a Schematic of CRISPR/Cas9 strategy targeting exon3. b The cfap20^ua5025 allele (cfap20^−/−) has a 1 bp deletion, 8 bp insertion (c.209delGinsTCGAGCTA; p.Gly70Valfs*29) that produces a frameshift at residue 70 and predicts loss of the majority of this deeply conserved protein (see also Supplementary Fig. ¹). c cfap20^−/− mutants have dramatically reduced abundance of transcript compared to wildtype at 48 h post fertilisation (hpf) (n = 4, error bars = SEM; two-tailed unpaired T-tests, gapdh P = 0.2912; cfap20 P = 1.54 * 10⁻⁷). d, e cfap20^−/− larvae exhibit anterior-posterior body axis kinks/ventral curvature (arrows) displayed at 48 hpf and 7 days post-fertilisation (dpf). e cfap20^−/− larvae develop pronephric duct cysts (arrowheads) at 7 dpf. f cfap20^−/− mutants exhibit significantly increased incidence of left-right pattern defects compared to wildtype (cardiac situs, n-value per condition shown, Fisher’s exact test, P = 0.0387). g, h cfap20^−/− adult homozygotes develop severe spine curvature compared to age-matched wildtype at 4 months post fertilisation. These motile ciliopathy phenotypes are also observed following knockdown of cfap20 in zebrafish (see Supplementary Figs. ³, ⁴). CRISPR clustered regularly interspaced short palindromic repeats, WT wildtype. Source data are provided as a Source Data file.

a C. elegans CFAP-20 is expressed specifically in ciliated sensory neurons and localises to the inner junction-containing proximal segment of the ciliary axoneme. Shown are fluorescence microscopy images of GFP-tagged CFAP-20 (CFAP-20::GFP) localising to cilia in the head (including amphid/labial neurons) and tail (phasmid neurons). Schematics show the relative locations and structural features of different cilia at the dendritic ends of sensory neurons in the head and tail. PS, proximal segment; DS, distal segment. b Loss of C. elegans CFAP-20 leads to longer phasmid (Kruskal-Wallis and Dunn’s Test; P values: CHE-11::GFP: P values: WT vs cfap-20 = 0.0186; WT vs pcrg-1 = 0.9999; cfap-20 vs pcrg-1 = 0.1634; OSM-3::GFP: WT vs cfap-20 = 0.0001; WT vs pcrg-1 = 0.6374; cfap-20 vs pcrg-1 = 0.0336) but not ASE cilia (one way ANOVA and Tukey Mann test; GFP (ASE): WT vs cfap-20 = 0.4322; WT vs pcrg-1 = 0.1976; cfap-20 vs pcrg-1 = 0.0097). Graphs represent ciliary length measurements using the cilium-localised CHE-11::GFP and OSM-3::GFP fluorescence reporters in phasmid cilia from wild-type animals and pcrg-1 mutants, or soluble GFP marker for ASE cilia. Loss of PACRG (pcrg-1 mutant) does not affect ciliary length. Measurements are from basal body (BB) to ciliary tip for the intraflagellar transport (IFT) markers, and transition zone (TZ) to the ciliary tip for the GFP reporter. Box plots represent minima, 25^th percentile, median, 75^th percentile, maxima. c C. elegans CFAP-20 is required for the structural integrity of the axoneme inner junction. The schematic shows normal ciliary/microtubule ultrastructures at the level of the transition zone, middle segment, and distal segment. Transmission Electron Microscopy (TEM) cross-sections of C. elegans amphid (head) cilia reveal ultrastructure defects in the cfap-20 mutant, namely a break at the inner junction (IJ) in transition zone and middle segment doublet microtubules (examples of the defects shown with white arrows). Source data are provided as a Source Data file.

a C. elegans cfap-20 and pcrg-1 single and double mutants show defects in gustatory plasticity. Wild-type animals pre-exposed to a high NaCl concentration (100 mM) display a reduced attraction to NaCl in chemotaxis assays involving 0 mM and 25 mM NaCl quadrant plates. After exposure to high-NaCl conditions, cfap-20 and pcrg-1 single and double mutants still have significantly increased attraction to NaCl compared to wild-type (one-way ANOVA and Bonferroni test P values: WT vs cfap-20 = 0.005; WT vs. pcrg-1 = 0.0001; WT vs cfap-20;pcrg-1 = 0.006; cfap-20 vs pcrg-1 = 0.9999; cfap-20 vs cfap-20;pcrg-1 = 0.9999; pcrg-1 vs cfap-20;pcrg-1 = 0.9318). b C. elegans cfap-20 and pcrg-1 single and double mutants have reduced lifespans. Staged animals are cultured on plates at 20 °C and monitored every 1–2 days for signs of life. Animals with impaired intraflagellar transport and/or prominent cilia structure defects, such as the che-11 control mutant shown, exhibit an enhanced lifespan. The cfap-20 single and double mutants (with pcrg-1) undergo a statistically-significant reduction in longevity (Log rank Mantle-Cox test; P values: WT vs cfap-20 = 0.0001; WT vs pcrg-1 = 0.0001; WT vs cfap-20;pcrg-1 = 0.0001; WT vs che-11 = 0.0001; cfap-20 vs pcrg-1 = 0.0001; cfap-20 vs cfap-20;pcrg-1 = 0.0755; cfap-20 vs che-11 = 0.0001; pcrg-1 vs cfap-20;pcrg-1 = 0.0024; pcrg-1 vs che-11 = 0.0001; cfap-20;pcrg-1 vs che-11 = 0.0001). c C. elegans cfap-20 and pcrg-1 single and double mutants exhibit a reduced locomotion (roaming defect) common in cilia mutants. The tracks of individual animals are scored with the help of a grid after 24 h. The che-11 mutant, which has IFT defects and major cilia structure anomalies, is shown together with wild-type as a control (one-way ANOVA and Tukey’s test; P values: WT vs cfap-20 = 0.0066; WT vs pcrg-1 = 0.0005; WT vs cfap-20;pcrg-1 = 0.0001; WT vs che-11 = 0.0001; cfap-20 vs pcrg-1 = 0.9426; cfap-20 vs cfap-20;pcrg-1 = 0.0532; cfap-20 vs che-11 = 0.0001; pcrg-1 vs cfap-20;pcrg-1 = 0.2742; pcrg-1 vs che-11 = 0.0001; cfap-20;pcrg-1 vs che-11 = 0.0001). d C. elegans cfap-20 and pcrg-1 single and double mutants have small body sizes (one-way ANOVA and Tukey’s test; P values: WT vs cfap-20 = 0.0001; WT vs pcrg-1 = 0.0001; WT vs cfap-20;pcrg-1 = 0.0001; WT vs che-11 = 0.0001; cfap-20 vs pcrg-1 = 0.4094; cfap-20 vs cfap-20;pcrg-1 = 0.0001; cfap-20 vs che-11 = 0.0001; pcrg-1 vs cfap-20;pcrg-1 = 0.0001; pcrg-1 vs. che-11 = 0.0001; cfap-20;pcrg-1 vs che-11 = 0.0001). The body lengths of staged animals were measured at young adulthood (72 h). All box plots represent minima, 25^th percentile, median, 75^th percentile, maxima Source data are provided as a Source Data file.

a Four families (F1 through F4) presenting with early-onset, progressive Retinitis Pigmentosa (RP) from two Ophthalmology Centres (UK & Canada). In each pedigree, biallelic variants in CFAP20 (M1 thru M6) segregate with RP. Family 1 also exhibits polycystic kidney disease that segregates separately from RP and CFAP20. Arrow indicates proband in each pedigree (Details of symptoms are in Table 1 and summarized in Supplementary Table ³; further details of CFAP20 gene variants are in Supplementary Table ²). b The residues altered in each of the four patient variants (white) are predicted to alter residues in the CFAP20 protein (red) that are on the surface of the protein and therefore may impact upon interactions with PACRG (blue). The variants likely alter key components of the microtubule inner junction (see Fig. 1b for context). Protein structure derived from PDB file 6VE7²¹. Colour fundus photographs of the left eye (c, j; F1:1 and F4:1 respectively) show a generalized retinopathy with peripheral pigment deposition, attenuated retinal vasculature and marked chorioretinal atrophy (c) and peripapillary atrophy (j). Autofluorescence imaging (d, f, h, k; F1:1, F2:1, F3:1, F4:1 respectively) demonstrates widespread decreased autofluorescence (d, f) indicating significant atrophy of retinal pigment epithelium (RPE) and focal lesions of RPE atrophy at the macula and midperiphery (k) and predominantly the inferior retina (h). Optical coherence tomographic imaging (e, g, i, l; F1:1, F2:1, F3:1, F4:1, respectively) shows significant loss of outer retinal layers (e, g) with some surviving outer retina at the fovea (i, l).

a Half life calculations assuming first-order decay (ln([relative protein]t = x/[relative protein]t = 0)). Lysates from transfected HEK293T cells revealed a significant (one-way ANOVA) decay for all variants, except for p.(Arg102His), compared with Wildtype (WT) protein (half-life of ~523 min). The fastest decay was variant p.(Arg113Trp) with a half-life of ~93 min (n = 3; error bars = SEM). b In C. elegans, human CFAP20 patient variants mislocalise in comparison to wildtype CFAP20. The GFP-tagged CFAP20 proteins bearing patient mutations concentrate mostly to the proximal segment (PS) of the axoneme, which is similar to wild-type CFAP20 protein expressed in C. elegans. However, patient variants also mislocalise (asterisks) in the distal segment (DS) and/or the dendritic region to the left of the phasmid cilia (especially p.(Arg102His)). XBX-1 is an IFT reporter that marks the basal body (BB) and entire axoneme. c Wildtype human CFAP20 mRNA rescues zebrafish cfap20 mutants, as compared to control mRNA (mCherry) or CFAP20 mRNA bearing patient variants. A representative cfap20^+/− incross is displayed, injected with WT/patient variant (p.(Arg113Trp)) CFAP20, or control mRNA. Mendelian ratios of 25% body axis curvature (red arrows) were expected where treatments failed to rescue the cfap20^−/− homozygotes. Larvae displaying mRNA toxicity phenotypes (i.e. oedema/dorsalization) occur rarely (black arrowhead). d The frequency of rescued (straight tail) and not rescued (curled tail) cfap20^−/− homozygotes were quantified for each missense human variant. p.(Tyr86Cys), which is associated with syndromic inherited retinal disease (IRD), failed to rescue body curvature. Conversely, the p.(Arg102His), p.(Arg113Trp), and p.(Arg153Gly) variants, that we associated with non-syndromic IRD, are not significantly different from WT CFAP20 mRNA (Fisher’s exact test; p.(Tyr86Cys) P = 0.0021; sample sizes, i.e. the number of cfap20^−/− homozygous mutant larvae, are indicated at the bottom of each bar). See also Supplementary Fig. ⁸. Source data are provided as a Source Data file.

a The larval, 7 days post-fertilisation, cfap20^−/− homozygote retina is indistinguishable from wildtype (WT) animals (green bracket = rod and blue/red/green cone OS and RPE, white arrowhead = UV cone OS). By 1.5 months post-fertilisation (mpf) the homozygote retina is degenerate with loss of organised cone subtype layers (green bracket = rod OS and RPE, white arrowhead = UV cone OS, black arrowhead = blue/red/green cone OS) and poorly defined OPL (asterisk). b, c By 4 mpf, the cone mosaic is lost in homozygotes and hyperreflective blebs (arrows) are observed in the outer retina (Optical Coherent Tomography B-scans (cross section) and en face projections). The photoreceptor and OPLs are thinner in homozygotes (n = 5; error bars = SEM; two-tailed unpaired Mann-Whitney, pR P = 0.0079, OPL P = 0.0238). Representative full-field electroretinogram traces from 4 mpf dark (DA) or light adapted (LA) animals presented in d. e At both ages, the DA b-wave is reduced in homozygotes (two-tailed unpaired T-tests; 1.5 mpf: n = 8 WT, 4 cfap20^−/−; P = 0.0382; 4 mpf n = 19 WT, 7 cfap20^−/−; P = 0.0017) whereas the LA b-wave is reduced only at 4 mpf (two-tailed unpaired T-tests; 1.5 mpf: n = 14 WT, 4 cfap20^−/−; P = 0.6010; 4 mpf: n = 20 WT, 7 cfap20^−/−; P = 5.60 * 10⁻⁵). Box plots: minima, 25^th percentile, median, 75^th percentile, maxima. f Rod and cone layers are shorter and became disorganised over time, with cone OSs detachment at 4 and 8 mpf (arrowheads). g False-coloured TEM micrographs reveal reduced density and OS dysmorphia of homozygotes at 8 mpf. Detached OSs (arrows) are engulfed by RPE. h Apoptotic cells (arrowheads) are observed in the outer retina of homozygotes but never WTs controls. IPL inner plexiform layer, INL inner nuclear layer, OPL outer plexiform layer, ONL outer nuclear layer, pR photoreceptor layer, IS inner segment, OS outer segment, RPE retinal pigment epithelium. See also Supplementary Figs. ⁹ and ¹⁰. Source data are provided as a Source Data file.

Ciliopathies most commonly arise from the dysfunction of well-characterised functional modules (ciliopathy modules), namely the basal body, the transition zone (TZ) ciliary gate, intraflagellar transport (IFT) or IFT-associated BBS cargo-trafficking system, a variety of signalling proteins, and a motility apparatus specific to motile cilia. Distinct from these macromolecular complexes are proteins within and near the IJ including CFAP20, PACRG and potentially CFAP52, which may represent a previously unknown ciliopathy module that influences ciliary beat patterns (motility) likely by acting via radial spokes (RS) and the Nexin-Dynein Regulatory Complex (N-DRC). Our study confirms a role for zebrafish CFAP20 in metazoan motile cilia, implicating it as a strong candidate for the motile ciliopathy, primary ciliary dyskinesia (PCD). In C. elegans, CFAP20 and PACRG independently regulate the signalling functions of different classes of non-motile cilia. Lastly, mutations in human CFAP20 result in retinal dystrophy, a ciliary photoreceptor degeneration phenotype that is phenocopied in zebrafish CFAP20 mutant animals. In these non-motile cilia, the IJH may influence, structurally and/or functionally, currently unknown signalling proteins, and represents a previously unknown disease locus for photoreceptor degeneration. IDA Inner Dynein Arm, ODA Outer Dynein Arm, BBS Bardet-Biedl syndrome. Shapes and symbols in this schematic are also represented in the complementary Fig. 1.