All animal experiments were approved by the Animal Research Committees of Kyoto University, Osaka Metropolitan University, and Tokyo University of Agriculture and were conducted according to their regulations on animal experimentation.

F344-kk/kk rats were obtained from the National BioResource Project for the Rat (NBRP-Rat, Kyoto, Japan) (Serikawa et al., 2009). BN/SsNSlc and F344/NSlc rats were purchased from Japan SLC, Inc., (Hamamatsu, Shizuoka, Japan).

Lengths and widths of gaits in kk-homozygous (n = 4) and wild-type (WT) (n = 4) female rats were examined at 9 weeks of age (Teunissen et al., 2001; Nishitani et al., 2020). Foot pads of the hind paws were immersed in the black ink and the rats were placed on the white absorbing paper (12 cm × 100 cm). Step lengths were measured for the right and left legs. Step widths were measured for every successive step.

Rats (3, 5, 7, 8, 9, and 10 weeks of age) were euthanized under isoflurane anesthesia. We examined a total 112 rats for histopathology (homozygous; n = 79 and WT; n = 33). Tissue samples from the central nervous system (CNS) were fixed in 10% neutral buffered formalin, and embedded in paraffin. Sections of 4 μm thick were cut and stained with hematoxylin and eosin (HE). We also counted the number of spheroids (exceedingly 5 μm in diameter) in the transverse sections of the cervical (C5 level) and lumber (L1–2 level) spinal cord at 10 weeks of age by microscopic observation. The areas of 2.37 mm² (high-power fileld) in the dorsal cord of the spinal cord from five different animals were evaluated in each experimental group. The data are presented as the number of spheroids/mm². For transmission electron microscopy (TEM), two formalin-fixed tissues of kk-homozygous rats at 5 and 10 weeks of age were stored in 2.5% glutaraldehyde in 0.1 M phosphate buffer (pH 7.4), post-fixed with 1% osmium tetraoxide at 4°C overnight and embedded in epoxy resin. Ultrathin sections were stained with uranyl acetate and lead citrate and examined with a Hitachi H-7500 electron microscope (Hitachi, Tokyo, Japan).

We conducted immunohistochemistry (IHC) using formalin-fixed paraffin sections of the lumbar spinal cord from 10 weeks of age. For IHC, we used the primary antibodies listed in Supplementary Table S1. After dewaxing and pretreatment, immunohistochemical staining was performed using a HISTOSTAINER 36A (Nichirei Biosciences, Tokyo, Japan). Sections were treated with 5% skimmed milk in phosphate-buffered saline for 15 min and reacted with each primary antibody for 1 h at room temperature. After incubation in 3% H₂O₂ for 15 min, application of horseradish peroxidase-conjugated secondary antibodies (Histofine Simple Stain MAX PO; Nichirei Biosciences) was performed for 1 h. Signals were visualized with 3,3′-diaminobenzidine (DAB Substrate Kit; Nichirei Biosciences).

Because kk/kk homozygous rats died before sexual maturity, we used kk/+ heterozygous rats for breeding. We intercrossed kk/+ heterozygous (BN/SsNSlc × F344-kk/+)F1 rats to produce F2 intercross or backcrossed to F344-kk/+ rats to produce backcross progeny. Genotyping for the kk locus was performed by observation of rats exhibiting abnormal gaits and wasting postures by 10 weeks of age. Only kk/kk homozygous rats were used for genetic mapping of the kk gene. To localize the kk locus to a specific chromosomal region, we performed genome-wide scanning on DNA samples from 22 kk/kk homozygous rats using a panel of 106 simple sequence length polymorphism (SSLP) markers that covered all autosomal chromosomes (Chrs) (Supplementary Figure S1). To narrow down the kk locus, we additionally used a SSLP marker (D8Rat46) and a SNP marker (J683662). Genotyping of the SNP marker was performed by direct sequencing of PCR products amplified with the following primer set; 5′-AGGCTCCTGAGCAAGTTCAA-3′ and 5′-TGCAGTCCTAGGTATCCCTTT-3′.

Total RNA was isolated from brains using ISOGEN II (Nippon Gene, Tokyo, Japan). reverse transcription–polymerase chain reaction (PCR) and direct sequencing of PCR products were carried out as described previously (Kuramoto et al., 2011). Details of primers are listed in Supplementary Table S2. The PCR products overlapped each other and spanned the entire coding sequence of the rat Hspa8 gene. All sequencing was performed by Macrogen Japan Corporation (Kyoto, Japan).

Genotyping of the T to A substation of the Hspa8 gene was performed by the Amp-FTA method (Nakanishi et al., 2009). Templates were prepared on the FTA card and amplified with the following primer set: 5′-ATTAAATATGGGACATTGCTTC-3′ and 5′-CCTTTGTATTCGACTTGGAC-3′. The substitution was detected by Cycleave PCR™ Assay (Takara Bio Inc., Kusatsu, Shiga, Japan), in which fluorescence-labeled DNA–RNA chimeric probes were used. Sequences of the probes were as follows: 5′-ATGGTGG(rA)GA-3′ for the mutant allele and 5′-TGGTGGT(rG)AA-3′ for the WT allele.

Genome editing by CRISPR-Cas was performed as described previously (Yoshimi et al., 2016). Guide RNAs (gRNAs) were designed by Optimized CRISPR Design (crispr.mit.edu) and synthesized by Integrated DNA Technologies, Inc., (Coralville, IA, United States). Long ssODNs (lsODNs) were prepared using a LsODN Preparation Kit (Biodynamics Laboratory Inc., Tokyo, Japan). Cas9 protein was purchased from Integrated DNA Technologies. Pronuclear-stage embryos of F344/Jcl (CLEA Japan, Inc., Tokyo, Japan) rats were produced by natural mating. The oviducts of female rats with vaginal plugs were removed after euthanasia by CO₂ and cervical dislocation, and embryos were flushed out from the ampullae with culture medium. Cas9 protein, gRNAs, and lsODN were introduced into the embryos using a super electroporator NEPA 21 (NEPA GENE Co., Ltd., Ichikawa, Chiba, Japan). Embryos that developed to the two-cell stage were transferred into the oviducts of pseudopregnant females that were anesthetized using isoflurane. Offspring were genotyped by the Amp-FTA method with the following primer set: 5′-CGGGTTTCAGAGATGGAAGA-3′ and 5′-ATTTTCATTGACAGGTCCGG-3′ and Ampdirect Plus buffer (Shimadzu Corporation, Kyoto, Japan). Founder rats were mated with F344/Jcl rats and F1 heterozygous rats were intercrossed to obtain F2 progeny.

Thoracic spinal cords of Hspa8 knock-in (KI) homozygous and WT rats at 9 weeks of age were removed and homogenized in a cell lysis reagent (CelLytic MT, Sigma Aldrich, St. Louis, MO, United States) with proteinase inhibitor cocktail (Nacalai tesque, Kyoto, Japan). The supernatants were collected after centrifugation at 13,000 ×g for 10 min and protein concentrations were determined by an absorption spectrometer using the Bradford protein Assay (Bio-Rad Laboratories, Hercules, CA, United States). The supernatants were boiled for 5 min with SDS sample buffer (Cosmo Bio, Tokyo, Japan) with 5% 2-mercaptoethanol (Bio-Rad Laboratories). Samples were separated on 5–20% gradient polyacrylamide gels (ATTO Corporation, Tokyo, Japan) and transferred to polyvinylidene difuoride (PVDF) membranes (Bio-Rad Laboratories). Membranes were incubated overnight at 4°C with the following antibodies: rabbit monoclonal anti-HSC70 (clone EP1531Y, ab51052, 1:2,000; Abcam, United Kingdom) and mouse monoclonal anti-β-actin (clone AC-15, A54411, 1:30,000; Sigma Aldrich). After washing, the membranes were treated with peroxidase-conjugated secondary antibody (Histofine Simple Stain MAX PO; Nichirei Biosciences) for 30 min (for β-actin) or 1 h (for HSC70) at room temperature. Signals were visualized with ECL-prime (GE Healthcare, United Kingdom) and quantified with a luminescent image analyzer (LAS-4000; GE Healthcare). β-actin was used as an internal control.

Human WT and V95E HSC70 (HSPA8) were expressed and purified as previously described (Rauch et al., 2016; Johnson et al., 2022). In brief, inoculated bacterial cultures were grown to OD600 = 0.6 and cooled to 20°C. At this point, protein expression was induced with 200 μL IPTG and cultures were incubated with shaking overnight. Cells were harvested and the resulting pellet was either frozen at-80°C or carried forward immediately to purification. Cell pellets were lysed by sonication and cleared lysate was applied to Ni-NTA resin (Thermo Fisher Scientific, Waltham, MA, United States). The Ni-NTA elution was treated with TEV protease (MacroLab; University of California, Berkeley) to cleave the 6His-tag and the cleavage product was further purified by an ATP-agarose (Sigma Aldrich) column. DnaJB4 was expressed and purified as previously reported (Rauch et al., 2016). Inoculated bacterial cultures were grown, induced, and harvested as above. Cell pellets were lysed by sonication and cleared lysate was applied to Ni-NTA resin (Thermo Fisher Scientific). The Ni-NTA elution was treated with TEV protease (MacroLab; University of California, Berkeley) to cleave the 6His-tag and the cleavage product was further purified by a Superdex S200 (GE Healthcare) size exclusion column. BAG1 was expressed and purified as described elsewhere (Rauch et al., 2016). Briefly, cultures were handled as above and then cells were lysed by sonication and lysate was applied to a HisPur Ni-NTA resin (Thermo Fisher Scientific). The Ni-NTA elution was treated with TEV protease (MacroLab; University of California, Berkeley) and the cleavage product was subsequently purified by ion-exchange chromatography (MonoQ, GE Healthcare).

ATP hydrolysis activity was assessed in a 96-well plate format (Fisher #12565501) using a malachite green assay that measures the generation of inorganic phosphate upon ATP hydrolysis as previously described (Chang et al., 2008; Rauch and Gestwicki, 2014; Rauch et al., 2016; Taylor et al., 2018). In short, 1 μM HSC70 was titrated with increasing concentrations of the co-chaperone of interest (DnaJB4 or BAG1). Reactions were initiated by the addition of excess ATP and incubated at 37°C for 1 h. Malachite green reagent was then added to each well and reactions were immediately quenched with sodium citrate. Absorbance was measured at 620 nm using a SpectraMax M5 plate reader (Molecular Devices). Experiments were performed in triplicate. DnaJB4 data were fit by the Michaelis–Menten equation to derive pseudo-K_m and pseudo-V_max values. DnaJB4 and BAG1 titration data were normalized as percentages by taking the lowest and highest value in each subcolumn as 0 and 100%, respectively, in GraphPad Prism 9 (GraphPad Software, Boston, MA, United States) software.

The luciferase refolding assay was performed in a 96-well format (Corning #3912) as reported previously (Rauch and Gestwicki, 2014; Rauch et al., 2016; Taylor et al., 2018). Briefly, Renilla luciferase (0.5 mg/mL, Promega, Madison, WI, United States) was denatured in buffer A (25 mM HEPES; pH 7.2, 50 mM potassium acetate, and 5 mM dithiothreitol) containing 8 M guanidine hydrochloride (GuHCl) at room temperature for 120 min. The denatured protein was diluted 1:40 in buffer A and placed on ice for 20 min. Refolding reactions were then prepared by adding 2 μL of the denatured luciferase stock into 48 μL of refolding buffer (28 mM HEPES; pH 7.6, 120 mM potassium acetate, 12 mM magnesium acetate, 2.2 mM dithiothreitol, 0.1 mM ATP, 8.8 mM creatine phosphate, and 35 U/mL creatine kinase) containing HSC70 (1 μM) and varying concentrations of the indicated co-chaperone (DnaJB4 and/or BAG1). Refolding reactions were initiated by the addition of excess ATP (1 mM) and incubated at 37°C for 1 h. After this incubation, SteadyGlo (Promega) reagent was added and the luminescence signal was immediately measured using a SpectraMax M5 plate reader. Titration data were normalized as percentages by taking the lowest and highest luminescence values as 0 and 100%, respectively, in GraphPad Prism 9 software. DnaJB4 data were fit by the Michaelis–Menten equation to derive half-maximal effect values and no statistically significant difference (p > 0.05) was observed in the WT HSC70 vs. HSC70 V95E results. Experiments were repeated on three separate protein samples (biological replicates) and each experiment was separated into three wells (technical triplicates; total n = 9).

FP saturation assays were carried out as previously described (Rauch and Gestwicki, 2014; Rauch et al., 2016). Briefly, 20 nM N6-(6-Amino)hexyl-ATP-5-FAM (ATP-FAM) (Jena Bioscience) was incubated with a titration of HSC70 in black, round-bottom, low-volume, 384-well plates (Corning #4511) for 30 min at room temperature. Fluorescence polarization was measured (excitation, 485 nm; emission, 535 nm) using a SpectraMax M5 plate reader. Experiments were performed in triplicate and data were fit using a sigmoidal dose–response (variable slope) curve in GraphPad Prism 9. Data was transformed as (x = log [HSC70 (μM)]).

Data are expressed as the mean ± SD. Data were analyzed using GraphPad Prism 9. The statistical significance of differences among multiple groups was determined by two-way ANOVA with Bonferroni’s post-hoc test. Comparisons between the two groups only were determined by Student’s or Welch’s t-test. p < 0.05 were considered statistically significant (see figure legends).

To find the onset of clinical symptoms of the F344-kk/kk rats, we examined when abnormal gaits appeared. At 3 weeks of age, the kk/kk homozygous rats showed an unsteady hindlimb gait compared with control rats (Supplementary Video S1). Almost all kk/kk rats (94% of kk/kk rats) developed abnormal gaits appearing in the hind limbs between 7 and 8 weeks of age. The abnormal gaits were characterized by the significantly shorter step lengths: the step lengths of the homozygous rats were shorter than the WT rats (7.7 ± 0.37 cm vs. 10.0 ± 1.43 cm, p < 0.001) but the step widths were not different between the homozygous and WT rats (4.3 ± 0.23 cm vs. 4.0 ± 0.51 cm, p = 0.13). The kk/kk homozygous rats exhibited the complete ataxia of the limbs and stooping position, hindlimb splay/extension, and eventually fell into a complete prone position (Figures 1A,B). After the onset of the abnormal gaits, kk/kk rats could not survive longer than 4 weeks. We observed mortalities of male (n = 8) and female (n = 5) kk/kk rats for 12 weeks and found that the mortalities were not different between male and female rats (9.5 ± 0.69 vs. 10.2 ± 1.10 weeks of age, p = 0.135). The average mortality of the kk/kk rats was 9.79 ± 0.89 weeks of age. No WT rat died during the 12-weeks observation periods. Additionally, kk/kk rats exhibited significantly smaller body size than the WT rats from 3 weeks of age (Figure 1A; Supplementary Figure S2).

To examine histopathological alterations in the CNS, we performed light microscopic observation. We found dystrophic swollen axons throughout the CNS from at least 3 weeks of age in the kk/kk homozygous rats (Figure 1). The axonal spheroids varied in size (approximately 5–50 μm) and were heterogeneous in morphology: they showed a homogeneous or granular appearance with or without clefts and vacuoles (Figure 1G). Such axonal swellings were not observed at all in the WT rats. In the spinal cord, axonal spheroids were observed in both white and gray matter, located mainly in the dorsal cord (fasciculus gracilis and dorsal corticospinal tract) (Figures 1D,F). In gray matter, lesions locate in the dorsal horn rather than the ventral horn. The spinal cord lesions were more severe in the posterior than the anterior of the spinal cord; the latter part of thoracic, lumber, and sacral parts exhibited more severe lesions than the cervical parts of the spinal cord (Supplementary Figure S3). In the brain stem, numerous large axonal spheroids were observed from at least 3 weeks of age, especially in the nucleus gracilis, nucleus cuneatus, and nucleus cuneatus accessorius of the medulla oblongata (Figure 1I). In the cerebellum, cerebellar white matter, cortex (predominantly in the granular cell layer), and the inferior and middle cerebellar peduncle were primarily affected. Some spheroids were associated with large vacuoles in the granular layers (Figure 1K). Based on these neuropathological findings, the F344-kk/kk rats were histopathologically diagnosed with NAD.

To examine morphology of the swollen axons of F344-kk/kk rats in detail, we performed TEM experiments on the dorsal cord of the lumber spinal cord in the kk/kk homozygous rats (Figure 2). We found the accumulation of electron-dense bodies (Figure 2: white arrows), degenerated or swollen abnormal mitochondria (Figure 2: white arrowheads), membranous or tubular structures, edematous vacuoles (Figure 2: asterisks), and neurofilaments (NFs) in the swollen axons (Figure 2). Swollen axons sometimes lacked a myelin sheath. These findings were consistent with the ultrastructural observations previously reported in patients and animals with NAD (Yagishita and Kimura, 1974; de Leon and Mitchell, 1985; Beck et al., 2011; Tanaka et al., 2017).

To characterize the spheroids and predict the pathogenesis of NAD in F344-kk/kk rats, we performed IHC analysis (Figure 3). Axonal spheroids were strongly positive for synaptophysin, a synapse-associated glycoprotein on presynaptic vesicles (Figure 3B). The amyloid β precursor protein (APP) is thought to be an effective marker for axonal injury (Gentleman et al., 1993; Hayashi et al., 2015). Although the APP was not accumulated in axons of the WT rats (Figure 3C), it was accumulated in swollen axons and even in some non-swollen axons in F344-kk/kk rats (Figure 3D). This finding indicated that both swollen and non-swollen axons were injured in the F344-kk/kk rats. In addition, the spheroids were also positive for neurofilament (neuron-specific intermediate filaments) proteins (Supplementary Figure S4) and ubiquitin (Figure 3F), an important factor for post-translational modifications (Lee et al., 1987; Yuan and Nixon, 2021). These proteins were not accumulated within axons of the WT rats.

To identify the causative mutation of the kk, we performed a positional cloning approach. We obtained 171 kk/kk homozygous rats from F2 intercross and 87 kk/kk homozygous rats from backcross progeny. Genome-wide scanning using 22 kk/kk rats mapped kk to the rat Chr 8. Fine mapping using the remaining kk/kk progeny mapped the kk between J683662 and D8Rat188 (Figure 4A). The kk was physically mapped within a ~ 1.02-Mb region defined by J683662 and D8Rat188, in which nine genes were included. We searched the Rat Genome Database for genes that were expressed in the CNS and found that Clamp (CXADR-like membrane protein), Hspa8 [heat shock protein family A (Hsp70) member 8], Bsx (brain-specific homeobox), Jhy (junctional cadherin complex regulator), and Ubash3b (ubiquitin associated and SH3 domain containing, B) were expressed in the CNS.

To find the mutation, we sequenced coding regions of these candidate genes of F344-kk/kk rats. Although no mutation was found in the Clamp, Bsx, Jhy, or Ubash3b genes, we found a nucleotide substitution from T to A at nucleotide 284 of the coding sequence (c.284 T > A) of Hspa8 (Figure 4B). The substitution was predicted to change valine (Val) to glutamate (Glu) at amino acid 95 located in the nucleotide-binding domain (NBD) of the rat HSC70 protein (Figure 4B; Supplementary Figure S5). Missense prediction analysis with the PROVEAN software predicted that Val95Glu (V95E) was deleterious (score: −3.629) (Choi et al., 2012). All of the 171 kk/kk progeny were homozygous for the V95E missense mutation.

To examine whether the V95E missense mutation caused NAD in F344-kk/kk rats, we produced Hspa8 knock-in (KI) rats that harbored the mutation. Knocking in of the mutant allele was performed by lsODN-mediated knock-in with the CRISPR-Cas9 system (Yoshimi et al., 2016; Supplementary Figure S6). We obtained 24 pups from pseudopregnant rats that electroporated embryos were transferred into. Direct sequencing analyses revealed that 4 rats were potential founders. We developed two Hspa8-KI rat strains. Hspa8-KI rats, homozygous for the V95E mutation, exhibited abnormal gaits of the hind limbs after 5 weeks of age and deteriorated with increasing age (Figure 5A). Heterozygous and WT rats did not show these clinical signs. The expression level of HSC70 protein was significantly decreased in the spinal cord of Hspa8-KI homozygous rats compared with the WT rats, which may contribute to the disease phenotype in the Hspa8-mutant rats (Figure 6). Histopathologically, the Hspa8-KI homozygous rats exhibited spheroids in the CNS, which was very similar to those of F344-kk/kk rats (Figures 5B–E). Thus, we concluded that the V95E missense mutation of the Hspa8 gene was the causative factor of NAD in rats.

HSC70 is composed of two domains: the N-terminal NBD and the C-terminal substrate-binding domain (SBD). The NBD, where the V95E mutation is located, hydrolyzes ATP and this activity is coupled to the binding and release of its peptide/protein substrates in the SBD. The SBD itself may be further distilled to an α-helical lid and a β barrel that directly interacts with its substrates (Figure 7A). Hydrolysis of ATP in the NBD is known to cause a conformation change that “closes” the lid and strengthens affinity for misfolded proteins (termed “clients”); therefore, the V95E mutation may affect ATP hydrolysis or functional coupling between the NBD and SBD. Furthermore, this hydrolysis cycle is assisted by co-chaperone proteins, including J-domain proteins (JDPs), which catalyze ATPase activity, and nucleotide exchange factors (NEFs), including those which reset the nucleotide cycle to promote iterative ATP binding, hydrolysis, and ADP release from the NBD (Figure 7A). Thus, another outcome of the V95E mutation may be disruption in the functional interaction of HSC70 with its co-chaperone partners.

To generate a working hypothesis for how the V95E mutation might impact HSC70 or its interactions with co-chaperones, we mapped the location of the mutation onto a structure of the NBD of human HSC70 (pdb 4H5T). First, we noticed that the V95 position is far from residues known to be critical for nucleotide binding or hydrolysis. To test this idea, we measured binding of a fluorescent ATP analog to recombinantly expressed and purified WT and the V95E mutant proteins and found that, indeed, both proteins had similar affinity (WT Kd 350 ± 68 nM; V95E Kd 470 ± 120 nM; Supplementary Figure S7). Likewise, when we measured steady state ATP hydrolysis in malachite green assays, the WT and V95E proteins had very similar initial velocity (V₀) rates (~16 pmol ATP/μM chaperone/min), demonstrating that the V95E mutation does not meaningfully affect the basal enzymatic activity of HSC70.

Next, we tested whether the V95E mutation might impact collaboration with co-chaperones. When we titrated DnaJB4 into hydrolysis reactions containing recombinantly purified HSC70, we found that the ATPase activities of both WT and V95E were stimulated (Figure 7B; left). By fitting the results to the Michalis-Menten equation, we derived pseudo-Km values, showing that the apparent affinities for DnaJB4 were comparable (WT = 0.092 ± 0.018 μM; V95E = 0.088 ± 0.022 μM). We did notice a modest decrease in the pseudo-V_max, with WT having a maximum hydrolysis rate of 42.90 ± 0.365 μmol P_i/min, and V95E having a pseudo-V_max of 36.63 ± 1.41 μmol P_i/min (^***p = 0.0001). Taken together, these data suggest that V95E HSC70 has a modest effect on DnaJB4-stimulated hydrolysis of ATP. Next, we explored the response of WT and V95E to a representative of the NEF family of co-chaperones: BAG1. In these experiments, BAG1 was titrated into a mixture of HSC70 (1 μM) and DnaJB4 (0.1 μM). The results revealed that BAG1 stimulates WT HSC70’s ATPase activity, but that V95E was relatively insensitive to BAG1 (Figure 7B; right), only reaching approximately 25% stimulation at the highest concentration tested. These data suggest that V95E HSC70 has perturbed interactions with co-chaperones, especially BAG1. We next tested the effects of V95E on a more complex chaperone function: refolding of the denatured client. In these assays, chemically denatured Renilla luciferase is added to a mixture of chaperones, co-chaperones, and ATP and the ability to refold the client protein is measured by production of light (luminescence). With either DnaJB4 (Figure 7C; left) or the combination of DnaJB4 and BAG1 (Figure 7C; right), both WT HSC70 and V95E HSC70 were able to refold luciferase to a similar extent. Thus, the effects of V95E on ATPase assays do not seem to dramatically impact refolding activity.