Characterization of NarJ, a System-specific Chaperone Required for Nitrate Reductase Biogenesis in Escherichia coli *

The narGHJI operon encodes the three subunits, α, β, and γ, of the respiratory nitrate reductase complex in Escherichia coli. A fourth open reading frame of the operon encodes a putative protein, NarJ, which is not present in purified nitrate reductase, but is required for biogenesis of the membrane-bound complex. NarJ was identified with a T7 expression system and was produced at significantly less than stoichiometric levels relative to the three enzyme subunits. A functional His-tagged NarJ fusion protein was overexpressed from a multicopy plasmid, purified by Ni2+ affinity chromatography, and characterized. Western blot analysis with antibodies raised against the fusion protein demonstrated that NarJ remained in the cytosol after assembly of the active membrane complex. The cytosolic αβ complex accumulated in anarJ insertion mutant was rapidly degraded after induction, but was stabilized by NarJ expressed from a multicopy plasmid. Overproduction of the His-tagged NarJ fusion protein in the same mutant led to the formation of an αβ·NarJ complex, which was resolved by Ni2+ affinity chromatography. The NarJ protein therefore has the properties of a system-specific (private) chaperone that reacts directly with and modifies the properties of the cytosolic αβ subunit complex, but remains in the cytoplasm after the assembly of the active αβγ complex in the membrane.

Respiratory nitrate reductase in Escherichia coli is a membrane-bound electron transfer complex that is composed of three subunits, ␣, ␤, and ␥, encoded by the narG, narH, and narI genes of the narGHJI operon, respectively (1)(2)(3). The ␣ and ␤ subunits are hydrophilic proteins that contain molybdopterin and non-heme iron cofactors and the active site for nitrate reduction. In the membrane complex, these subunits are tightly associated with the hydrophobic ␥ subunit, a hemeprotein (cytochrome b NR ) that is required for the transfer of electrons from physiological donors through the membrane quinone pool to nitrate (4 -6). An ␣␤ dimer has been released from the membrane by heat treatment and purified as an active complex that reduces nitrate to nitrite with artificial electron donors such as reduced methyl viologen (6 -8). The ␣␤␥ complex has been purified from membranes in the presence of detergents (2,4,5); it utilizes quinol analogues as well as reduced viologen dyes as electron donors for nitrate reduction (6).
The fourth gene product of the narGHJI operon, NarJ, is required for biogenesis of the membrane-bound complex (9, 10), but is not present in the purified preparations of nitrate reductase (2). In a narJ mutant, the ␥ subunit was inserted normally into the membrane, whereas the ␣␤ complex accumulated in the cytosol in a form that contained a normal complement of cofactors, but was completely inactive (10). In contrast, the ␣␤ complex that accumulated in narI mutants in the presence of NarJ appeared to express low levels of reduced methyl viologen-nitrate reductase activity (9,10). These results suggested that, during the biogenesis of nitrate reductase, NarJ may modify the ␣␤ complex or facilitate its interaction with the membrane-bound ␥ subunit. It was not clear, however, whether NarJ was a subunit of the complex that was lost during the purification procedures or a chaperone-like component that was required only for assembly of the complex. The experiments presented here were undertaken to identify and localize NarJ in the cell and to clarify its possible role in the expression of nitrate reductase.
Plasmids pSL962 (narGHJI), pMV4 (narJI), and pES203.1 (narJ) were previously described (2). Plasmid pGP1-2 (Kan R , T7 RNA polymerase) and the Amp R T7 expression vectors pT7-4 and pT7-6 were as described by Tabor and Richardson (12). The Cam R T7 expression vector pSU24 was kindly supplied by Valley Stewart (Cornell University). Specific fragments of the narGHJI operon were subcloned into a polycloning site behind the T7 promoter of the T7 expression vectors as diagrammed in Fig. 1A. To construct plasmid pXL241, a 7186-base pair fragment containing the narGHJI genes, isolated on an agarose gel after digestion of plasmid pSL962 with EcoRI (partial) and DraI (complete), was ligated with plasmid pSU24 previously digested with SmaI and EcoRI. To construct plasmid pXL243, a 1166-base pair fragment containing the narJ open reading frame, isolated on agarose after digestion of plasmid pMV4 with PstI and XmnI, was ligated with plasmid pSU24 previously digested with PstI and SmaI. For construction of plasmids pXL742 and pXL762, a 2162-base pair fragment containing the narJI open reading frames, isolated after digestion of pMV4 with PstI and EcoRI, was ligated with pT7-4 and pT7-6, respectively, previously digested with PstI and EcoRI. For construction of plasmids pXL744 and pXL746, a 1559-base pair fragment containing the narI open reading frame, isolated after digestion of pMV4 with BamHI and EcoRI, was ligated with pT7-4 and pT7-6, respectively, previously digested with BamHI and EcoRI. The structure of each constructed plasmid, isolated from colonies of strain MV1190 transformed to ampicillin resistance, was confirmed by restriction enzyme mapping.
To construct plasmids expressing the His-tagged NarJ fusion protein, a HindIII-EcoRI fragment carrying the narJI open reading frames was subcloned into phagemid pTZ19R (Bio-Rad), and the sequence around the translation initiation codon on narJ was modified by site-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. directed mutagenesis to create a NcoI restriction site. This modification (A to G at nucleotide 4 of the open reading frame) resulted in the substitution of Val for Ile at residue 2 of the putative NarJ protein and had no effect on the ability of the plasmid to complement a narJ mutant. The resulting plasmid was digested with NcoI, the ends were blunted by the Klenow filling reaction, and two fragments were produced by digestion with XmnI (narJ open reading frame) and EcoRI (narJI open reading frames), respectively. The narJ fragment was ligated with plasmid pTrcHis (Invitrogen) that had been digested with BamHI, followed by the Klenow filling reaction. In the resulting plasmid, pXLJ-His 6 , the entire modified narJ sequence was fused in-frame to an open reading frame that encoded a 36-residue peptide containing 6 sequential His residues and a enterokinase cleavage site. The narJI fragment was ligated with pTrcHis treated as described above but in addition digested with EcoRI. In the resulting plasmid, pXLIJ-His 6 , the same fused narJ open reading frame was constructed followed by the narI open reading frame. The general structure of each of these plasmids was confirmed by restriction enzyme mapping.
T7 Expression Procedure-The selective expression of the narGHJI components from the T7 promoter was carried out essentially as described by Tabor and Richardson (12) in strain K38(pGP1-2) transformed with the recombinant T7 expression plasmids described above. The strains were grown in 5 ml of L-broth supplemented with 50 g/ml kanamycin and either 100 g/ml ampicillin or 25 g/ml chloramphenicol to 150 Klett units (540-nm filter), treated essentially as described (12), and labeled with [ 35 S]methionine (5 C/ml) at 30°C. Cells were harvested by centrifugation, dissolved in SDS sample buffer (16 mM Tris-HCl, pH 6.8, 20 mM dithiothreitol, 0.4% SDS, 2% glycerol, and 0.0008% bromphenol blue), and analyzed by SDS-PAGE 1 followed by autoradiography.
The same procedure was followed for the pulse-chase experiment except that 0.1% nonradioactive methionine was added after the 5-min labeling period and the incubation continued at 30°C. Aliquots were removed at 10-min intervals and analyzed as described above.
Cell Disruption and Fractionation-Crude extracts of cells were prepared by passing cells suspended in Buffer A (20 mM sodium phosphate, pH 7.8, 0.5 M NaCl, 0.1 mM phenylmethylsulfonyl fluoride, and 0.1 mM N ␣ -p-tosyl-L-lysine chloromethyl ketone) through a French press twice at 15,000 p.s.i. and removing cell debris by centrifugation at 10,000 ϫ g for 10 min. To determine the localization of nitrate reductase components, crude extracts were centrifuged for 90 min at 100,000 ϫ g. The resulting supernatants were assumed to represent the cytosolic fraction, and the pellets suspended in the same volume of Buffer A represented the membrane fraction.
Overproduction and Purification of NarJ-His 6 -Strain MV1190(pXLJ-His 6 ) was grown on a rotary shaker at 37°C in 100 ml of L-broth to midlog phase (150 Klett units, 540-nm filter); 1 mM IPTG was added; and the culture was grown for an additional 3 h. A crude extract was prepared as described above, and the overproduced fusion protein was purified on a 2-ml column of Ni 2ϩ -charged ProBond resin (Invitrogen) washed two times with 10 ml of Buffer A. The crude extract (10 ml) was loaded by gravity, and the unabsorbed proteins were removed by washing with 50 ml of Buffer A. The column was then eluted consecutively with 10-ml volumes of Buffer B (Buffer A at pH 6.0) supplemented with 0.8, 8, 40, 80, and 300 mM imidazole. Fractions (1.0 ml) were analyzed by SDS-PAGE, and the peak fractions containing the fusion protein were pooled, concentrated on an Amicon filtration unit, and finally washed and suspended in water.
The affinity-purified sample was further resolved by chromatography on a Sephadex G-100 column (1 ϫ 57 cm) that was equilibrated and run with 50 mM Tris-HCl, pH 8.0, containing 1 mM EDTA at a rate of 2 ml/h. The column was calibrated with a mixture of standard proteins (50 g each of bovine serum albumin, carbonic anhydrase, and cytochrome c plus 5 g of blue dextran in 200 l).
To remove most of the N-terminal peptide containing the polyhistidine tag from the fusion protein, samples of the purified fusion protein was digested overnight at 37°C with calf intestine enterokinase (Biozyme Laboratories International Ltd.) in 40 mM Tris-HCl, pH 8.0, and 5 mM CaCl 2 , and the resulting products were purified by gel filtration on the Sephadex G-100 column described above.
General Analytical Procedures-The N-terminal amino acid sequences were determined on an Applied Biosystems Model 477A gasphase sequencer by Chris Chen in the Analytical Chemistry Center of this institution. Protein masses were determined with the matrix-assisted laser desorption ionization mass spectrometry system in the Analytical Chemistry Center.
Protein was determined by the procedure of Lowry et al. (14) or was followed by absorption at 280 nm. SDS-PAGE was carried out on gels prepared and run according to the procedures of Laemmli (15). Whole cells or proteins were dissolved in SDS sample buffer and heated for 3 min at 95°C before loading. If necessary, the undissolved residue from cells was removed by brief centrifugation.
Antisera against the purified NarJ-His 6 fusion protein were produced commercially by Bethyl Laboratories, and specific antibodies used for Western blot analyses were purified by the procedure of Olmstead (16). Purified antibodies against the ␣ and ␤ subunits of nitrate reductase were from an earlier study (9) and were purified by the same procedure.
Western blot analyses employed detection by enhanced chemiluminescence following the protocol described by the manufacturer (Amersham International). After SDS-PAGE, the proteins were transferred from gels to nitrocellulose paper with a semi-dry Electroblotter (Millipore). The washing and labeling procedures were as described using horseradish peroxidase-labeled second antibodies. Purified nitrate reductase used as standard was prepared previously (2).

Identification of NarJ-
The four open reading frames of the narGHJI operon (2, 3) encode proteins with calculated masses of 140 kDa (narG, ␣ subunit), 58 kDa (narH, ␤ subunit), 26.5 kDa (narJ), and 25.5 kDa (narI, ␥ subunit). To identify NarJ, fragments of the narGHJI operon were cloned into T7 expression vectors (Fig. 1A) and selectively expressed in the presence of [ 35 S]methionine as described under "Materials and Methods." Parallel constructions were made for each fragment in vectors with the ␤-lactamase gene oriented in opposite directions so that the processed fragments of ␤-lactamase could be identified and used as internal controls.
When the narJ and narI genes were expressed together in cells carrying either plasmid pXL742 or pXL762, two bands corresponding to apparent 29-and 23-kDa proteins were observed (Fig. 1B, lanes 3 and 4) along with ␤-lactamase and its processed fragments (lane 1). The 23-kDa band was identical to that observed for the narI gene product expressed alone from plasmids pXL744 and pXL764 (lanes 5 and 6), whereas the 29-kDa band, which apparently corresponded to NarJ, migrated closely to one of the processed products of the ␤-lactamase gene (lanes 1 and 3). Although the putative NarJ protein migrated with an apparent mass that was higher than expected, its expression from pXL762 (lane 4), which contained the ␤-lactamase gene oriented in the opposite direction, confirmed its identification as the narJ gene product. The level of NarJ appeared to be significantly less than that of NarI when both were expressed from the same plasmid, even when adjusted for the 3:1 (NarI/NarJ) ratio of methionine content.
A fragment expressing NarJ only and one expressing all four operon proteins were also cloned into the T7 expression vector pSU24 (Fig. 1A). The selective marker for this vector, cat, encodes a product that is not processed and migrates faster than NarJ on SDS-PAGE, facilitating the identification of NarJ in the selective expression system (Fig. 1C). NarJ appeared to be formed in less than stoichiometric amounts relative to the levels of NarI (␥ subunit) even when expressed with the entire operon (fifth through eighth lanes) and at even lower levels when expressed alone (first through fourth lanes). The pulsechase experiment shown in Fig. 1C demonstrated that the turnover rate of NarJ was not significantly different from those of the other subunits, eliminating a rapid turnover rate as the explanation for the relatively low levels of NarJ.
We conclude that NarJ migrates on SDS-PAGE less rapidly than would be predicted from its mass calculated from the narJ DNA sequence. It appeared to be produced with the T7 expression system at less than stoichiometric levels compared with the other products of the narGHJI operon in amounts that were not compatible with the development of a facile purification procedure.
Expression and Purification of NarJ as a His-tagged Protein-To facilitate the purification of NarJ, the narJ gene was cloned into a vector to create a fused protein with a polyhistidine tag as described under "Materials and Methods." The resulting plasmid, pXLJ-His 6 , encoded a protein with a 36residue peptide, containing six tandem His residues and an enterokinase cleavage site, fused to the N-terminal Met of the NarJ sequence. Induction of expression of the fused gene by IPTG in a wild-type strain carrying plasmid pXLJ-His 6 led to the overexpression of an apparent 33-kDa protein, which was located almost exclusively in the supernatant fraction after high speed centrifugation of crude extracts (data not shown).
The His-tagged NarJ fusion protein functionally replaced NarJ for expression of nitrate reductase in a narJ mutant. Mutant MD100 produces only ␣ and ␤ subunits and no nitrate reductase activity as the result of an insertion in the narJ gene (9). Transformation of MD100 with pXLJ-His 6 or an analogous plasmid, pXLIJ-His 6 , which encoded NarI as well as the Histagged NarJ fusion protein, led to the expression of nitrate reductase activities on plate assays (17) similar to those formed with MD100 transformed with plasmids expressing NarJ and NarJ plus NarI (9).
The His-tagged fusion protein was readily purified on a Ni 2ϩ affinity column (Fig. 2A). The overproduced fusion protein appeared to be partially cleaved by endogenous proteases following breakage of cells in the French press (lanes 2 and 3). Much of the overproduced fusion protein and its major cleavage products were absorbed to the Ni 2ϩ resin and eluted by increasing concentrations of imidazole. After most of the contaminating proteins were washed off the column at lower imidazole concentrations ( lanes 5-8), the fusion protein and its cleavage products were eluted at 300 mM imidazole (lanes 9 -13). The pool of the peak fractions, concentrated and desalted, contained three major components (I-III) on SDS-PAGE (Fig. 2B, lane 2). N-terminal sequence analysis yielded only one sequence, GGSHHHHHHG, which corresponded to that expected for residues 2-11 of the peptide containing the polyhistidine and enterokinase cleavage site motifs. The mass of the largest component (30,639 Da), determined by matrix-assisted laser desorption ionization mass spectrometry, corresponded reasonably closely to that calculated for the fusion protein with the N-terminal Met removed (30,315 Da). Because only one Nterminal sequence was observed for the fraction that was absorbed and eluted from the Ni 2ϩ resin, it seemed most likely that the other two smaller fragments were generated by removal of C-terminal sequences, leaving the N-terminal sequence intact.
Treatment of the pooled purified fraction with enterokinase generated three smaller fragments (Fig. 2B, lane 3). N-terminal sequence analysis revealed only one N-terminal sequence, DRWGSMVLV, which corresponded to the last 5 residues of the fusion peptide and the first 4 residues of the narJ open reading frame. This result indicated that all three peptides contained the same N-terminal sequence after enterokinase cleavage. The mass of the largest fragment resulting from enterokinase cleavage (23,525 Da), as determined by mass spectroscopy, corresponded closely to the mass calculated for the cleavage of protein III at the enterokinase site (23,232 Da). These results verify that the purified protein is the expected His-tagged NarJ fusion protein. In extracts, the fusion protein appears to be very sensitive to endogenous proteases, which degrade the protein mainly in the C-terminal region.
The undegraded form of His-tagged NarJ was resolved from most of the partially degraded forms by gel filtration under nondenaturing conditions (Fig. 3A). The affinity-purified protein was resolved into two peaks. Peak I eluted at a position that corresponded to a dimer (66 kDa) and was composed almost exclusively of the undegraded 33-kDa form of the protein (Fig. 3B). Peak II eluted at the monomeric position and appeared to contain both undegraded and degraded forms (Fig.  3B).
Western blots employing antisera prepared against Histagged NarJ demonstrated that little, if any, proteolysis of the fusion protein occurred in vivo (Fig. 4). The purified antiserum did not significantly cross-react with other proteins in the cell extracts (lanes 2 and 3), and only the intact protein was visualized in cells that overproduced the His-tagged protein (lanes 4 and 5).
Localization of NarJ-NarJ, produced at relatively low levels under most conditions, could also be visualized on Western blots using the antiserum prepared against the His-tagged protein. In Fig. 5A, low levels of NarJ were observed in extracts of wild-type cells (lane 2) and significantly increased levels in  5). The band identified as NarJ was not present in cells of mutant MD100 (lane 3), which contains an insertionally inactivated narJ gene. When crude extracts from the wild-type strain RK4353 were fractionated by high speed centrifugation, essentially all of the NarJ protein was present in the supernatant fraction (Fig. 5B). Under these same conditions, the nitrate reductase subunits, as assessed by Western blotting with antibodies against the ␣ and ␤ subunits, were located chiefly in the membrane fraction (9).
Formation of an ␣␤⅐NarJ Complex-Although NarJ is not associated with the assembled membrane-bound nitrate reductase complex, it seemed possible that it interacted directly with an intermediate in the assembly pathway. Previous studies with mutants (9, 10) had established that, in the absence of the integral membrane ␥ subunit, the ␣ and ␤ subunits accumulated as a dimeric complex. In the absence of NarJ, this complex had a normal complement of cofactors (10), but was completely devoid of reduced methyl viologen-nitrate reductase activity. In the presence of NarJ, the complex expressed significant reduced methyl viologen-linked activity, but at very re-  Fig. 2B, lane 2) from a Sephadex G-100 column as described under "Materials and Methods"; B, analysis of column fractions (equal volumes) by SDS-PAGE (15% acrylamide). The gel was stained with Coomassie Blue. The first lane contained the protein standard mixture, and the position of carbonic anhydrase is indicated (31kD).

FIG. 4. Western blot analysis of the His-tagged fusion protein accumulated in vivo.
The blot was stained using antibodies against purified NarJ-His 6 . Cells were grown in the presence of IPTG to induce full expression of the His-tagged fusion protein from the plasmids. Preparation and electrophoresis of whole cell extracts were as described under "Materials and Methods." Lane 1, purified NarJ-His 6 (60 g); lane 2, undiluted whole cell extract of strain MV1190 (untransformed); lane 3, undiluted whole cell extract of strain MV1190 transformed with the unmodified vector pTrcHisA; lane 4, 100-fold dilution of whole cell extract of strain MV1190 transformed with pXLJ-His 6 (expressing the His-tagged fusion protein); lane 5, 10-fold dilution of strain MV1190 transformed with plasmid pXLIJ-His 6 (expressing both the NarJ fusion protein and the NarI protein). duced levels (15-20%) compared with that of the membranebound complex (9). As shown in Fig. 6, the presence of NarJ appeared to stabilize the ␣␤ complex that accumulated in the absence of the ␥ subunit. In this experiment, mutant MD100 was grown aerobically to midlog phase and then shifted to anaerobic conditions with the addition of nitrate to induce expression of the two subunits, and the levels of the complex were assessed in extracts of cell samples by Western blotting using anti-␣ subunit antibodies. In wild-type strains, this induction procedure leads to the rapid formation of nitrate reductase activity, with steady-state specific activities being reached in ϳ90 min (18). In MD100 (Fig. 6A), the ␣ subunit accumulated rapidly, reaching a maximum in 60 min and then declining after 90 min. To examine the effects of NarJ on the accumulation of the complex, MD100 was transformed with plasmids that expressed either NarJ (pES203.1) or NarJ and NarI (pMV4) under the control of the tac promoter (9), and induction was carried out with the addition of IPTG to simultaneously induce these components. When both NarJ and NarI were induced with the ␣ and ␤ subunits (Fig. 6C), the ␣ subunit levels increased for 60 min and then remained constant, as might be expected for the formation of the stable, membranebound complex. Similarly, the induction of NarJ alone with the ␣ and ␤ subunits led to the stabilization of the accumulated ␣ subunits (Fig. 6B). This result suggested that NarJ may react directly with and stabilize the ␣␤ complex prior to its interaction with the membrane-inserted ␥ subunit.
To look directly for an ␣␤⅐NarJ complex, we took advantage of the special characteristics of the functionally active Histagged NarJ fusion protein. The ␣␤ complex and His-tagged NarJ were co-induced in strain MD100(pXLJ-His 6 ), and the resulting cells were processed and fractionated as described above for the overproduced His-tagged fusion protein. SDS gels prepared from the fractions were stained directly for protein (Fig. 7A) or blotted and separate regions of the paper stained with specific antisera against the ␣ and ␤ subunits (Fig. 7B) or His-tagged NarJ (Fig. 7C). In this short induction procedure, His-tagged NarJ was not overproduced to the same degree as in the cultures used for purification, but based on both the protein stain and the Western blot (Fig. 7, A and C), it is clear that the fusion protein was absorbed and eluted in an almost completely undegraded form. The ␣ and ␤ subunits, induced at much lower levels, were also absorbed to the Ni 2ϩ affinity column and eluted at high imidazole concentrations in the same fractions with His-tagged NarJ (Fig. 7B). A presumed partially degraded form of the ␣ subunit, also present in the eluted fractions, was apparently formed during the column procedure since it was not present in the crude extract (lane 3). A similar induction and fractionation procedure was carried out with strain MD100(pES203.1), which produced both NarJ and the ␣␤ com-plex; in this case, most of the NarJ protein and ␣ and ␤ subunits were not absorbed to the column, and the remainder of each was washed off at low concentrations of imidazole (data not shown). These results demonstrate that overproduced Histagged NarJ formed a complex with the ␣␤ complex that was sufficiently stable to be purified by binding to and elution from the Ni 2ϩ affinity resin.  (expressing narJ and narI). B, localization of NarJ in wild-type strain RK4353. Fractions were separated by centrifugation after breakage in the French press as described under "Materials and Methods," and equivalent volumes of each fraction were analyzed using antibodies against NarJ-His 6 . Lane 1, whole cell extract; lane 2, supernatant fraction; lane 3, membrane fraction.
FIG. 6. Effect of NarJ on the stability of the ␣ subunit. Cells were grown aerobically in L-broth to midlog phase and then shifted to anaerobic conditions with the addition of 1% nitrate to induce expression of ␣ and ␤ subunits from the chromosome and NarJ (pES203.1) or NarJ plus the ␥ subunit (pMV4) from the plasmids. Cultures were sampled at the indicated times after the anaerobic shift, and whole cell extracts were analyzed by Western blotting using antibodies against the purified ␣ subunit as described under "Materials and Methods." The same level of purified ␣ subunit was loaded in each panel to control for exposure time in developing the blots. A, MD100; B, MD100(pES203.1); C, MD100(pMV4).

FIG. 7.
Identification of an ␣␤⅐NarJ-His 6 complex. Strain MD100(pXLJ-His 6 ) was grown aerobically to midlog phase in the presence of IPTG to induce the overproduction of NarJ-His 6 and then shifted to anaerobic conditions in the presence of nitrate for 60 min to induce the formation of the ␣ and ␤ subunits. Extracts of the cells were prepared and fractionated on a Ni 2ϩ affinity column as described in the legend to Fig. 2A. Equal volumes of all fractions were analyzed after SDS-PAGE (7.5-15% acrylamide gradient) by staining with Coomassie Blue (A) or by Western blotting as described in the legend to Fig. 4 using antibodies against the purified ␣␤ complex (B) or against purified NarJ-His 6 (C).

DISCUSSION
The narJ gene is the third of four contiguous open reading frame in the narGHJI operon, and its start and stop codons overlap by 1 base pair each of the adjacent open reading frames (19). In general, this type of overlapping structure has been thought to lead to translational coupling and the formation of similar amounts of each of the operon products (20). This did not seem to be the case for the NarJ protein; less than stoichiometric levels of NarJ accumulated with the T7 expression system compared with the narG, narH, and narI gene products even though NarJ appeared to be just as stable as the other operon proteins. Furthermore, only low levels of NarJ could be accumulated when fragments containing the narJ gene were expressed from multicopy plasmids under the control of the tac promoter. These results suggested that, relative to the other operon products, the expression of NarJ is restricted at the level of translation and that less than stoichiometric levels of NarJ are required for the assembly of nitrate reductase.
Based on its properties, the NarJ protein would appear to fit the formal definition of a molecular chaperone. It is a protein that is required for the assembly of the nitrate reductase complex on the membrane; it forms a complex with the ␣ and ␤ subunits prior to their assembly with the membrane-integrated ␥ subunit; and after assembly of the ␣␤␥ complex on the membrane, NarJ remains in the cytosol.
In contrast to general molecular chaperones (21), NarJ would have to be defined as a system-specific chaperone since it appears to be required only for the biogenesis of nitrate reductase. As a component of the narGHJI operon, it is expressed only when nitrate reductase formation is induced, and its deletion appears to have little effect on the growth and physiology of the affected cells except that which is attributable to the absence of functional nitrate reductase. Genes related to narJ have been identified in other operons that encode membranebound nitrate reductase complexes (22)(23)(24), but data base searches have not revealed significant homologies to any other known genes that might provide clues concerning the role of NarJ or its homologues in the biogenesis of nitrate reductase.
What possible role might NarJ play in the assembly of the nitrate reductase complex? Previous studies have established that a cytoplasmic ␣␤ complex, accumulated in mutants that produce no membrane-integrated ␥ subunit, is partially active when NarJ is present, but is inactive in mutants that lack NarJ as well (9,10). When purified, the inactive form contained a full complement of cofactors (10) and appeared to be readily degraded by endogenous (9) and exogenous (10) proteases. The ␣␤ complex accumulated in the presence of NarJ catalyzed the reduced methyl viologen-nitrate reductase reaction at ϳ15-20% the rate catalyzed by the membrane-bound complex (9), and it appeared to be less sensitive to endogenous protease degradation than the inactive ␣␤ complex accumulated in the absence of NarJ (9). Furthermore, Palmer et al. (25) have shown recently that, during the in vitro incorporation of molybdopterin cofactor, NarJ is one of the components required to form active nitrate reductase. Together, these observations suggest that NarJ facilitates a maturation or refolding of the inactive ␣␤ complex to an active, protease-resistant complex. Since the inactive ␣␤ complex does not form a complex with the membrane-bound ␥ subunit in the absence of NarJ (9, 10), it seems likely that either the active "matured" form of ␣␤ or the form bound to NarJ is required for interaction with the membrane-integrated ␥ subunit and assembly of the active ␣␤␥ complex. A detailed characterization of the structure and ac-tivity of the ␣␤⅐NarJ complex should provide more definitive information about the steps involved in assembly of nitrate reductase and the role of NarJ in this process.
Putative system-specific or "private" chaperones have been identified for several other unrelated systems, including extracellular lipase formation (limA) in Pseudomonas cepacia (26), urease formation (ureD) in Klebsiella aerogenes (27), pilus formation (papD) in E. coli (28), and F 1 -ATPase assembly (ATP11, ATP12) in Saccharomyces cerevisiae (29). In each case, the identified component, like NarJ, is required for assembly or processing, but is not found associated with or required for activity or function of the final product. None of these systemspecific chaperones have structural features or apparent activities in common that might imply general functional homologies. It is possible that, in each case, a specific problem in biogenesis can be solved only by the participation of a specific protein factor. It is also possible that some general function is shared by these factors, such as protection or formation of otherwise reactive intermediates in the assembly process or facilitation of an interaction of intermediates with the general molecular chaperones of the cell to complete folding steps that are critical to the assembly process.