We genotyped 5 cases and 15 controls with the Affymetrix v2 Canine array. We found a 15-Mb segment of autozygosity on chromosome 34 that was shared by all cases (genome-wide P < 0.001; Fig. 1b). The shared region contained 151 genes. We mined the ciliary proteome⁴ and ciliome⁵ databases and identified ten proteins that had been discovered in at least two independent genomic or proteomic studies of cilia enrichment. We sequenced the coding exons and intron-exon boundaries of six of these candidates in cases and controls and identified a stop codon (p.Arg96X) in CCDC39 (Gene ID: 488089) in the affected dogs that was predicted to truncate 90% of the 976–amino acid CCDC39 protein (Fig. 1b). All of 10 additional cases were homozygous for the p.Arg96X alteration, all of 10 obligate carriers were heterozygous for it and 8 of 102 randomly sampled healthy Bobtails were heterozygous for it; we did not find the alteration in 80 healthy animals from 9 other breeds. We sequenced CCDC39 RT-PCR products from the tracheal RNA of a carrier and found a mutant to wildtype allelic ratio of about 0.25, compatible with nonsense-mediated RNA decay of transcripts containing the p.Arg96X alteration (Fig. 1b).

FAP59, the Chlamydomonas ortholog of CCDC39, was predicted to be essential for motile ciliary function, as orthologs do not occur in nonciliated organisms (‘CiliaCut’) or in Caenorhabditis elegans (‘MotileCut’)⁶. FAP59 was also among the top 50 of the 650 proteins that were detected by mass spectrometry in purified flagella of Chlamydomonas, consistent with ciliary localization⁷. Mouse Ccdc39 was predicted to be present in cilia because Ccdc39 is strongly expressed in tissues rich in ciliated cells; Ccdc39 was also shown by in situ hybridization to be expressed in olfactory and vomeronasal sensory neurons and the respiratory epithelium⁸. To extend these findings, we performed in situ hybridization at different stages of mouse development and identified specific expression of Ccdc39 in node cells carrying motile cilia, in upper and lower airways, and in ependymal and choroid plexus cells, consistent with Ccdc39 having a functional role in motile cilia (Fig. 2a). In tissue from adult humans, quantitative RT-PCR (qRT-PCR) showed predominant expression of CCDC39 in nasal brushings and, to a lesser extent, in lungs and testes. However, this expression was systematically lower than that of other PCD genes (expression of DNAI1 > DNAI2 > LRRC50 ≈ C14orf104 (also called KTU) > CCDC39; Supplementary Fig. 2a,b).

To provide further support for the role of CCDC39 in ciliary motility, we used morpholino-based suppression of ccdc39 in zebrafish embryos. Using reciprocal BLAST, we identified the only Danio rerio ortholog of CCDC39 (LOC555319; 53% identity and 73% similarity). We amplified this transcript readily from embryonic complementary DNA (cDNA) as early as the shield stage (data not shown). Injection of either a translation-blocking or a splice-blocking morpholino at the two-cell stage caused a dose-dependent increase in heart-looping defects at 36 hours post-fertilization (rightward or absent looping; Fig. 2b,c and Supplementary Fig. 3) and bilateral or lack of spaw expression at the left lateral plate mesoderm in 14 somite embryos (Fig. 2d,e). These phenotypes were likely specific, as they were reproduced at similar frequencies by both morpholinos and were rescued by co-injection with wildtype human CCDC39 mRNA (Fig. 2d). These findings recapitulate laterality defects seen in other PCD morphants (for example, C14orf104, also known as ktu, and lrrc50)^9,10, consistent with compromised fluid flow at the Kupffer's vesicle due to impaired ciliary motility.

These findings prompted us to screen for mutations in unresolved human PCD cases (Supplementary Note). We sequenced the 20 coding exons and intron-exon boundaries of human CCDC39 (Gene ID: 339829) in 53 cases (from 50 families) with axonemal defects reminiscent of those observed in Bobtails, that is, the absence of inner dynein arms in all examined cilia and coexistence of axonemes with various ultrastructural defects within the same section (Fig. 3a and Supplementary Fig. 4). This group accounts for 5–15% of PCD cases¹¹. We detected 14 unambiguous loss-of-function mutations in 19 of the 50 families (38%): four nonsense mutations, six frameshifts leading to premature stop codons and four splice-site variants (located within 3 bp of an intron-exon boundary) (Fig. 3b and Table 1). Notably, no nonsense or frameshift mutations in CCDC39 were reported in the latest 1000 Genomes Project data release (see URLs) for 60 sequenced individuals. Transmission of mutations was consistent with autosomal recessive inheritance (Table 1 and Supplementary Fig. 5). Fifteen cases (ten of whom were consanguineous) were homozygous and five were compound heterozygous. Two cases carried only one pathogenic allele (DCP580 and DCP481). Nasal epithelial cells were available for DCP580, allowing us to investigate the integrity of the CCDC39 mRNA by RT-PCR. One primer pair yielded a unique 537-bp amplicon in control subjects as well as an additional 644-bp product in DCP580 because of a 116-bp insertion between exons 9 and 10. Sequencing of the corresponding genomic region showed that this insertion resulted from activation of a pseudo-exon in intron 9 (designated ψExon9). The sequence 5′ of ψExon9 matches a perfect acceptor splice site, and its 3′ boundary contains a near-perfect splice donor site that is activated by an AT>GT transition involving nucleotide c.1167+1261 (Supplementary Fig. 6). This mutation (which, if translated, would lead to the frameshift alteration p.Glu390SerfsX6) was not found in any other case. A second loss-of-function mutation has not yet been identified for DCP481, but two previously unknown candidate variants are worth mentioning: a missense variation involving a highly conserved residue (p.Thr594Ile) and a putative branch site variant in intron 9 (c.1168-32A>G). We also detected two rare missense variations in the heterozygous state among the 31 remaining cases (p.Cys803Tyr and p.Thr182Ser), but their effect on CCDC39 function is unknown.

Of the 15 unambiguous disease-causing mutations, 11 were private, whereas 4 were shared by several families not known to be related (Table 1). Genotyping of ten microsatellites that flank CCDC39 and span ~9 Mb of the disease locus supports a founder effect for three of the mutations and/or alterations (p.Thr358GlnfsX3, p.Ser786IlefsX33 and c.357+1G>C). We found the p.Glu731AsnfsX31 alteration on a shared haplotype in three families, but on a distinct haplotype in the fourth family (DCP91), which suggests a recurrent event (Supplementary Fig. 7).

The PCD phenotype of individuals with recessive CCDC39 mutations is characterized by chronic upper and lower airway infections that cause severe morbidity. Nine cases (41%) showed situs solitus and ten cases (45%) had situs inversus, whereas three cases (14%) had heterotaxia. Two of the cases with heterotaxia had documented polysplenia (Ivemark syndrome; MIM208530). This is consistent with previous reports that randomization of left-right asymmetry not only results in situs solitus or situs inversus but can also cause situs ambiguous, including Ivemark syndrome^12,13. Four affected males had oligoasthenospermia.

To determine whether CCDC39 mutations underlie PCD with distinct ultrastructural defects or heterotaxia in the absence of respiratory symptoms, we used massive parallel sequencing to scan the 20 coding exons for mutations in 24 independent cases of PCD with absence of inner dynein arms only (6 cases) or absence of both inner and outer dynein arms (18 cases) without axonemal disorganization, and in 216 sporadic heterotaxia cases and 216 ethnically matched controls from the United States (Supplementary Note). We amplified 25 ~250-bp amplicons that spanned the 20 CCDC39 exons from DNA pools of ~20 cases or controls, mixed and sequenced on a Roche FLX instrument. The sensitivity and specificity of a similar protocol have been estimated at 83% and 98%, respectively (Y.M. & M.G., unpublished data). Contrary to PCD cases with absence of inner dynein arms and axonemal disorganization, we detected no nonsense, frameshift or splice site variants in CCDC39 in these cohorts.

To gain insights into CCDC39 function, we analyzed its expression at the protein level. In protein blotting, antibodies to CCDC39 recognized a band of the expected size (~110 kD) in extracts from respiratory epithelial cells of a control subject but not of a CCDC39-deficient individual (Supplementary Fig. 8). High-resolution immunofluorescence analysis of respiratory cells showed predominant axonemal staining in controls (in agreement with the mass spectrometry analyses of purified flagella of Chlamydomonas⁷) but not in cells containing mutations in CCDC39 (Fig. 4).

We used immunofluorescence to characterize the effect of CCDC39 deficiency on cilia structure. TEM analysis pointed toward defective inner dynein arms, nexin links and radial spokes but normal outer dynein arms (Fig. 3a). Because outer dynein arm complexes (ODA type 1 and type 2) vary along the length of ciliary axonemes and TEM cannot easily identify incomplete or regional outer dynein arm defects^9,13, we used antibodies to the outer dynein arm components DNAH5 (type 1 and 2), DNAI2 (type 1 and 2) and DNAH9 (type 2) in five CCDC39-deficient individuals. The localization of outer dynein arm components was normal in CCDC39-deficient cilia (shown for DNAH5; Fig. 5). We then analyzed inner dynein arm components using antibodies to DNALI1. DNALI1, which is normally found throughout respiratory axonemes, was completely absent from CCDC39-deficient cilia (Fig. 5). The nexin links correspond to the dynein regulatory complex (DRC)¹⁴. Using antibodies to GAS11 (also known as GAS8), the human ortholog of the trypanin or DRC4 subunit of the protist DRC^15–18, we found that GAS11 was localized along the whole axoneme in controls but was confined to the cytoplasm and ciliary base in CCDC39-deficient cilia (Fig. 5). qRT-PCR and protein blot analysis showed that the levels of GAS11 transcripts and protein were unaffected in individuals with PCD (Supplementary Fig. 9).

We investigated the effect of CCDC39 deficiency on ciliary function. Optic microscopy on nasal or bronchial biopsies showed ciliary immotility in 17 of 22 cases with CCDC39 mutations and residual dyskinetic motility in the other 5 cases (DCP414, DCP323, DCP481, DCP552 and OP122). High-speed videomicroscopy analyses of respiratory cells obtained by nasal brushing biopsy (OP122) identified a beating pattern characterized by reduced amplitude with rigid axonemes that showed fast, flickery movements, which suggests defective beat regulation (Supplementary Videos 1 and 2). This pattern differs from that imparted by PCD variants (DNAH5, DNAI1, DNAI2, LRRC50 and C14orf104 (KTU)) that cause defects in the dynein arms and enable beat generation. In these cases, the cilia appear paralyzed but not rigid.

In summary, we have shown that CCDC39 is an axonemal protein whose absence results in failure to correctly assemble DNALI1-containing inner dynein arm complexes, the DRC and the radial spokes, thereby causing axonemal disorganization and dyskinetic beating. Loss-of-function mutations in CCDC40, which encodes another axonemal protein with coiled-coil domains, cause PCD with identical ultrastructural and biochemical defects (including ciliary depletion of CCDC39 and GAS11)¹⁹. Both CCDC39 and CCDC40 may be integral components of the DRC, whose precise composition remains to be defined¹⁴. Mutations in the Chlamydomonas PF2 gene (which encodes the coiled-coil domain-containing DRC4 component) also cause structural defects of both the DRC and inner dynein arms, with failure to assemble DRC components 3–7 (refs. ^20–22). Therefore, as was proposed for PF2, CCDC39 could encode a protein that contributes to the stability of the DRC by interacting with one or several DRC subunits¹⁸. Alternatively, CCDC39 could participate in the transport of the inner dynein arms and DRC. Although the intraflagellar transport mechanism for components located at the interior of the axonemal shaft remains poorly understood, studies in Chlamydomonas have shown that components of the axonemal matrix are required for positioning inner axonemal components but not the outer dynein arms²³. CCDC39, mutations of which affect the internal part of the axoneme with no apparent deleterious effect on outer dynein arms, could therefore be involved in this process. In support of this, several ciliary and centrosomal proteins involved in intraflagellar transport contain SMC-like domains, as does CCDC39 (Fig. 3b). As previously suggested²⁴, SMC-like domains might represent signature sequences that are recognized during these transport processes.

We have also shown that dogs are useful for accelerating the identification of genes in which mutations cause inherited diseases in humans. Diseases that are characterized by locus heterogeneity in humans will almost always involve a single founder mutation in a given dog breed, which facilitates mapping. The highly inbred structure of the domestic dog population, combined with increasingly attentive medical care, has resulted in a long list of breed-specific inherited conditions that advantageously complements phenotype-driven screens in model organisms, including mice. Our results provide useful information for the counseling of previously orphan PCD families and open possibilities for the study of new PCD therapies in the dog model.

URLs. 1000 Genomes Project, http://www.1000genomes.org/.