Synthesis of pyrroloquinoline quinone in vivo and in vitro and detection of an intermediate in the biosynthetic pathway

In Klebsiella pneumoniae, six genes, constituting the pqqABCDEF operon, which are required for the synthesis of the cofactor pyrroloquinoline quinone (PQQ) have been identified. The role of each of these K. pneumoniae Pqq proteins was examined by expression of the cloned pqq genes in Escherichia coli, which cannot synthesize PQQ. All six pqq genes were required for PQQ biosynthesis and excretion into the medium in sufficient amounts to allow growth of E. coli on glucose via the PQQ-dependent glucose dehydrogenase. Mutants lacking the PqqB or PqqF protein synthesized small amounts of PQQ, however. PQQ synthesis was also studied in cell extracts. Extracts made from cells containing all Pqq proteins contained PQQ. Lack of each of the Pqq proteins except PqqB resulted in the absence of PQQ. Extracts lacking PqqB synthesized PQQ slowly. Complementation studies with extracts containing different Pqq proteins showed that an extract lacking PqqC synthesized an intermediate which was also detected in the culture medium of pqqC mutants. It is proposed that PqqC catalyzes the last step in PQQ biosynthesis. Studies with cells lacking PqqB suggest that the same intermediate might be accumulated in these mutants. By using pqq-lacZ protein fusions, it was shown that the expression of the putative precursor of PQQ, the small PqqA polypeptide, was much higher than that of the other Pqq proteins. Synthesis of PQQ most likely requires molecular oxygen, since PQQ was not synthesized under anaerobic conditions, although the pqq genes were expressed.

Pyrroloquinoline quinone (PQQ) is a cofactor of several bacterial dehydrogenases and transfers redox equivalents to the respiratory chain. The physiological electron acceptors vary from ubiquinone in the case of membrane-bound glucose dehydrogenase (e.g., glucose dehydrogenase of Acinetobacter calcoaceticus) to a cytochrome c in the case of methanol dehydrogenases (e.g., methanol dehydrogenase of Methylobacterium extorquens AM1) (for a review, see reference 2). The chemical structure of PQQ has been determined (13,33), but the biosynthetic pathway of PQQ has not yet been solved. From 13 C nuclear magnetic resonance studies with Hyphomicrobium X and M. extorquens AM1, it was suggested that the amino acids tyrosine and glutamic acid are the precursors for PQQ (19,44). Studies to detect intermediates in PQQ biosynthesis in A. calcoaceticus, Methylobacterium organophilum, and Pseudomonas aureofaciens have been negative thus far (43).
Genes involved in PQQ biosynthesis have been cloned from several organisms. Five A. calcoaceticus pqq genes, pqqIV, V, I, II, and III (15,17), and six Klebsiella pneumoniae pqq genes, pqqA, B, C, D, E, and F (25,26), were cloned and sequenced. Comparison of the deduced amino acid sequences showed that the proteins encoded by the first five genes of the K. pneumoniae pqq operon (pqqABCDE) show similarity to the proteins encoded by the corresponding A. calcoaceticus genes (49 to 64% identical amino acid residues). The K. pneumoniae pqqF gene encodes a protein that shows similarity to Escherichia coli protease III and other proteases (26), but its equivalent has not yet been found in A. calcoaceticus. Recently, three M. extorquens AM1 pqq genes, pqqD, G, and C, have been cloned and sequenced (28); pqqC was only partly sequenced. The encoded proteins showed similarity to the K. pneumoniae PqqA, B, and C proteins and the A. calcoaceticus PqqIV, V, and I proteins, respectively. Four additional pqq genes have been detected in M. extorquens by isolation of mutants and complementation studies. From similar studies, six (possibly seven) pqq genes have been postulated in M. organophilum DSM760 (4). Finally, a DNA fragment cloned from Erwinia herbicola contained a gene encoding a protein similar to K. pneumoniae PqqE and A. calcoaceticus PqqIII (22). Except for the K. pneumoniae PqqF protein, none of the Pqq proteins shows similarity to other proteins in the database.
One of the pqq genes is small and may encode a polypeptide of 24 amino acids (PqqIV, A. calcoaceticus), 23 amino acids (PqqA, K. pneumoniae), or 29 amino acids (PqqD, M. extorquens AM1). Interestingly, these putative polypeptides contain conserved glutamate and tyrosine residues (positions 15 and 19, respectively, in K. pneumoniae and the equivalents in A. calcoaceticus and M. extorquens). Those residues have been suggested previously as precursors in PQQ biosynthesis. Replacement of Glu-16 by Asp and Tyr-20 by Phe in A. calcoaceticus PqqIV abolished PQQ biosynthesis (16). A frameshift in K. pneumoniae pqqA had the same result (26). It was suggested that the PqqA/PqqIV polypeptide might act as a precursor in PQQ biosynthesis (15,16,26).
Our aim is to elucidate the route of PQQ biosynthesis and the role of each of the six known K. pneumoniae pqq genes in this process. We have taken advantage of the fact that E. coli is unable to synthesize and excrete PQQ unless supplied with the six K. pneumoniae pqq genes (25,26). Using plasmids in which one of the six pqq genes is inactivated at the time, we have investigated PQQ synthesis in vivo and in vitro. We also examined the expression of the different pqq genes of K. pneumoniae, especially pqqA.
(Part of this work was presented in a preliminary form at the 9th Meeting on Vitamin B 6 and Carbonyl Catalysis and the 3rd Meeting of PQQ and Quinoproteins at Capri, 22-27 May 1994.)

MATERIALS AND METHODS
Bacterial strains, phages, plasmids, and growth media. The bacterial strains, phages, and plasmids used in this study are listed in Table 1. The growth media used were Luria broth (LB; 1% tryptone, 0.5% yeast extract, 0.5% NaCl in demineralized water [pH 7]) and minimal medium A (36) supplemented with 0.4% gluconate and the required amino acids and vitamins (25 g/ml). Ampi- VOL. 177, 1995 BIOSYNTHESIS OF PQQ 5089 cillin and kanamycin were used at 50 g/ml, chloramphenicol was used at 34 g/ml, and tetracycline was used at 20 g/ml. Isopropyl-␤-D-thiogalactopyranoside (IPTG) was used as an inducer when pqq genes were placed under control of the tac promoter and to induce T7 RNA polymerase, which was under control of the lacUV5 operator. The methods used for preparing cells for phage stocks and assaying phage and phage DNA were those described by Arber et al. (3). Transformation, digestion, and ligation were performed by standard procedures (34). Restriction and modification enzymes and buffers were obtained from Pharmacia, Biozym, and Gibco BRL. Plasmid DNA was isolated by the alkaline lysis method (34). For large-scale DNA isolations, RNA was removed by LiCl precipitation followed by RNase treatment (29). Construction of K. pneumoniae KA196, KA220, and KA222. To isolate a Tn10 insertion in the K. pneumoniae lacZ gene, K. pneumoniae NCTC418 was made sensitive to bacteriophage by the introduction of plasmid pAMH62 and then infected with NK1098, as described by Way et al. (45). White colonies were selected on Luria agar plates containing tetracycline, 5-bromo-4-chloro-3-indoyl-␤-D-galactopyranoside (X-Gal; 40 g/ml), and IPTG (40 g/ml). The ␤-galactosidase activity of one of these mutants, KA196, was reduced to background levels.
A K. pneumoniae strain defective in only pqqB or pqqC was constructed by transferring the pqqB38::Tn5tac1 allele (from pBCP272) or the pqqC40::Tn5tac1 allele (from pBCP274) to the chromosome of K. pneumoniae KA56 as described elsewhere (24). The resulting strains were designated KA220 and KA222, respectively.
Construction of pqq-lacZ operon fusions. Several pqq-lacZ operon fusions were constructed by incubating E. coli W3350/pBCP138 with b20 (containing a Tn5 with a promoterless lacZ gene in the left-end inverted repeat of Tn5 [IS50L] [37]) for 30 min at 37ЊC and isolating kanamycin-and chloramphenicol-resistant colonies on Luria agar plates. Plasmid DNA from the pooled mutants was transformed into E. coli MC1060, and blue transformants were selected on Luria agar plates containing kanamycin, chloramphenicol, and X-Gal. The location of the Tn5lacZ insertion was determined by restriction analysis. The fusions were transferred to the chromosome of KA196 with ES1 as described elsewhere (24), yielding KA197 (pqq-18::lacZ; insertion between pqqA and pqqB), KA204 (pqqB24::lacZ), and KA202 (pqqE22::lacZ; insertion in the middle of pqqE) (Fig.  1A). The exact positions of the lacZ fusions in KA197 (98 bp downstream of the pqqA start codon in the pqqA-pqqB intercistronic space) and KA204 (340 bp downstream of the pqqB start codon) were determined by sequencing the fusion points.
Construction of plasmids. (i) Plasmids with an incomplete set of pqq genes. Nonpolar insertions of the Tn5tac1 element (9) in the pqqB and pqqC genes of pBCP138 were isolated by infection of E. coli W3350/pBCP138 with ::Tn5tac1. The insertion point of the Tn5tac1 element of the resulting plasmids, pBCP272 (pqqB38::Tn5tac1, insertion approximately 200 bp downstream of the start codon) and pBCP274 (pqqC40::Tn5tac1, insertion approximately 700 bp downstream of the start codon), was determined by restriction enzyme analysis.
Plasmid pBCP168, containing the complete K. pneumoniae pqq operon, was constructed by ligating the 2.6-kb XhoI-HindII fragment of pBCP162 (25) (this fragment contained pqqF and part of pqqE) into pBCP138 digested with XhoI and HindIII. Tn5tac1 insertions were isolated in the pqqB, pqqC, and pqqE genes of pBCP168 by using E. coli W3350/pBCP168. The insertion points of the Tn5tac1 elements in pqqB (pBCP328, approximately 500 bp downstream of the start codon), pqqC (pBCP329, approximately 600 bp downstream of the start codon), and pqqE (pBCP330, approximately 100 bp downstream of the start codon) were determined by restriction enzyme analysis (Fig. 1C).
To construct a plasmid in which Tn5tac1 was inserted closer to the pqqB start codon than in pBCP328, phage ES1 was grown on E. coli ED8654/pBCP272, and the lysate was used to infect E. coli W3350/pBCP164. In the resulting plasmid, pBCP324 (Fig. 1C), correct integration of the Tn5tac1 element was confirmed by restriction enzyme analysis.
A plasmid with a defective pqqD gene was constructed by deleting the internal 63-bp AatII fragment of pqqD of pBCP168. This deletion caused a nonpolar mutation in the pqqD gene, and the resulting plasmid was designated pBCP338 (Fig. 1C).
A plasmid containing all pqq genes except pqqF was constructed by digestion of pBCP168 with AccI and filling in the AccI site with Klenow polymerase. After a second digestion with NheI, the resulting 3.7-kb NheI blunt-ended fragment, which contained the pqq promoter (up to 220 bp upstream of the pqqA start codon), the pqqABCDE genes, and 320 bp of pqqF (i.e., one-seventh of the gene), was ligated into the vector pJF119HE digested with XbaI and SmaI, resulting in pBCP499 (Fig. 1C).
A plasmid containing only the pqqA gene was constructed by ligation of the 1.0-kb PstI-EcoRV fragment of pBCP165 (25), containing pqqA and the pqq promoter, into the vector pJF118HE digested with PstI and SmaI, resulting in pBCP335 (Fig. 1C).
Plasmid pBCP390 (Fig. 1C), containing only the pqqC gene, was constructed by digesting pBCP165 with Tth111I and filling in the Tth111I site with Klenow polymerase. After a second digestion with SphI, the resulting 0.97-kb SphI blunt-ended fragment, which contained pqqC and 213 bp of pqqD, was ligated into the vector pJF119HE digested with SphI and SmaI.
To construct pBCP176 (Fig. 1C), containing only functional pqqA and pqqB genes, the pqqC2::Tn10Km allele from pBCP141 (26) was transferred to pBCP164 by using ES1 as described for the construction of pBCP324 in this section.
A plasmid containing only the functional pqqA, B, and C genes was constructed by deleting the internal HindIII fragment of a pqqD::Tn5lacZ insertion (containing the kanamycin resistance and part of the IS50R sequence [23]). After digestion with EcoRI and SalI, the 6.0-kb fragment (containing the complete pqqA, B, and C genes and part of pqqD::Tn5lacZ) was ligated into pBR322 digested with EcoRI and SalI, resulting in pBCP337 (Fig. 1C).
To construct a plasmid containing pqqABCD, the internal HindIII fragment of a Tn5lacZ insertion in pqqE was deleted. Ligation of the 6.5-kb EcoRI-SalI fragment of this plasmid (containing the pqqA, B, C, and D genes and part of pqqE::Tn5lacZ) into pBR322 digested with EcoRI and SalI resulted in pBCP341 (Fig. 1C).
To obtain a pqqA-lacZ protein fusion, a 0.5-kb DNA fragment containing part of ORFX, the pqq promoter, and 51 nucleotides of pqqA (encoding 17 amino acids) was amplified by PCR. The primer downstream of the pqqA start codon contained a BamHI site, and the primer upstream of the pqqA start codon (in ORFX) contained a SalI site. This PCR fragment was digested with SalI, and the SalI site was filled in with Klenow polymerase. After a second digestion with BamHI, the resulting 0.5-kb fragment was ligated into pNM481 which had been digested with SmaI and BamHI, resulting in pBCP361.
A pqqC-lacZ protein fusion was constructed by ligating the 2.2-kb BamHI fragment of pBCP138, containing the pqq promoter, the pqqA and pqqB genes, and part of the pqqC gene, into pNM481 digested with BamHI. This resulted in pBCP362.
To construct a plasmid with a lacZ fusion in pqqE, pBCP168 was digested with SalI, and the SalI site was filled in with Klenow polymerase. After a second digestion with BglII, the resulting 3.2-kb blunt-ended BglII fragment (containing the pqq promoter, the pqqA, B, C, and D genes, and part of the pqqE gene) was ligated into pNM480 digested with SmaI and BamHI, resulting in pBCP363.
(iii) Plasmid with pqqA behind the T7 promoter. To construct a plasmid with the pqqA gene under control of the T7 promoter, an NdeI site was created at the pqqA start codon by site-directed mutagenesis (21). The mutated fragment was sequenced to confirm that no additional mutations had occurred and cloned (together with the other pqq genes) in pRE1. The resulting plasmid, pBCP352, in which all six pqq genes were under control of the heat-inducible p L promoter, produced PQQ upon heat induction (data not shown). The 1.3-kb NdeI-BamHI fragment of pBCP352 (containing pqqAB and part of pqqC) was ligated into pET-3b (41) digested with NdeI and BamHI. In the resulting plasmid, pBCP364, pqqA was under the control of the T7 promoter.
Preparation of cell extracts. To prepare cell extracts to study in vitro synthesis of PQQ, E. coli JA221 cells containing one or more plasmid-borne pqq genes were grown overnight at 37ЊC in minimal medium A containing gluconate and centrifuged at 10,000 ϫ g for 10 min. The pellet was washed twice with 0.5 volume of 0.9% NaCl and resuspended in 1/100th of the original culture volume of 25 mM potassium phosphate buffer, pH 7, containing 0.5 mM EDTA. The cells were ruptured by passage through an Aminco French pressure cell at 1,000 kg/cm 2 , and the cell extracts were centrifuged for 10 min at 13,000 rpm in an Eppendorf centrifuge to remove cell debris. The supernatant was divided into Eppendorf vials, frozen in liquid N 2 , and stored at Ϫ80ЊC until use. Portions of cell extract were thawed on ice directly before use and used only once. For cell cultures harboring plasmids with a tac promoter in front of pqq genes (except pBCP335), IPTG to a final concentration of 50 M was added during growth. Plasmid pBCP335 contained the tac promoter upstream of the pqq promoter and the pqqA gene. Since the pqq promoter alone was sufficient to direct expression of pqqA, no IPTG was added.
Preparation of culture supernatant for measurement of PQQ and intermediate of PQQ biosynthesis. To measure PQQ production and excretion into the medium, E. coli JA221 containing one or more plasmid-borne pqq genes or K. pneumoniae NCTC418/pBCP165 was grown overnight at 37ЊC in minimal medium A containing gluconate, in the presence or absence of 50 M IPTG. The cells were centrifuged at 13,000 rpm in an Eppendorf centrifuge for 5 min, and the culture supernatant was used for the measurement of PQQ.
To measure the production and excretion of the PQQ biosynthesis intermediate, E. coli JA221/pBCP329 or K. pneumoniae KA222/pBCP329 cells were grown in minimal medium A containing gluconate in the presence of 50 M IPTG at 37ЊC until the late exponential-early stationary phase (final optical density at 600 nm was 0.8 to 1.1). The cells were centrifuged at 13,000 rpm in an Eppendorf centrifuge for 5 min. The supernatant was immediately stored on ice and assayed within 1 h for the presence of the intermediate.
In vitro synthesis of PQQ. To measure PQQ synthesis in vitro, 50 l of cell extract (one extract or a combination of two extracts; 0.1 to 0.6 mg of protein) was added to 150 l of 100 mM 1,4-piperazinediethane sulfonic acid (PIPES) in a 2-ml Eppendorf vial and incubated at 37ЊC with shaking. At various times, the reaction was stopped by adding HClO 4 to a final concentration of 5% (vol/vol). After incubation on ice for 20 to 60 min followed by neutralization with 5 M KOH to pH 7, the reaction mixture was incubated on ice for another 10 min and centrifuged for 5 min at 13,000 rpm in an Eppendorf centrifuge to remove the KClO 4 (12). The assays were slight modifications of those described elsewhere (42) and allowed the determination of PQQ concentrations in the range of 0.6 to 15 nM (apo-GCD) and 2 to 50 nM (apo-EDH). A calibration curve was made with PQQ (Fluka) dissolved in minimal medium A. For cell extracts, the amount of PQQ was expressed as picomoles per milligram of protein. As a consequence, the detection level was 0.4 pmol/mg of protein with apo-GCD and 1.3 pmol/mg of protein with apo-EDH.
(i) GCD assay. First, the sample (50 l) was mixed with 120 l of 0.1 M Tris-HCl (pH 7.5) containing 3 mM CaCl 2 and 0.02 M apo-GCD and incubated for 5 to 15 min at room temperature. Then, a 0.1 M Tris-HCl (pH 7.5) solution containing 3 mM CaCl 2 , 1.2 mM phenazine methosulfate, and 0.063 mM 2,6dichlorophenolindophenol was added to give a total volume of 950 l. The reaction was started by adding 50 l of 1 M glucose in demineralized water, and the decrease in A 600 was measured.
(ii) EDH assay. PQQ was determined with apo-EDH on a Cobas Bio automatic analyzer (Hoffmann-La Roche). The sample (80 l) and demineralized water (15 l After incubation for 30 min at 37ЊC with shaking, the mixture was transferred to a 1-ml cuvette, and the assay was continued as described above for the PQQ assay. To correct for the possible PQQ already present, the sample was also assayed for PQQ with apo-GCD. ␤-Galactosidase assay. ␤-Galactosidase activity was measured as described elsewhere (27), and the activity was expressed as nanomoles of o-nitrophenyl-␤-D-galactopyranoside (ONPG) hydrolyzed per minute per milliliter of cells (optical density at 600 nm is 1).
Synthesis of PqqA. E. coli BL21(DE3), carrying on its chromosome the gene for T7 RNA polymerase under lacUV5 control, was transformed with pBCP364, containing pqqA behind the T7 promoter. The transformed cells were grown in LB at 37ЊC to an optical density at 600 nm of 0.8. After induction of the T7 polymerase gene by the addition of IPTG to a final concentration of 400 M, the cells were grown for an additional hour, harvested by centrifugation, and subjected to tricine-sodium dodecyl sulfate-polyacrylamide gel electrophoresis (tricine-SDS-PAGE) (35).
Protein determination. The amount of protein was determined with the bicinchoninic acid (Sigma) method (39). The assay was carried out according to the instructions of the manufacturers, on a Cobas Bio automatic analyzer (Hoffmann-La Roche), with bovine serum albumin as a standard.

RESULTS
Expression of pqq-lacZ operon fusions. The expression of the K. pneumoniae pqq operon was studied with the help of several chromosomal pqq-lacZ operon fusions (see Fig. 1A). Cells were grown in LB and harvested at the exponential phase. Table 2 shows that the fusions located close to the pqq promoter had a higher ␤-galactosidase activity than the fusions further downstream. The highest ␤-galactosidase activity, that of the pqq-lacZ fusion located in the intercistronic space between pqqA and pqqB (KA197), was 15-fold lower than the induced wild-type ␤-galactosidase activity in K. pneumoniae NCTC418.
PQQ synthesis under aerobic and anaerobic conditions. To investigate the role of molecular oxygen in PQQ biosynthesis, we measured PQQ production under aerobic and anaerobic culture conditions. To switch the culture to anaerobic conditions, the cells were diluted 1:50 into fresh medium and flushed with N 2 for 30 min. Because the PQQ level in a wild-type K. pneumoniae strain is low and close to the detection level, we used wild-type K. pneumoniae NCTC418 harboring pBCP165, containing the complete pqq operon. Under anaerobic conditions, little PQQ was detected in the culture supernatant (12 nM) compared with aerobic conditions (540 nM). The small amount of PQQ detected under anaerobic conditions could be derived from the (aerobically grown) preculture.
Since the failure to synthesize PQQ under anaerobic conditions could be due to the lack of expression of the pqq genes, two chromosomal pqq-lacZ operon fusions were investigated. However, anaerobiosis had no significant effect on the ␤-galactosidase activity in KA197 (pqq-18::Tn5lacZ) and KA204 (pqqB24::Tn5lacZ) ( Table 2).
Expression of pqqA. To investigate whether the pqqA gene encoded a polypeptide, the pqqA gene was cloned behind the strong, inducible T7 promoter. E. coli BL21(DE3) cells containing the resulting plasmid, pBCP364, produced a polypeptide of the size predicted for PqqA (2.7 kDa) upon induction with IPTG, whereas uninduced cells did not produce such a polypeptide (Fig. 2).
If PqqA is the precursor for PQQ biosynthesis, it would be  expected that pqqA would encode a polypeptide which is produced in higher amounts than the other Pqq proteins. To compare the expression of the different pqq genes, lacZ fusions were constructed with pqqA (pBCP361), pqqC (pBCP362), and pqqE (pBCP363) (see Fig. 1B). The activity of the fusion proteins was measured in E. coli MC1060. The lacZ fusion in pqqA resulted in a 20-fold-higher ␤-galactosidase activity (500 nmol of ONPG/min/ml of culture) than the lacZ fusions in pqqC and pqqE (23 and 19 nmol of ONPG/min/ml of culture, respectively).

In vivo complementation and growth studies.
In vivo complementation studies were used to investigate whether all six pqq genes (pqqABCDEF) were necessary for PQQ production and excretion and to test the functionality of the plasmids used in this study. All plasmids used are shown in Fig. 1C. For in vivo complementation, two compatible plasmids, each containing an incomplete set of pqq genes, were transformed together into the E. coli recA strain JA221. As a control, each of the plasmids was transformed separately. The cells containing the various pqq plasmids were grown overnight in minimal medium containing gluconate, and PQQ was measured in the culture supernatant. Table 3 shows that all plasmid combinations in which at least one copy of each of the six pqq genes was present resulted in PQQ synthesis and excretion. No PQQ was detected in supernatants from cell cultures harboring only a single plasmid which lacked either pqqA, C, D, or E. In super-natants of cell cultures harboring plasmids lacking pqqB (pBCP324 and pBCP328) or pqqF (pBCP186 and pBCP499), small amounts of PQQ, only slightly above the detection level, were measured (Table 3).
To study whether all pqq genes are required for growth on glucose minimal medium via glucose dehydrogenase, we transformed E. coli ZSC112, which is unable to grow on glucose because of a ptsM and ptsG mutation, with various plasmids lacking one of the six pqq genes. Growth on glucose was not stimulated by any of these plasmids, whereas the control plasmid pBCP165 (pqqABCDEF) stimulated growth (Table 3).
In vitro PQQ synthesis. The role of the various Pqq proteins in PQQ biosynthesis was studied with the help of an in vitro system in which a cell extract containing all but one of the Pqq proteins was combined with an extract containing the missing Pqq protein. All plasmids used for the in vitro studies are shown in Fig. 1C. The presence of PQQ was detected with two different apo-enzymes specific for PQQ, apo-GCD and apo-EDH. The PQQ values determined with apo-GCD and apo-EDH agreed. A cell extract lacking all six Pqq proteins contained less than 0.4 pmol of PQQ per mg of protein, whereas a cell extract with all six Pqq proteins, PqqA, B, C, D, E and F, contained approximately 12 pmol of PQQ per mg of protein.
In the latter case, the amount of PQQ did not increase with prolonged incubation (Table 4 and Fig. 3A). The intracellular PQQ concentration was calculated to be approximately 3.5 The cells were grown overnight in minimal medium A containing 0.4% gluconate in the absence or presence of 50 M IPTG. Cells were harvested at an optical density at 600 nm of 1.2. PQQ was measured enzymatically in the culture supernatant with apo-GCD and apo-EDH. Only the PQQ concentrations obtained with apo-GCD are given. Growth on glucose minimal medium plates was judged after incubation at 37ЊC for 48 h; ϩ, growth; Ϫ, no growth. Deletion or inactivation of a particular pqq gene is indicated by a dash at the appropriate position.
VOL. 177, 1995 BIOSYNTHESIS OF PQQ 5093 M, assuming that 1 mg of total cell protein is equivalent to an internal volume of 3.3 l (40) and that all PQQ is localized in the cytoplasm. In the case of E. coli JA221/pBCP165 (pqqABC DEF) , the PQQ concentration in the medium was 180 nM (the optical density at 600 nm was 1.2 when the cells and the supernatant were harvested). This means that more than 98% of the PQQ produced by the culture was present in the medium, assuming that an optical density at 600 nm of 1.0 corresponds to an internal volume of 600 l per liter of culture (40). In vitro complementation. Using cell extracts that contained all Pqq proteins except one, we investigated whether PQQ synthesis could be restored by adding a second extract containing the missing Pqq protein. Cell extracts lacking PqqA, D, or E did not contain or synthesize PQQ and could not be complemented in vitro by a cell extract containing the missing protein (Table 4).
In cell extracts lacking PqqF (pBCP186), the amount of PQQ was near the detection level (Table 4). Since pBCP186 contained a Tn10 in the middle of pqqF, possibly resulting in a truncated but partially active PqqF protein, pBCP499, which contained only 320 bp of pqqF (one-seventh of the gene), was constructed. Table 4 shows that small amounts of PQQ were present even when pqqF was almost completely deleted. In a cell extract lacking PqqF (pBCP186), PQQ synthesis could not be restored by the addition of a second extract containing PqqF (Table 4).    Table 4 also shows that extracts lacking PqqC could be complemented with extracts containing PqqC and that extracts lacking PqqB produced PQQ. We will discuss this in more detail below.
In vitro complementation of a cell extract lacking PqqC. An extract containing all Pqq proteins except PqqC could be complemented in vitro by an extract containing PqqA, B, and C (pBCP337, Table 4). The production of PQQ reached its maximum within 30 min (Fig. 3B). A plasmid that produced only PqqC, pBCP390, also restored PQQ synthesis. The truncated pqqD gene from pBCP390 was not functional, since it could not complement pBCP338 [pqqABC(⌬D)EF] in vivo, the PQQ concentration in the culture supernatant being less than 0.6 nM.
We determined the amount of PQQ produced and its production rate in extracts containing all Pqq proteins except PqqC, supplemented with an extract containing PqqC as the only Pqq protein. The rate of PQQ production increased with increasing amounts of PqqC-containing extract when the amounts of PqqA, B, D, E, and F were kept constant. The same amount of PQQ was produced (Fig. 4A). When the amount of cell extract containing all Pqq proteins except PqqC was increased while keeping the amount of PqqC-containing extract constant, the rate of PQQ production increased. The amount of PQQ produced also increased with the amount of extract containing all proteins except PqqC (Fig. 4B). These results suggested that cells which lacked PqqC formed an intermediate in PQQ biosynthesis which could be converted into PQQ by a PqqC-containing cell extract. This was studied in more detail by measuring the amount of the intermediate and PQQ at the start and at the plateau (after 30 min) of the reaction. Table 5 shows that during this in vitro complementation reaction, the intermediate was converted into PQQ. PQQ production by cell extracts lacking PqqB. Since strains lacking PqqB produced amounts of PQQ barely above the detection level, it came as a surprise that an extract from E. coli carrying a plasmid lacking PqqB (pBCP328) produced PQQ (6.5 pmol of PQQ per mg of protein; Table 4). The maximal  a Cell extract (one extract or a combination of extracts) was incubated as described in Materials and Methods. The reaction was stopped by HClO 4 /KOH treatment at the time indicated (0, 30, or 45 min). After removal of the KClO 4 precipitate, the supernatant was assayed for PQQ and assayable PQQ biosynthesis intermediate, as described in Materials and Methods, with apo-GCD. Values are based on the protein contents of the sample, e.g., single extract or the total protein content of the extract lacking PqqC combined with the PqqCcontaining extract.

Excretion of intermediate by PqqC-lacking cells. Studies with E. coli cells harboring a plasmid that encoded all
b Deletion or inactivation of a particular pqq gene is indicated by a dash at the appropriate position.
VOL. 177, 1995 BIOSYNTHESIS OF PQQ 5095 amount of PQQ produced was reached after 45 min. A small amount of PQQ was detectable at the start of the experiment (Table 4 and Fig. 3C). Since in pBCP328 the Tn5tac1 element was inserted in the middle of the pqqB gene possibly producing a truncated but still active PqqB protein, pBCP324 [pqqA(B38:: Tn5tac1)CDEF], in which a Tn5tac1 element was inserted 200 bp downstream from the start codon of pqqB, leaving only one-fifth of the functional gene intact, was constructed. Extracts made from cells harboring pBCP324 produced amounts of PQQ comparable to those in an extract made from cells harboring pBCP328 (Table 4). Addition of an equal amount of a cell extract containing PqqB (pBCP176) to a pBCP328-derived extract resulted in a small stimulation of PQQ production ( Table 4) because the Tn10 insertion in the pqqC gene was not completely polar (data not shown). This increase in PQQ production became evident when the amount of PQQ was expressed as picomoles per milligram of protein of the PqqBlacking extract rather than per milligram of protein of the sum of the protein contents of both extracts (as is done in Table 4). Calculated in this way, combination of a PqqB-lacking cell extract (pBCP328) with an extract containing PqqB (pBCP176) produced twice as much PQQ (9 pmol of PQQ per mg of protein in 30 min). Cell extracts of K. pneumoniae KA220, which lacks PqqB, containing pBCP324 produced PQQ in amounts comparable to the amounts produced in a cell extract from E. coli JA221/ pBCP328, varying from 0.6 pmol of PQQ per mg of protein at the start of the experiment to 13.5 pmol of PQQ per mg of protein after 45 min. The supernatant of K. pneumoniae KA220/pBCP324 cells contained little PQQ (concentration of PQQ was less than 5 nM) compared with K. pneumoniae NCTC418 harboring pBCP165 (concentration of PQQ was 540 nM).
Studies with E. coli and K. pneumoniae cell extracts lacking PqqB showed that they contained the same intermediate of PQQ biosynthesis as cell extracts from PqqC-lacking cells. In vitro, PQQ production in PqqB-lacking extracts was stopped at different times by the addition of HClO 4 , and the amount of PQQ and PQQ biosynthesis intermediate in the supernatant (after KClO 4 removal) was determined. At the start of the experiment, this supernatant contained very little PQQ (see above), but when PqqC was added, PQQ was formed (

DISCUSSION
The synthesis of PQQ and its role as a cofactor in several dehydrogenases have been demonstrated in a number of bacteria (for a review, see reference 20). Although a number of pqq genes involved in PQQ biosynthesis have been isolated from several bacteria, including A. calcoaceticus (15,17), K. pneumoniae (25,26), M. extorquens (28), and Erwinia herbicola (22), the function of these genes in PQQ biosynthesis is unknown at present. The six K. pneumoniae pqq gene products show no similarity to other proteins in the database except for PqqF, which shows similarity with protease III from E. coli and some insulin-degrading enzymes (26). Interestingly, the three pqq operons that have been analyzed in some detail all contain a small gene (pqqA in K. pneumoniae) that could encode a polypeptide of 23 to 29 residues. All three polypeptides contain a glutamate and a tyrosine residue at conserved positions. Possible pathways for PQQ biosynthesis starting with a tyrosine and a glutamate residue have been proposed (19,44).
In this paper, we have examined the role of each of the six K. pneumoniae pqqABCDEF genes in PQQ biosynthesis. Using an in vitro system, we have also detected an intermediate in PQQ biosynthesis, and we have shown that the PqqC protein probably catalyzes the last step in PQQ synthesis.
The role of each of the K. pneumoniae pqq genes in PQQ biosynthesis in intact cells and in metabolism via a PQQ-dependent pathway was studied in E. coli since E. coli can synthesize apo-glucose dehydrogenase, which oxidizes glucose to gluconate, but not its cofactor, PQQ. Consequently, an E. coli pts mutant, which cannot metabolize glucose via the phosphotransferase system (the major pathway for glucose metabolism), grows slowly on glucose when PQQ is added to the growth medium or when a plasmid which contains the pqq operon from K. pneumoniae is present (25). Our studies revealed that each of the six K. pneumoniae pqqABCDEF genes is required for growth on glucose via the glucose dehydrogenase-dependent pathway and for substantial PQQ secretion into the medium. It is important to note that the pqqA gene complemented in trans and was required for PQQ synthesis and excretion. This is in agreement with the hypothesis that the pqqA gene encodes the precursor polypeptide for PQQ.
Our data show that almost no PQQ was synthesized by K. pneumoniae harboring a plasmid containing the K. pneumoniae pqqABCDEF genes under anaerobic growth conditions, although the expression of several pqq-lacZ operon fusions was not impaired, suggesting that the Pqq enzymes were synthesized under anaerobic conditions. Most likely, a hydroxylase, requiring molecular oxygen, is involved in the biosynthesis of PQQ for the formation of the quinone groups (19,44). We cannot presently exclude the possibility, however, that one or more enzymes involved in PQQ biosynthesis are inactive in the absence of oxygen.
Using pqq-lacZ operon fusions localized on the K. pneumoniae chromosome, we also studied the expression level of the K. pneumoniae pqq genes. The ␤-galactosidase activity decreased about sevenfold within the pqq operon, fusions located at the end of the operon having the lowest activity. These results confirm our earlier conclusion that besides the pqqA promoter, which was mapped by primer extension analysis to lie upstream of pqqA (26), no other strong internal promoters were present. The ␤-galactosidase activity of the different pqq-lacZ operon fusions indicated that the transcription of the pqq genes was low, the highest activity being that of the pqq-lacZ fusion located between pqqA and pqqB. The value was 5-to 10-fold lower than that of lacZ fusions to K. pneumoniae genes encoding metabolic enzymes, such as the sor (sorbose) and gut (D-glucitol) genes (46).
To study in more detail the role of the various Pqq proteins in PQQ biosynthesis, we have developed an in vitro system for PQQ synthesis by combining extracts containing all but one of the Pqq proteins with an extract containing the missing protein. An E. coli cell extract made from cells in which all six Pqq proteins were present contained 12.0 Ϯ 3.0 pmol of PQQ per mg of protein. Extracts lacking the PqqA, PqqC, PqqD, PqqE, or PqqF protein contained no PQQ or amounts below the detection level (except maybe in the case of PqqF; see below). A certain amount of PQQ was detected in extracts of PqqBdeficient cells, however.
In vitro complementation could be clearly demonstrated in the case of PqqC. PQQ was produced when a cell extract containing all Pqq proteins except PqqC was combined with a cell extract that contained PqqC. The separate extracts produced no PQQ. These results strongly suggest that an intermediate in PQQ synthesis had accumulated in cells lacking PqqC. The putative intermediate was also detected in the culture medium of E. coli and K. pneumoniae cells lacking PqqC and could be converted into PQQ with a cell extract containing only PqqC. This result suggested that PqqC is the last enzyme of the pathway and that the intermediate is a PQQ-like molecule rather than a polypeptide resembling PqqA. However, it cannot be completely excluded at present that other enzymes, not encoded by the known pqq genes but present in E. coli and K. pneumoniae, are required for the conversion of the putative intermediate into PQQ. At present, we are purifying and characterizing the detected intermediate.
In all other cases, reconstitution of PQQ biosynthesis by combining the various extracts was not successful. It is important to note that in all cases, in vivo complementation was observed with the same plasmids from which the Pqq proteins in these cell extracts were derived. Possibly, complexes between two or more Pqq proteins have to be formed for proper functioning, a process that may occur only during the synthesis of these proteins in the intact cell. Alternatively, the concentration of one or more Pqq proteins may be too low in the extracts compared with their concentration in an intact cell. Finally, some intermediates in PQQ biosynthesis, when accumulated in the various pqq mutants, as well as one or more of the Pqq proteins might be unstable under the conditions used to prepare and incubate the extracts.
Our studies on the role of the PqqB protein have yielded unexpected results. E. coli cells containing (on a plasmid) all pqq genes except pqqB excreted little if any PQQ into the growth medium (Table 3). Similarly, an E. coli mutant unable to grow on glucose via the phosphotransferase pathway could not metabolize glucose via the PQQ-dependent glucose dehydrogenase pathway if the pqqB gene was lacking. These results point to an essential role for PqqB in PQQ biosynthesis. To our surprise, however, a cell extract, containing all Pqq proteins except PqqB could produce PQQ in vitro in a time-dependent manner. It should be noted, however, that the rate of PQQ production in a PqqB-lacking cell extract is relatively low compared with that catalyzed by the PqqC-containing extract (compare Fig. 3B and C). PqqB homologs have been found in A. calcoaceticus (PqqV) and M. extorquens (PqqG), but conflicting results about their role have been reported. The A. calcoaceticus PqqV protein was reported not to be necessary for growth via a PQQ-requiring pathway (15,17), whereas in the case of M. extorquens AM1, it was concluded that the PqqG protein was required for PQQ biosynthesis (28), similar to PqqB in E. coli.
These conflicting results may be explained by our observations with extracts made from cells lacking the pqqB gene. These cells contained the same intermediate which we have detected in PqqC-deficient cells but could not convert it into PQQ, although functional PqqC was present. Furthermore, the intermediate could hardly be detected in the growth medium of these PqqB-deficient cells, although the intracellular concentration was comparable to that of PqqC-lacking cells. Pos-sibly, PqqB is involved in the transport of PQQ across the cytoplasmic membrane into the periplasm. Since there is no evidence that PqqB contains hydrophobic stretches, it is unlikely that PqqB itself can transport PQQ across the membrane. However, PqqB could modify an existing transport system so that secretion of PQQ becomes possible. Lack of PqqB could cause accumulation of PQQ in the cytoplasm and subsequent inhibition of PqqC activity, resulting in an increased concentration of the intermediate in the cytoplasm. This hypothesis would also explain why in a cell extract made from PqqB-deficient cells, in which the cell contents (e.g., PQQ) become diluted, PqqC would become active. The PqqB-dependent transport system might also recognize the intermediate. As a consequence, the intermediate would be secreted by cells lacking PqqC but containing PqqB. This is in agreement with our findings.
We have shown that E. coli cell extracts made from cells containing all Pqq proteins except the protease III-like PqqF protein contained a small amount of PQQ, just above the detection limit. These cells also produced some PQQ in the culture medium, although the final concentration was at least 100-fold lower than that produced by cells harboring all six Pqq proteins, suggesting that small but measurable amounts of PQQ might be produced in the absence of PqqF. In previous studies, we reported that the PqqF protein is necessary for the substantial conversion of glucose into gluconate via PQQ-dependent glucose dehydrogenase, which is required for growth of a K. pneumoniae ptsI mutant on glucose via this pathway (26). We have recently observed, however, that a plasmid containing the K. pneumoniae pqqABCDE genes but lacking the pqqF gene restored growth on glucose of ptsI derivatives of E. coli JM109 and HB101 but not of some other E. coli K-12 strains (3a). Possibly, protease III or other protease III-like enzymes can, to a limited extent, substitute for PqqF in PQQ biosynthesis, i.e., produce small amounts of PQQ. This might explain the observation by Goosen and coworkers (15) that a plasmid containing the five known A. calcoaceticus pqq genes, which showed similarity to pqqA, B, C, D, and E of K. pneumoniae, restored growth of an E. coli ptsI mutant on glucose minimal medium. It is important to note, however, that not all E. coli strains supplied with all pqq genes except pqqF on a plasmid can synthesize PQQ in amounts sufficient to support growth on glucose via glucose dehydrogenase. Thus, this proposal requires that the enzyme substituting for PqqF be present in some E. coli strains at higher levels than in others. We are presently in the process of identifying this PqqF-substituting enzyme.
We have mentioned previously the hypothesis that the small PqqA polypeptide might be a precursor in PQQ biosynthesis. This would require synthesis of PqqA in stoichiometric amounts rather than catalytic amounts compared with the other Pqq proteins. Using a plasmid with the pqqA gene cloned behind an inducible T7 promoter, we could demonstrate that a polypeptide with a mobility expected for PqqA is indeed synthesized. The level of expression of various pqq-lacZ protein fusions demonstrated clearly that expression of the pqqA gene was much higher (at least 20-fold) than the expression of other pqq genes like pqqC and pqqE. The drop in expression of the genes downstream of pqqA might be caused by transcriptional termination within the operon. This is supported by analysis of the mRNA sequence between pqqA and pqqB, which revealed a hairpin structure (between nucleotides 1034 and 1053 of the published sequence of the pqq operon [26]). A hairpin was also found downstream of pqqIV (15) and pqqD (30), the genes corresponding to pqqA in A. calcoaceticus and M. extorquens AM1, respectively. In M. extorquens AM1, the transcript en-VOL. 177, 1995 BIOSYNTHESIS OF PQQ 5097 coding PqqD was more abundant than the transcripts encoding PqqD and PqqG (the homolog of K. pneumoniae PqqB) together (30). In conclusion, we think it is likely that the mRNA which terminates at the hairpin codes for a PqqA polypeptide. Together with the relatively high expression of the pqqA gene compared with that of the other pqq genes, this supports the hypothesis that PqqA is the precursor for PQQ synthesis.