(A) Schematic illustrating the null allele generated by a −7 bp CRISPR deletion in Zmynd10 exon 6. (B) Immunoblots from testes extracts from postnatal day 26 (P26) control and mutant male mice show loss of ZMYND10. (C, D) Immunostaining for ZMYND10 reveals a complete loss of signal in multiciliated ependymal cells (C) and lung cryo-sections (D). Multicilia are marked with acetylated α- tubulin (C). (E) Neonatal Zmynd10 mutants display hydrocephaly; the white arrowhead points to doming of the head. See also . (F) TEM of tracheal ciliary axoneme cross-sections shows a lack of axonemal outer (ODA: white arrowhead) and inner (IDA: black arrowhead) dynein arms in mutants. (G) Representative image of a gross dissection of lungs shows situs inversus totalis in mutants. See also . (H) H&E staining of coronal sections of nasal turbinates reveals mucopurulent plugs in mutants. Scale bars in (C) = 5 μm, in (D) = 100 μm, in (F) = 100 nm.

(A) Schematic of Zmynd10 mouse locus (ENSEMBL: ENSMUSG00000010044) and guide design targeting critical exon six which were cloned into px330 (Addgene:#42230) (). (B) Validation of activities of gRNAs to generate double-stranded breaks were initially tested in HEK293 cells by using a reporter to assay reconstitution of EGFP upon cleavage (Addgene: #50716) (). (C) Table summarizing pronuclear microinjection rounds for generating Zmynd10 mutants.

(A) Representative image of a testis from Zmynd10 mutant and wild type control (P150). (B) As a result of defects in flagellar development, mutants have smaller testes (mean ± s.e.m, n = 3 males/genotype). Zmynd10 mutants have lower sperm density (mean ± s.e.m, n = 3 males/genotype). (C) Zmynd10 mutant sperm have reduced motility. Arrows point to static/paralyzed sperm in mutants. Scale bars = 50 μm. (D) By immunofluorescence, staining of DNAI2 and DNALI1 is absent along Zmynd10^-/- flagella Arrowhead in wild type point to the annulus region and flagellar tip. In the mutant, the arrowhead points to the annulus. (E) Coronal vibratome sections of brains show Zmynd10 mutants display dilated lateral ventricles (arrowhead) due to hydrocephaly. (F) Zmynd10 mutants display heterotaxy defects. Arrowheads denote direction of the heart’s position, which is reversed in the mutants; numbers denote lobes of the lungs (1,2 and 3 correspond to upper, mid and lower lobes of the right lung; 4 and 5 correspond to the upper and lower lobes of the left lung).

(A) Transmission electron micrographs (TEM) of tracheal multiciliated cells from wild type and mutant female mice (P9). Arrows point to cilia on luminal surface of the cells; basal bodies appear correctly docked in the mutant cells similar to controls. Scale bar = 2 μm. (B) Immunostaining for cilia (acetylated α-tubulin) and apical F-actin (phalloidin) reveals no cytoskeletal gross defects in Zmynd10 mutants. Scale bars = 10 μm.

Immunoblots of whole protein extracts from trachea P26 (A) and P30 oviducts (B) show reduced abundance of ODA subunits HC-γ, HC-β (DNAH5 and DNAH9), IC1 and IC2 (DNAI1 and DNAI2). Immunofluorescence of trachea (C) and oviduct (D) tissue sections show loss of axonemal DNAH5 and DNAI2 staining as well as reduced total abundance in Zmynd10 mutants compared to controls. Brightfield is included in oviduct merge panel (D) to highlight absence of staining in cilia in Zmynd10 mutants. Scale bars in (C) and (D) = 50 μm. (E) No significant changes are detected in levels of dynein transcripts by quantitative RT-PCR of (Dnah5, Dnali1) normalized to (Tbp) in P12 Zmynd10 mutant oviducts (n = 3/genotype, dark grey Zmynd10 mutants). (F) During early motile ciliogenesis, mildly reduced levels and laddering consistent with degradative, misfolded intermediates (arrowheads) of ODA HC-γ DNAH5 are detected in Zmynd10 mutant oviducts (P7). These will be subsequently cleared as tissue differentiation proceeds.

(A) Z-projection through healthy human donor nasal brush immunofluorescence shows a mix of ‘mature’ motile ciliated cells (lower inset, arrow) with exclusively axonemal staining of dynein subunits (DNAI2 green, DNALI1 magenta) and strong foci of SENTAN (red) at cilial tips, as well as ‘immature’ cells having cytoplasmic staining of dynein subunits and no SENTAN (right inset, arrowheads). Scale bar = 10 μm. (B) Nasal brush immunofluorescence from Zmynd10 mice shows components of outer arm dyneins (DNAH5, DNAI2) are initially expressed in apical cytoplasm of immature mutant cells (arrowheads) but subsequently undergo ‘clearance’ in mature cells (lower panels, arrows), whilst all complexes exclusively translocate into cilia in control mature cells (middle panels, arrows). (C) Inner arm dynein component DNALI1 is expressed and apically arrested in immature mutant cells (arrowhead) and unlike controls, will get subsequentlycleared along with the ODA IC, DNAI2, in mature mutant cells. Scale bars = 5 μm.

(A,B) Z-projections of Proximity Ligation Assay (PLA) on human donor (A) or mouse P7 (B) nasal brush biopsies confirming ODA subunits (mouse IC2/DNAI2 and HC-γ/DNAH5) are pre-assembled in the cytoplasm of mammalian multiciliated cells. Control (antibody only control, Ab control) sections incubated with only DNAH5 show no PLA signal. Red spots denote individual ODA complexes (<40 nm) that appear as peri-nuclear foci in immature cells (arrowhead) and translocate to cilia in mature cells (arrow). (B) In Zmynd10 mutant cells, reduced number of foci are observed and restricted only to the cytoplasm, highlighting defects in cytoplasmic pre-assembly. GRP94 was used as a pan-cytosolic marker (A) or phalloidin for apical actin ring (B). Nuclei are stained with DAPI (blue). See also for mouse IC2/DNAI2 and HC-β/DNAH9 PLA. Scale bars in (A) = 5 μm and (B) = 10 μm. (C) Endogenous Immunoprecipitation of DNAI2 from P26 Testes extracts reveals no defects in DNAI2 association with its heterodimeric partner DNAI1 in Zmynd10 mutants. (D) Endogenous Immunoprecipitation of DNAI2 from P7 oviduct extracts show disruption in subsequent association between DNAI2 and DNAH5 in mutants compared to controls as quantified by intensities of the DNAH5 pull-down bands (E). Arrowheads show predicted degradative or misfolded intermediates of DNAH5 polypeptide in the mutants only. Numbers to the right of panels denote protein molecular weight in kDa. See also for analysis in P7 and P90 trachea. (E) Ratio of mutant versus wild type fold enrichment (from 4D, : IP/input) for DNAI2 and DNAH5, normalized for differences in stability in input, as well as amount of DNAH5/DNAI2 complexes. (F) Schematic of axonemal ODA showing the intermediate chain heterodimers (IC) bind normally to heavy chains (HC) to form the entire motor complex in controls (green) and that this association is perturbed in mutants (red).

(A) Z-projections of proximity ligation assay (PLA) on mouse P7 nasal brush biopsies confirming that ODA subunits (mouse IC2/DNAI2 and HC-β/DNAH9) are pre-assembled in the cytoplasm. Control (antibody only control, Ab control) sections were incubated with only DNAH9 antibody and show no PLA signal. Red spots denote individual ODA complexes (<40 nm) in control cells and cilia. In Zmynd10 mutant cells, reduced number of foci are observed and restricted only to the cytoplasm, highlighting defects in cytoplasmic assembly. Phalloidin was used to mark the apical actin ring. Nuclei are stained with DAPI (blue). Scale bar = 10 μm. (B, C) Immunoprecipitation of DNAI2 from P7 trachea extracts show disruption in subsequent association between DNAI2 and DNAH5 in mutants compared to controls, as quantified by intensities of the DNAH5 pull-down bands (). In contrast, in P90 adult Zmynd10 mutant trachea (C), levels of DNAI2 are profoundly affected (input), and negligible complexes can be purified.

(A) Immunofluorescence of ZMYND10 in control and mutant adult testes (P150) in asynchronous seminiferous tubules (arrowhead). (A) ZMYND10 is strongly expressed in primary spermatocytes and spermatids, where it is restricted to the cytoplasm and never in developing sperm tails. This staining is lost in mutants. Nuclei are stained with DAPI. Scale bar = 100 μm. (B) Cross-sections of similarly staged seminiferous tubules reveal similar developmental staging of sperm between adult control and mutants, but loss of DNAI2 (ODA) and DNALI1 (IDA) proteins from cytoplasm and axonemes (arrowhead) of mutant sperm. Nuclei are stained with DAPI. Scale bars = 50, 100 μm. (C) Schematic summarizing mouse spermatogenesis which is initially synchronized postnatally (stages shown to right), before occurring in asynchronous waves across seminiferous tubules. Axonemal dynein pre-assembly in the cytoplasm occurs from spermatid stage around P25 and continues till flagellogenesis at P30. (D) Unbiased quantitative proteomics of control and mutant testes (elongated spermatid: P25) reveal that loss of ZMYND10 leads to a primary reduction in abundance of all dynein HC subunits during cytoplasmic pre-assembly, whilst other components remain initially unaffected at this stage. Yellow diamonds highlight significantly (p<0.05) changed hits based on LFQ intensity between Zmynd10 mutants and wild type littermates (n = 3/genotype). Schematic below highlights fold change of specific subunits on given dynein arms. Some heavy chains had specific multiple isoforms detected (DNAH7b/c and DNAH12 (E9QPU2*, F6QA95, Q3V0Q1†). See also . Key: IAD: inner arm dynein; OAD: outer arm dynein; DNAAFs: dynein axonemal assembly factors; RSP: radial spoke protein; DRC: dynein regulatory complex.

Intensities of peptides detected by mass spectrometry from P25 testes whole proteome analysis () are plotted across the polypeptide lengths for individual axonemal dynein subunits (ODA and IDA-HCs, ODA-ICs, ODA-LC). TTC25 (ODA-docking complex) and cytoplasmic dynein 1 and 2 heavy chain polypeptides are used as controls. Axonemal dynein heavy chains specifically show reduced peptide intensities in Zmynd10 mutant testes samples (red) compared to controls (blue). All the other proteins tested have comparable peptide intensities between mutant and control samples.

(A) Summary schematic of affinity purification-mass spectrometry (AP-MS) analysis of endogenous ZMYND10 interactomes from P30 mouse testes, overlapping between two commercial polyclonal Sigma (S) and Proteintech (P) antibodies. See also . (B) Endogenous ZMYND10 affinity purification with two validated ZMYND10 antibodies from Zmynd10 control and mutant P30 testes extracts confirm the ZMYND10 interaction with FKBP8 and HSP90AB1 (S only) in control samples. These interactions are not found in mutant samples lacking ZMYND10, serving as specificity controls. An in vivo interaction between ZMYND10 and LRRC6, as well as with DNAAF1, was not detected in reciprocal endogenous immunoprecipitations in P30 control testes extracts, using rIgG as a control. (C) Endogenous FKBP8 and DNAI1 immunoprecipitations in differentiating healthy human donor tracheal epithelial cultures (D17 ALI) both show binding of client DNAH5, the rest of the complexes are distinct, suggesting they act at sequential steps of assembly. (D) Endogenous ZMYND10 immunoprecipitation of client DNAH5 and chaperone HSP90, but not HSP70 from differentiating oviduct epithelial tissue (P7) using rIgG as a control. Protein molecular weights displayed in KDa.

(A) Summary of mutations generated in ZMYND10-tGFP by site-directed mutagenesis to disrupt the binding interface for FKBP8, including the W423A mutation predicted to functionally disrupt one of two Zn²⁺-fingers in the MYND domain and the T379C PCD patient mutation (), lying just before the MYND domain. (B) Extracts from transiently transfected HEK293 cells affinity purified against turboGFP shows C-terminal mutations interrupt endogenous FKBP8 and HSP90 binding to ZMYND10. (C) Healthy human donor tracheal epithelial cultures (MucilAir) before ciliation (D17, post-ALI) or fully ciliated (D60, post-ALI), were cultured for 24 hr in control (vehicle only) or DM-CHX (concentrations indicated in μM) before harvesting protein extracts. Immature cultures (D17) were more sensitive to effects of specific PPIase inhibitor DM-CHX, destabilizing dynein components, whilst mature cultures were minimally affected. (D) Quantification of band intensities for DNAH5 (blue) or DNAI1 (yellow) from (C) were normalized to loading control, and plotted as a fold change after 24 hr. (E) Z projections of whole-mount immunofluorescence of mouse tracheal epithelial cultures (MTECs) stained with DNAH5 (magenta), DNAI2 (green) and DAPI (blue) after treatment of three days post-airlift (ALI) with either 100 μm DM-CHX or vehicle control for 14 days in culture to differentiate, followed by switching cultures from control to 100 μm DM-CHX (mature: cilial beat visualized) or from DM-CHX into vehicle control (no cilia beat) and culturing an additional 7 days. Treatment of mature ciliated cultures had little effect on DNAH5 and DNAI2 levels, whilst treatment during differentiation disrupted expression of dynein subunits, as evidenced by lack of axonemal staining in most cells (right panels) but release allows recovery of dynein pre-assembly, mostly cytoplasmic after 7 days.

(A) Extracts from transiently transfected RPE-1 cells with ZMYND10-turboGFP variants and human LRRC6-myc were affinity purified against turboGFP (upper panels). Inclusion of LRRC6 destabilizes binding of wild type ZMYND10 to endogenous FKBP8. C-terminal ZMYND10 mutations do not affect its interaction with LRRC6. Expression of interactors is confirmed in input cell lysates (lower panels). (B) During dynein arm assembly in the cytoplasm, ZMYND10 interaction with co-chaperone PPIase FKBP8 and HSP90 is required for the stabilization and folding of dynein heavy chains (DNAH). Competitive binding of LRRC6 to ZMYND10 may advance the HSP90-FKBP8 chaperone cycle taking the client DNAHs to the next stage of assembly. We propose either R2TP or a specialized R2TP-like complex may function in parallel to assemble, scaffold or tether the DNAI1-DNAI2 heterodimeric complex via the PIH-domain proteins, such as DNAAF6, DNAAF2 and PIH1D2. The TPR-domain containing SPAG1 or DNAAF4 could recruit HSP90 via its MEEVD domain and together with RUVBL1/RUVBL2 these assembly factors could form a stable platform to promote HC-IC subunit interactions. This R2TP or R2TP-like complex likely operates distinctly to ZMYND10 as DNAI1/2 heterodimers are detected in Zmynd10 mutants, however, these are likely degraded if fully functional mature complexes (with heavy chains) cannot be assembled. This chaperone-relay system comprises several discrete chaperone complexes overseeing the folding and stability of discrete dynein subunits. Folding intermediates are handed off to successive complexes to promote stable interactions between subunits all the while preventing spurious interactions. Stable dynein complexes, once formed are targeted to cilia via transport adaptors and intraflagellar transport (IFT).

We agree on the importance of clarifying this point of initial attempts at assembly, followed by clearance when this fails in mutants. Figures 3 and 4 are now better annotated and a more detailed description in the figure legends is provided. We have also used an independent marker for motile ciliary maturation, SENTAN (Kubo et al., 2008), which is enriched in the apical structure of mature motile ciliary tips. Using two commercial SENTAN antibodies on healthy human donor nasal brushings (here, new Figure 3A), we have demonstrated that the immunostaining pattern for two axonemal dynein subunits DNAI2 and DNALI1, transitions from a predominantly cytoplasmic (SENTAN negative, immature cell stage) to exclusively ciliary (SENTAN positive, mature cell stage) staining during motile ciliated cell differentiation for. We received a small aliquot of un-purified serum against mouse SENTAN (Kubo et al., 2008), which also showed enrichment in cilial tips but less precisely during differentiation (), as dyneins progressively translocate from the cytoplasm to cilia. We have not included the mouse data in the paper (see , right panel) as it requires much more optimization.

We appreciate the reviewer’s keen eye but note that the perceived ‘presence’ of DNAH5 in cilia axonemes of immature mutant cells is likely due to the images being acquired with increased pixel dwell time to better visualize the complete absence of DNAH5 signal in neighboring mature mutant cells, further highlighting the robust clearance of DNAH5. We have amended this panel to clarify the absence of DNAH5 signal in mutant cilia. We reiterate that we never observe DNAH5 staining in the ciliary compartment in Zmynd10 mutant cells. The representative images acquired at comparable imaging conditions are provided to additionally emphasize this point ().

We have significantly overhauled Figures 6 and 7, as well as added Figure 4—figure supplement 1 and changed Figure 6—figure supplement 1 to remove ‘negative IP data’, as we concur with the reviewers that these types of studies have limited insight. We refer to the comprehensive DNAAF4 IP-MS dataset generated by Tarkar et al., 2013, which detects DNAAF2 but not ZMYND10 as an interaction partner. More recently, Paff et al., 2017, reported that PIH1D3 interacts with DNAAF2 and DNAAF4 in vitro, suggestive of a ternary complex between the three in vivo. Their interaction studies also failed to detect ZMYND10 as an interaction partner. Taken together, recent findings suggest that ZMYND10 functions separate to the other DNAAFs. This is why the main focus of this study was the discovery of a novel functional chaperone complex comprising of ZMYND10, FKBP8 and HSP90, the existence of which we have validated in multiple ways. The interaction with LRRC6 is interesting. As shown in our initial submission, we failed to detect ‘endogenous’ interaction between ZMYND10 and LRRC6 by reciprocal IPs during cytoplasmic dynein pre-assembly (now new Figure 6B). Both Zariwala et al., 2013 and Moore et al., 2013 had shown interaction only with overexpression systems. However, LRRC6 is the only ‘DNAAF’ to be significantly down-regulated in our P25 Zmynd10 mutant testes MS dataset (see Figure 5D), suggesting some type of functional interaction exists. In response to the reviewers' comments, we have now tried a commercially available LRRC6 antibody from Novus (NBP1-82816, as used in the Zariwala paper) but in our hands this antibody fails to recognise endogenous mouse LRRC6 and no enrichment is seen in ZMYND10 pull downs, although a clear interaction with FKBP8 is detected (see ).

We also tested the custom mouse LRRC6 rabbit polyclonal that was a gift from Professor Hiroshi Hamada (used in this paper and panels shown in ). Although testes at different developmental stages were used (P24 with synchronised seminiferous tubules and P60 with asynchronised seminiferous tubules), we failed to observe a consistent interaction between ZMYND10 and LRRC6. Instead bands of different molecular weight were enriched in each of the two cases where a potential co-purification was observed (, asterisks). All the while the interaction between FKBP8-HSP90-ZMYND10 is maintained. This suggests that any interaction between these two DNAAFs may be extremely transient in vivo.

The reviewer raises a very interesting point, about the specificity of FKBP8 inhibition on stability of a single ‘critical’ component as opposed to a broader effect on several dynein components during cytoplasmic pre-assembly. Our model would have predicted the former, where only stability of HC subunits would be primarily affected. However, we also observed effects on DNAI1 and DNAI2 (). In our 24-hour treatment window, it is possible that DNAI1 and DNAI2 are destabilized secondarily, and as these proteins are much smaller compared to the ~500KDa HCs, we would expect to see faster turnover. Regardless, we show that inhibiting the PPIase activity of FKBP8 phenocopies ZMYND10 loss of function in terms of ODA subunit stability.