Dynll1 is essential for development and promotes endochondral bone formation by regulating intraflagellar dynein function in primary cilia

Abstract

Mutations in subunits of the cilia-specific cytoplasmic dynein-2 (CD2) complex cause short-rib thoracic dystrophy syndromes (SRTDs), characterized by impaired bone growth and life-threatening perinatal respiratory complications. Different SRTD mutations result in varying disease severities. It remains unresolved whether this reflects the extent of retained hypomorphic protein functions or relative importance of the affected subunits for the activity of the CD2 holoenzyme. To define the contribution of the LC8-type dynein light chain subunit to the CD2 complex, we have generated Dynll1-deficient mouse strains, including the first-ever conditional knockout (KO) mutant for any CD2 subunit. Germline Dynll1 KO mice exhibit a severe ciliopathy-like phenotype similar to mice lacking another CD2 subunit, Dync2li1. Limb mesoderm-specific loss of Dynll1 results in severe bone shortening similar to human SRTD patients. Mechanistically, loss of Dynll1 leads to a partial depletion of other SRTD-related CD2 subunits, severely impaired retrograde intra-flagellar transport, significant thickening of primary cilia and cilia signaling defects. Interestingly, phenotypes of Dynll1-deficient mice are very similar to entirely cilia-deficient Kif3a/Ift88-null mice, except that they never present with polydactyly and retain relatively higher signaling outputs in parts of the hedgehog pathway. Compared to complete loss of Dynll1, maintaining very low DYNLL1 levels in mice lacking the Dynll1-transcription factor ASCIZ (ATMIN) results in significantly attenuated phenotypes and improved CD2 protein levels. The results suggest that primary cilia can maintain some functionality in the absence of intact CD2 complexes and provide a viable animal model for the analysis of the underlying bone development defects of SRTDs.

Introduction

Most non-hemopoietic cells contain a primary cilium, a short antenna-like signaling organelle on the cell surface required for transducing a wide range of chemical, physical and mechanical signals from the environment to the cell body (1,2), with particular importance for the hedgehog signaling pathway (3,4). Primary cilia are built around a specialized microtubule bundle, the axoneme, which provides tracts for the intra-flagellar transport (IFT) of cilia components by the anterograde kinesin-2 motor and associated IFT-B cargo-binding complexes and the retrograde cytoplasmic dynein-2 (CD2) and IFT-A complexes, respectively. Genetic loss of anterograde IFT components typically leads to the complete absence of cilia (because essential building blocks cannot be moved to the growing ‘plus’ end of axonemal microtubules). In contrast, mutations in retrograde IFT components cause a variable range of structural aberrations (lengthening, shortening, bulging at the base or bulging at the tip), but usually not the loss of primary cilia (1).

Congenital defects in cilia function can lead to a range of disease syndromes that are collectively referred to as ‘ciliopathies.’ The importance of the CD2 complex is highlighted by findings that mutations in its subunit genes are the most common cause of a spectrum of skeletal chondrodysplasias, the short-rib thoracic dystrophies (SRTDs) (5–13), which affect approximately 1/200 000 newborns (14). The defining clinical feature of these syndromes is severe shortening of bones that is formed by endochondral ossification, especially of the ribs and in the limbs; less penetrant symptoms include polydactyly and kidney, liver and retinal defects (5–12). In about half the cases, the severe narrowing of the thoracic cavity leads to perinatal death from acute respiratory failure, and more patients die as infants. However, among patients who survive past 2 years, respiratory problems progressively diminish with age as their ribs continue to grow, and some patients even reach normal height as adults (14,15).

CD2 contains eight subunits, and mutations in five of these genes have so far been identified in SRTD patients: the catalytic heavy chain DYNC2H1, which is the most prevalent SRTD locus; the intermediate chains WDR34 and WDR60; light intermediate chain DYNC2LI1; and the light chain TCTEX1D2 (5–13). To date, no SRTD patient mutations have been found in the light chains DYNLL1/2, DYNLRB1/2 and DYNLT1/3. Different mutations in the affected CD2 subunits can lead to different SRTD disease severities, from Jeune syndrome, which is often compatible with life, to the perinatal lethal short-rib polydactyly syndromes I–IV (OMIM 208500). It is unresolved whether this variability in clinical outcomes reflects differences in the relative importance of the affected CD2 subunits for the stability and activity of the CD2 holoenzyme complex or differences in the residual biochemical functionality of the mutated subunits.

A similar phenotypic heterogeneity is also observed among the currently available mouse mutants for CD2 subunits. Although homozygous mutations in all subunits investigated so far consistently lead to embryonic lethality, there are some notable differences in the severity of developmental defects and the reported cilia structure and signaling defects (1). Apart from different genetic backgrounds, a reason for this phenotypic variability may be that the majority of available CD2 alleles contain N-ethyl-N-nitrosurea (ENU)-generated point mutations, or randomly generated gene traps, which may not completely disrupt the function of the affected gene. However, even the phenotypes of the two targeted, bona fide null mutations in CD2 subunits differ in severity. Although Dync2li1 knockout (KO) mice die early in development [approximately Embryonic Day 10.5 (E10.5)], mice lacking the intermediate chain WDR34 exhibit a milder phenotype where some mice survive until about E16.5 and then present with polydactyly (16,17).

Mechanistic insight into the physiological importance of cilia for bone development has been gained from conditional mouse KO models that lack the anterograde motor protein KIF3A or the IFT-B subunit IFT88 in limb bud mesoderm-derived cells (18–21). In these mice, complete absence of cilia in the targeted cells leads to impaired signaling in response to the hedgehog proteins sonic hedgehog (SHH) and indian hedgehog (IHH), and results in limb patterning defects with severe polydactyly, and extreme bone shortening due to premature hypertrophy of proliferating chondrocytes (18–21). In this context, the prominent skeletal abnormalities in human SRTD patients most likely also relate to the essential role of cilia in hedgehog signaling during endochondral bone formation. However, because of the relatively early embryonic lethality of all currently available animal models with CD2 subunit mutations, the role of the retrograde IFT motor complex in bone development has remained difficult to study.

In contrast to the subunits affected by SRTD mutations, the light chains DYNLL1/2, DYNLRB1/2 and DYNLT1/3 are also shared with the other dynein families, the cytoplasmic dynein-1 (CD1) complex and the axonemal dyneins. DYNLL1 is unique among these subunits in that it also interacts with and regulates the function of dozens of other proteins that are involved in a wide range of dynein-independent cellular functions, including protein kinases (NEK9, PAK1 and TLK1/2), apoptosis regulators (BIM and BMF), DNA repair factors (53BP1 and MRE11) and several transcription factors (e.g. ER-α, TRPS1 and ASCIZ; 22,23).

To better understand the role of DYNLL1 as a subunit of the CD2 motor complex in an in vivo context, we have here generated germline and conditional Dynll1 KO mouse models. We show that mice lacking Dynll1 in the germline are developmentally arrested from approximately E8.5 and die at mid-gestation with incomplete embryonic turning and severe neural tube and cardiac defects. This is strikingly similar to the phenotype of mice containing a targeted null allele for the CD2 light intermediate chain, Dync2li1, and phenocopies the defects of entirely cilia-deficient Kif3a and Ift88 mice (16,24–26). In addition, we report that limb bud mesoderm-restricted conditional Prx1-Cre Dynll1-deleted mice exhibit severe bone shortening and thereby represent the first viable animal model that accurately replicates the defining clinical feature of human SRTD patients. We show that loss of DYNLL1 leads to partial reductions in the levels of other CD2 subunits, severe retrograde IFT defects, widening of primary cilia and impaired hedgehog signaling in vivo and in vitro. Overall, the phenotypes of Dynll1-deficient mice are similar to Kif3a- or Ift88-deficient mice but are in some aspects attenuated, indicating that some cilia-specific signaling functions are preserved despite the severe retrograde transport defects. Finally, our data also reveal that the phenotypes of targeted Dynll1 KO models are considerably more severe than those of mice lacking the essential Dynll1 transcription factor ASCIZ (27) as well as previously reported gene-trapped Dynll1^GT mice (28), indicating that very low levels of DYNLL1 are sufficient for largely normal cilia signaling functions.

Results

Germline Dynll1 KO results in severe developmental delays and mid-gestation lethality

We recently reported the characterization of B lymphoid-specific conditional Mb1-Cre Dynll1-deleted mice (29,30), in which the recombined Dynll1^fl allele lacks the majority of the promoter region and two of three exons and where the remaining exon lacks any in-frame ATG codon that could give rise to a truncated protein product. To generate a targeted germline Dynll1 null allele, we took advantage of the low-level ectopic activity of Mb1-Cre in germ cells (31) and selected incidental Dynll1^+/− mice obtained during routine breeding as founders for the Dynll1 KO line.

Germline Dynll1^−/− embryos were developmentally grossly delayed and growth retarded at E9.5, resembling a normal embryo at around E8.5 (Fig. 1A). All Dynll1 null embryos analyzed at E9.5 exhibited severe cardiac edema (ballooning of the pericardial sac indicative of terminal heart failure; Fig. 1A), and no KO embryos were detected past E11.5. At E9.5, most Dynll1 KO embryos failed to initiate embryonic turning (class I mutants), about a third arrested turning approximately half-way (class II mutants) and only a small fraction (class III mutants) had completed turning to adopt the fetal position (Fig. 1A and B). In addition to invariant cardiac edema, a large fraction of Dynll1 null embryos exhibited neural tube closure and tail development defects (Fig. 1B) and ~25% exhibited reversed heart looping as a sign of inverted left-right axis formation (Supplementary Material, Fig. S1A and B).

Figure 1

Dynll1 null mice exhibit severe ciliopathy-like features. (A) Classification of Dynll1^−/− embryos into three groups according to the extent of embryonic turning at E9.5. Class I embryos [bottom right 53% (n = 20)] exhibit the most severe phenotype and fail to initiate embryonic turning. Class II embryos [bottom left, 37% (n = 14)] arrest turning half way with the tail to the left or right side, and Class III embryos [top right, 10% (n = 4)] complete turning. (B) Frequency of Dynll1^−/− embryo classes, cardiac oedema and neural tube closure defects. (C) Whole embryo extract western blot analysis of Dynll1 WT, heterozygous and KO littermates. (D) qPCR expression analysis of Dynll1 and Dynll2 mRNA levels in WT (+/+) and Dynll1 null (^−/−) embryos at E10.5 (n = 3–5 per group). Error bars represent the mean ± SEM.

As expected, DYNLL1 protein was undetectable in lysates of whole Dynll1^−/− embryos in western blot experiments (Fig. 1C). In addition, levels of Dynll1 mRNA using a probe directed at the non-deleted exon 3 were reduced by ~ 25-fold in KO embryos compared to wildtype (WT) littermate controls (Fig. 1D), consistent with impaired activity of the truncated promoter sequence in the KO allele or increased turnover of the truncated mRNA by the non-sense mediated RNA decay pathway, or both. However, the severity of the developmental defects in Dynll1 KO embryos was in a way surprising, considering that mammals also contain a Dynll2 gene, whose gene product shares >90% identity with DYNLL1 (83 of 89 amino acid residues) and might therefore be able to at least partially compensate for the loss of Dynll1. Remarkably, quantitative PCR (qPCR) expression analyses of E9.5 whole embryo extracts revealed that Dynll2 is normally expressed at ~ 500-fold lower levels than Dynll1, and its expression remained unchanged in Dynll1^−/− embryos (Fig. 1D). Thus, the extremely low basal levels of Dynll2 expression, and the lack of its upregulation in Dynll1 null embryos, indicate that there is no compensation by Dynll2 for loss of Dynll1 during early embryonic development in mice.

Altogether these results demonstrate that Dynll1 is essential for mouse development to progress past the stage of embryonic turning. It should be noted that this phenotype—including the cardiac edema, reversed heart looping, open neural tubes and the proportional distribution of the turning defects (classes I–III)—is strikingly similar to the defects observed in Dync2li1^−/− embryos (16), which until recently represented the only targeted KO mutant of any CD2 subunit in the mouse. Moreover, the Dynll1 KO phenotype is also remarkably similar to targeted null mutations of the anterograde IFT motor genes, Kif3a and Kif3b (16,24–26), supporting the notion that the defects in Dynll1 KO embryos are to a large part due to impaired cilia functions.

Figure 2

Prx1-Cre ΔDynll1 mice exhibit profound skeletal and hair follicle defects. (A) Prx1-Cre and Prx1-Cre ΔDynll1 littermates at 8 weeks of age. (B) Alizarin-red/Alcian-blue stained forelimbs of 8-week-old Prx1-Cre and Prx1-Cre ΔDynll1 mice. (C) Bone caliper measurements of Prx1-Cre and Prx1-Cre ΔDynll1 long bones (n = 8 per group). Error bars represent the mean ± SEM. (D) Frozen skin sections of Prx1-Cre and Prx1-Cre ΔDynll1 mice at P1 and P7. (E) Alizarin-red/Alcian-blue stained forelimbs from Prx1-Cre and Prx1-Cre ΔDynll1 mice at 8 weeks.

Conditional Prx1-Cre Dynll1-deleted mice exhibit ciliopathy-like bone shortening and hair follicle defects

To avoid mid-gestational lethality and expand the genetic analysis of Dynll1 toward ciliopathy-associated bone defects, we adopted a conditional KO approach using the mesoderm-specific Prx1-Cre transgene, which had been instrumental in establishing the bone-specific functions of the ciliogenic anterograde IFT components Ift88 and Kif3a (18,19). Prx1-Cre expression is relatively specific to the limbs and sternum (which avoids lethal complications of impaired rib and vertebrae development), parts of the cranial and craniofacial mesenchyme and the dermal mesenchyme on parts of the ventrum (32). PCR analysis of Prx1-Cre^TG Dynll1^fl/fl embryos confirmed complete deletion of Dynll1 in the forelimb buds between E9.5 and E10.5 and less efficient deletion in the hindlimb bud (Supplementary Material, Fig. S2A), consistent with the previously reported timing of Prx1-Cre expression (32,33). qPCR analysis of Dynll1 mRNA in limb buds at E11.5 revealed significantly decreased Dynll1 expression, similar to that seen in the germline Dynll1 KO embryos, when compared to Prx1-Cre control limb buds (P = < 0.0001) (Supplementary Material, Fig. S2B), and western blot analysis confirmed the complete loss of DYNLL1 protein in mutant limb buds at this time (see Fig. 6D).

In contrast to the mid-gestational lethality seen in constitutive Dynll1 mutants, Prx1-Cre Dynll1-deleted mice survive to adulthood but exhibit a number of developmental defects. The most striking defects in the Prx1-Cre Dynll1-deleted mice are an extreme shortening of all long bones and a severe hypotrichosis on the ventrum and around the limbs (Fig. 2A–C). Bone caliper measurements at 8 weeks revealed significantly shorter long bones, with between 30% and 40% shortening compared to controls (P < 0.0001) (Fig. 2C). Histological analysis of ventral skin sections of Prx1-Cre Dynll1-deleted mice revealed normal hair follicle morphogenesis and numbers at Postnatal Day 1 (P1) but arrested follicle maturation at Stages 2–3 by P7, when follicles in littermate controls had matured to Stages 7–8 [Fig. 2D; for staging see (34)]. These bone shortening and hypotrichosis phenotypes are comparable to published phenotypes for Prx1-Cre Kif3a-deleted or Ift88-deleted mice, which lack cilia in the limb bud mesoderm as well as on dermal fibroblasts, and the hair defects are also similar to mice lacking Shh or Gli2 (19,35,36).

In contrast to the comparable bone length and skin defects, there are two important phenotypic differences between our Prx1-Cre Dynll1-deleted mice and the Prx1-Cre Kif3a/Ift88 precedent. First, although both the Prx1-Cre Kif3a-deleted and Ift88-deleted mice exhibit severe forelimb polydactyly with six to nine digits and complete loss of the anterior–posterior polarity of digit patterning (18,19), the Prx1-Cre Dynll1-deleted mice never contained more than five digits and digit identity and asymmetry were maintained (Fig. 2E). However, Alizarin-red/Alcian-blue skeletal staining of both forelimbs and hindlimbs of Dynll1 mutants revealed severely dysmorphic carpal and phalangeal bones, with a range of bone malformations and fusions. These defects were more severe in the forelimbs, presumably related to the delayed and irregular onset of Prx1-Cre driven recombination in the hind limb (32,33). Among 20 forelimbs (10 mice) collected at 8 weeks of age, 20% appeared oligodactylous (appearance of only 4 digits), 10% exhibited syndactyly (the fusion of 2 or more digits), 45% exhibited forking of a metacarpal or phalangeal bone (45%) and all limbs presented with synphalangism (fusion of phalangeal joints) (Fig. 2E and Supplementary Material, Fig. S2D). Similar defects were also observed in forelimbs collected from Prx1-Cre Dynll1-deleted embryos at E16.5 and E18.5, including syndactyly and forked phalanges, but again without any cases of polydactyly (Supplementary Material, Fig. S2C). Thus, these data indicate a unique role for Dynll1 in digit formation and patterning that is different from the cilia-deficient anterograde IFT mutants.

While ribs appeared normal in length and width, as expected based on the Prx1-Cre expression pattern, a second difference between Prx1-Cre Dynll1-deleted and Prx1-Cre Kif3a-deleted mice (18) was that Dynll1-deleted mice did not exhibit the abnormal asymmetrical rib attachment to the sternum (Supplementary Material, Fig. S2E). Finally, although Prx1-Cre is also expressed in the anterior cranial mesoderm (32), similar to the Kif3a mutant mice (18), Dynll1 mutant mice had no overt cranial or craniofacial defects when compared to Prx1-Cre controls (Supplementary Material, Fig. S2F).

Impaired endochondral bone formation and altered bone structure in Prx1-Cre Dynll1-deleted mice

In light of the severe bone shortening, we monitored a key stage of bone development in frozen sections of the femur and tibia of E18.5 Prx1-Cre Dynll1-deleted embryos and littermate controls. Longitudinal bone growth is driven by a process called endochondral bone formation, during which proliferating chondrocytes gradually differentiate into quiescent hypertrophic chondrocytes that synthesize osteogenic extracellular matrix components (37,38). The proliferating chondrocytes are normally stacked into neatly organized columns along the axis of bone growth; during growth they divide perpendicular to this axis, and then ‘rotate’ back, into their assigned growth column (37,38).

In contrast to littermate control sections, Prx1-Cre Dynll1 mutant growth plates were much shorter, with fewer proliferating chondrocytes, and were severely disorganized (Fig. 3A–H): chondrocytes in growth columns were discontinuous and aligned at twisting angles relative to the longitudinal bone axis, and numerous round, pre-hypertrophic and hypertrophic cells were interspersed between the flattened proliferating chondrocytes throughout the region that would normally be the proliferative zone (e.g. arrowhead in Figure 3F). Possibly as a consequence of chondrocyte column deviation from the normally strict longitudinal orientation, the diaphysis did not narrow during bone lengthening in Prx1-Cre Dynll1 mutants (Fig. 3B and D) compared to controls (Fig. 3A and C). There were also striking polarity defects in the anterior–posterior axis of the mutant long bones with ectopic chondrocyte proliferation observed emanating from the perichondrium near the bone collar around the center of the diaphysis (Fig. 3D, box and arrowhead), whereas in control bones the growth plates were confined to the proximal and distal ends (Fig. 3A and C). Again, this disorganization of the growth plate and the ectopic growth areas along the bone collar is strikingly similar to the phenotype of Prx1-Cre Ift88-deleted long bones (19) and thus implies a similar cilia-related defect as the underlying cause for the bone defects of the conditional Dynll1 KO mutants.

Figure 3

Endochondral bone formation is impaired in the absence of Dynll1. (A) Frozen H&E stained femur (A and C) or tibia (B and D) sections and proximal end growth plates from femur (E and F) or tibia (G and H) sections from E18.5 Prx1-Cre (A, C, E and G) and Prx1-Cre ΔDynll1 (B, D, F and H) mice. Ectopic chondrocytes are present in long bones of Prx1-Cre ΔDynll1 mice (D, box and arrowhead). (I–O) μCT analysis of bone dimensions in femurs of 8-week-old Prx1-Cre and Prx1-Cre ΔDynll1 mice. (I) Femur length, (J) trabecular thickness, (K) trabecular number, (L) periosteal perimeter, (M) cortical area and (N) cortical thickness (n = 6–7 per group). Error bars represent the mean ± SEM. (O) Representative μCT scan images from femurs of 8-week-old Prx1-Cre and Prx1-Cre ΔDynll1 mice.

To complement this histological analysis and determine the long-term outcome of this defect in bone development, we also performed micro-computed tomography on femurs of 8-week-old male mice (Fig. 3O and Supplementary Material, Fig. S3). Confirming the caliper measurements (Fig. 2C), femur length was significantly lower in the mutants (Fig. 3I). While Dynll1 mutants contained a similar density of trabecular number in the secondary spongiosa compared to controls, trabecular thickness was significantly lower (Fig. 3J and K). In the cortical region, consistent with the lack of diaphyseal narrowing observed during development, periosteal perimeter was significantly greater in Prx1-Cre Dynll1-deleted mice (Fig. 3L), and although slightly lowered, normal cortical thickness and area were maintained (Fig. 3M and N).

Primary cilia are preserved but thickened in the absence of Dynll1

Because of the extensive phenotypic similarities between our germline and conditional Dynll1 KO mice and the corresponding entirely cilia-deficient Kif3a/Ift88 mutants, we next investigated the impact of Dynll1 deletion on cilia formation.

Immunofluorescence microscopy of frozen sections of E12.5 Prx1-Cre Dynll1-deleted limb buds for the ciliary marker acetylated tubulin revealed that cilia were still present within the mutant limb buds, albeit at a reduced density compared to Prx1-Cre controls (Fig. 4A and B). To assess whether or not this may reflect a cell-intrinsic propensity for reduced cilia formation, we freshly isolated primary murine embryonic fibroblasts (MEFs) from E12.5 limb buds from Prx1-Cre Dynll1-deleted embryos and littermate controls for in vitro ciliogenesis assays. Following 60 h serum starvation, cilia formation was modestly but significantly reduced in the Dynll1-deleted MEFs compared to the Prx1-Cre controls; however, cilia length was unaffected by the loss of DYNLL1 (Fig. 4C–E).

Figure 4

Altered cilia morphology in the absence of Dynll1. (A) Frozen sections from Prx1-Cre and Prx1-Cre ΔDynll1 limb buds at E12.5 were stained with acetylated tubulin (green) to identify cilia in the limb bud mesoderm. (B) Quantification of cilia density in control and Dynll1 deficient limb bud sections (four imaged areas per limb bud and thre mice per genotype). (C) In vitro ciliogenesis in primary MEFs following ~ 60 h serum starvation. Acetylated tubulin (red) marks the ciliary axoneme, and gamma-tubulin (green) marks the centrosome. (D) Percentage of ciliated cells, and (E) average cilia length in Prx1-Cre and Prx1-Cre ΔDynll1 primary MEFs quantified from three independent mice/genotype, each in triplicate. (F–H) Scanning electron microscopy of neural tube cilia of E10.5 WT and Dynll1^−/− littermate embryos. (F) Representative images (each group of three cilia is from a different embryo). (G) Cilia length and (H) cilia width measured using ImageJ software (n = 51–76 cilia from three embryos per group). Width was measured a quarter way up from the base of the cilia.

As loss of Dynll1 did not affect the length of cilia, we wanted to determine whether it caused other changes to cilia morphology at higher microscopic resolution. For this purpose, scanning electron microscopy was employed to image neural tube cilia in the germline Dynll1 KO embryos and littermate controls at E10.5. Interestingly, while there was no difference in the percentage of ciliated cells in the neural tube and in the average length of the cilia, cilia in Dynll1 mutants appeared misshapen and significantly wider when compared to littermate controls (Fig. 4F–H).

Altogether, these results indicate that Dynll1 is not essential for ciliogenesis per se but that its absence affects the structural organization of primary cilia and the incidence of cilia formation in some tissues.

Figure 5

Cilia signaling is impaired in the absence of Dynll1. (A) Retrograde IFT was assessed in primary Prx1-Cre and Prx1-Cre ΔDynll1 MEFs following ~ 60 h serum starvation. MEFs were stained with acetylated tubulin (red) to mark the ciliary axoneme and IFT88 as a cargo of the retrograde IFT machinery (green). Prx1-Cre ΔDynll1 MEFs showed an accumulation of IFT88 at the tip and base of the cilium, while Prx1-Cre controls showed positive IFT88 staining solely at the base. Images of individual tubulin and IFT88 channels and are shown in Supplementary Material, Figure S4. (B–E) Western blot analysis of (B and C) E10.5 whole embryo lysates (n = 3–6 per group) and (D and E) fore limb bud lysates (3 independent biological replicates with 10 pooled buds per sample (i.e. 15 embryos per group). Error bars represent the mean ± SEM. (F and G) qPCR analysis for Gli1 and Ptch1 expression in (F) E10.5 in WT or Dynll1 KO whole embryo lysates (n = 4–5) and (G) fore limb bud lysates of conditional Prx1-Cre ΔDynll1 mice and controls (n = 5). (H) qPCR expression analysis for Gli1 and Ptch in three independent sets of serum-starved primary MEFs following 24 h treatment with SAG (100 nm). Error bars represent the mean ± SEM. Note that a paired t-test was used for panel H, and unpaired t-test was used for all other panels.

Dynll1 is required for retrograde IFT and cilia-dependent hedgehog signaling

Given that Dynll1-deficient mice were still able to form primary cilia, we next investigated how cilia functions were affected. To determine if the retrograde IFT function of the CD2 complex was impacted by the loss of DYNLL1, we performed immunofluorescence microscopy analyses of IFT88 in serum-starved primary limb bud MEFs. IFT88 is normally rapidly transported back from to the tip of the cilium by the IFT-A/CD2 machinery and, consequently, was localized predominantly near the base of the cilium in Prx1-Cre control cells. In contrast, in Prx1-Cre Dynll1-deleted limb bud MEFs, IFT88 exhibited a bi-modal distribution with similar staining intensities at the cilia tip and the base (Fig. 5A and Supplementary Material, Fig. S4A), indicating that Dynll1 is essential for effective retrograde transport by the CD2 motor complex. Furthermore, semi-quantitative intensity plots of the IFT88 signals along the cilium showed a clear accumulation of IFT88 at the two ends of the cilium without increased staining along the length of the axoneme (Supplementary Material, Fig. S4A and B), which may suggest that the absence of DYNLL1 primarily impairs the ability of the CD2 motor to get started rather than its processivity during cargo transport.

Primary cilia are essential for the hedgehog signaling pathway, and the premature differentiation of chondrocytes in the growth plate (Fig. 3A–H) and the arrested hair follicle development (Fig. 2D) in Prx1-Cre Dynll1 mice are consistent with impaired responses to IHH and SHH, respectively. Thus, to determine whether the retrograde IFT defect in the absence of Dynll1 translates to a cilia signaling defect, we monitored molecular markers of the hedgehog pathway in our mice.

Figure 6

Reduced levels of CD2 subunits in the absence of Dynll1. Western blot analysis of WDR34 and DYNC2LI1 protein levels in WT and Dynll1^−/− embryos at E10.5 (A–C) and Prx1-Cre and Prx1-Cre ΔDynll1 fore limb buds at E11.5 (10 buds from 5 embryos pooled per sample) (D–F). Quantification was performed using ImageJ.

Cilia are involved in relaying both the absence (through the conversion of the GLI3 transcription factor from its full-length transcriptional activator form, GLI3-FL, to its cleaved transcription repressor form, GLI3-R) and the presence of hedgehog ligands (via the activation of hedgehog pathway target genes such as Gli1 and Ptch1) (20,39). In western blots of E10.5 whole embryo extracts, processing of GLI3-FL to GLI3-R was significantly impaired by ~ 2.5-fold in Dynll1 KO embryos compared to matched WT and heterozygous controls (Fig. 5B and C). This effect was even more enhanced in lysates of E11.5 forelimb buds, with an > 4.5-fold reduced GLI3-R/FL ratio in Prx1-Cre Dynll1-deleted samples compared to Prx1-Cre controls (Fig. 5D and E). Overall, these changes are quantitatively comparable to previously reported results for limb buds of Prx1-Cre Kif3a-deleted mice as well as E11.5 Ift88 mutant embryos (18,19).

Similar to the impaired processing of GLI3, the expression of hedgehog target genes was also impaired in Dynll1-deficient embryos. Gli1 expression was reduced by ~60%, and Ptch1 was reduced by ~25%, in the absence of Dynll1 compared to controls in both the germline KO and the conditional Prx1-Cre mouse models (Fig. 5F and G). These differences are milder than the reported ~80% reduction of Gli1 expression levels in Prx1-Cre Kif3a/Ift88 cKO tissues (18,40).

While these GLI3 processing and Gli1/Ptch expression results reflect the steady state level of hedgehog pathway activity in the developing embryo and limb buds, we also monitored the impact of Dynll1 on ligand-induced activation of the pathway in a homogenous cell population in vitro. For this purpose, primary limb bud MEFs were serum starved for 48 h and then treated with smoothened agonist (SAG) for 24 h. Again, induction of Ptch1 (P = 0.0380) and Gli1 (P = 0.0325) was significantly reduced by ~ 2-fold in Prx1-Cre Dynll1-deleted MEFs compared to Prx1-Cre controls (Fig. 5H).

Taken together, these data indicate that Dynll1 is required for efficient retrograde transport within cilia and the function of the cilia-specific hedgehog signaling pathway.

DYNLL1 is required for normal CD2 subunit levels

The most plausible explanation for the severe IFT88 transport defect (Fig. 5A) in Dynll1-deleted MEFs, and also the overall strikingly similar phenotypes of germline Dynll1 and Dync2li1 KO mice, was that loss of the DYNLL1 subunit interferes with the function of the CD2 motor complex. In cells from human SRTD patients with DYNC2LI1 mutations it has recently been shown that reduced levels of this subunit lead to the destabilization of the remaining CD2 complex and concomitant loss of the DYNC2H1 heavy chain (13,17). We therefore tested whether loss of DYNLL1 could have a similar effect on other CD2 subunits in our mice.

As a surrogate for CD2 complex stability in the absence of DYNLL1, we determined protein levels of the WDR34 and DYNC2LI1 subunits (for which mouse-reactive antibodies are commercially available). In both Dynll1^−/− whole embryo lysates and Prx1-Cre Dynll1-deleted limb bud extracts, there was a clear ~50% reduction of DYNC2LI1 protein levels compared to the relevant control (Fig. 6A, B, D and E). Levels of the WDR34 subunit, to which DYNLL1 binds directly based on yeast two-hybrid assays (Supplementary Material, Fig. S5), were modestly reduced by ~25% in western blot analyses of Dynll1 KO embryos compared to matched controls (Fig. 6A and C). However, the reduction of WDR34 was more noticeable (>50%) in limb bud extracts of the conditional Dynll1-deleted mice compared to Prx1-Cre controls, where a higher signal-to-background ratio allowed for a more reliable quantification (Fig. 6D and F).

Figure 7

Prx1-Cre ΔAsciz mice exhibit an attenuated phenotype. (A) Prx1-Cre, Prx1-Cre ΔAsciz and Prx1-Cre ΔDynll1 mice at 3 weeks of age. (B) Alizarin-red/Alcian-blue stained forelimbs of 3-week-old Prx1-Cre, Prx1-Cre ΔAsciz and Prx1-Cre ΔDynll1 mice. (C) Bone caliper measurements of Prx1-Cre, Prx1-Cre ΔAsciz and Prx1-Cre ΔDynll1 long bones at 8 weeks (n = 8 per group). Error bars represent the mean ± SEM. (D) qPCR expression analysis of Dynll1 mRNA following in Prx1-Cre, Prx1-Cre ΔAsciz and Prx1-Cre ΔDynll1 limb buds at E11.5 (n = 4–10 per group). Error bars represent the mean ± SEM. (E–G) Western blot analysis of DYNLL1, WDR34 and DYNC2LI1 protein levels in Prx1-Cre control, Prx1-Cre ΔDynll1 and Prx1-Cre ΔAsciz limb buds at E11.5 (five embryos pooled per sample). (F and G) Quantification of relative WDR34 and DYNC2LI1 protein levels using ImageJ. (H and I) qPCR expression analysis for Gli1 and Ptch1 in anterior limb bud lysates of Prx1-Cre control and Asciz- or Dynll1-deleted embryos at E11.5 (n = 4–10 per group). Error bars represent the mean ± SEM.

These results suggest that loss of DYNLL1 leads to a partial destabilization of the remaining CD2 complex components. However, considering that the phenotype of germline Dynll1 KO embryos (Fig. 1) is virtually identical that of Dync2li1 embryos (16), this indicates that the residual DYNC2LI1 protein is de facto inactive, which implies that DYNLL1 affects the in vivo function of the CD2 complex beyond its role in subunit assembly and stabilization.

Preservation of very low DYNLL1 levels in mice lacking the Dynll1-transcription factor ASCIZ leads to attenuated phenotypes

Dynll1 expression is regulated in a phylogenetically highly conserved manner by the transcription factor ASCIZ (also known as ATMIN and ZNF822) (28,41–43), and conditional loss of Asciz or Dynll1 leads to virtually identical phenotypes during most parts of B cell development and B cell lymphomagenesis (29,30). However, the germline Dynll1 KO phenotype described above (Fig. 1) is considerably more severe than that of targeted Asciz KO mice on the same C57BL/6 genetic background, which survive until ~E16.5 (27), suggesting that even very low DYNLL1 levels can sustain a significant proportion of its functions during embryonic development. Thus, to further explore how these low DYNLL1 levels affect its cilia-related functions during bone development, we performed conditional Prx1-Cre Asciz deletion analyses similar to those described above.

Interestingly, Prx1-Cre Asciz-deleted mice presented with a macroscopically evident but attenuated phenotype compared to age-matched Prx1-Cre Dynll1-deleted mice on the same C57BL/6 genetic background (Fig. 7A–C). Hypotrichosis was still evident in Prx1-Cre Asciz-deleted mice but noticeably milder than in the conditional Dynll1 mutants (Fig. 7A), and the length of the long bones clustered approximately halfway between the Prx1-Cre controls and the Dynll1-deleted mice of the same age (Fig. 7B and C). Importantly, except for the relatively modest shortening of the femur, trabecular and cortical bone structure of adult Prx1-Cre Asciz-deleted femurs were in the normal range (Supplementary Material, Fig. S6A–D), including the periosteal perimeter that was most significantly increased in the corresponding Dynll1 mutants (Fig. 3J). Finally, Prx1-Cre Asciz-deleted mice also showed a modest shortening of metacarpal and phalangeal bones, but in contrast to the Dynll1-deleted mice lacked any cases of oligodactyly, syndactyly or forking of metacarpal or phalangeal bones (Supplementary Material, Fig. S6E).

As expected, in qPCR analyses of E11.5 limb buds, Dynll1 mRNA levels in Prx1-Cre Asciz embryos were significantly reduced compared to Prx1-Cre controls but remained at slightly higher levels than the residual transcript in the corresponding Prx1-Cre Dynll1 mutants (Fig. 7D). Likewise, in contrast to the absence of detectable DYNLL1 protein in the Dynll1-deleted limb buds, very low amounts of DYNLL1 (~15% of control levels) were preserved in Prx1-Cre Asciz-deleted limb buds (Fig. 7E). Interestingly, this was associated with a markedly improved protein levels of the CD2 subunits DYNC2LI1 and WDR34 (Fig. 7E–G), compared to their ~2-fold reduction in the Prx1-Cre Dynll1-deficient cohort (Fig. 6D–F). Consistent with the improved CD2 subunit levels, Ptch1 expression was significantly improved in the Asciz-deleted limb buds compared to their Dynll1-deficient counterparts, albeit not quite to the level of the Prx1-Cre control (Fig. 7H), although there was only a very marginal improvement in Gli1 expression (Fig. 7H). While the reasons for this differential effect remain unclear, overall these results support the notion that the very low levels of DYNLL1 that remain in the Asciz mutant are sufficient to support a significant proportion of the CD2-related cilia functions during bone development.

Finally, we also tested if further reducing the Dynll1 gene dosage in Prx1-Cre Asciz-deleted mice would shift this attenuated phenotype more toward the phenotype of Prx1-Cre Dynll1-deleted mice. However, heterozygous loss of Dynll1 in the Prx1-Cre Asciz-deleted cells (Prx1-Cre Asciz^fl/flDynll1^fl/+ mice; blue symbols in Supplementary Material, Fig. S6F) led only to very modest additional shortening of the humerus (P = 0.0099) and a trend toward decreased femur length (0.09058) with no effects on the tibia or ulna compared to Prx1-Cre Asciz^fl/fl Dynll1^+/+ mice (red symbols in Supplementary Material, Fig. S6F) and did not cause any severe digit malformations (Supplementary Material, Fig. S6G).

Discussion

The data presented here demonstrate that loss of the CD2 subunit DYNLL1 leads to severe ciliopathy-like phenotypes as a result of impaired retrograde IFT and associated hedgehog pathway signaling defects. Notably, even though primary cilia can still form in the absence of DYNLL1, the phenotypes of germline and Prx1-Cre mediated conditional deletions of Dynll1 are overall remarkably similar to phenotypes that result from the complete absence of cilia in the corresponding KOs of the anterograde IFT components, Kif3a and Ift88 (18,19,24–26,44). The main difference between these phenotypes is that the severe polydactyly present in conditional Kif3a/Ift88 mutants (18,19) is never observed in Prx1-Cre Dynll1-deleted mice. At the molecular level, processing of GLI3 to its repressor form was similarly impaired in the Dynll1 mutant compared to the Kif3a and Ift88 mutants, but defects in the induction of the downstream hedgehog pathway targets Ptch1 and Gli were much milder (0.5–2-fold in Dynll1 mutants compared to ~5-fold in Kif3a and Ift88 mutants). Mechanistically, these differences imply that as long as cilia are present they are still able to fulfill a subset of their functions in hedgehog signaling, even when retrograde IFT is severely impaired. This partial activity may simply involve passive diffusion of cilia-dependent signaling intermediates back into the cytoplasm, or it may reflect an altered equilibrium in the localization of signaling components, such as the hedgehog pathway activator Smoothened, which are normally kept out of the cilium by the IFT-A machinery. Likewise, impaired GLI3-R-mediated repression of hedgehog target genes is believed to be responsible for the severe polydactyly in Kif3a/Ift88 mutants; however, although the GLI3-A/GLI3-R balance is similarly impaired in our Dynll1 mutant, without efficient retrograde transport, GLI3-A may be trapped within the cilium and would therefore not need to be antagonized by GLI3-R.

Among the members of the CD2 complex, the phenotypes reported here for Dynll1 KO mice are most similar to those of Dync2li1 KO mice, which until recently represented the only targeted null mutation of any CD2 subunit (16), but are considerably milder than those of targeted Wdr34 KO mice (17) and the available point mutation and gene-trap alleles of the catalytic CD2 subunit Dync2h1 (45). Given that the Dynll1 KO mice still contain ~50% of normal DYNC2LI1 protein levels, the phenotypic similarity to Dync2li1 null mice is remarkable as it indicates that the remaining DYNC2LI1 is essentially inactive, which in turn implies that DYNLL1 has additional roles in CD2 activity beyond complex assembly and stability. On the other hand, considering that WDR34 and DYNLL1 interact directly with each other in two-hybrid assays (Supplementary Material, Fig. S5), it is surprising—with the caveat of different genetic backgrounds—that the recently reported phenotype for targeted Wdr34 null mice (17) is slightly different from the Dynll1 KO mice. While the precise function of DYNLL1 in the CD2 complex remains to be determined, it has been assumed to be similar to its role in the related CD1 motor, where it is believed to facilitate the homodimerization and folding of the intermediate chain for its assembly into the dimeric holoenzyme complex (46,47). However, the CD2 complex differs from CD1 in that it contains an asymmetric set of non-identical intermediate chains, WDR34 and WDR60 (48), and so far, DYNLL1 has only ever been shown to homodimerize its targets (49). Furthermore, based on fractionation assays of overexpressed subunits, it has recently been proposed that CD2 is comprised of three distinctive subcomplexes, among which DYNLL1 was associated with the WDR34 component but not the WDR60 component (50), which would be consistent with our yeast two-hybrid assays in which DYNLL1 selectively interacted with WDR34 but not WDR60 (Supplementary Material, Fig. S5). In this sense, one explanation for the divergent Dynll1 and Wdr34 KO phenotypes could be that, in the absence of DYNLL1, WDR34 still associates with CD2, but in an aberrant manner that affects the topology of the complex in a more detrimental (dominant-negative) manner than if WDR34 were simply missing. Thus, differential functions of different subunits for CD2 complex assembly and activity could explain phenotypic differences between different subunit mutations, including why some mutants exhibit polydactyly (17,45,51) while others do not. Another emerging possibility could be that the targeted Wdr34 KO allele still retains partial activity (52).

It should be noted that the phenotype of the germline Dynll1 KO mice shown here is more severe than that of mice containing a recently reported Dynll1^GT/GT gene-trap mutation, which supports survival until at least E13.5 and does not impair embryonic turning (28). One possible explanation for this discrepancy could be that the two different Dynll1 mutants are on different genetic backgrounds [C57BL/6 for the Dynll1 KO versus a mixed C3H/HeH background for Dynll1^GT/GT mice (28)]. However, the gross phenotype of Dynll1^GT/GT embryos is very similar to that of two independent Asciz point mutations on C3H/HeH backgrounds and also quite similar to the targeted Asciz KO phenotype on C57BL/6 background, whereas the targeted Dynll1 KO phenotype is dramatically more severe than the Asciz KO (on the same C57BL/6 background). Thus, consistent with our data that the very low residual DYNLL1 protein levels in targeted and conditional Asciz KO mouse models lead to noticeably attenuated cilia-related phenotypic defects compared to the corresponding Dynll1 KO mice, it would be conceivable that alternative splicing around the gene trap in the Dynll1^GT/GT allele may support the production of some DYNLL1 protein resulting in a milder, hypomorphic phenotype.

A surprising component of the phenotype of the Prx1-Cre Dynll1-deleted mice is the unique digit patterning defects, including synphalangism, syndactyly, digit forking and oligodactyly. The phalangeal and metacarpal bone fusions present in all of the Prx1-Cre Dynll1-deleted forelimbs are similar to the digit joint fusions reported in mice lacking the 5′-Hoxd cluster (53,54). As the 5′-Hoxd cluster regulates joint formation antagonistically to Gli3 (55), the synphalangism in Prx1-Cre Dynll1-deleted mice could be related to altered GLI3 processing. However, overall these digit patterning defects are fundamentally different from the phenotypes of other cilia-related mutants, so that it is likely that they are due to a mechanism that may involve one of the dozens of CD2-independent DYNLL1 targets.

In terms of CD2-independent functions, in some ways, it is also surprising that the phenotypes reported here are not more severe considering that DYNLL1 is also part of the ubiquitous CD1 complex that exerts fundamentally important functions during each and every cell cycle. CD1 has been implicated in chromosome movements, spindle organization and positioning as well as checkpoint silencing during mitosis (56–58), and loss of the Dynll1 ortholog ctp results in mitotic defects in Drosophila (42). However, during our recent conditional Dynll1 deletion analyses in mouse B lymphoid cells, we found that innate B cell subsets were still able to rapidly expand by self-renewal without DYNLL1, and B cell cancers also maintained their ability to proliferate in its absence as long as the pro-apoptotic DYNLL1 target BIM is simultaneously disabled (29,30). One possibility could be compensation by the almost identical DYNLL2 protein, but based on our expression analyses, Dynll2 is expressed at extremely low levels before mid-gestation (Fig. 1D) and in the E11.5 limb bud samples DYNLL2 protein (which cross-reacts with the non-discriminatory DYNLL1 antibody used) was undetectable in Prx1-Cre Dynll1-deleted samples (Figs 6D and7E). With the caveat that we cannot rule out the possibility that DYNLL2 is expressed at higher levels at other times than those tested here, an alternative explanation would be that, for unknown reasons, DYNLL1 may be more important for the structural integrity of the CD2 complex than the CD1 complex. Finally, it should be noted that while our results clearly support a crucial role of CD2-related cilia signaling, it is possible that dynein-independent roles of DYNLL1, for example in the regulation of BIM (59,60) and DNA repair-related proteins (61–63), may also contribute to the phenotypes of Dynll1-deficient mice reported here.

In conclusion, this study has provided the first-ever conditional KO mouse model for any CD2 subunit gene. The Prx1-Cre Dynll1-deleted mice accurately replicate characteristic clinical features of the bone defects observed in human SRTD syndrome patients and thereby represent a first step toward viable animal models to better understand the pathogenesis and prognosis of this disorder, which is predominantly caused by CD2 subunit mutations. While Prx1-Cre Dynll1-deleted mice do not exhibit polydactyly, it should be noted that this feature is also only present in a minority of human patients with CD2 mutations (14). Future studies using Dermo-Cre, which is expressed throughout the limb mesoderm but also extends to the rib anlagen, may be useful to extend this model to the rib cage defects, which are a hallmark of SRTDs and responsible for their life-threatening perinatal respiratory complications.

Materials and Methods

Ethics statement

Animal experiments were performed according to the Australian Code for the Care and Use of Animals for Scientific Purposes, 8th edition (2013), and approved by the St. Vincent’s Hospital Melbourne Animal Ethics Committee, approval numbers 019/13 and 002/17.

Mice

Mice were housed in specific pathogen-free micro-isolators. All mouse mutations were generated on an inbred C57BL/6 background or had been backcrossed to the C57BL/6 background for at least 10 generations. Conditional Asciz (27) and Dynll1 (29) alleles and Prx1-Cre (32) mice have been described before. Mice were analyzed at the ages indicated in the figure legends.

Embryo and neonate analyses

The time of pregnancies defined as E0.5 on the morning vaginal plugs was observed. Embryos were dissected from the uterus in cold phosphate-buffered saline (PBS) and immediately processed for histology, protein or RNA extraction or MEF isolation, and genotyped by PCR using yolk sac or tail DNA.

Mice at P1 or P7 were euthanized and subsequently scalded at 65°C–70°C for 20–30 s to allow for the removal of skin and visceral tissue. Ventral skin was processed for histology immediately and mice were processed for skeletal staining.

Skeletal staining

Skin, muscles and visceral organs were removed from embryos, neonatal animals or limbs from 8-week-old mice prior to fixation in 95% (v/v) EtOH for 1–2 days. Samples were transferred to acetone overnight before incubation in a solution containing one part Alcian-blue (Sigma Aldrich) [0.3% (w/v) in 70% (v/v) EtOH] and one part Alizarin-red (Sigma Aldrich) [0.1% (w/v) in 95% (v/v) EtOH] in a 1:17 mixture of glacial acetic acid and 70% (v/v) ethanol, for 2 days at 40°C with continuous shaking. Skeletal samples were cleared by incubation in 1% KOH and stored in 100% glycerol (64). The skeletal preparations were analyzed and photographed in glycerol solution.

Histology and immunofluorescence microscopy of tissue sections

Hind limbs from E18.5 embryos were removed and fixed in 4% (v/v) paraformaldehyde overnight at 4°C. Fixed tissues were then infiltrated with 30% sucrose in PBS overnight at 4°C. Sucrose-equilibrated samples were flash frozen in OCT and 10–20 μm sections were cut on a Leica CM1900 cryostat (Leica Microsystems GmbH, Wetzlar, Germany). Unfixed ventral skin sections were flash frozen in OCT, and 40 μm sections were cut as above. Sections were stained with hematoxylin and eosin yellow (H&E) using standard protocols. Slides were visualized using a Leica DM 2000 microscope, and images were taken using an Olympus DP72 camera and Olympus cellSens entry software.

For analysis of cilia in vivo, limb buds were dissected from WT embryos (E12.5), embedded in OCT and snap frozen. Sections of 20 μm were cut and fixed for 10 min in 4% (v/v) paraformaldehyde, 0.2% Triton X-100 in PBS. Sections were incubated in blocking buffer (10% horse serum in PBS) for 30 min and in acetylated α-tubulin overnight (Sigma Aldrich, T7451) at 4°C. After three washes in 0.2% Triton-X in PBS slides were incubated in secondary antibody for 1 h. Slides were washed three more times and mounted with Dako fluorescent mounting medium containing 2 μg/ml 4′,6-diamidino-2-phenylindole (DAPI) and sealed with nail polish. Slides were visualized on a Zeiss Axiovert 25 inverted phase contrast/fluorescent microscope, and images were taken using an AxioCam MRC camera (Zeiss) and Axiovision software.

Cell culture and confocal microscopy

MEFs were derived from E12.5 embryo limb buds using standard protocols as previously described (27). Cells were cultured in Dulbecco’s Modified Eagles medium containing 10% fetal calf serum for up to 10 passages.

For in vitro analysis of cilia, confluent MEF cultures were ‘serum starved’ for ~ 60 h to induce ciliogenesis. Cells grown on coverslips were fixed for 5 min in 2% (v/v) paraformaldehyde (PFA) in PBS followed by 1 min in 1:1 methanol:acetone mix at −20°C. Coverslips were blocked for 1 h in 10% horse serum in PBS and incubated with the appropriate primary antibody diluted in blocking solution overnight at 4°C. The following antibodies were used: Acetylated α-Tubulin (Cell Signaling Technology, 5335S), Acetylated α-Tubulin (Sigma-Aldrich, T7451), γ-Tubulin (Sigma-Aldrich, T5192) and IFT88 (Proteintech, 13 967–1-AP). Single-color immunofluorescent staining was visualized on a Zeiss Axiovert 25 inverted phase contrast/fluorescent microscope, and images were taken using an AxioCam MRC camera (Zeiss) and Axiovision software. Multi-color immunofluorescent staining was visualized on a Nikon A1-R confocal microscope (Nikon Corporation) and processed by NIS Elements AR (version 4.13) software (Nikon Corporation) for imaging and co-localization analysis. Cilia length was quantified using ImageJ (NIH Image).

For hedgehog pathway signaling analysis, confluent MEF cultures were serum starved for 48 h to induce ciliogenesis and treated with 100 nM SAG (Merck, 566 660) for 24 h. Cells were then processed for quantitative PCR.

Optical projection tomography

Staged embryos were stained for optical projection tomography (OPT) [Sharp 2002] with an antibody to E-cadherin (Abcam, AB-11512) as described (27), with 48 h primary and secondary antibody incubations interspersed with extensive 12 h washes to remove unbound antibody. Samples were imaged on a Bioptonics 3001 OPT machine (Bioptonics, UK), and datasets were reconstructed by NRecon (Skyscan, Belgium) and visualized using Drishti (http://anusf.anu.edu.au/Vizlab/drishti/).

Immunoblot and quantitative PCR analyses

Western blot analysis was performed as described (30) using antibodies against Actin (EMD Millipore/Merck, MAB1501; loading control), DYNLL1 (Abcam, ab51603), DYNC2LI1 (Proteintech, 15 949–1-AP), GLI3 (R&D Systems, AF3690), WDR34 (Sigma-Aldrich, HPA040764), horse-radish peroxidase-coupled secondary antibodies and ECL reagents (GE Healthcare). Western blots were quantified using ImageJ software (NIH, Bethesda). Total RNA was isolated using Isolate II RNA Micro Kit (Bioline) and reverse transcribed using High-Capacity cDNA Reverse Transcription kit (Applied Biosystems) following the manufacturer’s protocol. Real-time PCR was performed using AmpliTaq Gold DNA Polymerase (Applied Biosystems) and a LightCycler 480 (Roche). TaqMan primers for Dynll1, Dynll2, Gli1, Ptch1 and Actin were purchased from ThermoFisher Scientific. Ct values were normalized to Actin as the endogenous control and expressed as deltaCt (dCT).

Micro-computed tomography

Limbs from 8-week-old mice were dissected and fixed in 4% (v/v) PFA overnight at 4°C and transferred to 70% (v/v) EtOH until required for micro-CT. Micro-CT was performed on femora using the SkyScan 1076 system (Bruker-microCT, Kontich, Belgium). Images were acquired using the following settings: 9 μm voxel resolution, 0.5 mm aluminum filter, 50 kV voltage and 100 μA current, 2600 ms exposure time, 0.5 degree rotation, frame averaging = 1. Images were reconstructed and analyzed using NRecon (version 1.6.9.8), Dataviewer (version 1.4.4) and CT Analyser (version 1.11.8.0), as previously described (Ansari et al, JBMR 2018), with the following adaptations: femoral trabecular analysis was carried out at the distal femur in a region of 15% of the total femur length commencing at 7.5% of the total femur length toward the mid-shaft. Cortical analyses were performed in a region of 15% of the femoral length commencing from 30% proximal to the distal end of the femur. The lower thresholds used for trabecular and cortical analyses were equivalent to 0.20 and 0.642 g/mm³ calcium hydroxyapatite (CaHA), respectively.

Scanning electron microscopy

E10.5 embryos were dissected into cold PBS, and segments of the neural tube at the level of the forelimb were split in two using 30 gauge needles to reveal the inner face of each half. Samples were fixed in 2.5% gluteraldehyde in 0.1 M sodium cacodylate for 1 h, washed 3 × 10′ in 0.1 M sodium cacodylate and then incubated in 1% osmium tetroxide in 0.1 M sodium cacodylate for 1 h. Following washing in MilliQ water, samples were dehydrated in an ethanol series and then transferred to hexamethyldisilazane (HMDS) through an HMDS/ethanol series. Samples were then air dried, mounted on stubs and gold coated for 70 s. Microscopy was performed using a Nova NanoSEM (FEI/Thermo Fisher).

Yeast two-hybrid assays

Human WDR34 and WDR60 cDNAs were cloned into the yeast expression vector pGADGH (Clontech), and DYNLL1 was cloned into pGBT9 (Clontech). Empty vectors were used as negative controls and pGADGH-ASCIZ was used as a positive control for yeast two-hybrid assays in the Saccharomyces cerevisiae strain PJ69-4A as described (41).

Statistical analysis

Statistical analysis was performed using GraphPad Prism software (San Diego). If not indicated otherwise, P values were calculated by two-tailed unpaired Student’s t-test or in the case of SAG treatment experiments two-tailed paired Student’s t-test. Error bars represent the mean ± SEM. Numbers of mice per analysis and numbers of independent experiments are indicated in the figures or legends.

Acknowledgements

The authors thank Nora Tenis for genotyping and helping with mouse colony management, the St. Vincent’s Hospital Melbourne Biomedical Resource Centre for animal care, Carol Wicking for reagents and advice and Louise Purton for Prx1-Cre mice. The authors acknowledge use of the facilities and the assistance of Joan Clark at the Monash Ramaciotti Centre for Cryo Electron Microscopy.

Conflict of Interest statement. None declared.

Funding

National Health and Medical Research Council of Australia (NHMRC) (Senior Research Fellowship, GNT1022469, to J.H.; GNT1081242 to N.A.S.; Senior Research Fellowship, GNT1106516, and GNT1082051 and GNT1098654 to I.M.S.); the 5-point Foundation (to J.H.); the Margaret Walkom Bequest (to J.H.); an anonymous philantropic foundation (to J.H.); Australian Postgraduate Award to A.K.; SVU Brenda Shanahan Fellowship (to N.A.S.); Australian Research Council (DP160103100 to I.M.S.); and NHMRC Independent Research Institutes Infrastructure Support and Victorian State Government Operational Infrastructure Support grants.