1 (J compared with H). not closely associated and the distance between cells was variable.
In view of the incomplete loss of caveolae in the notochord, despite the inhibition of Cav1 expression as judged by western blotting (Fig. 4K), we proposed that Cav3 could contribute to caveolae formation in the notochord. This was examined by co-injection of cav1-MO2 and cav3-MO1 (Nixon et al., 2005). This caused a more severe notochord defect with round cells in the notochord still observed at 48 hours (Fig. 4N). In addition, a greater loss of morphological caveolae was observed (Fig. 5G, 84% reduction). With the unexpected localisation of both Cav1 (this study) and Cav3 (Nixon et al., 2005) in the zebrafish notochord, and in view of the functional role of Cav1␣ in notochord development described above, we examined whether the expression of caveolin in the notochord was a feature of other vertebrates. Both mouse and chicken notochord show strong caveolin expression as judged by antibody labelling (Fig. 2F,G) consistent with an evolutionarily conserved role for caveolin in notochord function or development.

Loss of Cav1␣ expression disrupts neuromast maturation
As noted above, neuromasts show strong expression of Cav1 (Fig. 1I,J,L,M and Fig. 6A,B). We investigated whether Cav1 knockdown would affect neuromast development. We used DASPEI, a vital dye, to label hair cells within the neuromasts. Injection of embryos with morpholinos targeted to cav1␣, at low concentrations that result in normal embryonic morphology, causes a dramatic decrease in the number of neuromasts in the posterior lateral line at 72 hours ( Fig. 4M and Fig. 6E,F). The few neuromasts that form often contain fewer hair cells than normal neuromasts at the equivalent age. The number of neuromasts in the anterior lateral line is also reduced but not to the same extent as those in the posterior lateral line. To confirm the absence of neuromasts (and not just the loss of hair cells that label with DASPEI) we performed scanning electron microscopy (SEM) of 72 hour zebrafish. Although these animals appear relatively normal, there are fewer neuromasts in embryos injected with cav1-MO1 at 1.5 ng/embryo (Fig. 6H,J). Quantification showed that 60% of embryos injected with cav1-MO1 have three or fewer neuromasts in the posterior lateral line (n=71) whereas 100% of WT embryos have more than seven (Fig. 4M). Only 22% of the embryos injected with cav1-MO1 have an equivalent number of neuromasts (7+) equivalent to control MO-injected embryos (Fig. 6D) or WT embryos (Fig. 6G,I). These studies show a vital role for Cav1␣ in neuromast development.

Discussion
The existence of small membrane invaginations termed caveolae has been known for over 50 years but to date their precise function is unknown. The production of knockout mice has furthered our understanding of this organelle, but it remains  surprising that an organism can function without such an abundant feature of the plasma membrane. We have used the zebrafish to further our knowledge of the function of this elusive organelle. The ease of manipulating genes and the accessibility of the zebrafish embryo makes this system ideal for studying caveolin and caveolae function. We found that Cav1 expression in signalling tissues is important for axis formation. Interestingly, we also found caveolin expressed in the notochord and neuromasts. We examined the notochord at the ultrastructural level and discovered an incredible density of caveolae in this tissue. Consistent with the observed pattern of Cav1 expression, we show that knockdown of this protein leads to disruption of both notochord and neuromast development.
Cav1 is localised to key signalling tissues Previous studies have examined the expression of Cav1 in the zebrafish at limited stages. These studies localised cav1 to the intestinal epithelium, somite borders and the heart ventricle (Smart et al., 2004) along with the notochord, neuromast, skin and vasculature (Fang et al., 2006), but did not report details of earlier Cav1 expression. We find that Cav1 is spatially restricted from 12 hours in the developing zebrafish embryo. As in the previous study, we find cav1 expression in the intestine, inter-somite region and the heart but, in addition, we find expression in tissues important for developmental signalling such as the notochord (described by Fang and co-workers) and tailbud, including the Kupffer's vesicle. The notochord is an early embryonic structure that is the source of signals important for axis formation, neural differentiation and somite patterning (Adams et al., 1990;Currie and Ingham, 1996;Ericson et al., 1995;Lewis et al., 1999). The expression of Cav1 in the notochord may reflect a role of caveolae in regulating developmental signalling events or, as we discuss later, a role for caveolae in mechanosensation. The tailbud is critical for posterior body development and contains the tail organiser regions of the zebrafish (Kanki and Ho, 1997;Melby et al., 1996). Caveolin expression in the tailbud is dynamic, with labelling evident at 16 hours, becoming more prominent by 18 hours, and then disappearing soon after 24 hours, correlating with the transient appearance of this structure. This is consistent with a role of caveolin in the regulation of signalling events at the tailbud during development.

Functional characterisation of Cav1 during development
We have carefully assessed the role of Cav1␣ in zebrafish development using various MOs at a range of concentrations and by optimising conditions for knockdown of Cav1␣ protein as judged both by western blotting and by immunofluorescence. This allowed us to minimise possible toxic effects. Previous studies (Fang et al., 2006;Smart et al., 2004) showed defects in axis elongation and somite patterning, but the high levels of MO used make it difficult to rule out toxic effects. We also see these defects (including head necrosis) when we inject the cav1␣ MOs at high concentrations, but when used at lower concentrations and with a second non-overlapping MO to cav1␣ we see less-severe defects allowing us to dissect further the possible functions of caveolin and caveolae. We also designed and injected a splice-blocking MO directed to the exon2-intron2 splice junction of cav1 which caused a reduction, but not loss, of full-length cav1 mRNA, as judged by RT-PCR. This is consistent with contribution of maternal mRNA to caveolin expression. This MO did not cause any significant phenotype suggesting that maternal mRNA for caveolin is important at early time points.
Journal of Cell Science 120 (13) Our results provide new insights into Cav1␣ function and identify novel roles for caveolin in the notochord and neuromasts of the fish.

Cav1 is required for neuromast maturation
Neuromasts are lateral line organs that form superficial sensory organs in amphibians and fish (Dijkgraaff, 1989;Northcutt, 1992). The neuromasts are a unit of sensory hair cells with a ring of support cells surrounding them. The neuromasts detect water movements and facilitate schooling, prey capture and predator avoidance (Dijkgraaff, 1989). There are two major components of the lateral line, the anterior and posterior lateral line (including neuromasts and sensory neurons) (David et al., 2002;Metcalfe, 1985). In zebrafish, the posterior lateral line primordium begins migrating from a position posterior to the otic placode. By the end of embryogenesis, the posterior lateral line consists of seven or eight neuromasts regularly spaced along the tail . Cav1 expression is absent from the migrating primordium but is detected in the mature neuromast from 48 hours. Expression of cav1 appears to occur in a ring-like shape that may represent the supporting cells of the neuromast. The function of Cav1 in these cells is unknown but we have now shown that Cav1␣ is required for their formation. Using a voltage-sensitive dye that selectively labels sensory hair cells (Harris et al., 2003), as well as scanning electron microscopy (SEM), we show that downregulation of Cav1␣ results in reduced maturation of the neuromasts. Cav1␣ is not required for migration of the primordium, because by 72 hours, we observe some neuromasts in the tail but not along the body of the fish. Lateral line neuromasts have striking similarities to the structure and function of the hair cells of the inner ear (Platt et al., 1989). In most mammals, it is thought that the loss of hair cells in the inner ear that results in deafness is permanent (Saunders et al., 1991;Saunders et al., 1985). However, in birds and fish, auditory hair cells are replaced by new hair cells when they are damaged (Corwin and Cotanche, 1988;Williams and Holder, 2000). A better understanding of the mechanisms involved in neuromast maturation and regeneration may therefore provide insight into triggers that may be capable of stimulating regeneration in mammalian sensory epithelia. Whether caveolae play a role in the mature neuromast is unclear from our studies, although enrichment of Cav1 in neuromasts is still apparent after 5 days of development. However, this remains an interesting possibility in view of the postulated role of caveolae in mechanosensation (Parton and Simons, 2007;Rizzo et al., 1998;Rizzo et al., 2003;Sedding et al., 2005;Yu et al., 2006).

Cav1 in the developing notochord
The notochord is an embryonic structure in vertebrates that functions in both tissue patterning and as a supportive element. It is eventually replaced by a backbone (Stemple, 2005). In some vertebrates, such as some fish, the notochord can function as a structural element throughout the life of the animal. In zebrafish, the notochord functions as a structural element during embryogenesis and early larval development (Kimmel et al., 1995). Without a fully differentiated notochord, the embryonic axis fails to elongate correctly (Stemple et al., 1996). Consistent with the high level of Cav1 expression in the notochord, caveolae are extremely abundant, indicating an important role for these structures in notochord function and the importance of the notochord as a powerful system to dissect caveolae function. Caveolae function in the notochord may be conserved in evolution as indicated by our observations of high levels of caveolin in the notochords of both the mouse and chicken. In addition, examination of images from an earlier ultrastructural analysis of the intervertebral joint of the teleost, Perca flavenscens reveals abundant caveolae. This tissue is derived from the notochord and contains the sheath and highly vacuolated notochord cells (Schmitz, 1995). Extensive intermediate filament (IF) networks were also observed within these cells, and in later studies of the intervertebral joint of Perca flavenscens, they were identified as keratins (Schmitz, 1998). Consistent with these studies, we now show the association of 10-nm-diameter filaments with the caveolae lining the regions of cell-cell contact between the notochord cells and co-staining of these regions with anti-keratin antibodies. By electron tomography, we show for the first time that these intermediate filaments appear to have an intimate connection with caveolae, and we speculate that these components may work together to keep the notochord functioning as a structural component of the embryo. In keratinocytes, keratin filaments provide mechanical integrity (Fuchs and Coulombe, 1992) and are particularly prominent in the cytoplasm of cells that are subject to mechanical stress. Normally intermediate filaments attach to sites of cell-cell adhesion, or desmosomes (Fuchs and Cleveland, 1998), but here, rather than connecting to desmosomes, we see close association with caveolae. The components involved in this link are now amenable to dissection and functional characterisation.
Downregulation of Cav1␣ expression causes severe defects in notochord morphology. The tails of the embryo are curved, the embryo is shorter and notochord cells contain smaller vacuoles. These changes may reflect decreased internal pressure in the notochord. At the ultrastructural level, we observed a dramatic reduction in density of caveolae, a looser apposition of the plasma membranes of neighbouring cells and a change in the size of the medial layer of the sheath with the filaments apparently disorganised and less densely packed. Whether this reflects a perturbation of sheath component deposition, as observed in sneezy, happy and dopey mutants (loci that encode three subunits of the coatomer vesicular protein coat complex) (Coutinho et al., 2004) or a secondary effect of reduced internal pressure is as yet unclear. It is also worth noting that both Cav1 (this study) and Cav3 (Nixon et al., 2005) are expressed in the notochord of the zebrafish. This may reflect the importance of caveolae in this tissue, because both Cav1 and Cav3 can generate caveolae. In fact when we inject both MOs together we see an even greater effect on the notochord cells with the persistence of small round vacuoles and a greater loss of caveolae at 48 hours. The presence of caveolae in the strut-like interconnecting strands of closely apposed membranes in the notochord as well as their association with intermediate filaments suggest that caveolae may play a vital structural, or mechanosensory role in the notochord. When endothelial cells are subjected to fluid shear stress, the number of caveolae increases (Boyd et al., 2003;Rizzo et al., 2003). The high density of caveolae in the notochord might therefore be related to the high hydrostatic pressure in the notochord as the cells swell in response to water uptake into the glycosaminoglycan (GAG)-enriched vacuoles (Adams et al., 1990). Caveolae could also contribute a huge area of membrane that can be released into the bulk plasma membrane when surface expansion is required. The extremely high density of caveolae in the notochord means that this mechanism could provide up to an 80% increase in plasma membrane surface area if the caveolae were to flatten out into the bulk membrane.
In conclusion, these studies reveal a vital role for caveolin in notochord and neuromast development. How caveolae contribute to these vital developmental processes can now be tested in this highly tractable experimental system.

Embryo collection
Zebrafish embryos were raised, removed from their chorions and fixed in 4% paraformaldehyde as described previously (Westerfield, 2000). Animals were staged according to standard criteria (Kimmel et al., 1995) or as hours post fertilisation (h). Fixed embryos were stored at -20°C in methanol until required.

mRNA in situ hybridisation of whole mount zebrafish embryos
In situ hybridisation was carried out as previously described (Nixon et al., 2005).

Morpholino (MO) injection
cav1-MO1 antisense oligonucleotide was designed for flanking sequence to the ATG of zebrafish cav1␣ with the sequence 5Ј-TCCCGTCCT TGTATC CGCT -AGTCAT-3Ј, a second MO (cav1-MO2) 5Ј-TCATGCTGTCC CGTGCCTGAAG -TGC-3Ј and a standard control morpholino; control MO 5Ј-CCTCTTACC -TCAGTTACAATTTATA-3Ј. The morpholinos were obtained from Gene Tools, LLC (Philomath, OR) and diluted for injection in 1ϫ Danieau solution [58 mM NaCl, 0.7 mM KCl, 0.4 mM MgSO 4 , 0.6 mM Ca(NO 3 ) 2 , 5 mM HEPES, pH 7.6] with Phenol Red as an indicator. The cav1␣ morpholinos (0.75-4.5 ng) or control MO at the same concentration was injected into fertilised zebrafish eggs between the one-and two-cell stage. To confirm that the defects we see are specific we injected either MO alone at a relatively low concentration (0.75 ng/embryo) and saw little effect, but when we injected them together we saw the same effect on the notochord as we saw with the higher concentration.
The cav1-MOs were co-injected with a MO to p53 (Langheinrich et al., 2002) at concentrations from 0.3-3 ng/embryo and saw no rescue of the caveolin MO phenotype. A splice-blocking MO was designed to the exon2-intron2 splice junction with the sequence 5Ј-AAAGTTTGTCCTACCTTCACCACAT-3Ј and injected up to 9 ng/embryo.

Analysis of embryos
Live embryos were immobilised in embryo medium containing tricaine (0.2% 3aminobenzoic acid ethyl ester) and examined with an Olympus SZ12 dissecting microscope, images were captured using a DP-70 camera and Olympus DP controller program. Figures were generated using Adobe Photoshop 7 or CS2; modifications were applied to whole images only except Fig. 4B where images of two different focal planes were pieced together to show the whole embryo in focus.
The total body lengths (Zebrafish lengths) were measured in mm directly with an Olympus dissecting microscope, or images of curved embryos were captured and lengths determined using the measure tool in Adobe Photoshop CS2.

Western blot analysis of morpholino-injected embryos
Control-MO, cav1-MO1 or cav1-MO2 (3.0 ng/embryo) injected embryos at 48 hours were incubated in 100 l lysis buffer (150 mM NaCl, 1.0% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS, 50 mM Tris-HCl, pH 8.0) on ice for 30 minutes. The embryos were sonicated at 4°C for 15 minutes, incubated on ice for an additional 30 minutes and mixed by pipetting. The lysed mix was centrifuged at 1000 rpm for 1 minute. Protein concentration was determined using the BCA assay (Pierce). 10 g of sample was analysed using SDS-PAGE and immunoblotted using a polyclonal Cav1 antibody from Transduction Laboratories at 1:2000. The membrane was stained for 1 minute in Coomassie Blue followed by destaining in 50% methanol to show equal protein loading.

Antibody labelling of whole mount zebrafish embryos
Embryos were labelled as previously described (Nixon et al., 2005).

DASPEI labelling of whole mount embryos
Live embryos were labelled with the fluorescent dye 2-[4-(dimethylamino)styryl]-N-ethylpyridinium iodide (DASPEI; Molecular Probes, Eugene, OR) as described (Balak et al., 1990;Whitfield et al., 1996). Embryos were immobilised in Tricaine and neuromasts were counted at 72 hours. All clusters of two or more cells along the posterior lateral line labelled with DASPEI were counted as neuromasts. Images were captured using an Olympus AX-70 microscope and NIH image 1.62 software.

Electron microscopy
Embryos were fixed in 2.5% glutaraldehyde in PBS and then processed as described previously for SEM (Davies and Forge, 1987;Williams and Holder, 2000). Immunogold labelling of frozen sections was carried out on 3-day-old embryos fixed in 4% paraformaldehyde/0.1% glutaraldehyde and processed as described previously (Parton et al., 1997). Sections were labelled with a pan-cytokeratin antibody and an anti-caveolin antibody followed by labelling with PAG 15 nm for cytokeratin and anti-rabbit 10 nm gold for the caveolin according to standard techniques. Transmission electron microscopy was performed as described previously (Nixon et al., 2005). Transverse sections were prepared in the region of the distal portion of the yolk extension to compare different embryos at the same position along the rostral-caudal axis. Caveolae density was measured by examining random sections through the notochord and counting the number of surfaceconnected caveolae profiles per linear length of membrane (ignoring non-connected caveolae and coated pit profiles). The number of caveolae/m 2 was determined by taking the mean number of caveolae necks per 1 m length (30 hours, n=35; 48 hours, n=47) and multiplying by the section thickness (0.06 m). This method was also used to approximate the number of caveolae/m 2 in mouse adipocytes and was compared with a 20 m 2 glancing section from the same cells, where the number of caveolae were counted and 38 caveolae/m 2 was the result.
For electron tomography, 300 nm sections of 3-day post fertilisation embryos were stained and labelled with 15 nm gold as a fiducial marker. Dual axis tilt series from +60 to -60 degrees were obtained in a Tecnai F30 300 kV TEM using the SerialEM software package. Reconstruction and modelling of the data was performed using the IMOD suite of programs (Mastronarde, 1997).

Immunofluorescence of paraffin-embedded sections
Paraffin sections were de-waxed according to standard methods. Antigens were retrieved using a modified retrieval method as follows. Samples on slides were boiled three times for 3 minutes on high in a standard microwave in 10 mM sodium citrate (pH 6.0), allowed to cool to room temperature and then treated for 2 minutes with 0.05% trypsin. For labelling sections with the cytokeratin antibody, sections were treated with proteinase K at 20 g/ml for 10 minutes at room temperature. Slides were then labelled according to standard protocols. Briefly, slides were incubated in 0.1% saponin for 10 minutes, quenched in 50 mM NH 4 Cl for 10 minutes and blocked for 1 hour in 2% horse serum/0.2% BSA. Sections were incubated in primary antibody for 1 hour at 37°C [Pan-caveolin (Luetterforst et al., 1999), Transduction Laboratories polyclonal caveolin-1 antibody or DakoCytomation polyclonal wide spectrum cytokeratin antibody (DakeCytomation, CA)], washed four times for 5 minutes, incubated with secondary antibody at room temperature for 1 hour (anti-mouse/rabbit CY3). Samples were washed four times for 5 minutes and mounted in N-propyl gallate. Slides were imaged using an Olympus BX51 fluorescent microscope and captured using a DP-70 camera and Olympus DP controller program.