Division of Developmental Biology, The Roslin Institute and R(D)SVS, University of Edinburgh, Easter Bush, Midlothian, United Kingdom
Correspondence to: Megan G. Davey, Division of Developmental Biology, The Roslin Institute and R(D)SVS, University of Edinburgh, Easter Bush, Midlothian, UK EH25 9RG. E-mail: email@example.com
Cilia are microtubule-based organelles that project from the surface of the most cells in the embryo and adult body. They play an essential yet unelucidated role in signalling and physiological roles during development and loss leads to catastrophic consequences and death for the embryo (Hildebrandt et al., 2011). Cilia consist of a specialised plasma membrane surrounding a microtubule-based axoneme, which projects from a modified centriole, the basal body, which is orientated and docked to the plasma membrane (Ishikawa and Marshall, 2011). Cilia are classed as having either a 9+0 or 9+2 axoneme structure indicating whether the axoneme consists of nine outer microtubule doublets alone or also contains a central singlet pair. While 9+0 cilia are most often mono or primary non-motile cilia, and are only present as one per cell, 9+2 cilia are motile and have dynein arms and other modifications attached to the microtubules of the axoneme facilitating movement. A further important distinction between 9+0 and 9+2 cilia is that motile 9+2 cilia are typically found on multiciliated cells; a multiciliated cell may have as few as two 9+2 cilia, (e.g., ependymal cells of the adult spinal cord) (Alfaro-Cervello et al., 2012) or several hundred such as in the trachea (Dawe et al., 2007). Cells with multiple cilia are terminally differentiated, whereas a primary cilium can form transiently in G1 of a proliferating cell (Dawe et al., 2007).
Multiciliated cells are characteristic of tissues requiring movement of fluids or particles, with cilia orientated as such to allow particles to be moved by a coordinated beat (Satir and Christensen, 2007). Multiciliated epithelia have been studied postnatally in mammalian trachea and oviduct (Noreikat et al., 2012; Vladar and Stearns, 2007) whilst studies using Xenopus skin and kidney and the chicken choroid plexus offer an insight into embryonic ciliogenesis (Doolin and Birge, 1966; Stubbs et al., 2012). The process of ciliogenesis to produce a primary cilia is dramatically different from that in a multiciliated cell. Whereas primary cilia use an existing centriole as a basal body, a multiciliated cells must undergo de novo centriologenesis in order to produce multiple basal bodies to give rise to multiple cilia (Dawe et al., 2007; Vladar and Stearns, 2007). Multicilin is a master regulator of multiciliate fate, inducing the formation of multiple centrioles, cilia, and expression of FOXJ1 (Stubbs et al., 2012), which is itself a master regulator of motile cilia development of both single motile cilia like those of the node and ventral neural tube and of motile cilia of multiciliated cells such those of the respiratory tract (Yu et al., 2008; Cruz et al., 2010).
TALPID3 (KIAA0586) is an essential gene for vertebrate development. TALPID3 was first mapped in the recessive, embryonic lethal, polydactylous talpid3 (ta3) chicken breed at The Roslin Institute, Edinburgh (Davey et al., 2006). The TALPID3 protein localises to the centrosome in human, chicken, mouse, and zebrafish cells and mutations in model species cause a loss of primary, non-motile 9+0 cilia formation and, subsequently, a loss of Hedgehog pathway regulation (Bangs et al., 2011; Ben et al., 2011; Davey et al., 2006). A loss of normal processing and nuclear localisation of the family of GLI proteins (Ben et al., 2011; Davey et al., 2006), primary cilia-dependent processes, result in a loss of Hh signalling phenotype in some organs, such as the dorsalisation of the neural tube as well as a gain of Hh signalling phenotype in others, e.g., polydactyly (Davey et al., 2006) depending on the required combinations of GliRepressor (GliR) and GliActivator (GliA) proteins, which pattern specific tissues (Eggenschwiler and Anderson, 2007). ta3 chicken and talpid3−/− (ta3−/−) mouse embryos have one major difference in phenotype: left-right axis specification is randomised in talpid3−/− mice but normal in ta3 chickens (Bangs et al., 2011). In mice, specification of the left-right axis is dependent on motile 9+0 monocilia (McGrath et al., 2003). Chickens do not appear to have long motile cilia on node cells (Essner et al., 2002; Gros et al., 2009), which also lack FOXJ1 expression (Davey, unpublished data). However, long 9+0 monocilia, normally present on chick floorplate cells, which are presumed to be motile, are not present in ta3 embryos and cannot be rescued by FOXJ1 expression (Cruz et al., 2010) suggesting that as in the mouse, TALPID3 is also required for motile 9+0 ciliogenesis. It is clear, therefore, that TALPID3 is essential for 9+0 cilia formation. However, TALPID3 has not yet been shown to be necessary for 9+2 motile ciliogenesis, which is the aim of this study. In order to study 9+2 cilia cells in the chicken embryo, we first undertook an expression analysis of FOXJ1, to determine embryonic tissue that had the potential to have multiciliated cells. This analysis (below) suggested the embryonic prosencephalic choroid plexuses would be the most likely tissues to first have multiciliated cells.
The choroid plexuses (ChP) are secretory organs that develop from the roof plate of the telencephalon, diencephalon, and hindbrain soon after neural tube closure (Cohen and Davies, 1937; Dziegielewska et al., 2001). In the chick and mouse, a continuous ChP develops straddling the telencephalic-diencephalic boundary through the intraventricular foramina. The first morphological signs of differentiation occur at similar developmental stages: E3.5 in the chick and E11 in the mouse (Cohen and Davies, 1937; Sturrock, 1979). However, in the chicken, prosencephalic ChP develops in advance of the hindbrain ChP, whereas in mammals this order is reversed (Dziegielewska et al., 2001). In the mouse, specification of the choroid plaque prior to choroid plexus outgrowth is dependent on Wnt and Bmp expression from the roofplate and cortical hem, promoted by Gli3R and inhibited by Shh signalling (Hébert and Fishell, 2008) and in mice lacking Gli3, the telencephalic ChP is lost to dorsal expansion of the ventral telencephalon (Grove et al., 1998; Theil et al., 1999). However, in Ftm−/−/RPGRIP mutants, which lack normal cilia and have neural tube patterning very similar to that of that of ta3 chicken (Vierkotten et al., 2007) and similarities in forebrain patterning to Gli3 mutants (Besse et al., 2011; Delous et al., 2007), ChP genes Ttr1 and Bmp4, are maintained in the presumptive ChP epithelium and a telencephalic ChP appears to develop (Besse et al., 2011; Delous et al., 2007), suggesting dorsomedial telencephalic patterning in cilia mutants is present, if disorganised. The evagination of pseudo-stratified presumptive ChP epithelia grows and branches throughout embryonic development to produce a highly convoluted structure by day 8 of chicken development (Dziegielewska et al., 2001; el-Gammal, 1983). Studies of the development of the ChP in chicken embryos have demonstrated that multiciliated cells appear during embryonic development, although the exact stage of development has not been shown (Doolin and Birge, 1966; el-Gammal, 1981, 1983). In the mouse, FOXJ1 is expressed in the ChP from E13 (Lim et al., 1997) and cells subsequently become multiciliated at E17 in mouse (Banizs et al., 2005; Sturrock, 1979).
The Chicken Choroid Plexus Expresses FOXJ1 From 20HH (E3.5)
We examined the expression of FOXJ1 between E3.5–E8 in wild type (wt) and ta3chicken embryos, to determine regions of the developing chicken embryo that potentially could develop multiciliated cells, particularly during the period in which ta3 embryos are viable and easily identifiable. As has been described (Cruz et al., 2010), FOXJ1 is expressed in the floorplate and roofplate of the developing neural tube and is particularly evident in the dorsal hindbrain (Fig. 1A). These areas are known to develop long, motile 9+0 mono cilia but not multiple 9+2 cilia. However, an additional area of expression in the dorsal telencephalon and diencephalon was evident in wt at all stages examined (20/20; arrow = telencephalon, arrowhead = diencephalon, Fig. 1A, C, E), and in most ta3 embryos (5/7; arrow = telencephalon, arrowhead = diencephalon, Fig. 1B, D, F, G) although the telencephalic expression always appeared weaker than in wt. The variable loss of FOXJ1 correlates with the range of telencephalic patterning defects reported in ta3 chick embryos (Buxton et al., 2004; Ede and Kelly, 1964). Of the ta3 embryos that did express diencephalic FOXJ1, expression was not uniform and some embryos exhibited FOXJ1 domains that were smaller (arrowhead, Fig. 1G) or larger and disorganised (arrowhead, Fig. 1F) than wt FOXJ1 expression (arrowhead, Fig. 1E). In section, FOXJ1 was shown to localise to the epithelium of the choroid plaque in both wt and ta3 embryos (arrows Fig. 1H, I; see Supp. Movie, which is available online). This is similar to the Ftm−/− mouse mutant, which also lacks cilia and in which the telencephalic ChP forms but is disorganised (Besse et al., 2011). The ChP is well documented to contain multiciliated ependymal cells in human, mouse, and chicken and motile 9+2 cilia have been shown to develop during embryogenesis in the chicken (Doolin and Birge, 1966; el-Gammal, 1981, 1983) although the specific stage of development had not been determined.
The Choroid Plexus Transitions From Monocilia to Multicilia at E8
We then examined the morphology and localisation of cilia and centrosomes at 23HH (E4) in the forebrain ChP of wt and ta3 embryos, using antibodies against acetylated α tubulin (axonemes; green) and γ tubulin (centrosomes; red). At 23HH, the epithelial luminal surface of the ChP is highly ciliated, but these axonemes are short, suggestive of non-motile cilia (Fig. 2A, magnification of boxed area in A', arrows indicate short cilia). The ta3 ChP also has a high concentration of centrosomes at the luminal surface, but the accumulation of disorganised a tubulin at the lumen of the ChP obscured the presence of any axonemes although we have previously shown that ta3 embryos lack primary cilia (Yin et al., 2009; Fig. 2B). At 23HH, the ChP, in common with the neuroepithelium of the nervous system, is pseudo-stratified, so it was likely that these were monocilia present on small apical surfaces of neuroepithelial cells. To distinguish if these luminal cells were multiciliated, we examined the number of ciliary axonemes (acetylated α tubulin; green) or centrosomes (γ tubulin; red) per cell, visualising cell boundaries with phalloidin (actin; blue) at 23HH (E4) and 29HH (E6). We found that between E4–E6, short cilia are present at one per cell (Fig. 2C, magnification of boxed area in C' cells with single cilia indicated by arrow, two cilia in a single cell indicated with arrowheads; of 50 cells, 31 could be assigned a single cilia, 3 may have had two cilia, and other cells had no visible cilia), while most cells maintained one centrosome pair/cell (Fig. 2D, magnification of boxed area in D'; arrows indicate single centrosome/centrosome pair, arrowheads indicate a cell with two separate centrosomes; of 100 cells, 86 had one centrosome pair; 11 had two separate centrosomes, 3 had no visible centrosome); thus the forebrain ChP does not exhibit multiple cilia per cell before 29HH (E6).
We then undertook a similar analysis at 34HH–38HH (E8, E10, and E12) in telencephalic ChP using acetylated α tubulin (green) to visualise ciliary axonemes and phalloidin to determine cell boundaries (blue). During this period of development, the telencephalic and diencephalic ChP become morphologically distinct and both become increasingly folded and complex structures (el-Gammal, 1981, 1983; Fig. 3A). At 34HH (E8), we observed a dramatic transition in the ChP towards the formation of multiple cilia in single cells, with tufts of cilia seen frequently localising to single cells at the apex of the ChP (arrow, Fig. 3B). These cilia are long, frequently measuring more than 5 μm long (Fig. 3B), suggesting that they are motile cilia. Between 36–38HH (E10–E12), an increasing number of cells on the surface of the ChP become multiciliated (arrows, Fig. 3C), although there are still areas with long single cilia (arrowhead, Fig. 3C). At 36–38HH, axonemes projecting from epithelial cells on the surface of the ChP appear closely associated and co-ordinated (Fig. 3D). In conclusion, 34HH (E8) ChP presented an excellent stage and tissue in which to examine multiciliated cells in ta3embryos.
As has been shown earlier, ta3 cells have an abnormal localisation of tubulin, which makes it difficult to determine if cilia are present. We, therefore, undertook scanning electron microscopy (SEM) on wt (Fig. 4A) and ta3 34HH (E8) ChP to visualise the surface of the ependymal cells for cilia. Wt telencephalic ChP confirmed what we had observed with antibodies. At 36HH (E8), the surface of the ependymal cells, the epithelial component of the ChP, are characteristically pine cone–shaped (Doolin and Birge, 1966) and are covered in microvilli and have long cilia, with bulb-like tips (Fig. 4C, arrow indicating pine cone–shaped cell, covered in short microvilli, visible through a surrounding mass of long cilia). We observed areas, particularly on the lateral surfaces of the ChP, where cilia were very closely associated (Fig. 4B, B' arrows), suggesting these were multiple cilia projecting from a single cell, interspersed with other cells projecting long monocilia (Fig. 4B arrowhead). This differed from cilia at the apex of the ChP, which appeared more evenly multiciliated (Fig. 4C). ta3 forebrain ChP were smaller and less branched than wt ChP, and due to abnormalities in brain morphology it was difficult to assess if ChP were associated with the telencephalon or diencephalon. On the surface of the ta3 forebrain ChP, although the microvilli and “pine cone” appearance are evident, there are almost no cilia, although infrequently a lone monocilia was seen (arrow, Fig. 4D, D').
ta3 Fails to Form Motile Cilia But Does Form Multiple Centrioles
EM analysis of 34HH (E8) telecephalic ChP confirmed that the cilia observed in Wt choroid plexus are motile cilia with a 9+2 structure and that there are multiple cilia per cell (asterisks mark axonemes, red arrowheads mark basal bodies, arrows mark tight junction at cell boundary Fig. 5A, B). We also observe ciliary pits that contained multiple closely associated cilia (Fig. 5C), which likely represent the tightly associated axoneme bundles observed in IHC and SEM (Figs. 3D, 4B, B'). At 34HH (E8), the axonemes and basal bodies of Wt cilia were always observed to have the same basal-apical orientation, although these bundles of cilia did not have a uniform proximodistal axis (probably as a result of the pine cone–shaped ependymal cell) and, therefore, sections of axonemes and basal bodies arising from the same ciliary pit revealed subtle changes in the structure of the basal bodies, particularly as they enter the transition zone, the point at which the microtubules of the basal body transform into those of the ciliary axoneme (Fig. 5D; basal bodies observed in various different proximo-distal levels 1, 2, 4–6, with 1 the most proximal before the transition zone and 6 the most distal, as an axoneme emerging from the ciliary pit). For clarity, we have reproduced the structures in higher resolution in Figure 5D', including a basal body in the transition zone (3) from another ciliary pit and a mature axoneme, and include a full description in the legend to Figure 5. However, in short, we observed a wide variety of basal body, transition zone, and axoneme structure including microtubule triplets (1), incomplete C-tubules (2), subdistal appendage (3), the transition zone and distal appendages (4), Y-fibres (5), and 9+2 axoneme structure (5).
EM sections of ta3 choroid plexus confirmed that cilia do not form although ta3 cells form multiple centrosome pairs per cell. The failure of ciliogenesis appears to lie in the inability of ta3 centrosomes to translocate or become orientated to the apical surface of the cell. ta3 centrosomes were rarely found near the cell surface and were often mislocalised deep within the cell and did not show co-ordinated orientation within a group (Fig. 5E, asterisks indicate centrioles, arrowheads indicate microvilli, arrows indicate tight junctions between cells). Occasionally, long basal bodies were observed projecting axoneme-like structures into the cell (asterisk in Fig. 5F).
Having determined that the choroid plexus has motile multiciliated cells at 34HH (E8), we were able to show that motile cilia, as well as primary cilia (Yin et al., 2009), are absent in ta3 embryos. Importantly, we show that the lack of cilia in the ta3 embryo is not due to a failure of centriologenesis, nor axoneme formation but is likely due to the inability of the ta3 centrioles to translocate to or orientate with the apical surface of the cell. We are primarily interested in the behaviour of the ta3 centrosome, where the TALPID3 protein normally localises and is thought to act (Ben et al., 2011; Yin et al., 2009). Future comparison of ta3 centrosomes to the stereotypical behaviour and localisation of ta3 basal bodies during multiciliated cell ciliogenesis will help us determine the function of the TALPID3 protein.
Phenotypes of human ciliopathic conditions are complex; a large number of organs are affected due to abnormal patterning through loss of Hedgehog and Wnt signalling during embryogenesis, while loss of motile cilia is generally thought to cause to health problems in the adult, such as respiratory disease and infertility (Hildebrandt et al., 2011). Motile cilia also contribute to patterning during embryogenesis, such as to the development of polarised stereocilia in the inner ear (Eggenschwiler and Anderson, 2007) but except for the specialised 9+2 inner ear kinocilium, the role of motile cilia in development has not been as well explored as that of the primary cilium. Specialised primary cilia in the embryo can have osmosensory and mechanosensory functions, such as the cholangiocytes of the liver, which are essential for normal bile production (Masyuk et al., 2008). The motile cilia of the ChP have a similar specialised role in cerebrospinal fluid production and physiology and a loss of motile ChP contributes to development of hydrocephalus (Banizs et al., 2005) and it is possible that motile cilia may fulfill this role in other embryonic locations. Having determined that ta3 embryos lack motile cilia will allow us to understand the various contributions 9+0 or 9+2 cilia make during embryonic development, particularly to less well-understood organ systems such as the pancreas.
The ta3 chicken phenotype has been consistent with a mammalian ciliopathy phenotype, in all respects, except that of left-right axis specification (Bangs et al., 2011). As is observed in all organ systems analysed thus far, the ta3 telencephalon lacks expression of PTCH1 and GLI1, while expression of GLI2 and GLI3 are normal (Buxton et al., 2004). The loss of normal GliA:GliR protein ratio is the same in different species, zebrafish, chicken, and mouse, in which a loss of TALPID3 has been modelled (Ben et al., 2011; Davey et al., 2006; Davey, unpublished data) and is similar to other mouse models that lack primary cilia and have a similar phenotype (Huangfu and Anderson, 2005). This suggests that abnormal patterning in the ta3 telencephalon is also due to the same loss of both GliA and GliR function, consequential to a loss of primary cilia formation, which underpins abnormal patterning of the other organs (e.g., limb). The loss of GliR alone, as is seen in the Gli3−/− mouse or Gli3 hypomorphic Pdn/Pdn allele, causes a loss of telencephalic ChP (Kuschel et al., 2003; Theil et al., 1999) due to a loss of dorsomedial telencephalon. In contrast, forebrain ChP is maintained in both ta3and the Ftm−/−/RPGRIP mutants, although patterning is not normal and the dorsomedial telecephalon appears disorganised. In Gli3−/− and Pdn/Pdn mice, the telencephalic-diencephalic boundary is lost, as shown by a loss/reduction of Emx1 and Emx2 in the telencephalon as a consequence of an increased domain of Fgf8 expression (Kuschel et al., 2003; Theil et al., 1999). Interestingly, Bmp4, a regulator of ChP fate, is lost from the dorsomedial telecephalon but not the diencephalon (Kuschel et al., 2003) demonstrating that diencephalic ChP gene expression is not sufficient to rescue ChP formation in the absence of a telencephalic-diencephalic boundary. In Ftm−/− mutants, Fgf8 in the Anterior Neural Ridge (ANR) is unchanged (Besse et al., 2011) and decreased in the ta3 ANR; the telencephalic-diencephalic boundary is thought to be intact in both mutants (Besse et al., 2011; Buxton et al., 2004) and this seems to be central to the maintenance of ChP formation. Genes that control the telencephalic-diencephalic boundary have been shown to have dual roles in ChP induction; overexpression of Emx1 and Emx2 negatively regulates ChP formation (Von Frowein et al., 2006) while a loss of Emx2 also causes a loss of ChP (Subramanian and Tole, 2009). Therefore, it is not the EMX genes themselves that supress ChP but the context of the molecular anatomy they confer. The hindbrain ChP forms at the boundary between the roof plate organiser tissue and lateral neuroepithelial tissue, and is controlled by Delta-Notch signalling (Broom et al., 2012). Forebrain ChP cells probably also derived from the cortical hem, the dorsal organiser of the prosencephalon, through an unelucidated mechanism (Hébert and Fishell, 2008; Subramanian and Tole, 2009). Considering the location and dependence on the telencephalic-diencephalic boundary for ChP development, it seems likely that a similar context-dependent boundary organiser will be found to be essential for forebrain ChP induction.
Eggs from talpid3 flock (M.G. Davey, The Roslin Institute) were incubated at 38°C for 3–12 days, and staged as per Hamburger and Hamilton (1951). For SEM and EM, embryos were dissected into PBS, avoiding contamination with yolk, heads were removed and placed into 4% PFA, 2.5% Glutaldehdye in 0.1M cacodylate buffer. The choroid plexus was immediately removed and placed into fresh fixative (as previous) for 24 hr. Tissue was prepared and visualised for EM as per Davey et al. (2007) and the axoneme/basal body structure was compared to the observations of Paintrand et al. (1992). Specimens were rinsed 2 × 15 min in 0.1M cacodylate buffer, then placed in 1% osmium tetroxide 1 hr, rinsed 2 × 15 min in dH2O, then dehydrated through a grade ethanol series, and critically point dried using a BAL-TEC CPD 030. Samples were then mounted on Al stubs using carbon adhesive tabs and coated with 45 nm Au/Pd using a Cressington 208HR sputter coater and examined using a Philips XL30 ESEM operating at an accelerating voltage of 5 or 15 kV. For section immunocytochemistry, embryos were dissected into PBS, fixed, sectioned, and stained as per Davey et al. (2006). Antibodies used were acetylated α tubulin (Sigma-Aldrich T7451, St. Louis, MO), γ tubulin (Sigma-Aldrich T5192), anti-mouse (Life Technologies A11017, Carlsbad, CA), anti-rabbit (Life Technologies A21207). Phalloidin (Life Technologies A22284) was used to visualise actin structures. In assigning centrosome/cilia number to cells, cells were only included if the entire border of the cell was visible with phalloidin in section. Cells without clearly defined cell borders were not counted. Cells were then assessed for centrosome number according to the number of γtubulin foci observed within the cell border. Association of cilia to individual cells was assessed by proximity of an acetylated α tubulin–positive axoneme to phalloidin boundary usually independently of centrosome association, although the centrosome was used to confirm cell association in locations when the cell:axoneme association was not clear. All comparisons between wt and ta3 embryos were undertaken in staged matched siblings.
Wholemount RNA In Situ Hybridisation and OPT
Wholemount RNA in situ hybridisation was carried out as per Nieto et al. (1996). FOXJ1 probe was produced as by Cruz et al. (2010). For OPT, embryos were washed extensively in TBST mounted and visualised as per Bangs et al. (2010). All comparisons between wt and ta3 embryos were undertaken in staged matched siblings.
The talpid3 flocks were maintained at the Roslin Institute, Edinburgh, under a Home Office licence and after ethical review. The justification for their maintenance is primarily that they are used as animal models for human conditions.
M.G.D., L.A.S., L.M., G.M.D. are supported by BBSRC Career Track Fellowship funding to M.G.D. (BB/F024347/1). K.E.M. was supported by a summer project studentship from the Nuffield Foundation. A.B. is supported by BBSRC funding to the e-Chick Atlas (BB/G000816/1). The Roslin Institute receives Institute Strategic Grant funding from the BBSRC.