To access the private area of this site, please log in.

Username: 

Password:  

 

Chapter 17

Respiration

ROBERT B. GENNIS and VALLEY STEWART

 

INTRODUCTION

Since the first edition of this book, there has been very substantial progress in our understanding of both the aerobic and anaerobic respiratory systems in Escherichia coli. These systems allow the cells to oxidize a wide variety of organic substrates (e.g., NADH, succinate, lactate), passing the electrons ultimately to any of a number of oxidants (e.g., oxygen, nitrate, dimethyl sulfoxide [DMSO]). Membrane-bound enyzmes are crucial for coupling the redox chemistry to the generation of a proton electrochemical potential gradient (i.e., the proton motive force) across the cytoplasmic membrane. The proton motive force, in turn, is utilized by other membrane proteins to do work, such as solute transport, flagellar rotation, or ATP synthesis. The bioenergetic principles are the same for both the aerobic and various anaerobic respiratory systems of E. coli and are no different than the principles considered in understanding mitochondrial function in eukaryotes. Indeed, one of the major advances of the last several years has been the demonstration that several of the aerobic respiratory components of E. coli are closely related to those in the mammalian mitochondrion. Mitochondrial NADH dehydrogenase (complex I), succinate dehydrogenase (complex II), and cytochrome c oxidase (complex IV) all have homologs in E. coli. This has increased the interest and motivation for a detailed examination of the E. coli respiratory system. By contrast, the enzymes involved in anaerobic respiration in E. coli have generally not been as thoroughly examined biochemically, although noteworthy exceptions include DMSO reductase and fumarate reductase.

In the past decade, the structural genes for virtually all of the aerobic and anaerobic respiratory enzymes have been identified, cloned, and sequenced, thereby providing invaluable structural information and material for biochemical studies. Sequence information also enhances studies of membrane protein topology and genetic regulatory elements. The coming years will see increased application of this information to studies of enzyme function for a broad variety of respiratory enzymes.

Unlike the mitochondrion, E. coli must adapt to a wide variety of growth conditions and environmental challenges in order to survive. The respiratory system has a modular design to facilitate the fine tuning of the system to meet the physiological needs of the organism. A complex regulatory network adjusts the composition and thus the properties of the respiratory system to allow the organism to compete effectively in the environment. This chapter reviews the characteristics of the respiratory components with emphasis on how the system works. This information provides the basis for understanding the physiology and genetic regulation.

 

MODULARITY OF RESPIRATORY COMPONENTS

As part of its ability to adapt to different growth conditions, E.coli alters the composition of its respiratory system. The design of the system itself is relatively simple and has a modular character. That is, different components can be substituted in the membrane in place of or in addition to other components as they are needed and can be fully functional parts of the entire system. Figure 1 illustrates the three types of respiratory components: (i)substrate-specific dehydrogenases, which carry out the oxidation of organic substrates and feed electrons into the mobile quinone pool; (ii) quinones, which deliver reducing equivalents to the terminal oxidoreductases; and (iii) terminal oxidoreductases, which reduce the terminal electron acceptors being utilized. The amount of each component is strictly regulated to optimize the respiratory chain according to the substrates present and the physiological needs of the cell. The genetic regulatory mechanisms are briefly described below (see Regulation of Enzyme Synthesis) and also elsewhere in this collection by Lynch and Lin (see chapter 95).

The modular construction of the respiratory system is an advantage in that it enables the cell to make minimal substitutions or alterations to adapt to different growth conditions. For example, in the presence of nitrate under anaerobic conditions, nitrate reductase is synthesized and is used as the terminal oxidoreductase. Nitrate reductase accepts electrons from the quinone pool just as do the terminal oxidases; it is the mobile quinone that facilitates the modular design of the respiratory system. All of the dehydrogenases are quinone reductases (ubiquinone and/or menaquinone), and all of the terminal oxidoreductases are quinol oxidases (ubiquinol and/or menaquinol; see Quinones, below).

One important aspect of the modularity of the respiratory system is its apparent redundancy. For example, E. coli can make two different NADH dehydrogenases and two different terminal oxidases. A closer look, however, reveals that these multiple components perform distinct tasks. Figure 2 shows the electron flux diagram with the two NADH dehydrogenases (NDH-I and NDH-II) and the two oxidases (cytochrome bd and cytochrome bo ₃). These enzymes are coupled to the generation of the proton motive force to different degrees (66, 297). NDH-I is a primary proton pump and results in translocating 2H⁺/e^–, whereas NDH-II is not coupled (H⁺/e^– = 0). The two oxidases are also different in their efficiency of proton translocation, with H⁺/e^– = 2 for cytochrome bo ₃ and H⁺/e^– = 1 for cytochrome bd. This means that the number of protons delivered to the periplasm can vary from one to four per electron for the aerobic oxidation of NADH, depending on the extent to which each of the four enzymes is used (Fig. 2). The electron flux through these enzymes will depend on the concentrations of the enzymes in the membrane, the NADH, quinone, and oxygen concentrations, and the steady-state characteristics of the enzymes, i.e., the V _max and K_m values for NADH, quinone, and oxygen. The different bioenergetic efficiencies of these different components have been experimentally demonstrated by comparing the growth yields in steady-state cultures of isogenic strains forced by genetic manipulation to utilize various combinations of these enzymes for NADH respiration (66).

Hence, E. coli can combine its respiratory components to vary the bioenergetic efficiency of NADH oxidation (measured by the H⁺/e^– ratio) over a large range. This suggests that the optimized respiratory chain is not always that with the highest value of H⁺/e^–. One important function of the respiratory chain independent of its bioenergetic efficiency is the maintenance of redox balance and the regeneration of NAD⁺ from NADH. Under growth conditions where fermentation provides an ample supply of energy, the net flux through the chain to regenerate NAD⁺ might be of overriding importance. The respiratory chain under semianaerobic conditions might be important to protect enzymes that are sensitive to dioxygen by reducing its steady-state intracellular concentration. The cytochrome bd ubiquinol oxidase, which is synthesized under low aeration conditions, has an extraordinarily high affinity for dioxygen(K_m = 0.02 _μM in whole cells) and appears to be well suited for functioning at low oxygen concentrations (170). The apparent trade-off, however, is that this enzyme has a lower bioenergetic efficiency (H⁺/e^– = 1) than does cytochrome bo ₃ (H⁺/e^– = 2) (297). The cells under these growth conditions appear to compensate for this by repressing the expression of the ndh gene which encodes the uncoupled NDH-II, so that a greater proportion of the electron flux would go through NDH-I, which is coupled (H⁺/e^– = 2) (66).

Thus, the modular design of the respiratory system allows the organism to readily optimize the system to changing growth conditions. The cells can adapt rapidly to the changing availability of substrates and electron accepting compounds and can also find the correct balance between the competing needs for bioenergetic efficiency, NAD⁺ regeneration and net electron flux, and oxygen scrubbing.

 

MECHANISMS OF PROTON TRANSLOCATION

One of the major functions of the respiratory chain is to generate a proton electrochemical gradient, usually referred to as the proton motive force, across the cytoplasmic membrane. The respiratory system is designed to oxidize a wide variety of substrates and to utilize any of several different terminal electron acceptors, as described in the preceding section. To utilize the free energy available from these reactions, the redox chemistry which is catalyzed by the enzymes must be coupled to the movement of charge and/or protons across the membrane (reviewed in reference 380). The enzymes in the E. coli respiratory system exhibit several mechanisms by which this is accomplished, and these are illustrated schematically in Fig. 3 for representative respiratory enzymes. Also shown in Fig. 3 are the number of protons per electron (H⁺/e^–) that appear in the periplasm for each of these enzymes. Each of these enzymes catalyzes two half-reactions, often at different sites in the protein. For example, NDH-II (Fig. 3C) catalyzes the two-electron oxidation of NADH and the two-electron reduction of ubiquinone. This enzyme has no transmembrane elements, and all of the chemistry occurs on the cytoplasmic side of the membrane. The net result, therefore, is that one proton (per NADH) is consumed from the cytoplasm, but there is no transmembrane voltage generated by this enzyme. The bioenergetic contribution of this enzyme to the proton motive force is therefore negligible. Succinate dehydrogenase (succinate:ubiquinone oxidoreductase, complex II [Fig. 3B]) does contain membrane-spanning subunits, but the protons used in the reduction of ubiquinone again come from the cytoplasm (167). The net result is no change in the cytoplasmic proton concentration and no generation of a proton motive force. The similar enzyme fumarate reductase is unlikely to contribute to the proton motive force, for the same reason.

 

Scalar Chemistry and Substrate Protons

The simplest way by which a respiratory enzyme can generate a proton motive force is to catalyze the two half-reactions (and access protons) on opposite sides of the membrane. Essentially, one half-reaction consumes protons from the cytoplasm, and the second delivers protons to the periplasm. This is sometimes referred to as a scalar mechanism, because the proton motive force results from the chemistry occurring at the two separated active sites and no proton-conducting channel is required.

The best-characterized respiratory enzyme which generates a proton motive force by utilizing substrate protons is the cytochrome bd ubiquinol oxidase (Fig. 3D). The proposed mechanism by which this enzyme results in a proton motive force is consistent with a large amount of experimental data (235, 257). The two-electron oxidation of ubiquinol results in the release of two protons to the periplasm. The two electrons are transferred to the second active site, where they are used to reduce oxygen to water, a reaction that consumes one proton from the cytoplasm for each electron. The net result is that substrate protons from ubiquinol appear in the periplasm and protons used to generate water are consumed from the cytoplasm, i.e., a net of one proton translocated across the membrane per electron. There is net charge movement across the membrane concomitant with this reaction, and the charge carriers are either the electrons, being transferred between the active sites, or the protons moving into or out of the active centers. The overall reaction yields a transmembrane voltage, positive outside, and a net translocation of protons with a stoichiometry of 1H⁺/e^–. It is very likely that nitrate reductase (NarGHI enzyme [262]) and DMSO reductase (DmsABC enzyme [412]) generate proton motive force in the same manner.

 

True Proton Pumps

In contrast to the enzymes that generate proton motive force by the simple separation of the active sites, enzymes which are authentic proton pumps utilize the redox free energy to remove protons from the cytoplasm, actively transport them through a proton-conducting channel, and deliver them to the periplasm. This is sometimes referred to as a vectorial mechanism. E. coli contains at least two true proton pumps: the cytochrome bo ₃ ubiquinol oxidase (Fig. 3E) and NDH-I (NADH:ubiquinone oxidoreductase) (Fig. 3A). The best characterized of these enzymes is cytochrome bo ₃. This enzyme catalyzes the same reaction as does cytochrome bd and delivers a substrate proton to the periplasm by the same mechanism, as pictured in Fig. 3D. This mechanism predicts a value for H⁺/e^– of 1, as found experimentally for cytochrome bd (297); however, the measured value for cytochrome bo ₃ is H⁺/e^– = 2 (297). The most reasonable explanation is that this enzyme has a second mechanism by which protons are transported to the periplasm. It is postulated that there is a proton-conducting channel and that the energy to move protons across this channel against an electrochemical gradient is provided by the chemistry at the oxygen-reductase site. This is pictured schematically in Fig. 3E. As described in more detail below (see Oxygen), cytochrome bo ₃ is a member of a large superfamily of oxidases which have been characterized as primary proton pumps (60, 172). The mechanism by which this occurs is not known.

The second respiratory enzyme that must be a true proton pump is NDH-I (408). This enzyme is a homolog of the mitochondrial NADH dehydrogenase (complex I), which is known to pump two protons across the membrane per electron. The E. coli enzyme is a coupling site and generates a proton motive force, but the H⁺/e^– ratio has not been measured. It is likely that H⁺/e^– = 2. The mechanism of proton pumping is unknown.

 

THERMODYNAMIC CONSIDERATIONS

Thermodynamics as applied to bioenergetics allows one to make estimates of the relative amounts of energy produced by different respiratory chains. These estimates help to rationalize observations of preferential electron flow and regulation of enzyme synthesis when considering multiple respiratory pathways. This section seeks to provide a basis for understanding these thermodynamic arguments and to emphasize that the numerical values presented are often only crude estimates. These estimates provide useful guidelines for comparison but must be evaluated carefully in the context of specific experimental observations (402). In particular, one must consider both the actual redox potentials of the donor and acceptor couples and the proton stoichiometries of the reactions being considered. The thermodynamic aspects of respiration have been elegantly and clearly described for both the general reader (165, 272) and the specialist (402). Only essential highlights, drawn largely from the treatment of Nicholls and Ferguson (272), are covered here.

Respiration-driven proton translocation generates a proton electrochemical gradient (

, measured in kilojoules per mole) across the cytoplasmic membrane. The free energy of the proton electrochemical gradient is used to energize ATP synthesis, motility, and solute uptake, as described elsewhere in this volume by Harold and Maloney (chapter 19). The proton electrochemical gradient consists of two distinct elements: an electrical gradient, resulting from the steady-state charge separation across the membrane (ΔΨ; negative inside), and a chemical gradient, resulting from the asymmetric proton distribution across the membrane (ΔpH; alkaline inside):

where F is the Faraday constant, R is the gas constant, and T is the absolute temperature. A related term, proton motive force (Δ p; measured in millivolts), allows for an easier comparison with the redox potential changes that are associated with respiration:

It must be emphasized that many calculations in bioenergetics are based on values determined under "standard conditions," i.e., when all reactants and products are present at a concentration of 1 M (solutes) or 1 atm (gases). Thus, the standard Gibbs energy (ΔG, measured in kilojoules per mole) for a reaction is as follows:

ΔG = –2.3RT log₁₀ K

where K is the equilibrium constant (i.e., the ratio of products to reactants at equilibrium). This value itself does not predict the actual Gibbs energy of a reaction that takes place within a cell, operating under nonstandard conditions. The actual Gibbs energy (ΔG, measured in kilojoules per mole) of a reaction is a measure of its capacity to do work, and can be estimated by

ΔG = –2.3RT log₁₀ (K/Γ)

or

ΔG = ΔG + 2.3RT log₁₀ Γ

where Γ is the observed mass action ratio (i. e., the observed ratio of products to reactants).

 

Redox Potential

Proton motive force is generated by coupled oxidation-reduction reactions, catalyzed by components of the respiratory chain. For example, the oxidation of NADH to NAD⁺ can be coupled with the reduction of O₂ to H₂O. The equilibrium values for the half-reactions

are described by their respective midpoint redox potentials at pH = 7 (E_m,7, measured in millivolts). This midpoint potential is calculated from the standard redox potential (E, measured in millivolts), which is measured at 1 M concentrations of both the oxidized and reduced components (including protons; i.e., pH = 0), in reference to a hydrogen electrode.

Most redox reactions in respiration involve protons:

where ox is the oxidized species and red is the reduced species. This results in a pH dependence of the value of E by –59 mV per pH unit when m = n, by –30 mV per pH unit when m = 1 and n = 2, and by 0 mV when m = 0. For example, E for the H⁺/H₂ couple is 0 mV, whereas E_m,7 is 7 (–59) = –414 mV. Thus, the actual redox potential at pH = 7 (E_h,7, measured in millivolts), can be calculated from the Nernst equation:

  (1)

where n is the number of electrons transferred.

A final yet significant consideration concerns the value of pH (periplasmic or cytoplasmic) to be used in calculating E. The pH of the compartment in which scalar protons are formed should be used, rather than the compartment in which the carrier is located (P. C. Hinkle, in Bioenergetics, a Practical Approach, in press).

Values of E_m,7 for couples relevant to this chapter are shown in Table 1. The difference in redox potentials between a donor and acceptor couple (ΔE_m,7) can provide a crude estimate of the available energy resulting from the coupled oxidoreduction reaction. For example, the ΔE_m,7 for NADH oxidation coupled to O₂ reduction is (–0.32) – 0.82 = –1.14 V.

However, the E_m,7 values provide only a rough approximation of the relative efficiencies of different redox reactions. The actual Gibbs energy change (ΔG) that results from electron transfer between donor and acceptor couples is a function of the actual redox potential difference (ΔE_h):

ΔG = –nFΔE_h

For a redox span ΔE_h = 1,000 mV, ΔG is –96.5 kJ mol^–1 for a one-electron transfer and –193 kJ mol^–1 for a two-electron transfer (272).

The difficulty in determining the intracellular concentrations of reactants and products complicates calculations involving the actual redox potential E_h. Therefore, thermodynamic arguments are commonly based on values for the standard (midpoint) redox potential, E_m,7. For example, the ΔE_m,7 value for NADH → O₂ respiration is –1.14 V, whereas the ΔE_m,7 value for HCO₂ ^- → NO₃ ^– respiration is –0.86 V. Thus, one might conclude that the latter respiratory chain provides only 75% as much free energy as the former. However, these values are based on the standard-state concentrations, which may have little relationship with the actual concentrations. For example, consider a two-electron transfer redox couple for which the concentrations of the oxidized and reduced forms are equal. From the Nernst equation (equation 1), we find that E_h,7 = E_m,7. However, if the concentration of the oxidized form is 1,000 times that of the reduced form at 25C, the Nernst equation yields

For a one-electron reaction, the value would be 176 mV. Thus, the actual redox potential in this hypothetical example is substantially different from the standard midpoint potential.

Finally, if the donor and acceptor couples are on opposite sides of the membrane and the electron is transferred from the negative compartment to the positive compartment, then the term for electrical potential (ΔΨ) must also be considered:

Δ G = –nF (ΔE_h + ΔΨ)

 

Proton Stoichiometry and Proton Motive Force

A specific oxidation or reduction reaction catalyzed by a respiratory enzyme may result in net proton translocation across the cytoplasmic membrane. Therefore, calculations of bioenergetic efficiency must also account for the stoichiometry of proton translocation. For a hypothetical membrane that is impermeable to proton leakage, n electrons passing through a redox span ΔE will translocate m protons against a proton motive force Δ p (272):

m Δ p = n ΔE  (2)

Thus, a higher proton stoichiometry results in a lower proton motive force at equilibrium. For both the NADH/ubiquinone (NDH-I) and ubiquinol/O₂ (cytochrome bo ₃oxidase) couples, the H⁺/e^– ratio is 2, whereas this ratio is 1 for both the formate/ubiquinone and ubiquinol/nitrate couples (see Mechanisms of Proton Translocation, above). Substituting the appropriate ΔE_m,7 values in equation 2, one calculates that the proton motive force for NADH → O₂ respiration (H⁺/e^– = 4) is –285 mV, whereas the value for HCO₂ ^– → NO₃ ^– respiration (H⁺/e^– = 2) is –430 mV, quite a different picture from that provided by consideration of the ΔE_m,7 values alone (see above). Again, these estimates are constrained by not knowing the actual redox potential differences, ΔE_h,7. Indeed, experimental estimates of proton motive force for O₂ and NO₃ ^– respiration yield values of around –200 and –160 mV, respectively (55, 300, 448). The measured value for proton motive force also depends on kinetic parameters in the steady state, so that the rates and degree of coupling are also factors under working conditions.

If increased proton stoichiometry results in decreased proton motive force, why does NADH → O₂ respiration at high aeration preferentially involve the proton-pumping cytochrome bo ₃ rather than the scalar cytochrome bd? Apparently, the strategy is not to attain the highest possible proton motive force but, rather, to generate sufficient proton motive force in the most economical manner. Sufficient proton motive force is that which is necessary to drive ATP synthesis, active transport, and motility. The enzymes involved in these processes probably operate at fixed stoichiometry (e.g., 3 H⁺/ATP), and therefore an increased proton motive force will not increase the efficiency of these enzymes. Assuming that the phosphorylation potential (ΔG_p; the Gibbs energy for the ATP synthase reaction) is about 600 mV (about 60 kJ mol^–1 [165, 272]) and that three protons are needed for synthesis of each molecule of ATP, ATP synthesis will occur at a proton motive force (Δ p) of about

or about 200 mV. This is within the range of the above theoretical and measured estimates for proton motive force. We emphasize that all of these calculations yield crude approximations that involve numerous assumptions and ad hoc postulates, which are thoroughly considered by Harold (165).

Therefore, the proton motive force generated by the proton-pumping NADH → O₂ respiratory chain is sufficient to drive ATP synthesis, and it is more efficient than a scalar chain because the pumping mechanism results in more extruded protons per molecule of substrate (66, 269). Additionally, equation 2 predicts the equilibrium state with no proton leakage, whereas growing cells under substrate-limited conditions may never attain such an equilibrium. Therefore, maximizing the number of protons extruded per substrate molecule consumed may represent the best strategy for energy generation with limited available substrate resources.

In summary, then, several constraints limit our ability to predict the bioenergetic efficiency of a given respiratory chain. The reactions occur in chemically complex cellular compartments, thereby complicating extrapolation from standard measurements made with aqueous solutions of pure reactants (402). Additionally, the stoichiometry of proton translocation further influences the maximum attainable proton motive force, a factor that is not revealed by simply examining redox potentials of the couples involved.

 

QUINONES

Quinones are lipid-soluble molecules that mediate electron transfer between protein components of respiratory chains. E.coli synthesizes three types of quinones: a benzoquinone, ubiquinone (Q), and two napthoquinones, menaquinone (MK) and demethylmenaquinone (DMK). All three quinone species contain an octaprenyl side chain (C₄₀), and thus quinones are dissolved within the lipid bilayer of the cytoplasmic membrane. Reduction and oxidation reactions involving two-electron transfers form the quinol and quinone species, respectively (Fig. 4). Some reactions (e.g., succinate:ubiquinone oxidoreductase and menaquinol:fumarate oxidoreductase) may involve sequential one-electron transfers proceeding through a semiquinone radical intermediate (409, 418). The biosynthesis of quinones is described elsewhere in this volume by Meganathan (see chapter 39).

The functions of the different quinones have been examined through mutant analysis. Mutants specifically defective in Q (ubi) and MK/DMK (men) biosynthesis have been isolated and characterized (see chapter 39). ubiE mutants lack both Q and MK but retain DMK (424). Three general types of experiments have been conducted to examine quinone functions in these mutants: examination of growth with different respiratory substrates (155, 218, 247, 423), examination of coupled enzyme activities in membrane preparations from different mutants (423, 424), and reconstitution of coupled enzyme activities by incorporation of quinones into membranes of quinone-deficient mutants (401, 424).

The general pattern that has emerged from these studies is that Q is used for oxygen respiration, both Q and MK are used for nitrate respiration, and both MK and DMK are used for anaerobic respiration with acceptors other than nitrate (Table 2). However, results from quinone reconsitution studies must be interpreted cautiously. Reconstitution of membranes from a menA mutant led to the conclusion that fumarate respiration and DMSO respiration specifically require MK (423), whereas growth studies and measurements of hydrogen-dependent enzyme activities with a ubiE (Q^– MK^– DMK⁺) mutant show that DMK can support both fumarate and DMSO respiration (424).

Aerated cells contain about four or five times more Q than MK plus DMK, whereas anaerobic cells contain about one-third as much Q as MK plus DMK (401, 424). This provides some degree of specificity for the enzymes that are coinduced under anaerobic conditions and also has a rationale in that the midpoint potential of the MK/menaquinol couple in the membrane is about 200 mV less than for the Q/ubiquinol couple (Table 1). Hence, MK is better suited for a respiratory chain with lower potential electron acceptors. Even so, examination of relative midpoint potentials does not satisfactorily explain all the observations. For example, the DMK/demethylmenaquinol couple seems well situated to donate electrons to nitrate but apparently fails to do so (Table 2). The basis for the regulation of quinone levels is unknown, although it is established that MK and DMK synthesis does not require an active fnr gene (385). The relative proportions of MK and DMK also vary with growth conditions (385, 401, 423, 424).

The specificity of a particular quinone for a given respiratory chain may also reflect specific structural features of quinones. Certain mutants blocked in late steps of Q biosynthesis accumulate octaprenyl Q analogs that lack various ring substituents. At least one of these precursors (MMQ) functions in oxygen respiration with NADH, d-lactate, and glycerol 3-phosphate as electron donors but not with succinate as the electron donor (400). However, it is also possible that MMQ has altered redox properties that make it unsuitable for use with the fumarate/succinate couple, which has a rather high midpoint potential (Table 1).

Reconstitution experiments led to the conclusion that Q is specifically required for NADH oxidation with either oxygen or nitrate as the electron acceptor, although consideration of midpoint potentials suggests that MK should also be a satisfactory carrier from NADH to nitrate and other anaerobic electron acceptors. Additionally, the Q concentration is lowered by about 10-fold in anaerobic cultures (401, 424). Thus, the quinone species that couples NADH oxidation to anaerobic electron acceptors warrants further examination.

 

HEMES

Many of the respiratory enzymes are heme proteins (141). Heme biosynthesis is considered elsewhere in this volume by Beale (see chapter 49). Figure 5 shows the structures of hemes that are in respiratory components: hemes B, C, D, and O. Capital letters are used to designate the heme itself (e.g., heme B), whereas lowercase letters refer to the protein-bound species (e.g., heme b or cytochrome b). Hemes C, D, and O are derived from heme B. In addition to these four heme types, E. coli contains two different hemes in nonrespiratory proteins. Siroheme (266) is found in the biosynthetic sulfite reductase (CysIJ enzyme) and in the cytoplasmic anaerobic nitrite and sulfite reductases (NirBD and AsrABC enzymes, respectively). A variant of heme D is present in catalase II (HP-II [84]). Whereas the hydroxyls in heme D in cytochrome bd are trans (Fig. 5), those in HP-II are cis.

Heme B (protoheme IX) is found in the membrane anchoring polypeptides in at least three classes of respiratory enzymes, the aerobic succinate dehydrogenase, the anaerobic nitrate reductases, and the respiratory formate dehydrogenases. Hydrogenase-1 may also contain heme B (see Cytochromes and Membrane Anchor Subunits, below).

Heme C refers to heme B which is covalently bound to the protein by two thioether linkages (Fig. 5) between the heme vinyl groups and the two cysteines within a Cys-X-X-Cys-His motif in the protein sequence. In E. coli and Salmonella typhimurium (official designation, Salmonella enterica serovar Typhimurium), all of the cytochromes c are expressed anaerobically and have no role in aerobic respiration (see Cytochromes and Membrane Anchor Subunits, below). All cytochromes c are either located within the periplasm as soluble proteins or anchored in the cytoplasmic membrane with the heme facing the periplasm. Three anaerobic respiratory enzymes are associated with cytochromes c: respiratory nitrite reductase, trimethylamine N-oxide (TMAO) reductase, and periplasmic nitrate reductase.

Heme D (350, 376) and heme O (431) (Fig. 5) are each associated with one of the terminal oxidases. Heme d is the dioxygen binding site in the cytochrome bd ubiquinol oxidase, and heme o is the dioxygen binding site in the cytochrome bo ₃ ubiquinol oxidase. Whereas heme B is found in numerous enzymes, heme D and heme O have been found only in the prokaryotic respiratory oxidases. As indicated in Fig. 5, heme O has been shown to be the biosynthetic precursor of heme A (314, 368), which is present in the respiratory oxidases of many bacteria, and is also the heme component of the eukaryotic cytochrome oxidase. However, E. coli cannot convert heme O to heme A (368). It has been suggested that in HP-II, heme D may be formed nonenzymatically (234). The reason why heme O and heme D are present at the dioxygen-binding sites of the terminal oxidases is not clear, since there are examples other bacterial oxidases in which heme B can serve this function (138).

 

ELECTRON DONORS

 

Formate

Formate (HCO₂ ^–) is a product of anaerobic pyruvate cleavage by pyruvate-formate lyase, as described elsewhere in this volume by Kessler and Knappe (see chapter 15). E. coli and S. typhimurium both metabolize formate by two distinct routes. In the absence of exogenous electron acceptors, formate is converted to carbon dioxide and dihydrogen by the enzyme complex formate-hydrogen lyase, as described elsewhere in this volume by Bck and Sawers (see chapter 18). In the presence of exogenous electron acceptors, formate oxidation to carbon dioxide is coupled to acceptor reduction, forming an anaerobic respiratory chain. Oxygen, nitrate, nitrite, DMSO, TMAO, and fumarate all serve as electron acceptors for respiratory formate oxidation in E. coli (1, 286, 423, 441).

In vitro assays for formate dehydrogenase activity are carried out with artificial electron acceptors. Formate dehydrogenase-H (FdhF-HycB enzyme), a component of the formate-hydrogen lyase enzyme complex, efficiently reduces benzyl viologen, whereas the respiratory enzymes, formate dehydrogenase-N (FdnGHI enzyme) and formate dehydrogenase-O (FdoGHI enzyme), exhibit strong phenazine methosulfate-mediated dichlorophenolindophenol (PMS-DCPIP) and methylene blue reductase activities (reviewed in reference 357). All three formate dehydrogenase large subunits contain selenocysteine, which is encoded by an opal (UGA) codon within the respective structural genes, as described in chapter 18. The large subunits also contain molybdenum cofactor (227, 286). Both Q and MK serve as the immediate electron acceptors for respiratory formate oxidation (see Quinones, above).

Formate dehydrogenase-N synthesis is induced during anaerobic growth with nitrate (227) but is not stimulated by other anaerobic electron acceptors such as nitrite and TMAO (32). Formate dehydrogenase-O is synthesized at a low level that is indifferent to anaerobiosis and nitrate (288), whereas formate dehydrogenase-H synthesis requires anaerobiosis and is inhibited by nitrate, as described in chapter 18. Therefore, formate dehydrogenase-N is the dehydrogenase component of the formate-nitrate respiratory chain. Which formate dehydrogenases are involved in respiration with other electron acceptors? This question has been addressed in the context of nitrite respiration, with mutants defective in one or more of the formate dehydrogenases. The results indicate that all three formate dehydrogenases contribute to formate-nitrite respiration, although the involvement of formate dehydrogenase-O is relatively minor (103). It is also hypothesized that formate dehydrogenase-H can donate electrons to fumarate reductase via hydrogenase (see Hydrogen, below). Thus, it seems that formate dehydrogenase-H can participate in anaerobic respiration, even though it is usually considered to be a fermentation enzyme. Formate dehydrogenase-H is further described in chapter 18 of this volume.

Formate Dehydrogenase-N (Formate:Quinone Oxidoreductase-N; FdnGHI Enzyme).

Formate dehydrogenase-N contains three subunits: a large selenomolybdoprotein containing the catalytic site (FdnG), and smaller iron-sulfur (FdnH) and cytochrome b (FdnI) subunits (Table 3) (31, 124). Properties of these subunits are summarized below (see Common Features of Respiratory Enzymes).

Formate oxidation in nitrate-grown cells supports transmembrane proton translocation with an estimated ratio of H⁺/e^– = 2 with quinone analogs as electron acceptors and H⁺/e^– = 4 with nitrate or oxygen as electron acceptors (199). This shows that formate dehydrogenase-N catalyzes net proton translocation from the cytoplasm to the periplasm, presumably by a scalar mechanism (see Mechanisms of Proton Translocation, above). However, the topology of formate dehydrogenase-N within the cytoplasmic membrane is not established, thereby complicating attempts to explain the mechanism of proton translocation.

Formate dehydrogenase-N is encoded by the fdnGHI operon located at 32 min on the E. coli genetic map (31, 32). Enzyme synthesis is maximal during anaerobic growth with nitrate, and analysis of fdnG-lacZ fusion expression shows that anaerobic regulation and nitrate regulation are mediated by the Fnr and NarL proteins, respectively (32) (see Regulation of Enzyme Synthesis, below).

Formate Dehydrogenase-O (Formate:Quinone Oxidoreductase-O; FdoGHI Enzyme).

An isoenzyme of formate dehydrogenase-N, formate dehydrogenase-O, has recently been recognized (288, 322). This enzyme has also been termed formate dehydrogenase-Z (288). Formate dehydrogenase-O is encoded by the fdoGHI operon located at 88 min on the E. coli genetic map (287). Formate dehydrogenases O and N are structurally and immunologically closely related (287, 288). Formate dehydrogenase-O is synthesized at relatively low levels irrespective of the presence of oxygen or nitrate (288). The physiological role of this isoenzyme is not known, but it may serve to couple formate oxidation to the reduction of a variety of electron acceptors, including oxygen (see above).

Other Proteins Involved in Formate Metabolism.

All three formate dehydrogenases contain selenocysteine, cotranslational incorporation of which is catalyzed by the selA–D gene products as described elsewhere in this volume by Bck and Sawers (see chapter 18). Genetic studies have revealed two additional genes required for formate metabolism in both E. coli and S. typhimurium, fdhD and fdhE (termed fdnC and fdnB, respectively, in S. typhimurium [242, 281, 327, 360]). Genetic and physical mapping shows that the fdhE gene is probably part of an fdoGHI-fdhE operon whereas the fdhD gene is immediately upstream of and divergently transcribed from the fdoGHI-fdhE cluster (287, 327). Evidence suggests that both gene products act posttranslationally in the formation of active formate dehydrogenase, but their mode of action is unknown (242, 281, 288, 360).

 

Glucose

d-Glucose Dehydrogenase (d-Glucose:Quinone Oxidoreductase; Gcd Enzyme).

d-Glucose dehydrogenase is a single-subunit enzyme (87 kDa) encoded by the gcd gene located at 3 min on the E. coli genetic map (87). Topological studies indicate the enzyme has five transmembrane helical spanning segments (439). The active site of d-glucose dehydrogenase faces the periplasm. The enzyme oxidizes glucose to gluconate and reduces ubiquinone. The cofactor used by this enzyme is pyrroloquinoline quinone (PQQ), which E. coli is not capable of synthesizing. Hence, for this enzyme to function, exogenous PQQ must be provided in the medium. In the absence of PQQ, d-glucose dehydrogenase is present as an apoprotein. The predicted binding site for PQQ is located on the periplasmic side of the membrane (439). No proton motive force is generated by the reaction catalyzed by this enzyme, so it is presumed that the quinone reductase site also faces the periplasm (246, 439). The enzyme has been purified and reconstituted along with the cytochrome bo ₃ quinol oxidase and Q-8 in phospholipid vesicles to form a minimum functioning respiratory chain (246, 439). Similar experiments have been reported with both d-lactate dehydrogenase (245) and pyruvate oxidase (210).

 

Glycerol 3-Phosphate

E. coli and S. typhimurium expresse two respiratory glycerol-3-phosphate dehydrogenases. Both enzymes oxidize sn-glycerol-3-phosphate to dihydroxyacetone phosphate and reduce quinone in the cytoplasmic membrane. Their function is to salvage glycerol and glycerol phosphates generated from the breakdown of phospholipids and triacylglycerol. Glycerol oxidation is obligately coupled to electron acceptor reduction, so E. coli and S. typhimurium must be provided with an electron acceptor for anaerobic growth with glycerol as the sole carbon source. This simple test is widely used to determine if a particular strain or mutant is capable of anaerobic respiration with a specific electron acceptor. Other enterobacteria, such as Klebsiella spp., can ferment glycerol in the absence of exogenous electrons acceptors, as described elsewhere in this volume by Bck and Sawers (see chapter 18).

Aerobic Glycerol-3-Phosphate Dehydrogenase (Glycerol-3-Phosphate:Quinone Oxidoreductase; GlpD Enzyme).

A single-subunit enzyme (57 kDa) catalyzes glycerol 3-phosphate oxidation with either oxygen or nitrate as the electron acceptor. The GlpD enzyme is maximally synthesized under aerobic growth conditions and is necessary for aerobic growth on glycerol. It has been purified and shown to be a dimer of identical subunits (331), and it uses noncovalently bound FAD as a cofactor. The sequence does not suggest that this is a transmembrane enzyme (10).

Aerobic glycerol-3-phosphate dehydrogenase is encoded by the glpD gene located at 75 min on the E. coli genetic map, in an operon with two genes of unknown function (332). Expression of the glpGED operon is controlled by glycerol and by catabolite repression (444). Enzyme synthesis is maximal during aerobic growth, and anaerobic repression is mediated by the ArcA protein (185) (see Regulation of Enzyme Synthesis, below).

Anaerobic Glycerol-3-Phosphate Dehydrogenase (Glycerol-3-Phosphate:Quinone Oxidoreductase; GlpABC Enzyme).

A three-subunit enzyme catalyzes glycerol 3-phosphate oxidation coupled to electron acceptors other than oxygen. The GlpABC enzyme is loosely associated with the membrane, and a two-subunit form (GlpAC) has been isolated and characterized. The GlpA protein (62 kDa) contains noncovalently bound FAD, and it is suggested that the GlpC protein (43 kDa) may bind flavin mononucleotide (FMN) as a cofactor (93). The GlpB polypeptide (44 kDa) has sequences suggesting two iron-sulfur clusters (see Iron-Sulfur Subunits, below), and this putative third subunit may also act as a membrane anchor for the GlpAC complex (93). However, the sequence of the GlpB protein does not indicate any transmembrane hydrophobic helices, so the mode of attachment of this enzyme to the membrane and the role of GlpB remain unknown. It has been suggested that putative amphipathic helices in GlpB may be transmembrane elements (93), leaving open the possibility that this enzyme functions as a coupling site. This has yet to be demonstrated.

Anaerobic glycerol-3-phosphate dehydrogenase is encoded by the glpACB operon (216) located at 49 min on the E. coli genetic map (93). (Note that the glpACB operon is designated as glpABC in reference 93, but the former terminology is most often used.) Expression of the glpACB operon is controlled by glycerol and by catabolite repression (216). Enzyme synthesis is maximal during anaerobic growth with fumarate, and analysis of glpA-lacZ fusion expression shows that anaerobic expression is mediated largely by the Fnr protein (185) (see Regulation of Enzyme Synthesis, below).

 

Hydrogen

E. coli and S. typhimurium can both use hydrogen as an electron donor for anaerobic respiration, with nitrate, DMSO, TMAO, and fumarate as electron acceptors (35, 240, 326, 423, 442). Three distinct membrane-bound hydrogenase isoenzymes are involved in hydrogen metabolism: hydrogenase-1 (HyaABC enzyme), hydrogenase-2 (HybABC enzyme), and hydrogenase-3 (HycEG enzyme). All three enterobacterial hydrogenase isoenzymes contain nickel, incorporation of which is discussed elsewhere in this volume by Bck and Sawers (see chapter 18). The properties and structures of hydrogenases from diverse bacterial species have been reviewed (296, 429). Hydrogen metabolism from the perspective of fermentation is also discussed in chapter 18. Earlier studies on hydrogen metabolism in enterobacteria have been reviewed (357).

Hydrogenases 1 and 2 oxidize dihydrogen to protons, donating the electrons to the quinone pool (hydrogen uptake). Q, MK, and DMK all serve as electron acceptors for dihydrogen oxidation (see Quinones, above). Both enzymes contain nickel, iron, and acid-labile sulfide (18, 325).

Hydrogenase-3 is a component of the cytoplasmic membrane-associated formate-hydrogen lyase complex, which helps to maintain pH balance during fermentative growth by converting formate to carbon dioxide. This enzyme complex uses the electrons derived from formate oxidation to reduce protons to dihydrogen (hydrogen evolution). Several proteins of the formate-hydrogen lyase complex share structural similarity with proteins of the proton-translocating NADH dehydrogenase (NDH-I; Nuo enzyme), including the presumed Q-binding subunit (see NADH, below). Whether the formate-hydrogen lyase complex plays a role in generating proton motive force is unknown.

Hydrogenases 1 and 2 contain homologous nickel-binding large subunits, HyaB (66 kDa) and HybC (63 kDa), respectively. The predicted amino acid sequences of these subunits do not reveal potential membrane-spanning regions (252, 253), and both sequences are homologous to soluble periplasmic hydrogenases from a variety of organisms (reviewed in references 296 and 429). Thus, both enzymes presumably face the periplasm.

The small subunits, HyaA (41 kDa) and HybA (34 kDa), contain amino-terminal cleaved leader sequences that include the Arg-Arg-X-Phe-X-Lys motif found in hydrogenase small subunit leader sequences (reviewed in references 392 and 429). This motif is also found at the amino termini of periplasmic nitrate reductases and some formate dehydrogenases. The large subunits do not contain cleaved leader sequences, and the mode of hydrogenase export to the periplasm is not known; one idea is that the two subunits form a cytoplasmic complex prior to export. The small subunits also contain presumed membrane anchor regions in their carboxyl termini, consisting of a series of hydrophobic residues followed by a region containing several charged (mainly basic) residues. Finally, possible third subunits of both enzymes, HyaC (28 kDa) and HybB (44 kDa), may be cytochromes b (252).

Hydrogen oxidation supports transmembrane proton translocation with an estimated H⁺/e^– ratio of 2 and 4 with fumarate and nitrate as the electron acceptors, respectively (200). This suggests that hydrogenase catalyzes net proton translocation from the cytoplasm to the periplasm, presumably by a scalar mechanism: two protons are released in the periplasm as a result of dihydrogen oxidation, and two protons are consumed from the cytoplasm to reduce quinone to quinol (see Mechanisms of Proton Translocation, above).

Hydrogenase-1 (Dihydrogen:Quinone Oxidoreductase 1; HyaABC Enzyme).

The physiological role of hydrogenase-1 in anaerobic metabolism remains unclear (reviewed in reference 357). The close proximity of the operons encoding hydrogenase- 1 (hyaABCDEF) and a cytochrome bd homolog (cyxAB), as well as their coordinate regulation (6, 58), may suggest that the respective enzymes form an H₂ → O₂ respiratory chain (see also Oxygen, below).

The HyaB (small) subunit probably contains two [4Fe-4S] clusters (253), on the basis of analogy to hydrogenases from other organisms (296), but the arrangement of Cys residues is atypical in comparison with ferredoxin-type iron-sulfur proteins (see Iron-Sulfur Subunits, below). Nonetheless, 10 Cys residues that may be involved in forming iron-sulfur clusters are conserved in small subunits from a variety of organisms (253).

Hydrogenase-1 is encoded by the hyaABCDEF operon, located at 22 min on the E. coli genetic map (253, 254). The hyaDEF genes encode proteins involved in hydrogenase maturation as described in chapter 18. Analysis of hya-lacZ operon fusion expression indicates that enzyme synthesis requires anaerobic growth conditions and is repressed by nitrate. Anaerobic regulation is mediated by the ArcA and AppY proteins (58) (see Regulation of Enzyme Synthesis, below). The Fnr protein only indirectly regulates anaerobic hyaA operon expression, by controlling expression of an operon necessary for nickel uptake (430). The mechanism of nitrate regulation is unknown but seems not to involve the NarL protein (58).

Hydrogenase-2 (Dihydrogen:Quinone Oxidoreductase 2; HybABC Enzyme).

Hydrogenase-2 makes up the majority of hydrogenase activity, particularly in glycerol-fumarate respiring cultures (reviewed in reference 357). Threrefore, it is presumed that hydrogenase-2 is mainly responsible for hydrogen respiration, although hydrogenase-1 may also play a significant role under certain conditions.

The sequence of the HybB (small) subunit (252) predicts that it contains four ferredoxin-type [4Fe-4S] cluster motifs (see Iron-Sulfur Subunits, below). Most other known hydrogenase small subunits, including the HyaA subunit, do not contain analogous motifs.

Hydrogenase-2 is encoded by the hybABCDEFG operon located at 65 min on the E. coli genetic map (252). The hybDEFG genes encode proteins involved in hydrogenase maturation, as described in chapter 18. Enzyme synthesis requires anaerobic growth conditions and is repressed by nitrate (58, 324, 326). The regulators involved in anaerobic and nitrate repression have not been reported. The Fnr protein only indirectly regulates anaerobic hybA operon expression, by controlling expression of an operon necessary for nickel uptake (430).

 

d- and l-Lactate

d-Lactate Dehydrogenase (d-Lactate:Quinone Oxidoreductase; Dld Enzyme).

d-Lactate dehydrogenase catalyzes the oxidation of d-lactate to pyruvate coupled to the reduction of Q. The enzyme has been purified and extensively studied as a model for examining protein-lipid interactions (reviewed in reference 171). d-Lactate dehydrogenase consists of a single subunit (64 kDa) encoded by the dld gene located at 47 min on the E. coli genetic map, and the enzyme uses noncovalently bound flavin adenine dinucleotide (FAD) as a cofactor. The deduced amino acid sequence (311) indicates no transmembrane hydrophobic segments, so it is assumed that the enzyme attaches to the surface of the cytoplasmic membrane, presumably interacting with the lipid bilayer.

Although not considered to be a transmembrane protein, d-lactate dehydrogenase is firmly attached to the membrane and requires detergents or harsh chaotropic agents to be released. The mode of membrane binding is not known, but there are data to suggest a conformational change upon interaction with lipids, and enzyme turnover is stimulated about fivefold in the presence of a variety of lipids and detergents (171, 381). The natural electron acceptor is quinone, and all the chemistry is catalyzed on the cytoplasmic surface of the membrane. Hence, electron transfer from d-lactate to quinone is not coupled to the generation of a proton motive force. Furthermore, the enzyme can be reconstituted on the outer surface of E. coli membrane vesicles and reduce the quinone pool as efficiently as it does from the cytoplasmic surface (245). The enzyme can also be reconstituted with lipid vesicles and shown to reduce Q-8 (245). There is no evidence to suggest any additional subunit in the membrane to which the single-subunit d-lactate dehydrogenase might dock .

L-Lactate Dehydrogenase (L-Lactate:Quinone Oxidoreductase; LldD Enzyme).

L-Lactate dehydrogenase participates in respiration with a variety of terminal electron acceptors, including oxygen, nitrate, and fumarate (273). L-Lactate dehydrogenase has been purified (136) as a single subunit, FMN-dependent enzyme (43 kDa). As is the case with most of the single-subunit dehydrogenases, L-lactate dehydrogenase does not appear to have transmembrane elements. The deduced amino acid sequence is homologous to both eukaryotic and prokaryotic FMN-dependent enzymes but shares no similarity with the FAD-dependent dehydrogenases in E. coli, including D-lactate dehydrogenase (Dld enzyme) and the aerobic glycerol-3-phosphate dehydrogenase (GlpD enzyme).

l-Lactate dehydrogenase is encoded by the lldD (formerly lctD) gene, located at 80 min on the E. coli genetic map, in the lldDRP operon with genes for a regulator and a permease (113). Enzyme synthesis is maximal during aerobic growth with l-lactate, and analysis of lldD-lacZ fusion expression shows that anaerobic repression is mediated by the ArcA protein (113) (see Regulation of Enzyme Synthesis, below). Induction is specific for l- rather than d-lactate, and an lldD null strain retains the ability to grow with d-lactate as the sole carbon source (113).

 

NADH

One of the areas in which there has been substantial recent progress has been the clarification of the respiratory NADH dehydrogenases, or NADH:ubiquinone oxidoreductases (reviewed in references 127, 134, 345, 414, 436, and 437). It is now evident that there are two distinct NADH dehydrogenases, which are called NDH-I and NDH-II, whose properties are summarized below. NDH-I is a 14-subunit intrinsic membrane protein, which is a homolog of the eukaryotic mitochondrial complex I (408). This enzyme utilizes the free energy available from the redox reaction to pump protons across the cytoplasmic membrane. Although this has not het been measured directly, by analogy with the eukaryotic enzyme (414) it is likely that NDH-I pumps 2H⁺/e^–, as indicated below. This is shown schematically in Fig. 3. The natural electron acceptor is Q within the cytoplasmic membrane:

5H⁺ _in + NADH + Q

NAD+ + QH₂ + 4H⁺ _out

Much of the research on NDH-I is motivated by the fact that it is a primary proton pump and is essentially a stripped-down version of the 42-subunit mitochondrial complex I. By contrast, NDH-II is a one-subunit peripheral membrane enzyme (194) that carries out the NADH:ubiquinone oxidoreductase reaction at the inner surface of the cytoplasmic membrane and is not coupled to the generation of a proton electrochemical gradient across the membrane. Its function is simply to oxidize NADH and feed the electrons into the respiratory chain. It seems to be common for bacteria to possess both coupled (NDH-I) and noncoupled (NDH-II) NADH dehydrogenases (436, 437), whose concentrations are optimized for the physiological needs under different growth conditions. This is discussed in greater detail above (see Modularity of Respiratory Components).

NADH Dehydrogenase-I (NADH:Ubiquinone Oxidoreductase-I; NDH-I; NuoA-N Enzyme).

Over the past 8 years, the status of NDH-I has risen from a point when its existence was questioned to the current position, in which its genetic locus (nuoA–N; located at 49 min on the E. coli genetic map) has been cloned and sequenced (408) and the protein has been purified and subjected to biochemical and biophysical characterization (223).

Early bioenergetics studies indicated that E. coli must contain a respiratory NADH dehydrogenase which is coupled to the transmembrane proton electrochemical gradient. For example, membrane vesicles demonstrate uncoupler-sensitive, ATP-dependent reduction of NAD⁺ by substrates such as succinate (290), and this "reversal" of electron transfer is possible only with assistance from the proton motive force acting on the NADH dehydrogenase, so that it functions as an NAD⁺ reductase. Also, it was shown that under some conditions, oxidation of NADH by the respiratory chain in whole cells results in two additional protons pumped across the membrane in comparison with oxidation of other substrates, such as succinate (221). This also indicates a primary proton pump associated with NADH dehydrogenase. This is not always observed, however, and is easily explained in the light of current knowledge by the existence of two separate NADH dehydrogenases, only one of which is a proton pump.

The study with membrane preparations by Matsushita et al. (245) was the first definitive biochemical demonstration of two different NADH dehydrogenases coexisting in the E. coli membrane. These studies relied on the ability of NDH-I to utilize the substrate deamino-NADH, whereas NDH-II does not oxidize this substrate. Early efforts to purify the membrane-bound NADH dehydrogenases were fraught with difficulty, demonstrating lability problems and multiple species depending on the substrate used as an electron acceptor. The purification of NDH-II was facilitated by its overproduction by a strain harboring the cloned ndh gene (194). The purification of NDH-I was hindered until very recently by lability problems.

Genetic studies at first indicated a single NADH dehydrogenase, now known as NDH-II (445). A mutant was isolated on the basis of the presumption that cells lacking NADH dehydrogenase activity should be unable to eliminate excess NADH produced when the cells metabolize mannitol and hence should be unable to grow when mannitol is the sole energy source. Such cells should, however, be able to grow fermentatively on glucose. Some mutants obtained from such a screen had the expected deficiency in respiratory NADH dehydrogenase and were used to clone the ndh gene (mapped at 22 min) which encodes NDH-II. Subsequent studies (62), however, showed that this mutant also carries a second lesion at 49 min, corresponding to the nuo locus encoding NDH-I.

   Cloning and sequencing of the nuo locus. The operon encoding the subunits of NDH-I was cloned by using DNA probes based on the protein sequence of the FMN- and NADH-binding subunit of the mitochondrial complex I (408). The genomic fragment forming the nuo (NADH-ubiquinone oxidoreductase) locus contains 15,862 bp and encodes 14 polypeptides, all of which are found in the purified NDH-I preparation. All 14 subunits, named NuoA through NuoN (originally Nuo1 to Nuo14), have homologs in the 40-plus subunit mitochondrial complex I. In addition, the 14-gene nuo operon from E. coli corresponds to the recently reported 14-gene nqo cluster encoding the subunits of NDH-I from Paracoccus denitrificans (432, 433). The sequence relationships between the E. coli Nuo subunits and those of the mitochondrial and P. denitrificans complex I have been helpful in assigning functional roles to individual subunits (Table 4).

  The NuoF subunit. Photoactivable derivatives of NAD⁺ have been used to identify the 51-kDa subunit of both the bovine enzyme (FP51 [108]) and the corresponding subunit from P. denitrificans (Nqo1 [438]) as the NADH-binding subunit. The E. coli homolog is NuoF, which is therefore presumed to have the same function in NDH-I. It is presumed that the FMN-binding site is on the same subunit, since FMN appears to be the immediate oxidant of NADH. In the bovine enzyme, a soluble subcomplex called the FP (flavoprotein) complex, consisting of the FP51, FP24, and FP10 subunits, contains both the FMN- and NADH-binding sites. Neither the FP24 nor the FP10 subunits are attractive candidates for the FMN-binding site, so it is likely that the bovine FP51 and the corresponding NuoF subunit are FMN-binding proteins. NuoF is also predicted to contain a single [4Fe-4S] cluster, based on sequence analysis, as is the corresponding FP51 protein of the mitochondrial enzyme. Recent gene interruption experiments with the corresponding subunit from Neurospora crassa (128) have identified this subunit as the site of the FMN and center N3, one of the electron paramagnetic resonance (EPR)-detectable iron-sulfur clusters in complex I (345). This same iron-sulfur cluster has been identified in E. coli NDH-I (345), so, again by analogy, it is likely that iron-sulfur cluster N3 is also located within NuoF.

  The NuoH subunit. The NuoH subunit corresponds to subunit ND1 in bovine complex I (408). This hydrophobic subunit (eight predicted transmembrane spans) has been identified as the binding site of a potent inhibitor of complex I, rotenone, based on labeling with photoactivatable derivatives of rotenone and dihydrorotenone (118). On the basis of the arguable assumption that the rotenone-binding site should at least partially overlap the Q-binding site, ND1 has been tentatively assigned as the Q-binding subunit (127, 408). Sequence similarities of ND1 and the bacterial glucose dehydrogenase (glucose:ubiquinone oxidoreductase) from Acinetobacter calcoaceticus (132) have also been used to buttress the claim that ND1 contains the Q-binding site, but the degree of sequence similarity is weak and its significance has been questioned (127). It should be noted that a photoactivable derivative of Q has been used to directly identify its binding site in complex I from N.crassa (168). The subunit labeled, however, is a 9.6-kDa nucleus-encoded subunit that has no Nuo homolog. Hence, it has been argued that the conclusion that this 9.6-kDa subunit is the primary determinant of the Q-binding site in complex I must be incorrect (133).

  Other Nuo subunits. The sequence analysis of the Nuo polypeptides (408) reveals additional information (Table 4). Five of the subunits contain cysteines in motifs suggesting that they contain iron-sulfur clusters (see Iron-Sulfur Subunits, below). Two of these five subunits are predicted to contain multiple clusters: NuoG is predicted to have two [2Fe-2S] and probably two [4Fe-4S] clusters, and NuoI is expected to have two [4Fe-4S] clusters. One cysteine motif in NuoG is not conserved in other species, making it a reasonable candidate for ligation of the [2Fe-2S] cluster N1c, which is unique for the E. coli enzyme. Analysis of NDH-I in situ (i.e., in membrane preparations) by EPR spectroscopy has identified six distinct signals (375), and the iron content of the purified enzyme is also consistent with a large number of iron-sulfur clusters (223) (see below).

The hydrophobicity analyses of the deduced amino acid sequences (408) predict that 7 of the 14 Nuo subunits contain transmembrane helices (Table 4). These subunits correspond to all seven of the mitochondrially encoded subunits of the eukaryotic complex I; most of the corresponding genes are clustered at the end of the nuo operon. Two subunits, NuoB and NuoI, have segments that can be interpreted as single transmembrane spans, but both NuoB and NuoI correspond to subunits in the bovine enzyme that are considered hydrophilic (127). The bovine subunit corresponding to NuoI is clearly part of a soluble subcomplex and is unlikely to be transmembranous (129). Furthermore, NuoB and NuoI, together with NuoG and NuoD, have been identified as components of a membrane-associated subcomplex of E.coli NDH-I (H. Weiss et al., personal communication), and they are unlikely to have transmembrane elements.

Figure 6 is a diagrammatic model of NDH-I, showing the arrangement of subunits in a transmembrane segement associated with a hydrophilic peripheral segment. This aspect is consistent with what is known about the mitochondrial enzyme and is also consistent with biochemical studies on NDH-I.

  Sequence relationships with other proteins. Eight of the subunits of NDH-I are related by sequence analysis to subunits of other enzymes. These eight subunits are predicted to contain all of the iron-sulfur clusters in NDH-I as well as the FMN-, NADH-, and Q-binding sites. NuoE, NuoF, and NuoG are related to portions of the α and γ subunits of the soluble NAD⁺-reducing hydrogenase of Alcaligenes eutrophus (127, 280, 285). These two subunits of this four-subunit hydrogenase are associated with its NAD⁺ reductase activity (378). This would argue that NuoE, NuoF, and NuoG should be considered to represent a structural and functional unit within NDH-I that is involved in the oxidation of NADH. As discussed previously, NuoF has been implicated as the NADH- and FMN-binding subunit by other arguments, and all three subunits are predicted to contain a total of four iron-sulfur clusters. The relationship between complex I and the NAD⁺-reducing hydrogenase has been discussed in some depth previously (127, 285).

There is also a clear relationship between the Nuo subunits (and their corresponding components of the eukaryotic complex I) and subunits of the hydrogenase associated with the E. coli formate-hydrogen lyase (48, 127). It has been suggested that NuoB, NuoD, NuoH, NuoI, and NuoL are related to the HycG, HycE, HycD, HycF, and HycC subunits of this hydrogenase. Two of these subunits are predicted to be intrinsic membrane proteins (NuoH and NuoL), and NuoB and NuoI are predicted to contain a total of three iron-sulfur clusters. NuoH is also implicated as being a Q-binding site, as discussed above.

The functional meaning of the relationship with the subunits of formate-hydrogen lyase is not clear. Nevertheless, the data do suggest that NuoB, NuoD, NuoH, NuoI, and NuoL may form a discrete functional and structural unit within NDH-I. The relationships of the subunits in Fig. 6 have been drawn to reflect this.

  Biochemical studies of NDH-I. Only preliminary biochemical characteristics of NDH-I have been reported (223). The enzyme is very labile, but this problem appears to have been overcome to some extent. The purified enzyme contains all 14 subunits encoded by the nuo operon, and there is no evidence of additional polypeptide components. The hydrodynamic properties of the purified enzyme are consistent with the predicted total molecular mass of 525 kDa, and the purified enzyme contains one equivalent of bound FMN and 20 to 24 equivalents of nonheme iron. This latter value is in the expected range, given the six spectroscopically resolved iron-sulfur clusters (three [2Fe-2S] plus three [4Fe-4S] clusters) and the number predicted from the sequence analysis (seven putative clusters [Table 4]). The purified enzyme oxidizes NADH (K_m, 10 μM) and reduces Q-2 (K _m, 2 μM) with a turnover number of about 10 s^–1.

It also appears that a water-soluble subcomplex of NDH-I can be isolated, different from those of the mitochondrial complex I (Weiss et al., personal communication). Initial data indicate that a peripheral component with structural integrity contains three of the subunits: NuoD, NuoE, and NuoF. This information has been incorporated in the drawing of the peripheral component indicated in Fig. 6.

  Inhibitors. The mitochondrial complex I is inhibited by a number of naturally occurring and synthetic compounds that block electron transfer from the high potential iron-sulfur cluster (center N2) to Q (Fig. 6). Recent studies have shown that these inhibitors can be classified in two groups (131, 133). Class I inhibitors act in a partially competitive manner with quinone and also block the activity of glucose:ubiquinone oxidoreductase from Gluconobacter oxidans. Included in this group are piericidin A and annonin VI. Presumably, these act directly at the quinone binding site. Class II inhibitors have no effect on the glucose:ubiquinone oxidoreductase and are noncompetitive with Q in complex I. Rotenone is a class II inhibitor. Since piericidin A and rotenone are reported to displace each other from the bovine enzyme, it is likely that the binding sites partially overlap. E. coli NDH-I is significantly less sensitive to inhibition by the classical inhibitors of complex I, piericidin A and rotenone (K_i, 4.5 and 8.5 μM, respectively, for membrane-bound NDH-I). The most potent inhibitor of the E. coli enzyme is annonin VI, which has a K_i of 0.03 μM. Other inhibitors that block NDH-I are capsaicin (435), galloylglucose (212), and dicyclohexylcarbodiimide (DCCD) (434).

  Iron-sulfur clusters. The sequence analysis of the polypeptides predicts the possibility of seven iron-sulfur clusters in NDH-I, and six have been resolved by EPR spectroscopy of the enzyme in situ (345). There are three [2Fe-2S] clusters (N1a, N1b, and N1c) and three [4Fe-4S] clusters (N2, N3, and N4). Center N1c appears unique to the E. coli enzyme in comparison with other NADH:ubiquinol oxidoreductases, but each of the other centers has counterparts in the bovine and other enzymes (345). Generally, the characteristics of the clusters in E. coli NDH-I follow the pattern found in all the related enzymes: (i) a very low-potential cluster (N1a, E_m,7 = –430 mV), which is not reduced by NADH but is in apparent equilibrium with the bound FMN (346); (ii) a group of iron-sulfur clusters with similar midpoint potentials that may serve to conduct electrons to the Q-binding site (N1b, N1c, N3, and N4; E_m,7 = –290, –290, –270, and –270 mV, respectively); and (iii) a high-potential tetranuclear cluster which is probably the direct reductant of ubiquinone (N2, E_m,7 = –230 mV). The only unambiguous assignment of any of these centers to a particular subunit is center N3 in NuoF, as discussed above. Figure 6 shows tentative, speculative assignments which are reasonable but not unique.

  Mechanism and coupling to proton pumping. There are no data that directly address the kinetic mechanism of NDH-I or the proton-pumping function of this enzyme. The same mechanistic schemes that have been postulated and discussed in relation to the mitochondrial complex I enzymes, however, would seem to apply equally to the E. coli enzyme. These have been reviewed (414), but generally the experimental data are insufficient to define a unique, compelling mechanistic model.

NADH Dehydrogenase II (NDH-II; Ndh Enzyme).

NDH-II appears to have been designed to provide biochemists some relief from the complexities of NDH-I. NDH-II contains only a single subunit (47 kDa), which has no transmembrane elements (194). It utilizes noncovalently bound FAD as a cofactor and readily oxidizes NADH and reduces Q, but it is not coupled to proton translocation across the membrane (Fig. 3). This enzyme contains no metal prosthetic groups, so it must catalyze the direct electron transfer from the reduced flavin to Q (71). The mode by which it attaches to the inner surface of the cytoplasmic membrane to reduce Q-8 within the bilayer is not known. Other flavoprotein dehydrogenases, such as pyruvate oxidase and lactate dehydrogenase, are similar in their lack of a transmembrane anchor and any metal prosthetic groups.

NDH-II is not affected by any of the classical inhibitors of complex I (e.g., rotenone and piericidin A) or by capsaicin or DCCD (436, 437). Also, NDH-II in E. coli does not oxidize deamino-NADH, although NDH-II isolated from other species in some cases does function with this artificial substrate. It appears that NDH-II homologs or analogs are quite widespread in bacterial species (437), providing an alternate, uncoupled pathway for the reduction of NADH. A version of NDH-II has also been found in the mitochondrion of Saccharomyces cerevisiae (244).

In E. coli, the level of NDH-II in the membrane is regulated by the Fnr protein, which represses ndh gene transcription under anaerobic or low aeration conditions (149) (see Regulation of Enzyme Synthesis, below). Hence, it would appear that under such conditions the cell relies on NDH-I for most of its NADH oxidation activity. This is consistent with the phenotype of nuo strains (5, 295, 446), which show defects particularly when the cells are in the late exponential or stationary phase of growth, i.e., low dissolved-oxygen supply.

 

Proline

Proline Oxidase (Proline:Quinone Oxidoreductase; PutA Enzyme).

Proline is required for protein biosynthesis but can also be used by E. coli and S. typhimurium as the sole carbon and nitrogen source. Proline oxidase (233) is a dimer containing a single subunit (144 kDa), and it contains noncovalently bound FAD as a cofactor (59). PutA catalyzes the two-electron oxidation of proline to pyrroline-5-carboxylic acid and reduces the quinone pool in the membrane. The pyrroline-5-carboxylic acid is also oxidized by PutA, in an NAD⁺-dependent reaction.

The proline utilization (put) operon, located at 22 min on both the E. coli and S. typhimurium genetic maps, encodes a specific permease (PutP) as well as the dehydrogenase (PutA). In addition to being a membrane-associated bifunctional enzyme (233), PutA autogenously regulates expression of the put operon (111, 241). When the proline concentration is high and there are ample binding sites on the cytoplasmic membrane, PutA is enzymatically active as a dehydrogenase (110, 426). In the absence of excess proline, PutA accumulates in the cytoplasm and binds to the put operon control region DNA, thereby repressing put operon expression (110, 111, 241). The binding of proline to PutA results in increased hydrophobicity of the protein and an increased affinity for membrane-binding sites (110). The shift of PutA from the DNA site to the membrane requires both proline and an electron acceptor, such as oxygen (110). Whether the membrane binding is simply to the lipid surface or is protein mediated is not clear, although there is evidence that such binding can be saturated. There are no compelling data to support an intrinsic membrane protein providing a docking site for PutA or the other single-subunit nontransmembranous dehydrogenases.

The elaborate put autogenous regulatory system allows the cells to use proline as a carbon source when it is abundant but to reserve it for its essential role as a component required for protein biosynthesis as a priority when it is present at lower levels (120).

 

Pyruvate

Pyruvate Oxidase (Pyruvate:Quinone Oxidoreductase; PoxB Enzyme).

Pyruvate oxidase catalyzes the oxidative decarboxylation of pyruvate to acetate and carbon dioxide (36, 37) with the concomitant reduction of Q. Under laboratory growth conditions, this enzyme is not essential (74). The conversion of pyruvate to acetate is wasteful of energy compared with its conversion to acetyl coenzyme A. Under aerobic conditions, pyruvate dehydrogenase (AceEF-Lpd enzyme) converts pyruvate to acetyl coenzyme A, and under strictly anaerobic conditions, this is done by pyruvate-formate lyase (Pfl enzyme). Neither of these is a respiratory enzyme. However, it has been suggested that under microaerophilic conditions, pyruvate oxidase bridges the gap in providing a source of energy from pyruvate catabolism (74). Consistent with this suggestion, it has been demonstrated that poxB gene expression is maximal in early stationary phase and is strictly dependent on the rpoS (katF) gene (74). The poxB gene is located at 19 min on the E. coli genetic map (147).

Pyruvate oxidase has been of interest primarily as a model for studying protein-lipid interactions. The enzyme is a water-soluble tetramer of identical subunits (62 kDa) and does not require detergents or harsh conditions for release from the membrane. Each subunit contains noncovalently bound FAD, and the enzyme also requires a second cofactor, thiamine PP_i, for activity. In the presence of the substrate and cofactor, the enzyme changes its conformation to expose a hydrophobic lipid-binding site (312). Under such conditions, pyruvate oxidase binds to E. coli membrane vesicles and to phospholipid vesicles (328). To reduce Q-8 which is dissolved in the lipid bilayer, pyruvate oxidase must be bound in this manner to the membrane (146, 210).

In addition to facilitating access to the lipophilic substrate, lipid binding activates the enzyme and increases the turnover with artificial electron acceptors by 20- to 30-fold (312). Proteolysis and site-directed mutagenesis studies have pointed to the membrane-binding region being at the extreme carboxyl-terminal portion of the polypeptide (145, 146). It has been suggested that a putative amphipathic helix at the carboxyl terminus interacts at the membrane surface and facilitates partial penetration of the bilayer by the protein (145). The cooperative interaction of the membrane-binding domains of at least two of the four subunits appears necessary for this to occur (403). The recently determined structure of prostaglandin synthase provides a model of how a protein might interact at the surface of the membrane and utilize lipophilic substrates (283). In the case of pyruvate oxidase, however, the membrane binding is conditional and depends on the presence of the substrate. In the absence of pyruvate and thiamine PP_i, pyruvate oxidase appears to be a water-soluble cytoplasmic enzyme. As is the case for proline oxidase (see above), partitioning of the enzyme between the cytosol and the membrane is dependent on the substrate concentration.

 

Succinate

Succinate Dehydrogenase (Succinate:Ubiquinone Oxidoreductase; Complex II; SdhABCD Enzyme).

Succinate dehydrogenase, also known as succinate:ubiquinone oxidoreductase or complex II (by analogy with its mitochondrial homolog), catalyzes the oxidation of succinate to fumarate and reduces Q in the membrane; it is a critical component of the tricarboxylic acid cycle. The reaction proceeds without generating any charge separation across the membrane, and it is not a proton pump. The two protons released in the cytoplasm upon the two-electron oxidation of succinate are balanced by the two protons required for the reduction of Q (Fig. 3). This enzyme is thus not a coupling site in the respiratory chain. Considering that the function of this enzyme is relatively simple, it is difficult to rationalize why E. coli succinate dehydrogenase is so complicated (Table 5), consisting of four subunits with five different prosthetic groups (FAD, three iron-sulfur clusters, and a heme B). Although much simpler than NDH-I, which is a primary proton pump, succinate dehydrogenase is much more complex than NDH-II or other simple dehydrogenases that have similar functions.

Succinate dehydrogenase has been purified to homogeneity (209), and the sdhCDAB operon encoding all of the subunits has been cloned and sequenced (100, 425). The enzyme is part of a large family of related enzymes that includes succinate dehydrogenases from other prokaryotes, mitochondrial complex II, and the fumarate reductases (quinone:fumarate oxidoreductases). The properties of this family of enzymes have been reviewed in detail (167). Although relatively few biochemical and biophysical studies have been performed with E. coli succinate dehydrogenase, considerable detail is available from studies on the related enzymes.

All of the enzymes in this family (167) have in common two large subunits that are hydrophilic and contain a covalently bound FAD (SdhA subunit) as well as three iron-sulfur clusters (S1, S2, and S3; SdhB subunit). These two subunits are attached to the membrane by one or two hydrophobic subunits, which act as the membrane anchor. It is likely that the SdhB subunit provides the contact to the hydrophobic subunits and is essential for the attachment of SdhA to the membrane (K. Kita, personal communication). The hydrophobic subunits can be associated with zero, one, or two heme B prosthetic groups in different enzymes in this family. These hydrophobic subunits are essential for the quinone oxidoreductase activity, whereas the two hydrophilic subunits alone are sufficient for the succinate dehydrogenase/fumarate reductase activities with artificial electron acceptors and donors. For this reason, the name succinate dehydrogenase is often used to refer simply to the subcomplex of the two hydrophilic subunits. The intact E. coli succinate:ubiquinone oxidoreductase consists of four different subunits, which are assumed to be present in the complex in 1:1:1:1 stoichiometry (Table 5). A schematic showing the prosthetic groups and subunits is shown in Fig. 6.

All four subunits are encoded by the sdhCDAB operon located at 17 min on the E. coli genetic map, within a cluster of genes encoding four different tricarboxylic acid cycle enzymes. Enzyme synthesis is maximal during aerobic growth, and analysis of sdh-lacZ fusion expression shows that anaerobic repression is mediated by the ArcA protein (187) (see Regulation of Enzyme Synthesis, below).

The SdhA subunit contains the covalently bound FAD which, by analogy with other members of the family (167), is probably attached to residue His-45 by a His-N(3)-8α-riboflavin linkage. The electrochemical properties of the FAD group of E. coli succinate dehydrogenase have not been explored in any detail, but by analogy to studies on the mammalian enzyme (167), it appears that the FAD is reduced in a concerted two-electron step by succinate, and the FADH₂ can then transfer its electrons to the iron-sulfur clusters in two sequential one-electron steps. The FAD of the mammalian succinate dehydrogenase has an E_m,7 of –79 mV for the two-electron process, which is more than 100 mV more negative than that of the succinate/fumarate couple (E_m,7 = 30 mV), i.e., apparently unfavorable for rapid electron transfer from succinate to FAD. However, it is possible that substrate binding alters the electochemical properties of the bound flavin.

SdhB contains all three iron-sulfur clusters (Table 5). The groups of cysteines ligating these centers and a proposed ligation scheme (identical to that proposed for FrdB) are discussed below (see Iron-Sulfur Subunits). Spectroscopic studies (167) on other enzymes in the family indicate that center S1 is near the bound FAD (12 to 18 ) (1.2 to 1.8 nm), which is on the SdhA subunit, and is also estimated to be only 9 to 12 (0.9 to 1.2 nm) from center S2 and 10 to 20 (1.0 to 2.0 nm) from center S3. Center S3 appears to be associated with the quinone reductase site, which is likely, at least in part, to be within the hydrophobic subunits.

The SdhC and SdhD subunits, which make up cytochrome b ₅₅₆, are described below (see Cytochromes and Membrane Anchors). Little is known from direct experimentation about the quinone active site of E. coli succinate dehydrogenase, but work on the mammalian complex II (167) shows that it is associated with a highly stabilized ubisemiquinone pair, presumably bound to the hydrophobic anchor subunits. Spectroscopic studies indicate that the bound semiquinones interact with center S3 and/or the cytochrome b component.

 

ELECTRON ACCEPTORS

 

Fumarate

E. coli and S. typhimurium can both use fumarate as a terminal electron acceptor for anaerobic respiration. Fumarate reductase is quite similar in many respects to succinate dehydrogenase (see above), except that it preferentially catalyzes the reverse reaction, reducing fumarate to succinate. This step is part of the reverse tricarboxylic acid cycle that operates in nonaerated cells, as described in chapter 16 of this volume. Fumarate reductase is the subject of a recent review (391).

Fumarate Reductase (Menaquinol:Fumarate Oxidoreductase; FrdABCD Enzyme).

Fumarate reduction with a variety of electron donors results in the generation of proton motive force, as revealed by experiments measuring amino acid uptake, atebrin fluorescence quenching, and proton extrusion both in cells and in membrane vesicles. These experiments have been thoroughly and critically reviewed by Ingledew and Poole (176). However, none of these studies have directly addressed the question whether menaquinol:fumarate oxidoreductase results in net proton translocation across the cytoplasmic membrane. Measurements of proton extrusion with hydrogen as the electron donor yielded estimated H⁺/e^– ratios of 1 and 2 with fumarate and nitrate as the electron acceptor, respectively (200). Hydrogen:quinone oxidoreductase and quinol:nitrate oxidoreductase are each thought to catalyze net proton translocation with an H⁺/e^– ratio of 1, so the lower overall H⁺/e^– ratio for the coupled hydrogen-to-fumarate reaction implies that the menaquinol:fumarate oxidoreductase reaction does not result in proton translocation. Likewise, oxidation of glycerol 3-phosphate coupled to fumarate reduction yields an estimated H⁺/e^– ratio of 1 (256). Furthermore, the well-studied fumarate reductase from Wolinella succinogenes apparently does not act to form a transmembrane proton gradient (140). These observations, along with analogy to succinate dehydrogenase, suggest that fumarate reductase does not itself catalyze net transmembrane proton translocation.

The enzyme contains four subunits: a large FAD-containing catalytic subunit, a smaller iron-sulfur subunit, and two rather small intrinsic membrane anchor subunits (Table 6) (91, 94, 154, 226, 410). The FrdAB subunits are homologous to the SdhAB subunits. The membrane anchor polypeptides of the respective enzymes share no obvious sequence similarities, but apparently form analogous structures (see Succinate, above, and Cytochromes and Membrane Anchor Subunits, below).

The FrdA and FrdB subunits are extrinsic membrane proteins that face the cytoplasm (225). The FrdC and FrdD subunits are intrinsic membrane proteins; topological models based on theoretical predictions illustrate each subunit as having three transmembrane-spanning regions (92, 418). All four subunits are required for proper membrane localization and for fumarate respiration (73, 96).

Formate, glycerol 3-phosphate, hydrogen, and NADH are all effective electron donors for fumarate reduction (54, 162, 200, 255, 441, 442), whereas lactate is not (155, 162). MK and DMK serve as the intermediate electron carriers (see Quinones, above). FrdABCD enzyme apparently does not contain heme (418); earlier studies indicating the participation of a cytochrome b involved in fumarate reduction have not been pursued (53, 343). By contrast, the W. succinogenes FrdC subunit is a diheme cytochrome b (213).

The FrdA subunit contains covalently bound FAD (410). Mutational alteration of residue His-44 in the FrdA subunit results in an enzyme that contains noncovalently bound FAD. This mutant enzyme retains substantial fumarate reductase activity but has lost its succinate dehydrogenase activity (46). Other residues important for catalysis have also been identified by site-specific mutational analysis (329). Studies of the FrdB subunit are described below (see Iron-Sulfur Subunits).

The FrdC and FrdD anchor subunits are both essential for assembly of FrdABCD enzyme into the cytoplasmic membrane (73, 96) and for transfer of electrons from menaquinol to fumarate (73, 409). Association with the anchor subunits also modifies the properties of the FrdAB catalytic dimer (226). The FrdCD complex is also involved in mediating electron flow from menaquinone to the FrdAB complex (see Cytochromes and Membrane Anchor Subunits, below).

Fumarate reductase is encoded by the frdABCD operon, located at 93 min on the E. coli genetic map. Enzyme synthesis requires anaerobic growth conditions, is not further induced by fumarate, and is repressed by nitrate (197, 198, 219, 351). Analysis of frdA-lacZ gene fusion expression shows that anaerobic regulation and nitrate regulation are mediated by the Fnr and NarL proteins, respectively (186, 198) (see Regulation of Enzyme Synthesis, below).

  Fumarate transport. Aerated cells take up dicarboxylic acids via the periplasmic binding protein-dependent DctAB-Cbt system. However, dct and cbt mutants are unimpaired for anaerobic fumarate uptake, and the kinetic parameters for uptake are also quite different for aerated and anaerobic cultures (121). Further analysis suggests that fumarate-respiring cells express a fumarate/succinate antiporter (Dcu activity), allowing for the electroneutral and energy-efficient uptake of fumarate coupled to extrusion of succinate, a compound that accumulates in anaerobic culture media (122).

Recent studies indicate that two genes, termed dcuA and dcuB (dicarboxylate uptake), are involved in anaerobic fumarate uptake (344). These genes are linked to the aspA (aspartase) and fumB (fumarase B) genes, encoding enzymes involved in the reverse tricarboxylic acid cycle as described elsewhere in this volume (see chapter 16). Expression of Dcu activity requires anaerobiosis and is inhibited by nitrate; these effects are mediated by the Fnr and NarL proteins, respectively (121, 122) (see Regulation of Enzyme Synthesis, below).

 

Nitrate

Nitrate (NO₃ ^–) respiration is widely distributed among bacterial species, including the enterobacteria. Nitrate appears to be among the most energetically favorable of the anaerobic electron acceptors, because it inhibits the synthesis of enzymes involved in the respiration of most other acceptors. This special position in the hierarchy of electron acceptors has focused much attention upon nitrate respiration over the years. The product of nitrate reduction, nitrite (NO₂ ^–), is also an efficient electron acceptor; nitrite respiration is considered below (see Nitrite).

Microbiology textbooks generally recognize two forms of nitrate reduction, nitrate assimilation and nitrate "dissimilation." The latter term is a misnomer and was coined in earlier times when nitrate reduction was thought to simply serve for the disposal of excess reducing equivalents. The more accurate term, nitrate respiration, recognizes the fact that anaerobic nitrate reduction is coupled to oxidative phosphorylation (reviewed in reference 357).

E. coli K-12 has multiple respiratory nitrate reductases. The major E. coli nitrate reductase (NarGHI enzyme) is a cytoplasmic membrane-bound enzyme whose synthesis is induced by nitrate during anaerobic growth. A homologous enzyme (NarZYV enzyme) is constitutively synthesized at low levels. S. typhimurium LT2 also contains at least two respiratory nitrate reductases (22), neither of which has been extensively characterized. Recently, structural genes for a periplasmic nitrate reductase (NapABC [YojC-YejYX] enzyme) have been identified as part of the E. coli genome project (P. Richterich, N. Lakey, G.Gryan, J. Jaehn, L. Mintz, K. Robison, and G. M. Church, unpublished DNA sequence, GenBank accession number U00008). The physiological roles of the NarZYV and NapABC enzymes are unclear.

Nitrate Assimilation.

Assimilatory nitrate reduction is the aerobic conversion of nitrate (via nitrite) to ammonium, allowing the use of nitrate (and nitrite) as a sole nitrogen source. In all species studied to date, the structural and regulatory genes for nitrate respiration and assimilation are distinct. Neither E. coli K-12 nor S. typhimurium LT2 can assimilate nitrate during aerobic growth. However, nitrate assimilation has been studied in the related enterobacterium Klebsiella pneumoniae M5al. Enzymes and transport components for nitrate assimilation are arranged in the nasFEDCBA operon, whose transcription is subject to general nitrogen control mediated by the Ntr system and to nitrate and nitrite induction mediated by the NasR protein (144, 232).

The combined operation of respiratory nitrate reductase and NADH-nitrite reductase (see Nitrite, below) allows enterobacteria to grow anaerobically with nitrate as the sole nitrogen source. Indeed, such cultures actually excrete ammonium into the medium (89, 90). However, this ammonium production is simply a by-product of nitrate respiration and occurs only during anaerobic growth when these enzymes are synthesized. Synthesis of the anaerobic nitrate and nitrite reductases is not controlled by the availablilty of fixed nitrogen.

Chlorate and Azide.

Chlorate (ClO₃ ^–), a structural analog of nitrate, is a substrate for many molybdoenzymes (including some nitrate reductases), and selection for chlorate resistance yields mutants that are unable to synthesize the molybdenum cofactor (see Molybdenum Cofactor Subunits, below). Azide (N₃ ^–) is also an inhibitor of some nitrate reductases. Early studies identified three enzymes that are active on nitrate and/or chlorate (reviewed in reference 357). Enzyme A (nitrate and chlorate reductase activities; inhibited by azide) is the membrane-bound respiratory nitrate reductase. Enzyme B (nitrate reductase activity; inhibited by chlorate) is presumed to correspond to assimilatory nitrate reductase. Enzyme C (chlorate but not nitrate reductase activity; insensitive to azide) probably represents several activities, including the DMSO and tetrathionate reductases (see Molybdenum Cofactor Subunits, below). Finally, the more recently described periplasmic nitrate reductase (NapABC enzyme) of Thiosphaera pantotropha reduces only nitrate and is inhibited by neither chlorate nor azide (34).

Nitrate Reductase-A (Quinol:Nitrate Oxidoreductase-A; NarGHI Enzyme).

E. coli membrane-bound nitrate reductase has been comprehensively reviewed (51, 357). The enzyme contains three subunits: a large molybdoprotein containing the catalytic site (NarG), and smaller iron-sulfur (NarH) and cytochrome b (NarI) subunits (Table 7) (76, 124, 237, 238, 239, 349). Properties of these subunits are summarized below (see Common Features of Respiratory Enzymes). A fourth polypeptide, NarJ, is also encoded by the narGHJI operon but is not found in the purified enzyme. The NarJ protein apparently plays a role in enzyme assembly (see below).

Formate, glycerol 3-phosphate, hydrogen, NADH, d- and l-lactate, succinate, and malate are all effective electron donors for nitrate reduction (183, 423; reviewed in reference 357). Both Q and MK serve as intermediate electron carriers for respiratory nitrate reduction (see Quinones, above).

Nitrate reduction with a variety of electron donors results in the generation of proton motive force, as revealed by measurements of both proton extrusion and membrane potential in spheroplasts and membrane vesicles (55, 139, 202, 357). Estimated H⁺/e^– ratios of 1 with artificial electron donors and 2 with physiological donors such as formate have been observed (139, 202). This and all other available evidence indicate that nitrate reductase catalyzes net proton translocation from the cytoplasm to the periplasm by a scalar mechanism (see Mechanisms of Proton Translocation, above) (201, 202, 262; reviewed in reference 357). Chemical labeling studies, coupled with comparisons of membrane-permeant and impermeant electron donors, show that the NarG and NarH subunits, and thus the active site for nitrate reduction, face the cytoplasm (201; reviewed in reference 357). The predicted amino acid sequences of the NarG, NarH, and NarI subunits are also consistent with this topology (42, 348).

Thus, in this model for nitrate reductase action, quinol oxidation occurs on the NarI (cytochrome b) subunit, with protons released to the periplasm, whereas nitrate reduction occurs on the NarG subunit, with protons consumed from the cytoplasm. This mechanism for net proton translocation parallels that for cytochrome bd (see Oxygen, below) and for the respiratory nitrate reductase of Paracoccus denitrificans (19).

Nitrate reductase is encoded by the narGHJI operon, located at 27 min on the E. coli genetic map (42, 52, 348, 361). Enzyme synthesis is induced by nitrate during anaerobic growth (337). Analysis of narG-lacZ operon fusion expression shows that anaerobic and nitrate regulation are mediated by the Fnr and NarL proteins, respectively (82, 356) (see Regulation of Enzyme Synthesis, below).

  Biogenesis of nitrate reductase-A. A soluble NarGH complex accumulates in strains lacking the NarI (cytochrome b) subunit (45, 238, 361). This soluble complex can also be purified from heat-treated cytoplasmic membrane fractions, in which limited proteolysis releases the NarGH complex from the membrane (237, 239). Evidence suggests that interaction with the membrane-intrinsic NarI subunit leads to membrane association of the NarGH complex (45, 75, 238, 349). Thus, the NarI subunit appears to serve also as a membrane anchor, analogous to the FrdCD and DmsC subunits of fumarate reductase and DMSO reductase, respectively. The homologous NarV subunit from nitrate reductase-Z (see below) also serves as a membrane anchor (45). More specific details of this association—for example, does the NarI subunit interact with the NarG subunit, the NarH subunit, or both?—have not been investigated.

The NarJ and homologous NarW (see below) proteins also play essential roles in the formation of active, membrane-bound enzyme, although the NarJ protein has never been detected as a structural component in purified nitrate reductase. Mutants lacking the NarJ protein accumulate substantial NarGH complex in the cytoplasmic membrane, but such membranes contain very low nitrate reductase activity (as measured with artificial electron donors). Conversely, mutants lacking the NarI protein accumulate high-activity NarGH enzyme largely in the cytoplasm (45, 115). NarGH complex purified from a strain lacking both the NarJ and NarI proteins contains molybdenum cofactor and iron-sulfur clusters, so these prosthetic groups appear to be incorporated in the absence of the NarJ protein (45). However, this complex has low activity and is unstable. Thus, the NarJ protein is essential for the formation or stability of active nitrate reductase, but its role(s) in this process is unknown. The P. denitrificans nitrate reductase operon also contains a narJ homolog (33), so NarJ-dependent assembly appears to be a common feature of membrane-bound nitrate reductases.

  Nitrate uptake and nitrite extrusion. Studies of nitrate transport are technically formidable, because the radioisotope ¹³N has a very short half-life (about 10 min). Studies with nitrate-responsive electrodes have shown that nitrate uptake is blocked by oxygen, a phenomenon explained at least in part by oxygen inhibition of nitrate reduction (109, 275).

Immediately upstream of the narGHJI operon lies the narK gene, encoding an intrinsic cytoplasmic membrane protein (274, 361). An appealing hypothesis is that the NarK protein serves as an electroneutral nitrate/nitrite antiporter (107, 274). Thus, rapid nitrate uptake would be coupled to its reduction, and the resultant (toxic) nitrite would be efficiently removed from the cytoplasm. However, detailed studies with [¹³N]nitrate lead instead to the conclusion that the NarK protein is a ΔΨ-driven system for nitrite extrusion (308). In this view, the NarK protein serves solely to export nitrite, preventing the intracellular buildup of toxic concentrations. It is suggested that nitrate uptake may be catalyzed by a Δ p-driven symport (308), although such a nitrate uptake system has not yet been identified genetically or biochemically.

Nitrate Reductase-Z (Quinol:Nitrate Oxidoreductase-Z; NarZYV Enzyme).

One explanation for the residual nitrate reductase activity detected in narG insertion mutants (178, 361) came with the discovery of a distinct yet homologous enzyme, termed nitrate reductase-Z (49, 178). Biochemical and DNA sequence analyses show that the NarZYV enzyme is only subtly different from the NarGHI enzyme (43, 180). Indeed, active hybrid enzymes are assembled in strains carrying appropriate combinations of genes from the narGHJI and narZYWV operons, although at least one of the hybrid enzymes (NarGYV enzyme) is less stable than either of the normal enzymes (44).

A narK homolog, the narU gene, is located immediately upstream of the narZ gene and, indeed, may be the first gene in a narUZYWV operon (unpublished data cited in reference 51). Analysis of narZ-lacZ fusion expression reveals unregulated, low-level expression that is indifferent to the presence of either oxygen or nitrate (unpublished data cited in reference 51). Thus, the physiological role of the NarZYV enzyme is unknown. One idea is that preexisting nitrate reductase Z could allow for at least a low level of nitrate respiration during a sudden transition from aerobiosis to anaerobiosis, thereby providing energy for biosynthesis of nitrate reductase A and other components needed for anaerobic respiration (49, 51).

Periplasmic Nitrate Reductase (NapABC Enzyme).

Many bacteria contain a respiratory nitrate reductase located in the periplasm. This enzyme has been thoroughly characterized in Thiosphaera pantotropha (also designated as P. denitrificans), in which it is responsible for the observed aerobic denitrification activity (30). The enzyme as purified consists of a large molybdoprotein (NapA) and a small cytochrome c (NapB) (34). The large subunit also contains an iron-sulfur cluster (57). Recent DNA sequence analysis suggests that a second cytochrome c (NapC) and two additional polypeptides (NapD and NapE) are also involved in enzyme activity (B. C. Berks, S. J. Ferguson, and D. J. Richardson, personal communication). The Alcaligenes eutrophus napA and napB genes have also been characterized (338).

Recent large-scale sequencing of the E. coli genome has identified genes that are homologous to the nap genes of A. eutrophus and T. pantotropha (Richterich et al., unpublished). These genes correspond to the aeg-46.5 locus, which was identified as an anaerobically expressed gene that maps at 46.5 min on the genetic map (85). The nap-like genes are arranged in an apparent operon with genes whose products are probably involved in cytochrome c biogenesis (see Membrane Anchors and Cytochromes, below). As yet, there is no direct biochemical evidence for a periplasmic nitrate reductase in E. coli, although two cytochromes c are likely to be the products of the yejY (napB) and yejX (napC) genes (179) (see Membrane Anchors and Cytochromes, below).

Expression of the aeg-46.5 operon is induced by nitrite (and more weakly by nitrate) during anaerobic growth. Analysis of aeg-46.5-lacZ operon fusion expression shows that anaerobic regulation and nitrite regulation are mediated by the Fnr and NarP proteins, respectively (85, 102a, 299) (see Regulation of Enzyme Synthesis, below). This regulatory pattern is in contrast to those of the NapABC enzymes in T. pantotropha (constitutive aerobic synthesis [30, 303]) and A. eutrophus (nitrate-independent synthesis during stationary phase [405]). The physiological role(s) of periplasmic nitrate reductases is unclear, although one idea is that these enzymes serve to dispose of excess reducing equivalents (30, 303, 338).

 

Nitrite

Bacterial nitrite reduction follows two pathways. Denitrification is accomblished by a series of anaerobic respiratory enzymes that convert nitrite (NO₂ ^–) through nitric oxide (NO) and nitrous oxide (N₂O) to dinitrogen (N₂). Enterobacteria do not perform denitrification, although weak nitrous oxide production by E. coli is attributable to nitrate reductase (347). Nitrite can also be reduced directly to ammonium (NH₄ ⁺) via a six-electron transfer. Enterobacteria have three distinct enzymes that catalyze this reaction: the major anaerobic nitrite reductase (NirBD enzyme), which is present in most enterobacteria; assimilatory nitrite reductase (NasB enzyme), which is not present in E. coli K-12 or S. typhimurium LT2 but is found in other enterobacterial species (e.g., Klebsiella spp. [232]); and respiratory nitrite reductase (NrfAB enzyme), which is found in E. coli K-12 but not in S. typhimurium LT2 or in Klebsiella spp. (1).

NADH-nitrite reductase in E. coli (NirBD enzyme) contains FAD, siroheme, and two iron-sulfur clusters. This enzyme is encoded by the nirBD operon at 72 min on the E. coli genetic map (282). Enzyme synthesis is induced by nitrite or nitrate during anaerobic growth (278). Analysis of nirB-lacZ operon fusion expression shows that anaerobic regulation and nitrite/nitrate regulation are mediated by the Fnr and NarL/NarP proteins, respectively (384) (see Regulation of Enzyme Synthesis, below). Although NirBD enzyme is synthesized only during anaerobiosis, it serves no direct respiratory function (293). Rather, its physiological role seems to be twofold: it serves both to regenerate NAD⁺ and to detoxify nitrite (309) that accumulates from nitrate respiration (278).

Formate-Dependent Respiratory Nitrite Reductase (Cytochrome c₅₅₂ ; NrfAB enzyme).

Respiratory nitrite reductase is a periplasmic tetraheme cytochrome c ₅₅₂ (88, 102). Activity is relatively low in comparison with NADH-nitrite reductase and in fact is absent in many enterobacteria (including S. typhimurium LT2 [1]). Formate is the best-studied electron donor, although d-lactate, pyruvate, and ethanol can also serve (2, 294). Each of the three formate dehydrogenases (formate dehydrogenases F, N, and O) can participate in donating electrons from formate (103). Reduced quinone appears to be the immediate electron donor (2, 174).

Nitrite reduction results in the generation of proton motive force (264, 293), although the mechanism for net proton translocation is unclear. DNA sequence analysis of the nrfA operon (nitrite reduction by formate) has revealed genes for nitrite reductase (NrfA, 50 kDa), a second periplasmic cytochrome c (NrfB, 18 kDa), a cytoplasmic iron-sulfur protein (NrfC, 24 kDa), and a protein with homology to NADH:quinone oxidoreductases (NrfD, 37 kDa) (102, 174). These observations allow for speculative schemes for electron transfer but do not yet provide a model for proton translocation.

The nrfABCDEFG operon is located at 93 min on the E. coli genetic map (174). Enzyme synthesis is induced by nitrite during anaerobic growth and is repressed by nitrate (278). Analysis of nrfA-lacZ operon fusion expression shows that anaerobic regulation and nitrite/nitrate regulation are mediated by the Fnr and NarP/NarL proteins, respectively (384) (see Regulation of Enzyme Synthesis, below).

 

Oxygen

E. coli has two respiratory oxidases, cytochrome bd and cytochrome bo ₃. As already discussed (see Modularity of Respiratory Components, and Mechanisms of Proton Translocation, above), both of these enzymes are quinol oxidases and both are coupling sites and contribute to the proton motive force. Each enzyme forms water, not peroxide, as the product of oxygen reduction (260). Cytochrome bo ₃ is part of a large superfamily of proton- pumping heme-copper respiratory oxidases (67, 137) and has homologs in virtually all aerobic bacteria as well as in eukaryotes. Cytochrome bd is the only well-characterized respiratory oxidase that is not a member of the heme-copper oxidase superfamily and is found in a relatively small number of bacterial species including, but not limited to, the enterics (215).

As summarized below, these two enzymes are not structurally related and have distinct physiological roles. Cytochromes bo ₃ and bd are maximally synthesized during growth under high aeration and limited aeration, respectively. Studies with cyo-lacZ and cyd-lacZ fusions show that this control is mediated by the ArcA and Fnr proteins (see Regulation of Enzyme Synthesis, below).

Cytochrome bd.

Cytochrome bd is a heterodimer (Table 8) (258) encoded by the cydAB operon located at 16 min on the E.coli genetic map (152). Each subunit contains multiple putative membrane-spanning helices (271). The enzyme contains three heme prosthetic groups: heme b ₅₅₈ , heme b ₅₉₅ , and heme d (236). Heme b ₅₅₈ is contained within CydA (subunit I [148]), and site-directed mutagenesis has identified the likely ligands to the heme iron as residues His-186 (125) and Met-393 (T. Kaysser and R. B. Gennis, unpublished observations). Both of these ligands are predicted to be at the periplasmic side of the membrane near the ends of putative transmembrane helices (Fig. 7). Heme b ₅₅₈ is probably involved directly in quinol oxidation(166). Heme d (Fig. 5) is the site where oxygen (and CO) binds to the enzyme (169, 203). Spectroscopic studies indicate that heme d has only one strong amino acid ligand and that the iron is not directly bound to a nitrogen (196). The role of heme b ₅₉₅ is not known, but the spectroscopic properties indicate it is high spin, i.e., ligated to only one strong ligand, and is located near heme d (169, 236, 249). Hence, heme b ₅₉₅ and heme d may form a heme-heme bimetallic center involved jointly in the reduction of oxygen to water (169). Their heme planes are not parallel, however (177). It has been clearly demonstrated that turnover of cytochrome bd results in the generation of a transmembrane voltage (257) and the translocation of 1 H⁺/e^– (297).

Reconstitution experiments have shown that cytochrome bd efficiently reduces Q-8 in a lipid bilayer (210). The ability to oxidize quinol is specifically eliminated by mild proteolysis, which has been shown to cleave subunit I in an interhelical connection on the periplasmic side of the membrane called the Q-loop (117). Monoclonal antibodies that bind to an epitope in the Q-loop similarly inhibit quinol oxidase activity (116). These data suggest that the quinol-binding site directly involves elements of the Q-loop and is located at the periplasmic side of the membrane. This is also consistent with the scalar proton translocation model, where the substrate protons resulting from the oxidation of quinol are released to the periplasm (Fig. 3 and 7).

The reduction of oxygen to water is a four-electron process. By analogy with studies on cytochrome c oxidase (15, 164), one might expect a series of intermediate forms of the heme d-oxygen adduct as oxygen is reduced from O₂ to 2 H₂O. These would include an initial heme d Fe²⁺–O₂ adduct (oxy adduct) and an oxoferryl adduct, heme d Fe⁴⁺–O^2–. Both the oxy (203) and oxoferryl (204) adducts can be prepared in vitro with the purified cytochrome bd and are stable. Remarkably, both the oxy and oxoferryl adducts are present in the purified enzyme as it is isolated. Studies with model heme D complexes with oxygen have been reported (276, 277) and should facilitate future efforts to understand the kinetic mechanism of the oxidase.

  CydCD proteins. In addition to the cydAB operon, two other genes, cydC and cydD, are known to be essential for the assembly of the active form of cytochrome bd (23, 142, 291). The cydC and cydD genes appear to be equivalent to the surB (339) and htrD (105, 291) genes, respectively. These two genes appear to be an operon (cydDC, located at 19 min on the E. coli genetic map [291]). Mutations in these genes do not influence the transcription of cydAB or the assembly of the CydA or CydB polypeptides in the membrane (23, 142, 291). The insertion of heme d, however, is blocked in these mutants (23, 142), and the insertion of hemes b ₅₅₈ and b ₅₉₅ may similarly be inhibited (291). The heme D isomer, cis-heme D, which is the prosthetic group of catalase HP-II (234), however, is synthesized, so cydDC is unlikely to be required for heme D biosynthesis (23). The sequences of cydC and cydD suggest that they encode an ATP-dependent (ATP-binding cassette) transporter in the cytoplasmic membrane (291). By analogy with the putative heme transport systems required for the assembly of cytochromes c in Bradyrhizobium japonicum and Rhodobacter capsulatus (25, 375), it has been suggested that CydC and CydD are required for the transport of heme to the periplasm of E. coli (291). Consistent with this is the observation that cydDC mutants cannot assemble cytochromes c (289).

  Physiological roles of cytochrome bd. Mutants that cannot make active cytochrome bd have pleiotropic phenotypes (23, 105, 289, 291, 339, 399). Since cytochrome bd is relatively resistant against inhibition by either azide or cyanide, cydAB or cydDC mutants are sensitive to azide and cyanide, reflecting the sensitivity of cytochrome bo ₃ to these inhibitors. These mutants are also sensitive to hydrogen peroxide, to cysteine, and to the presence of the dye toluidine blue (105). Furthermore, these mutants are temperature sensitive, consistent with the observation that cydAB transcription is under heat shock regulation (399). The temperature sensitivity is suppressed by either a multicopy plasmid carrying the cyo operon, leading to overproduction of cytochrome bo ₃, or arcA or arcB mutations which derepress expression of the chromosomal cyoABCDE operon (399). Hence, some of these mutant characteristics can be suppressed by increasing the respiratory activity of the cells under conditions where cytochrome bd is normally present. It has also been shown that surB mutants (probably cydC) are defective in resuming growth after starvation (339). Although all this is far from being fully understood, it is clear that cytochrome bd has a distinct physiological function, as reflected by its complex regulation.

Cytochrome bd is present at low levels when E. coli is grown with high aeration, but it is maximally present under microaerophilic conditions (135). The apparent K_m for oxygen is remarkably low (20 nM [302]), and this enzyme appears to be well suited for functioning when the dissolved oxygen concentration is very low. Experimental data have suggested that one role of cytochrome bd in E. coli may be to scavenge oxygen and inhibit the degradation of oxygen-sensitive enzymes present under anaerobic or microaerophilic growth conditions (170). A similar protective role of cytochrome bd has been postulated for Azotobacter vinelandii, for which it has been shown that cytochrome bd is essential for maintaining a functional nitrogen-fixing nitrogenase (112, 263). Interestingly, the measured K_m for O₂ of the A.vinelandii cytochrome bd is 4.5 μM, and cytochrome bd in this organism is induced under conditions of high aeration (112). The oxygen-scrubbing functions of cytochrome bd in E. coli and in A. vinelandii appear to be similar, but they are tuned to operate at very different oxygen concentrations.

In addition to its induction under conditions of low aeration, cytochrome bd has been reported to be induced when E. coli is grown under alkaline or other unfavorable growth conditions (12, 47) and to possibly be a primary sodium pump (11). The conclusion that cytochrome bd is a sodium pump has not been universally accepted and awaits further experimental verification.

  CyxAB proteins. The genes cyxAB (appCB), located at 22 min on the E. coli genetic map, are homologous to cydAB (6, 104). The complex cyxAB appA operon (appA encodes acid phosphatase) is immediately downstream of the hyaABCDEF operon encoding hydrogenase-1 (6) (see Hydrogen, above). It is not clear whether the polypeptides encoded by the cyxAB genes constitute an oxidase that plays an important functional role. However, the cyxAB genes are expressed in response to oxygen deprivation, and the cyxAB⁺ cyo cyd double mutant is less sensitive to oxygen than is the triple mutant. Expression of the cyxAB genes in the cyo cyd mutant background allows E. coli to form microcolonies after extended growth on nonfermentable substrates, indicating some repiratory function, although clearly suboptimal under the conditions examined.

Cytochrome bo₃.

The cyoABCDE operon, located at 10 min on the E. coli genetic map, encodes all of the subunits of the cytochrome bo ₃ ubiquinol oxidase (Table 9). The DNA sequence of this operon revealed an unexpected strong relationship with the eukaryotic mitochondrial cytochrome c oxidases (80). This finding has since been elaborated to confirm the existence of a large superfamily of proton-pumping respiratory oxidases (67). (For more detailed reviews, see references 60, 72, 137, 172, and 388).

  Heme-copper oxidase superfamily. The enzymes in the heme-copper oxidase superfamily have in common a subunit with strong sequence similarity to the largest subunit (subunit I) of the mammalian oxidase. This subunit contains the heme-copper bimetallic center, common to all the oxidases in the superfamily, which is the site where oxygen binds and is reduced to water. The pairwise identity between the E. coli CyoB sequence and the equivalent subunit of the bovine oxidase (not shown) is about 40% (80), demonstrating strong structural conservation, despite the fact that the substrates of these two oxidases are different (ubiquinol versus cytochrome c) and the heme prosthetic groups are different (hemes B and O versus heme A) (60, 67, 172).

The more than 80 sequenced oxidases in the superfamily can be subdivided into five classes, three of which are cytochrome c oxidases and two of which are quinol oxidases (137). Their similarities reside primarily in subunit I (CyoB), the most highly conserved subunit (Table 9). This subunit not only contains the site where the oxygen chemistry is catalyzed but also must contain the major elements of the proton-conducting channel(s) through the protein and the machinery for coupling the oxygen chemistry to proton pumping. The heme prosthetic groups associated with subunit I in different enzymes in the superfamily display considerable variation, being combinations of hemes A, B, and O, but the functional significance of this variation is not known (137, 388).

The major functional variation between the oxidases appears to reside in subunit II. In most of the cytochrome c oxidases, subunit II is the binding site for cytochrome c (319), and this subunit also contains a two-copper redox center (Cu_A), which is the initial oxidant of reduced cytochrome c (207, 220, 394). The quinol oxidases, including cytochrome bo ₃, contain subunit II (CyoA) with the same general topology as in the cytochrome c oxidases (79), but the Cu_A redox center and the cytochrome c- binding site are not present (261). The quinol oxidases interact with quinol instead of cytochrome c, and work from the Anraku laboratory indicates the presence of to quinol-binding sites in cytochrome bo ₃ (321). Affinity labeling with a photoreactive analog of Q has indicated that CyoA is directly involved in the interaction with ubiquinol (415). Hence, one role of subunit II in most enzymes in the superfamily is to provide the substrate- binding site (cytochrome c or quinol).

A phylogenetic analysis suggests that the E. coli cytochrome bo ₃, along with the other quinol oxidases in proteobacteria, may be derived by a gene transfer mechanism originating from the cytochrome c oxidases of the bacilli (72). Interestingly, the sequence of the NorB subunit of nitric oxide reductase from Pseudomonas stutzeri is clearly related to subunit I of the heme-copper oxidases (320, 388), suggesting that all these oxidases may have evolved from the denitrification enzymes prior to the advent of high oxygen concentrations in the atmosphere from photosynthesis (72, 320).

  Subunits. Typically, prokaryotic members of the oxidase superfamily contain subunits that are homologous to the three mitochondrion-encoded subunits (subunits I, II, and III) of the mammalian oxidase. However, there are exceptions in which subunit II and/or III is lacking, suggesting that these subunits are not generically necessary for the catalytic function of these enzymes (138, 388). None of the 10 nucleus-encoded subunits of the mammalian oxidase has a prokaryotic homolog. The cyo operon encodes four subunits that have been identified in the purified enzyme (Table 9). CyoA, CyoB, and CyoC are homologs of subunits II, I, and III, respectively, of the mammalian oxidase (80). CyoD is unrelated to any of the subunits in the mammalian oxidase, and its function is not known. Several other bacterial oxidases, however, contain subunits homologous to CyoD (181). In some preparations of cytochrome bo₃, there is a fifth subunit (261, 382), but its origin and function (if any) are not known; the four-subunit preparation (261) is fully functional. The cyo operon encodes a fifth polypeptide, CyoE, which has been shown to be a farnesyl transferase required for the conversion of heme B to heme O (313, 314), as discussed above (see Hemes) (Fig. 5). Although it has been suggested that CyoE might be a loosely bound fifth subunit (261), there is no evidence to support this contention. In E. coli, heme O is present only in the cytochrome bo₃ complex, so the presence of the gene required for heme O biosynthesis in the cyo operon is easily rationalized.

A major focus of attention in the study of cytochrome bo ₃ has been CyoB, which has been subjected to considerable site-directed mutagenesis analysis to identify residues critical for binding the metal prosthetic groups and/or for function. The topological analysis of CyoB indicates the presence of 15 transmembrane spans, including 12 putative transmembrane helices that are common to all members of the superfamily (79). The sequence alignment of all members of the superfamily of subunit I (CyoB) shows six fully conserved His residues (67, 137, 388), which are all located in putative transmembrane helices and which are all implicated in metal ligation (172).

CyoA, implicated in ubiquinol binding, has two putative transmembrane helices and a large, hydrophilic periplasmic carboxyl-terminal domain (79). In the mammalian cytochrome c oxidases, it is the homologous domain in subunit II that contains the ligands to Cu_A and the cytochrome c-binding site (319). It is not known whether this domain, located at the periplasmic interface, contributes to the quinol-binding site(s). By site-directed mutagenesis, the putative ligands to Cu_A have been placed back in CyoA, and the resulting mutant protein binds Cu with spectroscopic properties similar, though not identical, to those of authentic Cu_A (207, 390). This experiment demonstrates the similarity of the tertiary structures of CyoA and the equivalent domain in the cytochrome c oxidases. The hydrophilic domain of CyoA has been successfully crystallized, and a high-resolution structure should be forthcoming (389).

CyoC and CyoD have five and four putative membrane- spanning helices, respectively (79). Although the functions of these subunits are not known, functional cytochrome bo ₃ is not assembled in the membrane if these genes are deleted from the operon (268).

  Metal redox centers. Cytochrome bo₃ has three metal prosthetic groups, which are all contained within CyoB. There is a six-coordinate heme b₅₆₂, ligated to residues His-106 and His-421 in CyoB (224, 259), which is responsible for virtually all of the absorption in the visible region of the spectrum and, hence, the red color of the oxidase. Because of its spectroscopic properties, this is referred to as the low-spin heme component of the oxidase (261). The presumed function of heme b₅₆₂ is the oxidation of ubiquinol. Electrons are thought to be transferred from heme b₅₆₂ to the bimetallic center, consisting of heme o₃ and a copper, called Cu_B. This is the site where oxygen binds and is reduced to water, and the heme-copper bimetallic center is almost certainly involved directly in the mechanism of proton pumping, as illustrated in Fig. 7. The heme iron and copper are within about 5 (0.5 nm) of each other (60, 172), and these two metals are electronically coupled (315) and form a functional unit. Heme o₃ is ligated to residue His-419 within CyoB (68, 387), and Cu_B is also ligated to residues within CyoB (probably His-333, His-334, and His-284 [65, 372, 387]).

The subscript 3 (e.g., heme o ₃) is used to indicate the oxygen- (and CO-) binding component of the oxidase, analogous to heme a ₃, the universally used designation of the oxygen-binding heme component of mammalian cytochrome c oxidase. Heme o ₃ has also been designated as simply heme (or cytochrome) o or o ', and the enzyme has also been referred to as cytochrome bo or bo ' (292). Heme o ₃ is also referred to as the high-spin heme component of the oxidase. The Cu_B designation is also made by analogy to the equivalent copper in the mammalian cytochrome c oxidase.

All members of the heme-copper oxidase superfamily contain within subunit I a low-spin heme component and a heme-copper bimetallic center. The mammalian enzyme contains two equivalents of heme A, with a low-spin heme a and the heme a ₃-Cu_B bimetallic center (60, 172). Hence, heme b ₅₆₂ and heme o ₃ in the E. coli oxidase are analogous to heme a and heme a ₃, respectively, in the mammalian oxidase. Considerable effort has been dedicated to examining the biophysical properties of the metal centers in cytochrome bo ₃, with the particular goal of demonstrating the extent to which these metals are similar to the well-characterized redox centers in the mammalian cytochrome c oxidase. The data overwhelmingly show that the properties of the metal centers in the E. coli quinol oxidase are very similar to those in the mammalian cytochrome c oxidase (63, 64, 77, 78, 404, 406, 407). It is a reasonable working hypothesis that the mechanisms by which these two enzymes catalyze the oxygen chemistry and pump protons across the membrane are very similar, although a contrary opinion has been voiced (267). The similarity between cytochrome bo ₃ and the mammalian oxidase has motivated much of the interest in the E. coli enzyme.

The combination of spectroscopic techniques and site-directed mutagenesis has led to the assigment of the specific amino acid ligands to each of the three metal centers in CyoB (heme b ₅₆₂, heme o ₃, and Cu_B [67, 137, 172]). This assignment has been supported by parallel studies on the aa ₃-type cytochrome c oxidase from Rhodobacter sphaeroides (68). The main point of interest is that the resulting model places all the metals near the periplasmic surface. Hence, even though the protons used to generate water at the bimetallic heme-copper center must originate from the cytoplasm, the active site is actually placed near the opposite side of the membrane. This necessitates a proton-conducting channel leading from the cytoplasm to the heme-copper center.

  Proton translocation. For each electron used to reduce oxygen to water, two protons are taken from the cytoplasm. One "chemical" proton is used to generate water at the heme-copper site, and the second "pumped" proton is destined for the periplasm (Fig. 3) (16, 420). This requires a minimum of one proton-conducting channel. Experiments with site-directed mutations in polar residues in transmembrane helix VIII suggest that these residues might facilitate the movement of protons required for the formation of water (373). Helix VIII has conserved amphiphilic character, with one face having polar residues. Replacing these polar residues by nonpolar amino acids (e.g., Lys-362 to Met) results in inhibiting turnover of the enzyme, although little or no perturbation at the metal centers is apparent. The simplest conjecture is that these residues form part of a proton-conducting channel leading to the heme-copper center. Blocking this channel inhibits the formation of water and slows or eliminates enzyme turnover.

One set of results suggests that the protons that are pumped across the membrane may utilize a different pathway, at least in part. Residue Asp-135 is located in a cytoplasmic loop connecting transmembrane helices II and III. Replacement by Asn causes cytochrome bo ₃ to decouple the electron transfer function from proton pumping (374). This mutant enzyme and others with changes in this region of CyoB have about 60% of the turnover of the wild type but no proton pumping. The molecular explanation is not clear, but the data clearly show that proton pumping is not an essential element of the chemistry by which oxygen is reduced to water. Possibly, this region of CyoB is specifically involved in forming the input part of a channel specifically for pumped protons. Considerably more work is necessary to confirm this hypothesis, but it is a reasonable interpretation of the data.

Superoxide Production.

A side effect of aerobic respiration is the production of superoxide (O₂ ^–) and other partially reduced oxygen species. These highly reactive compounds cause various types of harm, including damage to DNA (126) and to the iron-sulfur clusters of dehydratases (130). By contrast, the iron-sulfur clusters of most respiratory chain enzymes are not damaged by superoxide, because the iron atoms in these clusters are fully cysteine liganded, unlike those in the dehydratases. A variety of defense and repair mechanisms (including synthesis of superoxide dismutase) are induced in response to superoxide, as described elsewhere in this book by Lynch and Lin (see chapter 95).

In vitro studies indicate that the respiratory chain is the likely source of most endogenous superoxide, at least in complex media (175). Superoxide is generated during the autoxidation of flavin-containing dehydrogenases, particularly NADH dehydrogenase II and, to a lesser extent, succinate dehydrogenase (J. A. Imlay, submitted for publication). The role of NADH dehydrogenase I is uncertain. Cytochrome oxidases and quinones appear to be insignificant sources of superoxide (Imlay, submitted).

In vitro measurements suggest that about one molecule of superoxide is formed for every 1,000 electrons delivered to oxygen, although this value varies with different substrates because of different rates of dehydrogenase autoxidation. If these in vitro measured rates are relevant to respiring cells, several micromoles of superoxide would be formed per second in vivo (175).

The anaerobic respiratory enzyme fumarate reductase has an extremely high superoxide turnover number, suggesting that the transition of an enterobacterium from an anaerobic to an aerobic environment will result in a large superoxide flux, as fumarate reductase routes electrons from the quinone pool to oxygen. Indeed, membrane vesicles containing fumarate reductase generate much more superoxide than do vesicles lacking the enzyme, irrespective of the electron donor supplied (206). It is speculated that relatively high and low electron densities around the flavin components of fumarate reductase and succinate dehydrogenase, respectively, lead to the observed relatively high and low rates of superoxide formation (J. A. Imlay, personal communication).

 

S- and N-Oxides

E. coli can use a variety of S- and N-oxide compounds as electron acceptors for anaerobic respiration, including DMSO (dimethyl sulfoxide), TMAO (trimethylamine N-oxide), and adenosine N-oxide (reviewed in references 21 and 412). Other electron acceptors include tetramethylene sulfoxide (tetrahydrothiophene 1-oxide [248]), methionine sulfoxide, picoline N-oxide, and pyridine N-oxide (317, 440). Two well-characterized molybdoenzymes are involved in respiration with these compounds: a membrane-bound constitutive enzyme, termed DMSO reductase (DmsABC enzyme), and a periplasmic inducible enzyme, termed TMAO reductase (TorACD enzyme) (317, 341, 411). A third, distinct enzyme is responsible for respiration with adenosine N-oxide (317). S. typhimurium LT2 can also respire with TMAO (217, 218) and DMSO (V. Stewart, unpublished observations).

The earlier literature suggested that four distinct TMAO reductases are synthesized by E. coli: three inducible enzymes, and one that is constitutively expressed (336). More recent analysis has demonstrated that the three inducible species actually correspond to different association forms of the same enzyme (TorACD enzyme) and that the constitutive form corresponds to the DmsABC enzyme (41, 340). S. typhimurium LT2 also expresses inducible and constitutive enzymes with TMAO reductase activity (218); presumably, the constitutive activity corresponds to DMSO reductase.

Studies with E. coli dms and tor structural gene mutants have shown that either the DmsABC or the TorACD enzyme is sufficient to support respiration with TMAO, pyridine N-oxide, and picoline N-oxide, whereas the DmsABC enzyme is specifically required for respiration with DMSO or methionine sulfoxide. This suggests that TorACD enzyme does not contribute to DMSO respiration (317), a conclusion supported by an independent study (101). Confoundingly, the purified TorA subunit is about one-third as active on DMSO as it is on TMAO (with an artificial electron donor), and torCAD operon expression is induced equally well by DMSO and TMAO (340; but also see reference 443). Therefore, the possible involvement of the TorACD enzyme in DMSO respiration may deserve further attention.

The dms tor double mutant retained the ability to respire with adenosine N-oxide but failed to respire with the other compounds tested (317). Thus, respiration of most S- and N-oxide compounds is accomplished with only two broad-specificity enzymes. A third enzyme is apparently involved in adenosine N-oxide respiration, whereas the genetic requirements for respiration of some other compounds (such as tetramethylene sulfoxide) remain to be determined.

DMSO Reductase (Menaquinol:DMSO Oxidoreductase; DmsABC Enzyme).

E. coli DMSO reductase is the subject of a recent comprehensive review (412). DMSO reduction with a variety of electron donors results in the generation of proton motive force, as revealed by proton extrusion measurements in whole cells. Estimated H⁺/e^– ratios of 1.5 with glycerol as the electron donor and either DMSO or nitrate as the electron acceptor were observed (40), suggesting that DMSO reductase, like nitrate reductase, catalyzes net transmembrane proton translocation. DMSO-dependent proton translocation is sensitive to the uncoupler carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP) and also to the quinone analog 2-n-heptyl-4-hydroxyquinoline N-oxide (HOQNO), but it is insensitive to inhibitors of cytochrome function, i.e., azide and cyanide (40). The enzyme contains three subunits: a large molybdoprotein containing the catalytic site (DmsA), and smaller iron-sulfur (DmsB) and intrinsic membrane (DmsC) subunits (Table 10) (38, 41, 411). Properties of these subunits are summarized below (see Common Features of Respiratory Enzymes).

Extensive physical analysis has established that the DmsA and DmsB subunits are extrinsic membrane proteins that face the cytoplasm (307, 316), whereas results with alkaline phosphatase and β-lactamase fusions indicate that the DmsC subunit is an intrinsic membrane protein with eight membrane-spanning helices (413). This structural topology suggests that the enzyme catalyzes transmembrane proton translocation by a scalar mechanism (see Mechanisms of Proton Translocation, above) (reviewed in reference 412). All three subunits are required for DMSO respiration (318).

Formate, glycerol 3-phosphate, hydrogen, and NADH are all effective electron donors for DMSO reduction, whereas d- and l-lactate are not (39, 40). MK and DMK serve as the intermediate electron carriers (see Quinones, above). DmsABC enzyme does not contain heme, and there is no evidence for the participation of a cytochrome in DmsABC-catalyzed reactions (reviewed in reference 412).

Purified DMSO reductase has a very broad substrate specificity, exhibiting substantial activity with a variety of S- and N-oxides including DMSO, methionine sulfoxide, tetramethylene sulfoxide, TMAO, pyridine N-oxide, picoline N-oxide, and adenosine N-oxide. The enzyme also reduces chlorate but not nitrate (411).

DMSO reductase is encoded by the dmsABC operon, located at 20 min on the E. coli genetic map (38, 41). Enzyme synthesis requires anaerobic growth conditions, is not further induced by DMSO, and is repressed by nitrate during anaerobic growth (39, 41). Analysis of dmsA-lacZ operon fusion expression shows that anaerobic regulation and nitrate regulation are mediated by the Fnr and NarL proteins, respectively (98) (see Regulation of Enzyme Synthesis, below).

TMAO Reductase (TorACD Enzyme).

TMAO reduction results in the generation of proton motive force (369), but it is not established that TorACD enzyme itself catalyzes net transmembrane proton translocation. Both E. coli and S. typhimurium use TMAO as an electron acceptor for anaerobic respiration (reviewed in reference 21).

DNA sequence analysis (250) has revealed a torCAD operon encoding a large molybdoprotein (TorA, 94 kDa) (340, 443), a periplasmic cytochrome c (TorC, 43 kDa), and a presumed intrinsic membrane subunit (TorD, 22 kDa). Analysis of the cytochrome c content has revealed only a single TMAO-inducible species (179). Amino-terminal protein sequence analysis shows that the mature TorA subunit is derived by cleavage of an amino-terminal signal sequence (250), and biochemical studies show that the TorA subunit is localized preferentially to the periplasm (341). The TorC (cytochrome c) subunit is apparently anchored to the periplasmic face of the cytoplasmic membrane by an amino-terminal hydrophobic region (250).

Formate, glycerol 3-phosphate, hydrogen, l-lactate, and NADH are all effective electron donors for TMAO reduction (441, 442; reviewed in reference 21). MK and DMK serve as the intermediate electron carriers (see Quinones, above), and formate- and NADH-dependent TMAO reduction is sensitive to the quinone analog HOQNO (reviewed in reference 21). Spectroscopic studies have revealed both cytochromes b and c to be associated with TMAO reduction, in a proposed arrangement of substrate → cytochrome c → cytochrome b → TMAO (56); this scheme is likely to be modified upon further analysis.

The purified TorA subunit has a broad substrate specificity, exhibiting substantial activity with a variety of N- and S-oxides including TMAO, pyridine N-oxide, DMSO, and tetramethylene sulfoxide. The enzyme also reduces chlorate and exhibits only very weak nitrate and methionine sulfoxide reductase activities (340, 341).

TMAO reductase is encoded by the torCAD operon, located at 22 min on the E. coli genetic map (250). Enzyme synthesis requires anaerobic growth conditions and is not repressed by nitrate during anaerobic growth (279, 340). Analysis of tor-lacZ operon fusion expression shows that anaerobic regulation is not mediated by the Fnr protein; the mechanism of anaerobic regulation is unknown (279).

Enzyme synthesis during anaerobic growth is induced strongly by TMAO, DMSO, and tetramethylene sulfoxide and somewhat more weakly by pyridine N-oxide (341). TMAO induction requires an intact torR gene, located immediately adjacent to the torCAD operon. The torR gene encodes the response regulator protein of a two-component regulatory system, but the cognate sensor protein has not yet been identified (342).

 

Tetrathionate and Thiosulfate

Tetrathionate (S₄O₆ ^2–) reduction supports anaerobic respiration in three genera of enterobacteria: Citrobacter, Proteus, and Salmonella (reviewed in reference 20). However, interest has waned in the past 20 years, and therefore tetrathionate respiration will not be considered further.

The reduction of thiosulfate (S₂O₃ ^2–) to hydrogen sulfide (H₂S) during growth in media such as triple sugar iron agar is a familiar diagnostic test in distinguishing Salmonella spp. from other enterobacteria (including E. coli). S. typhimurium expresses two distinct enzyme systems that independently produce H2S. The first enzyme is thiosulfate reductase, which is described in more detail below. The second enzyme system is the anaerobic sulfite reductase, which requires at least three subunits for activity (AsrABC enzyme) (173). This soluble enzyme system reduces sulfite to sulfide, thus providing an alternate pathway for generating H₂S during anaerobic growth. It is conceivable that anaerobic sulfite reductase serves roles similar to those of NADH-nitrite reductase (see Nitrite, above), i.e., regenerating oxidized nicotinamide cofactor and also serving to detoxify sulfite (E. L. Barrett, personal communication).

Thiosulfate Reductase (PhsABC Enzyme).

The contribution of thiosulfate reduction to anaerobic respiration is obscure. The E_m,7 of the thiosulfate/sulfide plus sulfite couple (less than –400mV [Table 1]) suggests that thiosulfate is not suitable to act as an electron acceptor. However, standard redox potentials provide only crude estimates of the suitability as an electron acceptor, and rapid diffusion of the product H₂S could serve to raise the effective redox potential in vivo (Barrett, personal communication).

DNA sequence analysis indicates that thiosulfate reductase contains three subunits: a large molybdoprotein containing the catalytic site (PhsA, 83 kDa) and smaller iron-sulfur (PhsB, 21kDa) and intrinsic membrane (PhsC, 28 kDa) subunits (168a). These proteins are homologous to those of theWolinella succinogenes polysulfide reductase (PsrABC enzyme), which catalyzes the reduction of polysulfide or elemental sulfur to hydrogen sulfide, allowing for sulfur respiration (214).

Formate serves as an electron donor for thiosulfate reduction in washed cells of S. typhimurium (168a); other compounds apparently have not been tested. S. typhimurium menB and hemA mutants, deficient in MK and heme biosynthesis, respectively, both fail to produce H₂S from thiosulfate, suggesting that the electron transport system requires these two electron carriers (86).

Thisoulfate reductase is encoded by the phsABC operon, located at 41 min on the S. typhimurium genetic map (86; 168a). Expression of phs-lacZ operon fusions is induced by anaerobic growth conditions and only weakly repressed by nitrate during anaerobic growth (3, 86). However, anerobic expression does not require the Fnr (OxrA) protein (86). Expression also depends on the growth phase, carbon source, and presence of reduced sulfur compounds in the medium (86).

 

COMMON FEATURES OF RESPIRATORY ENZYMES

 

Molybdenum Cofactor Subunits

Molybdenum cofactor in enterobacteria consists of an organic moiety (molybdopterin guanine dinucleotide [MGD]) complexed with molybdenum (Mo-MGD). The molybdenum cofactor in most known eubacterial and archaeal enzymes contains a dinucleotide form of molybdopterin (MPT), whereas eukaryotic enzymes contain simply Mo-MPT. In this review, we use the term molybdenum cofactor to denote the bacterial form, Mo-MGD. The structure, synthesis, and function of molybdenum cofactor are described elsewhere in this volume by Rajagopalan (see chapter 42).

Subunits from at least nine enterobacterial anaerobic enzymes contain molybdenum cofactor: the three formate dehydrogenases (FdhF, FdnG, and FdoG), the three nitrate reductases (NarG, NarZ, and NapA [YojC]), the two N- and S-oxide reductases (TorA and DmsA), and thiosulfate reductase (PhsA). Molybdenum cofactor is also present in subunits of at least two aerobically expressed nonrespiratory enzymes, biotin sulfoxide reductase (BisC [106]) and assimilatory nitrate reductase (NasA [231]). Molybdenum cofactor is not covalently attached to the polypeptide backbone, but it is released only upon denaturation of the enzyme.

Genes in five distinct loci are required for molybdenum cofactor biosynthesis or function. These loci were originally identified through analysis of chlorate-resistant mutants. Chlorate (ClO₃ ^–) is reduced to chlorite (ClO₂ ^–) by several different molybdoenzymes, and the resulting chlorite is toxic. Thus, mutations that block formation of molybdenum cofactor allow cells to survive anaerobic incubation in the presence of chlorate. These genes were originally designated as chl (for chlorate resistance; Chl^r phenotype). The subsequent discovery of multiple genes at most of the chl loci has made this nomenclature cumbersome, so these genes have been renamed (333). The generic designation mol is used to describe these genes without reference to a specific locus. The functions of the various gene products are described in chapter 42.

Selection for Chl^r was originally thought to yield some E. coli mutants (termed chlC) specifically defective in the major membrane-bound nitrate reductase (NarGHI enzyme). However, all such chlC mutants examined carry secondary lesions in one of the mol genes (Stewart, unpublished observations), and defined narGHJI structural gene mutants remain fully sensitive to chlorate (357, 361). This indicates that molybdoenzymes other than nitrate reductase contribute to the Chl^s phenotype. Indeed, the DMSO and TMAO reductases (DmsABC and TorACD enzymes, respectively) have substantial chlorate reductase activity (217, 340, 411), and thiosulfate reductase in S. typhimurium (PhsABC enzyme) contributes to the Chl^s phenotype in that organism (304). These observations explain why only mol mutants survive the chlorate resistance selection. A corollary is that molybdoenzymes can be identified as such by the fact that their activity is absent in mol mutants (see, e.g., references 39 and 106).

Molybdenum cofactor-containing subunits are large, ranging from about 80 kDa (PhsA protein) up to about 140 kDa (NarG and NarZ proteins). Despite these differences in mass, amino acid sequence comparisons reveal recognizable homology (43, 412, 428). The sequences share similarity in at least six distinct regions of contiguous residues (termed regions A to F in reference 428). These regions are separated by spacer sequences of different lengths in the different enzymes. Thus, the large NarG and NarZ polypeptides have relatively large spacer regions, whereas the smaller PhsA and DmsA polypeptides have relatively short spacer regions. Specific residues in region A of the DmsA protein are essential for enzyme function (379) (see Iron-Sulfur Subunits, below), and the Sec (selenocysteinyl) residue in region B in the FdhF subunit is essential for its normal catalytic activity (13). Mutational analysis of the other regions has not been reported. It is presumed that these homologous sequences reflect a common basis for binding molybdenum cofactor in the appropriate conformation. Eukaryotic enzymes that contain the Mo-MPT form of molybdenum cofactor are structurally distinct (reviewed in reference 428).

Recent sequence comparison of molybdoenzymes from a variety of prokaryotes (Berks et al., personal communication) has revealed four distinct subclasses of molybdoenzymes based on particular sequence characteristics within region B. This region includes the Sec residue in formate dehydrogenases that provides a ligand to the molybdenum atom in molybdenum cofactor (143). Nitrate reductases have a Cys residue at this position to provide a sulfur ligand, whereas S- and N-oxide reductases probably provide an oxygen ligand from a Ser residue (143). Interestingly, the soluble nitrate reductase sequences (NapA and NasA) in region B are more similar to those of the formate dehydrogenases than to the membrane-bound nitrate reductases (Berks et al., personal communication). Finally, the sequence of polysulfide reductase from Wolinella succinogenes defines a fourth subclass (Berks et al., personal communication), which also includes S. typhimurium thiosulfate reductase (PhsA subunit [168a]).

Molybdenum can exist in three distinct redox states: fully reduced, Mo(IV); partially reduced, Mo(V); and oxidized, Mo(VI). The intermediate state, analogous to the semiquinone state of FAD, allows molybdenum cofactor to accept electrons from iron-sulfur clusters, which are one-electron donors (reviewed in reference 412). The E_m,7 values for the Mo(IV/V) and Mo(V/VI) transitions are about +180 and +220 mV, respectively, for nitrate reductase (393) and about –75 and –90 mV, respectively, for DMSO reductase (70). Thus, the specific protein environment can greatly influence the redox characteristics of the Mo center, allowing it to function with substrates that span a range of redox potentials.

 

Iron-Sulfur Subunits

Iron-sulfur clusters are present in a number of respiratory enzymes, particularly those that function during anaerobic growth. Iron-sulfur clusters function in electron transfer reactions. The properties, structures, and functions of iron-sulfur clusters in diverse proteins have been studied (reviewed in references 26 and 69).

Iron-sulfur clusters consist of covalently bonded iron and sulfur and are positioned within the protein by additional covalent bonds between the iron molecules in the cluster and specific Cys residues in the polypeptide chain. Binuclear ([2Fe-2S]) clusters consist of two atoms each of iron and sulfur, with each iron additionally bonded to one or two Cys residues (Fig. 8). Tetranuclear ([4Fe-4S]) clusters consist of four atoms each of iron and sulfur, bonded in a cubane pattern, with each iron additionally bonded to a single Cys residue (Fig. 8). Some enzymes contain [3Fe-4S] clusters, which are similar in structure to [4Fe-4S] clusters except that one of the Cys residues (and therefore one of the iron atoms) is not present (27) (Fig. 8). For some time, it was thought that some clusters consist of [3Fe-3S] clusters. Subsequent work showed that the prototypical [3Fe-3S] cluster in the Azotobacter vinelandii 7Fe ferredoxin is actually a [3Fe-4S] cluster (364, 365), and no documented examples of [3Fe-3S] clusters exist for the enzymes considered in this chapter.

Primary Structures of Iron-Sulfur Clusters.

Amino acid sequences involved in specifying tetranuclear clusters usually contain a group of four Cys residues. In ferredoxins whose structure is known, the first three Cys residues provide iron ligands for one cluster whereas the fourth Cys provides an iron ligand for a second cluster (reviewed in references 26 and 69). The Pro residue following the fourth Cys residue is thought to promote the formation of a sharp bend in the polypeptide chain. It is presumed that the analogous sequences in larger respiratory proteins contribute iron ligands in similar fashion (reviewed in reference 412).

Representative Cys groups involved in forming [4Fe-4S] and [3Fe-4S] clusters in E. coli respiratory enzymes are compiled in Fig. 9. Individual Cys groups are designated by their position in the polypeptide chain, with the Cys group closest to the amino terminus defined as group I, the next defined as group II, and so on. The sequences of most of these Cys groups resemble one of four distinct motifs based on similarities in Cys spacing and other features of the amino acid sequence. These are arbitrarily designated as motif 1, motif 2, motif 3, and motif 4 for the purposes of this discussion. Cys motif 1 (Cys-X₂-Cys-X₂-Cys-X₃-Cys-X) resembles the ferredoxin-like motif, except that the final residue is not Pro (Fig. 9). Cys motif 2 (Cys-X₂-Cys-X₄-Cys-X₃-Cys-Pro) contains two more residues between the second and third Cys residues. Cys motif 3 (Cys-X₂-Cys-X₂-Cys-X₃-Cys-Pro) is the ferredoxin-like motif. The four [3Fe-4S] clusters, in the FrdB, NarH, NarY, and SdhB polypeptides, are specified by Cys groups that resemble motif 3 in sequence, with the exception that the second Cys residue is replaced by Trp, Ile, or Val (Fig. 9). Finally, Cys motif 4 contains a larger, variable number of residues between the second and third Cys residues (Cys-X₂-Cys-X_8–15-Cys-X₃-Cys-Pro).

Many iron-sulfur subunits of respiratory enzymes contain four Cys₄ groups, in the order amino terminus-motif 1 (group I)-motif 2 (group II)-motif 3 (group III)-motif 4 (group IV)-carboxyl terminus (e.g., DmsB, FdnH, FdoH, HybA, HycB, NrfC, and PhsB [Fig. 9]). The term "four-cluster ferredoxins" has been suggested to describe these proteins (R. A. Rothery, personal communication). Several variations of this general arrangement have been found.

The sequences that specify [2Fe-2S] clusters are more difficult to recognize, in part because the spacing between the Cys residues is rather variable (reviewed in references 61 and 69). The FrdB and SdhB [2Fe-2S] clusters (Fig. 9) are well documented (see below), whereas Cys groups I in the NuoE and NuoG proteins are at present only hypothesized to form [2Fe-2S] clusters on the basis of comparisons with NADH:ubiquinone oxidoreductases from other organisms (see NADH, above).

Iron-Sulfur Clusters in Molybdoprotein Subunits?

The amino-terminal portions of most respiratory molybdoproteins also contain Cys groups of the form Cys-X2–3-Cys-X3-Cys-X26–34-Cys, with the exception that the NarG and NarZ sequences substitute His in place of the first Cys residue (Fig. 9). It is striking that these Cys groups are found mainly in those molybdoproteins that interact with an electron transport chain containing at least one additional iron-sulfur protein (31, 412); only the TorA (250) and biotin sulfoxide reductase (284) sequences do not contain analogous Cys groups. An apparent exception is K. pneumoniae assimilatory nitrate reductase (NasA), which also contains an amino-terminal Cys group but is not thought to interact with a separate iron-sulfur protein (231, 232). However, NasA differs from other molybdoenzymes in containing an apparent carboxyl-terminal iron-sulfur cluster, which may provide an analogous electron transfer function.

It has been tempting to speculate that these Cys groups are involved in ligating iron-sulfur clusters (31). If these molybdoproteins contain a single iron-sulfur cluster, the fourth liganding Cys residue would be expected to reside some distance away in the polypeptide chain (reviewed in reference 395). The substitution of His for the first Cys residue in the NarG and NarZ sequences could conceivably result in formation of a [3Fe-4S] cluster or could represent a rare case of a non-Cys ligand to iron (reviewed in reference 69).

Direct evidence for this idea comes from analysis of T. pantotropha periplasmic nitrate reductase (NapA), which contains an EPR-detectable [4Fe-4S] cluster (57). Additionally, purified E. coli formate dehydrogenase H (FdhF) contains at least 3.3 g-atoms of iron per mol of protein, again suggesting the presence of an iron-sulfur cluster (14).

Mutational analysis of the Cys group in the DmsA protein reveals that substitutions of Ser for residues Cys-38 and Cys-42 (corresponding to the second and third Cys residues in this group [Fig. 9]) lead to enzymatic defects, including failure to accept electrons from the menaquinol pool (379). However, the iron-sulfur cluster content of DmsABC enzyme is fully accounted for by the four [4Fe-4S] clusters in the DmsB subunit (70), and, furthermore, extensive EPR analysis of the mutant enzymes has failed to reveal alterations in the iron-sulfur clusters (379; Rothery, personal communication).

Thus, while these Cys residues are important for electron transfer within the DmsABC enzyme, available evidence indicates that they do not specify a iron-sulfur cluster. One possible reason for the difference between the NapA and DmsA proteins may reside in the relative geometries of the Cys groups: Cys-X₂-Cys-X₃-Cys for NapA but Cys-X₃-Cys-X₃-Cys for DmsA. The difference in spacing between the first two Cys residues might drastically alter the ability of the Cys group to direct cluster formation (Rothery, personal communication). Further work with other enzymes should help to clarify these issues.

Mutational Analysis of Iron-Sulfur Clusters.

Site-specific mutagenesis of the frdB gene has been used to study the [2Fe-2S] cluster in the FrdB protein. Substutions of Ser for Cys at positions 57, 62, and 77 (the first, second, and fourth Cys residues [Fig. 9]) lead to altered EPR and/or redox properties of the [2Fe-2S] cluster (cluster 1) without altering the either the [3Fe-4S] or the [4Fe-4S] cluster (416). Although the [2Fe-2S] cluster still forms in each mutant, its redox potential is decreased and the mutant enzyme activity drops concomitantly. This implicates these Cys residues in providing ligands for the [2Fe-2S] cluster. By contrast, substituting Ser for Cys at position 65 results in a [2Fe-2S] cluster with a slightly increased redox potential (416). A clue to the interpretation of this result comes from the sequence of the SdhB subunit, which contains an Asp residue at the correpsonding position (Fig. 9). Together, these observations indicate that a Cys residue at this position is not needed to assemble a functional [2Fe-2S] cluster with appropriate redox characteristics. One suggestion is that ligation by the carboxyl oxygen might be important at this position (417).

The formation of [3Fe-4S] clusters has also been studied through site-specific mutagenesis of the dmsB, frdB, and narH genes. Cys-102 in the DmsB protein is the second Cys residue in group III and corresponds to the position occupied by Trp-220, Val-207, or Ile-208 in Cys group III of the the NarH, FrdB, and SdhB proteins, respectively (Fig. 9). Changes of Cys-102 to Ser, Phe, Trp, and Tyr result in mutant DmsABC enzymes that fail to transfer electrons from menaquinol, but retain the ability to accept electrons from reduced benzyl viologen. EPR spectroscopy indicates that the mutant enzymes contain a novel [3Fe-4S] cluster along with three [4Fe-4S] clusters (306). A complementary finding was provided by the change of Val-207 to Cys, in group III of the FrdB protein. The resulting mutant FrdABCD enzyme retained some in vivo function. Furthermore, the [3Fe-4S] cluster was replaced with a much lower-potential [4Fe-4S] cluster (E_m,7 = –70 and –350 mV, respectively [243]). Together, these results establish the roles of Cys group III in these enzymes in forming the respective [4Fe-4S] and [3Fe-4S] clusters, and also suggest that the primary sequence of the polypeptide chain is a major determinant in specifying the type and midpoint potential of the resulting iron-sulfur clusters.

In striking contrast, a change of residue Trp-220 to Cys, in group III of the NarH protein, did not result in the conversion of the [3Fe-4S] cluster to a [4Fe-4S] cluster (9). Furthermore, the magnetic interactions of all four iron-sulfur clusters were little changed. Thus, simple interconversion of [3Fe-4S] and [4Fe-4S] clusters is not a general phenomenon. This work highlights the contributions made by other aspects of protein sequence and structure to the formation of iron-sulfur clusters.

Additional studies of site-specific mutants have explored the roles of other Cys residues in the formation of [4Fe-4S] clusters. Changes of residues Cys-204, Cys-210, and Cys-214 to Ser in group III of the FrdB protein resulted in loss of both the [3Fe-4S] and the [4Fe-4S] clusters (243). Apparently, these changes also cause general structural perturbations, as the mutant enzymes are no longer associated with the cytoplasmic membrane. Similarly, Ala substututions for the first Cys residue in each of groups II, III, and IV of the NarH protein (Cys-184, Cys-217, and Cys-244, respectively) cause the loss of all four iron-sulfur clusters (as well as the molybdenum cofactor), although in this case the mutant enzymes remain membrane bound (9). Ser substitutions at positions 217 and 244 have similar effects.

By contrast, the Cys-184 → Ser NarH mutant protein retains all four iron-sulfur clusters and reduced but significant enzyme activity. The reasons for these differences are not currently clear, although reasonable explanations are formulated (9). The EPR properties of the Cys-184 → Ser mutant may also suggest that Cys-184 is involved in ligating [4Fe-4S] cluster 3 (E_m,7 = –200 mV), although this idea requires further tests (9).

A different situation was found for Cys group I in the NarH protein (8). Ser or Ala substitutions for residues Cys-16 and Cys-19, the first two Cys residues in group I, resulted in mutant enzymes with similar properties. All are stable, abundant, and membrane associated, and all have reduced but substantial enzymatic activity. EPR spectroscopy revealed that these group I mutants contain only three iron-sulfur clusters and specifically lack cluster 1 (E_m,7 = +80 mV). These results clearly identify group I Cys residues as ligands for cluster 1.

Together, these results with the FrdB and NarH proteins, coupled with structural studies of ferredoxins, suggest specific ligand assignments for each of the respective iron-sulfur clusters (8, 9, 243). These proposed patterns may provide one guide toward a more detailed understanding of iron-sulfur cluster assembly and structure in these proteins.

Functions of Iron-Sulfur Clusters.

At least some iron-sulfur clusters are clearly involved in transferring electrons from quinol to the active site (8, 9, 243, 306, 416; R. A. Rothery, G. Giordano, F. Blasco and J. H. Weiner, submitted for publication). The mechanisms by which these electron transfers occur remain to be determined. Questions arise, however, about the participation of specific iron-sulfur clusters in electron transfer reactions.

The E_m,7 of the MK/menaquinol couple is about –74 mV (Table 1), whereas the E_m,7 values for electron acceptors range from about +30 mV(fumarate/succinate) to about +430 mV (nitrate/nitrite). One would then expect the E_m,7 values of the iron-sulfur clusters to lie somewhere between –70 mV and the value for the specific electron acceptor in question. However, several of the characterized iron-sulfur clusters have midpoints that are substantially lower than –70 mV: DmsB clusters 3 (–240 mV) and 4 (–330 mV), FrdB cluster 2 (–320 mV), and NarH clusters 3 (–200 mV) and 4 (–400 mV). Likewise, SdhB cluster S2 (–175 mV) is not succinate reducible (Table 5). This is always the case for center S2 of members of the succinate dehydrogenase family, and the function of this iron-sulfur cluster is not known. Finally, center N1a in NDH-I (complex I) also has a very low potential. Do these clusters play roles in electron transport in the respective enzymes?

Mutational alterations of FrdB iron-sulfur clusters 1 and 2 that result in lowered midpoint potentials yield enzymes with reduced yet detectable catalytic activity. It is thus proposed that electron transfer can bypass either of these clusters if necessary but that functional clusters 1 and 3 are necessary for optimal enzymatic activity (243, 416). Thus, at least iron-sulfur clusters 1 and 3 in FrdABCD enzyme seem to be involved in electron transfer from menaquinol to FAD. Likewise, the NarH protein contains two relatively high-potential iron-sulfur clusters. Selective loss of cluster 1 (+80 mV) through mutation leaves an enzyme with sufficient activity to promote growth with nitrate as an electron acceptor. This indicates that either cluster 1 or the [3Fe-4S] cluster 2 (+60 mV) can serve in independent but parallel pathways for electron transfer (8).

Electrochemical studies of the FrdAB catalytic dimer are interpreted to show that the FrdB iron-sulfur cluster 2 serves as a second, slower pathway for this electron transfer when electron donor supply is limiting (367). An alternative view holds that cluster 2 is an evolutionary vestige that does not participate directly in electron transfer but that it may play a role in enzyme structure and stability (243). Similar ideas have been advanced with respect to the NarH and DmsB protein iron-sulfur clusters 3 and 4 (8, 412). However, a recent preliminary report suggests that oxidation and reduction of the low-potential iron-sulfur centers in the NarH protein are involved in inhibiting and activating enzyme activity, respectively, in response to aerobic and anaerobic conditions (B. Bennett and R. C. Bray, Biochem. Soc. Trans. 22:283S, 1994). This provocative idea will undoubtedly be subjected to further tests with site-specific mutant enzymes.

 

Cytochromes and Membrane Anchor Subunits

Membrane-associated respiratory enzymes contain membrane anchor subunits, which also play roles in electron transfer reactions with quinols and quinones. In some enzymes, these membrane anchors are cytochrome b species. Periplasmic enzymes are not known to contain membrane anchor subunits, although at least one intrinsic membrane protein is encoded in each of the structural gene operons (NrfD, TorC and TorD, and NapC [YejX]). Perhaps such proteins serve a role in at least loosely attaching the enzyme to the periplasmic face of the membrane. Periplasmic respiratory enzymes are associated with cytochrome c species, whose properties are also summarized in this section.

Cytochromes b.

Cytochromes b are often found as intrinsic membrane proteins that mediate electron flow between quinone pools and respiratory enzymes. Many respiratory enzymes contain cytochromes b, and the properties of those associated with cytochrome bo ₃ and cytochrome bd are described in the respective sections above (see Oxygen). This section summarizes available information on the cytochromes b that are associated with the other respiratory enzymes covered in this chapter, because these cytochromes appear to share some structural and functional similarities.

  Nitrate reductases A and Z. Nitrate reductase-A contains a cytochrome b with an absorption maximum of approximately 556 nm, cytochrome

(NarI subunit [75, 161, 238, 310, 349]). Redox titration experiments identified two heme b species, carrying E_m,7 values of +20 and +120 mV, associated with nitrate reductase (159, 161). Likewise, chemical measurements estimate the presence of four hemes per NarGHI dimer (75). Since only one polypeptide corresponding to a cytochrome has been identified (348, 349), it is apparent that the NarI subunit must be a diheme cytochrome b. Respiratory nitrate reductase from Paracoccus denitrificans respiratory nitrate reductase likewise contains a diheme cytochrome b with E_m,7 values of +95 and +210 mV (19).

Recent sequence analysis of the T. pantotropha narI gene allows identification of features and residues that are conserved with the E. coli narI and narV (cytochrome b of nitrate reductase Z) genes (33). Four His residues within presumed transmembrane helices are conserved and probably provide the ligands for heme b binding. All four His residues are on conserved (presumably internal) faces of the helices. The overall structure of the NarI subunits is predicted to contain five transmembrane helices, with the amino terminus in the periplasm and the carboxyl terminus in the cytoplasm (33). This topological model awaits direct experimental tests.

The two hemes are modeled as being essentially perpendicular to the surfaces of the membranes, with the lower-potential heme near the periplasmic face and the higher-potential heme near the cytoplasmic face. Such a geometry would allow the electrical potential difference between the two hemes to provide some of the work required for transmembrane electron transfer (33).

  Formate dehydrogenases N and O; hydrogenase-1; thiosulfate reductase. Formate dehydrogenase-N also contains a cytochrome b with an absorption maximum of approximately 556 nm, cytochrome

(FdnI subunit [123, 124, 160, 310]). Attempts to resolve distinct heme b species, for example by redox titration, have not been reported. However, the heme:protein stoichiometry in purified formate dehydrogenase N (124) is consistent with the FdnI subunit being a diheme cytochrome (33).

The NarI structural model described above has been extended for the formate dehydrogenase-O and -N (FdoI and FdnI) cytochromes b, the presumed cytochrome b from hydrogenase-1 (HyaC subunit), and the presumed cytochrome b from thiosulfate reductase (PhsC subunit) (33). These proteins are modeled to contain only four transmembrane helices, with both the amino and carboxyl termini in the cytoplasm. Conserved His residues are predicted to ligate a pair of hemes, one near the periplasm and one near the cytoplasm. Again, experimental tests of this model have not been reported.

  Succinate dehydrogenase. SdhC and SdhD are each small hydrophobic subunits (Table 5) which contain between them a single heme B prosthetic group known as cytochrome b ₅₅₆ (209). In the E. coli succinate dehydrogenase, this heme b is reduced by succinate (167, 209), but in the mammalian enzyme, the midpoint potential of the cytochrome b component is –180 mV, much too low to be reducible by succinate. Site-directed mutagenesis studies (C. Vibat and R. B. Gennis, unpublished observations) have demonstrated that the cytochrome b ₅₅₆ component of E. coli succinate dehydrogenase is not essential for either the assembly of the enzyme or the succinate:quinone oxidoreductase activity. It is likely that the heme facilitates the electron transfer to Q, even if it is not essential for this activity.

Although cytochrome b ₅₅₆ has been reported to be within the SdhC subunit (265) it is likely that both SdhC and SdhD are required and that the heme iron is ligated by one histidine residue in each of the small hydrophobic subunits. Work from the laboratory of Kiyoshi Kita (personal communication) has shown that expression of both the sdhC and sdhD genes is required to overproduce cytochrome b ₅₅₆ in the E. coli membrane (in the absence of the SdhA and SdhB subunits), whereas no heme protein is apparent when sdhC alone is expressed. This is consistent with the results of the site-directed mutagenesis studies (Vibat and Gennis, unpublished) and with the model of Hederstedt et al. (C. Hgerhll and L. Hederstedt, personal communication) based on sequence comparisons. This model is pictured in Fig. 10, which shows three variants of the membrane anchors in this family of enzymes. Type A contains two subunits and two heme components (Thermoplasma acidophilum succinate dehydrogenase [17]), type B contains a single subunit and two hemes (Bacillus subtilis succinate dehydrogenase [163]), and type C contains two subunits and one heme (E. coli succinate dehydrogenase [209]).

It is interesting that the proposed histidyl residues required for the ligation of the heme in E. coli succinate dehydrogenase are also present in the sequence of E. coli fumarate reductase, in which there is apparently no heme prosthetic group. If this model is correct, other aspects of the structure of FrdCD must prevent heme binding in this enzyme. The two histidyl residues which are implicated in the heme binding in SdhCD are about the only common aspects of the sequences when the anchor polypeptides are aligned. These proteins exhibit no recognizable homologies, although they clearly have related functions and combine with the recognizably homologous hydrophilic subunits.

Other Membrane Anchor Subunits.

  Fumarate reductase. Fumarate reductase is anchored to the membrane by the FrdC and FrdD polypeptides (see Fumarate, above). Two somewhat different topological models depict each polypeptide as having three membrane-spanning regions (92, 418), but neither model has been subjected to experimental analysis. The actual topologies probably resemble those of the membrane anchor subunits (SdhCD) of succinate dehydrogenases (Fig. 10).

A mutational alteration of residue His-82 to Arg in the FrdC subunit eliminates fumarate reductase activity when assayed with a menaquinol analog as electron donor but has little effect on the activity measured with an artificial electron donor, reduced benzyl viologen (409). Additional mutations affecting oxidation have been isolated in both the frdC and frdD genes; most of these were constructed by site-specific mutagenesis strategies based on comparative analysis of the photosynthetic reaction center (418, 419). The effects of these changes are interpreted to reveal two quinol-binding sites, involved in one-electron transfers to the iron-sulfur clusters of the FrdB polypeptide. At least one of these quinol sites requires residues from both the FrdC and the FrdD subunits. FrdC residue His-82 is proposed to stabilize the quinone radical, whereas residue Glu-29 may be involved in quinol deprotonation (419). All of these substitution mutants retain significant fumarate reductase activity when assayed with reduced benzyl viologen as the electron donor, whereas certain frdD deletion mutants are defective in this activity (418, 419).

  DMSO reductase. The DmsC subunit of DMSO reductase is an intrinsic membrane protein with eight membrane-spanning helices (413) and is essential for membrane association of the DmsAB complex. Like the FrdCD complex, the DmsC subunit stabilizes the soluble catalytic and iron-sulfur subunits (318).

The DmsAB dimer retains activity with an artificial electron donor but is inactive with a menaquinol analog as the electron donor (318). As is the case with fumarate reductase, this suggests that the DmsC subunit is important for mediating electron flow from menaquinol to the DmsAB complex. Further evidence for this comes from a site-specific mutational change of residue His-65 to Arg, which is analogous to the His-82 → Arg change described for FrdC (see above). This change in the DmsC subunit affects growth on DMSO but leaves activity with an artificial electron donor intact (unpublished data cited in reference 412). Again, this suggests that the DmsC subunit is involved in electron transfer from menaquinol.

Cytochromes c.

Cytochromes c are located in the periplasm; some are soluble proteins, whereas others are anchored to the cytoplasmic membrane. The heme C is covalently attached via thioether linkages to two adjacent Cys residues, located in the motif Cys-X-X-Cys-His (where X is any residue [reviewed in reference 375]). The axial ligands to the iron atoms are provided by the adjacent His and a second proximal Met (or His) residue.

E. coli and S typhimurium aerobic respiratory chains do not contain cytochromes c. However, three anaerobic respiratory enzyme complexes are known or suspected to use cytochromes c in electron transfer reactions. All three involve periplasmic enzymes: respiratory nitrite reductase (NrfAB enzyme), periplasmic TMAO reductase (TorACD enzyme), and periplasmic nitrate reductase (NapABC [YojC-YejYX] enzyme).

Analysis of the cytochrome c content in anaerobically grown E. coli has revealed five cytochromes c (179). A 43-kDa species is induced by anaerobic growth with TMAO, whereas 50-, 24-, 18-, and 16-kDa species are induced by anaerobic growth with nitrite. Genetic and physiological evidence demonstrates that the 43- and 50-kDa species correspond to the TorC and NrfA proteins, respectively. The 18-kDa species is most probably the NrfB protein, whereas the 24- and 16-kDa species probably correspond to the NapC (YejX) and NapB (YejY) proteins, respectively (179).

Little is known about cytochrome c biogenesis in E. coli. However, genetic analysis of B. japonicum and R. capsulatus mutants defective in cytochrome c formation has begun to reveal details of this process (reviewed in reference 375). The apocytochrome c is thought to be translocated into the periplasm through the conventional sec-dependent protein secretion apparatus. An apparent ATP-binding cassette-type transporter is proposed to transport the heme group into the periplasm. A thioredoxin-like protein may be involved in maintaining the critical Cys residues in a reduced state prior to heme ligation. Finally, additional proteins may constitute the cytochrome c-heme lyase activity that ligates the heme to the apocytochrome (24, 25, 301, 305; reviewed in reference 375). Many of these proposed functions are based on primary sequence analysis and await the results of direct experimental tests.

DNA sequence analysis has revealed homologs for most of these genes in E. coli. Two homologous genes are present in the nrfABCDEFG operon, encoding the Nrf enzyme, and homologs for many of these genes are also present in the aeg-46.5 operon, which also encodes the NapABC (YojC-YejXY) enzyme (see Nitrate, above). The ATP-binding cassette transporter encoded by the cydCD genes is also involved in cytochrome c biogenesis (289). The presence of these presumed biogenesis genes adjacent to the structural genes for known or suspected cytochromes c further suggests that they play roles in cytochrome c formation. This genetic arrangement ensures coexpression of cytochrome c structural and biogenesis proteins under the appropriate physiological conditions (174). It also provides a rationale for the observed regulation of the aeg-46.5 operon, which apparently encodes periplasmic nitrate reductase yet is more strongly induced by nitrite than by nitrate (299). This induction by nitrite may reflect the need for the cytochrome c biogenesis genes in Nrf enzyme synthesis.

 

REGULATION OF ENZYME SYNTHESIS

The transcriptional regulatory proteins Fnr, ArcA, and perhaps AppY control respiratory enzyme synthesis in response to anaerobiosis. A further level of transcriptional control, in response to the availability of nitrate and nitrite, is provided by the regulators NarL and NarP. These regulatory pathways have been extensively reviewed (157, 158, 188, 191, 192, 230, 353, 354, 358, 363) and are only summarized here. Aspects of this topic are also described elsewhere in this collection by Lynch and Lin (see chapter 95).

 

Response to Anaerobiosis

The Fnr Protein.

The fnr (fumarate and nitrate reductase) gene was originally identified by isolating pleiotropic mutants that lack most anaerobic respiratory pathways (83, 219, 270), and studies with lacZ operon fusions showed that the Fnr protein is an activator of anaerobic respiratory gene expression (82, 197, 356). The S. typhimurium fnr gene, termed oxrA, was identified by studies of mutations that affect regulation of pepT gene expression (366). The fnr gene is located at 29 min on the E. coli genetic map (219). Most of the anaerobic respiratory enzymes described in this chapter require the Fnr protein for synthesis; exceptions include the HyaABC, TorACD, and PhsABC enzymes (Table 11). The Fnr protein also acts as an anaerobic repressor of ndh gene expression (149, 355) and of cyoABCDE operon expression (see below).

Mutational studies identified the Fnr core consensus binding half-site as NTTGAT, similar to the Crp (cyclic AMP receptor protein) core consensus binding half-site of TGTGAN (195; reviewed in reference 353). In vitro studies confirm that Fnr protein binds to these sequences (151, 222, 335). Indeed, the fnr and crp genes are homologous, and appropriate directed changes in the Fnr and Crp DNA target sites or in the DNA-binding domains result in generally predictable interconversions of binding specificity (29, 229, 352, 447). These and other observations (28, 208, 421, 447a) suggest that the Fnr and Crp proteins may have common mechanisms of transcription activation.

The Fnr protein contains four essential Cys residues, three of which are located near the amino terminus (150, 251, 334, 377). Iron, presumably liganded by these Cys residues, is essential for Fnr protein activity (207a, 355, 377), and it is postulated that redox transitions of the bound iron atom mediate Fnr protein action (150, 207a). Fnr protein activity seems not to be directly controlled by molecular oxygen but, rather, to be controlled through a redox-responsive component (presumably the liganded iron within Fnr) with a midpoint potential of about +400 mV (386). Genetic and biochemical evidence indicates that active (anaerobic) Fnr is a dimer whereas inactive (aerobic) Fnr is a monomer (207a, 222, 447a).

The ArcA Protein.

The arcA (aerobic respiration control) gene, located at 0 min on the E. coli genetic map, was identified by mutations that affect sdhCDAB operon expression (187). The ArcA protein is a response regulator; the cognate sensor (histidine protein kinase) is encoded by the arcB gene located at 69 min on the E. coli genetic map (193). The corresponding S. typhimurium genes are located at homologous positions (4). The ArcA protein acts primarily as a transcriptional repressor of genes whose products are involved in aerobic metabolism (187) (Table 11). However, the ArcA protein also activates the transcription of some genes whose products are involved in anaerobic metabolism (e.g., pfl and cob [4, 323]). Initial studies of ArcA-DNA interactions have been reported (370), but much work remains to define ArcA-binding sites and ArcA protein action at various target operon control regions.

The ArcB protein controls ArcA protein activity through transphosphorylation reactions (182, 189, 190). The ArcB protein does not appear to respond directly to molecular oxygen but may be controlled instead by accumulation of fermentation products and other metabolites in anaerobic cells (182, 184). Thus, ArcA and Fnr protein activity is mediated in response to distinct physiological signals (183, 184).

Dual Regulation by the Fnr and ArcA Proteins.

Expression of some operons is affected by null alleles of both fnr and arcA. Two examples are provided by the cyoABCDE and cydAB operons, encoding the cytochrome bo ₃ and bd complexes, respectively. Anaerobic repression of cyoA operon expression is relieved in arcA null strains, indicating that the ArcA protein acts to repress cyoA operon expression (99, 184). Further results were interpreted to reflect Fnr protein repression of cyoA operon expression as well (97, 99, 184). The ArcA protein apparently activates cydAB operon expression (99, 135); the Fnr protein was suggested to act either as an activator (135) or as a repressor (99) of cydAB operon expression (reviewed in reference 158). Thus, one concept is that the ArcA and Fnr proteins both act as direct regulators of cyoA and cydAB operon expression (99, 158).

An alternate view, i.e., that the Fnr protein is only an indirect regulator of cyoA and cydAB operon expression, is provided by three recent observations. First, arcA gene expression is activated by the Fnr protein (95). Thus, the relief of anaerobic cyoA operon repression (for example) observed in fnr null strains could be explained by decreased synthesis of the ArcA repressor. Second, arcA null alleles are epistatic to fnr null alleles with respect to anaerobic repression of sdhCDAB and lctDRP operon expression (95, 183), observations inconsistent with an independent, direct role for Fnr protein in repression of these operons. Finally, physiological studies suggest that fnr null strains fail to accumulate the appropriate balance of ArcB protein signal metabolites, thereby influencing ArcA repressor activity (183). Thus, much work remains to dissect the relative roles of the ArcA and Fnr proteins in the anaerobic regulation of diverse target operons.

AppY Protein.

The appY gene at 13 min on the E. coli genetic map was identified in a search for positive regulators of appA (acid phosphatase) gene expression (7). Recent studies indicate that the AppY protein is involved in anaerobic induction of the hyaABCDEF (hydrogenase 1) and cyxAB appA (cydAB homologs and acid phosphatase) operons (6, 58). Evidence suggests that the AppY protein may also be involved in regulation of several other operons (7). These observations will undoubtedly stimulate much activity in the next few years to determine the scope and nature of AppY action in regulating respiratory enzyme synthesis.

 

Response to Nitrate and Nitrite

The NarL and NarP Proteins.

Nitrate is the preferred electron acceptor for anaerobic cells, presumably because of the relatively high midpoint potential of the nitrate/nitrite couple (Table 1). Thus, nitrate acts to induce the synthesis of nitrate respiratory chain components and simultaneously to repress the synthesis of other anaerobic respiratory chain components (reviewed in reference 357). The narL gene was initially identified as encoding a nitrate-responsive positive regulator of narG operon expression (356), and subsequent work demonstrated that the narL gene is also required for nitrate repression of frdA operon expression (186, 205, 359). The NarL protein is a response regulator, and a cognate sensor (histidine protein kinase) is encoded by the narX gene (reviewed in reference 358). The complex narXL operon maps at 27 min on the E. coli genetic map, adjacent to the narK and narGHJI operons (119, 356).

Further work led to the discovery of a second nitrate-responsive sensor encoded by the narQ gene at 53 min on the E. coli genetic map (81, 119, 298) and a second nitrate-responsive response regulator encoded by the narP gene at 46 min on the E. coli genetic map (299). The predicted NarX and NarQ proteins are homologous, as are the predicted NarL and NarP proteins. Evidence suggests that either the NarX or the NarQ protein is sufficient for nitrate-responsive positive regulation (phosphorylation) of both the NarL and NarP proteins, which themselves act to positively or negatively regulate target operon expression (299; reviewed in references 358 and 363). Transphosphorylation reactions have been studied in vitro (330, 397).

Nitrite-responsive gene expression is also mediated by the Nar regulatory proteins (Table 11). For example, expression of the nrfABCDEFG (formate-dependent nitrite reductase) operon is repressed by the NarL protein in response to nitrate but activated by the NarL and NarP proteins in response to nitrite (299, 384). Several different patterns of Nar-mediated regulation operate at other target operons (299; reviewed in references 358 and 363). It is hypothesized that these dual two-component regulatory systems mediate a complex variety of regulatory adjustments in response to a dynamic ratio of two alternate electron acceptors, nitrate and its reduction product, nitrite (reviewed in reference 363).

Mutational studies identified the NarL protein core consensus binding site as TACYNMT (where Y is C or T and M is A or C) (114, 229, 383, 384). In vitro studies confirm that the NarL protein binds to these sequences (228, 398). So-called NarL heptamer sequences are present in diverse orientations and locations within different target operon control regions (reviewed in references 358 and 363). The NarP protein binds to some of the same sites, but it does so with different affinity depending on the specific sequence (102a, 384).

All known NarL- and NarP-regulated target operons are also controlled by the Fnr protein in response to anaerobiosis (Table 11), raising interesting questions concerning the nature of dual activation by both the Fnr and NarL or NarP proteins (208, 383, 396; reviewed in reference 358). This and many other issues await further experimentation.

Nitrate Regulation of ArcA-Controlled Operon Expression.

The ArcA-repressed sdhCDAB (succinate dehydrogenase) and lldDRP (l-lactate dehydrogenase) operons are expressed at low levels during anaerobic growth (187). However, these operons must be efficiently expressed during anaerobic growth with nitrate, because both succinate and l-lactate serve as electron donors for nitrate respiration (see Nitrate, above). Recent analysis indicates that the addition of nitrate significantly relieves anaerobic repression of sdh-lacZ and lct-lacZ operon fusion expression (183). This relief appears to be independent of the NarL protein but requires an active nitrate respiratory chain, suggesting that nitrate control is mediated by its effects on the accumulation of metabolites that act to signal the ArcB sensor (183). This provocative idea demonstrates how physiological effects can indirectly influence patterns of gene expression. These observations also further underscore the very different types of physiological responses that are mediated by the Arc and Fnr regulatory systems (183).

ACKNOWLEDGMENTS

Many colleagues generously proofread portions of the manuscript, engaged in helpful discussions, and provided information in advance of publication: Ericka Barrett, Stuart Ferguson, Thorsten Friedrich, Lars Hederstedt, Peter Hinkle, James Imlay, Michael Johnson, E. C. C. Lin, Tomoko Ohnishi, Robert Poole, Alan Przybyla, David Richardson, Richard Rothery, James Shapleigh, and Hanns Weiss.

 

NOTE ADDED IN PROOF

The crystal structures of cytochrome c oxidase from Paracoccus denitrificans and bovine heart mitochondria have recently been reported at 2.8 (0.28 nm) resolution (S. Iwata, C. Ostermeier, B. Ludwig, and H. Michel, Nature [London] 376:660–669, 1995; T. Tsukihara, H. Aoyama, E. Yamashita, T. Tomizaki, H. Yamaguchi, K. Shinzawa-Itoh, R. Nakashima, R. Yaono, and S. Yoshikawa, Science 269:1069–1074, 1995). The proposed structures are consistent with the mutational and biophysical analyses of the E. coli cytochrome bo ₃ complex as described in this chapter.