A novel IBA57 variant is associated with mitochondrial iron–sulfur protein deficiency and necrotizing myelopathy in dogs

Introduction: Hereditary necrotizing myelopathy (HNM) in young Kooiker dogs is characterized by progressive ataxia and paralysis with autosomal recessive inheritance. The basic genetic defect is unknown. We investigated the possible cause by a genome-wide analysis using six affected and 17 unrelated unaffected Kooiker dogs and by functional follow-up studies. Method: The HNM locus was mapped by a case–control study using a dense SNP array and confirmed by linkage analysis of two pedigrees. The gene exons in the critical region were analyzed by next-generation sequencing. The functional effect of the candidate canine IBA57 pathogenic variant was biochemically examined in an established HeLa cell culture model in which the endogenous IBA75 gene product was depleted by RNAi. Results: The basic defect was localized in the centromeric 5 Mb region of canine chromosome 14. The most associated SNP co-segregated fully with HNM and reached an LOD score of 6.1. A candidate pathogenic mutation was found in the iron–sulfur cluster assembly gene IBA57 and led to the amino acid substitution R147W. The expression of human IBA57 harboring the canine R147W exchange could only partially restore the biochemical defects of several mitochondrial [4Fe-4S] proteins upon IBA57 depletion, showing that the mutant protein is functionally impaired. Discussion: Pathogenic variants in human IBA57 cause multiple mitochondrial dysfunction syndrome 3 (MMDS3), a neurodegenerative disorder with distant similarities to HNM. The incomplete functional complementation of IBA57-depleted human cells by IBA57-R147W identifies the DNA mutation in affected Kooiker dogs as the genetic cause of HNM. Our findings further expand the phenotypic spectrum of pathogenic IBA57 variants.


Introduction
A variety of inherited neurological disorders have arisen from spontaneous mutation events in dog breeds. These diseases present models for human diseases with shared molecular defects (van Steenbeek et al., 2016). As most dog breeds tend to be inbred and subjected to artificial selection by the breeders, these spontaneous mutations can reach undesirable frequencies.
A notable example is degenerative myelopathy in dogs, caused by mutations in SOD1, known for its involvement in amyotrophic lateral sclerosis in humans (Zeng et al., 2014). Less well characterized is the hereditary necrotizing myelopathy (HNM) affecting young dogs of the breed Nederlandse Kooikerhondje (named Kooiker dogs hereafter), which has been described earlier by Mandigers et al. (1993). The clinical signs, paresis and ataxia, start in most dogs around the age of 3-12 months in the hind limbs and progresses to tetraparalysis before the age of 2 years. Post-mortem examination performed in these dogs revealed a symmetric bilateral necrotizing myelopathy with malacia in the ventral and dorsal white matter. Pedigree analysis indicated that the disease is inherited as an autosomal recessive trait, probably originating from a single founder (Mandigers et al., 1993). Here, we present the identification of an associated variant and the functional evaluation of the defect gene product, a variant of IBA57 encoding a component involved in mitochondrial  protein assembly (Lill and Freibert, 2020). According to our findings, the IBA57 mutation is the likely cause of the observed disease phenotype.

Dogs
Young Kooiker dogs born in Europe, which displayed neurological signs suggestive of myelopathy, were referred for clinical examination to the first author (PM) at the Department of Clinical Sciences of the Faculty of Veterinary Medicine of Utrecht University (Table 1). If the neurological signs were conclusive for myelopathy and the clinical features were in agreement with HNM (Mandigers et al., 1993), additional diagnostics was limited to a routine blood examination, a toxoplasmosis titer, and on request of the owner, an MRI scan, under generalized anesthesia, of the cervicothoracic spinal cord. For the generalized anesthesia, the dogs were premedicated 15 min prior to induction with atropin at a dose of 0.03 mg/kg intramuscularly. The dogs were induced with 0.5 mg/kg of methadone (methadone, Pharmacy Department, Utrecht University) intravenously and 0.5 mg/kg of midazolam (midazolam in Dormicum Roche Netherlands BV, Mijdrecht) intramuscularly. The dogs were further anesthetized using propofol (propofol, Fresenius Kabi, Netherlands BV, Den Bosch) (based on an effect of approximately 2 mg/kg) and, after intubation, maintained on 0.5 L of oxygen, 1 L of air, and approximately 1% of isoflurane (Abbott Animal Health). The dogs remained under the clinical supervision of the first author, and at deterioration of neurological signs, an elective euthanasia, using 5 mL of EUTHASOL ® intravenously (AST Pharma BV, Oudewater, Netherlands), was performed. For post-mortem examination, the brain and spinal cord were fixed in 4% buffered formaldehyde and examined to confirm the diagnosis of HNM as described earlier (Mandigers et al., 1993 (Table 1) and 17 unrelated, unaffected Kooiker dogs that were older than 1 year and clinically unaffected were genotyped using the Illumina CanineHD SNP array. The DNA samples and arrays were processed at the Centre National de Génotypage, Paris, France, at the time in the frame of the LUPA European project 1 . A case-control comparison of allele frequencies was performed with the PLINK software version 1.07.
The SNP BICF2G630517911 was genotyped in available samples from two pedigrees that segregated HNM (Supplementary Figure  S1). The oligonucleotides for PCR amplification and chain termination DNA sequence analysis, performed as outlined in the following section, are included in Supplementary Table S1. The LOD score for linkage was calculated with a stand-alone version of the SuperLink version 1.7 (Fishelson and Geiger, 2004).

DNA sequence analysis
One affected Kooiker dog and two unaffected control dogs were selected for targeted resequencing of genomic DNA. After DNA was sheared, fragments of all exons and adjacent intronic 35 bp of genes from the critical region were enriched for and amplified using an Agilent SureSelect in-solution enrichment design. Library preparation was performed using the Illumina TruSeq ® Nano DNA Prep Kit. The enriched libraries were sequenced on an Illumina MiSeq. Mapping and variant calling of the paired-end reads of 150 bases were achieved using a custom bioinformatic pipeline based on the Burrows-Wheeler Aligner algorithm (Li and Durbin, 2009). The SAMtools was used for annotation and initial variant calling against the reference genome CanFam3.1 . The variants were curated by comparing the allele frequencies of detected variants in a cohort of 590 dogs (Supplementary Table S2). Two variants of interest were further analyzed in all available dogs of the HNM pedigrees by chain termination sequencing using the BigDye v3.1 (Applied Biosystems) on a Genetic Analyzer 3500xL (Applied Biosystems). The oligonucleotides used for PCR amplification and sequencing reactions are specified in Supplementary Table S1.
Tissue culture-based IBA57 complementation assays and plasmid generation HeLa cells were cultured at standard conditions and transfected thrice in a 3-day interval by electroporation as described by Stehling et al. (2009), Sheftel et al. (2012, and Ajit Bolar et al. (2013), allowing for RNAi-mediated depletion and concomitant plasmidbased expression of silently mutated, RNAi-resistant, wild-type IBA57 (smIBA57 wt ). The smIBA57 wt -encoding plasmid (Debray et al., 2015) used in this procedure further served as a template to introduce the Kooiker dog p.Arg147Trp (R147W)-encoding genomic mutation into the human IBA57 open reading frame. The resulting plasmid smIBA57 R147W was used to study the HNM-related mutation in the complementation assays. Plasmid modification was achieved using the Q5 ® Site-Directed Mutagenesis Kit (New England Biolabs) according to the protocol of the manufacturer. The mutagenesis primer sequence was TCCGGTGGAAGGTCAC, with the mutant base underscored, and the reverse primer sequence was TCCTGTATAGCGCGA. The DNA sequence of the complete IBA57 coding region of the mutant plasmid was verified by dideoxy DNA sequencing. For subsequent biochemical analyses, the transfected HeLa cells were harvested by trypsination, and sample aliquots were subjected either to immunoblotting or digitonin-based cell fractionation, followed by spectrophotometric determination of enzyme activities as published (Biederbick et al., 2006;Stehling et al., 2009;Stehling et al., 2018b).

Statistical analysis
Data are presented as mean ± SD and analyzed by one-way repeated measures ANOVA, followed by Dunnett's post hoc test. The smIBA57 wt data served as the control in order to individually determine the significance levels of IBA57 depletion phenotypes as well as of smIBA57 wt and smIBA57 R147W complementation phenotypes. *p < 0.05, **p < 0.01, and ***p < 0.001.

Dogs and clinical findings
Eight young Kooiker dogs with clinical signs of HNM were screened between 2001 and 2014 (Table 1). All eight dogs (six males and two females) had a history of progressive hind limb paresis and ataxia with an onset between 4 and 12 months of age. Neurological examination revealed postural deficits and exaggerated spinal reflexes in the hind limbs in all eight dogs. Two dogs also showed ataxia of the front limbs and postural deficits. Based on the neurological examination, the neurolocalization was in the cervicothoracic/thoracolumbar region. As the clinical presentations were highly similar to cases observed earlier (Mandigers et al., 1993), the diagnostics was limited to a routine blood examination and toxoplasmosis titer, which revealed no abnormalities in all eight dogs. In one dog, an MRI scan was performed that revealed abnormalities suggestive of spinal cord disorder (Truar et al., 2012). The health condition of all dogs deteriorated within the following months, pressing the owners to elect euthanasia. Post-mortem examination performed in these eight dogs revealed similar pathological lesions as described earlier (Mandigers et al., 1993), revealing a symmetric bilateral necrotizing myelopathy with malacia in the ventral and dorsal white matter, thus confirming the clinical diagnosis.

Gene mapping and mutation detection
The genome-wide association analysis of six HNM-affected Kooiker dogs from different litters and 17 unrelated controls from the same breed suggested involvement of the region adjacent to the centromere of chromosome 14 (CFA14). The p-value of the chi-square test for this region reached 1.7 10 −7 (Figure 1). The SNP allele frequencies in the region showed that the cases were identically homozygous from the beginning of the chromosome to position 4719754 (SNP BICF2G630519065, CanFam3.1) on CFA14. The segregation of SNP BICF2G630517911 on position 3021087 was analyzed in two extended pedigrees, which included two more patients in addition to the six cases that were included in the genome-wide association analysis (Supplementary Figure S1). The allele of this SNP shared by the cases had a frequency of 0.14 in the controls of the genome-wide analysis. The LOD score for linkage between HNM and the SNP in the two pedigrees was 6.1 without recombinations.
The critical region contained 38 genes (Supplementary Table S3). The exons and bordering DNA sequences of these genes from one case and two controls were enriched and analyzed by next-generation sequencing techniques. One variant in IBA57 and one in OBSCN remained the causative candidates for HNM after quality control and filtering of all observed variants.
The observed variant in OBSCN changed an arginine codon to a stop codon.
The annotation was CFA14:g.693260G>A, resulting in XM_ 038686046.1:c.17228C>T and XP_038541974.1:p.Arg5690*, truncating the largest isoform of obscurin by 3,325 amino acids. Obscurin is a large cytoskeletal protein in the striated muscles, and variants have been associated with cardiomyopathies in humans, although the relevance of the latter has been questioned (Grogan and Kontrogianni-Konstantopoulos, 2019;Fukuzawa et al., 2021). The available DNA samples of dogs from the extended pedigrees were genotyped for the two candidate variants, and both co-segregated with HNM. However, the two variants were not in complete linkage disequilibrium. Of the 136 heterozygous carriers of the IBA57 variant that were detected in our DNA bank, three dogs did not carry the OBSCN mutation. A total of 1,522 Kooiker dogs were analyzed, yielding an allele frequency of 0.04, which is not unusual in dog genetics.
Frontiers in Genetics frontiersin.org 04 .1:p.Arg147Trp. The mutated arginine residue is part of a short loop between β-strand 6 and α-helix 2 of the protein that is conserved in animals but not in, e.g., plants and fungi ( Figure 2) and also not in bacterial YgfZ relatives (Waller et al., 2010;Weiler et al., 2020).

IBA57 complementation assay
To test whether the IBA57-R147W amino acid exchange causes a mitochondrial phenotype and thus relates the Kooiker dog HNM to the MMDS3 spectrum in humans, we applied a well-established HeLa tissue culture complementation approach (Ajit Bolar et al., 2013;Debray et al., 2015;Lossos et al., 2015). In this assay, endogenous IBA57 was depleted by three consecutive rounds of RNAi, and concomitantly, a siRNA-resistant version of human IBA57 (smIBA57) harboring the HNM c.439C>T mutation was ectopically expressed from a plasmid. After a total tissue culture time of 9 days, RNAi decreased the protein level of endogenous IBA57 by 60%-80%, whereas the abundance of the ISC factor NFU1 was not altered (Figures 3A, B). Expression of wild-type IBA57 (smIBA57 wt ) or R147W mutant protein (smIBA57 R147W ) restored IBA57 to even higher than the control levels, enabling the analysis of mutant smIBA57 R147W for its capacity to complement the mitochondrial defects resulting from RNAi-mediated IBA57 depletion. The enzyme activity measurements in IBA57-deficient cells revealed defects in the [4Fe-4S] cluster-containing proteins mitochondrial aconitase (ACO2) and succinate dehydrogenase (SDH) by more than 30% and 75%, respectively, but not in the Fe/S cluster-devoid reference enzyme citrate synthase ( Figures 3C, D). Particularly, the decrease in SDH activity was so severe that it also elicited a decline in the steady-state level of the Fe/S cluster-binding subunit SDHB (Figures 3E, F), which is in line with previously observed instability of Fe/S proteins in their apo-form (Stehling et al., 2018b).
The expression of smIBA57 wt could recover both ACO2 and SDH enzyme activities as well as the amount of SDHB up to the control levels ( Figures 3C, F). By contrast, the Kooiker dog HNM mutation-containing protein smIBA57 R147W was inefficient in restoring normal mitochondrial [4Fe-4S] protein assembly. The mutant protein elicited no improvement of ACO2 activity and corrected SDH activity to not more than 50% of wild-type smIBA57.

FIGURE 3
IBA57-R147W amino acid exchange compromises mitochondrial [4Fe-4S] proteins. HeLa cells were transfected thrice at a 3-day interval with scrambled control siRNAs (SCR) or with an IBA57-directed siRNA (siIBA57) in combination with complementing (compl.) plasmids encoding RNAiresistant wild-type or HNM-related IBA57 (smIBA57 wt or smIBA57 R147W , respectively). As a control, cells were mock-transfected. Analyses were performed with cells harvested after the third round of transfection (9 days of IBA57 depletion). (A) Total cell lysates were subjected to immunoblotting and analyzed for the protein levels of the indicated ISC proteins and β-actin as a loading control. (B) Immunoblot signals from part (A) were quantified relative to β-actin levels, and the ratio was normalized to mock-transfected control cells (dashed line). (C, D) Specific activities of mitochondrial aconitase (ACO2), SDH (RCC-II), and COX (RCC-IV) (C) were determined in mitochondria-containing membrane fractions prepared by digitonin-based cell fractionation, related to citrate synthase (CS) activity (D), and normalized to mock-transfected control cells. (E) Total cell lysates were subjected to immunoblotting as in (A) and analyzed for the steady-state levels of mitochondrial proteins and RCC subunits as indicated. (F) Immunoblot signals from (E) were quantified, and the protein per β-actin ratios (see A) were normalized to mock-transfected control cells. Representative blots are shown. All values are given as the mean ± SD (n = 3-4); *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant; dashed lines, 100% value of mock-transfected control cells.
Frontiers in Genetics frontiersin.org 05 IBA57 deficiency compromises TCA cycle function not only directly via impaired aconitase and SDH function but also indirectly by impairing protein lipoylation catalyzed by the mitochondrial [4Fe-4S] protein lipoic acid synthase (LIAS) (Sheftel et al., 2012;Debray et al., 2015). Probing for the lipoyl cofactors bound to the E2 subunits of the alpha-ketoglutarate dehydrogenase complex (KGDHc) and of the pyruvate dehydrogenase complex (PDHc) by immunoblotting revealed a severe defect in IBA57-deficient cells that was restored to the reference levels by smIBA57 wt but was only insufficiently complemented by the smIBA57 R147W mutant protein ( Figures 4A, B). By contrast, the levels of PDH-E2 (DLAT) protein were neither altered by RNAi nor by complementation. Together, these results indicate a limited LIAS maturation activity in the IBA57-R147W mutant protein-expressing cell.
Similar to SDHB, other Fe/S cluster-containing respiratory chain complex (RCC) subunits are also less stable without their cofactors (Sheftel et al., 2012;Torraco et al., 2017). Accordingly, protein levels of the two RCC-I [4Fe-4S] subunits NDUFS1 and NDUFS8 were decreased upon RNAi-mediated IBA57 deficiency (Supplementary Figure S2A, B). Since both NDUFS1 and NDUFS8 join the N and Q module assembly intermediates of RCC-I at early assembly steps (Guerrero-Castillo et al., 2017), their decay impaired RCC-I formation in general and resulted in the destabilization of other sensitive protomers, which includes the [2Fe-2S] subunit NDUFV2 and non-Fe/S subunits NDUFA13 and NDUFB4, whereas more stable subunits like NDUFA9 and NDUFB6 remained unaffected (Supplementary Figure S2A-D), which is in line with previous observations (Sheftel et al., 2009;Sheftel et al., 2012). Complementation of RNAi-mediated IBA57 deficiency by wild-type smIBA57 restored the majority of tested RCC-I subunits to the control levels, whereas the IBA57-R147W mutant protein only partially recovered NDUFS1 and NDUFS8 and entirely failed to recover the other sensitive RCC-I subunits. This difference in complementation between wild-type and mutant IBA57 again suggests a detrimental effect of the R147W amino acid exchange on mitochondrial [4Fe-4S] cluster assembly.
A yet ill-explained hallmark of IBA57 deficiency is the decrease in RCC-IV (COX) activity (Sheftel et al., 2012;Ajit Bolar et al., 2013;Debray et al., 2015;Stehling et al., 2018b), which is associated with a decline of COX2 and COX6A/B subunit levels ( Figures 3C, E, F). Complementation by wild-type smIBA57 restored both RCC-IV activity and subunit levels, but smIBA57 R147W only recovered the COX enzymatic activity to some extent, while the COX2 and COX6A/B levels remained diminished (Sheftel et al., 2012). By contrast, neither RNAi-mediated depletion of IBA57 nor complementation by smIBA57 wt or smIBA57 R147W substantially affected the [2Fe-2S] subunit UQCRFS1 (Rieske Fe/S protein) of RCC-III (Supplementary Figure S2E), which is consistent with the specific function of IBA57 in the maturation of mitochondrial [4Fe-4S] but not of [2Fe/2S] proteins (Sheftel et al., 2012). Accordingly, the UQCRC2 subunit level also remained unchanged. Immunostaining against the F1α/β subunits of complex V (F 1 F o -ATP synthase) did not reveal any alterations (Figures 3C, E, F) as well, arguing against a general OXPHOS defect caused by the IBA57-R147W amino acid exchange. Overall, these results are fully consistent with an MMDS3-related phenotype in the HNM-affected Kooiker dogs.
Frontiers in Genetics frontiersin.org 06 lactate dehydrogenase (LDH; Supplementary Figure S3A, B), while the levels of IRP1 protein remained unaffected (Supplementary Figure S3C-F). Likewise, probing for the amounts of several cytosolic and nuclear Fe/S and control proteins by immunoblotting did not reveal major alterations, with the notable exception of the CIA component CIAO3, which is known to be sensitive to disturbances in mitochondrial homeostasis (Torraco et al., 2017). Taken together, the IBA57-R147W mutant protein only partially and specifically complemented the mitochondrial [4Fe-4S] protein defects elicited by IBA57 depletion. Collectively, the findings strongly suggest that the IBA57 Kooiker dog mutation is causative for the observed HNM phenotype by impairing the function of canine IBA57 in mitochondrial [4Fe-4S] protein assembly.

Discussion
We have shown here, by genetic association and linkage analysis, that the gene causing HNM in Kooiker dogs is located near the centromere of chromosome 14. The p-value obtained for the association of this region was almost 4 orders of magnitude lower than the p-value for the next best region (Figure 1). The LOD score of 6.1 was well above the threshold value of 3, generally considered the proof of linkage.
Two of the 38 genes in the HNM candidate region are known for causing neurodegenerative phenotypes in humans. First, mutations in human GJC2, coding for gap junction protein gamma-2 (OMIM 608803), are the cause of hypomyelinating leukodystrophy 2 (OMIM 608804), lymphatic malformation 3 (OMIM 613480), and autosomal recessive spastic paraplegia 44 (OMIM 613206). However, no mutations were found in the coding DNA sequences of GJC2 of affected Kooiker dogs. Second, mutations in human IBA57 have been implicated in neurological diseases such as spastic paraplegia 74 (OMIM 615316) and multiple mitochondrial dysfunctions syndrome 3 (MMDS3, OMIM 615330) (Ajit Bolar et al., 2013;Debray et al., 2015;Lossos et al., 2015;Ishiyama et al., 2017;Lebigot et al., 2017;Torraco et al., 2017;Hamanaka et al., 2018;Liu et al., 2018). We identified a XM_ 025454932:c.439C>T mutation in canine IBA57 that co-segregated with HNM in the Kooiker dog pedigrees. The resulting XP_ 025310717:R147W mutation affected the second of two consecutive arginine residues that are conserved in animals but not in other species (Figure 2). IBA57 is essential for the assembly of [4Fe-4S] clusters in the mitochondria and cooperates with the mitochondrial ISC factors ISCA1 and ISCA2. Consistent with the tight functional cooperation of these three proteins, mutations in their genes have been associated with similar disease phenotypes summarized as MMDS (Ajit Bolar et al., 2013;Debray et al., 2015;Lossos et al., 2015;Ishiyama et al., 2017;Lebigot et al., 2017;Torraco et al., 2017;Hamanaka et al., 2018;Liu et al., 2018). Mutations in human IBA57 seem to occur more frequently than in the other two genes, and the phenotypical spectrum spans from relatively mild neurological symptoms to fatal outcome at birth. The severity of the disease cases correlates well with the associated specific biochemical defects in mitochondrial [4Fe-4S] proteins. Due to the neurodegenerative consequences of dysfunctional IBA57, the IBA57 c.439C>T mutation appeared to us as the likely candidate for the cause of HNM, yet direct biochemical testing in dog material was not feasible.
In order to characterize the Kooiker dog IBA57 amino acid exchange R147W in more detail, we applied an established depletion-complementation assay in human tissue culture (Ajit Bolar et al., 2013;Debray et al., 2015;Lossos et al., 2015). Endogenous IBA57 was depleted in HeLa cells by RNAi, followed by plasmid-based expression of the mutant protein IBA57-R147W carrying the canine mutation at the homologous position. Transfection with the plasmid containing mutant IBA57 resulted in markedly less efficient complementation of the IBA57 deficiency phenotypes than the expression of wild-type IBA57, unequivocally demonstrating that the HNM-related R147W amino acid exchange strongly compromised the activity of IBA57 in the mitochondrial [4Fe-4S] cluster assembly. The strongest defects were observed in the integrity and activity of respiratory chain complexes I and II (SDH), as well as in LIAS-dependent lipoyl cofactor formation on the pyruvate dehydrogenase complex subunit PDHc-E2. In addition, an MMDS3-typical deterioration of respiratory chain complex IV (Sheftel et al., 2012;Ajit Bolar et al., 2013;Debray et al., 2015;Stehling et al., 2018b) was observed, collectively resulting in a severe impairment of mitochondrial energy metabolism. The complex IV (COX) defects generally observed upon depletion of the ISC components, which includes the ISCA and IBA57 proteins, still lack a molecular explanation (Lill and Freibert, 2020). Either this impairment is indirectly caused by defects in the supercomplex partners RCC-I and RCC-II. Alternatively, it may be explained by a maturation defect of an uncharacterized mitochondrial [4Fe-4S] protein involved in, e.g., subunit synthesis or assembly of COX. In line with the specific function of IBA57 in mitochondrial [4Fe-4S] but not [2Fe-2S] cluster formation (Sheftel et al., 2012;Lill and Freibert, 2020), the [2Fe-2S] cofactor-containing RCC-III, as well as extra-mitochondrial Fe/S proteins, is not significantly affected by the HNM-related mutation.
Whereas the IBA57-R147W substitution has not been described before as pathogenic, two substitutions of the preceding conserved arginine residue either to tryptophan (R146W) or to proline (R146P) were found homozygous or compound heterozygous, respectively, in human patients with MMDS3 (Zeng et al., 2014;Weiler et al., 2020). Plasmid-based expression of human IBA57-R147W did not indicate that the amino acid substitution substantially affected the protein's stability, similar to the human disease-causing Trp substitution of the neighboring Arg146 (Debray et al., 2015). Both arginine residues are a part of the surface-exposed loop located not far from IBA57's conserved active site cysteine (Debray et al., 2015;Camponeschi et al., 2022). Structural modeling of the R146W substitution has proposed that the alteration might interfere with the interaction of IBA57 and its partner protein ISCA2 (Camponeschi et al., 2022), and this might decrease the efficiency of the [4Fe-4S] cluster assembly. In contrast to the neighboring Arg146, residue Arg147 appears to hardly protrude from IBA57 (Gourdoupis et al., 2018) and might thus be less relevant for interactions with ISCA2, explaining why in contrast to the fatal human R146W amino acid exchange, the present HNM-related R147W substitution elicits a rather mild phenotype. The fact that this residue is not conserved in, e.g., plants, fungi, and bacteria further fits the relatively mild consequences caused by this mutation.
In addition to the IBA57 mutation, we found a nonsense mutation in OBSCN, which also co-segregated with HNM to some extent. Large obscurin isoforms are only found as constituents of the sarcomeric cytoskeleton of striated muscle cells. Smaller obscurins have been detected in several rat and mouse tissues, but the significance of these is not clear (Ackermann et al., 2018). Frameshift mutations in the central part of human OBSCN have been implicated in cardiomyopathies with an apparent dominant effect (Grogan and Kontrogianni-Konstantopoulos, 2019). Over the years, about 10% of the Kooiker dog population have been heterozygous carriers of the OBSCN stop codon mutation, yet-according to the meticulously maintained health records of the breed club (Snels, 2022)-this group was not predisposed to cardiomyopathies. Interestingly, several losses of function mutations in human OBSCN have recently been shown to be associated with a recessive form of recurrent rhabdomyolysis, clearly distinct from the neurologic phenotype of HNM (Cabrera-Serrano et al., 2022). Therefore, we did not consider the OBSCN mutation to be primarily responsible for the HNM phenotype. It is very well possible that dogs that are homozygous for the OBSCN mutation would develop rhabdomyolysis at a later age if they would have no IBA57 mutation. On the other hand, the clinical phenotype of HNM in Kooiker dogs exhibits similarities to the leukoencephalomyelopathy that occurs in Rottweilers (Gamble and Chrisman, 1984;Wouda and van Nes, 1986) and Leonbergers (Oevermann et al., 2008) where the causative mutation has been recently identified in the gene NAPEPLD (Minor et al., 2018). However, NAPEPLD is involved in the synthesis of N-acylethanolamines and not in the Fe/S protein assembly, excluding any direct involvement in the observed biochemical alterations.
Human patients with an IBA57 deficiency suffering from a leukodystrophy-like phenotype were first described in 2013 (Ajit Bolar et al., 2013). Since then, a variety of deleterious IBA57 mutations have been identified in multiple individuals (Camponeschi et al., 2022), now commonly subsumed to MMDS3. The severity of MMDS3 varies widely and is probably related to the level of residual IBA57 activity afforded by gene alterations, although large phenotypic variability has also been observed in two siblings with the same IBA57 mutations (Hamanaka et al., 2018). Based on the clinical phenotype, human MMDS3 patients may be classified into three categories (Liu et al., 2018): first is the most severe congenital type, leading to perinatal death and only observed in the two siblings; second is the most prevalent infantile type, with an age at onset of at least 4 months and variable outcome; and third is the comparably least affected childhood-type, with an age at onset of at least 3 years, slowly progressing spastic paraplegia, optic atrophy, and peripheral polyneuropathy, summarized as the SPOAN-like phenotype (Lossos et al., 2015). The childhood type was only observed in a consanguineous, multiplex family and linked to a homozygous splice site mutation leading to a low-level expression of the nonmutated IBA57 protein. The clinical manifestation of the affected Kooiker dogs appears to be as mild as in the human childhood type, suggesting that the R147W amino acid exchange affected the function of IBA57 only moderately. However, the genetic alterations causing the human SPOAN-like and the canine HNM phenotype are not comparable, which may explain why optic atrophy is not observed in dogs. The MRI examinations of human patients invariably show lesions in the white matter of the brain and sometimes of the spinal cord. In the Kooiker dogs, the white matter lesions are mainly seen in the spinal cord and rarely in the brain. This divergence may reflect different energy requirements or a dedicated vulnerability of different parts of the central nervous system, particularly among man and dog. Overall, species specificity may also explain some of these subtle differences.

Data availability statement
The data sets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found at https://www.ncbi.nlm.nih.gov/, PRJNA948354.

Ethics statement
Ethical review and approval were not required for the animal study because the dogs involved were referred patients. Written informed consent was obtained from the owners for the participation of their animals in this study.

Author contributions
PM was responsible for the identification and clinical work-up of the dogs. Statistical analysis, data analysis, and manuscript writing were performed by PM, OS, FV, RL, and PL. The cell biological and biochemical parts were performed by MV-L and OS under the supervision of RL. The molecular analysis was performed by MV-L, FV, and PL. All authors contributed to the article and approved the submitted version.

Funding
This research was supported in part by the LUPA project, which was funded by the European Commission under the Health theme of the Seventh Framework Program (FP7-LUPA, GA: 201370). RL acknowledges generous financial support from Deutsche Forschungsgemeinschaft (SFB 987 and SPP 1927).