Denitrification constitutes one of the main branches of the global nitrogen cycle sustained by bacteria. The cycle manifests the redox chemistry inherent to the principal natural inorganic nitrogen species dinitrogen, ammonia, and nitrate, as shown in its essential, minimized version in Fig. 1. Nitrogen is intro-duced into the biosphere by biological and chemical fixation of dinitrogen (N2 ) and removed from there again by denitrifica-tion. In doing this, denitrification catalyzes successively N-N bond formation in the transformation of its intermediates ni-tric oxide (NO) and nitrous oxide (N2 O) to the next-lower oxidation state. The bacterial process is nearly exclusively a facultative trait. Its expression is triggered in the cell by the environmental parameters low oxygen tension and availability of an N oxide. Denitrification is part of the bioenergetic apparatus of the bacterial cell, where the N oxyanions nitrate and nitrite and the gaseous N oxides NO and N2 O serve in lieu of dioxygen (O2 ) as terminal acceptors for electron transport phosphorylation. It is clearly the role of denitrification in the global N cycle and in cellular bioenergetics that makes a detailed knowledge of this process essential. More recent concerns related to denitrification begin to fos-ter research in this area. Nitrate, irrespective of its role as essential plant nutrient, has become a pollutant of groundwa-ter and surface water, causing a major problem for the supply of drinking water. N2 O is next to CO2 and CH4 in its impor-tance as a potent greenhouse gas (its efficiency is much higher than that of CO2 ), and, together with NO, it is of much concern in terms of the ozone chemistry of the atmosphere (167, 204). Wastewater treatment plants may contribute to the emission of N2 O and increase the greenhouse effect (624). The residence time of N2 O in the atmosphere is estimated to be 150 years (453). Over the past decades, a constant increase has been observed; fertilizer denitrification is thought to contribute sig-nificantly to this increase. Biomass burning and nylon manu-facture are among the more recently discovered significant anthropogenic N2 O sources (152, 812). The flux of gases be-tween the soil and the atmosphere due to bacterial activities has been described in detail, with specific attention being paid to NO and N2 O (154). The distribution of denitrification among the prokaryotes does not follow a distinct pattern. The reaction is carried out by a diversity of bacteria belonging taxonomically to the vari-ous subclasses of the Proteobacteria. Denitrification also ex-tends beyond the bacteria to the archaea, where it is found among the halophilic and hyperthermophilic branches of this kingdom and may have evolutionary significance. Intriguingly, the NO- and N2 O-utilizing enzymes share structural elements with certain terminal oxidases of the aerobic respiratory chain. An entirely new development is the recognition of the core enzymes of denitrification in the mitochondria of certain fungi (460). Emphasis will be given here to the denitrification process in the strict sense. Nitrite reductase is the key enzyme of denitri-fication in catalyzing the first committed step that leads to a gaseous intermediate. For the denitrification process to lead to dinitrogen formation, the nitrite reductase reaction is comple-mented by the activity of two distinct metalloenzymes, which use NO or N2 O as substrates. Because much work on denitri-fication has been done with pseudomonads and several strains of Paracoccus, members of these groups will be prominent throughout this article, but most findings from other denitrifi-ers are included to present the entire scope and the manifold cell biological ramifications of the field. The process embraces important biological problems of structure and function in microbial bioenergetics and requires highly interdisciplinary approaches for its study. This review centers on the interlacing of denitrification with other cellular processes, emphasizing the concepts emerging from the sub-stantial advances in the biochemistry and genetics of this pro-cess during the past decade. Denitrification will be covered in terms of the structural and functional aspects of the partici-pating enzymes and with a focus on general cellular aspects to include the genetic basis and regulation of the process, the biological chemistry of activation and transformation of small inorganic molecules, transport and assembly processes related to metalloprotein biogenesis, cofactor biosynthesis, and anaer-obic gene expression and gene activation by environmental factors. Denitrification has been covered twice in this journal in the past (459, 642); in addition, a monograph has described the earlier developments in this area (643). Bioenergetic (793, 794) and environmental (678, 820) aspects of the process have also been emphasized in reviews elsewhere. The following abbreviations are used in this article: ABC, ATP-binding cassette; ALA, 5-aminolevulinic acid; asc, ascor-bate; BV, benzyl viologen; cAMP; cyclic AMP; COX, cyto-chrome- c oxidase; CRP, cyclic AMP receptor protein; CuNIR, Cu-containing respiratory nitrite reductase; DDC, diethyl di-thiocarbamate; DMSO, dimethyl sulfoxide; EXAFS, extended X-ray absorption fine structure; EPR, electron paramagnetic resonance; IHF, integration host factor; MCD, magnetic cir-cular dichroism; MGD, molybdopterin guanine dinucleotide; MPT, molybdopterin; MV, methyl viologen; ORF, open read-ing frame; PMS, phenazine methosulfate; SDS, sodium dode-cyl sulfate; SOD, superoxide dismutase.
Denitrification originally described a phenomenon, i.e., the loss of fixed nitrogen from the viewpoint of the N balance of a fermenting microbial community. The term was preserved by Kluyver when he formulated his unifying concept of cellular bioenergetics and recognized that denitrification allows bacte-ria a respiratory way of anaerobic life (458). An N oxide, instead of oxygen, serves as the electron acceptor for the gen-eration of an electrochemical gradient across the cytoplasmic membrane. Because a bacterial cell often disposes electrons over several terminal oxidoreductases that use different N ox-ides, their concerted action results in the sequential transfor-mation of nitrate to N2 . Denitrification is also found in chemo-lithotrophic organisms. The oxidation of an inorganic source of reductant coupled to N oxide reduction is sometimes termed autotrophic denitrification. The phenomenological definition of denitrification is thus the dissimilatory transformation of nitrate or nitrite to a gas species concomitant with energy conservation, as opposed to the assimilatory reduction of the same oxyanions to ammonia for biosynthetic purposes. Dissimilatory and assimilatory pro-cesses are juxtaposed in Table 1. The dissimilatory branch comprises ammonification in addition to denitrification, and both processes are initiated by respiratory nitrate reduction. Ammonification is the reduction of nitrate to ammonia that does not serve the purpose of N autotrophy. Sometimes the term is used to describe the liberation of ammonia from an organic molecule or even for the conversion of N2 to ammonia (N2 fixation), but those are the less preferred choices. A bacterium is either denitrifying or ammonifying; appar-ently there is no option within the cell to proceed either way. The ammonifying pathway is mostly not electrogenic, detoxi-fies nitrite, and serves as an electron sink. Ecological, physio-logical, biochemical, and genetic aspects of ammonification have been covered elsewhere (153, 820). Denitrification will be considered here as the assemblage of nitrate respiration, nitrite respiration combined with NO re-duction, and N2 O respiration: NO3 -NO2 -NO -N2O - N2 NO has recently been proven to be an intermediate (354, 940, 966). Both N2 O and NO fulfill the criteria expected from obligatory and free intermediates, i.e., kinetic competence, the possibility to feed the intermediates as precursors into the process and to detect them upon chemical or mutational block-age of the pathway, and biochemical and genetic evidence for the enzymes responsible for metabolizing these intermediates. The physiology and regulation of NO metabolism are closely interlaced with nitrite reduction (see the section on regulation, below). Whether NO reduction is able to sustain a respiratory process in its own right, independent of nitrite reduction, re-quires further investigation. Even in a complete denitrifier, the three respiratory complexes maintain a certain degree of inde-pendence since they respond to combinations of different ex-ternal and internal signals. The entire process is thus best described as of modular organization. When it is possible to distinguish the individual process, use of the specific over a more global term is preferred. For example, the qualifier ¡Èdis-similatory¡É does not distinguish between the respiratory or ammonifying modes of nitrate reduction; note also that seen as isolated processes, neither nitrate respiration nor N2 O respi-ration is denitrifying in the original sense of the term. Trace gas metabolism depending on nitrate or nitrite and usually yielding N2 O without having bioenergetic significance is observed for both the ammonifying and assimilatory branches of nitrate reduction and is usually referred to as denitrification sensu lato (83, 775, 776, 947). In the ecophysi-ological context, the possibility of chemodenitrification, i.e., the nonenzymatic conversion of nitrate or nitrite to a gas species, has to be considered. Organismic diversity is an important aspect of denitrification and encourages comparative and evolutionary interpretations. This requires a consolidated taxonomic basis upon which to formulate common principles. New names or modifications of familiar ones reflect the progress in clarifying the systematic relationship among the denitrifying bacteria. Relevant nomen- clatorial transfers of denitrifying taxa are tabulated in the final section, and the currently valid names are used throughout this article. Invalid taxa, following convention, are placed in quo-tation marks. The N oxide species NO • and N2 O are commonly termed nitric oxide and nitrous oxide in the denitrification literature, based on the continuing practice to amend the more highly oxidized, oxygen-rich species with the suffix -ic and the more highly reduced one with the suffix -ous. International Union of Pure and Applied Chemistry nomenclature rejects this conven-tion and recommends the terms ¡Ènitrogen monoxide¡É for NO • and ¡Èdinitrogen monoxide¡É for N2 O; these terms, although more accurate than the trivial names, have not yet found ac-ceptance. NO • is a radical, but for simplicity will be written below without the radical designation. Nitrification is a property of both chemolithotrophic and heterotrophic bacteria. The expression ¡Èheterotrophic nitrifi-cation¡É is used here to describe the oxidation of ammonia associated with organotrophic metabolism to contrast it with the better-known variant of autotrophic nitrification. Note, however, that the same term has been used to describe the conversion of reduced nitrogen in amines, oximes, hydroxam-ates, and other N compounds to a higher oxidation state of nitrogen or the liberation of nitrate and nitrite.
GENETIC BASIS OF DENITRIFICATION
Access to denitrification genes was first sought by conjuga-tional and transductional mapping in Pseudomonas aeruginosa. An important outcome of the early genetic analysis was the finding that P. aeruginosa encodes the respiratory (nar) and the assimilatory (nas) nitrate reductase systems from distinct gene sets (761). This was also shown for Ralstonia eutropha (former-ly Alcaligenes eutrophus) (892) and the nitrate respirer Kleb-siella pneumoniae (515, 901) and is assumed to be the rule for nitrate-assimilating denitrifiers. The distinct genetic basis for the respiratory and assimilatory process manifests itself in reg-ulatory responses. Genes for nitrate assimilation are repressed by ammonia and do not respond to oxygen, while the expres-sion of nar genes occurs at low oxygen concentrations and does not respond to ammonia. Oxygen inhibits nitrate uptake for nitrate respiration of denitrifiers or nitrate respirers (335) but has no effect on nitrate assimilation. Genes encoding the five N oxide reductases of denitrifica-tion have been identified from random transposon Tn5 mu-tagenesis, complementation analysis, and screening of cosmid and expression libraries. The genes currently identified as as-sociated with the process are listed in Table 2. Once the anal-ysis of the nitrate-reducing system of denitrifiers has advanced to the level achieved with enterobacteria, the total number of genes necessary for denitrification might well increase to about 50 for a single organism.
Gene Clusters and Nature of Denitrification Genes
The genes for denitrification encoding functions for nitrate respiration (nar), nitrite respiration (nir), NO respiration (nor), and N2 O respiration (nos) are assembled in clusters in Pseudo-monas stutzeri (100, 429), P. aeruginosa (22), Paracoccus deni-trificans (70, 185, 186), Sinorhizobium (formerly Rhizobium) meliloti (357), and ¡ÈAchromobacter cycloclastes¡É (556). The nir and nor genes are closely linked in the pseudomonads and in Paracoccus denitrificans. Figure 2 shows the comparative gene organization of the three denitrifiers for which these clusters have been analyzed to a significant extent. The nir-nor gene clusters harbor the structural information for both reductases and the functions for metal processing, cofactor synthesis, elec-tron donation, protein maturation, assembly processes, and regulation. So far only in P. stutzeri are the nos genes linked with the nir and nor genes, forming a supercluster of about 30 kb comprising 33 genes. Clustering of nir and nor functions may be common among denitrifiers, at least those depending on cytochrome cd1 nitrite reductase, and may facilitate genetic analysis. In the following, a brief characterization of genes in catego-ries of the function of the gene products is given (Table 2). The isolation of genes for N2 O utilization marked the beginning of the current stage of genetic analysis. The nos gene cluster of about 8 kb was found by mapping Tn5 mutations in mutants defective in N2 O reductase expression or processing. nosZ, the structural gene for N2 O reductase, was identified by heterolo-gous expression in Escherichia coli and was the first denitrifi-cation gene of known structure (878). The nosZ probe derived from P. stutzeri hybridized with homologous genes of P. aerugi-nosa (974), R. eutropha (974), Paracoccus denitrificans (349), and S. meliloti (357), indicating a high degree of conservation among N2 O reductases, which was confirmed by sequencing these genes. The nirS gene for cytochrome cd1 nitrite reductase was first isolated from P. aeruginosa with oligonucleotide probes de-signed from the amino acid sequence of the purified enzyme (770). It was found independently by targeting nirM, the gene for cytochrome c551 , which is located in P. aeruginosa just downstream of nirS (25, 613). The cytochrome cd1 genes from P. stutzeri ZoBell and JM300 were isolated with a phage ex-pression library and protein-deduced oligonucleotide probes, respectively (430, 774). The structural gene, nirK, for the Cu-containing nitrite re-ductase (originally also termed nirU or nir) was identified in Pseudomonas sp. strain G-179 by targeting the wild-type gene with Tn5 (941), in Alcaligenes faecalis S-6 by using protein-derived oligonucleotides as screening probes (608), and in Pseudomonas aureofaciens by screening an expression library with an anti-NirK antiserum (282). The norCB genes encoding the NO reductase complex were found by screening a cosmid bank of P. stutzeri with an oligo-nucleotide probe derived from the N terminus of the purified NorC subunit (99). The nor genes were shown by cosmid map-ping to be closely linked to nirS (429). Anticipating a similar clustering of nir and nor in P. aeruginosa (23) and Paracoccus denitrificans (186), homologous norCB genes were identified by sequencing nirS-carrying DNA fragments. By now, a small pool of primary structures of the reductases from various sources is known. The sequence relationship among the denitrification enzymes is depicted in phylogenetic unrooted trees in Fig. 3. NirS and NosZ sequences form rela-tively tight clusters, whereas the plasmid-encoded NorB se-quence of R. eutropha is somewhat removed from its homologs. While the NorB proteins and NosZ proteins of the pseudo-monads are found at neighboring branches of their respective trees, the pseudomonadal NirS proteins are separated. This suggests an evolution of NirS independent from the other reductases. Also on the basis of other arguments (see the section on N2 O respiration, below), the utilization of N2 O is conceived as a separate respiratory system. The NirF and NirN proteins, which are presumed to be paralogous with NirS, are found on separate lineages on the NirS tree and are only distantly related to NirS. The NirK proteins show no well-defined clustering. The NirK proteins of Rhodobacter spha-eroides and P. aureofaciens evolved along a separate lineage and are distant both from the rest of the NirK proteins and from each other. The nar genes for respiratory nitrate reduction are not linked to the denitrification genes proper in P. aeruginosa and P. stutzeri. The structural genes for the nitrate reductase com-plex, narGHJI, have been sequenced from the nitrate respirers E. coli (79) and Bacillus subtilis (168, 350) but remain to be completed for the first example of a denitrifier (70, 646). The structural genes encoding the two subunits NapA and NapB of the periplasmic nitrate reductase were first identified by screening a cosmid library of the pHG1 megaplasmid of R. eutropha with oligonucleotides derived from the N termini of the purified subunits (762). The two structural genes were sequenced, and the basic properties of this enzyme derived. A more extended nap locus, comprising napEDABC, was cloned and sequenced from a genomic cosmid library of Paracoccus denitrificans GB17 (71). The probe for screening was based on gene amplification from internal peptide sequences of the pu-rified NapA subunit. Homologs of napA have been detected by hybridization in P. aeruginosa and P. stutzeri; in both cases, these genes are not linked to the nar genes (883). Different types of regulators are encoded within the nir-nor clusters by the genes nnr, dnr, fnrD, nosR, nirI, and nirY. The genes nosR and nirI encode putative regulators for nosZ and nirS expression, respectively, and are predicted to possess iron-sulfur centers (see the section on regulation, below). nnr, dnr, and fnrD are each found downstream of and in the opposite direction to the nor operons (Fig. 2). The encoded proteins belong to the FNR family of transcription factors (22, 870, 882). Regulatory genes for denitrification belonging to the FNR family also reside outside the known denitrification loci. anr of P. aeruginosa encodes a global transcriptional activator for anaerobic metabolism including denitrification (264). The structurally although not functionally homologous gene fnrP of Paracoccus denitrificans maps adjacent to cco, which encodes a cytochrome cbb3 -type oxidase (187, 871). This is also the case for the anr homolog fnrA of P. stutzeri (882); however, this gene does not affect denitrification directly. In Rhodobacter sphaeroides, the nir and nor regions are not closely linked on the chromosome. Instead, they form a regu-lon under the common control of NnrR, another member of the FNR family (497). The nnrR gene is located there imme-diately upstream of norCB and is transcribed in the same direction. The narXL genes, which encode a two-component system and mediate the nitrate response, are part of the narG locus in P. stutzeri (320). The intergenic region of the nos and nir operons of P. stutzeri harbors nirY (5 orf286), whose deduced product is similar to LysR-type regulators (283). The LysR family is distinguished by a consensus DNA-binding motif in the N-terminal domain and comprises factors that control a broad variety of processes including the oxygen stress regulon (725). nirY is part of a gene region encoding components of heme D1 synthesis and pro- cessing of cytochrome cd1 , but the target of its product is unknown. To assign an electron donor to its cognate reductase will require genetic evidence. At present, only a small number of structural genes for electron carriers have been identified. nirM, immediately downstream of nirS of P. aeruginosa, en-codes cytochrome c551 , an electron donor for NirS (613). P. stutzeri harbors the gene sequence nirSTBM, of which nirM is homologous to the P. aeruginosa gene and nirT encodes a tetraheme cytochrome with a putative electron donor function (430). The relationship between the nirT and nirM products with respect to electron donation to NirS is unknown. nirT is cotranscribed with nirS in P. stutzeri, which indirectly supports the view of a function in nitrite reduction (319). nirT and nirB have yet to be identified in other denitrifiers; neither gene is part of the currently analyzed gene clusters of P. aeruginosa and Paracoccus denitrificans (Fig. 2). However, nirT is similar to napC, encoding the putative electron-transferring cyto-chrome c for the periplasmic nitrate reductase (see the follow-ing section). Several homologs of nirT are present in nondeni-trifiers, thus ascribing a broader importance to this gene (66, 116, 213, 560). Genes encoding electron donors are not necessarily part of the denitrification gene clusters. The cycA gene encoding cy-tochrome c550 of Paracoccus denitrificans lies adjacent to the isogene for the subunit I of COX (872). For the genes encod-ing the electron carriers for nitrite reductase, azurin (azu) and pseudoazurin (paz), no link to a denitrification gene has yet been recognized. Genes for ancillary functions in denitrification were found on sequencing the vicinities of the structural genes for the reductases. Heme D1 is the cofactor of denitrifiers that depend on the cytochrome cd1 nitrite reductase. Genes for heme D1 biosynthesis are part of the denitrification gene clusters in the three organisms studied. The first gene with a likely function in heme D1 biosynthesis, nirE, was found immediately down-stream of nirS in Paracoccus denitrificans (Fig. 2). The derived protein shows high similarity to methyltransferases acting on uroporphyrinogen III (185). The genes for heme D1 biosyn-thesis are distributed in P. stutzeri over two loci, nirJEN and nirCFDLGH, which lie upstream and downstream of nirS, re-spectively (283, 634). The same genes of P. aeruginosa, nirCFDLGHJEN, are clustered in one locus downstream of nirS and may encode the entire set of proteins required for the reaction steps leading to heme D1 (444). Genes encoding functions for metal processing, protein as-sembly, or maturation represent a further category of recog-nized denitrification genes (Fig. 2). Of Tn5-derived mutants defective in N2 O respiration, a distinct group synthesized an apo-N2 O reductase only. The affected genes of these mutants, nosDFY, encode a metal insertion apparatus for the reductase (984). Homologous genes have been found in other denitrifi-ers. In P. aeruginosa (974) and Paracoccus denitrificans (349), nosD genes also exist downstream of nosZ. In more extended analyses, nosRZDFY sequences have been identified in ¡ÈA. cycloclastes¡É (556) and S. meliloti (357), suggesting a high degree of organizational conservation. In R. eutropha, these ancillary genes have not yet been identified since they are not located in the immediate vicinity of nosZ (974). The functions of the genes nosL and nosX (homologous to nirX) are still insufficiently understood. The nosL product is tentatively considered an outer membrane protein featuring the signal peptide of a lipoprotein (224). The signature of a disulfide isomerase deduced from the P. stutzeri protein is not present in NosL of ¡ÈA. cycloclastes¡É (556) and S. meliloti (134), which invalidates the original proposal and requires a new functional assignment. Unfortunately, a nosL mutation in P. stutzeri is phenotypically silent. nosX is required for N2 O utilization by S. meliloti in an as yet unspecified function (134). The gene is also part of the nos region of ¡ÈA. cycloclastes¡É (556). The homolog nirX, with '36% sequence identity, has been found as part of the nir gene region of Paracoccus denitrificans (Fig. 2) (674). The gene is absent from the nos and nir gene clusters of P. stutzeri and was also not picked up by random mutagenesis (877, 972). It is very likely that there will be denitrification genes specific for a distinct species or for groups of denitrifiers; yet it would be premature to pinpoint those candidates from the limited ge-netic information available. The orf2 product of P. aeruginosa and its homologs (the orf175 product and NorE of P. stutzeri and Paracoccus denitri-ficans, respectively) are structurally related to subunit III of COX, and an analogous role is feasible (21). Given the simi-larity of the cytochrome b subunit of NO reductase to subunit I of COX, a function for the orf2 product in NO reduction does not seem unlikely (186, 978). The nirQ gene, located upstream of nirS in the pseudo-monads (and its homolog norQ, located downstream of norB in Paracoccus denitrificans) may code for a maturation or assem-bly function for a denitrification enzyme. Mutations of nirQ or norQ affect both NO reduction and nitrite reduction (186, 431). The immediate target gene(s) or reaction partners of the nirQ product are not known. Because of the sequence similar- ity of NirQ to regulators and because of a putative nucleotide-binding region and the pleiotropic nature of a nirQ mutation, a regulatory role has also been considered. Further nirQ ho-mologs are cbbQ and gvpN, which are part of the genes for C autotrophy of Pseudomonas hydrogenothermophila (944) and gas vesicle formation of Anabaena flos-aquae (454), respec-tively; partial sequence information indicates the presence of a nirQ homolog in Chromatium vinosum also (876). Thus, mem-bers of the NirQ protein family certainly must have varied functions affecting other processes besides denitrification. The presence of several more genes is indicated from un-characterized proteins whose expression is denitrification de-pendent. Periplasmic (35-kDa) and cytoplasmic (32-kDa) pro-teins are induced by nitrous oxide (or nitrate) in Rhodobacter sphaeroides IL106. Neither protein reveals similarity to known denitrification components based on partial amino acid se-quences (707). Under denitrifying conditions, certain nosA mutants of P. stutzeri JM300 express a soluble protein (NosE '70 kDa) of unknown function (588). The nosA gene encodes an outer membrane protein that was recognized from its prop-erty as a phage receptor (150). nirR of P. stutzeri affects nitrite reduction (429). The gene is located outside the denitrification gene cluster and encodes a protein of 25.6 kDa that has no noteworthy similarity to known proteins in data banks (unpub-lished data).
Genetic Organization and Gene Expression
Of the 33 genes in the denitrification cluster of P. stutzeri,23 are transcribed in the same direction; only two groups, com-prising 3 and 7 genes, are transcribed oppositely (Fig. 4). Ten transcriptional units can currently be defined with confidence, but since not all promoters have been mapped, the definitive number will be greater. The nir-nor gene clusters of Paracoccus denitrificans (186) and P. aeruginosa (24) each comprise cur-rently about four or five recognizable transcriptional units. The 11 genes nirS through nirN of P. aeruginosa (Fig. 2) are claimed to be transcribed as a single operon (444). Other presumed operons are the gene sequences nirQ-orf2-orf3 from P. aerugi-nosa and norCBQDEF of Paracoccus denitrificans. Polycis-tronic transcripts have been identified experimentally so far only for norCB and nirSTB of P. stutzeri (319, 968). Several genes of the nir region in the pseudomonads appear to be duplicated. The products of nirF and nirN both have sequence similarity to NirS, including a heme-binding motif in NirN (283, 444). They may be members of a paralogous gene family. Mutational evidence shows that NirF and NirN do not function as nitrite reductases. The nirDLGH products show pairwise similarity indicative of two putative gene duplication events (444, 634). Identification of promoters and the transcriptional organi-zation within the denitrification gene cluster allow first gener-alizations. Transcript initiation sites have been mapped for azu (30); nosR, nosZ, and nosD (171); norCB (968); nirS (319); and cycA (792). Even among this very small number, it was found that certain genes are transcribed from more than one pro-moter. Denitrification is an environmentally regulated process, and the use of multiple promoters may be a means of adjusting gene expression differentially with respect to the oxygen sup-ply, the presence and nature of an N oxide, and perhaps fur-ther external factors such as metal ions. It is anticipated that the promoters of principal denitrification genes are organized specifically for the integration of multiple environmental sig-nals. nosZ is expressed from six promoters; anaerobic expression occurs preferentially from promoter P3, whereas the weak constitutive expression observed in aerobic cells is likely to depend on promoter P2 (171). Cytochrome c551 (the nirM product) and cytochrome c550 (the cycA product) are expressed both under O2 -respiring and denitrifying growth conditions (95, 96, 518), a puzzling fact in light of their electron donor role in denitrification. Rather than resorting to two sets of genes, the existence of two promoters, as shown for cycA (792), pro-vides a satisfactory explanation for the dual expression of these cytochromes. Expression of the gene for cytochrome c2 (the cycA product) from Rhodobacter sphaeroides 2.4.1 is regulated in response to anaerobic photosynthetic and aerobic chemo-heterotrophic conditions from two promoters; it is negatively affected by heme or an intermediate of tetrapyrrole synthesis (727). Strain 2.4.1, although nondenitrifying, has a nitrate reductase and an N2 O reductase (576), of which the latter depends on cytochrome c2 . It is not known whether cycA ex-pression responds to anaerobic nitrate- or N2 O-respiring con-ditions. Anaerobic promoter activity in P. aeruginosa has been de-tected for nirS by a xylE gene fusion (20) and for nirS, nirQ, and norCB by lacZ fusions (22). All upstream regions of the se-quenced nirS and norC genes exhibit binding motifs for an FNR-like factor, which suggests the presence of anaerobically controlled promoter structures. The FNR recognition sites in the promoters of denitrification genes are probably recognized by the members of the FNR family mentioned above (see also the section on regulation, below). Heterologous expression of genes for site-directed mutagen-esis or biotechnological applications will have to reckon with diverse regulatory and maturation requirements of denitrifica-tion components. For instance, a lack of ancillary genes results in nonfunctional cytochrome cd1 and N2 O reductase on het-erologous expression of nirS and nosZ in Pseudomonas putida and E. coli, respectively (282, 767, 878). Genes for the blue copper proteins nitrite reductase (282, 608), azurin (137, 438, 860), and pseudoazurin (144, 928) are expressed as metal-containing holoproteins in E. coli. Either ancillary functions for metalloprotein biogenesis are provided by the host or Cu is incorporated spontaneously. Azurin can be expressed from pUC vectors to comprise '27% of the total protein in the periplasmic space. The material contains an inactive, probably chemically modified variant of azurin lacking Cu-binding prop-erties (860). Only recently was E. coli shown to have a periplas-mic, Cu-containing SOD (41, 64). This opens the possibility that the biosynthesis of a foreign periplasmic Cu protein in E. coli takes advantage of the innate Cu-processing capabilities of the host cell. Expression of nirK in E. coli at 37¡ëC yielded only insoluble material within inclusion bodies. At 27¡ëC, some soluble mate-rial was produced but still the enzyme was not processed cor-rectly and exported, since several immunoreactive peptides showed up on SDS-electrophoresis of the cytoplasmic fraction (488). A construct from which a high-activity enzyme was ex-pressed and exported into the periplasm required truncation of the native export signal by seven amino acids, an increase of the N-terminal charge by 1, and an increase of the overall hydrophobicity of the signal peptide. With this construct the successful expression of native and recombinant nitrite reduc-tase was possible (488, 490). Recombinant NosZ proteins have been expressed in a nosZ deletion mutant from a pSUP vector (973). In contrast to the by now sizable number of products of known denitrification genes, few gene products have been iso-lated. In addition to the reductases, only cytochrome c551 , cytochrome c550 , cytochrome c2 , cytochrome c552 (a diheme protein), azurin, and pseudoazurin have been purified. Gene products involved in a regulatory function, assembly, or bio-synthetic reactions have yet to be isolated and characterized biochemically.
Genome Size and Gene Maps of Denitrifiers
The advances in generating and handling macrorestriction fragments have led to chromosomal maps for several denitri-fying bacteria and to accurate assessments of their genome size. Genomes among denitrifying genera vary within a wide range. The highest and lowest extremes are represented by Bradyrhizobium japonicum (8.7 Mb) (493) and Neisseria gonor-rhoeae (2.2 Mb) (195), respectively. Comparative data are available for pseudomonads to show that genome size varies considerably within this genus. The range extends from 2.6 to 6.8 Mb for Pseudomonas lemoignei and Pseudomonas glathei, respectively (neither species denitrifies). The genomes of P. aeruginosa PAO (5.9 Mb) (356), P. aeruginosa C (6.5 Mb) (731), and P. fluorescens SBW25 (6.63 Mb) (664) are at the upper end of this range. A determination of the genome size of 20 strains of P. stutzeri with the rare-cutting restriction enzyme CeuI, which results in only four DNA fragments, gives a range of 3.7 to 4.6 Mb (278). A value of 4.3 Mb was extrapolated for the ZoBell strain. The most detailed chromosomal map of a denitrifying bac- terium is that of P. aeruginosa PAO, which comprises close to 200 identified markers; among them are about a dozen related to anaerobic nitrate respiration and denitrification (356). Genes for denitrification have been recently mapped to find whether they are organized as an ¡Èanaerobic respiration is-land¡É that could propagate this facultative trait by lateral gene transfer among the prokaryotes (883). Although the genes are concentrated in the 20- to 36-min segment of the chromosome, they are organized as individual clusters harboring genes for nitrite and NO reduction (19.9 to 20.5 min), separate from those for nitrate reduction (27.2 to 28.4 min) and N2 O reduc-tion (33.7 to 35.8 min) (Fig. 5). Only the nir and nor genes are commingled to come close to an anaerobic respiration island. However, it should also be noted that this appears to be limited to denitrifiers that depend on cytochrome cd1 , whereas in the copper nitrite reductase-containing denitrifiers the nir and nor genes are separated (39, 832). The anr gene for the anaerobic global regulator maps independently from denitrification genes at 60 min (264, 883). Further separate loci comprise the genes for the outer membrane protein NosA, the periplasmic nitrate reductase (nap), and azurin (azu). The hemL (372) and hemA (373) genes, necessary for anaerobic heme biosynthesis, map at 0.3 to 0.9 and 24.1 to 26.8 min, respectively. Genes involved in heme D1 biosynthesis, with nirE as the marker gene, are part of the nir locus (Fig. 5). Several more loci dispersed over the chromosome and asso-ciated with nitrate utilization have been mapped by classical techniques. These loci require further analysis to reveal their encoded functions (Fig. 5). Only the loci nir-9006 (21 min) and nar-9001 (30 min) are close enough to the nirS and narGH loci to make identity feasible. The nar-9011 locus affects the reduc-tion of nitrate to nitrite (355); narA, narB, narC, narD, and narE affect anaerobic growth on nitrate (405). For narD, a role in molybdenum transport or processing for nitrate reductase is suggested, since the respective gene defect is suppressed by a high dose of molybdate (867). ntmA and ntmB are loci that affect growth on distinct sources of nitrogen. Functions re-quired for assimilatory nitrate reduction, nasABC, have been mapped at three loci: nasC was suggested to harbor the struc-tural gene for the assimilatory reductase, and nasAB was re-lated to molybdenum cofactor biosynthesis. Genes for the as-similatory nitrate reductase have not yet been isolated from any denitrifier. Chromosome maps with a limited number of markers have been generated for Bacillus cereus (122), Bradyrhizobium ja-ponicum (493), S. meliloti (360), N. gonorrhoeae (195), Neisseria meningitidis (263), P. fluorescens (664), Rhodobacter sphaer-oides (800), and Rhodobacter capsulatus (251) and for the ar-chaeon Haloferax mediterranei (genome size, 2.9 Mb) (17). They provide a basis for locating denitrification genes on these maps. Genome sequencing projects are under way for N. gon-orrhoeae, N. meningitidis, P. aeruginosa, and the archeon Pyro-baculum aerophilum. Eventually, this will allow us to test hypo-theses about horizontal gene transfer and the evolutionary relationship among the denitrifiers. A remarkable case is represented by Rhodobacter sphaer-oides 2.4.1 in its possession of two circular chromosomes (3 and 0.9 Mb) in a 1:1 stoichiometry in addition to several plasmids (801). The distribution of denitrification genes with respect to both chromosomes represents an interesting problem. Strain 2.4.1, although nondenitrifying, has a nitrate reductase and an N2 O reductase that cross-react with antibodies raised against the Rhodobacter sphaeroides IL106 proteins (576). Its nondeni-trifying property is apparently due to an inability to reduce nitrite because the structural gene for nitrite reductase, nirK,is missing (496, 832). The chromosome structure of strain IL106 is unknown. A 108-kb plasmid found in this strain is not thought to be of importance for denitrification (576). Denitrification has a sporadic but recurrent record of being an unstable trait (404, 965), yet a frequent plasmid location of denitrification genes, which could explain this instability, is not the case. The well-studied sources P. aeruginosa, P. stutzeri, and Paracoccus denitrificans have their denitrification genes located chromosomally. The 0.45-Mb plasmid pHG1 for chemolitho-trophic growth of R. eutropha H16 on hydrogen has for a long time been the paradigm for plasmid-encoded denitrification (696, 732). However, even there the principal denitrification genes nar, nirS, and norZ are located on the chromosome (163, 672), whereas nosZ, norB, the nap genes, and certain support-ing functions for anaerobic growth are found on the megaplas-mid (762, 763, 974). On transfer of pHG1 to other hosts, it is maintained stably only in another strain of R. eutropha (732). A second case of plasmid-borne denitrification genes was found with S. meliloti (357), where the 1.4-Mb pNOD plasmid for nitrogen fixation and symbiosis carries a nosRZDFY cluster.
The complete denitrification process leading to N2 forma-tion starts with the nitrate-reducing system. Knowledge about this system in denitrifiers is required to reveal the regulatory factors directed at nitrate reduction per se and those interlac-ing this reaction with denitrification in the strict sense. Several factors of a regulatory network combining the initiator reaction with the main process are just being recognized, and a full experimental penetration of this network is anticipated to be a long-term task. Nitrate respiration has long been studied in E. coli. Given the prevalence of nitrate-respiring organisms sensu stricto and the array of available genetic tools for E. coli, it is not surpris-ing that most of our knowledge about nitrate reductase has come and will continue to come from this nitrate-respiring but nondenitrifying enterobacterium. E. coli ammonifies nitrite with enzymes encoded by the nir and nrf systems but has no coding capability for the assimilatory nitrite reductase, a siro-heme protein like the nirBD-derived dissimilatory nitrite re-ductase. The properties of the respiratory nitrate reductase of E. coli have been reviewed against a biochemical and genetic background (88, 273, 381). Respiratory and assimilatory nitrate reductases both have been amply covered in the context of catalysis by oxo molybdoenzymes (338). Also, the nature and function of molybdenum and iron-sulfur centers in electron transfer by the respiratory and periplasmic nitrate reductases, together with the pertinent biochemistry, have been discussed in detail, often in the context of sequence-derived predictions and hypotheses (69). As this status has not changed signifi-cantly since then, the reader is referred to these comprehensive articles. When a denitrifying bacterium is able to assimilate nitrate, the reaction may proceed simultaneously with nitrate respira-tion (855). Since the first reaction step in assimilatory and respiratory nitrate utilization is identical, the question arises whether just one enzyme would catalyze both reactions. This is not the case. As discussed in the preceding section, different genes, which are distributed over several loci on the chromo-somal map of P. aeruginosa, exist for the two processes (Fig. 5). Nevertheless, in spite of the distinct genetic basis for respira-tory and assimilatory nitrate reduction, the respiratory reduc-tase was found to serve a quasi-assimilatory function under anaerobic conditions and to provide nitrite in nas mutants as the substrate for nitrogen assimilation (760). This function ceases under aerobic conditions. Gene sharing between the assimilatory and the respiratory pathway may involve only a few functions, possibly related to molybdenum cofactor syn-thesis or to nitrate transport (287). Since Pichinoty¡Çs pioneering contributions on bacterial ni-trate metabolism, the generally held view is that of a mem-brane- bound respiratory reductase and of a soluble, pyridine nucleotide-dependent variant confined to the assimilatory branch of nitrate reduction. This generalization has to be mod-ified and broadened to accommodate a new type of dissimila-tory nitrate reductase. The finding by Satoh of a novel nitrate reductase activity located in the periplasm of R. sphaeroides IL106 and associated with a c-type cytochrome (715, 719) was initially rationalized in terms of peculiarities of phototrophic metabolism (554) but is now seen as a new more general aspect of dissimilatory nitrate metabolism not restricted to pho-totrophs. Studies from several laboratories have provided ev-idence for a broad distribution of this third type of nitrate reductase encoded by the nap genes.
Properties of Dissimilatory Nitrate Reductases
Many bacteria have more than one of the three types of nitrate reductases which comprise the soluble assimilatory-type nitrate reductase and two dissimilatory reductases, subdivided into the respiratory and the periplasmic nitrate reductases. Evidence for the coexistence of all three nitrate reductases has been provided for R. eutropha and Paracoccus denitrificans by mutational studies, gene sequencing, and, in part, isolation of these enzymes (740, 892). Purification and biochemical studies of membrane-bound respiratory nitrate reductase from denitrifiers have been di-rected at various sources, for which representative data are shown in Table 3. In addition, nitrate reductases at various stages of characterization have been isolated from the denitri-fiers Halomonas halodenitrificans, Thiobacillus denitrificans, Rhodobacter sphaeroides IL106, Bacillus licheniformis, and Bacillus stearothermophilus (348). Some of these nitrate reduc-tases are susceptible to proteolytic modification during isola-tion (84, 386), similar to the E. coli enzyme (191). Respiratory nitrate reductases are complexes of two or three subunits depending on the method of isolation. The enzyme is anchored to the cytoplasmic membrane by the NarI or g sub-unit and has to be solubilized by heat or detergent. The respi-ratory reductase is encoded by the narGHJI operon. narG encodes the large or a subunit of 112 to 145 kDa. This subunit carries molybdenum in the form of the MGD cofactor (255, 256). In the enzyme as isolated, molybdenum is detectable as a Mo(V) species via EPR by a series of resonances around g '1.9 (Fig. 6; Table 3). The molybdenum cofactor is the active site of the reductase (338). Nitrate reductases solubilized by heat show the featureless spectrum of an Fe-S protein with a small peak or a shoulder around 400 to 415 nm, which disappears on reduction (Fig. 6). The narH gene of E. coli encodes the small or b subunit (58 kDa) which binds three [4Fe-4S] clusters and one [3Fe-4S] cluster (79, 308). These clusters have markedly different redox potentials. Clusters 1 and 2 have potentials of 180 and 160 mV, respectively, and clusters 3 and 4 have redox potentials of 2200 and 2400 mV, respectively. Cluster 2 is the 3Fe center. The ligation of these clusters by cysteine residues of the b subunit has been probed by a combination of site-directed mutagenesis, EPR spectroscopy, and redox potentiometry (309). The cysteines are positioned in the protein in four groups. Alternative models have been developed where the Fe-S clusters are arranged pairwise in centers 1 and 4 and in centers 3 and 2 or in a supercluster arrangement (309). In each case, a high-potential cluster and a low-potential cluster are juxtaposed. Cys3Ala substitutions of the putative ligands of clusters 2 and 3 result in the loss of the Fe-S centers and additionally also the Mo cofactor, showing that these centers play a strong structural role. Removal of cluster 1 or cluster 4 does not affect the remaining three clusters either spectroscop-ically or in their redox properties; therefore, clusters 1 and 4 are believed to be structurally less important. Nitrate reductases isolated in detergent carry a b-type cyto-chrome, as seen in the form of a third or g subunit (19 to 23 kDa), encoded by the narI gene (Table 3). Such heterotrimeric enzymes exhibit the electronic spectrum and absorbance inten-sities of a diheme protein. The NarI function is ascribed to quinol oxidation and electron transport to the b subunit. The topology of the g subunit of the Paracoccus denitrificans en-zyme has been predicted by analogy to the NarI subunit from E. coli (192) as a transmembrane anchor that immobilizes the complex of the a and b subunits at the cytoplasmic side of the membrane (70). Heme ligation was proposed to be either by a single g subunit or, based on known heme ligation of b-type cytochromes, by two equivalent subunits involving identical transmembrane helices (865). The minimal subunit stoichiom-etry is suggested to be heterotrimeric. The ab complex of the heat-released enzyme of E. coli has menaquinone-9 as a further constituent in a stoichiometry of one quinone molecule per protein dimer (103). The intramo-lecular electron transfer in nitrate reductase is thought to in-volve sequentially the two hemes of the g subunit, the two Fe-S centers of the b subunit, and the molybdenum cofactor. The quinone is proposed to transfer electrons between the subunits or between the Fe-S clusters within the b subunit. In a thus far singular case, the respiratory nitrate reductase of E. coli encoded by the narGHJI operon, enzyme A, is du-plicated in the narZYWV operon encoding a similar enzyme Z (75% sequence identity between the a subunits) (80, 385). The subunits of the two enzymes are interchangeable and hence can form enzymatically active hybrid species such as a A b Z g Z or a Z b AgA (81). The purified a A b Z enzyme is less stable and shows a somewhat decreased activity and looser membrane association. Preliminary evidence has been provided for two nitrate-reducing activities in B. japonicum whose underlying molecular entities are not clear (245). It should be noted that a nitrate-reducing activity can be associated in certain bacteria with another physiological activity, as is the case for the high-molecular- mass flavoheme protein nitrilotriacetate dehydroge-nase (401). Whereas the membrane-bound respiratory nitrate reductase is expressed only under anaerobic growth conditions, the periplasmic nitrate reductase is synthesized and active in the presence of oxygen (58, 762). Both enzymes are under nitrate control exerted via the sensor protein NarX or NarQ (see the section on regulation, below). In contrast to the membrane-bound nitrate reductase, the periplasmic enzyme does not re-duce chlorate and nap mutants cannot be selected for by chlor-ate resistance (58). The physiological role of the periplasmic nitrate reductase is thought to consist of dissipating excess reducing power and providing nitrite for aerobic denitrifica-tion. In doing so, the enzyme may also function in the transi-tion from aerobiosis to anaerobiosis. Since oxygen is believed to inhibit nitrate transport (335, 612), the periplasmic form of the enzyme could foster the transition to anaerobiosis. Demonstration of the coexistence of the three types of ni-trate reductases was achieved first with R. eutropha by physio-logical and mutational means (892). The periplasmic nitrate reductase of this organism was purified, and its structural genes were identified and sequenced (762). The enzyme was shown to be plasmid encoded and to consist of two subunits, NapA (93.3 kDa) and NapB (18.9 kDa) (Table 3). Both subunits are synthesized with signal peptides conforming to the periplasmic location of the mature gene products. The enzyme was later also isolated from a narH mutant of Paracoccus denitrificans GB17, and the analysis of the nap locus was extended to a napEDABC cluster (71). The NapA subunit binds the molybdenum cofactor; a four- cysteine motif near the N terminus was proposed to bind a [4Fe-4S] cluster (101). NapA shows sequence relatedness to the respiratory nitrate reductase of E. coli and other molyb-doenzymes such as the assimilatory nitrate reductase, formate dehydrogenase, and DMSO reductase of various organisms. Binding sites for Fe-S clusters and Mo were proposed from comparative sequence analysis (762). The sequence alignment of NapA with enzymes that bind MGD revealed two conserved regions that may be relevant for substrate specificity and con-tribute to the active site (71). The small subunit NapB does not show similarities to other enzymes, but two potential heme C-binding sites are present in the sequence. Additional histidine residues are thought to ef-fect a bis-histidine coordination of the hemes. The redox po-tentials of these hemes differ by about 100 mV (180 and 215 mV) (72). NapC belongs to a homologous family of tetraheme c-type cytochromes first reported as NirT of P. stutzeri (430). The putative role of NapC is in the electron transfer between a quinol and the periplasmic nitrate reductase. Homologs of napC exist in R. eutropha in the sequence reported adjacent to the napAB locus (762), in the E. coli napFDAGHBC locus linked to the genes for cytochrome c biogenesis (303), in the Haemophilus influenzae ORF HI0348 as part of a nap locus (250), and in Rhodobacter sphaeroides (679). The functions of NapD and NapE for the NapAB complex are unknown, as are the functions of the napGH products of E. coli, which are predicted to be Fe-S proteins. Sequence-derived considerations place NapE in the membrane and NapD in the cytoplasm. Periplasmic nitrate reductases have been studied in parallel in two strains of Paracoccus denitrificans. The activity and ex-pression of periplasmic nitrate reductase in P. denitrificans PD1222 was demonstrated by physiological means (739) as well as by detecting a Mo(V) EPR signal in whole cells (738). The enzyme from P. denitrificans GB17 was subjected to bio-chemical and spectroscopic studies of its molybdenum (58, 61) and Fe-S (101) centers. Molybdenum coordination in the oxi-dized and reduced enzymes is different from that observed for the respiratory nitrate reductase. However, it resembles mo-lybdenum coordination in assimilatory nitrate reductase, lead-ing to the suggestion that these two soluble enzymes have similar Mo coordination spheres (62).
Molybdenum Cofactor
The hypothesis of an organic Mo cofactor as part of molyb-denum- containing enzymes arose from the finding of pleiotro-pic effects on fungal assimilatory nitrate reductase and xan-thine dehydrogenase. This concept gained momentum after interchangeability of low-molecular-mass compounds among molybdoenzymes had been demonstrated. A pivotal role in the experimental development of the field was played by the mu-tant nit-1 of Neurospora crassa, of which a defective nitrate reductase could be reconstituted by cofactor donors. The prin-cipal findings until 1990 have been summarized (340). Molyb-doenzymes carry the metal as part of a pterin cofactor, with the single exception of nitrogenase, where Mo is part of a polynu-clear iron-sulfur cluster. Rajagopalan and coworkers have pi-oneered the elucidation of the cofactor molecule and inferred its structure from urothionine (667). The molecule, MPT (415), does not carry Mo, even though its name suggests oth-erwise, and the same cofactor molecule is found in tungsten-containing enzymes (131). It has also been pointed out that MPT does not fulfill the role of a true cofactor since it does not dissociate from the enzyme but, rather, is the prosthetic group of the respective molybdo- or tungstoenzymes (340). The basic structure of the cofactor of the eukaryotic molyb-doenzymes is a 6-alkyl pterin derivative with a phosphorylated, four-carbon side chain (Fig. 7) (470). In its active cofactor form, MPT is complexed with Mo by the sulfur atoms of the dithiolene configuration at the 6-alkyl side chain. The cofactor molecule of prokaryotic molybdoenzymes is composed of the pterin moiety and an additional group, lending a higher mass to the cofactor. The name ¡Èbactopterin¡É had been proposed initially for the modified bacterial Mo cofactor (570). Evidence for such a cofactor was provided for the respiratory nitrate reductase of P. stutzeri (480). The name ¡Èbactopterin¡É was not upheld when subsequent work with Mo enzymes revealed the existence of several cofactors distinguished by their nucleoside moiety. DMSO reductase from the denitrifier Rhodobacter sphaer-oides IL106 played a decisive role in unraveling the chemical nature of the bacterial cofactor as MGD (Fig. 7) (412). The cofactor structure was subsequently proven from the crystal structure of this enzyme (728). However, the first crystal struc-ture of an MPT cofactor-containing protein, the tungstoen-zyme aldehyde ferredoxin oxidreductase from Pyrococcus fu-riosus (131), unexpectedly revealed a three-membered ring system with a pyran ring and not an open configuration as thought previously. The pyran ring was also found in the co-factor of DMSO reductase. It is possible that ring closure is part of a redox process taking place on the enzyme-bound cofactor (222). MGD is found in a variety of molybdoenzymes, among them formate dehydrogenases of both bacteria and archaea (400, 413), formylmethanofuran dehydrogenase of the methanogens (94, 439), and polysulfide reductase (400). Shortly after gua-nine was found, cytosine, adenine, or hypoxanthine was de-scribed for diverse prokaryotic Mo enzymes, of which some even seem to have more than one type of cofactor within the same protein molecule (94, 418, 667). MGD is the cofactor of the respiratory nitrate reductases from the denitrifiers P. carboxydoflava and P. stutzeri (255, 256), the same cofactor as in the respiratory nitrate reductase from E. coli (416). Also, the periplasmic nitrate reductase of Rhodobacter sphaeroides is among the MGD-containing en-zymes (667). A fluorescent compound, probably a pterin, has been extracted from the homologous enzyme of Paracoccus denitrificans GB17 (72), and since sequence comparison re-veals similarity to other MGD-containing enzymes, NapA is considered to carry an MGD cofactor (69, 338). In xanthine dehydrogenase of P. aeruginosa, only MPT, not the dinucleotide form, was found, which represents an excep-tional case for bacterial molybdenum cofactors (414). How-ever, MGD is present in cell extracts, as evidenced from the extraction of its oxidized, fluorescent derivative (427). Al-though the source of this MGD was not established and cells were not grown denitrifying, P. aeruginosa has the capability to synthesize MGD, and this cofactor is anticipated as part of its nitrate reductase. The evidence for MGD supports previous findings of restoration of nitrate reductase activity by mixing extracts of P. aeruginosa and a cofactor-deficient mobB strain of E. coli (713). DMSO reductase of R. sphaeroides IL106 (a monomeric enzyme of 86.5 kDa with no other prosthetic group but the Mo cofactor) was shown to bind a single Mo atom via two MDG molecules (339). The crystal structure confirmed this stoichi-ometry and revealed that in the oxidized enzyme, two MDG molecules donate four sulfur ligands from the dithiolene groups to a six-coordinate Mo atom (728). The cofactors are extended and oriented in a mirrored and rotated opposite position. They differ by a 10¡ë bend at the pyran ring and hence provide a slightly asymmetric environment. An oxo group and serine provide additional ligands for a distorted trigonal pris-matic geometry around the Mo atom. The binding of Mo by MGD appears not to be uniform in MGD-containing enzymes. Although DMSO reductase of Rhodobacter capsulatus is a bis-MGD enzyme, only one MGD molecule complexes Mo, the other being too far removed from the metal site (733). In the pyrogallol transhydroxylase of the strict anaerobe Pelobacter acidigallici, the stoichiometry suggests a single Mo atom com-plexed between a heterodimer, with each subunit carrying an MGD molecule (673). Based on sequence relatedness, dissimilatory nitrate reduc-tases have been placed in the DMSO reductase family (69, 338). In respiratory nitrate reductase, three or four S ligands and a long MoOO distance were detected by X-ray absorption to suggest a ligation similar to that of DMSO reductase (161, 275). On the other hand, the stoichiometry of MGD and Mo determined for nitrate reductase is 1:1, which implies that the additional S ligands are not from a second MGD molecule (255). Since it was initially not anticipated that one Mo atom can be complexed by two cofactor molecules in a molybdoen-zyme, the stoichiometry of the cofactor and metal and the nature of the proximal Mo ligands of respiratory nitrate reduc-tase warrant additional investigations. The role of the cofactor in the molybdoenzymes is thought to modulate the redox be-havior of the metal and aid in the electron transfer from or to other redox centers without the pterin undergoing a redox process itself (338). The biosynthesis of the Mo cofactor in E. coli has been intensively investigated and provided the principal picture shown in Fig. 8. It is hypothesized that the cofactor synthesis proceeds along the same basic pathway in other nitrate respir-ers including the denitrifying bacteria. Five distinct regions on the E. coli chromosome, moa (chlA), moe (chlE), mob (chlB), mod (chlD), and mog (chlG) participate in establishing a func-tional molybdoenzyme. They comprise loci necessary for MPT biosynthesis and Mo uptake. The former chl designations, re-sulting from chlorate-resistant mutants with pleiotropic effects in molybdoenzyme activities, have been replaced by combining the mo symbol with the old chl allele (741). MPT synthesis starts from a phosphorylated guanosine spe-cies (Fig. 8). Its synthesis resembles that of other pterins and riboflavin insofar as a guanosine derivative (usually GTP) serves as the template whose carbon atoms are incorporated into the pterin precursor. Whereas in the cyclohydrolase reac-tion for riboflavin and folate synthesis the C-8 atom of guanine is eliminated and does not appear in the products, the C-8 position is converted to the first C atom of the alkyl side chain of MPT. A novel route of biosynthesis is thus suggested for the Mo cofactor (923). In the early steps of MPT synthesis, the products of the moa and moe loci are required (419, 688). moaE and moeB mutants accumulate the desulfo-MPT precursor Z (Fig. 7 and 8). In this precursor, the terminal phosphate is in a diester linkage to C-29 and C-49 of the side chain, forming a six-membered ring (922). Other than in riboflavin biosynthesis, where the ribose of GTP forms the ribityl side chain, both the ribose and the ring car-bons of guanine are incorporated into the precursor Z of MPT. The moa locus of E. coli consists of a putative five-gene operon, moaABCDE (688). The MoaA protein (39 kDa) has been purified from Arthrobacter nicotinovorans. It is an Fe-S protein with a [3Fe-3S?] cluster whose coordinating cysteine residues are located in both the N- and the C-terminal regions. The candidate ligands, Cys32, Cys39, Cys277, and Cys294, have been identified by site-directed mutagenesis (564). The protein is suggested to act as an oxidoreductase or as the sulfur donor to MPT synthase. The N-terminal region of MoaA with the cysteine cluster shows some similarity to the NifB protein, which is involved in the FeMo cofactor synthesis of nitrogenase (563, 688). For a discussion of the similarity of MoaA to the NirJ and Pqq proteins, see the section on heme D1 biosynthe-sis, below. The products of the moaD and moaE genes (8.8 and 16.9 kDa, respectively) are the two subunits of the MPT synthase (formerly the converting factor). The protein has been puri-fied, but since the subunits are not tightly bound, the degree of oligomerization is not known (650). The function of the syn-thase is the introduction of the dithiolene sulfurs into precur-sor Z (Fig. 8). The sulfur is provided from an unknown sulfur source to the MoeB protein, from where it is transferred to the small subunit of the synthase. The presence of reactive sulfur is traceable to MoaE from its reactivity with iodoacetamide. In the following steps of cofactor synthesis, the products of the mod and mog loci are involved. The mod (chlD) locus encodes components used for molybdenum uptake (281) and is regulated by the availability of molybdate in the medium (580). Molybdenum transport and processing systems must be inti-mately associated with the biosynthesis of the denitrification apparatus to provide the metal to the respiratory and periplas-mic nitrate reductases. The narD locus of P. aeruginosa may correspond to chlD, since a mutation is rescued by exogenous molybdate (867). It is likely that the molybdenum uptake pro-cess in diazotrophic denitrifiers is shared for the two different types of molybdenum cofactors before more specific functions take over to yield either one. Transport studies have been undertaken with Bradyrhizobium japonicum (529) and Azospi-rillum brasilense (141), which are both diazotrophic denitrifiers. Bacteroids of B. japonicum have a high-affinity uptake system (Km , 0.1 mM) that is anaerobically more active. Molybdate uptake is thought to occur by proton symport since the system is sensitive to uncouplers and ionophores. Molybdate uptake is competitively inhibited by tungstate. Radioactive [ 99 Mo]mo-lybdate is very strongly bound and appears to be immediately channeled to intermediate binding proteins or to the respective target enzyme. A 100-fold excess of unlabeled molybdate is not able to exchange Mo even after only 1 min of uptake of the latter (529). The modABCD operons encoding molybdenum transporters have been unraveled in Azotobacter vinelandii (525), Rhodo-bacter capsulatus (890), and E. coli (551, 670). Homologous genes were also detected in the genome of H. influenzae (250). The uptake system of E. coli is an energy-dependent high-affinity system with a Km of 25 to 27 nM. It depends on a peri-plasmic binding protein with a low dissociation constant for molybdate (520). mod mutants are affected in molybdate up-take (330, 520, 701). A high exogenous molybdate concentra-tion can rescue a mod mutation, with molybdate uptake then proceeding via sulfate or selenate transporters with a lower efficiency (701). A partial sequence of the transport operon, consisting of modC and the upstream and downstream flanking genes, re-vealed that modC encodes a 39-kDa protein with similarity to the ATPases of the high-affinity ABC transporters (410). Members of the family of ABC transporters export and import a wide range of substances across membranes (336). Their architecture consists of variations around a basic model of a periplasmic substrate-binding protein, the membrane-bound transporter, and a cytoplasmic ATPase. ABC transporters are also relevant to other aspects of denitrification. ModA (chlX) is the periplasmic binding protein (26.4 kDa) for the molybdate transporter. It has been overexpressed and purified, which allowed its biochemical characterization. On binding of molybdate, ModA undergoes alterations in its elec-trophoretic mobility and exhibits a new absorption maximum at 287 nm. Purified ModA binds molybdate with a Kd of 3 mM (671); a previous estimate obtained with unfractionated periplasmic material gave a value of only 9 nM (520). It is not clear whether the discrepancy of the determinations is due to the differences in materials tested. Cells grown in medium containing ,10 nM molybdate still synthesize appreciable lev- els of cofactor and nitrate reductase, which implies a high affinity of the transporter to molybdate (736). modB (chlJ) encodes a hydrophobic transmembrane protein, considered to be the transmembrane carrier for molybdate. The role of modD is still unclear; at least the gene does not appear to be essential. The mod operon is under the control of the ModE (5 ModR), repressor whose gene is located immediately up-stream of and transcribed oppositely to modA (304, 887). In the presence of molybdate, ModE binds preferentially to a 9-bp palindromic recognition sequence, CGTTATATA-N4–12 -TATATAACG, at the modA promoter from 218 to 110 rel-ative to the start of transcription (14, 557). The protein acts as a dimer and binds two molecules of molybdate with a Kd of 0.8 mM. Homologs of modE, but no corresponding regulatory regions, are present in the mod region of the diazotrophs Azotobacter vinelandii and Rhodobacter capsulatus B10. Instead of a single modE homolog, the region upstream of modA of R. capsulatus harbors the divergently transcribed mopA (34% identity to E. coli ModE) and mopB genes, which are both suggested to encode pterin-binding proteins (890). It is not known whether R. capsulatus B10 reflects the situation of deni-trifying members of the genus Rhodobacter and other nonpho-tosynthetic denitrifiers, of which none have yet been investi-gated for their molybdate uptake system. The chelatase for Mo insertion into the cofactor is thought to be encoded by the mog locus (Fig. 8). If MPT is not loaded with the metal or if the cofactor is not inserted into an acceptor enzyme, it is rapidly degraded, such as in mog and mod mutants (426). The mob genes are involved in a late step of MGD synthesis (416). The mob locus comprises mobA and mobB, whose ex-pression is constitutive at a low level (384). The inactive solu-ble precursor of nitrate reductase from mob mutants already contains MPT. The in vitro activation of purified mob nitrate reductase depends on the mobA product, MgGTP, and a fur-ther activity ascribed to a factor X (712). A mobB mutant has no recognized phenotype, but overexpression of mobB en-hances factor X activity (635). The gene-deduced protein has a putative GTP-binding site to suggest a role for MobB in MGD synthesis, yet its function appears to be redundant, since a mobB deletion mutant continues to synthesizes active nitrate reductase, trimethylamine N-oxide reductase, and formate de-hydrogenase. All three enzymes contain MGD (635). The pro-tein has been purified from an overexpressing strain. It is a dimeric molecule (subunit mass, 19.5 kDa) that binds '0.8 mol of GTP/mol with a Kd of 2 mM (229). Its properties are com-patible with a role of binding the guanine nucleotide that becomes incorporated into MGD. mobA encodes the FA pro-tein (association factor), which has also been purified to ho-mogeneity. The factor is a low-molecular-mass monomeric protein (21.6 kDa) and activates inactive molybdoenzymes of mob mutants (229, 636). There is a limited sequence similarity of 80 C-terminal residues of MobA to the regions of NarG and NarZ of E. coli, which are suggested to bind the cofactor. MobA may therefore interact with the cofactor and attach the nucleotide moiety (416). Cofactor synthesis must incorporate an element for discrim-ination among different types of Mo cofactors present in deni-trifying bacteria. MGD is the cofactor of Hydrogenophaga (for-merly Pseudomonas) pseudoflava nitrate reductase, but MPT cytosine dinucleotide is the cofactor of the CO dehydrogenase of the same organism (569). MPT and MGD coexist in P. aeruginosa (427) and have to be provided for xanthine de-hydrogenase and nitrate reductase, respectively. Specificity could be part of the insertion process or could reside in dif- ferent mob-encoded proteins. The sequence of the late events has not been explicitly established. Charging of MPT with molybdenum, insertion into a target protein, and addition of the nucleotide moiety can be deduced from the phenotypes of mog, mod, and mob mutants. Recently it has been found that NarJ, encoded within the structural operon for nitrate reduc-tase, is part of factor X activity and renders the maturation process of the mob-encoded proteins specific for nitrate reduc-tase (635).
Transport of Nitrate and Nitrite
The catalytic site of nitrate reductase is oriented toward the cytoplasm and generates nitrite at the inner face of the mem-brane (411, 424, 471). In contrast, nitrite reductases are periplasmic enzymes in gram-negative bacteria. With the sites of reduction of the two oxyanions at opposite faces of the membrane, transport systems for nitrate and nitrite are re-quired. As yet, the prokaryotic nitrate/nitrite transporter has not been identified at the molecular level in any denitrifying bacterium. An antiport system is thought to control the move-ments of nitrate and nitrite across the membrane during ni-trate respiration (93, 180). Since no change in net charge af-fecting the membrane potential is associated with the counter movement of these anions, an antiporter is energetically the most favorable proposition. A nitrate/proton symport, driven by the energized membrane, has been proposed, both as the principal uptake process and to initiate nitrate uptake prior to the functioning of the antiport system (93, 473). Recent data confirm an energy requirement for nitrate uptake that is sen-sitive to protonophores and may favor a symport mechanism (482, 706, 917). However, there are also contradictory findings (334, 639), and a passive nitrate-specific pore has been postu-lated to account for the lack of energy requirement and missing unequivocal evidence for an antiporter. Perhaps the most ef-ficient nitrate uptake system of a denitrifier has been found in the vacuolated sulfide- and elemental sulfur-oxidizing bacte-rium Thioploca sp., where nitrate is concentrated from 25 mM in seawater to 0.5 M inside the cell (253). Transport systems for nitrate and nitrite are better known outside the denitrifying bacteria. In E. coli, a nitrite exporter (the narK product) is encoded as part of the narLXKGHJI gene cluster for respiratory nitrate reduction (80, 611). Homologs of NarK are anticipated in denitrifiers, since they are also found in the gram-positive nitrate respirers Bacillus subtilis (168) and Staphylococcus carnosus (237). The 12-span membrane-bound NarK protein of E. coli was initially thought to be a nitrate transporter since deletion of narK affected nitrate uptake. It was later believed to constitute the NO3 2 /NO2 2 antiporter (193). A more recent investigation making use of membrane vesicles (rather than intact cells) and of sensitive techniques (the use of [ 13 N]nitrate and quenching of a fluorescent dye by NO2) suggests that the physiological role of NarK is that of a nitrite exporter and leaves the search for the proper nitrate uptake system open again (706). Sequence analysis indicates that NarK belongs to the major facilitator superfamily of trans-porters. This family has a broad substrate spectrum that in-cludes H, sugars, and antibiotics; its members share 12 mem-brane- spanning segments and several sequence motifs (535, 837). NarK is thus related to the nitrate transporter and bispe-cific nitrate/nitrite transporter of eukaryotic microorganisms of this family (265, 838, 846). A different type of nitrate transporter which belongs to the ABC family of transporters has been identified in cyanobacte-ria and heterotrophic nitrate-assimilating bacteria and has been partially characterized biochemically (515, 622, 637). This transport system is repressed and immediately inhibited by ammonia (76). The cyanobacterial and enterobacterial uptake systems are homologous. NrtA (homologous to NasF) repre-sents the presumed periplasmic binding protein for nitrate, which is membrane inserted in Synechococcus sp. strain PCC 7942. NrtB (homologous to NasE) is the membrane-bound transporter, and NrtD (homologous to NasD) is the cytoplas-mic ATPase. The synechococcal transporter has an additional cytoplasmic component not found in Klebsiella pneumoniae (622). Besides nitrate, the Nrt system also transports nitrite (526).
Biological Redundancy in Nitrite Reductases
For the reduction of nitrite, one finds in denitrifying bacte-ria, although never within the same cell, two entirely different enzymes in terms of structure and the prosthetic metal. About three-quarters of strains collected worldwide, with a preva-lence of pseudomonads among them, have the tetraheme pro-tein cytochrome cd1 as the respiratory nitrite reductase (266). The same reaction is catalyzed by a CuNIR in a greater variety of physiological groups and bacteria from different habitats. In 23 strains from culture collections, a slight predominance of CuNIR was found, whereas the numerically dominant isolates and denitrifiers from aerobic soil enrichments have mostly cytochrome cd1 (158). Whether the numerical prevalence of strains with cytochrome cd1 reflects a real dominance over CuNIR or only a preferential isolation of such denitrifiers is unknown. Methods for assessing the full denitrification poten-tial of soil and water samples are still under development. Improvements of DNA probes recognizing denitrification genes and the definition of diagnostic sequences is necessary to reach meaningful data for natural, noncultured assemblages. Eventually, this should provide a detailed picture of the true denitrification potential and its underlying bacterial diversity of natural samples. Table 4 lists the sources of the two types of nitrite reductases for which an unequivocal identification by partial purification and spectral evidence has been reported. Within the Proteo-bacteria, neither CuNIR nor cytochrome cd1 is found in exclu-sive association with a particular subclass, but both types are found in each one of the alpha, beta, and gamma subclasses (965). The genetic information for either nitrite-reducing sys-tem coexists occasionally at the genus level but, as far as we currently know, not at the species level (Table 4). The distri-bution of the two types of enzymes has been studied by DNA-DNA hybridization with the structural genes nirS (774) and nirU (5 nirK) (941) and by cross-reaction with an antiserum against CuNIR (158). In most instances, the immunochemical data agree quite well with those from DNA hybridization. Since cytochrome cd1 is not inhibited by DDC, a preliminary diagnosis of whether a new isolate possesses a CuNIR can be made by inhibiting nitrite reduction in cell extract with this chelator (743). However, azurin and pseudoazurin are among the electron donors for cytochrome cd1 and are also affected by DDC. A further complication is the effect of this chelator on electron transfer components of the respiratory chain (481). The use of size-fractionated soluble material to ensure the removal of the low-molecular-weight Cu proteins and mem-brane vesicles and the use of an artificial electron donor system make this test less prone to error. Thus, in the case of Para-coccus denitrificans GB17, the initial conclusion of a CuNIR based on the use of DDC did not survive closer scrutiny (585). Preliminary evidence for a CuNIR from inhibition by DDC has been reported for Bacillus firmus (751), B. stearothermophilus (346), Bradyrhizobium japonicum (158), Nitrobacter vulgaris (6), and Rhodopseudomonas palustris (658). The interchangeability of a CuNIR with cytochrome cd1 has been studied by using nirK from P. aureofaciens and nirS from P. stutzeri (282). CuNIR was found to be active in a mutation-ally cytochrome cd1 -free background of P. stutzeri and restored the interrupted denitrification activity. In contrast, expression of nirS in P. aureofaciens resulted in an inactive enzyme species because the requirements for heme D1 biosynthesis are not met by the host organism. The capability of P. stutzeri to ex-press nirK functionally in vivo is remarkable in light of the electron donor specificity of the enzyme. NirK activity of P. au-reofaciens is dependent on its indigenous azurin (977), but a cupredoxin is not present in P. stutzeri. Thus, electron donation to heterologous NirK must involve a foreign electron carrier, in all likelihood a c-type cytochrome taking advantage of a low recognition specificity of the reductase. The underlying phe-nomenon of the interchangeability of electron donors for ni-trite reductases is discussed below in more detail in the section on electron donation.
Cu-Containing Nitrite Reductase
Enzyme properties. On the basis of their optical and EPR properties, Cu(II) centers in proteins are assigned to one of three classes, type 1 (blue) Cu, type 2 (nonblue) Cu, and type 3 (binuclear and EPR-inactive) Cu (530). The first nitrite re-ductase with Cu as prosthetic metal was discovered in Mori¡Çs laboratory (393). The basic properties established then were evidence for Cu from chemical analysis, restoration of catalytic activity of a metal-depleted enzyme by Cu, and inhibition of the enzyme by DDC. The enzyme, although not difficult to purify and in spite of its importance in the N cycle, did not enter the mainstream of Cu-protein research for many years. This has changed dramatically since. The crystal structure of CuNIR, the first of any denitrification enzyme (284), was a breakthrough and motivated subsequent studies on the reac-tion mechanism, electron transfer, and the interaction of elec-tron carriers with their cognate reductases. A survey of the enzymes purified from various sources re-veals apparent differences in absorbance properties (blue ver-sus green proteins), molecular masses, the subunit composition of the holoenzymes, and catalytic activities (Table 5). A seem-ing heterogeneity among the CuNIR species had been a gen-eral feature until recently and contrasted with the more uni-form family of cytochrome cd1 nitrite reductases. Evidence from three X-ray structures and four gene-derived amino acid sequences clearly reveal CuNIR species to be members of the same protein family. Crystal structures and X-ray scattering have also firmly established that CuNIR species are trimeric enzymes (208, 284, 301, 323, 488). The size of subunits determined by SDS-electrophoresis is close to 40 kDa in each case, independent of the suggested quaternary structure (Table 5). Some CuNIR species are pro-teolytically sensitive. Three peptides were observed in Bacillus halodenitrificans (198). The Rhodobacter sphaeroides IL106 en-zyme yields a single subunit or two different species, depending on the purification procedure (575). As far as sequences are known, all nitrite reductases have a related primary structure of a single-type subunit with a positional amino acid identity of 61 to 81%. The first amino acid sequence was determined from overlapping proteolytic peptides of CuNIR from ¡ÈA. cyclo-clastes¡É (244). Homologous structures were deduced from the nitrite reductase genes of strain G-179 of Pseudomonas sp. (941), Alcaligenes faecalis S-6 (608) and P. aureofaciens (282). The most distant proteins obtained from P. aureofaciens and ¡ÈA. cycloclastes¡É still have 61% sequence identity. The Cu-coordinating peptides are identical in the four enzymes with respect to their general location in the protein and the sur-rounding regions of conserved amino acids (Fig. 9). The presence of type 1 and type 2 Cu is manifest in the small and large hyperfine splitting in the EPR spectrum of nearly all nitrite reductases (Fig. 10; Table 5). The analytically deter-mined amount of Cu, however, is smaller than the expected number of six Cu atoms (Table 5). Often, only half of the theoretically amount has been detected by chemical means. Close to five Cu atoms per trimer are present in the ¡ÈA. cycloclastes¡É enzyme when the reported number is corrected for a mass of 109.4 kDa (516). A metal-depleted enzyme was reconstituted with Cu(II), and 5.3 to 5.6 Cu atoms were found by colorimetry and EPR spectroscopy (514). A lower than stoichiometric content of Cu may be due to the possibility that part of the Cu sites is occupied by Zn (1). The hydrodynamic properties of nitrite reductase indicate rapid dissociation-reassociation equilibria (284, 301). With a monomeric species forming part of these equilibria, loss of Cu may result during enzyme purification and explain why in two instances a protein with only the type 1 Cu in place has been isolated (539, 977). A 20-fold loss of enzyme activity is ob-served on cell breakage with the Alcaligenes xylosoxidans en-zyme, which is counteracted by incubating the crude extract with 1 mM CuSO4 (1). Spontaneous depletion of the type 2 Cu occurs on storing the enzyme from ¡ÈA. cycloclastes¡Éfor 2 to 3 days at 4¡ëC before ammonium sulfate precipitation, an effect that has been used to demonstrate by metal reconstitution the requirement of type 2 Cu for CuNIR catalysis (514). The Alcaligenes xylosoxidans enzyme is activated by a factor of up to 40 by freeze-thawing (538). Slow freezing may cause a rearrangement of the Cu centers and stimulate enzyme activity by filling part of the type 2 center. Unfortunately, the EPR characteristics of the material before and after treatment were not reported. Repeated chromatographic passage over hy-droxylapatite and TSK column material produces type 2 Cu signals in the EPR spectrum of the P. aureofaciens enzyme that were absent in the original material (977). Since type 1 Cu is usually very tightly bound and is removable only under drastic conditions, the source of the filling of the type 2 Cu site is unknown and warrants further scrutiny. All CuNIR species exhibit electronic absorbance around similar wavelengths (Fig. 11 and Table 5). Usually there is a prominent set of absorbances in the visible region around 460, 590, and 700 to 750 nm, with additional shoulders at shorter and longer wavelengths. CuNIR from Alcaligenes xylosoxidans and P. aureofaciens are blue enzymes and show very little absorption in the 450-nm region, whereas the other enzymes are green. The type 1 site is the chromophoric center since the type 2-depleted enzyme from ¡ÈA. cycloclastes¡É remains green (4). The set of ligands of the type 1 center in the green reduc-tases, (His)2 -Cys-Met, is the same as in azurin or other cupre-doxins. In the green reductases, the absorption at 590 nm is reduced and that around 460 nm is greatly enhanced. This difference reflects the degree of distortion of the tetrahedral, type 1 Cu-binding site (316, 499). Methionine and cysteine are slightly moved with respect to their position in a classical type 1 site. The Cu-S distance is somewhat shorter in the green than in the blue nitrite reductase (4). Green azurins can be engi-neered by introducing rhombic distortion into type 1 Cu sites by a strong axial ligand, for instance a Met121His substition (reviewed in reference 118).
Structure. The first three-dimensional structure analysis of a CuNIR was achieved with the enzyme from ¡ÈA. cycloclastes.¡É A detailed description of the refined structure, of crystal forms at various pH values, and of the substrate-bound form was pro-vided recently by Adman et al. (4). In the crystal, the enzyme subunits are tightly associated around a threefold axis to form a trimer around a central channel of 5 to 6 Å diameter (Fig. 12). Each subunit of the trimer comprises two domains. Their polypeptide fold is a ¡ÈGreek key¡É b barrel, similar to the type 1 blue proteins or cupredoxins (3, 4). The barrels are stacked onto each other, and two extended loops interact between domains I and II. Only nine hydrogen bonds exist between the two domains. Contributing to the stabilization of the molecule are surface interactions that comprise 28% of the monomer surface on trimer formation. The type 1 Cu site is formed from adjacent residues of domain I. The residues His95, Cys136, His145, and Met150 form a flattened tetrahedron similar to the Cu center in pseudoazurin (284). Type 2 Cu is bound by three histidine ligands (His100, His135, and His306) provided from domain II. However, His306 comes from the adjacent subunit; i.e., Cu is bound at the interface of two subunits (Fig. 12). The type 2 Cu is not required for the conformational stability of the trimer. Solution X-ray scattering was found to be unchanged for the type 2-depleted enzyme (796). Together with a water molecule, the ligands of the type 2 Cu form an unusual pseudotetrahedral geometry. A fourth histi-dine residue, His255, is nearby but is not a ligand in the crystal form obtained at pH 5.2. All histidines remain oriented in the same place on removal of the type 2 Cu. The distance to be bridged by electron transfer between the type 1 and type 2 Cu sites within the monomer is 13 Å. A second crystal structure of a green enzyme was obtained with the nitrite reductase from Alcaligenes faecalis (488). It confirms the Cu-binding residues and the trimeric nature of the enzyme and the other principal conclusions drawn from the ¡ÈA. cycloclastes¡É enzyme. The structure of the blue enzyme from Alcaligenes xylosoxidans was recently reported and con-firms that there are only subtle differences in the bond length of the Met-sulfur to Cu and angles with respect to those in the green form of CuNIR (208).
Mechanistic aspects. CuNIR forms in vitro NO from nitrite with PMS-asc as the electron donor system (Table 5). With the low-potential electron donor system of MV and dithionite, N2 O and ammonia formation have been reported also. The Alcaligenes faecalis S-6 and Rhodobacter sphaeroides IL106 en-zymes produce N2 O with MV and dithionite. A nonenzymatic further reduction of NO to N2 O takes place in this case. With PMS-asc as the reducing system the Alcaligenes faecalis enzyme generates NO and about 6% of N2 O (433); that from ¡ÈA. cycloclastes¡É generates about 3% N2 O besides NO (396). Sev-eral nitrite reductases were shown to form N2 O from nitrite with hydroxylamine, and some also reduce O2 (Table 5). The CuNIR of Alcaligenes xylosoxidans, in either a cell ex-tract or the soluble cell fraction, forms ammonia from nitrite with MV plus dithionite. The reaction is believed to comprise the conversion of NO to NH2 OH as a nonenzymatic step with cytochrome c9 as a hydroxylamine reductase (538). Ammonia formation from nitrite with MV plus dithionite and NO for-mation from nitrite with PMS plus asc have also been con-firmed to occur with the purified enzyme (1). A different mech-anism is required, since this excludes a catalytic contribution from cytochrome c9. The formulation of the active-site chem-istry will have to account for this enzyme versatility. The type 2 center is the substrate-binding site of nitrite reductase. Type 2-depleted enzyme is nearly inactive, but its activity increases markedly on reconstitution of the type 2 center (514). The spontaneous loss of Cu during the prepara-tion of the Alcaligenes xylosoxidans enzyme makes it feasible to compare species having only the type 1 Cu center occupied with species having both Cu centers filled. The EPR parame-ters of the type 2 Cu change on addition of nitrite from g i 2.355 to 2.290; A i increases by 2 mT. In contrast, there is no change in g i or the A tensor for type 1 Cu (366). The 1 H and 14 N electron nuclear double resonance (ENDOR) spectra of the type 2 Cu recorded at g i also change on addition of nitrite, whereas the equivalent spectra of the type 1 Cu are perturbed only to an insignificant extent. Both observations imply a re-arrangement of ligands at the type 2 center by the substrate nitrite. The extended X-ray absorption fine structure of the type 2-depleted enzyme closely resembles that for cupredoxins and remains unperturbed in the presence of nitrite, whereas that of the holoenzyme changes on nitrite binding (796). Gen-eral features for CuNIR thus comprise the trimeric structure, the electron entry via type 1 Cu, and the active site represented by type 2 Cu. The formation of a Cu-NO nitrosyl complex has been suggested to occur during the reaction of CuNIR with nitrite (371, 803). The cuprous nitrosyl is believed to be the key in-termediate in the reaction. The above reaction was formulated initially for the type 1-only nitrite reductase (803), but it is equally valid for the enzyme having its complete set of metal centers (396). The observed reactivity of the enzyme with hydroxylamine favors the exis-tence of an E-Cu-NO intermediate (371). When N2 O for-mation by CuNIR occurs, it is thought to result from the reaction of the nitrosyl intermediate with nitrite or the re-bound product NO. Evidence from 1 H ENDOR spectroscopy indicates that ni-trite displaces a water molecule from the type 2 Cu site. The lack of nitrogen coupling from 14 NO2 versus 15 NO2 as the substrate can be interpreted that nitrite binds to Cu via its oxygen atoms (366). The spectroscopic data are supported by the structural evidence obtained from nitrite-soaked crystals. Nitrite occupies the site of the water molecule with the oxygen atoms directed slightly asymmetrically toward the type 2 Cu at distances of 2.1 and 2.4 Å (4). The active site is located in a pocket formed by apposition of domains I and II of two dif-ferent monomers with the lining of domain II by hydrophobic amino acids and that of domain I by more hydrophilic residues. The mechanism is proposed to involve the displacement of a water molecule by nitrite, which leaves the active site as OH along the hydrophilic side, where it picks up a proton. The other proton stays on Asp98. This residue is assumed to be essential in protonating an oxygen from nitrite left at the Cu on cleavage of the Cu-ONO adduct. NO exits along the hydro-phobic side of the active site pocket (4). The reactivity of CuNIR with nitrite is mimicked by a syn-thetic mononuclear Cu 1 ONO2 2 complex that yields NO on protonation by glacial acetic acid (312). Different from the evidence of O-NO2 2 coordination by the enzyme, the nitrite coordination in the Cu complex is N-bonding, Cu 1 ON-NO2 2 . Nitrite could bind initially via its oxygen atoms as observed in the crystal, but rearrange during the catalytic cycle to give the Cu 1 ONO 1 intermediate. O-NO2 2 coordination has also been observed for active-site analogs. Questions remain whether nitrite binds in the enzyme initially to Cu(II) or Cu(I), the identity of the leaving group from the Cu-NO2 2 adduct, and whether oxygen transfer to Cu occurs. This point is also rele-vant to the mechanism of N2 O reduction by the Cu-containing N2 O reductase.
Electron donation. The redox potential for CuNIR from Alcaligenes xylosoxidans is 1260 mV at pH 7.2 (539). The measured midpoint potential is that of type 1 Cu, since type 2 Cu was absent from the enzyme preparation. For the ¡ÈA. cy-cloclastes¡É reductase, with both types of Cu, a midpoint poten-tial of 1240 mV was reported for the type 1 Cu (463) and 1260 mV was reported for the type 2 Cu (802). The principal elec-tron donors to CuNIR are azurin and pseudoazurin, while cytochromes appear less frequently involved. Azurins and pseudoazurins are distinguished by their optical spectrum and a number of structural elements, but both are members of the cupredoxin family of small (12- to 14-kDa) electron transfer proteins with a single type 1 (blue) Cu atom (2). Pseudoazurins have additional absorption bands around 450 and 750 nm. Azurins have a cysteine at the third position from the N ter-minus that is part of a disulfide bridge. Electrons enter CuNIR from a cupredoxin via the type 1 Cu. The intramolecular electron transfer rate from type 1 Cu to type 2 Cu, studied by pulse radiolysis, is slower than the re-duction of the type 1 Cu (802). The intramolecular electron transfer path in CuNIR shows analogies to that of ascorbate oxidase and involves the cysteine of type 1 Cu and His135 of the type 2 Cu of the same subunit (488). Azurins are electron donors for the blue variants of CuNIR found in P. aureofaciens (977) and Alcaligenes xylosoxidans (209). The latter bacterium synthesizes two similar azurins distinguishable in their primary structure but having the same redox potential (1305 mV) (209). Both proteins are active as electron donors. The spontaneously type 2 Cu-depleted en-zyme does not react with azurin but instead is active with reduced cytochrome c (583). Cytochrome c2 is the electron donor to CuNIR of Rhodobacter sphaeroides IL106 (720). Pseudoazurins are the electron donors for the green enzymes from Alcaligenes faecalis S-6 (432) and ¡ÈA. cycloclastes¡É (516). The cupredoxin-like donors have similar positive redox poten-tials as CuNIR to effect the electron transfer nearly isopoten-tially. The interaction of the electron donor with its cognate re-ductase involves a strong electrostatic element. This has been demonstrated by studying the docking of pseudoazurin and CuNIR by site-directed mutagenesis, which revealed a set of appropriately positioned complementary charges on the two proteins (489, 490). Pseudoazurin of Alcaligenes faecalis has a surface ring of lysine residues around the Cu atom (Lys10, Lys38, Lys57, and Lys77), which is required for the electron transfer to its cognate CuNIR. There is little change in the rate of electron transfer on substitution of these residues, but the Km increases substantially, particularly when Lys10 and/or Lys38 is replaced. Lys10 is crucial for electron transfer in this system. At the CuNIR surface, the residues Glu118, Glu197, Glu204, and Asp205 are conserved. Their substitution lowers the enzyme activity and increases the Km . The following pairing is possible: Glu197 with Lys10, Glu118 with Lys77 and/or Lys57, and Glu204 and/or Asp205 with Lys38. This pairing places the two type 1 Cu sites 14 to 15 Å away from each other. The electron transfer is thought to involve hydrogen bonds along a pathway that shows similarity to that at the CuA center of COX (490). In addition to a distinct charge recognition between the molecules, the overall charge distribution of the molecule is also important for the orientation of the molecule toward the reductase (489). The X-ray structures of pseudoazurin of Alcaligenes faecalis and azurin of P. aeruginosa are largely superimposable, al-though there is as little as 11% sequence identity between both proteins (491). In spite of the structural similarity, P. aerugi-nosa azurin reacts poorly with CuNIR from Alcaligenes faecalis. This can now be rationalized by the lack of the surface lysines. Introducing two lysine residues increases the affinity of azurin to the Alcaligenes reductase (491). It has also been noted that of the relevant charged residues for docking on CuNIR, only Asp205 is conserved in the enzyme of P. aureofaciens. The interaction of an azurin with its cognate CuNIR may therefore be of a different nature (490).