Function of Conserved Acidic Residues in the PSST Homologue of Complex I (NADH:Ubiquinone Oxidoreductase) from Yarrowia lipolytica *

Proton-translocating NADH:ubiquinone oxidoreductase (complex I) is the largest and least understood enzyme of the respiratory chain. Complex I from bovine mitochondria consists of more than forty different polypeptides. Subunit PSST has been suggested to carry iron-sulfur center N-2 and has more recently been shown to be involved in inhibitor binding. Due to its pH-dependent midpoint potential, N-2 has been proposed to play a central role both in ubiquinone reduction and proton pumping. To obtain more insight into the functional role of PSST, we have analyzed site-directed mutants of conserved acidic residues in the PSST homologous subunit of the obligate aerobic yeast Yarrowia lipolytica. Mutations D136N and E140Q provided functional evidence that conserved acidic residues in PSST play a central role in the proton translocating mechanism of complex I and also in the interaction with the substrate ubiquinone. When Glu89, the residue that has been suggested to be the fourth ligand of iron-sulfur center N-2 was changed to glutamine, alanine, or cysteine, the EPR spectrum revealed an unchanged amount of this redox center but was shifted and broadened in the gzregion. This indicates that Glu89 is not a ligand of N-2. The results are discussedin the light of structural similarities to the homologous [NiFe] hydrogenases.

eral arm protruding into the mitochondrial matrix. The seven central, nuclear-coded subunits are part of the peripheral arm (8,9) and include the subunits that bear all known redox groups of complex I, namely one noncovalently bound FMN and at least six iron-sulfur centers (10). The membrane arm contains the seven hydrophobic, mitochondrially encoded subunits.
The mechanism that couples proton translocation across the membrane to electron transfer from NADH through FMN and several iron-sulfur centers onto ubiquinone is unknown. All postulated mechanisms are rather speculative (11). However, it seems likely that iron-sulfur center N-2 is the immediate electron donor to ubiquinone. The pH-dependent midpoint potential of N-2 might indicate that this redox center is directly involved in proton pumping (11)(12)(13). However, the location of iron-sulfur center N-2 is still controversial because two approaches to resolve this issue gave contradictory results. Sitedirected mutagenesis in E. coli suggests the PSST homolog (14), and a similar study in Rhodobacter capsulatus (15) suggests the TYKY homolog as the most likely candidates to bind N-2.
These studies suffer from the fact that prokaryotic complex I tends to be rather unstable (16) and is down-regulated in deficient mutants from E. coli, the only bacterium that has an enzyme stable enough to be purified (17). Pure complex I from mitochondria of bovine heart and the fungus Neurospora crassa can be easily obtained in large quantities. However, genetic manipulation in these organisms is either impossible or rather difficult (18). Therefore, we have established the obligate aerobic yeast Yarrowia lipolytica as a model to study the structure and function of complex I. In contrast to brewer's yeast Saccharomyces cerevisiae, Y. lipolytica does contain complex I 1 and most of the powerful genetic tools that are routinely used in S. cerevisiae are available (19). Here we report the first application of this novel yeast genetic system to study the function of complex I.
To identify functionally important residues in the PSST homologue subunit of the Y. lipolytica complex I, we have introduced a series of single amino acid exchanges and replaced acidic residues that are strictly conserved among all known sequences. The altered protein complexes were analyzed for complete assembly in mitochondrial membranes and as purified enzymes. Functional changes were analyzed by enzyme kinetics, inhibitor sensitivity, and EPR spectroscopy. from New England Biolabs, and T4 DNA ligase was from Life Technologies, Inc. Hybond N ϩ membranes were obtained from Macherey and Nagel. The ABI Prism dye terminator cycle sequencing kit was purchased from Perkin-Elmer. NBQ 2 was synthesized according to the protocol of (20). N-methyl-N-(3, 4-dimethoxybenzyl)-4-(p-tert-butylphenoxy)benzamide (21)  Deletion of NUKM, the Y. lipolytica Homologue of PSST-The NUKM gene 1 was deleted by homologous recombination with a LEU2-marked deletion allele. A 3.5-kb genomic PstI/PstI fragment was subcloned into a pBluescript 228 SK(Ϫ) derivative from which part of the polylinker region had been deleted by digestion with KpnI and HindIII, blunt ending using Klenow polymerase, and religation. A region of roughly 760 base pairs, encompassing nearly the complete NUKM open reading frame except for three amino acids at the 3Ј-end and about 140 base pairs of the 5Ј-flanking region were removed by digestion with HindIII and KpnI and replaced with a 1.9-kb fragment carrying the complete LEU2 gene from Y. lipolytica. The resulting 4.7-kb PstI fragment in which the orientation of the LEU2 gene was opposite to the original NUKM gene was used for transformation of diploid Y. lipolytica GB1 cells. 26 LEU2 colonies were screened for homologous recombination at the chromosomal NUKM locus by polymerase chain reaction using combinations of inward primers derived from sequences outside the 4.7-kb PstI fragment (nPSST/1 and nPSST/2) and outward primers derived from the LEU2 sequence (ylleu2/us and ylleu2/ds) and by Southern blot analysis. Two heterozygous deletion mutant strains (PA1.1 and PA1.2) were identified.

Materials-Taq
Site-directed Mutagenesis of the PSST Homologue of Complex I-Point mutations were created using the QuikChange site-directed mutagenesis kit from Stratagene, using as a template a 2.3-kb EcoRI/ EcoRI fragment subcloned into the replicative vector pINA443 carrying the selection marker URA3 (in the following termed pNUKM mut / URA3). Mutated plasmids were confirmed by DNA sequencing and used for transformation of the heterozygous deletion strain PA1.2. Haploid strains carrying the desired mutation were obtained by sporulation followed by random spore analysis. LEU2, URA3 spores were tested for the absence of the genomic copy of NUKM by polymerase chain reaction or Southern blotting and the mutation was reconfirmed by sequencing of the entire open reading frame on the plasmid.
Analytical Methods-Protein was determined according to a modified Lowry protocol (22). Blue native polyacrylamide gel electrophoresis was as described in Ref. 23.
Purification of Complex I-NADH:ubiquinone oxidoreductase was purified by stepwise extraction of the mitochondrial membranes with dodecyl maltoside followed by ion exchange and size exclusion chromatography. 1 EPR Spectroscopy-Low temperature EPR spectra were obtained on a Bruker ESP 300E spectrometer equipped with a liquid helium continuous flow cryostat, ESR 900 from Oxford Instruments. Samples were mixed with NADH in the EPR tube and frozen in liquid nitrogen after a 30-s reaction time. The protein concentration of purified complex I was 3-9 mg/ml.
Determination of Catalytic Activities-Steady state dNADH:NBQ activity was recorded in a Shimadzu UV-300 spectrophotometer as NADH oxidation (⑀ 340 -400 nm ϭ 6.1 mM Ϫ1 cm Ϫ1 ) using a thermostatted cuvette (30°C). 100 M dNADH and 50 g/ml unsealed mitochondrial membranes were added to buffer containing 50 mM Tris/HCl, pH 7.4, and 2 mM KCN. The catalytic reaction was started by the addition of 60 M NBQ (25). Complex I activity could be selectively measured with dNADH as a substrate, because as previously shown for other species (26), it is not oxidized by alternative NADH:ubiquinone oxidoreductases. Inhibitors were added to the cuvette prior to NBQ. Michaelis-Menten parameters were determined by varying the concentration of NBQ (1-100 M). The pH dependence of the catalytic rate was measured in a multibuffer containing 10 mM each of Mes, Mops, Epps, Ches, and Caps and 2 mM NaN 3 to cover a pH range from 5.0 to 10.0. The buffer was adjusted to the appropriate pH by adding NaOH. Data were fitted according to the following equation (27), where Rate is the observed catalytic rate in the presence of saturating substrate concentrations (100 M dNADH, 60 M NBQ), [H ϩ ] is the concentration of protons, K A and K B are the dissociation constants of two protonable groups A and B, and C is the optimal catalytic rate that would be observed if group A was deprotonated and group B was protonated in all enzyme molecules. Detergent-and inhibitor-insensitive dNADH:HAR activity was measured using 200 M dNADH and 2 mM HAR, 2 mM NaN 3 in 20 mM Na ϩ /Hepes, pH 8.0 at 30°C (28). The reaction was started by the addition of 50 g/ml unsealed mitochondrial membranes.
Kinetic data were analyzed using the Psiplot software package version 4.61 (Poly Software International). The numerical procedure used to fit experimental data was the Marquardt algorithm (29).

Sequence Properties of the NUKM Gene Product-N-terminal
sequencing by Edman degradation identified SAPAGT as the first six amino acids of the mature protein. 1 Thus, 27 amino acids represent the mitochondrial import presequence, and the molecular mass of the mature protein is 20.3 kDa. Fig. 1 shows a partial sequence alignment of the PSST homologue from Y. lipolytica with bovine, fungal, and bacterial proteins. As only one of the two adjacent conserved cysteins in positions 85 and 86 could serve as a ligand for sterical reasons the motif CXXE-(X) 60 -C-(X) 29 -CP has been suggested to bind iron-sulfur center 2 The abbreviations used are: NBQ, n-nonylubiquinone; kb, kilobase;

N-2 (30).
Deletion and Site-directed Mutagenesis of NUKM-The reading frame of the NUKM gene was deleted by homologous recombination with a LEU2 marked deletion allele (Fig. 2) in the diploid Y. lipolytica strain GB1. Replicative plasmids carrying site-directed mutations (Table I), which altered four conserved acidic residues, including the putative glutamic acid ligand for the iron-sulfur center N-2 (Glu 89 in Y. lipolytica PSST) were used for transformation of the diploid deletion mutant. Haploid strains carrying only the mutated copy of NUKM on the plasmid were obtained through sporulation and subsequent random spore analysis.
Complex I Content and Characterization of Mitochondrial Membranes from Y. lipolytica-Specific dNADH:HAR activities of Y. lipolytica mitochondrial membranes were in the same range for all seven strains (Table I). Among the representative membrane preparations listed in Table I, only the membranes from mutants E89C and E89A that were prepared in parallel batches exhibited a 45% higher activity. As the nonphysiological dNADH:HAR activity only depends on a functional 51-kDa subunit (31), this demonstrated that the complex I content of the membranes was not dramatically altered by the mutations. The observed differences in HAR activity were obviously because of variations in growth conditions and quality of the membranes. Loading unsealed mitochondrial membranes from all mutant strains corresponding to equal amounts of total protein on blue native polyacrylamide electrophoresis gels confirmed that complex I was fully assembled and expressed at comparable levels in the parental and all mutant strains (data not shown).
To identify functionally relevant alterations in complex I, Michaelis-Menten parameters and pH dependence of dNADH: NBQ oxidoreductase activity and sensitivity toward several representative complex I inhibitors were determined. For membranes from the parental strain, a K m value close to 14 M was determined for the hydrophobic ubiquinone analog NBQ (Table I). The K m value for NADH has been reported earlier (24) and was not monitored in this study as the NADH binding site is known to be located on the 51-kDa subunit (32,33). The pH dependence of dNADH:NBQ activity in Y. lipolytica membranes was a bell-shaped curve with an optimum at around pH 7.6 (Fig. 3). Like for mitochondrial cytochrome bc 1 complex (27), this pH dependence could be simulated by assuming that two protonable groups A and B control the steady-state activity of complex I and that the optimal catalytic rate C would be reached if group A was deprotonated and group B was protonated in all enzyme molecules (see Equation 1 under "Experimental Procedures"). For the parental strain, pK A was 6.4 Ϯ 0.1 and pK B was 8.8 Ϯ 0.1 (Table II).
Y. lipolytica complex I was efficiently inhibited (Table III) by the type A inhibitors piericidin A and DQA and the type B inhibitor rotenone (see Ref. 34 for the classification of complex I inhibitors). Whereas the I 50 values for these three compounds were in the same range in Y. lipolytica membranes as for bovine heart submitochondrial particles (25), up to 10 M of an efficient type C inhibitor, the capsaicin derivative N-methyl-N-(3, 4-dimethoxybenzyl)-4-(p-tert.-butylphenoxy)-benzamide (21), did not have any effect on the activity of complex I from Y. lipolytica (data not shown).
We also monitored the low temperature EPR spectra of isolated complex I reduced with NADH (Fig. 4) that have been discussed in detail elsewhere. 1 Under these test criteria, characteristic alterations could be detected in all mutants analyzed here (see below) with the notable exception of D168N (Tables I and II), which exhibited virtually the same properties as the parental strain. This demonstrated that full complementation was achieved by the copy of the NUKM gene on the plasmid.
PSST Mutations D136N and E140Q Affect the Catalytic

FIG. 2. Construction of the nukm::LEU2 deletion allele (A) and deletion of the genomic copy by homologous recombination (B).
The NUKM open reading frame is shaded in dark gray. Polymerase chain reaction primer pairs ylleu2/us, nPSST/1 and ylleu2/ds, nPSST/2 were used to screen LEU2 colonies for marker insertion by homologous recombination. Restriction sites: E, EcoRI; H, HindIII; K, KpnI; P, PstI. To account for variations of complex I content in different batches of mitochondrial membranes the dNADH:NBQ activities were normalized to dNADH:HAR activities that were not affected by the mutations and reflected the complex I content.
FIG. 3. pH dependence of specific activity. The steady-state rates of dNADH:NBQ oxidoreductase activity were measured covering a pH range from 5.0 to 10.0 using unsealed mitochondrial membranes (50 g/ml total protein) at saturating substrate concentrations as described under "Experimental Procedures." The parameters used to plot the fitted curves are listed in Table II. q, parental strain; f, mutant strain E140Q; OE, mutant strain D136N.
Function of Complex I-The maximal dNADH:NBQ oxidoreductase activities at pH 7.4 for mitochondrial membranes from mutants D136N and E140Q were significantly decreased to about 20 and 50%, respectively, as compared with the parental strain, but the K m for NBQ was only slightly affected (Table I). Mutations D136N and E140Q also resulted in a narrowed pH optimum for the steady-state activity (Table II  and Fig. 3). pK A was shifted up by about half a pK unit to 7.0 and 6.8, respectively, and pK B was shifted down by about the same extent to 8.5 and 8.3, respectively (Table II). The I 50 values for DQA (type A inhibitor) shown in Table III revealed for both mutations a slight but significant resistance, whereas a 4 -5-fold hypersensitivity was observed for rotenone (type B inhibitor). Exchanging these two conserved acidic residues with the corresponding amide had no effect on the EPR spectra of the reduced iron-sulfur centers of complex I (data not shown).
Glu 89 Is Not a Ligand of Iron-Sulfur Center N-2-Exchanging Glu 89 , the glutamate that has been proposed as the forth ligand of iron-sulfur center N-2, with three different amino acids had only moderate effects on complex I activity in mitochondrial membranes (Table I). Complex I from all three mutants showed a normal pH profile that could be fitted with the same pK values as the enzyme from the parental strain (Table  II).
Whereas normal amounts of iron-sulfur center N-2 could be detected by EPR spectroscopy in all three mutant enzymes, a clear shift of the g z signal to higher and of the g xy signal to lower values (Fig. 4A) was observed. The 16 K spectra shown in Fig. 4B also revealed that in parallel to this shift the g z line of N-2 was significantly broadened. The corresponding line widths (L Z ) are listed in Table IV. Moreover, the power saturation characteristics at 12 K of center N-2, expressed as half saturation parameters (P 1 ⁄2), were altered in all three Glu 89 mutants (Table IV). Remarkably, the most conservative exchange E89Q had the most pronounced effect and was the only mutation that also caused a slight decrease of the K m for ubiquinone (cf. Table II). In contrast, exchanging Glu 89 for the potential iron-sulfur ligand cysteine and the small hydrophobic alanine caused much smaller but similar changes of the EPR parameters of center N-2 ( Fig. 4 and Table IV). As replacing Glu 89 by three different residues had only minor effects on the EPR spectra of the iron-sulfur center N-2, we conclude that it is not a ligand of this redox group. DISCUSSION With the work presented here we have established Y. lipolytica as a powerful yeast genetic model for the analysis of mitochondrial complex I. We have demonstrated efficient deletion of a gene for a nuclear coded subunit of complex I by homologous recombination using the LEU2 marker gene and the introduction of site-directed mutations using a replicative plasmid.
PSST, the subunit studied here, is one out of five complex I subunits that are homologous to subunits of bacterial membrane-bound [NiFe] hydrogenases (35) (Fig. 5). Those two that are conserved between soluble (36, 37) and membrane-bound [NiFe] hydrogenases, and complex I may be regarded as the catalytic core of complex I. The large subunit of soluble [NiFe] hydrogenases harbors the binuclear [NiFe] active site and is homologous to the 49-kDa subunit of complex I, and the small iron-sulfur subunit is related to PSST. PSST can be labeled specifically with a photoreactive pyridaben derivative (38) ,   FIG. 4. EPR spectra of purified complex I. EPR spectra of purified complex I from Y. lipolytica strains carrying mutations of the putative N-2 ligand Glu 89 are compared with spectra obtained with parental enzyme. Complex I was reduced with 6 mM NADH. A, EPR spectra of complex I obtained at 16 K; the g z and g xy signal positions for center N-2 are indicated; B, enlarged view of the g z region for iron-sulfur center N-2. EPR conditions: microwave frequency, 9.48 GHz; microwave power, 2 milliwatts; modulation amplitude, 1 millitesla (mT). whereas a piericidin A-resistant mutant of R. capsulatus was found to reside in the 49-kDa subunit (39).
In the present work, we have studied a series of site-directed mutations of acidic residues in the subunit of Y. lipolytica complex I homologous to PSST to gain insight into the proposed functional role of this critical subunit. Exchanging Asp 136 with the corresponding amide resulted in fully assembled and stable complex I that exhibited specific functional alterations when compared with the parental enzyme, namely lowered catalytic activity, slight resistance to DQA and hypersensitivity to rotenone. Most notable was a pronounced narrowing of the pH profile for the steady-state activity that could be expressed as shifts of 0.3-0.6 units in two pK values fitted to an equation that quantitatively described the pH dependence of activity (Fig. 3). Qualitatively, mutation E140Q had the same but somewhat less pronounced effects. Global structural changes caused by the mutations could be ruled out, as the EPR spectra of all iron-sulfur centers and especially of center N-2 were not altered at all. The fact that removing either one of these two adjacent conserved acidic residues caused such well defined and similar changes seems to indicate that Asp 136 and Glu 140 , which are not in the vicinity of any residues previously proposed to be functionally important within PSST, play a specific role in the mechanism of complex I. Although the limited effect on the catalytic rate excludes that these two residues correspond to the two pK A values controlling catalytic activity, it is tempting to speculate that they might contribute to a channel that transfers protons to or away from center N-2. The changed affinity for complex I inhibitors fits well with covalent labeling of this subunit with a derivative of pyridaben (38).
It is still controversial whether iron-sulfur center N-2 is located in subunit PSST or TYKY as removing the putative cysteine ligands by site-directed mutagenesis of TYKY in R. capsulatus (15) and PSST in E. coli (40) gave conflicting re-sults. Subunit PSST lacks a canonical motif for Fe 4 S 4 coordination, because only three of the four cysteine residues ligating the corresponding iron-sulfur center in the small subunits of [NiFe] hydrogenases (36,37) are conserved. However, it has been speculated that the conserved glutamic acid in position 89 downstream of the first cysteine may represent the fourth ligand for center N-2 (30) (see Fig. 1). This option can be excluded by the data presented here. If one assumes that despite the rather weak sequence homology between the small subunit of [NiFe] hydrogenases and PSST, the overall folding of the protein was preserved during evolution; the residue corresponding to Glu 89 in the structure of the hydrogenase from Desulfovibrio fructosovorans (36), or Desulfovibrio gigas (37) is found to be close but on the "wrong" side with respect to the location of a potential iron-sulfur center. In fact, based on this structural homology it is very difficult to identify conserved residues in PSST that could replace the cysteine missing in position 83 of the Y. lipolytica protein. As deduced from the hydrogenase fold, Asp 168 is the conserved acidic residue that comes closest to the correct position near the putative ironsulfur center in PSST. However, we could not find that replacing this residue by an asparagine had any effect on complex I and its EPR spectra. Thus, if N-2 were to be in PSST, the fourth ligand has yet to be identified.
Similarly, from the observed shift and broadening of the EPR line shape of iron-sulfur center N-2 by mutations in position Glu 89 it does not follow that N-2 is in PSST as this could also be interpreted as an indirect effect on the redox-center residing in TYKY. The membrane-bound [NiFe] hydrogenase from Methanosarcina barkeri (35) contains homologues of both of these subunits (Fig. 5), which both carry iron-sulfur centers, one Fe 4 S 4 center in the PSST and two Fe 4 S 4 center in the TYKY homolog. Apparently, already acquisition of a subunit homologous to TYKY by these membrane-bound hydrogenase was accompanied by the loss of two of the iron-sulfur centers found in the PSST homologous subunit of soluble hydrogenases. Therefore, the two Fe 4 S 4 centers in TYKY could have taken over the function of these iron-sulfur centers. This hypothetical arrangement is likely to be found in complex I and makes it very likely that TYKY and PSST are in close contact to each other. This would position the Fe 4 S 4 centers in TYKY within a couple of angstroms to Glu 89 in complex I and may also explain why studies with site-directed mutations in the more fragile bacterial enzymes lead to contradictory results.
Overall, our analysis of site-directed mutants of conserved acidic residues in the PSST subunit provides additional strong  evidence that this subunit plays a critical role in the catalytic mechanism of complex I. However, there is still no definite answer whether it contains iron-sulfur center N-2. Because we could exclude Glu 89 as the forth ligand for center N-2, models proposing that N-2 resides in PSST now have to solve the problem of the missing fourth ligand.