ZMYND10 functions in a chaperone relay during axonemal dynein assembly

Molecular chaperones promote the folding and macromolecular assembly of a diverse set of ‘client’ proteins. How ubiquitous chaperone machineries direct their activities towards specific sets of substrates is unclear. Through the use of mouse genetics, imaging and quantitative proteomics we uncover that ZMYND10 is a novel co-chaperone that confers specificity for the FKBP8-HSP90 chaperone complex towards axonemal dynein clients required for cilia motility. Loss of ZMYND10 perturbs the chaperoning of axonemal dynein heavy chains, triggering broader degradation of dynein motor subunits. We show that pharmacological inhibition of FKBP8 phenocopies dynein motor instability associated with the loss of ZMYND10 in airway cells and that human disease-causing variants of ZMYND10 disrupt its ability to act as an FKBP8-HSP90 co-chaperone. Our study indicates that primary ciliary dyskinesia (PCD), caused by mutations in dynein assembly factors disrupting cytoplasmic pre-assembly of axonemal dynein motors, should be considered a cell-type specific protein-misfolding disease.

32 Molecular chaperones promote the folding and macromolecular assembly of a diverse set of substrate 33 'client' proteins. How the ubiquitous chaperone machinery directs its activities towards a specific set 34 of substrates and whether this selectivity could be targeted for therapeutic intervention is of intense 35 research. Through the use of mouse genetics, imaging and quantitative proteomics we uncover that 36 ZMYND10 is a novel co-chaperone for the FKBP8-HSP90 chaperone complex during the 37 biosynthesis of axonemal dynein heavy chains required for cilia motility. In the absence of 38 ZMYND10, defects in dynein heavy chains trigger broader dynein motor degradation. We show that 39 FKBP8 inhibition phenocopies dynein motor instability in airway cells, and human disease-causing 40 variants of ZMYND10 disrupt its ability to act as FKBP8-HSP90 co-chaperone. Our study indicates 41 that the motile ciliopathy Primary Ciliary Dyskinesia (PCD) should be considered a cell-type specific 42 protein-misfolding disease and opens the potential for rational drug design that could restore 43 specificity to the ubiquitous chaperone apparatus towards dynein subunits.

61
Several PCD causing mutations are also found in a newly discovered set of genes, the "dynein 62 axonemal assembly factors" (DNAAFs), whose functions are poorly understood. DNAAFs are 63 proposed to assist Heat Shock Protein (HSP) chaperones to promote subunit folding and cytoplasmic 64 pre-assembly of dynein motors. DNAAFs are presumed to act as cilial-specific co-chaperones based 4 assembly, their precise molecular functions within the pre-assembly pathway still remain largely 81 unknown.

82
Previous studies had established a strong genetic link between loss of ZMYND10 and perturbations in 83 dynein pre-assembly 16 , however the putative molecular role of ZMYND10 as a DNAAF in this 84 process remains unclear. In order to systematically probe the mammalian dynein pre-assembly 85 pathway in greater molecular and cellular detail, we generated Zmynd10 null mice by CRISPR gene 86 editing. We used different motile ciliated lineages at distinct stages of differentiation from our 87 mammalian mutant model to pinpoint the precise stage at which ZMYND10 functions during dynein 88 pre-assembly.

89
Our protein interaction studies implicate a novel chaperone complex comprising of ZMYND10, 90 FKBP8 and HSP90 in the maturation of dynein HC clients. We postulate that a chaperone-relay

123
We analyzed expression of ODA components in different postnatal tissues by immunofluorescence 124 and immunoblotting from Zmynd10 mutants to assess whether ZMYND10 loss impacts ODA levels.

125
In postnatal trachea and oviducts, total levels of the ODA HCs DNAH9 and DNAH5 were reduced by 126 immunoblot (Figure 2A

133
We previously reported distinct morphological and dynein staining characteristics observed among 135 human multiciliated nasal brush biopsied cells depending on their maturity 11 . In cells isolated from 136 nasal turbinates of control mice, 'immature' multiciliated cells were rounder with higher cytoplasmic 137 dynein immunostaining and shorter cilia in keeping with cytoplasmic pre-assembly of the motility 138 machinery. In contrast, 'mature' cells have long, organized arrays of cilia intensely stained for ODA 139 subunits with a clear lack of cytoplasmic signal, suggesting that the motility machinery has 140 translocated into cilia and stably integrated into the axonemal ultrastructure ( Figure 3A,B upper 141 panels). In Zmynd10 mutants, no defects in ciliary length or number were observed ( Figure S3) 142 however outer or inner arm dyneins fail to incorporate into mature cilia axonemes. Importantly, no 143 cytoplasmic accumulations were noted in 'mature' ciliated cells (Figure 3A lower

151
As cytoplasmic DNAI2 and DNAH5 were detected in Zmynd10 -/-immature respiratory cells, 152 suggesting that they were initially synthesized, we sought to verify if they were assembled into 153 complexes using the in situ proximity ligation assay (PLA). In control immature human nasal brush 154 epithelial cells, we detected PLA signals consistent with DNAI2 and DNAH5 existing in both 155 cytoplasmic and axonemal complexes ( Figure 4A). However, we detected a highly reduced number 156 of PLA positive foci in nasal epithelial cells of P7 Zmynd10 -/-mice, with complexes restricted entirely 157 to the cytoplasm in contrast to the strong axonemal staining observed in similarly staged controls 158 ( Figure 4B). To directly examine the interactions between ODA IC and HC subunits, we 159 immunoprecipitated endogenous DNAI2 (IC2) from postnatal tracheal (P26) and oviduct (P7) extracts 160 from Zmynd10 -/-animals. DNAI2 co-precipitated DNAI1 (IC1) at similar levels from both wild type 161 and mutant P26 tracheal extracts ( Figure 4C). This indicated that loss of ZMYND10 does not 162 primarily impact IC subunit heterodimerization or stability during the assembly process. Importantly, 163 we observed significantly reduced co-immunoprecipitation of DNAH5 by DNAI2 in P7 oviduct 164 mutant extracts (mutant 0.55 vs wild type 1.1, normalized to total levels, Figure 4D,E). Moreover, we 165 observed similar degradative bands (arrowheads) for DNAH5 in the mutant samples indicating that 166 any DNAH5 that is incorporated may be poorly folded and unstable, in the absence of ZMYND10.

167
We hypothesize that this reduced association between the two subunits is due to the HC subunit being 168 in an assembly incompetent, unstable state such that any substandard complex would be targeted for 169 subsequent degradation ( Figure 4F). 170 7 ZMYND10 loss also leads to absent IDA motors from human, fly and mouse cilia 16 . To bypass the 172 limitation of robust immunoreagents for IDA detection, we used label-free quantitative proteomics 173 comparing postnatal testes extracts from P25 control and Zmynd10 mutant littermates (Figure 5A-C).

174
We hypothesized that synchronized spermiogenesis and flagellar extension at this stage would 175 correspond with cytoplasmic pre-assembly of flagellar precursors. Whilst protein expression profiles 176 were not different between mutant and controls for differentiation, meiosis and cell death markers 177 (Table S2) Figure S4). Next, we investigated whether ZMYND10 associates with chaperones 193 or other dynein assembly factors (Figure 6A), namely DNAAF2 which was shown to interact with 194 DNAAF4, which also co-precipitates several chaperone proteins including the CCT3 subunit of the 195 TriC chaperonin complex 8 as well as the non-canonical co-chaperone DNAAF5, whose HEAT 196 repeats have been proposed to function as scaffolds in multisubunit assembly 11,19 . We failed to detect 197 interactions with HSP70, DNAAF2, CCT3 or DNAAF5 by endogenous ZMYND10 affinity 198 purification from mouse testes or oviduct extracts (Figure 6A,C). Our findings suggest that 199 ZMYND10 functions at a stage of dynein assembly distinct from these previously described co-200 chaperone complexes. We were also unable to co-immunoprecipitate endogenous LRRC6, a protein 201 that has been previously shown to associate with ZMYND10 by over-expression studies 16 . Absence 202 of strong association was confirmed by a reciprocal pull-down using endogenous LRRC6 as bait. We 203 conclude that any interactions may be highly transient in vivo (Figure 6B).

205
We detected a specific interaction of ZMYND10 with HSP90, another major cytosolic chaperone 206 implicated in dynein pre-assembly ( Figure 6C). Our preliminary affinity purification mass-207 spectrometry (AP-MS) analyses to generate a ZMYND10 interactome ( Figure 7A, Table S1) also 208 revealed that ZMYND10 consistently co-precipitated the well characterised HSP90 co-chaperone 8 FKBP8 (Immunophilin FK506-binding protein (FKBP) family member; Uniprot: O35465), 20 as well To directly test for ZMYND10 dependent chaperone-client associations, we immunoprecipitated 237 endogenous FKBP8 and ZMYND10 from differentiating human tracheal epithelial cells (D17) and 238 oviducts (P7); both these stages correspond to active cytoplasmic dynein pre-assembly. We detected 239 associations of the client protein DNAH5 with both ZMYND10 and FKBP8 ( Figure 7G). In the case 240 of ZMYND10, it associated with HSP90 but not with HSP70 at a similar stage of cellular 241 differentiation ( Figure 6C). These findings show an association between a mammalian dynein 242 assembly factor and a client dynein heavy chain in vivo. Additionally, taking step-wise ODA 243 macromolecular assembly into account, we found that whilst DNAI1 co-immunoprecipitated 244 DNAH5, it did not immunoprecipitate either ZMYND10 or FKBP8 ( Figure 7G). This suggests that 245 the DNAI1-DNAH5 interaction occurs in a complex that is distinct and downstream from the FKBP8-9 DNAH5-ZMYND10 complex, as supported by our DM-CHX experiments in which both subunits are 247 destabilized after 24 hours drug treatment.

249
In summary, we propose that a complex or series of complexes of the cilial-specific ZMYND10 250 adaptor, the ubiquitous FKBP8 co-chaperone and chaperone HSP90 direct their activities towards 251 client axonemal dynein heavy chains including DNAH5 to promote their maturation and stability.

252
This 'folding/maturation' step is essential for the HCs to subsequently form strong associations with 253 other dynein subunits as observed between DNAH5 and the IC1-IC2 heterodimer and for productive 254 assembly to proceed.  (Table S1). As a filtering strategy to find 'true' interactions, we