The Enigma of Cobalamin (Vitamin B 12 ) Biosynthesis in Porphyromonas gingivalis IDENTIFICATION AND CHARACTERIZATION OF A FUNCTIONAL CORRIN PATHWAY*

The ability of Porphyromonas gingivalis to biosynthesize tetrapyrroles de novo has been investigated. Extracts of the bacterium do not possess activity for 5-aminolevulinic-acid dehydratase or porphobilinogen deaminase, two key enzymes involved in the synthesis of uroporphyrinogen III. Similarly, it was not possible to detect any genetic evidence for these early enzymes with the use of degenerate polymerase chain reaction. However, the bacterium does appear to harbor some of the enzymes for cobalamin biosynthesis since cobyric acid, a pathway intermediate, was converted into cobinamide. Furthermore, degenerate polymerase chain reaction with primers to cbiP , which encodes cobyric-acid synthase, produced a fragment with a high degree of identity to Salmonella typhimurium cbiP . Indeed, the recently released genome sequence data confirmed the presence of cbiP together with 14 other genes of the cobalamin pathway. A number of these genes were cloned and functionally characterized. Although P. gingivalis harbors all the genes necessary to convert pre-corrin-2

Anaerobic bacteria constitute almost 95% of the indigenous flora of the alimentary tract. Among these organisms, several opportunistic pathogens have been described, which are mainly Gram-negative anaerobic rods of the family Bacteroidaceae (1). These bacteria are also associated with the colonizing flora of the oral cavity and are quite similar in terms of their physiological properties, requirement for heme, and utilization of mainly amino acids and peptides as energy sources. Furthermore, disease caused by infection of anaerobic bacteria is an area of growing concern, although little is known about the fundamental biology of many of these organisms (2). Periodontal diseases result in gum tissue destruction and, in some cases, the spread of the infective anaerobic flora into the blood, which can cause soft tissue infections in the pulmonary tract, brain abscesses, and heart disease (3).
Porphyromonas gingivalis, one of the main causes of such infections, is an obligate anaerobe whose colonies are recognized by their characteristic heme pigments on blood agar plates (1). Heme is central to metabolism in nearly all organisms, especially in its role as a prosthetic group involved in electron transport. Its synthesis occurs via a branched biosynthetic pathway, which is also responsible for the synthesis of other modified tetrapyrroles including cobalamin (vitamin B 12 ) ( Fig. 1) (4). P. gingivalis exhibits an absolute growth requirement for heme in vitro (5), although it is not known whether this exogenous heme is used to complement an inability of the organism to make its own heme or whether, for instance, the exogenous heme is being used as a source of iron. Similarly, recent evidence also indicates that P. gingivalis requires adenosylcobalamin as a prosthetic group (6), but again it is not known whether the bacterium is able to make its own cobalamin. The biologically active forms of vitamin B 12 , methylcobalamin and adenosylcobalamin, are required in methylation and rearrangement reactions, respectively (7). The biosynthesis of cobalamin represents one of the most complex pathways in Nature, involving ϳ30 enzyme-catalyzed reactions. These include the genes encoding enzymes responsible for the synthesis of uroporphyrinogen III (the first macrocyclic intermediate in tetrapyrrole synthesis), the genes encoding enzymes for the transformation of uroporphyrinogen III into cobinamide, and the genes encoding enzymes for nucleotide loop assembly and attachment to the corrin ring (8 -11). The steps required for the synthesis of cobalamin are highlighted in Fig. 2.
Studies over the past decade have demonstrated that there are at least two quite distinct routes for cobalamin synthesis, representing oxygen-dependent (aerobic) and oxygen-independent (anaerobic) pathways (12,13). For the biosynthesis of cobinamide, the genes of the oxygen-dependent route are prefixed cob, whereas the genes for the oxygen-independent pathway are prefixed cbi (see Fig. 2). 1 The major differences between the two pathways relate to the requirement for molecular oxygen and the timing of cobalt insertion: in the * This work was supported by the Wellcome Trust and the Biotechnology and Biological Sciences Research Council. 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.
** To whom correspondence should be addressed. oxygen-dependent pathway, molecular oxygen is required to assist in the ring contraction process, and cobalt is added relatively late in the pathway, only after the synthesis of hydrogenobyrinic acid a,c-diamide, whereas in the oxygen-independent pathway, molecular oxygen is not required, and cobalt is inserted into the macrocycle at an early stage of the synthesis, at the level of precorrin-2 (see Fig. 2).
In this study, investigations into the ability of P. gingivalis to make tetrapyrroles (and in particular, cobalamin) are reported.
Although the organism appears to contain a functional oxygenindependent corrin biosynthetic pathway, which is able to transform precorrin-2 into cobalamin, the presence of the pathway is compounded by the lack of any enzymes for the synthesis of precorrin-2.

EXPERIMENTAL PROCEDURES
Materials-Horse blood was bought from Difco, and other media components were purchased from Oxoid (Hampshire, United Kingdom). The RapID ANA II diagnostic system was from Innovative Diagnostic Systems (Norcross, GA). Perchloric acid was purchased from Aldrich, and other chemicals were from Sigma (Dorset, UK). All general nucleic acid protocols were carried out as described (14). Protein was quantitated according to the Bradford procedure (15). Cobyric acid was synthesized as described by Bonnett et al. (16).
Bacterial Growth-All strains and plasmids are listed in Table I. The bacterial species P. gingivalis W83, W50, PgM3, PgM4, Hg405, Hg1241 and Hg189; Porphyromonas endodontalis Hg189; and Prevotella loescheii NCTC11321 were used in this study. Cells were routinely grown on Columbia blood agar plates (40 g of Columbia blood agar powder/liter and 5% (v/v) horse blood) or fastidious anaerobe agar in gas jars under anaerobic conditions (Anaerogen, Oxoid) at 37°C for 48 h. Broth cultures of Porphyromonas species (nutrient broth or defined medium (17) supplemented with 8 mM hemin plus 3 mM menadione) were inoculated with plate-grown cells and incubated anaerobically for 48 h. Cells were harvested either by scraping the colonies off the agar or by centrifugation at 5000 ϫ g for 15 min.
Genomic DNA Preparations-Genomic DNA was prepared according to standard procedures (14), except that excess proteinase K and RNase were added during the lysozyme treatment stage since many of these bacteria produce a high concentration of nuclease.
5-Aminolevulinic-acid (ALA) Dehydratase Assay-Harvested cells were washed twice in degassed assay buffer (50 mM CHES 2 (pH 8.5) containing 10 mM dithiothreitol, 5 mM MgCl 2 , and 100 mM ZnCl 2 ). Washed cells were harvested by centrifugation at 10,000 ϫ g for 10 min at 4°C. Cells were disrupted in a high-speed vibratory tissue disintegrator (Mickle Laboratory Engineering Co. Ltd., Surrey, UK). Samples were vibrated with 1 volume of Ballotini No. 10 glass beads for 12 min at 4°C. The resultant homogenate was centrifuged at 4000 ϫ g for 5 min at 4°C, and the clarified supernatant was removed. The beads were washed five times in 1 ml of assay buffer and recentrifuged, and the washings were combined with the crude homogenate sample.
ALA dehydratase activity was determined by the amount of porphobilinogen (PBG) formed using a stopped spectrophotometric assay. The method was essentially as described previously (18), except that ALA (5 mM final concentration) was preincubated for 2 min in assay buffer at 37°C, and the reaction was started by the addition of bacterial extracts (up to 2 mg of protein) to give a final volume of 0.5 ml.
Porphobilinogen Deaminase Assay-PBG deaminase activity was measured according to the method described by Jordan et al. (19). Briefly, crude cell extract (500 l) was incubated at 37°C in a final volume of 800 l containing 100 mM Tris-HCl (pH 8.0). The reaction was started by the addition of 0.1 mol of PBG. The reaction was stopped after either 30 or 60 min by the addition of 0.2 ml of 5 N HCl. Ten microliters of freshly prepared benzoquinone (1 mg/ml) was added to the solution. The mixture was centrifuged; and after 15 min, the solution was measured at 405 nm to quantify the amount of uroporphyrin formed using an extinction coefficient of 5.48 ϫ 10 5 M Ϫ1 cm Ϫ1 .
CbiK Assay-The activities of the long and short versions of CbiK, CbiK L and CbiK S , respectively, were determined using an in vitro linked assay similar to that described previously (20). The incubation contained 1 mg of recombinant Pseudomonas denitrificans uroporphyrinogen III methyltransferase (CobA), 0.1 mg of Escherichia coli PBG deaminase, 0.15 mg of E. coli uroporphyrinogen III synthase, 0.75 mg of S-adenosyl-L-methionine, and 0.1 mg of PBG. The assay components were mixed in 1 ml of degassed 50 mM Tris-HCl (pH 7.8). To this assay were added 0.1 mg of CbiK and 40 M CoCl 2 ⅐6H 2 O as required. Spectra were recorded using a Hewlett-Packard 352 photodiode array spectrophotometer over the range 300 -700 nm.
Cobalamin Bioassay-Cobalamin concentrations were determined FIG. 2. Cobalamin biosynthetic pathways. The aerobic and anaerobic pathways are outlined, highlighting the gene products required for the synthesis of adenosylcobalamin from uroporphyrinogen III. The aerobic pathway has been completely elucidated in P. denitrificans, and the intermediates are outlined on the right, whereas the anaerobic pathway has been partially resolved, and its intermediates are outlined on the left. The genes found in P. gingivalis are shown on the far left and correspond to the genes found in bacteria such as S. typhimurium that harbor an anaerobic pathway. Underlined proteins indicate those overproduced in this study. using a modified bioassay. Bioassay plates were prepared as described previously (21) with two different indicator strains (Salmonella typhimurium metE cysG, AR3612; and S. typhimurium cbiB metE, AR2680). AR2680 requires cobinamide or later intermediates for restoration of growth, whereas AR3612 can grow in the presence of the earlier intermediate cobyric acid. This metabolic difference provided a means of determining if exogenously provided cobyric acid was being taken up and converted into cobinamide by the bacteria under investigation.
Where possible, bacteria that were to be tested for the presence of cobalamin were grown in defined medium (17). However, not all strains were amenable to growth in defined medium and had to be grown on Columbia blood agar plates. In this case, the background level of cobalamins available from the media had to be taken into account, and the results were based on the conversion of exogenously added cobyric acid (100 or 1000 nmol/plate) into cobinamide.
PCR-Whole genomic DNA from P. gingivalis W83 was used as a template for initial isolation of porphinoid genes by PCR. Degenerate PCR was used to amplify a partial gene sequence of ϳ365 base pairs encoding a fragment of cbiP. Degenerate PCR primers were designed based on a highly conserved region of cbiP from known bacterial sequences (between amino acids 59 and 174 of S. typhimurium cbiP) using CODEHOP (consensus-degenerate hybrid oligonucleotide primers for amplification of distantly related sequences) (22). Degenerate primers to two other genes, hemB (encoding ALA dehydratase) and btuR (encoding the adenosyltransferase), were also designed (Table II); but although these worked on control samples of DNA (S. typhimurium genomic DNA), they failed to amplify a product with genomic DNA from P. gingivalis. Primers to known DNA sequences were used to isolate cbiGF, cbiJD, cbiL, and cbiK (Table II).
Reactions were carried out in 50 l containing NH 4 buffer, 1.5-4 mM MgCl 2 , 40 mM each deoxynucleotide, 0.5 mg of genomic DNA, and 50 pmol of each primer. Conditions were as follows: one cycle at 95°C for 3 min for denaturation and hot start with Taq polymerase (BioTaq); 35 cycles at 95°C for 20 s, 45°C for 30 s, and 72°C for 45 s; followed by one cycle at 72°C for 5 min. PCR products of the expected size were separated by agarose gel electrophoresis, extracted (QIAGEN Inc.), and cloned into pGEM-T Easy (Promega, Madison, WI). The subsequent plasmids were sequenced on an Applied Biosystems Model 373 DNA sequencer (PerkinElmer Life Sciences) to confirm the identities and sequences of the isolated genes.
Subcloning the cbi Genes for Expression Studies-Cloning procedures were performed as described (14). The isolated cobalamin biosynthetic genes from P. gingivalis, including cbiL, cbiK, cbiGF, and cbiJD, were separately cloned into pKK223-3 under the regulation of a tac promoter by directional cloning. A second set of plasmids were also constructed whereby each gene or gene fusion was cloned into pET14b behind a His tag encoding sequence under the control of a T7 promoter.
The pKK constructs were used to carry out functional complementation tests on either defined S. typhimurium or recombinant E. coli cobalamin mutants. The pET constructs were used for expression of protein for affinity purification by metal chelate chromatography. The plasmids and strains are described in Table I. Purification of His-tagged Proteins-The protein was expressed and purified as described in the Novagen pET manual. Briefly, a culture of the appropriate strain was grown in LB medium containing ampicillin and chloramphenicol at 37°C until an A 600 of 0.6 was reached. At this stage, isopropyl-␤-D-thiogalactopyranoside was added to a final concentration of 0.4 mM, and the cells were grown for a further 2 h before harvesting by centrifugation (10,000 ϫ g, 4°C). The bacterial pellet was resuspended in binding buffer (5 mM imidazole, 0.5 M NaCl, and 20 mM Tris-HCl (pH 7.9)) and sonicated at 8 microns for 4 ϫ 30 s bursts using an MSE Soniprep ultrasonicator. The solution was centrifuged again (10,000 ϫ g, 4°C), and the supernatant was loaded onto a charged His-Bind column (1 ϫ 2.5 cm). The column was washed in 10 column volumes of binding buffer and 6 column volumes of wash buffer (100 mM imidazole, 0.5 M NaCl, and 20 mM Tris-HCl (pH 7.9)), and finally, the protein was eluted in 6 column volumes of elution buffer (400 mM imidazole, 0.5 M NaCl, and 20 mM Tris-HCl (pH 7.9)). Protein was detected by the Bio-Rad protein assay and SDS-polyacrylamide gel electrophoresis. Fractions containing the His-tagged protein were pooled and desalted by passage through a PD-10 column that had been previously equilibrated in 50 mM Tris-HCl (pH 7.8).
Data Bases and Computer Programs-Sequences, alignments, and comparisons were performed with the GCG software package (Genetics Computer Group, Inc., Madison, WI). Searches of the Unfinished Bacterial Genome Data Base were performed using the BLAST suite. Evolutionary trees were constructed using PAUP Version 4.0.

RESULTS
Since the gene for methylmalonyl-CoA mutase, a vitamin B 12 -dependent enzyme, had previously been isolated and sequenced from P. gingivalis (6), we decided to investigate whether the bacterium is able to synthesize the coenzyme required for this reaction (adenosylcobalamin) de novo by determining the presence of any cobalamin biosynthetic genes.
This was achieved using a degenerate PCR approach with primers to btuR (11), the cobalamin adenosylation enzyme, and cbiP, cobyric-acid synthase (Fig. 2). Degenerate PCR with the btuR primers did not produce any DNA products of the correct size, although a negative result by this procedure does not mean that the gene is absent since the process is reliant on a high degree of similarity to gene homologs. However, the lack of success with these primers may also be explained by the fact that not all bacteria capable of de novo cobalamin synthesis appear to harbor btuR and must possess another, as yet unidentified, enzyme capable of adenosylation. Fortunately, degenerate PCR with the primers to cbiP produced a 350-base pair fragment of the expected size. Cloning and sequencing of this product revealed that the fragment has 54% sequence identity to cbiP from S. typhimurium. At this point in time, the non-annotated data from the P. gingivalis genome sequencing project became available on line. The cbiP sequence obtained from the degenerate PCR was found to match exactly that found in the P. gingivalis data base. Furthermore, the full sequence of cbiP in the P. gingivalis data base demonstrated clear similarity to cbiP from other organisms. A search of the P. gingivalis sequence data base with other known corrin biosynthetic enzymes revealed the presence of 14 genes involved in the cobalamin pathway (cbiL, cbiH, cbiK, cbiC, cbiET, cbiGF, cbiJD, cbiB, cbiP, cbiA, cobD, cobS, cobT, and cobU) (Figs. 2  and 3). These 14 genes encode enzymes that perform 17 different enzymatic transformations. The discrepancy between the number of genes and the number of enzymatic transformations is due to the fact that three of the genes encode bifunctional proteins, where two separate genes appear to have fused together (cbiJD, cbiET, and cbiGF). Of the 14 genes, 11 appear to be clustered into three operons, whereas the remaining three genes are dispersed randomly throughout the genome (Fig. 3). The identified genes appear to encode mainly enzymes that are responsible for the transformation of precorrin-2 into cobalamin (Fig. 2). An alignment of the cobalamin biosynthetic methyltransferases found in the P. gingivalis genome (CbiL, CbiH, CbiJ, CbiET, and CbiGF) with all other known cobalamin biosynthetic methyltransferases allowed a phylogenetic tree to be produced (data not shown) (23). Not surprisingly, the P. gingivalis methyltransferases segregated with enzymes associated with the oxygen-independent pathway. Indeed, three of the other cobalamin genes (cbiD, cbiG, and cbiK) can also be considered the genetic hallmarks of the anaerobic pathway since they are only found in organisms that contain the oxygenindependent route (13).

TABLE II
List of primers used in this study Primers for insertion of genes into pET14b were the same as the specific primers above, except that all 5Ј-primers started with NdeI (CAT ATG), and all 3Ј-primers started with either BamHI or BgIII. ALAD, ALA, dehydratase.

Gene
Primer Sequence 3. Genetic organization of the cbi (genes of cobinamide synthesis), cob (genes for nucleotide loop assembly) and hem genes in P. gingivalis. The genes in light gray are associated with both the aerobic and anaerobic pathways, whereas those in dark gray are found associated only with the anaerobic pathway.
Of significant interest was the absence of gene homologs that encode any of the enzymes of precorrin-2 synthesis. Furthermore, degenerate primers to ALA dehydratase failed to amplify any product, and enzymatic assays on crude cell extracts from P. gingivalis for ALA dehydratase and PBG deaminase failed to demonstrate any enzymatic activity. As for other modified tetrapyrrole synthesis, the only genes that were recorded in the genome were a few for heme synthesis, including hemN, which encodes an anaerobic coproporphyrinogen oxidase, and hemH, which encodes the protoporphyrin ferrochelatase. Thus, the P. gingivalis genome appears to harbor all the cbi genes together with a few late heme biosynthetic genes, but none of the genes for the synthesis of ALA and its transformation into precorrin-2.
Cloning and Functional Complementation-Although the P. gingivalis cbi genes appear to contain the genetic hallmarks of the oxygen-independent corrin pathway, there were a number of gene fusions that had not been observed previously. For instance, cbiJ and cbiD appeared fused together, as did cbiG and cbiF. Moreover, P. gingivalis also appeared to harbor cbiK, a gene now known to encode the anaerobic cobalt chelatase (24,25) and which has been detected in only four other organisms, S. typhimurium, Clostridium difficile, Clostridium acetobutylicum, and Clostridium tepidum. To evaluate the functionality of some of these genes, they were cloned and used in a series of complementation experiments (see below and Table III).
The genes cbiL, cbiJD, cbiK, and cbiGF were amplified from P. gingivalis genomic DNA and cloned into pKK223-3 and were then used to complement defined S. typhimurium or E. coli or recombinant E. coli cobalamin biosynthetic mutants. With cbiK, there was a certain ambiguity with the size of the protein since the gene had two potential translation start codons, the first of which appeared to give the protein an 18 -28-amino acid extension at the N terminus in comparison with other CbiK proteins. A motif search in the PROSITE dictionary of protein sites and patterns indicated that this extension displays some of the features associated with signal peptides found on membrane lipoproteins (26). Intriguingly, CbiK is the subject of two independent patent applications in Australia, with the authors citing CbiK as an immunoreactive antigen (41,42), supporting a theory that the protein is exported to the outer membrane. This is sustained by results detailing how CbiK is able to act as an outer membrane heme-binding protein that is important for the growth of P. gingivalis (27,28). It has been suggested that CbiK may be associated with a protective immune response against P. gingivalis-induced periodontitis (27).
To test whether P. gingivalis CbiK is active in the cobalamin biosynthetic pathway, cbiK was cloned such that it was expressed from the first methionine and is referred to as CbiK L . Initially, the cloned cbiK L was used to complement an E. coli cysG 3 mutant after the gene had been cloned in tandem with a uroporphyrinogen III methyltransferase from P. denitrificans (cobA). The plasmid was found to complement efficiently the cysG deficiency of the strain, indicating that cbiK L is functional as a precorrin-2 ferrochelatase in siroheme synthesis. However, the addition of exogenous cobalt to the medium prevented this complementation, suggesting that the ferrochelatase activity is inhibited by cobalt, consistent with the function of CbiK being a cobalt chelatase (24,25). If CbiK L does contain a signal peptide, it appears not to be transported since in E. coli, at least, CbiK L appears to be retained in the cytosol.  (38). Aside from an insert in the P. gingivalis sequence near the last ␤-strand (left side, middle), the structures align quite well. Residues conserved among all known CbiK sequences are highlighted in yellow and cluster around the active site. An N-mesomethylporphyrin molecule (green) has been positioned in the active site based on the known crystal structure of that compound with the related structure of ferrochelatase (39).
When P. gingivalis cbiK L was used to complement a cobalt chelatase-deficient E. coli corrin-producing strain, the gene was found to restore cobalamin levels to those previously observed with S. typhimurium cbiK (Table III) (24). Thus, the P. gingivalis and S. typhimurium cbiK genes are functionally equivalent, encoding precorrin-2 cobalt chelatases that can also act as precorrin-2 ferrochelatases. Moreover, the P. gingivalis protein can be modeled accurately onto the S. typhimurium CbiK structure (25), with complete conservation of the amino acids involved in catalysis and structure formation (Fig. 4). To investigate the role of the N-terminal extension, cbiK was expressed in both long (CbiK L ) and short (CbiK S ) versions as N-terminally His-tagged recombinant proteins in E. coli. This was achieved by cloning the respective genetic transcripts into pET14b. The two proteins were purified by metal chelate chromatography. Interestingly, the protein with the potential membrane lipoprotein signal sequence did not purify as easily as the protein without the signal and displayed a tendency to aggregate (Fig. 5). The two proteins were used for in vitro assays, which demonstrated that they both catalyzed the insertion of Co 2ϩ into precorrin-2 with similar efficiency (Fig. 6).
When cbiL, which encodes a cobalt-precorrin-2 methyltransferase (29), was cloned into pKK223-3, the resultant plasmid was found to complement an S. typhimurium cbiL strain (Table  III) (11,21). In this respect, the S. typhimurium and P. gingivalis cbiL genes are interchangeable. The two bifunctional genes cbiJD and cbiGF (encoding multifunctional enzymes of unknown function) were also cloned into pKK223-3. However, the presence of P. gingivalis cbiJD did not restore cobalamin synthesis to an S. typhimurium cbiD strain. This result is consistent with the appearance of at least two classes of cbiD, reflecting some subtle differences in the anaerobic cobalamin pathway. For instance, Bacillus megaterium cbiD does not efficiently complement the S. typhimurium cbiD strain (11). The ability of the fusion protein to complement a cbiJ mutant was not tested.
P. gingivalis cbiGF was found to complement an S. typhimurium cbiF mutant (11), indicating that the encoded protein is functional, in part, as the C-11 methyltransferase. The CbiF region of the fusion protein also modeled well onto the previously determined CbiF structure (30), in agreement with its functionality (data not shown). The cbiGF fusion was not tested for its ability to complement defined cbiG mutants. The functional complementation experiments demonstrated that the P. gingivalis genes represent part of an oxygen-independent pathway, broadly similar to those described for S. typhimurium and B. megaterium (Fig. 2) (8, 13).
Properties of the Expressed Proteins-As with CbiK (cobalt chelatase), the proteins CbiL (C-20 methyltransferase), CbiJD (unknown function), and CbiGF (unknown function) were overproduced as recombinant His-tagged variants in E. coli. This was accomplished by cloning the respective genes into pET14b, which, after transformation into BL21(pLysS), allowed the recombinant proteins to be expressed to a level of at least 10% of the total cellular protein. CbiK, CbiL, and CbiGF were all found to be largely soluble, whereas CbiJD was found to be largely insoluble. However, CbiJD could be resolubilized in 6 M guanidine hydrochloride and, after dialysis, remained in solution. The proteins could all be purified by metal affinity chromatography, and the purified proteins are shown in Fig. 5. CbiK L and CbiK S both ran as proteins with a molecular mass of 30 kDa, as did CbiL, whereas both CbiJD and CbiGF ran with a molecular mass of 66 kDa, consistent with their gene-derived molecular masses. The overproduced proteins are currently  and absence (thick line) of either CbiK L or CbiK S . In the absence of CbiK, a spectrum of partially oxidized precorrin-2 was obtained (broad absorption maximum between 380 and 400 nm) since the reaction was performed under aerobic conditions. In the presence of either CbiK L or CbiK S , the spectrum of cobalt sirohydrochlorin was obtained, consistent with the function of CbiK as a cobalt chelatase. being used for in vitro assays and in attempts to crystallize the proteins.
Can P. gingivalis Make Cobalamin de Novo?-The absence of any genes for the synthesis of precorrin-2 suggested that the bacterium was incapable of de novo synthesis of cobalamin. To address this possibility, cells were grown on a range of complex and defined media to ascertain whether there was any evidence for de novo synthesis of cobalamin.
When grown on rich medium (Columbia blood agar) known to contain exogenous vitamin B 12 , P. gingivalis strain W83 accumulated cobalamin to a level of ϳ5 pmol/10 9 cells (Table  IV). The bacteria were subcultured on defined medium (17) in the absence of vitamin B 12 , on which it was noted that the levels of cobalamin dropped to ϳ1 pmol/10 9 cells (Table IV). Further subculturing on defined medium resulted in poor growth and a lowering of the amount of cobalamin detected in cell extracts (data not shown). The addition of exogenous cobalt and ALA, the first universal intermediate of all porphinoid synthesis (4), to the defined medium did not result in significantly increased levels of cobalamin (data not shown). However, the addition of exogenous cobalamin to the defined medium resulted in an accumulation of cobalamin within the cells. Such uptake may be mediated by TonB-linked receptors, which are frequently involved in periplasmic translocation of hemin, iron, colicins, or vitamin B 12 in other bacteria and which have been reported in P. gingivalis (31). Based on all this available evidence, it appears that P. gingivalis is unable to synthesize its own cobalamin and relies on uptake to meet its coenzyme requirement.
Since the bacteria harbor genes capable of converting precorrin-2 into cobalamin, precorrin-2 was added to liquid cultures of P. gingivalis growing in defined medium. This did not result in any increase in endogenous cobalamin synthesis, merely suggesting that precorrin-2 is probably not taken up by the cells. Indeed, the first cobalamin pathway intermediate known to be taken up in other bacteria is cobyric acid. When this compound was added to the defined P. gingivalis medium, not surprisingly, it was found that it was converted into cobalamin (or at least an intermediate between cobinamide and adenosylcobalamin) (Table IV). This was deduced from the observed growth of an S. typhimurium indicator strain (AR2680) induced by extracts of P. gingivalis that had been supplemented with cobyric acid. The indicator strain cannot convert cobyric acid into cobinamide and hence cannot make cobalamin due to a mutation within cbiB. Growth of the indicator strain is therefore dependent upon the presence of exogenous cobinamide (or a later pathway intermediate). The conversion of cobyric acid into cobinamide (and later intermediates) in P. gingivalis was, as expected, dependent upon the concentration of added exogenous cobyric acid, i.e. the transformation of cobyric acid was dose-dependent (Table IV). This result demonstrates that at least the cobD (encoding L-threonine-phosphate decarboxylase) and cbiB (encoding cobinamide synthase) (9) gene products must be functionally active.
To test whether the presence of this cobalamin pathway is an isolated phenomenon of W83, we investigated the ability of a large number of other related strains to convert cobyric acid into cobinamide. To this end, several P. gingivalis strains were tested, along with closely related Bacteroides and Prevotella species. These species were selected so that a broad range of pigmented and non-pigmented species, which are often found in the same habitat, could be tested. All but P. loescheii were able to convert exogenously supplied cobyric acid into cobinamide (Table V). P. loescheii was capable of taking up exogenously supplied cobyric acid, but did not convert it into cobinamide.

DISCUSSION
The ability of P. gingivalis to biosynthesize tetrapyrroles de novo has been investigated. Extracts of the bacterium do not possess activity for ALA dehydratase or PBG deaminase, two key enzymes involved in the synthesis of uroporphyrinogen III. Similarly, it was not possible to detect any genetic evidence for these early enzymes with the use of degenerate PCR or by searching the data bases of the P. gingivalis genome sequencing project. However, biochemical and genetic investigations demonstrated that the bacterium does harbor the genes for the conversion of precorrin-2 into cobinamide. Some of these genes appear as fusions, including cbiJD and cbiGF, and these can be interpreted with respect to the previously unidentified functions for CbiD and CbiG since other fusion proteins in the cobalamin biosynthetic pathway appear to occur between enzymes that catalyze consecutive steps. For example, CbiET, which is also equivalent to CobL in the oxygen-dependent pathway, catalyzes the methylation at C-5/C-15 as well as the decarboxylation step in corrin synthesis (32), whereas CbiK L catalyzes the cobalt insertion step and methylation at C-20. On this basis, it is likely that CbiG would catalyze a step prior to C-11 methylation, possibly the ␦-lactone formation that gives rise to cobalt precorrin-4 ( Fig. 2) (33). Similarly, CbiD would therefore be likely to catalyze a step prior to corrin ring reduction, presumably the removal of the C-2 fragment and methylation processes at C-1 (Fig. 2) (13,32,33).
The research we have described on P. gingivalis reveals a number of interesting facts that are difficult to reconcile. First of all, there is no evidence that the bacterium can make uroporphyrinogen III, the universal macrocyclic primogenitor of heme, chlorophyll, siroheme, and cobalamin (4). On this basis, one would assume that the bacterium was unable to make any modified tetrapyrrole de novo, consistent with observations that exogenous heme and cobalamin are required for growth. Second, following from this point, it was somewhat surprising to discover that P. gingivalis contains a corrin pathway, with all the necessary enzymes to convert precorrin-2 into cobalamin. Some of these enzymes were shown to complement defined S. typhimurium and E. coli cobalamin biosynthetic mutants. P. gingivalis is able to take up exogenously added cobyric acid and convert it into cobalamin, indicating that at least some of these enzymes are expressed within the cell. Why then does the bacterium contain a functional corrin pathway if it is unable to make the precorrin-2 substrate for entry into the pathway? The least likely explanation is that the bacterium may have evolved a completely different method for the synthesis of precorrin-2. The recent discovery that precorrin-2 can be converted into heme demonstrates that there is still much to be learned about tetrapyrrole metabolism (34). Moreover, it is known that there are two routes for ALA synthesis (4); therefore, alternative routes for the synthesis of uroporphyrinogen III/precorrin-2 may also exist. However, the fact that we were unable to detect any de novo synthesis of cobalamin in P. gingivalis suggests that such an alternative pathway does not exist in this organism. It would be unlikely that the bacterium is able to convert heme back into precorrin-2 for thermodynamic reasons since the porphyrin macrocycle would have to be reduced back to the level of a hexahydroporphyrin and have its side chains modified to those observed in uroporphyrinogen III. Furthermore, all the P. gingivalis cultures were grown on medium containing exogenous heme, yet there was no evidence that any endogenous cobalamin was being produced.
A more likely explanation is that the bacterium has only relatively recently lost its early heme biosynthetic operon, but has not yet lost its remaining cbi genes. There is no evidence that any of these cbi genes have become inactive, indicating that little time has elapsed for gene alterations to have taken place, although there is some evidence to suggest that CbiK may have evolved a new role in heme binding (27,28). Indeed, the structure of CbiK is very similar to that of protoporphyrin ferrochelatase (30) and may well be able to bind protoheme when complexed with other proteins.
Moreover, the bacterium appears to have lost most of its heme biosynthetic genes since there is no hemD, hemE, or hemG. The remaining hemH gene may be retained to act as a dechelatase to aid in the removal of iron from exogenously imported heme, as has been suggested in Hemophilus influenza (35). The presence of the putative coproporphyrinogen oxidase HemN may be misleading since its function is not yet fully understood. As a pathogen living on a rich medium, P. gingivalis would appear to have lost its selection pressure in keeping the ability to make tetrapyrroles de novo. The bacterium has evolved efficient transport systems to import exogenous heme and cobalamin. The presence of the cobinamide biosynthetic enzymes in P. gingivalis means that the bacterium can salvage the corrin component of cobalamin and convert it back into the coenzyme form. A similar situation also exists in E. coli, where the organism has maintained its nucleotide assembly and attachment genes, but has lost the genes responsible for the biosynthesis of cobinamide (36,37).
In conclusion, P. gingivalis does not appear to have the ability to make either heme or cobalamin de novo. Both modified tetrapyrroles are somehow accumulated from the growth medium. A summary of the remaining pieces of the biosynthetic pathway in P. gingivalis is shown in Fig. 7.