The effect of Dnaaf5 gene dosage on primary ciliary dyskinesia phenotypes

Motile cilia are highly specialized organelles that transport fluids along the surface of the airway, brain ventricle, and fallopian tube and propel sperm. More than 2,000 genes function in the assembly and structure of motile cilia (1). It is not surprising that pathologic variants in over 50 genes are known to be causative of the human motile ciliopathy called primary ciliary dyskinesia (PCD) (2). Classic features of PCD directly reflect dysfunction of organs bearing motile cilia, including chronic upper and lower respiratory tract infection, bronchiectasis, and reduced fertility. Laterality defects may be present and associated with cardiac malformations, owing to the function of motile cilia present in the embryonic node during early development. Despite essential cellular functions of motile cilia, clinical features of patients with PCD vary. Mild disease without the classic features of PCD is increasingly recognized in a subset of patients (3, 4).

Limited epidemiology suggests broad genotype-phenotype relationships for several common PCD genes. For yet-undetermined reasons, severe disease is associated with variants in some genes (e.g., CCDC39, CCDC40) and mild disease in others (e.g., DNAH9, RSPH1) (3–5). Moreover, little is known about clinical features of patients with different variants in the same PCD gene. Mechanistic explanations for phenotypic differences among patients with the variants in the same gene may lie in modifier genes, microbiome, and environmental factors, among others. Airway inflammation and infection are dominant features of the PCD airway, which may be exacerbated by environmental exposures (e.g., cigarette smoke or allergens) (6). Phenotypic variations are well founded in cystic fibrosis, another genetic cause of chronic airway infection, where disease phenotype can be predicted by the class of mutation in the CFTR gene. Nevertheless, individuals with the most common CFTR mutation, the homozygous F508del variant, have a range of severity of disease, attributed in part to non-CFTR modifier genes (7, 8).

Compared with CF, PCD is less common and more genetically heterogenous, which leads to significant challenges in performing genotype-phenotype studies. With few exceptions, individuals with PCD have 2 pathogenic alleles within the same gene locus as autosomal homozygous recessive (9). Each of the variants may be identical or unique, leading to allele heterogeneity. Predicting the clinical phenotype is particularly difficult in the later condition. Compared with 2 identical variant alleles, if each allele of the same gene harbors a different mutation, the effect of residual function or total absence of the protein may ultimately dictate cilia function. This allele dosage affect has not been tested in PCD.

To test the contribution of gene dosage to genotype-phenotype relationships in PCD, we investigated mutations in the dynein axonemal assembly factor (DNAAF) DNAAF5 (HEATR2), a gene causative of PCD (10). DNAAF5 is a member of the HEAT-repeat–containing family of proteins (Huntingtin, elongation factor 3, PP2A, mTOR) (11, 12) characterized by ~30–40 amino acid–long modules within α helical repeats (13). We previously reported that DNAAF5 belongs to a group of at least 11 cytoplasmic proteins called DNAAFs (10, 14, 15). DNAAFs, including DNAAF5, are expressed only in the cytoplasm, are not found in the cilia, and are responsible for the assembly of the components of the ciliary motors within the cytoplasm, prior to transport to the cilia. Pathologic variants of any of the DNAAFs result in absence of the large motor protein complexes within the ciliary axoneme, called outer and inner dynein arms (ODA and IDA, respectively) (Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.168836DS1), and this absence results in ciliary dysmotility (10, 15–21). Unlike many of the PCD-associated proteins, which have variants throughout the cDNA sequence and evidence of deletion/null variants, DNAAF5 is unique as only missense and deletion variants have been identified and no evidence of homozygous human null variants was reported in ClinVar (22).

We previously described individuals with different DNAAF5 variants, their spectrum of clinical disease, and their abnormal protein products (14). We leveraged our patients’ clinical features and associated variants in DNAAF5 to develop mouse models harboring these mutations. Modeling specific human mutations in mice has been a powerful tool for understanding disease pathobiology while controlling for genetic background and environment (23–25). Models of human PCD variants in mice have not been reported. KO of genes causative of PCD in mice have a characteristic phenotype featuring sinus but not lung disease as well as infertility, and they are dominated by early hydrocephalus and death within about 2 months. To date, hydrocephalus provides the most robust marker of motile cilia dysfunction in mice (26). To investigate DNAAF5 variant phenotypes, we used genome editing to introduce a missense variant identified in patients with PCD in a conserved region of mouse Dnaaf5 as well as a frameshift (FS) deletion resulting in a null allele. We found that different pathogenic mutations in the same gene result in diverse phenotypes, similar to patterns observed in patients (14). Dnaaf5 missense variants demonstrated a gene dosage effect, with different allele numbers and combinations resulting in a range of disease phenotypes. Motile ciliated cells in some tissues were also differentially sensitive to the same mutations in Dnaaf5; sperm flagella were more affected than airway and fallopian tube cilia. Finally, proteomic, and transcriptional analysis suggested a broader effect of variants than previously appreciated.

Spectrum of PCD phenotypes in patients with pathologic variants in DNAAF5.

We have identified several variants in DNAAF5 as causative of PCD in our patient population (Supplemental Figure 1B and Supplemental Table 1). Most of the reported missense or deletion homozygous pathogenic variants fall within the highly conserved HEAT–repeat 10 at the C-terminus of the protein (10, 22, 27). There are reported pathogenic variants in other regions of DNAAF5, and some variants are of unknown significance (Supplemental Table 1). Homozygosity for a DNAAF5-null allele has not been reported and is not present in the patient population cared for in our centers (>15 patients). The Genome Aggregation Database (gnomAD 3.1.2) shows no homozygotes for any predicted loss-of-function variants in DNAAF5 (22). Furthermore, gnomAD Structural Variants (SVs) 2.1 shows no SVs affecting the coding gene itself (28). Furthermore, DECIPHER (a genotype and phenotype database) shows no loss-of-function sequence variants but lists 93 copy number variants (CNVs) and 6 “other” variants (such as uniparental disomy), all of which are present as heterozygous (29). In addition to a reported case (30), we identified a patient with compound heterozygosity of a null and missense alleles (Supplemental Table 1). These genomic data strongly support that DNAAF5 is an essential human gene, and patients with PCD must have some functional, albeit significantly reduced, DNAAF5.

Indeed, our patient population with DNAAF5 variants (>15) have a range of clinical features, including typical symptoms of PCD: chronic cough, nasal symptoms, and recurrent otitis media. Organ laterality changes (including situs inversus and congenital heart disease) are present in ~50% of patients, as predicted. When measured, nasal nitric oxide levels are uniformly low, consistent with PCD (31). Transmission electron microscopy (TEM) of DNAAF5 variants typically shows absence or truncation of the ODAs as well as absence of the IDA, thought to be due to a failure of the motor complex to assemble in the cytoplasm and to then move to the axoneme microtubules (14, 15, 19). Patients live in diverse environments, including urban settings and a cohort in a Mennonite farming community. These exposures could influence phenotypes (32).

Intriguingly, we identified a family with a DNAAF5 homozygous missense variant (c.14999G>T), resulting in the substitution of cysteine by phenylalanine at position 500 (C500F) (14). Affected siblings show typical diagnostic features of PCD including situs abnormalities, low nasal nitric oxide, and reduced numbers of IDA and ODA determined by TEM. However, compared with other patients with DNAAF5 variants, these children have less severe disease, and normal cilia beat frequency (CBF) was retained in about a quarter of the cells, as quantified in cultured nasal epithelium (14). Furthermore, while the numbers of IDA and ODA were reduced, the extent of reduction was less pronounced than in patients with other DNAAF5 variants (Figure 1, A and B). Given the lack of reports of individuals with biallelic DNAAF5-null alleles, as well as the existence of patients with clinically milder PCD features, we asked whether DNAAF5 variants represented an allelic series or whether the clinical variation could be attributed to environment or the presence of modifier genes.

Figure 1

The effect of NULL and MIS allele dose on survival. (A) Nasal epithelial cells were obtained from patients with DNAAF5 variants for transmission electron microscopy (TEM) to assess DNAAF5 function by the ultrastructure of the ciliary axoneme. Cross section of cilia show normal, truncated, or absent outer dynein arm (ODA) and inner dynein arm (IDA) (cilia motor complexes) on microtubule doublets. Arrow and arrowhead indicate the ODA and IDA, respectively. Scale bar: 100 nm. (B) Quantification of ODA and IDA identified in TEM cross sections of cilia from individuals with indicated variants in DNAAF5 (n = 1–2 individuals per genotype). (C) Intercross of WT/NULL mice produce no NULL/NULL offspring (n = 31 litters, 173 mice were analyzed). Data are shown as mean ± SEM. Percentage of genotype/litter: 36.9% ± 19.3%, 63.2% ± 19.3%, and 0% in WT/WT, WT/NULL, and NULL/NULL, respectively. (D) Intercross of male WT/MIS and female WT/MIS mice (n = 24 litters). Data are shown as mean ± SEM. Percentage of genotype/litter: 22.9% ± 4.2%, 54.2% ± 5.0%, and 22.9% ± 3.5% in WT/WT, WT/MIS, and MIS/MIS, respectively. (E) Intercross of WT/MIS males with MIS/MIS females (n = 35 litters). Data are shown as mean ± SEM. Percentage of genotype/litter: 57.2% ± 5.0% WT/MIS, 42.8% ± 5.0%, MIS/MIS. Fewer MIS/MIS offspring were produced than predicted by Mendelian genetics (χ², P = 0.04). (F) Kaplan survival curve for mice with indicated genotypes. (G) Breeding WT/NULL males with MIS/MIS females produce fewer than predicted MIS/NULL offspring (n = 13 litters). Data are shown as mean ± SEM. Percentage of genotype/litter: 81.0% ± 7.0%, WT/MIS; 19.0% ± 7.0%, MIS/NULL; χ², P = 0.0009. Mean ± SEM are shown in B–E and G; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 determined using Kruskal-Wallis test with Dunn’s multiple comparisons. Number of animals in each group is shown in parenthesis in C–E and G.

Different mutations in Dnaaf5 result in distinct phenotypes.

To investigate the phenotype-genotype relationships in an identical genetic background and environment, we used a CRISPR-Cas9 system to introduce the human c.14999G>T in the conserved Dnaaf5 mouse genome at position c.1493, resulting in a G>T and the equivalent p.C498F missense mutation (Supplemental Figure 2A). A synonymous mutation (c.1473C>T) was also introduced upstream to the missense mutation to prevent genomic recutting by Cas9 and improve the probability of homology-directed repair (Supplemental Figure 2A) (33).

Surrogate mothers implanted with edited blastocysts, all in the C57BL/6 strain, provided chimeric pups. Fifty-two founder pups were born, and genotypes were determined. Fifteen were WT with no disease phenotype, 1 of the pups showed a heterozygous missense mutation without a difference from the WT phenotype at birth (c.1493G>T), and 26 had indels resulting in multiple FS mutations. Mice with indels had features typical of PCD gene KO mice, especially hydrocephalus and runted growth (Supplemental Figure 2, B and C). Surviving mice with indels carried 1 of 3 different FS, each in trans with a WT allele (Supplemental Figure 2D). Germline transmission of the FS and missense C498F Dnaaf5 alleles (denoted MIS) were established after breeding with WT mice. The heterozygous F1 pups (Dnaaf5^WT/FS and Dnaaf5^WT/C498F) had normal postnatal growth and behavior. Germline transmission was confirmed in all 3 FS lines by breeding. We used the line with a c.1476delGGAGCAT FS* (p.Asp493fs*26) mutation for subsequent phenotypic analyses of the FS mutation. The FS mutation is predicted to result in RNA decay and a null allele. The 2 alleles, FS and missense, provided an opportunity to study phenotypes of each allele and the gene dosage effects in recessive genetic models (Table 1).

Table 1

Effect of Dnaaf5 variant gene dosage on cilia phenotype

An initial intercross of heterozygous Dnaaf5^WT/FS mice (denoted WT/NULL) yielded 31 litters and 173 animals (Figure 1C). However, no animals homozygous for the FS mutation (Dnaaf5^FS/FS; denoted NULL/NULL) were present at weaning. Lack of homozygous FS offspring was further confirmed by interbreeding the other 2 FS mutation lines. These observations suggest that loss of Dnaaf5 results in embryonic lethality. To determine when Dnaaf5 mutants die, independent editing of the critical mouse exon to generate a null allele was performed (Supplemental Figure 3A). In this line, Mendelian ratios at blastocyst stage were normal (Supplemental Figure 3B). However, by E6.5, homozygous mutant embryos had growth arrest and reduced numbers, and by E8.5, no surviving mutants were observed (Supplemental Figure 3, B and C). This early embryonic lethality, at a window before cilia function is required, suggests a nonciliary requirement for DNAAF5. The findings align with observations in Drosophila of adult lethality of CG31320/DNAAF5 mutant flies, where motile cilia are restricted to the sperm and chordotonal neuron (CH) in the fly and where motile cilia defects are not lethal (27, 34). The exact role of the DNAAF5 protein in nonciliary functions is not clear but is supported by the broad DNAAF5 expression beyond motile ciliated tissues (35). Consistent with data in gnomAD and ClinVar that no DNAAF5 homozygous null variants are reported, our observed early embryonic lethality in multiple null alleles suggests that DNAAF5 is an essential Metazoan gene.

Next, to assess the genetic fitness of the other Dnaaf5 mutations, we interbred Dnaaf5^WT/C498F (WT/MIS) mice; the live offspring included all predicted genotypes (Figure 1D). The genotype distribution was near Mendelian — with percentages of 23:54:23 — for WT, heterozygotes, and homozygotes, respectively. Breeding male WT/MIS with female MIS/MIS resulted in live MIS/MIS offspring but fewer than predicted, which may be related to in utero loss of MIS/MIS animals or subfertility of WT/MIS females (Figure 1E).

MIS/MIS animals showed gross phenotypes associated with motile cilia deficiency phenotypes commonly observed in mice, including development of hydrocephalus and situs abnormalities (Supplemental Figure 3, D and E). Most animals with 2 MIS alleles (MIS/MIS) survived beyond 150 days of life (a subset of animals was followed for up to 1 year) (Figure 1F). These findings suggest that homozygous missense mutants retain sufficient function to survive, in contrast to the early embryonic lethality of null alleles.

To study how dosage of a single MIS allele affects survival, we bred MIS/MIS female animals to WT/NULL male animals. Breeding resulted in viable litters (Figure 1G). However, only 22% of offspring were MIS/NULL (predicted 50%), suggesting in utero mortality of most MIS/NULL offspring. Compound mutant MIS/NULL mice that were born alive were runted and displayed early-onset gross macrocephaly (Supplemental Figure 3, F and G). Early lethality was observed, with 50% of these compound mutant mice dying by 30 days and none surviving beyond 90 days (Figure 1F).

Taken together, our results indicate that Dnaaf5 is essential for development and sensitive to gene dosage. While complete loss of function of Dnaaf5 (NULL/NULL) is embryonic lethal and homozygous missense C498F (MIS/MIS) survive postnatally, compound mutants (MIS/NULL) exhibit increased perinatal lethality, suggesting that we have generated mice representing an allelic series of Dnaaf5 function in vivo.

Combinations of Dnaaf5 alleles determine cilia motility and transport.

Both our patient phenotype and our MIS allele analysis suggest that missense mutant protein may be partially functional, with 2 MIS alleles providing an additive effect. To test the effect of allelic dosage on cilia phenotype, we evaluated the cilia function by measuring CBF and cilia transport of beads in the trachea of mice ex vivo (Figure 2A). In parallel, we also cultured primary tracheal epithelial cells at air-liquid interface (ALI) to exclude environmental factors (e.g., infection) (36).

Figure 2

The effect of NULL and MIS allele dose on airway cilia function and ultrastructure. (A) Schematic of primary culture airway cells and ex vivo isolated trachea mounts used for analysis of cilia function by cilia beat frequency (CBF), transport, and ultrastructure by transmission electron microscopy (TEM). (B) CBF of cultured cells from WT/WT and WT/NULL littermates, with mean ± SEM shown as 17.47 ± 0.80 Hz and 16.0 ± 0.47 Hz, respectively. (C) CBF of cultured cells from WT/WT, WT/MIS, MIS/MIS, and MIS/NULL littermates, with mean ± SEM frequency of 16.22 ± 0.29 Hz, 16.39 ± 0.33 Hz, 7.25 ± 0.53 Hz, and 0.06 ± 0.06 Hz, respectively. CBF was assessed in 5 random areas per Transwell; 5 Transwell cultures were performed for each genotype per experiment. n = 2–4 independent experiments comprising unique mice. (D and E) Mucociliary transport speed and bead displacement on the surface of isolated trachea from WT/WT, WT/MIS, and MIS/MIS mice, with mean ± SEM speeds shown as 39.04 ± 5.02 Hz, 39.29 ± 2.86 Hz, and 7.16 ± 0.45 m/sec, respectively. Mean bead displacement over 5 seconds is shown as 89.13 ± 4.42 μm, 86.82 ± 4.80 μm, 26.00 ± 1.36 μm, respectively (n = 8–10 animals per genotype). (F) Representative TEM images of airway cilia from WT/WT, MIS/MIS, and MIS/NULL mice. Arrows indicate presence (magenta) or absence (green) of outer and inner dynein arms (ODA, IDA). Scale bar: 100 nm. (G) Quantitation of outer dynein motor protein complexes detected by TEM of WT/WT, MIS/MIS, and MIS/NULL cilia cross section, with mean ± SEM shown number of ODA 6.0 ± 0.4, 3.8 ± 0.2, and 0.8 ± 0.4, respectively. (H) Quantification of inner dynein arms in ciliary axonemes from cultured tracheal epithelial cells from WT/WT, MIS/MIS, and MIS/NULL, with mean ± SEM shown number of IDA 3.8 ± 04, 0.3 ± 0.0, and 0.3 ± 0.2, respectively. n = 4 animals per genotype, n = 5–10 cross sections per genotype. *P < 0.05, **P < 0.01, ***P < 0.001 determined using Kruskal-Wallis test with Dunn’s multiple-comparison test.

We first compared the CBF of WT/WT with WT/NULL littermates, finding no difference in the CBF between these genotypes (Figure 2B). Next, we compared the effect of MIS dosage on CBF. The effect of a single MIS allele when in trans with a WT allele (WT/MIS) was not statistically different from WT/WT (Figure 2C). Similarly, bead transport across the surface of the trachea was not different between mice of these genotypes (Figure 2, D and E). In contrast, 2 MIS alleles (MIS/MIS) resulted in variable but distinctly lower mean CBF and decreased tracheal bead transport when compared with WT/MIS and WT/WT control littermates (Figure 2, C–E). The heterogeneity of cilia motility among cells in the MIS/MIS samples was similar to that observed in our patients with the DNAAF5 c.14999G>T (C500F/C500F) missense variant (14); most cilia from MIS/MIS animals were completely immotile, while others retained markedly reduced motility (Supplemental Videos 1–3).

Finally, we tested a single copy of the MIS mutation in trans with the NULL allele (MIS/NULL). MIS/NULL cultured cells showed a more pronounced cilia motility defect compared with MIS/MIS, as all cilia were immotile (Figure 2C). Together, these results suggest that a single copy of C498F results in a Dnaaf5 protein with either reduced levels or functions that are required to adequately equip motile cilia, likely due to inadequate dynein motor complex assembly.

Dnaaf5 mutations determine the degree of cilia ultrastructure defects.

Dnaaf5 contributes to the cytoplasmic assembly of the axonemal IDA and ODA motor complexes. The absence of ODA and IDA in the cilia indicates insufficient production of the motor complexes. To investigate the relationship between the CBF phenotype and genotypes, we examined the cilia ultrastructure using TEM.

Cilia were scored for normal or defective (absent, truncated) IDA and ODA, compared with WT/WT mice. We did not appreciate structural defects in cross sections of WT/NULL cilia compared with WT/WT. Consistent with our CBF results, normal axonemal ultrastructure suggests that defects in cilia motility do not arise from haploinsufficiency — i.e., 1 WT allele is sufficient for normal dynein motor assembly. Next, we evaluated the effects of levels of the MIS mutation on airway cilia ultrastructure. We observed a dose-dependent reduction in the mean number of ODAs and IDAs with successive loss of the WT allele (WT/WT to WT/MIS to MIS/MIS) (Figure 2, F–H). IDA were significantly reduced in cross sections of cilia from MIS/MIS. To assess MIS dosage, the ultrastructure of cilia from mice with a single MIS allele was assessed in isolation (MIS/NULL). Consistent with the reduced CBF, ODAs and IDAs were also mostly absent in the setting of a single-mutant allele (Figure 2, F and G).

As assessed by TEM, there was striking variability in the presence and morphology of ODA and IDA, among different doublet microtubules in the same cilia cross section of mice with a mutant allele (MIS/MIS or MIS/NULL). The ODA appeared normal on some microtubules while absent or truncated on other microtubules within the same cross section (Figure 2F). This variability was not unique to a specific microtubule of the 9 different A tubules (diagram, Supplemental Figure 1A) that dock the dynein motor complexes. Variable ODA morphology was similarly present in patients with the C500F mutation (Figure 1A). This observation may suggest that delivery of the dynein complex is not uniform along the length of the ciliary axonemal and that partially formed complexes may be transported into the same ciliary axoneme alongside fully assembled dynein complexes.

Tissue specificity of genotype-phenotype relationships in the airway.

Having observed differences in cilia ultrastructure in the airway, we examined the tissue specificity of phenotypes in mice with different Dnaaf5 genotypes. Motile cilia dysfunction is associated with upper airway sinus disease in patients and mice with KO of PCD gene (37–41). Unlike humans, mice deficient in PCD genes are not reported to develop lung disease. We did not observe lung inflammation in our mice of any genotype. We observed sinus inflammation in the MIS/MIS and MIS/NULL mice, compared with WT/MIS mice and WT/WT littermates (Supplemental Figure 4A). We took advantage of the long survival of MIS/MIS mice to test lung clearance using intratracheal Pseudomonas aeruginosa. There was a trend toward decreased clearance in the MIS/MIS mice compared with WT/MIS and WT/WT animals (Supplemental Figure 4, B and C). We were unable to perform these experiments in MIS/NULL animals due to limited survival beyond weaning age.

Figure 4

The effect of MIS allele dose on sperm and fallopian tube function and structure. (A) Sperm with motility was low in MIS/MIS (1.35% ± 1.0%) compared with WT/WT (44.46% ± 2.17%) and WT/MIS (33.04% ± 4.10%) littermates (n = 5–6 mice/genotype, P < 0.01). (B) Sperm showing forward progressive motility was significantly lower in MIS/MIS sperm (0.56% ± 0.5%) compared with WT/WT (22.90% ± 2.11%) and WT/MIS (20.12% ± 3.27%) littermates (n = 5–6 mice per genotype. P < 0.01). (C) Reduced dynein arms in MIS/MIS sperm compared with WT/WT littermates. Representative TEM images of cross sections from sperm flagella. Arrows and arrowhead indicate the presence (magenta) and absence (green) of ODA and IDA. Scale bar: 100 nm. (D and E) Reduction of motor complex ODA and IDA in sperm flagella in MIS/MIS compared with WT/WT littermates (n = 4 mice per genotype). (F) Reduced CBF of cilia from ex vivo, isolated fallopian tube from WT/WT, WT/MIS, and MIS/MIS littermates, with mean ± SEM shown as 5.65 ± 0.31 Hz, 8.85 ± 0.39 Hz, and 9.34 ± 0.28 Hz, respectively (n = 2–3 mice per genotype). *P < 0.05, **P < 0.01, ****P < 0.0001 determined using Kruskal-Wallis test with Dunn’s multiple-comparison test.

Tissue specificity of genotype-phenotype relationships in the ependyma.

Motile cilia are considered essential for creation of CSF flow networks (42), and defects in cilia-related genes are associated with enlarged brain ventricles — i.e., hydrocephalus. Mice deficient in genes coding for dynein assembly factors or dynein arm complex proteins (e.g., Dnah5) develop severe postnatal macrocephaly due to hydrocephalus, which is fatal within approximately a month (17, 41, 43, 44). We did not observe gross hydrocephalus in WT/WT, WT/MIS, or WT/NULL mice. In contrast, about half of MIS/MIS and all MIS/NULL animals developed at least mild hydrocephalus by 12 weeks of age, ultimately leading to death. However, in animals surviving longer (1–10 months), histologic and brain magnetic resonance imaging (MRI) showed varying degrees of ventriculomegaly in all MIS/MIS animals, independent of the presence of gross macrocephaly or sex (Figure 3, A and B). The earliest signs of macrocephaly were observed at 3 weeks of age (mean = 5.0 ± 3.0 weeks, n = 22). Histologic evaluation of 3-day-old pups did not show significant differences in the lateral ventricle area between WT/MIS and MIS/MIS animals (Supplemental Figure 5A). MRI of similarly aged pups showed a trend of increased lateral ventricle volume in MIS/MIS animals that did not reach significance (Supplemental Figure 5, B and C). By MRI, the lateral and third ventricle in MIS/MIS animals older than 3 weeks of age were significantly larger than in WT/MIS and WT/WT animals (Figure 3C and Supplemental Figure 5D), while there was no difference in fourth ventricle volumes (Supplemental Figure 5, E and F). Histologic evaluation did not show obstruction within the brain ventricle system or aqueduct (Supplemental Figure 5, G and H), suggesting that MIS/MIS and MIS/NULL animals develop nonobstructive, communicating hydrocephalus due to insufficient DNAAF5 function leading to insufficient motor complexes in the cilia.

Figure 3

The effect of MIS allele dose on hydrocephalus and ependymal cilia ultrastructure. (A) Coronal brain cross sections show hydrocephalus in MIS/MIS and MIS/NULL animals compared with other genotypes. Note variable degrees of hydrocephalus in MIS/MIS mice. Arrowhead indicates lateral ventricles. (B) MRI showing ventricular enlargement in MIS/MIS compared with WT/MIS and WT/WT animals. Arrow indicates lateral ventricles; arrowhead indicates third ventricle. (C) Quantification of lateral ventricle volumes from MRI images in B showing significant ventricular enlargement of MIS/MIS compared with WT/MIS and WT/WT mice (P < 0.01, n = 6 animals per genotype). (D) Bead velocity on isolated ventricles from WT/WT, WT/MIS, and MIS/MIS mice, with mean ± SEM speeds shown as 57.57 ± 1.2 μm/sec, 55.02 ± 0.78 μm/sec, and 35.93 ± 0.70 μm/sec, respectively; data were significantly different between WT/WT and MIS/MIS animals (P < 0.0001, n = 2 animals per genotype, measured on 2 lateral ventricles per animal). (E) Representative composite images of fluorescent bead velocity vectors showing transport on ex vivo lateral ventricles. (F) Representative TEM images of cross sections from ependymal cilia of MIS/MIS and WT/WT littermates showing reduced dynein motor protein complexes in ependyma ciliary axonemes of MIS/MIS mice. Arrows and arrowhead indicate the presence (magenta) and absence (green) of ODA and IDA. Scale bar: 100 nm. (G and H) Quantification of ODA and IDA in ependymal cilia cross sections from different genotypes (n = 4 animals per group). **P < 0.01, ****P < 0.0001, determined using Kruskal-Wallis test with Dunn’s multiple-comparison test.

In addition to brain abnormalities, MIS/MIS mice with hydrocephalus had cervical spine lordosis with thoraco-lumbar kyphosis, detected by MRI (Supplemental Figure 5I). Even MIS/MIS mice with mild ventricular enlargement had the same spinal abnormality observed in mice with severe ventricle enlargement. No skeletal deformity was noted, suggesting that the spine deformity may be secondary to neuromuscular changes. This spine phenomenon was observed in other mice that are null for motile cilia genes, including Foxj1^–/– animals (43, 45). In zebrafish, spinal curvature is often observed in motile ciliopathy and is linked to dysfunction of the subcommissural organ and Reissner’s fiber within the ventricular system (46, 47).

To examine the effect of Dnaaf5 mutant alleles in the cilia of the ependyma compared with the airway, we performed transport assays on the surface of extracted lateral ventricles of WT/WT, WT/MIS, and MIS/MIS animals (Figure 3D). MIS/NULL mice were not evaluated, since severe hydrocephalus limited extraction of the brain with intact ventricles. Cerebral spinal flow is directed by complex patterns of cilia-mediated transport (42). The rate and distance of movement of beads on the surface of ependymal cilia on the lateral ventricles of WT/WT and WT/MIS mice were similar; in contrast, MIS/MIS animals had comparatively slower bead velocity and shorter flow distance (Figure 3, D and E, and Supplemental Videos 4 and 5). The loss of ependyma ciliary axoneme IDA and ODA in MIS/MIS compared WT/WT were similar to those observed in airway cilia (Figure 3, F–H).

Taken together, these observations suggest a dose-dependent effect of the MIS allele on development of hydrocephalus. In the presence of a single C498F allele (MIS/NULL), hydrocephalus occurred earlier and was more severe than in MIS/MIS animals.

Tissue specificity of genotype-phenotype relationships associated with fertility.

Fertility is related to motile cilia in the fallopian tube and the function of the sperm flagella. We first tested the effect of gene dosage of the MIS allele in female mice by breeding Dnaaf5 WT/MIS or MIS/MIS females with WT/WT males, and this resulted in live offspring indicate that MIS/MIS female mice are fertile. To assess sperm function, WT/MIS or MIS/MIS males were bred with WT/WT females. Only WT/MIS males produced live offspring, indicating that 2 missense alleles in Dnaaf5 cause male infertility. Limited survival prevented testing the fertility of MIS/NULL mice of either sex. These findings suggest that the sperm are more sensitive to mutations in Dnaaf5 than the fallopian tube cilia.

To assess the effects of these mutations on sperm and fallopian tube cilia function, we evaluated the cilia motility in the relevant tissues. Unlike in the airway and ependyma cilia, where decreased motility was observed in MIS/MIS cilia, sperm from MIS/MIS animals were all immotile and could not be recovered by capacitation or the addition of 8-Bromo-cAMP, despite similar sperm viability between genotypes (Figure 4, A and B; Supplemental Figure 6A; and Supplemental Videos 6–10). There were no significant differences in sperm motility between WT/WT and WT/MIS mice.

Consistent with loss of function, ultrastructural evaluation of flagella of MIS/MIS sperm showed complete absence of ODA and IDA (Figure 4, C–E). Interestingly, approximately half of the sperm cross sections from MIS/MIS animals also showed abnormal numbers of doublet microtubules and absence of the central apparatus (Supplemental Figure 6, B and C). This ultrastructural abnormality was not observed in airway cilia from either mice or patients with mutations in Dnaaf5 and may be specific to sperm.

We also assessed fallopian tube cilia function. Unlike male Dnaaf5 MIS/MIS animals, female animals with MIS/MIS mutations were fertile. Indeed, analysis of CBF of dissected fallopian tubes showed motile cilia, with a frequency range similar to airway cilia (Figure 4F).

In summary, differences in cilia function and accompanying ultrastructural changes in sperm compared with airway suggest differences in the tolerance of mutations in dynein motor assembly proteins in different tissues.

Ciliary axoneme proteomics of Dnaaf5 mutant mice.

How Dnaaf5 variants cause different phenotypes could be related to the quantities or types of motor proteins missing in the axoneme of mutants. Cilia from patients with PCD variants in genes coding for DNAAFs have absent ODA and IDA ciliary dynein arm proteins — including DNAH5, DNAI1, DNALI1, and DNAH7 — in ciliary axonemes; however an unbiased analysis of proteins in the variant ciliary axoneme has not been performed (10, 27). We isolated cilia from cultured tracheal epithelial cells from mice with the MIS/NULL genotype and compared proteomics to WT littermates by tandem mass tag (TMT) mass spectrometry (Figure 5, A and B, and Supplemental Table 2).

Figure 5

Molecular phenotype of Dnaaf5 mutant multiciliated cells. (A) Schematic of strategy used for analysis of cultured airway cells from WT and mutant mice. Isolated motile cilia were analyzed by mass spectrometry and total cells by scRNA-Seq. (B) Proteomics of ciliary axonemes isolated from primary culture tracheal epithelial cells from WT/WT and MIS/NULL littermates. Heatmap comparing the expression level of cilia-associated proteins (n = 4 animals per genotype). (C) Reduction in DNALI1 in cilia of primary human airway cells from a normal individual and a patient with PCD due to DNAAF5 variant p.Leu795Pro by immunofluorescence analysis. Representative images shown. Scale bar: 100 μm. (D) Uniform Manifold Approximation and Projection (UMAP) comparing airway epithelial cell clusters of WT/WT and MIS/MIS littermates. Foxj1 expression is shown for reference (n = 3 animals per genotype). (E) Feature plot showing differential expression of selected genes in WT/WT and MIS/MIS littermates identified using scRNA-Seq analysis.

Compared with WT/WT, MISS/NULL axonemal cilia showed significant reduction in ODA proteins, DNAH11, DNAH5, and DNAI2, and IDA protein, DNALI1 (Figure 5B). Interestingly, we also observed reduction in proteins associated with the ODA docking complex, including CCCDC151 and ARMC4, suggesting that the docking proteins are interdependent on motor protein assembly, transport, or retention in the axoneme. Proteins in the cilia with less known function were also identified as reduced, including CFAP126 (Flattop), a microtubule inner protein, and NME9, a thioredoxin domain-containing protein (Figure 5B and Supplemental Table 2). Analysis of primary airway cells from our patients with a pathogenic variant in DNAAF5 showed decreased expression of some of these candidate proteins in the ciliary axonemes (Figure 5C and Supplemental Figure 7A). These missing proteins were not previously associated with Dnaaf5 mutations, and this may suggest a wider disruption of the integrity of dynein arm complexes or other roles for DNAAF5 (and other DNAAF proteins) in ciliary motor assembly and cilia assembly.

Dnaaf5 mutations are associated with increased cilia-related transcript abundance.

The effect of mutations in cilia assembly genes on the airway multiciliated cell transcriptome is not defined. We hypothesized that mutations in Dnaaf5 result in a compensatory response to decreased motor proteins in the ciliary axoneme. We used single-cell RNA-Seq (scRNA-Seq) to transcriptionally profile cultured tracheal epithelial cells from Dnaaf5 MIS/MIS and WT littermates (n = 3 unique mice for each genotype) (Figure 5D). We identified all known major cell epithelial types, including basal progenitor cells, secretory cells, multiciliated cells, ionocytes, and neuroendocrine cells (Supplemental Figure 7, B and C). MIS/MIS mutations did not affect the distribution of cell types or differentiation of culture primary cells compared with WT/WT, though we did observe a shift in the abundance of subclusters of basal cells between groups (Figure 5D). Gene Ontology (GO) analysis showed increased expression of cilia assembly as a group (GO: 0060271), with MIS/MIS cells showing increased differential expression of motile cilia–related genes (e.g., transcripts of outer dynein arm genes DNAH5, radial spoke gene RSPH9, and organization genes CCDC39; Figure 5E, Supplemental Figure 7D, and Supplemental Table 3). Downregulated transcripts in motile ciliated cells were related to glutathione metabolism and prostaglandin metabolism (Supplemental Table 4). To determine if the same molecular phenotype was present in patients with DNAAF5 variants, we examined bulk RNA from primary culture nasal epithelial cells obtained from a patient with a compound variant (DNAAF5; c.2353-2354del, p.S785 fs). The upregulated genes were similar to those observed in Dnaaf5 mutant mouse cells (Supplemental Figure 7E), suggesting a conserved feedback response in human disease, broadly affecting ciliogenesis.

Patients with PCD have pathologic allele variants leading to clinical features affecting specific organ systems: the respiratory tract, reproduction, cerebral spinal fluid circulation, and development of organ laterality. However, the disease of PCD is not invariable; instead, patients have a constellation of signs and symptoms that differ among individuals, across ages, and, we propose, by gene variant and mutation type. We also recognize that environmental influences, infection type, and variant suppressor gene may influence phenotypes. To address these issues, we investigated PCD genotype-phenotype relationships in a fixed genetic background and environment by modeling human PCD variants of Dnaaf5 in mice. We drew upon our observations of patients with DNAAF5 variants, experience with DNAAF5 function in ciliogenesis, and familiarity with the unique features of motile cilia dysfunction in mice (10, 14, 45). We were particularly interested in a human DNAAF5 variant associated with mild disease that we previously showed produced detectable protein, determined by immunoblot (14). Generating a clinically conserved allele to model a human missense variant and a null mutation in mice by CRISPR-Cas9 genome editing allowed formal testing of allele fitness and the impact of each allele alone and in combination with a WT allele. Using this approach, we identified a distinct gene dosage effect of MIS and NULL alleles (Table 1). In mice, gene dosage was associated with clinical phenotypes of motile cilia-dependent organs and quantifiable differences in cilia motor function and cilia ultrastructure, emphasizing the importance of knowledge of the specific mutation, the function of the gene, and, if relevant, the allele combinations.

Survival served as a quantifiable measure of allelic fitness. Mice with PCD gene KO uniformly develop early hydrocephalus and death by 2 months of age (or earlier) (39–41). The onset of hydrocephalus occurred in mice by 3 weeks of age, and newborn animals did not show significant differences in lateral ventricle volumes, though there was a trend of increased volumes in MIS/MIS animals. None of the patients with PCD with variants in DNAAF5 in our cohort or patients with variants in DNAAF5 described by others showed clinical evidence of hydrocephalus, consistent with most human gene variants causative of PCD. However, brain imaging was not performed to rule out clinically insignificant ventriculomegaly. While rare in humans with PCD, development of hydrocephalus is a well-established phenotype in mice, which we confirm is consistent with defective function and ultrastructure of cilia on the ventricular ependyma (48). The development of ventriculomegaly followed by an early death is similarly observed in KO of most motile cilia genes in the hydrocephalus-prone C57BL/6 strain, limiting the ability to identify differences in phenotypes (26, 49). In contrast to the KO approach, we opted to introduce a patient specific mutation (DNAAF5^C500F/C500F, MIS/MIS) and a null mutation to study the interaction of different variants (14). We first assessed allelic fitness of the NULL and MIS alleles by their survival. We demonstrate that, in the heterozygous condition, in a WT background, neither allele (WT/MIS or WT/NULL) significantly impacted the survival phenotype. The null allele (NULL/NULL) in the homozygous condition, however, resulted in early embryonic lethality, which, to our knowledge, has not been previously noted in PCD gene KO models. The homozygous missense mutation (MIS/MIS) resulted in decreased survival, while — in trans with a single null allele — the MIS/NULL genotype allowed live birth but death within 1 month. Thus, when not paired with the WT allele, each mutant allele alone was lethal but to a different degree. These results provide evidence that phenotypes should be interpreted relative to the specific variant, and this method of interpretation may impact our analysis of genotype-phenotype relationships in human disease.

The cause of embryonic lethality observed in animals with NULL/NULL mutations in Dnaaf5 is unclear, but it arises by E6.5, in a window prior to cilia function, earlier than motile cilia function in nodal cilia at E7.5 (50). The results point to a cilia-independent chaperoning function for DNAAF5 during embryogenesis. As noted, the orthologue of Dnaaf5 is also lethal in early Drosophila development (27). The early intrauterine death of mice suggests that DNAAF5 variants in humans have some partial function, at least pertaining to the survival phenotype. Indeed, we were able to detect reduced levels of DNAAF5 by immunoblot in cultured nasal epithelial cells from patients with DNAAF5 variants (14). Second, the majority of known DNAAF5 human deletion variants are within the 3′-terminus (Supplemental Figure 1 and Supplemental Table 1), likely retaining a sufficient, partially functioning protein. Third, homozygous null mutations of DNAAF5 are not found in ClinVar or gnomAD, indicating that pressures of allelic fitness are at work. Additional support for survival of variants with partial function comes from Chlamydomonas, where truncated abnormal appearing dynein complex are partially functional, despite the loss of the outer dynein arm motor protein heavy chain β (oda4) (51).

Fertility was used as a second determinant of allelic fitness. Most males with PCD are infertile, though subfertility occurs and is most often the case in females (52). As expected, male mice with a single copy of the WT allele were fertile, consistent with the carrier state in autosomal recessive genetics. Notable was the loss of fertility in MIS/MIS male mice and lack of sperm flagella motility, despite having less frequent hydrocephalus and longer survival than the MIS/NULL mice. Observing completely immotile MIS/MIS sperm contrasted with decreased retained motility of some cilia in the airway, ependyma, and fallopian tube, likely due to a different motor protein composition of the motor complex in different tissues. The components of the outer dynein motor complex within the axoneme of the airway, ependyma, and fallopian tube — including dynein heavy chains DNAH5, DNAH11, and DNAH9 — are similar. Sperm dynein heavy chain, however, differs, using DNAH8 and DNAH17 instead (53). Different cilia gene requirements for flagellar assembly than airway assembly may account for infertility without PCD (54). TEM examination of sperm tail ultrastructure of MIS/MIS mice showed significantly diminished outer dynein arms compared with the airway and ependymal cilia. Furthermore, sperm from MIS/MIS animals showed a defect manifest as an additional central apparatus microtubule that was not observed in airway or brain ciliary axoneme. These findings suggest that assembly of the axonemal dynein motors require different preassembly machinery that could not be served by the haploinsufficient MIS/MIS coded protein and that the sperm ODA motor assembly is more dependent on a fully functional DNAAF5 than other motile cilia.

Ultrastructural analysis of Dnaaf5 mutants also informed us of activities of dynein assembly factors and of motor protein transport to the axoneme. Dnaaf5 is expressed only in the cytoplasm and, in concert with other cytoplasmic factors (e.g., DNAAF2, SPAG1, DNAAF4, LRRC6), participates in the assembly of cilia motor proteins. Observation in Xenopus suggests that IDA and ODA are assembled in unique pools (15, 55), and we saw a tendency of loss of IDAs over ODAs, indicating that Dnaaf5 is perhaps more essential for IDA assembly. Analysis of the MIS/MIS cilia ultrastructure also led to several unexpected observations. First, although the exact composition of the “truncated” outer arm motor structures in DNAAF5 mutant cilia are unknown, they provide sufficient cilia motility only in the MIS/MIS and not MIS/NULL genotypes. We cannot assess the longitudinal numbers of ODA to determine if the abundance is related to the MIS allele dosage. Second, in the MIS/MIS cilia, the presence of normal-appearing outer dynein arms alongside absent or truncated outer dynein arms within the same cilia cross section is not microtubule pair specific (i.e., occurring on the same microtubule numbers 1–9), suggesting that the dynein assembly and transport machinery is both indiscriminate and may transport a partially assembled complex. Finally, assembly and transport of the dynein arm complexes may be disengaged from the microtubule assembly process (56) such that, in normal conditions, both processes occur in parallel; however, when one assembly factor stalls (i.e., Dnaaf5), the others continue to function, resulting in patchy assembly across the axoneme.

The observed ultrastructural changes in Dnaaf5 mutant cilia were supported by proteomics of cilia from MIS/NULL mice, showing decreases in components of outer and inner dynein motor complex proteins as well as docking proteins CCDC151 and ARMC4. Docking proteins are not known to be part of the cytoplasmic motor assembly process, though they may attach to the outer dynein arms during transport into the cilia (15). Other diminished proteins are those with lesser-known function, including FAP126 (57), an inner junction associated protein, and NME9, a thioredoxin-like protein predicted to be associated with motile cilia; the latter we previously showed interacts with DNAI2 during assembly and is present in the ciliary axoneme (55). The latter was missing in human cilia isolated from a patient with a DNAAF5 variant, supporting the proteomic findings in Dnaaf5 mutant cilia. These findings suggest a more complex role for DNAAF5 during ciliary axoneme assembly than previously understood.

There are several limitations of our findings. First, the gene dosage effect identified may be unique to Dnaaf5, and additional studies are required to identify similar patterns of genotype-phenotype relationships with other cilia-associated genes. Second, the function of DNAAF5 is unknown, and how the mutant proteins may interact in the axonemal motor pathway is not defined; however, we had previously shown that DNAAF5 variants continue to interact with SPAG1, a potential pathway binding partner (14). Third, while we also sought to determine if variants in Dnaaf5 result in transcriptional responses that may provide a molecular phenotype, the use of scRNA-Seq is limited and is exploratory, with no similar reported studies for comparison. In that regard, we do not know how MIS/MIS mutant or the PCD subjects with DNAAF5 variants can compensate for decreased motor proteins, though one possibility is the augmentation of the overall activity within the dynein axonemal particles that are proposed to control motor protein assembly (58).

In conclusion, our data support the contention that different combinations of pathologic variants of the PCD gene Dnaaf5 may lead to varying partial function, resulting in mild or severe disease, dependent on the protein function and independent of genetic background or environment. Moreover, each allelic variant of DNAAF5 and any of the more than 50 genes known to cause PCD may lead to a gene dosage effect that contributes to the phenotype. In cases of variants that do not result in RNA decay and null conditions, interpreting the effects of compound heterozygosity of variants may require a comprehensive evaluation including functional assays to establish the contribution of each variant to disease. Such considerations do not negate the importance of environmental exposures, access to medical care, and appropriate management of chronic conditions when trying to make broad genotype-phenotype links.

Supplemental Methods are available online with this article.

Generation of Dnaaf5 mutant mice

The Dnaaf5 C498F mouse was created in a C57BL/6 background using reagents designed and validated at the Genome Engineering & Stem Center at Washington University. Briefly, gRNAs were designed to cleave as close to the C498 position as possible. The gRNAs were produced as synthetic CRISPR RNAs (crRNAs) that were annealed with the trans-acting CRISPR RNA (tracrRNA), complexed with recombinant Cas9 protein, and validated as described in Supplemental Material. The gRNA/Cas9 complex with both single-stranded oligo DNA nucleotides were electroporated into single-cell C57BL/6 strain embryos. Embryos (20–25) were transferred to each pseudopregnant C57BL/6 mother (59). Live born mice were genotyped using next-generation sequencing (NGS), as during validation. Mice were contained in a microorganism barrier facility, and all lines were bred in a single room. Tail biopsies were used to extract DNA, submitted to NGS, to identify genotypes.

Airway epithelial cell culture

Human nasal epithelial cells were isolated from biopsy brushes as previously described (14). Mouse airway epithelial cells were isolated from trachea harvested from animals and grown as previously described (36, 60). Human and mouse basal epithelial cells were expanded in culture and were then differentiated on supported membranes (Transwell) using ALI conditions. Cell preparations were maintained in culture for 4–10 weeks.

TEM

Fresh or cultured airway epithelial cells, sperm, and fragments of brain lateral ventricles were fixed in 2% paraformaldehyde/2% glutaraldehyde in 100 mM sodium cacodylate buffer and processed for TEM as described in Supplemental Methods. Cilia cross sections were scored by 2–3 readers, blinded to the genotype. Axonemal doublets were not scored unless the A and B microtubules could be discerned in the images.

Epithelial cell immunofluorescence staining

Airway cells were fixed and immunostained as previously described (36, 61). Primary antibodies used were acetylated α-tubulin (1:500, clone 6-11-B1, Sigma-Aldrich), rabbit polyclonal DNALI1 (1:100, Sigma-Aldrich), CFAP126 (HPA045904, 1:100, Sigma-Aldrich), and NME9 (HPA040000, 1:100 Sigma-Aldrich). Primary antibodies were detected using fluorescently labeled secondary antibodies (Alexa Fluor, Invitrogen). Nuclei were stained using DAPI. Images were acquired using a Ti2 Nikon epifluorescence microscope interfaced to a CMOS camera and Elements imaging software (Nikon). Images were globally adjusted for brightness and contrast using Affinity Photo (Serif Ltd.).

High-speed video microscopy of multiciliated cells

Cells were imaged live and recorded using a Nikon Eclipse Ti-U inverted microscope modified with lenses that use phase contrast and Hoffman modulation contrast (NAMC, Nikon). The microscope was enclosed in a customized environmental chamber maintained at 37°C as described (10, 62). Images were captured by a high-speed video CMOS camera and processed with the Sisson-Ammons Video Analysis system (Ammons Engineering). CBF was analyzed in at least 5 fields obtained from each preparation, after visually confirming ciliated cells in the analyzed areas.

Computer-assisted sperm analysis

Sperm obtain from the cauda epididyma were assessed assess sperm motility by computer-assisted sperm analysis (CASA) performed using a Hamilton–Thorne digital image analyzer (HTR-CEROS II v.1.7; Hamilton–Thorne Research). Sperm capacitation tests were performed using 1 mM of 8-bromo-cAMP, with additional details in the Supplemental Methods.

Tissue histology

Mouse lungs were inflated via the trachea at 20 cm H₂O with formalin and submerged in buffer formalin overnight at room temperature. Heads from mice were immersion fixed in Bouin’s fixative until the bones were decalcified, after which coronal slices were obtained to capture the lateral ventricles of the brain and the maxillary sinuses. Sections were stained with H&E, and images were acquired by bright-field microscopy using a Nikon Ti2 microscope (Nikon). To measure the ventricle size in mice pups, animals were euthanized and fixed with 10% formalin, after which the brains were extracted and cut using a brain slicer (Zivic Instruments) and imaged as before. Area measurements were performed using the area function within NIS-Elements (Nikon).

Brain MRI imaging and ventricle volume quantification

MRI was performed using a 4.7T Varian and 9.4T (Bruker) MRI scanner (Varian Inc.) using settings described in the Supplemental Methods. The lateral and fourth ventricle volumes were calculated using ITK-SNAP software (Version 3.8.0), based on the slice thickness multiplied by the ventricle volume on each slice.

Cilia transport measurement

Transport on the surface of trachea and brain ventricle was determined by imaging fluorescent bead movement. Trachea were removed, fully opened across the length of the sagittal plane, and submerged in a well containing PBS within a 37°C temperature-controlled enclosure; the ciliated surface was visually confirmed. Beads (Fluoresbrite, 2 μm diameter; Polysciences) were diluted 1:500 in phosphate buffered saline. Then, 10 μL were added to the surface of the distal trachea. For functional imaging of the ependyma cilia flow network, whole mount of the lateral wall of lateral ventricle was prepared as previously described (63). For brain studies, the diluted microbeads were deposited with a micropipette on the dorsal side, toward the anterior region of the lateral wall of the lateral ventricle. Details of flow recording are in the Supplemental Methods.

Cilia isolation and mass spectroscopy

Ciliary axonemes were isolated from the surface of highly ciliated airway cells by application of cilia buffer as described, with some modifications (64) as described in the Supplemental Methods. TMT labeling was performed using the TMT 10-plex reagent kit (Thermo Fisher Scientific). Detailed methods of liquid chromatography–mass spectroscopy analysis is included in the Supplemental Methods.

ScRNA-Seq analysis of airway epithelial cells

Cultured primary airway cells were prepared for scRNA-Seq by dissociating the ALI cultures as previously described (65). Cell viability was maintained above 80% across all samples. Library preparation and sequencing was performed by the Genome Technology Access Center core at Washington University in St. Louis. Sequencing data are available at the Gene Expression Omnibus, under accession no. GSE229917. For each sample, 20,000 cells were loaded on a Chromium Controller (10× Genomics) for single-cell capture, and cDNA was prepared according to the 10× Genomics protocols as described in the Supplemental Methods. Paired-end sequencing reads were processed by Cell Ranger (10× Genomics software, version 2.0.0). For each sample, 5,010–9,910 cells were captured. The average sample had a mean of 85,446 reads per cell (ranging from 62,408 to 112,0111 depending on the library). The median number of genes detected per cell on the average sample was 4,487. The samples were preprocessed using Seurat package (66) and filtered to remove stressed or dead cells (those with mitochondrial gene content of more than 25%), potential doublets (cells with more than 7,500 genes detected), or low-quality cells (those with less than 2,500 genes detected). Clustering was performed using the FindClusters function in Seurat. Clusters of cells were manually annotated according to their known marker gene, and gene expression and clustering results were displayed on UMAP.

Statistics

Analysis was performed using GraphPad Prism (version 9). Differences between 2 groups were compared using the Mann-Whitney U test. Multiple medians were compared using the Kruskal-Wallis test, followed by Dunn’s multiple-comparison test. Paired comparisons were analyzed using the Wilcoxon signed-rank test. The χ² test was used to test genotype ratios for Mendelian fit. For airway clearance studies, 2-tailed unpaired t tests and 1-way ANOVA (Tukey’s multiple-comparison test) was performed to compare each condition. P < 0.05 indicated a statistically significant difference.

Study approval

Human studies. Human studies were performed with permission from the IRB of Washington University in St. Louis (IRB no. 201705095). Patients were seen at the St. Louis Children’s Hospital PCD and Rare Lung Disease Clinic. Consent from parent or guardians and assent was obtained from children, if above 10 years old. Airway epithelial cells were obtained by brush of the inferior surface of the middle turbinate.

Animal studies. All animal studies were performed with the permission of the IACUC of Washington University in St. Louis.