Pathogenic mutations in NUBPL affect complex I activity and cold tolerance in the yeast model Yarrowia lipolytica

Abstract Complex I deficiency is a common cause of mitochondrial disease, resulting from mutations in genes encoding structural subunits, assembly factors or defects in mitochondrial gene expression. Advances in genetic diagnostics and sequencing have led to identification of several variants in NUBPL (nucleotide binding protein-like), encoding an assembly factor of complex I, which are potentially pathogenic. To help assign pathogenicity and learn more about the function of NUBPL, amino acid substitutions were recreated in the homologous Ind1 protein of the yeast model Yarrowia lipolytica. Leu102Pro destabilized the Ind1 protein, leading to a null-mutant phenotype. Asp103Tyr, Leu191Phe and Gly285Cys affected complex I assembly to varying degrees, whereas Gly136Asp substitution in Ind1 did not impact on complex I levels nor dNADH:ubiquinone activity. Blue-native polyacrylamide gel electrophoresis and immunolabelling of the structural subunits NUBM and NUCM revealed that all Ind1 variants accumulated a Q module intermediate of complex I. In the Ind1 Asp103Tyr variant, the matrix arm intermediate was virtually absent, indicating a dominant effect. Dysfunction of Ind1, but not absence of complex I, rendered Y. lipolytica sensitive to cold. The Ind1 Gly285Cys variant was able to support complex I assembly at 28°C, but not at 10°C. Our results indicate that Ind1 is required for progression of assembly from the Q module to the full matrix arm. Cold sensitivity could be developed as a phenotype assay to demonstrate pathogenicity of NUBPL mutations and other complex I defects.


Introduction
Defects in respiratory chain function underlie a large share of mitochondrial disorders, of which a third are due to complex I deficiency (OMIM 252010) (1)(2)(3). Clinical presentations such as Leigh syndrome usually occur in infancy or early adulthood. Symptoms include skeletal muscle myopathy, cardiomyopathy, hypotonia, stroke, ataxia and lactic acidosis (3,4). Human com-plex I has 44 different structural subunits, of which 37 are encoded in the nuclear genome and 7 in the mitochondrial genome (5). Disease-causing mutations have been identified in 27 out of 44 structural genes. However, this only provides a diagnosis for about 50% of cases, as complex I deficiency may also be caused by mutations in assembly factors, tRNAs and other genes for mitochondrial gene expression or mitochondrial DNA maintenance (2,6). Assembly factors are defined as proteins which are required for the correct assembly and function of complex I but are not present in the mature structure. So far, 16 assembly proteins have been identified for complex I, and disease-causing mutations have been found in 11 of those (3,7). Particularly, advances in genome sequencing have led to the identification of several novel genes affecting complex I function, providing a genetic diagnosis for patients with these rare inheritable conditions.
The complex I assembly factor NUBPL (nucleotide binding protein-like) was first identified in the aerobic yeast Yarrowia lipolytica where it was named Ind1 (iron-sulfur protein required for NADH dehydrogenase) (8). Genomic deletion of IND1 resulted in a major decrease in complex I, to approximately 28% of wildtype levels, but no effect on other respiratory complexes. Ind1 has been proposed to insert iron-sulfur (FeS) clusters in complex I based on its homology to the cytosolic FeS cluster assembly factors Nbp35 and Cfd1, corresponding to NUBP1 and NUBP2, respectively, in human (9). However, it should be noted that Ind1 is not essentially required for complex I in Y. lipolytica, and it would therefore only have an auxiliary role in cofactor assembly.
Phylogenetic analysis showed that the IND1 gene is present in almost all eukaryotes, and closely matches the distribution of complex I (8). RNAi knockdown of NUBPL in human HeLa cells led to a specific decrease in complex I, as well as accumulation of assembly intermediates (10). Moreover, mutations in NUBPL were identified by exome sequencing in clinically described cases of complex I deficiency (1,11,12). Recently, NUBPL was found among 191 genes that are essential for oxidative phosphorylation in a genome-wide CRISPR/Cas9 screen of human leukemia K562 cells (13). The IND1 homologue in the model plant Arabidopsis thaliana is also required for complex I assembly (14). Combined, these studies show that NUBPL has an evolutionary conserved function in complex I assembly, but its precise molecular function remains to be demonstrated.
So far, 13 cases (in 11 families) with likely clinically relevant mutations in NUBPL have been reported or newly diagnosed (Table 1). Patients display various symptoms within the broad spectrum of mitochondrial disorders, including motor problems, increased lactate levels and some degree of complex I deficiency (1,11,12). In addition, NUBPL patients show a distinct magnetic resonance imaging pattern with abnormalities of the cerebellar cortex, deep cerebral white matter and corpus callosum (11). Genetic analyses revealed that all but one patient have inherited an intronic mutation in one copy of NUBPL, c.815-27T>C, together with a likely deleterious mutation in the other copy of the gene ( Table 1). The c.815-27T>C mutation affects a splicing branch site in intron 9, leading to aberrant mRNA splicing (15). Approximately 30% of wild-type NUBPL transcript remains, but two mis-spliced transcripts are also produced. One of these disappears by nonsense mediated decay. The other mis-spliced transcript lacks exon 10, leading to Asp273Gln, a frameshift and then a premature stop codon. The predicted protein product, p.(D273Qfs*31), is unstable, and NUBPL protein is almost undetectable in patient fibroblasts (15). The c.815-27T>C variant is often co-inherited with the missense variant c.G166>A (p.Gly56Arg), which, on its own, is not thought to be pathogenic (15), and is absent in at least two cases (Table 1, patient 2 and siblings 10 and 11). Recently, a clinical case of hereditary bilateral striatal necrosis was potentially attributed to biallelic missense variants in NUBPL, but neither the branch site mutation nor c.G166>A was found (Table 1, patient 9). It was suggested that this case represents the less severe end of the spectrum of NUBPL dysfunction, compared with the reported cases with the branch site mutation (16).
The haploid yeast Y. lipolytica has been used extensively in biotechnology, but it has also been developed as a model organism to study complex I biology (17)(18)(19)(20)(21). This has been necessary because the usual yeast model Baker's yeast, Saccharomyces cerevisae, does not have complex I. The D273Qfs*31 variant of NUBPL was recreated in the Ind1 protein of Y. lipolytica, which confirmed that the altered protein was unable to support the assembly of complex I (22). Here, we compare five missense variants in the NUBPL gene, two of which are newly identified in patients, in an effort to assign pathogenicity and to enhance our understanding of the molecular function of NUBPL/Ind1. The resulting protein variants were tested in Y. lipolytica for protein stability and complex I levels, oxidoreductase activity and assembly intermediates. Together, the results indicate pathogenicity of four missense mutations and reveal that ind1 mutants are sensitive to low temperature.

Selection of NUBPL mutations for analysis in Y. lipolytica
Thirteen cases of mitochondrial disorders associated with mutations in NUBPL have been published or were otherwise known to us ( Table 1). Mutations that result in frameshifts, insertions or deletions (patients 1-4 and 7) are almost certainly deleterious but are unlikely to shed light on the molecular function of NUBPL. Instead, we focused on assessing the functional impact of patient mutations resulting in a single amino acid change at a conserved position, such as Leu104Pro (L104P, patients 9, 10, 11 and 12), Asp105Tyr (D105Y, patients 5 and 6), Leu193Phe (L193F, patient 8) and Gly287Cys (G287C, patient 13). Many more missense variants in NUBPL have been uncovered by large-scale exome sequencing (gnomad.broadinstitute.org), of which Asn198Tyr and Val182Ala occur at a relatively high allele frequency (Table 2). While Asn198 and Val182 are not conserved in Y. lipolytica Ind1, the NUBPL Gly138Asp (G138D) variant with an allele frequency of 6.73 × 10 -4 does affect a conserved amino acid, corresponding to Gly136 in Y. lipolytica Ind1, and was included in this study. G138D is predicted to be damaging to NUBPL (Table 2) and has been found in two Parkinson's disease patients (P. Eis and E. Hatchwell, personal communication).
L104P and D105Y affect residues in the highly conserved Switch I motif (Fig. 1A), which is involved in ATP hydrolysis; L193F affects a residue just outside the Mrp family signature motif; and G287C introduces an additional cysteine to the Cterminus, which may create an unwanted disulphide bridge. G138 is not in a conserved motif, but the introduction of a negatively charged Asp could affect protein folding. The corresponding amino acids in the Ind1 protein of Y. lipolytica were identified by alignment using Clustal Omega ( Fig. 1A; Table 3), and this amino acid numbering is used throughout the study. A homology model of Ind1 was made, and the positions of the substituted amino acids are indicated in Fig. 1B.

Protein stability in Ind1 variants
In order to study the biochemical and physiological effects of the amino acid changes, site-directed mutations were introduced into a plasmid carrying the IND1 gene under the control of its native promoter and transformed into Y. lipolytica ind1 cells, a GB10 strain with a genomic deletion of the IND1 gene (8). The plasmid contains a chromosomal ARS/CEN region which is  Allele frequency data from gnomAD (http://gnomad.broadinstitute.org/) for the most commonly occurring missense mutations in NUBPL, as well as the clinically relevant variants listed in Table 1. A PolyPhen-2 score close to 1 indicates that the amino acid change is "probably damaging". maintained at ∼1 copy per cell. The GB10 strain was engineered to contain the NDH2i gene, encoding an alternative NADH dehydrogenase targeted to the matrix side of the inner mitochon-drial membrane, which bypasses the essential requirement of respiratory complex I in Y. lipolytica (23). The previously reported variant Ind1 protein N271Qfs*31, recapitulating the effect of the c.815-27T>C branch-site mutation in NUBPL, was included for comparison (22). Mitochondrial membranes of each strain were isolated and subjected to Western blot analysis to visualize Ind1 protein levels. The D103Y, G136D, L191F and G285C variants displayed levels of Ind1 protein similar to ind1 + IND1 (henceforth referred to as complemented wild type, cWT) ( Fig. 2A). The L102P substitution resulted in lower levels of Ind1 protein, to approximately 45% of cWT levels. The decrease in L102P protein was similar to that seen in N271Qfs*31. Antibodies against aconitase (Aco1) and subunit 2 of complex II (Sdh2) were used as a loading control and to confirm that other FeS cluster binding proteins are not affected ( Fig. 2A).
To rule out that lower amounts of the Ind1 L102P protein were due to decreased transcript levels, expression of IND1 was assessed by reverse transcription-polymerase chain reaction (RT-PCR) (Fig. 2B). While this technique is not quantitative, at low cycle numbers, major differences in transcript levels are easily detected. Normal levels of IND1 transcript in the L102P variant indicated that the protein undergoes post-translational degradation. In order to further investigate the effect of L102P on protein stability, Ind1 and Ind1 L102P were expressed in E. coli.  Separation of soluble and insoluble fractions showed that all of the Ind1 L102P protein is found in the insoluble pellet fraction, whilst most of the normal Ind1 protein can be found in the soluble fraction. This suggests that the L102P substitution causes protein misfolding, which would result in its degradation by proteases. Overall these data show that, with the exception of L102P, the selected amino acid substitutions in Ind1 do not affect protein stability.

Ind1 variants L102P, D103Y and L191F have significantly decreased levels of complex I
It has previously been shown that cells lacking IND1 have approximately 28% of fully active complex I compared with wild type (8). To investigate the effect of the amino acid variants in Ind1 on complex I, mitochondrial membranes were separated by blue-native polyacrylamide gel electrophoresis (BN-PAGE) to resolve the respiratory complexes. Incubation of the gel with NADH and nitro-blue tetrazolium (NBT) reveals NADH dehydrogenase activities, which is used as a proxy to estimate complex I levels (24). In cells expressing the Ind1 variants L102P and D103Y, complex I was strongly decreased, similar to the levels in the ind1 mutant (empty vector control, e.v.) and N271Qfs*31 (Fig. 3A). The L191F substitution resulted in a minor decrease in complex I, whereas G136D and G285C were indistinguishable from cWT. For comparison, a knockout strain of the NUBM subunit of complex I (nubm ) displayed no complex I activity, in agreement with the FMN cofactor of NUBM being the site of NADH oxidation. Another complex I mutant, where a key catalytic residue of the NUCM subunit has been substituted (Y144F) (25), displayed complex I levels similar to cWT. This is consistent with a mutation that affects catalysis but not assembly of complex I. Complex V levels were unaffected by mutations in IND1 (Fig. 3A, lower Coomassie-stained band).
To measure the oxidoreductase activity of the matrix arm of complex I, two different spectrophotometric assays were used. Electron transfer from NADH to the artificial electron acceptor hexaammineruthenium(III) chloride (HAR) involves only the primary electron acceptor FMN bound to the NUBM subunit of complex I and, like NADH:NBT staining, serves as a proxy for complex I content. The L102P, D103Y, L191F and N271Qfs*31 variants showed a significant decrease in NADH:HAR activity, whereas G136D and G285C were similar to cWT (Fig. 3B, Table S1).
(DBQ) was measured, encompassing all cofactors in the matrix arm of complex I. A significant decrease in electron transfer was seen with the L102P, D103Y, L191F and N271Qfs*31 variants, but not in G136D and G285C (Fig. 3C, Table S1). The nucm + NUCM Y144F mutant had low dNADH:DBQ activity (Table S1), despite wild-type levels of fully assembled complex I (Fig. 3A). Comparison of dNADH:DBQ activity between the Ind1 variants and nucm + NUCM Y144F shows that in the Ind1 variants there is still substantial activity remaining. For the Ind1 variants, the NADH:HAR activities correspond with those of dNADH:DBQ, suggesting that all the complex I present is enzymatically active.
Taken together, the data from three different methods are in agreement and indicate that the L102P and D103Y substitutions in Ind1 cause a defect in complex I assembly of similar severity as IND1 deletion. The L191F substitution caused a mild decrease, whereas G136D and G285C appeared to have no significant effect on complex I levels under standard growth conditions.

All ind1 mutants accumulate a Q module assembly intermediate
In human cell lines depleted of NUBPL, a subcomplex representing part of the membrane arm was observed (10). Similarly, the Arabidopsis indh mutant had a membrane arm assembly intermediate, but no full-size complex I (14). To investigate complex I assembly in Y. lipolytica cells lacking Ind1, the ind1 strain was compared with mutant strains of different subunits of the matrix arm of complex I, nubm , nucm and nukm . NUBM is the homologue of the human NDUFV1 protein (bovine 51-kDa subunit) in the N module. NUCM (human NDUFS2, bovine 49 kDa) and NUKM (human NDUFS7, bovine PSST) are in the Q module (Fig. 4A). In the ind1 mutant, the protein levels of NUBM are slightly decreased, but the levels of NUCM are not affected (Fig. 4B). Thus, compared with the decreased levels of complex I in ind1 , NUBM and NUCM are imported into the mitochondrial matrix, but the majority is not assembled. Decreased levels of NUBM are also seen in the nucm and nukm mutants.
Mitochondrial membranes were isolated from cWT, ind1 , nubm , nucm and nukm , and the respiratory complexes were separated by BN-PAGE, followed by in-gel activity staining and Western blot analysis with antibodies against NUBM and NUCM. This confirmed the lack of complex I in the nucm and nukm mutants (Fig. 4C). In the nubm mutant, NUCM antibodies detected a weak signal around ∼1 MD (Fig. 4C, bottom  panel), likely corresponding to a small amount of complex that lacks NUBM. NADH:NBT activity staining did not detect any full-size complex I in nubm , consistent with NUBM being the site of NADH oxidation. At longer exposure times to visualize assembly intermediates, a subcomplex of approximately 500 kDa containing both NUBM and NUCM was seen in the ind1 mutant, which was also found in cWT (Fig. 4C). This subcomplex is likely to represent the full matrix arm of complex I (26). An additional assembly intermediate containing NUCM, but not NUBM, was found at approximately 160 kDa. This intermediate can also be seen in nubm and nukm , but not in cWT, and likely corresponds to the Q module. Kmita (28)). Assembly factors would likely also be bound to this structure, resulting in the observed molecular weight of ∼160 kDa. These data show that while mitochondrial membranes of ind1 cells have some fully assembled complex I and a 500 kDa intermediate, there is accumulation of a Q module intermediate.
Next, we investigated the assembly of complex I in the Ind1 protein variants. Interestingly, the G136D and G285C variants that have normal complex I levels (Fig. 3A) visibly accumulated the Q module intermediate containing NUCM (Fig. 5B, lower  panel). Generally, the pattern of assembly intermediates containing NUCM is similar in all mutants, except that Ind1 D103Y accumulated relatively more Q module intermediate and very little of the 500 kDa intermediate. For intermediates containing the NUBM subunit, variants with low levels of complex I (N271Qfs*31, L102P and D103Y) closely resembled the empty vector control, whereas those with near normal complex I levels (G136D, L191F, G285C) more closely resemble cWT (Fig. 5B, upper panel).

ind1 mutants are sensitive to cold
During storage of Y. lipolytica strains, it was noticed that the ind1 mutant was sensitive to cold, exacerbating the mild growth defect. To investigate this further, the growth of ind1 mutants, as well as nucm + NUCM Y144F , nubm and nukm , was compared at 28 • C and 10 • C. Using a standard drop assay for yeast growth, cells were spotted onto agar plates in a dilution series. Each cell seeds a new colony, the diameter of which is determined by cell division rate and cell size. After incubation of the plates overnight at 28 • C plates to initiate growth, they were placed either at 28 • C for a further 2 days or moved to 10 • C for 10-14 days. At 28 • C all strains grew similarly, despite slight variations in colony size (Fig. 6, left panels). However, when grown at 10 • C, some strains displayed a dramatic growth retardation (Fig. 6, right panels). The ind1 strain expressing wild-type IND1, Ind1 variant G136D, nubm and nukm all displayed significant growth at 10 • C. However, all other Ind1 variants displayed very slow or no further growth at 10 • C. Interestingly, the nubm and nukm mutants that lack complex I completely were able to grow in cold conditions, but nucm + NUCM Y144F with an inactive form of complex I could not grow, similar to the ind1 mutant.
These data reveal a novel growth difference of certain complex I mutants grown at normal (28 • C) or cold (10 • C) conditions. Ind1 variants, except for G136D, also display this temperaturedependent growth defect.

Complex I assembly in the Ind1 G285C variant is temperature sensitive
The Ind1 variant G285C has normal levels and activity of complex I (Fig. 3A). However, it accumulated the Q module intermediate (Fig. 5B), and growth at 10 • C was impaired (Fig. 6). To investigate if assembly of complex I in the G285C variant is conditional on temperature, mitochondrial membranes were isolated from cells producing wild-type Ind1 (cWT) and the Ind1 G285C variant after overnight growth at 10 • C and compared with samples grown at 28 • C. Mitochondrial membranes were separated by BN-PAGE and stained with NADH:NBT to assess formation of complex I. cWT displayed full complex I assembly in cultures grown at both 28 • C and 10 • C (Fig. 7A). By contrast, the G285C Ind1 variant could not support complex I assembly when grown at 10 • C, with complex I levels decreased to those in the ind1 mutant (Fig. 3A). BN-PAGE followed by Western blotting and labelling with antibodies against NUBM confirmed the significant decrease in complex I (Fig. 7A). Immunodetection of Ind1 showed that the G285C protein variant is stable at 10 • C, meaning that the decrease in complex I levels is not due to a cold-dependent degradation of the Ind1 G285C protein.
Next, the effect of cold conditions on assembly intermediates of complex I was assessed. Mitochondrial membranes were separated by BN-PAGE followed by Western blotting with antibodies against NUCM. A short exposure of the signal confirmed the decreased levels of fully assembled complex I in cells expressing the G285C Ind1 variant grown at 10 • C (Fig. 7B,  top panel). A longer exposure revealed a different pattern of complex I assembly intermediates in mitochondria isolated at normal and cold temperatures (Fig. 7B). Ind1 G285C samples from cold conditions lacked a band corresponding to the assembly intermediate at 500 kDa, made up of the N and Q modules, and an increase in the Q module intermediate at 160 kDa. This suggests that the G285C Ind1 variant accumulates greater levels of the Q module at cold temperatures, as it is unable to assemble it into the 500 kDa intermediate. There is still a strong signal at the size corresponding to complex I; however, this is due to a long exposure time needed to see assembly intermediates. Taken together, these data show that the Ind1 G285C variant is more severely impaired in complex I assembly when grown in cold conditions.

Discussion
Since NUBPL was associated with mitochondrial disease resulting from complex I deficiency (1,11,12), the gene is routinely included in the list of candidate genes for genetic diagnoses. In children's hospitals around the world, additional mutations have been found in NUBPL that are likely to be pathogenic. In this study five protein variants of human NUBPL were recreated in the homologous Ind1 protein in the yeast Y. lipolytica to study the effect of the amino acid changes on complex I assembly and to confirm pathogenicity.
Ind1 variants L102P and D103Y showed the strongest decrease in complex I, similar to the effect of N271Qfs*31 and deletion of the IND1 gene (Fig. 3). The L102P substitution destabilized the Ind1 protein (Fig. 2), which most likely explains the complete loss of functionality. The D103Y substitution does not affect protein stability, suggesting that the amino acid change renders Ind1 inactive. Asp103 is situated at the Nterminal end of the Switch I motif (Fig. 1A). In molecular motors driven by ATP (or GTP) hydrolysis, the Switch I and II sequences form the nucleotide binding pocket, while a conserved lysine residue of the Walker A motif interacts with ATP (29). Upon ATP hydrolysis, Switch I and II undergo large conformational changes, facilitating monomer-dimer transitions (e.g. in ParA and MinD in bacterial cell division) or interactions with other proteins (e.g. NifH in nitrogenase). D103 aligns with D38 in MinD of Escherichia coli, which is the residue required for dimerization of MinD as well as for interaction between MinD and MinC (30). NUBPL and other members of the NBP35/Mrp subfamily of P-loop NTPases also form dimers (9). Lack of dimer formation or lack of interaction with another protein, for example a subunit of complex I, is likely to fully abolish the function of NUBPL/Ind1. Interestingly, the Ind1 D103Y variant appears to have a block in assembling the N module onto the Q module, as very little matrix arm assembly intermediate was observed and high levels of Q module accumulated (Fig. 5B). Possibly, a failure to use ATP hydrolysis for conformational changes of Ind1 may have a dominant effect and trap a specific assembly intermediate.
The Ind1 L191F variant caused a moderate decrease in complex I levels and activity. This residue is close to the Mrp family signature (Fig. 1A) found only in the subfamily of NBP35/Mrp Ploop NTPases. Members of this subfamily are found in archaea, bacteria and eukaryotes. The precise molecular function of this amino acid motif is not known.
The G136D substitution in Ind1 had no significant effect on complex I levels and redox activities (Fig. 3); however, the variant protein caused accumulation of the Q module assembly intermediate, albeit at low levels (Fig. 5). The equivalent G138D variant in human NUBPL has so far not been associated with a clinical case of complex I deficiency even though it is found at ∼5-fold higher frequency (Table 2) than the L104P variant (to date, the most frequent complex I deficiency missense mutation that has been found). Both G138D and c.815-27T>C variants have been observed in the heterozygous state in Parkinson's disease patients (P. Eis and E. Hatchwell, personal communication), albeit not at a significantly higher frequency compared with publicly reported subjects (gnomad.broadinstitute.org). Since the allele frequency for G138D and c.815-27T>C are moderately high (Table 2), homozygotes or compound heterozygotes for these variants are likely to exist, and they may have late onset symptoms linked to mild complex I deficiency that are clinically relevant in diseases involving mitochondrial dysfunction, such as Parkinson's disease, as suggested by others (15,31).
It is possible that in the case of 'mild' IND1 mutations, effects on complex I assembly only manifest themselves under certain conditions, as for the Ind1 G285C variant. At the normal growth temperature of Y. lipolytica (28 • C), there was no significant effect on complex I levels and activities, although accumulation of the Q module assembly intermediate was observed (Fig. 5). At low temperature (10 • C), the Ind1 G285C variant showed a dramatic decrease in complex I and increased levels of the Q module intermediate (Fig. 7). Glycine 285 is located at the end of two short alpha helices (Fig. 1C), and a cysteine may disrupt this structure. Alternatively, the cysteine, if exposed, may form an unwanted disulfide bridge with another cysteine, in particular the cysteines of the CxxC motif, which are important for the function of NUBPL/Ind1 (8,10). A Gly to Cys substitution in Nfu1, a mitochondrial protein involved in FeS cluster assembly that also carries a CxxC motif, causes a dominant genetic effect in yeast (32) and severe mitochondrial disease in humans (33). Why the effects of G285C are exacerbated at cold temperatures remains an open question.
At the normal growth temperature of 28 • C, Y. lipolytica containing Ind1 variants and complex I mutants grew similarly to cWT, because NADH oxidation in these strains is maintained by expression of ND2i, a single subunit NADH dehydrogenase on the matrix side. However, at 10 • C, functionally compromised Ind1 variants and nucm + NUCM Y144F displayed almost no growth, whereas the nubm and nukmΔ strains were comparable with cWT in cold tolerance (Fig. 6). An initial hypothesis for this difference is that strains with a non-functional complex I produce large amounts of reactive oxygen species (ROS) such as superoxide (5), which may be exacerbated in the cold. Interestingly, lack of the accessory NDUFS4 subunit of complex I in the model plant Arabidopsis was associated with increased ROS production and diminished cold tolerance (34). The nukmΔ mutant lacks complex I completely, and the nubm lacks the FMN site where most ROS is produced. Further investigation is needed to uncover the cause of this growth defect.
In summary, of the five NUBPL missense variants that were functionally assessed in Y. lipolytica, four (L104P, D105Y, L193F and G285C) were found to be deleterious to complex I and one (G138D) was found to exhibit a mild defect. Along with other studies (18,21,35,36), these results reconfirm the utility of Y. lipolytica as a model for human mitochondrial disorders. However, care must be taken in over-interpreting the results, as some effects seen in Y. lipolytica have not been found in humans. In human NUBPL RNAi HeLa lines, the levels of NDUFV1 subunit are substantially decreased (10), whereas in the Y. lipolytica ind1 mutant, the levels of the homologous NUBM protein are only slightly less than in cWT. Secondly, the Ind1 D103Y variant appears to have stable protein levels in Y. lipolytica, but human patients with D105Y have almost no NUBPL protein (11). Ideally, all clinically relevant NUBPL mutations should also be tested in human cell lines, although Y. lipolytica is useful as a cost-effective model to study potentially pathogenic mutations affecting complex I.
A defect in Q module assembly in Y. lipolytica ind1 mutants is consistent with studies of NUBPL in human cell lines (10,11). The Q module has 3 FeS clusters that may be specifically inserted by NUBPL. However, detailed proteomic analysis of complex I assembly intermediates did not find NUBPL associated with any subcomplexes; it only occurred as a dimer (37). Possibly, NUBPL interacts with an individual Q module subunit, or the protein interaction with the Q module assembly intermediate is transient. The Ind1 protein variants characterized in this study, in particular the D103Y variants, could serve as a tool to unravel exactly which step in complex I assembly is mediated by NUBPL/Ind1.

Molecular cloning and mutagenesis
The IND1 sequence from pUB4-IND1-strep (8) was cloned into the pGEM-T Easy vector (Promega, Madison, WI USA) and used for mutagenesis using a QuikChange ® II kit (Agilent, Santa Clara CA, USA), as per the manufacturer's instructions. Mutagenesis primers were designed using the online QuikChange primer design tool, hosted at http://www.genomics.agilent.com/ primerDesignProgram.jsp (Table S2). Mutated IND1-strep sequence was inserted into pUB4, between XbaI and NsiI restriction sites. Plasmids were confirmed by sequencing and diagnostic restriction digest.

Recombinant protein expression
IND1 sequences encoding the Ind1 variants L102P and D103Y, excluding the first 36 amino acids, were cloned from pUB4 into pET15b. Ind1-strep was expressed and purified as described in Bych et al. (8). Briefly, plasmids were transformed into E. coli Rosetta containing a plasmid with the genes for ironsulfur cluster assembly (pISC) and plasmidpLYS. Colonies were grown overnight at 37 • C in Lysogeny broth (LB) media, with required antibiotics, and then diluted 50× in 100 ml Terrific broth [47.6 g l -1 Teriffic broth, 0.8 ml 50% (v/v) glycerol] and grown until OD600 = 0.6. Protein expression was induced by addition of 1 ml l -1 benzylalcohol, 50 μm Fe-ammonium citrate, 100 μm L-cysteine and 1.2 ml l -1 isopropyl β-D-1-thiogalactopyranoside (IPTG). Cells were grown at 20 • C overnight, then harvested by centrifugation at 5000 × g and flash frozen in liquid nitrogen. Cells were then resuspended in buffer W (100 mm Tris-HCl pH 8.0, 150 mm NaCl) with 0.2% (w/v) dodecyl maltoside. The cell suspension was sonicated and separated into soluble and insoluble fractions by centrifugation at 16 100 × g.

Protein blot analysis
Mitochondrial membranes were mixed with Laemmli buffer [2% (w/v) SDS, 125 mm Tris-HCl pH 6.8, 10% (w/v) glycerol, 0.04% (v/v) β-mercaptoethanol, and 0.002% (w/v) bromophenol blue] and separated on a 15% SDS-PAGE gel. Proteins were transferred under semi-dry conditions to nitrocellulose membrane (Protran TM ). Equal loading and transfer was confirmed using Ponceau S stain. Proteins were labelled with antibodies and detected using secondary horseradish peroxidase-conjugated antibodies and chemiluminescence. Mouse monoclonal antibodies against the NUBM and NUCM subunit complex I were a gift from Volker Zickermann. Rabbit polyclonal antibodies against Sdh2 and Aco1 were raised against recombinant Saccharomyces cerevisiae proteins and previously reported (40); antibodies against Y. lipolytica Ind1 are described in Bych et al. (8).
Protein was transferred from the BN-PAGE gel to a PVDF membrane (0.2 μm, Millipore TM ) using wet transfer and BN cathode buffer, for 600 min at 40 mA (maximum voltage 100 V). After transfer, immunolabelling was carried out as described above.

NADH dehydrogenase assays
BN gels were stained for complex I, as described in Sabar et al. (24). Briefly, BN gels were equilibrated in 0.1 M Tris-HCl pH 7.4 and then incubated with new buffer containing 0.2 mm NADH and 0.1% (w/v) NBT. After the desired staining intensity was reached, the reaction was stopped using a solution of 45% (v/v) MeOH and 5% (v/v) acetic acid. Destaining using a solution of 45% (v/v) MeOH and 5% (v/v) acetic acid was performed for optimal enzyme stain visualization.
NADH:HAR oxidoreductase activity was assayed as described (8). Activity for 25 μg mitochondria was measured in a 1 ml volume of 20 mm HEPES-NaOH pH 8.0, 2 mm NaN 3 , 0.2 mm NADH, 2 mm HAR and 40 μg ml -1 alamethicin. Measurements were made in a 1-ml quartz cuvette using a UV/Vis spectrophotometer (Jasco, V-550) and started after addition of HAR. dNADH:DBQ oxidoreductase activity was measured as dNADH oxidation activity in the presence of the ubiquinone analogue n-decylubiquinone (DBQ) as an electron acceptor. Activity for 50 μg mitochondria was measured in a 1 ml volume of 20 mm MOPS-NaOH pH 7.4, 50 mm NaCl, 2 mm KCN, 0.1 mm dNADH and 0.1 mm DBQ. Measurement started after addition of DBQ. 0.01 mm of the complex I inhibitor Piericidin A was added at the end of the reaction to ensure that dNADH:DBQ oxidoreductase activity is due to complex I.

Supplementary Material
Supplementary Material is available at HMG online.