A novel hypomorphic allele of Spag17 causes primary ciliary dyskinesia phenotypes in mice

Cilia are centriole-derived, microtubule-based membranous extensions that exist in almost every cell (Gilula and Satir, 1972; Pedersen et al., 2008; Satir, 2005). Ciliary structures are highly conserved across the animal kingdom and can be classified broadly into two main types, largely based on the structure of the ciliary axoneme. The nonmotile primary cilia have a ‘9+0’ axonemal structure (Nogales et al., 1999; Satir, 2005) because they lack the central pair of microtubules and other molecular motors, such as dynein arms and radial spokes, responsible for ciliary movement. These primary cilia are mainly involved in mechanosensory functions (Goetz and Anderson, 2010; Pazour et al., 2000; Sánchez and Dynlacht, 2016). Nodal cilia are motile cilia that lack the central pair apparatus but do have inner and outer dynein arms and can cause a leftward fluid flow to convey an asymmetric signaling event, resulting in asymmetric expression of laterality genes, such as Pitx2 (Brueckner, 2001; Marszalek et al., 1999; Nonaka et al., 1998). The motile cilia develop a central pair of microtubule singlets in the center of the axoneme and are therefore described as having a ‘9+2’ arrangement of microtubule doublets (Nogales et al., 1999; Satir, 2005). The well-documented function of motile cilia is the coordinated rhythmic beating to move body fluids in the brain, respiratory tract and male and female genital ducts (Brightman and Palay, 1963; Dirksen, 1971; Jeffery and Reid, 1975). This is consistent with the localized tissue distribution of motile cilia to these organs.

Defects in the assembly or function of motile cilia can cause primary ciliary dyskinesia (PCD). PCD (OMIM: 244400) is a rare and highly heterogeneous condition, with variants in more than 40 causative genes reported to date (Bustamante-Marin et al., 2019; Cindrić et al., 2020; Lucas et al., 2019). In spite of this progress, a genetic diagnosis remains elusive in approximately 35% of PCD patients (Kurkowiak et al., 2015; Yang et al., 2018). PCD manifests as chronic respiratory tract infections, infertility, and laterality defects in around 50% of cases (Afzelius, 1981; Leigh et al., 2019). These laterality defects in PCD are thought to develop owing to defective nodal cilia in the developing embryo. Hydrocephalus occurs infrequently in individuals with PCD and might reflect dysfunctional ependymal cilia, although various other mechanisms have also been proposed (Berlucchi et al., 2012; Del Bigio, 2010; Kosaki et al., 2004; Lee, 2013; Vieira et al., 2012; Wessels et al., 2003). Intriguingly, most rodent models of PCD genes develop hydrocephalus, but there is significant evidence for phenotypic variability based on the genetic background (Abdelhamed et al., 2018; Brody et al., 2000; Fernandez-Gonzalez et al., 2009; Finn et al., 2014; Ha et al., 2016; Ibañez-Tallon et al., 2004; Lechtreck et al., 2008; Lee et al., 2008; Sironen et al., 2011).

Ciliary motility is an ATP-dependent process that results in activation of dynein motor proteins between the pairs of microtubule doublets that are the major structural component of the ciliary axoneme (Satir and Sleigh, 1990). A number of studies indicate that signals from the central pair propagate through radial spokes to modulate the dynein activity and, ultimately, affect ciliary beating and flagellar movement (Adams et al., 1981; Smith and Yang, 2004; Wirschell et al., 2009; Yang and Smith, 2009; Zhu et al., 2019). These studies generally conclude that dynein is a downstream effector of the central pair apparatus (Huang et al., 1982; Porter et al., 1994, 1992; Rupp et al., 1996). The central pair consists of the two 13-protofilament microtubule singlets (C1 and C2) interconnected by a bridge-like structure, in addition to several projections docked onto the C1 (C1a-C1f) and C2 (C2a-C2b) singlets. At least 23 polypeptides, ranging in molecular weight from 14 to 360 kDa, make up the central apparatus (Adams et al., 1981; Zhao et al., 2019). The C1 and C2 microtubule singlets and their associated projections are structurally and biochemically distinct. At least 10 different polypeptides are uniquely associated with the C1 microtubule, and seven are unique to the C2 microtubule (Dutcher et al., 1984; Lee, 2011). This biochemical and structural asymmetry is believed to have functional consequences for ciliary or flagellar beat and waveform (Inaba, 2011). Studies in Chlamydomonas reinhardtii indicate that the 240 kDa protein PF6 is the largest protein in the C1a projection (Goduti and Smith, 2012; Rupp et al., 2001), with multiple roles, such as acting as a scaffold protein essential for assembly of the smaller C1a components (Rupp et al., 2001), interaction with the calmodulin calcium-binding protein (Wargo et al., 2005) and, possibly, playing a role in modulating the activity of both inner and outer dynein arms (Goduti and Smith, 2012). As expected, Chlamydomonas pf6 mutants develop paralyzed flagella and lack the central pair C1a projection (Rupp et al., 2001).

Sperm associated antigen 17 (Spag17) is the mammalian homolog of Chlamydomonas pf6 (Zhang et al., 2005). SPAG17 protein is present in motile cilia and flagella with a ‘9+2’ axoneme structure (Zhang et al., 2005, 2007), including the ciliated ependymal cells (Gokce et al., 2016). Similar to its ortholog, the mammalian SPAG17 protein is present in the central pair apparatus of motile cilia and is essential for development of the C1a projection of the C1 microtubule singlet (Rupp et al., 2001; Teves et al., 2013; Zhang et al., 2007). SPAG17 and other C1 proteins, including SPAG6 and SPAG16L, are known to bind within the C1a projection (Zhang et al., 2005, 2007). Consistent with this biology, Spag17 null mice develop severe ciliary motility defects that lead to hydrocephalus, severe respiratory distress and death within 12 h after birth (Teves et al., 2013). Spag17 mutant mice also have phenotypes beyond the motile cilia, including skeletal defects, with shorter and fewer primary cilia on the Spag17 mutant chondrocytes, osteoblasts and embryonic fibroblasts (MEFs) (Teves et al., 2015). This role in mouse skeletal development is consistent with genome-wide association studies implicating SPAG17 in the control of human height (Kim et al., 2010; N'Diaye et al., 2011; Takeuchi et al., 2009; van der Valk et al., 2014; Weedon and Frayling, 2008; Weedon et al., 2008; Wood et al., 2014; Zhao et al., 2010). More recently, SPAG17 variants were reported to cause primary ciliary dyskinesia in human patients (Andjelkovic et al., 2018). These variants are linked to male infertility owing to severe asthenozoospermia (Xu et al., 2018) and affect almost all stages of spermatogenesis in mice (Kazarian et al., 2018).

This study presents a phenotypic analysis of a new Spag17 hypomorphic allele that we recovered from a 1-ethyl-1-nitrosourea (ENU) mutagenesis forward genetic screen. We named this allele primary ciliary dyskinesia only (Pcdo). Our data indicated that this Spag17^Pcdo allele was not perinatal lethal, but homozygous mutants had a shorter lifespan and developed PCD phenotypes attributable to defects in the structural development of motile cilia. We saw no effects of the Spag17^Pcdo variant on the development or function of primary cilia.

We identified the Pcdo mutant in a mouse ENU mutagenesis forward genetic screen for recessive alleles disrupting organogenesis. ENU mutagenesis was performed as described previously (Stottmann and Beier, 2014), and we phenotyped the mice to recover mutant alleles at early postnatal stages as part of an experiment to look for mutants with abnormal forebrain development. As described further below, Pcdo mutants were initially identified by an enlarged head and distended lateral ventricles, easily visible upon gross dissection. Further histological examination confirmed the occurrence of hydrocephalus and abnormal accumulation of mucus in the respiratory passages. There was no evidence of heterotaxy or any other abnormal organ laterality.

In order to identify the causal variant in Pcdo mutants, we performed whole-exome sequencing on three phenotypic mutants (Table S1). We filtered for variants that were homozygous for the ‘alternative’ allele in each mutant not present in the dbSNP database (https://www.ncbi.nlm.nih.gov/snp/), to exclude strain polymorphisms predicted to have a ‘high’ or ‘moderate’ impact on protein coding and common to all three sequenced mutants. Only three genes of predicted ‘high’ impact met these criteria: plexin C1 (Plxnc1), Sfi1 homolog, spindle assembly associated (yeast) (Sfi1) and Spag17. A null allele of Plxnc1 exists and does not have the same phenotype as the Pcdo mutants (Pasterkamp et al., 2003). Sfi1 seems to be highly polymorphic, given that this appears routinely in similar analyses in our laboratory; therefore, we initially excluded this as a candidate. The null Spag17 phenotype, however, was sufficiently similar to Pcdo to pursue as a candidate (Teves et al., 2013).

The Spag17-204 (ENSMUST00000164539) transcript has 49 exons and encodes for a full-length SPAG17 protein with 2320 amino acids (ENSMUSP00000134066; Fig. 1A). Another smaller splice variant, encoding a 97 kDa SPAG17 isoform, is found in the testis. This is known to be proteolytically processed during the process of spermatogenesis and sperm maturation to generate 72 and 28 kDa SPAG17 fragments (Silina et al., 2011; Zhang et al., 2005). The mouse SPAG17 protein has a small coiled coil domain, followed by two regions of compositional bias with lysine-rich and glycine-rich regions, followed by a PapD-like domain (flagellar-associated PapD-like, Pfam: PF14874). PapD-like domains are known to be highly conserved in mammals (Zhang et al., 2005; Fig. 1A).

Sanger sequencing of Pcdo DNA confirmed the A to T transversion at the start of exon 36, position c.5236, p.1746 (Fig. 1B). This mutation introduces a nonsense stop codon (TAA) instead of the wild-type lysine (K) amino acid at position 1746. The homozygous Spag17^c.5236A>T mutation segregated completely with the mutant Pcdo phenotype and was confirmed by genotyping assays in more than 50 mutant animals over multiple generations to date. Semi-quantitative RT-PCR analysis of Spag17 RNA using primers spanning exon 33 and 39 indicated high levels of Spag17 transcript in the testis compared with very limited expression in the lungs and brain of 3-month-old animals (Fig. 1C). Quantification of the band intensity and normalization to a corresponding β-actin band were performed. We found that Spag17 transcript was only marginally reduced in the Pcdo/Pcdo mutant tissues compared to wild type, suggesting that the early stop codon did not stimulate a nonsense-mediated decay response (Fig. 1D). We analyzed protein levels using two separate antibodies (Zhang et al., 2005). An antibody against an epitope in the C-terminus (Fig. 1A) reliably identified known isoforms with molecular weights of 97 and 72 kDa. Protein levels in the 1-month-old brain were low and similar in the wild type and mutant. In the testis at 1 month, both isoforms were expressed much more highly than in the brain, and they were completely missing in the mutant. With an antibody against an N-terminus epitope, we were able to identify the full-length 250 kDa species that was missing in the testis at both 1 month of age and in the adult mice. We saw only a slight reduction of this protein isoform in the brain (Fig. 1E). In addition to these isoforms, we also consistently saw an 88 kDa form with both antibodies in the adult testis. This seemed to be completely absent in mutant tissue (Fig. 1F). Given the more subtle effects on RNA, we suspect that this protein is degraded in the mutant tissues.

It was previously reported that the Spag17 germline null animals suffered respiratory distress and died within the first 12 h of postnatal life due to defects in the respiratory motile cilia (Teves et al., 2013). In the Spag17^Pcdo/Pcdo mutant animals, we observed frequent neonatal death within the first day of postnatal life (17.4%, n=4/23). Mutant animals that survived after the first day were viable to weaning age (n>37). Spag17^Pcdo/Pcdo animals that were allowed to age beyond weaning experienced a shortened lifespan, with premature death at around 1-4 months of age (n=6/8). These animals were either found dead or were euthanized for humane reasons due to excessive morbidity. This is most probably attributable to severe hydrocephalus and chronic respiratory insufficiency. Approximately 25% (n=2/8) of Spag17^Pcdo/Pcdo animals lived normally to an age of around 6 months before they were euthanized.

Previous reports indicated that SPAG17 is required for normal primary cilia function and for the growth and elongation of the long bones in mice (Teves et al., 2015). SPAG17 has also been linked to height in human studies (N'Diaye et al., 2011; van der Valk et al., 2014; Weedon and Frayling, 2008; Weedon et al., 2008; Wood et al., 2014). We therefore investigated skeletal development in the Spag17^control (n=16: 3 Spag17^wt/wt, 13 Spag17^Pcdo/+) and Spag17^Pcdo/Pcdo (n=11) animals. We performed Alizarin Red and Alcian Blue skeletal preparations and measured the length of the long bones, including the humerus, ulna, radius (Fig. 2A,B), femur (Fig. 2C,D) and tibia (Fig. 2E,F). Statistical analysis (Student's unpaired t-test) did not detect a significant difference in any long bone lengths when comparing data between Spag17^Pcdo/Pcdo mutants and control animals except for a slight increase in the Spag17^Pcdo/Pcdo mutant femur length (Fig. 2G). The biological significance of this finding remains to be identified. To assay primary cilia structural development further, we generated MEFs from Spag17^Pcdo/+ and Spag17^Pcdo/Pcdo animals and performed immunocytochemistry for the ciliary membrane marker, ARL13B. The number of ciliated cells and the morphology of those cilia in vitro were not different in the Spag17^Pcdo/Pcdo mutant cells when compared with control cells (Fig. 2H-J; 301 cells from Spag17^control and 436 cells from Spag17^Pcdo/Pcdo; four animals of each genotype were included in the analysis). These data indicate the Pcdo mutation does not affect the structure of primary cilia in MEFs or long bone development in the same way as total loss of Spag17.

Hydrocephalus is a prevalent phenotype in PCD mouse models. Likewise, Spag17^Pcdo/Pcdo mutant animals developed a fully penetrant, mildly progressive hydrocephalus. The hydrocephalus was first obvious shortly after birth and before the beginning of the second postnatal week (Fig. 3A-F). Measurement of the lateral ventricle area from mutants compared with wild-type control animals showed no difference at postnatal day (P)1 (Fig. 3A,B; n=3 wild type and 3 mutant). This measurement was significantly increased in the mutants by P7 (Fig. 3C,D; n=3 wild type and 3 mutant) and remained significantly higher at P15 (Fig. 3E-G; n=4 wild type and 3 mutant). No masses of abnormal growths indicative of obstructive hydrocephalus were observed within the ventricular system (Fig. S1). However, we did observe severe overt intracerebroventricular and subarachnoid hemorrhage in a small subset of Spag17^Pcdo/Pcdo mutant animals (n=3/32). As expected, this led to blood clots and obstruction of the ventricular system at various points and, ultimately, severe obstructive hydrocephalus and fatality (Fig. S2).

The hydrocephalus in the motile cilia models has usually been attributed to perturbed flow of cerebrospinal fluid (CSF) towards the next most caudal opening within the cerebroventricular system. A highly conserved function of the ependymal motile cilia lining the brain ventricles seems essential for directing this CSF flow. Given the Pcdo phenotypes, we hypothesized that ciliary beating might be severely compromised in Spag17 animals (Teves et al., 2013). We examined this using ex vivo high-speed video microscopy at P4 (Movies 1 and 2). We showed previously that the ependymal motile cilia follow a spatiotemporal pattern of development in the developing forebrain and lateral ventricle, and ependymal motile ciliogenesis was seen along the medial wall of the lateral ventricle at ∼P0, as opposed to the lateral wall, where ependymal ciliogenesis starts towards the end of the first postnatal week (Abdelhamed et al., 2018). We recorded the ciliary beating in both the lateral ventricle and the aqueduct. Surprisingly, we did not detect any reduction in the ciliary beat frequency in any of the areas we measured (Fig. 3H-J). On the contrary, we observed that the Pcdo mutant motile cilia of the medial walls of the lateral ventricle were slightly hyperkinetic (Fig. 3J). Detailed analysis of the kymographs of the beating cilia showed altered waveforms, as indicated by a shorter interval between forward strokes in Pcdo/Pcdo mutants (Fig. 3H,I). A quantification of this showed that the mutant cilia were beating at a rate significantly higher than control cilia in the lateral ventricle (P=0.002; Fig. 3J). Interestingly, kymographs of the beating cilia and analysis of the cilia beat frequency did not detect any significant difference in the aqueduct cilia (P=0.189; Fig. 3J). We also measured CSF localized flow with fluorescent microbeads introduced ex vivo into the forebrain slice immediately before the video microscopy recording. This analysis did not show abnormalities in flow speed in isolated forebrain lateral ventricle slices (P=0.30; Fig. 3K; Movies 3 and 4). These data are consistent with our conclusion that the Pcdo K1746* allele of Spag17 is a hypomorph, in comparison to the previously reported Spag17 null allele, which develops paralyzed cilia that affect mucociliary clearance and neonatal survival (Teves et al., 2013).

The aqueduct of Sylvius is a narrow channel for CSF flow connecting the third ventricle to the fourth ventricle. This channel is obstructed in mouse models with deficits in primary cilia (Town et al., 2008) and in motile ciliopathy attributable to collapse and fusion of the walls of the ventricle (Ibañez-Tallon et al., 2004). Histological characterization of the aqueduct at P4 from Pcdo mutant animals indicated fusion of the ependymal lining at multiple points of the aqueduct (Fig. 4A-D; n=3 wild type and 3 mutants). We also noted some enlargement of the subcommissural organ (SCO) in the Spag17^Pcdo/Pcdo mutant animals (Fig. 4E,F). Later in development, the mutant aqueduct appeared collapsed and shrunken (Fig. 4G-J; n=4 wild type and 4 mutants). The SCO is an area of highly differentiated ependyma located in the dorsocaudal region of the third ventricle at the entrance to the aqueduct of Sylvius and is well known for secreting high molecular weight glycoproteins necessary for CSF flow and circulation (Rodríguez et al., 1987). Hydrocephalus is a common feature in animal models with loss or defective development of the SCO (Blackshear et al., 2003; Cao and Wu, 2015; Sakakibara et al., 2002; Stoykova et al., 1996). Collectively, these studies and others suggest that the secretory activity of the SCO is responsible, at least in part, for the maintenance of an open aqueduct. We suspected that aqueductal stenosis could cause localized CSF flow abnormalities and hydrocephalus in the Spag17^Pcdo/Pcdo mutants. We tested this with video microscopy and fluorescent microbeads in the intact aqueductal lumen (Movies 5 and 6). We saw that the aqueductal stenosis significantly affected the directionality and speed of the moving beads in the P4 mutant aqueduct (Fig. 4K; P<0.0001; n=2 control and 4 mutant animals). We conclude from these data that CSF bulk flow is extremely disturbed along with aqueductal stenosis and that this is what ultimately leads to the hydrocephalus in the Pcdo mutants.

Gross anatomical assessment of the visceral organs in Spag17^Pcdo/Pcdo mutants showed that organ laterality was unaffected in this model in all animals examined (Fig. S3; n=5 animals each genotype). Histological and gross anatomical examination of the lung did not show any right or left lung isomerism, and gross lung development was not affected in the Spag17^Pcdo/Pcdo mutants (Fig. S3; n=18 control and 15 mutants). However, detailed histological examination of the lungs and trachea from P8 and P14 Spag17^Pcdo/Pcdo mutants showed abnormal accumulation of homogenous eosinophilic material filling the upper respiratory passages, the trachea and main bronchi, indicative of mucus accumulation, most probably attributable to defective mucociliary clearance (Fig. 5A-D; n=6 control and 6/7 mutants).

Another common PCD phenotype is male infertility. All male Spag17^Pcdo/Pcdo animals we tested (n=4) were infertile and were unable to produce any litters when mated with wild-type control female mice (when paired twice for 5 weeks at each attempt, n=8). Conventional histological analysis of the testis was performed at 5 weeks (n=4 wild type, 3 mutants), 10 weeks (n=3 wild type, 3 mutants) and 4 months of age (n=3 wild type, 3 mutants). In all stages of testicular development we examined, the Spag17^Pcdo/Pcdo mutant seminiferous tubules were devoid of any sperm, with a clear lumen, whereas all control sections showed evidence of normal spermatogenesis, with mature sperm observed in the center of the seminiferous tubules (Fig. 5E-J). No clear difference was noted at the level of the germ cells or the number or morphology of developing spermatocytes (Fig. 5E-J). The round spermatids failed to mature fully and failed to produce flagella (Fig. 5E-J). This indicates that the processes involved in spermiogenesis, including development of the sperm flagellum, require SPAG17.

Given the functional deficits we observed in the motile cilia, we performed a morphological analysis of the motile ciliary structure. We first performed immunohistochemical analysis with the acetylated α-tubulin antibody to stain the ciliary axoneme. No apparent morphological differences were detected in the Spag17^Pcdo/Pcdo mutant ependymal and/or respiratory cilia when compared with wild-type control cilia (Fig. 6A-D). We saw similar results with a more stringent scanning electron microscopy (SEM) analysis. The abundance of cilia and length of cilia appeared comparable in ependymal cilia lining the medial and lateral walls of the forebrain lateral ventricle, indicating that the Spag17^Pcdo allele does not perturb production of motile cilia in either the brain ependymal cells or the respiratory epithelial cells (Fig. 6E-H; n=4 wild type and 4 mutants). By contrast, Spag17^Pcdo/Pcdo mice showed deficits in formation of the sperm flagellum. Consistent with the histological analysis above, immunohistochemical staining of testicular sections from Spag17^Pcdo/Pcdo mutants showed a total lack of acetylated α-tubulin staining at 5 and 10 weeks of age in comparison to the age-matched controls, in which we observed clear and pronounced acetylated α-tubulin staining (Fig. 6I-L; n=3 wild type and 3 mutants for each age).

SPAG17 is a structural protein within cilia essential for development of the C1a process of the C1 microtubule structure of the central pair apparatus (Zhang et al., 2005). We therefore investigated the ultrastructure of the ependymal and respiratory motile cilia using transmission electron microscopy (TEM). Cross-sections of ependymal cilia appeared fairly normal, with no clear defects in either the structure or the orientation of the central pair apparatus or other motility structures, such as inner or outer dynein arms (Fig. 7A-C). This is consistent with the limited functional disturbance in ependymal ciliary beat frequency and the undisturbed localized (intraventricular) CSF flow observed in this model. By contrast, TEM analysis of the cross-sections of respiratory cilia showed predominantly abnormal respiratory tracheal cilia in the Spag17^Pcdo mutants. Two types of related deformities can be observed clearly. First, mutant respiratory cilia lack one of the central pair microtubule singlets in a subset of the mutant cilia, with the resulting appearance of ‘9+1’ microtubule organization (Fig. 7D-G). Second, most of the remaining cilia in the mutant trachea have an abnormal central pair apparatus, with one of the central pair singlets appearing smaller or incomplete compared with wild type (Fig. 7H-K). Rotational polarity of the motile cilia axoneme is dependent on the central pair orientation (Kunimoto et al., 2012). Orientation of the axonemal central pair of microtubules was measured by drawing a line through the central pair and measuring the angle of the line with respect to the 0-180° angle arbitrarily set for the first line drawn, as previously described (Hegan et al., 2015; Kunimoto et al., 2012). The deviation from this angle was larger in the Pcdo mutant cilia, and 22.2% of the Spag17^Pcdo mutant central pairs (n=42) in the respiratory cilia were not properly aligned (22.2% versus 16.7% in wild type, n=18; Fig. 7L,M). This indicates that rotational planar cell polarity of the central pair was perturbed in the Spag17^Pcdo/Pcdo mutant respiratory cilia. SPAG17 is not only essential for central pair development but could also be indispensable for establishing the central pair rotational polarity. The observed degree of central pair disorientation could have a strong negative influence on the highly coordinated ciliary beating and thereby contribute to impaired directional fluid flow and mucociliary clearance in the lung.

The observed central pair orientation defect prompted us also to assay the orientation of the basal body basal feet in the mutant respiratory cilia compared with wild type. The directionality of ciliary beating is determined by the orientation of the basal feet of the basal bodies, which are usually positioned at the fourth, fifth and sixth circularly arranged triplets of microtubules within the basal bodies. The basal feet associated with the basal bodies are usually oriented in the direction of the effective stroke of the ciliary beating, and this is controlled by the planar cell polarity pathway (Boisvieux-Ulrich et al., 1985; Gibbons, 1961; Kunimoto et al., 2012). However, we found that in the Spag17^Pcdo/Pcdo mutant cilia (n=16 cilia), the basal feet of the basal body appeared to be aligned normally when compared with wild-type control cilia (n=52 cilia; Fig. 7N,O). We did not observe further defects in other structures required for ciliary motility, such as inner and outer dynein arms.

Altogether, our data indicate that there is a differential requirement for SPAG17 among the different cell types that produce motile cilia. Based on the phenotypes we observe in the testis, lung and brain, we conclude that SPAG17 is crucial for initiation and formation of the sperm flagellum, whereas it has a more specific role (limited to central pair apparatus and C1a projection development and orientation) during structural development of the respiratory cilia and an even more limited role during differentiation and maturation of cilia of the brain ependymal cells.

In this study, we present a detailed characterization of the Pcdo mutant we recovered from an ENU forward genetic screen and identified as a nonsense allele of Spag17 (Fig. 1A,B). The Pcdo variant is a c.5236A>T nonsense mutation predicted to introduce a premature stop codon at position 1746 of the 2320 amino acid SPAG17 protein. Significantly higher levels of Spag17 RNA were expressed in the testis in comparison to the lungs and brain tissues of the adult mice. These data are consistent with previous Spag17 expression analyses (Zhang et al., 2005). The Pcdo mutation resulted in only a slight reduction in RNA expression in mutant tissues (Fig. 1C,D), indicating that the Pcdo mutation does not induce nonsense-mediated decay in any tissue examined. Protein analyses of mutant tissue samples from testis and brain indicated that the Pcdo mutation leads to the loss of multiple SPAG17 protein isoforms in testis but not in brain, although the ENU-induced variant is common to the full-length isoform of Spag17 known to be expressed in testis, brain and lung. Our analysis also identified very low levels of the SPAG17 97 kDa isoform in the brain, which did not seem to be affected in the Pcdo mutants. The precise mechanism leading to loss of the SPAG17 protein in the Pcdo mutants remains to be determined. Similar observations have been documented in studies of another motile cilia gene, adenylate kinase 7 (Ak7), in which mutation leads to loss of protein in the testis but not in the respiratory cilia (Lorès et al., 2018). Our data indicate that proteins expressed in both motile cilia and sperm flagella might have tissue-specific properties in each of these two highly related organelles. This might be attributable to the major differences between sperm flagella and simple motile cilia in the brain and the respiratory epithelium.

Spag17^Pcdo/Pcdo mice developed early postnatal communicating hydrocephalus and had accumulation of mucus in the respiratory passages, and male mutants were infertile. Although the majority of Pcdo mutant animals survived weaning, they had a shortened lifespan compared with controls. All these phenotypes are consistent with deficits of the motile cilia in the Pcdo mutants. Despite these indicators of deficits of the motile cilia, beating and fluid flow were not reduced in the mutant lateral ventricle. Instead, we determined the cause of the hydrocephalus to be progressive aqueductal stenosis leading to CSF fluid turbulence and disrupted ventricular bulk flow. Pcdo mutant mice did not have obvious defects in formation of the primary cilia in the MEFs that we examined directly. Furthermore, the Pcdo mutants did not have the hallmark phenotypes of mouse models lacking primary cilia (e.g. Marszalek et al., 1999). These findings that Pcdo mutants develop milder gross anatomical and ciliary motility phenotypes and survive longer than the Spag17 knockout mouse phenotypes (Teves et al., 2013) support our conclusion that Pcdo is a hypomorphic allele of Spag17. Previous data have shown some variability in PCD phenotypes based on genetic background; therefore, we cannot exclude the idea that this might contribute to the discrepancy with published results. Our experiments were performed in mice of a mixed C57BL/6J and FVB/NJ background, and the previous report of the null allele used mice of a mixed C57BL/6J and 129S4/SvJ background. However, at least one report has suggested that these phenotypes tend to be less severe on the 129 background (Finn et al., 2014). Taken together, we do not favor a model wherein the differences in phenotype are largely attributable to effects of the genetic background, but this should be considered.

Motile and primary ciliopathies have long been considered as two distinct dysfunctions of two related but functionally different organelles, the motile and nonmotile cilia, respectively. However, recent advances and further characterization of ciliopathy mutants have indicated that motile and primary ciliopathies can overlap in the same animal model and/or human patients (Bukowy-Bieryłło et al., 2013; Ferkol and Leigh, 2012; Fliegauf et al., 2007; Moore et al., 2006). Characterization of the Spag17 knockout mice provided new evidence that motile and primary ciliopathies can overlap or develop in the same model organism (Teves et al., 2013, 2015). SPAG17 mutations are reported to cause human PCD (Andjelkovic et al., 2018), and variants in the gene have been linked to human height (Kim et al., 2010; Takeuchi et al., 2009; Weedon and Frayling, 2008; Weedon et al., 2008; Zhao et al., 2010) and multiple body measurements in goats (Zhang et al., 2019). Interestingly, combined missense changes in SPAG17 and WDR35 cause complex neurodevelopmental malformations and skeletal cranioectodermal dysplasia, a primary ciliopathy (Córdova-Fletes et al., 2018). These studies confirm that Spag17 has a role in the development and function of both motile and immotile primary cilia, but the effects differ based on the precise allele in question.

A hallmark of a central pair PCD ciliopathy is the absence of laterality defects, subtle ciliary beating abnormalities, and unequivocal ultrastructural defects of the ciliary axoneme. Vertebrate organ asymmetry is well known to require the nodal cilia that generate leftward flow across the node. These cilia are motile because they express inner and outer dynein arms but do not have a central pair apparatus. It is therefore intriguing to see that almost no central pair protein mutants, including this allele of Spag17, develop laterality defects (Cindrić et al., 2020; Davy and Robinson, 2003; Lechtreck et al., 2008). However, mutants of the dynein complexes consistently develop various degrees of laterality defects, including dextrocardia and heterotaxy (Abdelhamed et al., 2018; Bartoloni et al., 2002; Dougherty et al., 2016; Loges et al., 2018; Ta-Shma et al., 2018). The lack of laterality phenotypes in the Spag17^Pcdo/Pcdo mutants further supports the model that central pair proteins, including Spag17, are dispensable for the development of organ asymmetry and body laterality.

The main function of the ependymal cilia is to beat and generate CSF flow. The forebrain ependymal cilia in the Spag17^Pcdo mutant have no apparent ultrastructural abnormalities, and the ciliary beating frequency was unaffected in the ependymal cilia of the aqueduct (Fig. 3J). Surprisingly, a subset of the ependymal cilia on the medial wall of the lateral ventricle in the Spag17^Pcdo/Pcdo mutants are hyperkinetic and beat with a rate significantly higher than that of the control cilia (Fig. 3J). Our data suggest, however, that this hyperkinetic beating does not disrupt the local cilia-mediated flow in the acute ex vivo isolated medial wall forebrain slice (Fig. 3K). It is possible that this could have a more prominent effect in vivo, where there are more limiting space constraints. Alternatively, the lack of an effect on the structural development and function of the ependymal cilia could be attributable to the expression of near-normal levels of SPAG17 proteins in brain ependymal cells (Fig. 1). Interestingly, the hyperkinetic cilia phenotype in the Spag17^Pcdo/Pcdo mutants is consistent with other motile cilia mutant phenotypes that have normal motile cilia ultrastructure but hyperkinetic cilia, such as DNAH11 PCD patients (Bartoloni et al., 2002; Dougherty et al., 2016; Knowles et al., 2012; Pifferi et al., 2010; Schwabe et al., 2008) and a mouse model (Handel and Kennedy, 1984). Assessment of this beating pattern of the CSF flow in vivo would be a better indicator of the effect of the hyperkinetic cilia on the fluid flow and development of hydrocephalus. This would also be helpful to explain whether the hyperkinetic cilia phenotype is causing the hydrocephalus in this model directly or whether other mechanisms are responsible.

The aqueduct of Sylvius in the Pcdo mutants appeared stenotic, with collapsed walls, and was associated with a slight enlargement of the SCO (Fig. 4). In other models, patency of the aqueduct is compromised in the presence of a dysfunctional SCO or SCO ependymal cilia (Pérez-Fígares et al., 2001; Swiderski et al., 2012), leading to aqueductal stenosis or occlusion and noncommunicating hydrocephalus. Ciliary beating function and directionality were defective when assayed in the intact aqueduct lumen. The stenosis could have led to a mechanical constraint on the ciliary beating and thus to significantly reduced flow of CSF into the fourth ventricle. We concluded that Pcdo mutants had communicating hydrocephalus because the aqueduct was stenotic and not completely occluded, indicative of a potentially milder dysfunction of the SCO.

Surprisingly, we noted severely defective structure of the respiratory motile cilia consistent with a central pair protein insult. The Spag17^Pcdo mutant cilia either lack one of the central microtubule pair or have an abnormal central pair structure. The orientation of the central pair also appeared not to align in one direction, indicative of disorganization of ciliary central pair polarity (Fig. 7L,M). This is expected to interfere with a directional fluid flow in the respiratory passages and lead to mucus accumulation and recurrent infection. The axis of polarity is controlled by signaling and mechanical cues. The mechanical cue is a cilia-generated fluid flow (Mitchell et al., 2007). The intense mucus accumulation observed in the Spag17^Pcdo/Pcdo mutants indicates that ciliary motility in the lung is defective. This is consistent with the structural defects and is likely to contribute significantly to increased morbidity and the shortened lifespan in these mutants. We should note a limitation to our ultrastructural studies. We were able to analyze multiple cilia, but these were from a relatively limited set of animals. Current world-wide experimental conditions are challenging our ability to collaborate in order to obtain more robust sample sizes for this portion of the study. Moreover, development of the sperm flagellum was completely inhibited in the Pcdo mutants, and no mature sperm with a flagellum can be detected at any postnatal stage of testicular development (Figs 5E-J and 6I-L).

Altogether, our data indicate that SPAG17 is necessary for sperm flagellum development and important for proper assembly of the respiratory motile cilia central pair apparatus but is not required in a similar manner in the development of ependymal cilia. Spag17 is highly expressed in the testis, with moderate expression levels in the lungs, and the lowest levels observed in the brain (Fig. 1C,D). We note a correlation between the basal Spag17 expression levels and the downstream results of Pcdo mutation in the different motile cilia examined. Spermatocytes and round spermatids seem to have an absolute requirement for SPAG17. Although other cell types of motile cilia can be more tolerant of Pcdo mutation, the respiratory cilia cannot compensate for the mild reduction in Spag17. At the protein level, SPAG17 is known to interact with other central pair proteins, such as SPAG16 and SPAG6, and to form a complex at the C1a microtubule projections (Fu et al., 2019; Wargo et al., 2005; Zhang et al., 2005; Zhao et al., 2019). One explanation of this finding is that the minimal levels of SPAG17 produced in the Pcdo mutant ependymal cells were sufficient to facilitate the interaction with other proteins and to form the C1a projection, but much more SPAG17 was needed in the respiratory cilia to form these protein complexes and thereby form the correct central pairs of the motile respiratory cilia. All isoforms of SPAG17 were completely abolished in the Pcdo mutant testis; therefore, it seems that the C1a projection interactome was unable to form, and this led to complete failure of formation of the sperm flagellum. Several testis-specific proteasome/ubiquitin enzymes have been reported in rodents (Mochida et al., 2000; Nishito et al., 2006) and humans (Mitchell et al., 1991; Uechi et al., 2014). It is therefore also possible that the mutant SPAG17 protein is degraded in the testis by a sperm-specific proteasome-mediated quality control mechanism (Morales et al., 2003; Sutovsky, 2011; Zimmerman and Sutovsky, 2009). Other possibilities include the simple explanation that not all tissues construct axonemes from the same protein set, and SPAG17 is functionally compensated for in cilia outside the testis. Altogether, these data are consistent with recent reports that effects of specific gene mutations on protein translation and thereby function might differ dramatically between various cell types expressing the same allele (Lucas et al., 2019).

In summary, this study describes a new allele of Spag17 that encodes the C1a projection SPAG17 protein. The Pcdo allele recapitulated most PCD phenotypes previously observed in other models due to central pair abnormalities. These are often the most difficult to diagnose in humans owing to lack of laterality phenotypes, only subtle beating defects and largely undisturbed ciliary ultrastructure (Edelbusch et al., 2017). We suggest that the Spag17^Pcdo allele is a very useful model for studying the pathogenesis and molecular mechanisms of human PCD subsequent to central pair defects.

ENU mutagenesis was performed as described by Herron et al. (2002) and Stottmann and Beier (2014). Briefly, 6- to 8-week-old C57BL/6 Mus musculus males (G0 males) were injected intraperitoneally with three weekly fractioned doses of ENU and then bred to FVB Mus musculus females (The Jackson Laboratory) to generate G1 heterozygous carrier males. These G1 males were then outcrossed to FVB females to generate G2 potentially heterozygous carrier females. These G2 females were backcrossed to their respective G1 male parent to generate G3 embryos or pups, which were screened for organogenesis phenotypes. The Pcdo mutant mice were first identified with dilated brain ventricles and hydrocephalus. All animals were maintained through a protocol approved by the Cincinnati Children's Hospital Medical Center IACUC committee (IACUC2016-0098), and animal care and use complied with all relevant local animal welfare laws, guidelines and policies. Mice were housed in a vivarium with a 12 h-12 h light-dark cycle, with food and water ad libitum.

Neonatal pups were sacrificed by decapitation. For histology of the adults, littermate animals underwent cardiac perfusion using cold heparinized PBS and formalin (Sigma-Aldrich) solution. Brains, lungs and testes were dissected and fixed for ≤72 h in formalin at room temperature, followed by immersion in 70% ethanol (for histology). Samples were then paraffin embedded and sectioned at 6 µm thickness. Sections were processed for Hematoxylin and Eosin staining with standard methods. Histological sections were imaged using a Zeiss Discovery.V8 Stereoscope or Nikon NIE upright microscope. The area of the lateral ventricles was measured from the coronal brain sections. We limited our analysis to sections that contained the anterior commissure as an anatomical landmark for consistency between wild-type and Spag17^Pcdo/Pcdo mutant animals. Analysis was performed using the area measurement tool within the Nikon Element 4.50 software.

Immunohistochemistry was performed as previously described by Driver et al. (2017). Briefly, fixed paraffin blocks from control and mutant animals were sectioned at a thickness of 6 µm. Sections were deparaffinized and rehydrated in graded ethanol. Antigen retrieval was performed by boiling in citrate buffer pH 6.00 for 1 min in a microwave. Nonspecific antigens in the tissue sections were blocked in 4% normal goat serum in PBS-Tween, and the anti-acetylated α-tubulin antibody (1:2000, Sigma-Aldrich, cat# T6793) was incubated with the tissue overnight at 4°C, followed by three stringent washes in PBS-Tween and application of Alexa Fluor 596-conjugated goat anti-mouse secondary antibody (1:500, Invitrogen, cat# A11005) for 1 h, followed by three washes in PBS-Tween. Nuclei were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) and mounted in ProLong Gold antifade mounting media (Invitrogen). Images were captured using a Nikon C2 confocal microscope.

Brain tissue from three obviously affected, postnatal male mice was used for exome analysis. Whole-exome sequencing was done with BGI Americas, using standard protocols. Annotation of single nucleotide polymorphisms was performed at BGI Americas with an in-house software package, ‘AnnoDB’, incorporating elements from analysis platforms such as Sorting Intolerant From Tolerant (SIFT), Polymorphism Phenotyping v.2 (PolyPhen-2), phylogenetic P-values (PhyloP), GERP (Genome Evolutionary Rate Profiling) and Functional Analysis through Hidden Markov Models (FATHMM).

The Spag17 nonsense variant in exon 36, identified by whole-exome sequencing, was confirmed by Sanger sequencing using the following primers: forward, 5′-TTTGCCATTTGGATTTACAGG-3′; and reverse, 5′-GTAGCACTTTGGGTCCTTCG-3′. PCR amplicons from Spag17^wt/wt, Spag17^Pcdo/wt and Spag17^Pcdo/Pcdo animals were purified and concentrated using the Zymo DNA Clean and Concentrator kit (Zymo Research Corporation). One hundred nanograms of concentrated DNA was used in each Sanger sequencing reaction.

All animals and pups obtained from the Spag17^Pcdo mouse line were genotyped using the TaqMan Sample-to-SNP kit (Applied Biosystems) for a single-nucleotide change at mouse chr1:100088282A>T (assay ID ANWCWCA). Spag17^Pcdo TaqMan Sample-to-SNP assays were performed according to the manufacturer's instructions and run using a QuantStudio 6 Real Time PCR machine (Applied Biosystems).

Total RNA was extracted from testes, lungs and whole brains of adult mice using TRIzol reagent (Thermo Fisher Scientific) according to the manufacturer's instructions. Five micrograms of total RNA was used for reverse transcription into complementary DNA (cDNA) using the SuperScript III First Strand synthesis system (Invitrogen) according to the manufacturer's directions. Final cDNA products were diluted 1:10, after which 2 µl of the diluted cDNA was used to amplify Spag17 transcript with primers that span exon 33 and exon 39 (forward, 5′-GATGGAGGGCTACGAAAGC-3′; reverse, 5′-AACTGTTAGGTGGGCTGCAA-3′). β-Actin was used as loading control (forward, 5′-GTGACGTTGACATCCGTAAAGA-3′; reverse, 5′-GCCGGACTCATCGTACTCC-3′). PCR products were run on a 2% agarose gel with ethidium bromide for 45 min at 120 V. PCR bands were imaged using a Bio-Rad Universal Hood II Molecular Imager w/CFW-1312M Camera equipped with Image Lab v.6.0.1 software. The image analysis tool in the Image Lab v.6.0.1 software was used to quantify the band intensity value. Data obtained from the Spag17 exon 33-39 bands were normalized against the β-actin value of the corresponding sample. Finally, wild-type and mutant data obtained for testes, lungs and brain were normalized against the average wild-type testis expression values. One-way ANOVA with multiple comparisons statistical analyses were performed using GraphPad Prism v.8.0.1 software.

Adult mouse brain and testis tissues were lysed in Pierce RIPA buffer (Thermo Fisher Scientific, cat# 89901) containing protease inhibitor cocktail (Roche, cat# 11697498001). The protein concentration in the whole cell extracts was determined with the BCA colorimetric assay (Thermo Fisher Scientific) according to the manufacturer's instructions. Denatured proteins were separated by electrophoresis on a gradient of 4-12% Tris-glycine gel. The protein was transferred to a polyvinylidene difluoride membrane, blocked in Odyssey blocking buffer and incubated overnight at 4°C with 1:3000 rabbit anti-N-terminus or C-terminus anti-SPAG17 antibodies (Zhang et al., 2005) and 1:2000 mouse anti-tubulin antibodies (Sigma-Aldrich, cat# T6199). Membranes were washed and incubated for 1 h in 1:15,000 goat anti-rabbit IRDye 680CW (LICOR) and 1:15,000 goat anti-mouse IRDye 800Rd (LICOR), and bands were visualized on the LICOR Odyssey imaging system.

The medial and lateral walls of the P8 forebrains were processed for SEM as previously described (Abdelhamed et al., 2018). Briefly, samples were fixed in electron microscopy grade 2% paraformaldehyde and 2.5% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.4) at 4°C overnight. Tissues were washed thoroughly in 0.1 M sodium cacodylate buffer (pH 7.4) then post-fixed in 1% osmium oxide (diluted in 0.1 M sodium cacodylate buffer) for 1 h. Samples were washed thoroughly in 0.1 M sodium cacodylate buffer and dehydrated before critical point drying in 100% ethanol. Brain tissues were then coated with gold palladium using a sputter coater (Leica EM ACE600) and scanned with a Hitachi SU8O1O scanning electron microscope.

Brain ependymal cells and tracheal epithelial cells were fixed in 2.5% glutaraldehyde and processed for TEM analyses by standard protocols as previously reported (Wallmeier et al., 2019). Sections were collected on copper grids, stained with Reynold's lead citrate and visualized using the Philips CM10 or Jeol 1400+.

High-speed video microscopy of the beating ependymal cilia, ciliary beat frequency and CSF flow analysis were performed as previously described (Abdelhamed et al., 2018). Briefly, brains were dissected out from P4 Spag17^Pcdo/Pcdo or wild-type control animals in Dulbecco's modified Eagle's medium: Nutrient Mixture F-12 (DMEM/F12) supplemented with l-glutamine (Gibco) and 1% N2 supplement (Gibco) at room temperature, and cut into serial 200-µm-thick coronal sections. Sections were obtained from the forebrain and aqueduct. Areas from the medial wall of the lateral ventricle were microdissected further and subjected to video microscopy recording at ∼100 frames/s. These videos were used to measure the ciliary beat frequency. For concurrent imaging of the beating cilia with green fluorescent microbeads (FluoSpheres, Thermo Fisher Scientific), the beads were introduced into slices immediately before video recording, and we used an inverted Nikon Ti-E wide-field microscope fitted with a Nikon ×40 Plan Apo 0.95 N.A. air objective and an Andor Zyla 4.2 PLUS sCMOS monochromatic camera. Light was channeled through a custom quad-pass filter, and 300 frames were collected at ∼60 frames/s. Bead tracking and ciliary beat frequency analysis were performed using NIS Element software v.4.5. Statistical analysis and pairwise comparisons were performed using GraphPad Prism software.

For skeletal preparations, adult animals were sacrificed by induction of deep anesthesia using isoflurane, followed by trans-diaphragmatic cardiac extrusion. Animals were eviscerated and fixed for 2 days in 95% ethanol. They were stained overnight at room temperature in Alcian Blue solution (Sigma-Aldrich, cat# A3157) containing 20% glacial acetic acid. Excess stain was cleared in 95% ethanol for 24 h, and skeletons were then slightly cleared in a 1% KOH solution overnight at room temperature. They were then stained overnight in Alizarin Red solution (Sigma-Aldrich, cat# A5533) containing 1% KOH. A second round of clearing was performed by incubating tissues in 20% glycerol/1% KOH solution for 24 h. Finally, they were transferred to 50% glycerol/50% ethanol for photography. Skeletal preparations were imaged using a Zeiss Discovery.V8 Stereoscope.

MEFs were generated from embryonic day (E)13.5 embryos. Embryos were dissected in PBS, decapitated and eviscerated. The remaining tissue was incubated in trypsin overnight at 4°C. Tissue fragments were incubated with the trypsin in a 5% CO₂ incubator for 30 min. Cells were then allowed to grow to confluency in complete DMEM containing 10% fetal bovine serum and penicillin/streptomycin. MEFs were stained within three passages of their isolation. Cells were stained for the ciliary membrane marker ARL13B (1:500, ProteinTech, cat# 17711-1-AP) overnight at 4°C, after blocking nonspecific antigenic sites with 4% normal goat serum in PBS-Tween for 1 h. Alexa Fluor 488-conjugated goat anti-rabbit secondary antibody (1/500, Invitrogen, cat# A11008) for 1 h, and nuclei were stained with DAPI.

MEFs were generated in the laboratory directly from mouse embryos with standard protocols. All antibodies (with the exception of anti-SPAG17) used in this study were commercially available, and our results matched multiple published reports. SPAG17 results were consistent with previous results, and we saw loss of SPAG17 immunoblot signals in Spag17 mutants, which is ideal validation of an antibody. Sample sizes were generally selected by historical precedent in the field of three or more mutant/control pairs from three different litters, as far as possible (see Discussion for study limitations for some sample sizes). This assignment of mutant/control from different litters also accounted for allocation to experimental groups. Most of the analyses presented in this article were performed by an investigator blinded to the genotypes until after initial data collection. This was possible because the initial tissue collections and genotyping PCRs were often done by a different member of the scientific team. Samples or animals were excluded from the study only in the case of pre-established and clear experimental failures; these failures included immunohistochemistry, immunocytochemistry or immunoblots where no appropriate signal was obtained for any specimens in the experiment. No animals in obvious distress or showing signs of morbidity were used in this study, with the exception of the observations of intraventricular hemorrhage presented in the supplemental data. GraphPad Prism v.8.0.1 software (GraphPad Software, LLC) was used to perform all statistical analysis presented in this study. Analysis also included an analysis for appropriate methodology and sample distribution(s) for the appropriate test. Paired comparisons were made with Student's t-test (two tailed), and multiple sample data sets were analyzed with an ANOVA followed by multiple comparison testing. Graphs all use lines to show the mean±s.d.

We thank Chelsea Menke for technical assistance throughout this project, as well as Maria Teves for helpful discussion and the anti-SPAG17 antibody.