NADH as Donor

THORSTEN FRIEDRICH* AND THOMAS POHL

[SECTION EDITOR: VALLEY STEWART]

Posted August 13, 2007

Institute of Organic Chemistry and Biochemistry, Albert-Ludwigs-University, Albertstrasse 21, 79104 Freiburg, Germany

*Corresponding author. Mailing address: Institute of Organic Chemistry and Biochemistry, Albert-Ludwigs-University, Albertstrasse 21, 79104 Freiburg, Germany. Phone: (49) 761 203 6060, Fax: (49) 761 203 6096, E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it

The aerobic respiratory chains ofEscherichia coli andSalmonella contain several primary dehydrogenases, including two NADH dehydrogenases, succinate dehydrogenase, and D-lactate dehydrogenase, which oxidize organic substances and deliver the electrons to membrane-bound ubiquinone-8. The ubiquinol is reoxidized by two oxidases reducing oxygen to water. Under oxic growth conditions mainly the quinol oxidase bo₃ is present, but this complex is replaced by the quinol oxidase bd under microoxic growth conditions. The exergonic electron transfer is coupled with the endergonic translocation of protons across the cytoplasmic membrane. The resulting membrane potential is used to drive energy-consuming processes such as ATP synthesis, solute transport, and flagellar motion. With NADH as donor the number of protons translocated per electron varies between one and four depending on the enzymes involved in electron transfer. This chapter describes the properties of the two NADH dehydrogenases, focusing on the mechanism of the energy-converting NADH dehydrogenase as derived from the high-resolution structure of the soluble part of the enzyme. The regulation of the gene expression of the two NADH dehydrogenases as well as the biosynthesis of their cofactors will be covered by other contributions.

INTRODUCTION

The number of NADH dehydrogenases and their role in energy transduction in Escherchia coli have been under debate for a long time. Now it is evident that E. coli possesses two respiratory NADH dehydrogenases, or NADH:ubiquinone oxidoreductases, that have traditionally been called NDH-I and NDH-II (112, 175). NDH-I is made up of 13 or 14 different subunits, contains one flavin mononucleotide (FMN) and several iron-sulfur (Fe-S) clusters, and couples its redox reaction with the translocation of protons across the membrane (59, 62, 149). It is homologous to the mitochondrial complex I, and therefore nowadays most investigators call this enzyme complex I as well. NDH-II is a single-polypeptide enzyme with flavin adenine dinucleotide (FAD) as the only redox group, and its redox reaction is not coupled with proton translocation (89, 182). The function of NDH-II is most likely the maintenance of a balanced NADH/NAD⁺ ratio by oxidizing excess NADH.

Initial confusion concerning the E. coli NADH dehydrogenases arose from a screen for mutants unable to grow with mannitol as the sole carbon source but able to grow fermentatively on glucose (183). These mutants were incorrectly believed to contain single-locus mutations. One mutation mapped at 22 min was used to clone the ndh gene coding for NDH-II (182). NDH-II restored the KCN-sensitive NADH oxidation of an ndh mutant strain without contributing to a membrane potential. On the basis of these experiments, it was concluded that E. coli contains only one NADH dehydrogenase not involved in energy conservation (89, 181). Later, it was demonstrated that the mutant also carried a mutation at 52 min, corresponding to the nuo locus encoding complex I (32). In addition, the presence of a single non-energy-converting NADH dehydrogenase contradicted the finding that E. coli cells respiring on malate show a higher H⁺/O ratio than those respiring on succinate and that E. coli is capable of reducing NAD⁺ by a reverse electron flow from succinate or glycerol 1-phosphate in an uncoupler-sensitive manner (103, 111, 134). Two different types of NADH dehydrogenases were also detected with immunological methods (128). It was demonstrated that only one NADH dehydrogenase reacts with NADH and deamino-NADH linked with the generation of a membrane potential in the presence of KCN and ubiquinone-1 (112). Finally, electron paramagnetic resonance (EPR) spectroscopic characterization of aerobically grown E. coli membranes showed the presence of several Fe-S clusters associated with an NADH dehydrogenase (114).

Containing both complex I and NDH-II to optimize growth under various conditions seems to be a common feature of bacteria (33, 120). This topic is covered in greater detail in the chapter dealing with the modularity of respiratory chains Chapter (The Aerobic and Anaerobic Respiratory Chain of Escherichia coli and Salmonella enterica: Enzymes and Energetics). Respiration not only generates a proton motive force but also removes excess reducing equivalents to level out the NADH/NAD⁺ ratio. Under certain growth conditions, the recycling of NAD⁺ is more important than a maximum energy yield (33, 120). In E. coli, complex I operates in aerobic and anaerobic respiration (33), while NDH-II is repressed under anaerobic growth conditions (74), resulting in high metabolic fluxes and growth rates under aerobic conditions (158, 159). It was shown that NDH-II is more important in aerobic respiration than complex I, while the latter is essential for respiration on fumarate and dimethyl sulfoxide (DMSO) (158, 159, 165). E. coli mutants lacking complex I are at a competitive disadvantage during stationary growth phase (184). This could be explained by the lower amount of energy that can be obtained from any carbon source given the poor availability of nutrients in the stationary phase. Furthermore, complex I mutants grew poorly on acetate and on some amino acids as the sole carbon source and failed to produce the L-aspartate chemotactic band on tryptone swarm plates (135). The insufficient recycling of NADH most likely resulted in excess NADH inhibiting tricarboxylic acid cycle enzymes and the glyoxylate shunt (135). Salmonella enterica serovar Typhimurium complex I mutants are unable to activate ATP-dependent proteolysis under starvation conditions (6).

NADH DEHYDROGENASE I (COMPLEX I; NDH-I; NuoA-N COMPLEX)

General Remarks

Like mitochondrial complex I, its bacterial counterpart is a multisubunit enzyme and couples the electron transfer from NADH to ubiquinone with a proton translocation according to the following overall equation:

NADH + Q + 5 H⁺_n → NAD⁺ + QH₂+ 4 H⁺_p

where Q refers to ubiquinone and H⁺_nand H⁺_p are the protons taken up from the negative inner side and delivered to the positive outer side of the membrane. The H⁺/e⁻ stoichiometry of the enzyme in the cytoplasmic membrane was determined at a minimum of 1.5 (22). In analogy to mitochondrial complex I, the number should equal 2 (69, 169). It was also proposed that the redox reaction of complex I from E. coli and Klebsiella pneumoniae is coupled with the translocation of 1 Na⁺ per NADH oxidized (71). This will be discussed at the end of the paragraph. Complex I is characterized by its multiple prosthetic groups, namely one FMN and nine Fe-S clusters; by its sensitivity to several naturally occurring compounds, e.g., the antibiotic piericidin A or the acarizide annonin (44, 116); and by its numerous subunits. Bacterial complex I is made up of 14 different subunits (62, 63). Two of these are fused in E. coli and serovar Typhimurium, resulting in a complex of 13 subunits. The mitochondrial complex from bovine heart mitochondria contains 31 additional polypeptides (35; mitochondrial complex I is reviewed in references 26, 63, 123, 166, 168, and 173). Therefore, bacterial complex I is considered to be a structural minimal form of a proton-pumping NADH:ubiquinone oxidoreductase (105, 167).

E. colicontains three different quinones, namely ubiquinone (E_m,7 = +110 mV), demethylmenaquinone (E_m,7 = +40 mV), and menaquinone (E_m,7 = −80 mV) (170). For thermodynamic reasons it was expected that only ubiquinone and possibly demethylmenaquinone were used by complex I. However, complex I is required for anaerobic respiration of NADH to fumarate, implying that it is able to reduce menaquinone (158).

Genes

The nuo operon (from NADH:ubiquinone oxidoreductase) containing the genes of E. coli complex I was cloned with DNA probes designed by the protein sequence of the FMN- and NADH-binding sites of mitochondrial complex I (167). The locus forming the nuo operon is made up of 15,020 bp at 52 min and encodes 13 subunits named NuoA to NuoN (originally Nuo1 to Nuo14), which all have homologues in mitochondrial complex I (167). Due to a sequencing error, a TGA stop-codon in the 3' region of nuoC was in-frame with the nuo sequence and it was initially believed that nuoC and nuoD constituted separate genes (167). This error was recognized by comparison with the E. coli genome sequence (21, 29). The genes nuoC and nuoD are also fused in other bacteria such as serovar Typhimurium (113), Buchnera aphidicola (37), Aquifex aeolicus (43), and Yersinia pestis (45). In E. coli, the nuo genes code for six hydrophilic subunits (NuoB, CD, E, F, G, and I) and seven hydrophobic (NuoA, H, J, K, L, M, and N) subunits. Sequence comparisons with the homologues in Paracoccus denitrificans (named Nqo1 to Nqo14 [171, 174]) and of the mitochondrial complex from bovine heart gave first clues to the function of individual subunits (Table 1).

Table 1Nomenclature, properties, and proposed function of the subunits of the E. coli complex I and nomenclature of the T. thermophilus subunits

In bacteria, the nuo genes are organized in one large operon or gene cluster. The order of the 14 nuo genes is conserved in most species. However, several clusters of nuo genes exist in Rickettsia prowazekii, which is closely related to mitochondria (5), and Sinorhizobium meliloti contains a second set of nuo genes made up of three gene clusters (62). The nuo gene clusters of Helicobacter pylori and Campylobacter jejuni belonging to the ε-proteobacteria are missing the homologues of nuoE and nuoF but include a modified nuoG (52, 151). Remarkably, these three genes code for subunits constituting the NADH dehydrogenase part, a conserved module for NADH oxidation in complex I (see below). The genes nuoE and nuoF and the gene nuoG are located at separate loci in the hyperthermophilic eubacterium A. aeolicus (43). The complex I gene cluster of P. denitrificans and Rhodobacter capsulatus contains several unassigned reading frames (URFs) in addition to the nuo genes (46, 175). Deletion of these URFs in R. capsulatus had no significant effect on the assembly and activity of complex I, indicating that these URFs are not essential for the complex (47).

Replacement of any of the E. coli nuo genes by a kanamycin resistance cartridge leads to a complete loss of complex I, indicating that the bacterial complex subunits indeed constitute the structural minimal form of an energy-converting NADH:ubiquinone oxidoreductase and that all subunits are needed for a proper assembly of the complex (T. Pohl, J. Walter, and T. Friedrich; submitted for publication). Nonpolar Δ(nuoF-L) and ΔnuoG deletions led to the same phenotype (50). Chromosomal nuo mutants have been generated by disrupting individual genes by in-frame deletion, by insertion of a resistance cartridge, or by replacement of the gene with a resistance cartridge (4, 13, 54, 55, 56, 95, 96, 119, 163). In all these cases, a functional complex I was obtained after complementing the mutant with the wild-type nuo gene in-trans although sometimes in reduced amounts. Recently, the construction of an expression vector comprising all nuo genes and the upstream control region was reported (3). The disadvantage of this expression system is the low protein production. This problem was solved by the development of phage Lambda-mediated recombinogenic engineering of chromosomal and episomal DNA, also known as λ-Red-mediated recombineering ( 39, 41, 186). With this technique, the introduction of any desired mutation in an extrachromosomal complex I overexpression system was realized, without the above-mentioned limitations associated with in-trans complementation or the need for the time-consuming construction of knockout vectors for chromosomal DNA mutagenesis (133). The versatility of this approach is exemplified by the insertion of a hexahistidine-coding sequence in the nuo genes, permitting the efficient isolation of E. coli complex I by immobilized metal affinity chromatography (T. Pohl and T. Friedrich, submitted for publication).

The regulatory region of the nuo operon contains two promoters located about 290 and 120 bp upstream of the coding sequence (24). Expression of the operon corresponds to oxygen and nitrate availability, mediated by the two-component sensor-regulators ArcBA and NarXL, and to the presence of C₄-dicarboxylates (24). In addition, the expression is stimulated in the early exponential growth phase by the growth-phase-responsive regulator Fis and to a much lesser extent by several factors in the late-exponential and stationary-growth phases (165, 185). The Fis-binding sites are located 237 bp, 197 bp, and 139 bp upstream of the start of the nuoA-N transcript (165). The integration host factor (IHF) protein was shown to repress the transcription of the nuo genes (24, 161). Due to its regulation, complex I is an essential enzyme in anaerobic respiration of fumarate and DMSO (158). The expression is not stimulated by growth on substrates supplying large amounts of NADH, which might be because maximal energy conversion takes place only under conditions of NADH limitation. The control of enzyme synthesis is covered in detail in Global Control of Respiratory Enzyme Synthesis.

Biochemistry

Purification.

The first preparation of the E. coli complex I was obtained from a wild-type B strain in the presence of an alkyl polyglucoside detergent (104). The preparation was made up of 13 subunits plus a proteolytically cleaved fragment of NuoCD. The peripheral subunits NuoB, E, F, G, and I and the proteolytic fragment of NuoCD were assigned to the corresponding nuo genes by N-terminal sequencing. The remaining seven subunits are hydrophobic proteins predicted to fold into 61 α-helices across the membrane (Table 1). Their homologues in eucaryotes are mitochondrially encoded. They have been assigned to the nuo genes by their apparent molecular mass in sodium dodecyl sulfate (SDS) (104). The preparation was split into a soluble NADH dehydrogenase fragment made up of NuoE, F, and G, an amphipathic connecting fragment containing NuoB, CD, and I, and a hydrophobic membrane fragment consisting of NuoA, H, J, K, L, M, and N (104). The expression of the nuo genes was enhanced approximately fourfold by replacing the 5'-promoter region of the nuo operon with the phage T7 RNA polymerase promoter and by expressing the genes with the T7 RNA polymerase coded on an inducible plasmid in a derivative of strain AN387 (152). Complex I was prepared from this strain in the presence of dodecyl maltoside, and subunit NuoCD was assigned to a 65-kDa polypeptide in the preparation (152). Other protocols for the isolation of complex I from strain MWC215, lacking NDH-II (42, 148), and from strain BL21 grown under limited oxygen availability (84, 142) were described. Most subunits in these preparations have been identified by means of mass spectrometry (42, 142).

Cofactors.

The hydrophilic subunits contain all known cofactors of the complex (Table 1). One noncovalently bound FMN, two binuclear Fe-S clusters (N1a and N1b), and seven tetranuclear Fe-S clusters (N2, N3, N4, N5, N6a, N6b, and N7) are present in E. coli complex I (Table 2). With the exception of N5 all cofactors have been detected in preparations of E. coli complex I, of defined fragments of the complex, and of overproduced single subunits and attributed to the corresponding binding sites by site-directed mutagenesis (13, 29, 56, 104, 119, 133, 137, 148, 152, 160, 163). N5 has not yet been detected in E. coli complex I, most likely due to a spin ground state different from S = 1/2 (180). However, the sequence motif coordinating the cluster is conserved. The proposed localization of the cofactors was proven to be true by the recently published structure of the soluble part of the Thermus thermophilus complex (143). The FMN is the primary electron acceptor of NADH. The electrons are transferred to the acceptor quinone by a chain made up of the Fe-S clusters N3, N1b, N4, N5, N6a, N6b, and N2 (143). N1a is not part of this chain and most likely is used to reduce the risk of radical formation at the flavin (68, 102, 141, 143). The distance from N7 to the electron transport chain is too long for biological electron transfer in a reasonable timescale (117, 143). It was shown that N7 is not reduced by NADH and therefore not involved in the electron transfer but that it is essential for the stability of the complex (133). The midpoint potential of N1a and N2 is pH dependent, but its significance for the mechanism of complex I is under debate (86, 189, 190). After reconstitution in artificial phospholipid membranes, all preparations show inhibitor-sensitive NADH:short-chain-ubiquinone oxidoreductase activity. However, the turnover of the various preparations reported in the literature differs for unknown reasons from 1 to 22 μmol of NADH/(min · mg)⁻¹ (42, 142, 148, 152, 155).

Table 2Localization and midpoint potential of the cofactors of the E. coli complex I

Electron Microscopy

Electron microscopy was mainly used to investigate the gross structure of the complex. It turned out that E. coli complex I has the same L-shaped conformation as its mitochondrial homologue (77, 142; for reviews, see references 61 and 76). It consists of a so-called peripheral arm made up of the hydrophilic subunits and a so-called membrane arm comprising the hydrophobic subunits. The peripheral arm is located in the aqueous phase, while the membrane arm is embedded in the lipid bilayer. Both arms are arranged perpendicular to each other (77, 142). An alternative conformation with both arms arranged side by side was also reported (25). This so-called horseshoe-shaped conformation was reversibly transformed into the L-shape by raising the ionic strength of the buffer. Both conformations were distinguished from each other by precipitation with poly(ethylene glycol) and analytical ultracentrifugation. The complex in the horseshoe-shaped conformation showed a high activity, which was prohibited by an addition of salt (25). However, these results were questioned in a subsequent study showing only a mild effect of salt on the enzyme activity (142). Single-particle analysis and two-dimensional crystallization indicated that the complex is L-shaped in detergent and in lipids (142). However, evidence for the horseshoe-shaped conformation was also found in complex I proteoliposomes (61). Parts of the complex were attributed to distinct domains by splitting the preparations and subsequent electron microscopy of the fragments obtained. Incubation of the preparation at pH 9.0 led to the dissociation of complex I into the NADH dehydrogenase fragment and the so-called quinone reductase fragment (25). Electron microscopy of complex I and the quinone reductase fragment revealed the position of the electron input part at the top of the peripheral arm (25). Treatment of the preparation with other detergents caused a loss of the membranous subunits NuoL and NuoM as well as the dissociation of hydrophilic subunits such as NuoCD and NuoB (84). From these experiments a model for the arrangement of the subunits was derived with the subunits NuoL and NuoM located at the most distal end of the membrane arm (12, 84; Fig. 1). The membrane arm prepared by incubating the complex at pH 3.5 turned out to be stable and allowed its two-dimensional crystallization, resulting in a projection map at 8 Å resolution (11). In this work about 60 α-helices, some perpendicular to the membrane and others tilted, were detected (11).

Fig. 1Putative arrangement of the subunits of E. coli complex I. The arrangement of the peripheral subunits is derived from the crystal structure of the peripheral arm of the T. thermophilus complex I (143). The arrangement of the hydrophobic subunits is derived from electron microscopic data obtained with E. coli complex I (11). The membrane runs horizontally.

Structure of the Peripheral Arm

The peripheral arm of the T. thermophilus complex was crystallized and the structure was determined at 3.3 Å resolution in a pioneering study (141, 143). The T. thermophilus peripheral arm is made up of eight different subunits, including the so far unrecognized subunit Nqo15, for which homologues have been identified only in organisms closely related to T. thermophilus (82, 141). In agreement with previous electron microscopic studies, the peripheral arm has an overall "Y" shape. The rough arrangement of the subunits is shown in Fig. 1.

NuoF contains the FMN and the tetranuclear Fe-S cluster N3, coordinated within a four-helix bundle. One Rossman fold comprises the binding sites for NADH and FMN in NuoF. The arrangement of the elements of the Rossman fold is unique compared with other NAD-dependent dehydrogenases, which might be of functional importance.

The C-terminal domain of NuoE is homologous to thioredoxins and contains cluster N1a. The hydrophobic environment of N1a might cause its low midpoint potential (Table 2). The N-terminal part of NuoE builds a four-helix bundle. NuoG shows homologies to [FeFe] hydrogenases and a superfamily containing molybdo-bis(molybdopterin-guanine-dinucleotide) (Mo-bisMGD) enzymes as predicted from sequence comparisons.

The N-terminal part of NuoG carries the binuclear Fe-S cluster N1b and the tetranuclear Fe-S clusters N4 and N5 in a fold also found in the [FeFe] hydrogenases (131). These hydrogenases usually contain a fourth, tetranuclear Fe-S cluster called FS4A, which is, however, not present in the structure of the T. thermophilus complex I due to the absence of the corresponding binding motif. The C-terminal part of NuoG is homologous to the superfamily containing Mo-bisMGD enzymes (16), but the structure clearly shows that this cofactor is not present in complex I. The tetranuclear Fe-S cluster N7, which is missing in complex I from most species, is present in this domain at a position equivalent to that of the homologous cluster in the Mo-bisMGD enzymes (17, 99).

NuoI is homologous to 2x[4Fe-4S] ferredoxins and bears the clusters N6a and N6b, which were detected by UV/visible light spectroscopy (137).

NuoB and NuoD are homologous to the small and large subunits of [NiFe] hydrogenases, respectively (1, 66, 140). The N-terminal part of NuoB contains a flavodoxin-like fold, but a second flavin is not present in the fold as was proposed earlier (2). NuoB harbors the tetranuclear Fe-S cluster N2 as predicted from site-directed mutagenesis (56). N2 is coordinated by four cysteines, with two of them being vicinal in location. Molecular dynamics simulation showed that such a unique coordination is principally possible (78). This coordination leads to an unusual geometry and a higher flexibility around the cluster (141). The highly polar environment of N2 may be responsible for its high midpoint potential (Table 2). Although N2 is entirely bound to amino acids of NuoB, it is located at the interface to subunit NuoD. The Ni-binding site of the large subunits of [NiFe] hydrogenases is conserved in NuoD (143), but the [NiFe] active site is missing, as proposed by Fourier transform infrared (FTIR) spectroscopy (162). This site is located close to a cavity extending to the suggested location of the membrane arm. It was proposed that this cavity builds the quinone-binding domain (141, 143) as indicated by site-directed mutagenesis in E. coli (13) and Yarrowia lipolytica (188). However, this proposal has to be confirmed by the structure of the entire complex I with bound quinone.

NuoC does not contain a cofactor, and its function seems mainly to be stabilization of the complex (75). NuoC is fused to NuoD in complex I from some organisms. The homologues in the E. coli formate:hydrogen lyase are also fused (9).

The presence of Nqo15 as part of complex I has been demonstrated so far only in T. thermophilus (82, 141). Nqo15 exhibits structural similarity to members of the frataxin family, and it is most likely involved in iron storage or iron release but has no direct function in electron transfer or ion translocation. It was shown that the preparation of E. coli complex I is not associated with CyaY, the E. coli frataxin homologue (133a).

Structure of the Membrane Arm

The hypothetical arrangement of the membranous subunits was derived mainly by electron microscopy (11, 12, 84) in combination with indications from phylogenetic analysis (62; see below). A model showing the putative arrangement of the hydrophobic subunits is shown in Fig. 1.

According to this model, NuoJ and H are located at the proximal end of the membrane arm. NuoJ is predicted to fold in five transmembranous (TM) helices. In general, point mutations did not affect the assembly of the complex and had a mild effect on electron transfer activity (90). However, mutation of conserved Val65 (numbering according to E. coli complex I), most likely located in TM helix 3, led to strongly reduced electron transfer and proton translocation activity (90). NuoH is predicted to contain eight TM helices, which was proven experimentally by using LacZ and PhoA fusions (139). It was proposed that the subunit is involved in the binding of N,N'-dicyclohexylcarbodiimide (DCCD), a strong inhibitor of complex I (155, 164, 172). Mutations of the conserved acidic amino acids in the P. denitrificans homologue, however, did not alter the sensitivity of the complex to DCCD (101). Mutations causing mitochondrial diseases in humans disturbed the assembly and reduced the activity when introduced in E. coli NuoH (95).

NuoA, K, and N are expected to be located at the same end of the membrane arm, probably in contact with the peripheral arm (11). NuoA is predicted to contain three TM helices and has been shown to interact with subunits of the peripheral arm in P. denitrificans (92). Mutations of conserved acidic amino acids in the E. coli NuoA led to slightly reduced electron transfer activity, while double mutants lost their activity nearly completely (91). NuoK is the smallest subunit of complex I, and three TM helices are predicted in this subunit. Site-directed mutagenesis studies revealed that the conserved E36 and E72 located in TM helix 2 and 3, respectively, are important for electron transfer activity and possibly proton translocation (93, 96). The enzymatic activity of E36/E72 double mutants was restored when two acidic amino acids were placed in either helix 2 or helix 3 at an interval of three amino acids (96). In addition, it was shown that simultaneous mutation of arginines 85 and 87 located on a short cytoplasmic loop strongly reduced electron transfer and proton translocation activities of complex I (93).

The residual hydrophobic subunits NuoL, M, and N are the largest hydrophobic subunits of complex I and constitute approximately two-thirds of the mass of the membrane arm. They are related to each other (51, 97) and to subunits of the multisubunit monovalent cation/proton antiporters (67, 110, 156). From electron microscopy the numbers of TM helices in the E. coli complex I were determined to be 16, 14, and 14 for NuoL, M, and N, respectively (11). By using the "majority vote" principle for the results of six different secondary structure prediction programs the numbers were predicted at 16, 14, and 12, respectively (T. Pohl and T. Friedrich, unpublished results) (Table 1). The numbers have been determined experimentally with LacZ and PhoA fusions in R. capsulatus (109). In this work, two helices of subunits NuoL, M, and N are discussed as surface helices (109). Site-directed mutagenesis of conserved amino acids of NuoN led to a decreased electron transfer and proton translocation activity (4). The affinity to the substrate quinone was altered in some of the mutants, indicating a possible interaction of NuoN with ubiquinone. Because of their homology to antiporter subunits (see below), it is believed that NuoL and M are involved in proton translocation (11, 58, 62, 67, 107, 142).

Relationship with Other Proteins and Modular Evolution of the Complex

Sequence analysis demonstrated the homology of other proteins to parts of complex I, and from this, the modular structure of complex I was recognized (63, 67). The proposed homology to enzymes of known structures was confirmed by the structure of the peripheral arm (141, 143). NuoB and NuoD are homologous to the small and the large subunit of [NiFe] hydrogenases. This homology was first discovered by Böck and coworkers (23, 140) and confirmed by others (1, 66). Later, a family of membrane-bound multisubunit [NiFe] hydrogenases was identified consisting of homologues of NuoB, CD, I, H, and one of the antiporter-like subunits, most likely NuoN (12, 57, 62, 67, 79, 80, 140, 157). Some members of the hydrogenase family contain additional subunits. From several experiments it was concluded that membrane-bound multisubunit [NiFe] hydrogenases are most likely involved in energy conversion (79, 80). The small and large subunits of these membrane-bound [NiFe] hydrogenases show a higher degree of sequence identity to the homologous complex I subunits than to those of the soluble [NiFe] hydrogenases (67). From these data it was concluded that complex I and the membrane-bound multisubunit [NiFe] hydrogenases possess a common ancestor (67). NuoE, F, and G share homology with the diaphorase part of several NAD-dependent hydrogenases and formate dehydrogenases (31, 67, 108, 122, 132). This part is called NADH dehydrogenase module, and from protein and cofactor composition, the module corresponds to the NADH dehydrogenase fragment obtained by splitting the E. coli complex I mentioned above. This module is present in soluble enzymes, and the free energy available from the redox reactions involving the cofactors of the module is not sufficient to drive proton translocation against a gradient (Table 2). Therefore, the redox reaction of the NADH dehydrogenase module is most likely not coupled with proton translocation, but this module may serve as a reversible switch to convert one two-electron transfer reaction in two one-electron transfer reactions (58, 67). NuoK, L, M, and N are homologous to subunits of multisubunit monovalent cation/proton antiporters (67, 83, 110, 136, 156). The antiporters consist of seven membranous subunits, three of them showing a striking homology to NuoK, L, and M. They translocate substrates across the membrane by means of conformational changes.

From these data two hypothetical schemes for the evolution of respiratory complex I were depicted (62, 67, 110). According to one scenario, an ancestral [NiFe] hydrogenase consisting of the progenitors of NuoB and NuoD evolved by addition of NuoC, H, and I and the progenitor of NuoL, M, and N, leading to the common progenitor of the complex I family and the family of multisubunit membrane-bound [NiFe] hydrogenases (62). The ancestor of the complex I family lost the [NiFe] active site and was equipped with a quinone-binding site instead. The membrane part was enlarged with further proteins by triplication of the progenitor of NuoL, M, and N and by addition of NuoA, J, and K. This complex is meant to be the last common ancestor for the complex I family with members present in today’s bacteria, archaea, and eucaryotes. E. coli complex I evolved by addition of NuoG and later NuoE and NuoF as an electron input module (62). In the other scenario, the ancestral [NiFe] hydrogenase joins an ancestral membrane-bound antiporter consisting of the progenitors of NuoK, L, and M (110). NuoH and NuoI were added either to one of these complexes or to the combination of both. Gene duplication of NuoM resulted in NuoN. As in the other scheme, the electron acceptor part was added to the progenitor of complex I in all three kingdoms of life (110). The membrane-bound hydrogenases would be a "degenerated" form of the last common complex I ancestor by having lost two of the antiporter-like subunits. Because of a possible convergent evolution of the homologues in complex I and the multisubunit monovalent cation/proton antiporter, it is not possible to rule out one scenario, although the genes of the antiporter subunits are more closely related to the homologous complex I subunits than to each other (109).

Mechanism

Electron transfer.

The most plausible route of electrons in complex I became evident mainly from the crystal structure of the peripheral arm of the T. thermophilus complex as well as from spectroscopic data and from site-directed mutagenesis (Fig. 2) (117, 141, 143). The electrons from NADH are transferred to the FMN most likely via hydride transfer. The reduced FMNH₂ is forced to deliver two electrons to one-electron-accepting Fe-S clusters. This is most likely done by an "electron disproportion." According to this mechanism, the first electron is given to N3 in a thermodynamically favorable way. The midpoint potential of the resulting FMN/flavosemiquinone couple is about −380 mV (102, 150), allowing the reduction of cluster N1a with a midpoint potential of −330 mV. It was proposed that this mechanism of flavin reoxidation reduces the risk of producing reactive oxygen species that might evolve at the flavosemiquinone site (102, 141, 143), because the FMN is exposed to the solvent while N1a is shielded from it. It is not known whether and how the electron from N1a is transferred to the quinone reduction site. From N3 the electron moves down a chain of Fe-S clusters made up of N1b, N4, N5, N6a, and N6b to N2, the proposed electron donor for the substrate quinone. The presence of N1b seems to be superfluous because the distance between N3 and N4 is less than 14 Å, implying a direct electron transfer. However, N3 and N4 have a spatial arrangement leading to an edge-to-edge distance close to 16 Å, when the mere distance between the Fe atoms is measured (141). Either the S atoms of the Fe-S clusters are involved in the electron transfer, which would lead to a possible branch in the electron pathway so that the electrons from N3 proceed via N1b or directly to N4, or the S atoms are not involved in the electron transfer and the electrons necessarily have to take the route via N1b (Fig. 2). From the structure and from functional studies it is obvious that N7 is not involved in the electron transfer, at least not in T. thermophilus and E. coli (133, 143). N7 is also present in the superfamily containing Mo-bisMGD enzymes and might be an evolutionary relic of the electron transfer route of these enzymes in complex I. It is possible that the last common ancestor of the complex I family (see above) was equipped with NuoG , allowing the reaction with another electron donor. The "old" substrate might be dehydrogenated by a molybdenum cofactor, and the electrons might be transferred to the quinone via Fe-S clusters N7, FS4A, N4, N5, N6a, N6b, and N2. With the adaptation of NuoE and NuoF the putative substrate binding site on NuoG and its connection to the electron transfer route became dispensable. However, in a few organisms like E. coli the presence of N7 is obligatory for structural reasons.

Fig. 2Edge-to-edge distances of the cofactors of the peripheral arm of complex I as deduced from the crystal structure of T. thermophilus complex I. The substrate quinone is abbreviated with Q. Electron transfer steps are indicated by solid arrows. Hypothetical electron transfer steps are shown as dotted lines.

Quinone reduction.

The mechanism of quinone reduction by complex I is still under debate because (i) the position of the quinone-reduction site is not yet known, (ii) the number of quinones bound to the complex is under discussion, (iii) the energetics with the various naturally occurring quinones exhibiting different midpoint potentials is not understood, and (iv) the electron donor to the substrate quinone, which is freely exchangeable with the quinone pool, is not known.

It has been proposed from labeling the complex with photoreactive quinone-site inhibitors and photoreactive quinones as well as by site-directed mutagenesis studies that subunits NuoB, D, and H are involved in quinone binding, under the assumption that the quinone-site inhibitors act at the quinone-binding site (40, 49, 94, 145). However, the possibility that the quinone-site inhibitors operate at a site distant from the quinone reduction site by preventing access or by inducing conformational changes cannot be excluded (87). From the structure of the peripheral arm the quinone-binding site was proposed to correspond to a cavity extending from an interface between NuoB and NuoD to the plausible position of the membrane arm involving NuoA, K, and N (11, 141, 143). The proposed localization of the quinone-binding site between NuoB and NuoD fits with site-directed mutagenesis in E. coli (13) and Y. lipolytica (188). However, from sequence analysis it was also proposed that NuoM and NuoL contain motifs for the binding of quinones (53). In addition, NuoM was labeled by a photoreactive quinone derivative (72) and NuoL was labeled by an inhibitor of the quinone-binding site (118). Thus, it cannot be completely excluded that the distal part of the membrane arm is involved in electron transfer, although the electronic connection to the chain of the Fe-S clusters is not evident.

It has been proposed that a Q-cycle mechanism operates in complex I, in analogy to the bc₁ complex (27, 48), but this has been questioned by others (147). Two semiquinones named SQ _Nfand SQ_Ns were identified in the mitochondrial complex I (124, 178, 179). The fast-relaxing SQ_Nf is sensitive to uncouplers and located 12 Å from N2. The slow-relaxing SQ_Ns was determined to be at least 30 Å from N2. These semiquinone radicals have so far not been detected in a bacterial complex.

Furthermore, there is spectroscopic evidence for an additional cofactor of unknown chemical structure that was proposed to be involved in quinone reduction (60, 146). It was suggested that the cofactor possibly originated from the posttranslational modification of an amino acid side chain similar to those found in the so-called quino-proteins (146). From the mass spectrometric characterization of subunits of bovine complex I, it was concluded that, if such a cofactor existed, it would not be covalently linked to the complex, at least not in bovine heart mitochondria (36).

Inhibitors.

Complex I is inhibited by a large number of naturally occurring and synthetic compounds. Rhein and diphenyleneiodonium block NADH oxidation most likely by interacting with the flavin site (106). ADP-ribose, another soluble inhibitor, was shown to inactivate mitochondrial complex I (187) but has only a minor effect on the E. coli complex I. The hydrophobic inhibitors block the electron transfer between Fe-S cluster N2 and the substrate quinone (44, 116). Because of their competition with quinone, the quinone-site inhibitors were classified in two groups (65). This classification was extended to three different classes of inhibitors on the basis of steady-state inhibition studies (44). For most of these compounds such as rotenone, piericidin A, and the acetogenins, the most potent inhibitors of complex I, the essential structural elements affecting the inhibitory potency are not clear (116). Thus, together with the fact that bulky short-chain ubiquinones are fairly good electron acceptors of complex I, it was proposed that the quinone- and inhibitor-binding site in complex I may be built up by a large cavity (126). This was supported by equilibrium studies indicating that all three classes of complex I inhibitors share one large binding pocket (127). Amilorides, known inhibitors of Na⁺/H⁺ exchangers, were found to inhibit complex I from mitochondria and bacteria (118, 153, 155; see below).

Proton translocation.

The coupling of the electron transfer with the proton translocation is not understood. The number of coupling sites and the type of coupling (redox driven or conformationally driven) is still under debate. It is now widely accepted that the redox reaction of complex I is accompanied by conformational changes (14, 28, 58, 107), but the extent to which the conformational changes contribute to proton translocation is not clear. The structure of the peripheral arm suggests that proton translocation is connected to quinone reduction but not to NADH oxidation. It was proposed that the electron transfer from N2 to the quinone is coupled with a proton translocation based on FTIR spectroscopic data and site-directed mutagenesis with the E. coli complex I (54, 55, 57, 81). The structure revealed that N2 is part of a hydrogen-bond network that extends from the surface of the protein to the proposed quinone-binding site (141, 143). A proton might be taken up by N2 from the surface via tyrosine 139 in NuoB, which was identified to be protonated upon reduction of N2 (54). The proton might be transferred to the quinone by the conserved tyrosine 277 in NuoCD (94) or by other proton-accepting amino acids such as glutamic acid 67 in NuoB (55). Both amino acids are essential for enzyme activity, and FTIR spectroscopy revealed that glutamic acid 67 becomes protonated during the redox reaction of the enzyme (55). In addition, replacing the conserved aspartic acid 94 on NuoB with the corresponding amide led to a complete loss of activity (55). This residue is located close to the protein surface and hydrogen bonded to the backbone amide of helix H1 of NuoB. It was proposed that the amphipathic helix H1 is attached to the membrane arm of the complex and possibly connects the electron transfer in the peripheral arm with the proton translocation in the membrane arm (143). The function of helix H1 has to be discussed carefully as it is involved in crystal contacts. Other amino acids that possibly protonate the quinone are aspartic acid 329 and 592 as well as glutamic acid 351 on NuoCD. In contrast, it was recently reported that the conserved histidine 359 on NuoCD, which is hydrogen bonded to N2, is the redox-Bohr group of complex I and that the redox reaction of N2 is not coupled with proton translocation (190).

An elegant mechanism for the translocation of two protons via redox-driven conformational changes of the quinone-binding protein has been proposed (125). According to this model a semiquinone anion is reduced to the dihydroquinone by the uptake of an electron from the electron transfer chain described above and two protons from the inside. After a conformational change, the protons are released to the outside via proton well, which might be the transporter-like subunit NuoN of the hydrogenase module. The electron is transferred to a quinone of the quinone pool, and the system is ready for the next turnover.

Because of their homology to subunits of the multisubunit monovalent cation/proton antiporters it was proposed that NuoL, M, and N are involved in proton translocation (58, 67). Electron microscopy revealed the localization of NuoL and M at the distal part of the membrane arm (11, 12, 84, 107). Because of the distance of NuoL and NuoM of more than 100 and 60 Å, respectively, from the suggested quinone-binding site it was proposed that the electron transfer is coupled with long-range conformational changes, leading to subsequent proton translocation (11, 84). The conformational changes that are induced by an addition of NADH are not yet understood (107). Electron microscopy of an NADH-treated sample revealed that both arms of the complex were widened, leading to a so-called open conformation. It was considered that the enzyme cycles between a "closed" and an "open" conformation during catalysis (107).

Sodium Translocation

It was reported that complex I from E. coli and K. pneumoniae couples the electron transfer reaction with the translocation of Na⁺ ions instead of protons (70, 71, 100, 154). With a partially purified enzyme from K. pneumoniae, a Na⁺/e⁻ stoichiometry of 1 was determined (71). An E. coli strain lacking the Na⁺/H⁺ antiporter genes nhaA and nhaB was unable to grow at higher Na⁺ concentrations except when cultivated under conditions with an elevated expression of the nuo genes, and it was shown that inverted E. coli membrane vesicles catalyzed NADH-driven Na⁺ uptake (154). However, the evidence for sodium pumping was not convincing due to the low catalytic rates reported (100, 154) and the incomplete inhibition of the reaction by complex I inhibitors (70). Studies with K. pneumoniae mutants individually lacking complex I, NDH-II, and the Na⁺-translocating NADH:quinone reductase (NQR) indicated that the activity of complex I is associated with proton translocation, while the activity of the NQR is coupled with Na⁺ translocation (18). It was demonstrated that the enzymatic activity of a highly purified preparation of the E. coli complex I was independent of the Na⁺ concentration at pH 6.0 and pH 7.8. The preparation of the E. coli complex I catalyzed the generation of a proton gradient across the membrane after reconstitution in proteoliposomes (155). This gradient was driven by the NADH:decyl-ubiquinone oxidoreductase activity of complex I and sensitive to complex I inhibitors and protonophores. A contribution of a Na⁺ gradient to the membrane potential could not be measured (155). This was supported by FTIR spectroscopy (64). The electron transfer reaction of complex I was inhibited by DCCD and different types of amilorides, which are discussed as blocking the ion translocation, independent of the Na⁺ concentration of the assay. However, the more basic the buffer the larger the 50% inhibitory concentration (IC₅₀) values, indicating inhibition of proton translocation (155). The magnitude of the pH gradient generated by complex I varied depending on the difference between the Na⁺ concentration inside and outside the proteoliposomes, showing that E. coli complex I may be capable of secondary Na⁺/H⁺ antiport (155).

NADH DEHYDROGENASE II (NDH-II; Ndh ENZYME; ALTERNATIVE NADH DEHYDROGENASE)

General Remarks

NDH-II is a single-subunit enzyme with a molecular mass of 47 kDa facing the cytosol (20, 89, 129, 182). Despite the absence of any predicted transmembrane segment (182) it has to be purified in the presence of detergents, and the activity of the preparation is stimulated by an addition of lipids (20). Many amphipathic helices have been detected in the sequence of the E. coli NDH-II, possibly explaining its membrane attachment (115). It contains FAD as the sole cofactor and catalyzes the oxidation of NAD(P)H coupled to the reduction of quinones (34). High V_max and low K_m values were reported for hydrophobic quinones, supporting its membrane-bound nature (20). The quinone-binding mechanism is not known, but it is reasonable to assume that the polar head of the quinone has to enter some kind of cavity inside the enzyme as it has been reported for other monotopic membrane proteins. From work with the enzyme of Saccharomyces cerevisiae it was proposed that NDH-II contains a binding site for tightly bound quinone and a second site for catalytic quinone (177). The absence of any metal prosthetic groups indicates that the electrons from the reduced FAD are directly transferred to the quinone. The electron transfer reaction of NDH-II is not inhibited by the quinone-site inhibitors of complex I, such as rotenone, piericidin A, capsaicin, and DCCD (176).

The structure of NDH-II is not yet known, but from comparative three-dimensional modeling a theoretical model of the enzyme has been derived (144). Recently, the crystallization and preliminary structure determination of the enzyme from the extremophile Acidianus ambivalens was reported (30), so that a refined model of the enzyme is anticipated.

The level of NDH-II in E. coli is regulated by Fnr (73, 74) repressing the ndh transcription under anaerobic or microaerobic growth. In this case E. coli relies on the activity of complex I for NADH oxidation, which is consistent with the phenotype of nuo strains showing defects in the late-exponential or stationary growth phase (6, 135, 184). The expression of ndh is also regulated by the growth phase-responsive transcription factor Fis, acting at three distinct sites of the promoter (88). Binding to Fis I and Fis II sites activates ndh transcription, while interaction with Fis III leads to a repression.

Classification of NDH-II

NDH-II has been found in plant and fungal mitochondria but not in animal mitochondria. It is widespread in bacteria, most likely to provide an uncoupled pathway for the oxidation of NADH (115). From sequence comparisons NDH-II enzymes were classified into three groups coined A, B, and C (115, 130). Group A includes the enzymes with two adenine dinucleotide-binding sites involved in the binding of NAD(P)H and the flavin. E. coli NDH-II belongs to this group. Members of group B contain, in addition to these binding sites, one EF-hand motif involved in calcium binding. This group is made up entirely of sequences from plants, fungi, and yeast (115). Group C NDH-II enzymes are marked by a single Rossman fold and the presence of a conserved histidine, most likely involved in covalent binding of the flavin. Members of the group C may contain FMN instead of FAD (10). NDH-II was also classified as a member of class IV of the superfamily of pyridine nucleotide disulfide reductases (7, 8).

Metabolic Functions

At the first glance NDH-II seems to be a redundant protein. However, it adopted many different and sometimes crucial functions in various organisms and under distinct growth conditions so that it significantly contributes to robustness of the species (8, 115). Several functions—from a simple replacement of NDH-I in Bacillus subtilis (15) and Zymomonas mobilis (98) to protecting nitrogenase in aerobic nitrogen fixation in Azotobacter vinelandii (19) to involvement in methanol metabolism (38) and sensing in cyanobacteria (85)—have been reported. Its main function in E. coliis most likely to ensure a well-balanced NAD⁺/NADH ratio under optimal energetic yields (33). This has also been shown in an E. coli K strain with a defective F₁-ATPase, which exhibits nearly fourfold enhanced NDH-II activity compared with the parental strain due, in part, to transcriptional upregulation when grown in a glucose-limited chemostat (121). Due to the increased amount of NDH-II, the mutant was able to recycle NADH at high rates. It has also been reported that NDH-II is involved in reducing cytosolic copper ions to prevent the formation of free radicals by Cu(II) via either flavin or the quinone (8, 138). Putative binding domains for Cu(I) and Cu(II) have been proposed in a subset of ndh sequences including that of E. coli by a bioinformatic approach (8).

ACKNOWLEDGMENTS

Work in our laboratory was supported by grants from the Deutsche Forschungsgemeinschaft and the Volkswagen Stiftung.

We thank Helga Lay and Linda Böhm for their help in preparing the manuscript.