{alpha}-5,6-Dimethylbenzimidazole adenine dinucleotide ({alpha}-DAD), a putative new intermediate of coenzyme B12 biosynthesis in Salmonella typhimurium

The late steps in adenosylcobalamin (AdoCbl) biosynthesis are referred to as the nucleotide loop assembly (NLA) pathway. Fig. 1 shows our current understanding of how the nucleotide loop of AdoCbl is assembled in Salmonella typhimurium. Genetic and biochemical evidence indicate that during de novo synthesis of AdoCbl, enzymes of the NLA pathway convert adenosylcobinamide phosphate (AdoCbi-P) to AdoCbi-GDP (catalysed by the CobU enzyme), activate the lower ligand base 5,6-dimethylbenzimidazole (DMB) to its nucleoside in two steps (catalysed by CobT and CobC), and join AdoCbi-GDP and N¹-(α-D-ribosyl)-5,6-dimethylbenzimidazole (α-ribazole) to yield AdoCbl with the release of GMP as byproduct (catalysed by CobS) (Maggio-Hall & Escalante-Semerena, 1999; Warren et al., 2002). The first step in the activation of DMB is catalysed by the CobT enzyme, whose well-documented nicotinate mononucleotide (NaMN)-dependent phosphoribosyltransferase activity yields N¹-(5-phospho-α-D-ribosyl)-5,6-dimethylbenzimidazole (α-ribazole-5'-P); α-ribazole-5'-P features a N-glycosidic bond in the α, rather than the more common β, configuration (Cheong et al., 1999, 2001, 2002; Friedmann & Harris, 1965; Trzebiatowski & Escalante-Semerena, 1997). The second step in DMB activation is catalysed by the CobC enzyme, whose phosphatase activity dephosphorylates α-ribazole-5'-P to yield α-ribazole (O'Toole et al., 1994). Fig. 1 also shows (broken arrows) the uncertainty that exists about the timing of the loss of the phosphate group modifying the 5' OH group of the ribosyl moiety of cobalamin. The alternative route for the removal of the 5' phosphate group of the ribosyl moiety shown in Fig. 1 is based on previously reported biochemical data (Maggio-Hall & Escalante-Semerena, 1999).

(20K): Fig. 1. Model for the NLA pathway previous to this work. In this model the CobT enzyme works as a phosphoribosyltransferase that activates DMB using NaMN as phosphoribosyl donor. Genetic and biochemical evidence support the reactions represented by solid arrows. Only in vitro biochemical evidence supports the reactions represented by broken arrows. Pi, orthophosphate; PPi, pyrophosphate; NTP, nucleoside triphosphate; NDP, nucleoside diphosphate; Na, nicotinic acid; Ado, 5'-deoxyadenosine. The question mark represents an unidentified enzyme.

Phenotypic analysis of cobT mutant strains of S. typhimurium revealed the unexpected phenotype of being able to synthesize AdoCbl if the medium was supplemented with DMB (Trzebiatowski et al., 1994). This observation led to the proposal of the existence in S. typhimurium of an alternative activity that could compensate for the lack of CobT enzyme in cobT mutant strains (Trzebiatowski & Escalante-Semerena, 1997). The requirement for DMB was interpreted to mean that the alternative function had a higher K_m for DMB, thus the medium had to be supplemented with DMB. The alternative activity was subsequently shown to be encoded by the cobB gene and as expected, cobT cobB double mutant strains were no longer responsive to exogenous DMB (Tsang & Escalante-Semerena, 1998). The CobB protein was identified as a member of the Sir2 family of eukaryotic regulatory proteins (sirtuins) involved in gene silencing and cell ageing (Brachmann et al., 1995; Lin et al., 2000; Rine & Herskowitz, 1987; Tsang & Escalante-Semerena, 1998). Sirtuins, including CobB, have NAD⁺-dependent ADPribosyltransferase activity (Frye, 1999; Tanny et al., 1999) and NAD⁺-dependent protein deacetylase activity (Imai et al., 2000; Smith et al., 2000; Starai et al., 2002; Tanner et al., 2000). Relevant to the work presented in this paper is the ADPribosyltransferase activity of the CobB enzyme.

The CobT reaction yields α-ribazole-5'-P when the enzyme transfers the phosphoribosyl moiety of NaMN to DMB (Trzebiatowski & Escalante-Semerena, 1997). Data reported here show that CobT can transfer the ADPribosyl moiety of NAD⁺ to DMB to yield α-5,6-dimethylbenzimidazole adenine dinucleotide (hereafter referred to as α-DAD) (Fig. 2). A recent report showed that the Escherichia coli CobB protein used NAD⁺ as substrate to derivatize DMB. The same report showed that the same activity was associated with the CobT protein from this bacterium (Frye, 1999). The identity of the products of these reactions was not established.

(16K): Fig. 2. Phosphoribosyltransferase and ADPribosyltransferase activities of the CobT enzyme of S. typhimurium. The ADPribosyltransferase activity of CobT generates a new dinucleotide referred to as α-DAD. In either of the reactions, if the CobT enzyme uses NaMN or NaAD as substrate, nicotinate (Na) is released. Similarly, if the enzyme uses NMN or NAD+ as substrate, nicotinamide (Nm) is released.

We report here the purification and characterization of the product obtained when dinucleotides [NAD⁺ or nicotinate adenine dinucleotide (NaAD)] were used as substrates for the CobT enzyme. This work shows that α-DAD can be used in vitro as substrate for enzymic conversion of adenosylcobinamide (AdoCbi) to AdoCbl. In vitro enzyme activity assays
Phosphoribosyltransferase and ADPribosyltransferase activities of CobT.
CobT phosphoribosyltransferase activity [when NMN (nicotinamide mononucleotide) or NaMN were used as substrates] and ADPribosyltransferase activity (when NaAD or NAD⁺ were used as substrates) were initially performed as described by Frye (1999), except the mixture contained 100 mM glycine/NaOH buffer, pH 10, and [2-¹⁴C]DMB (specific activity, 1 mCi mmol^-1; 37 MBq mmol^-1) in a final volume of 20 µl. Initial velocity of the reactions was measured in reaction mixtures containing NMN or NaAD (1, 5 or 10 mM), or NaMN (0·25, 0·5, 1, 5 or 10 mM) and 0·012 µg purified CobT protein in a 20 µl final volume (Trzebiatowski & Escalante-Semerena, 1997). The concentration of CobT in the mixture was increased 10-fold (0·12 µg) when NAD⁺ (1, 5, 10, 20 or 40 mM) was used as substrate. To study these activities at physiological pH, assays were performed in 100 mM MOPS buffer, pH 7. Assays were incubated for 10 min (NaMN and NMN), 20 min (NaAD) or 90 min (NAD⁺) at 37 °C and contained 80 µM [2-¹⁴C]DMB (final specific activity, 24·8 mCi mmol^-1; 917·6 MBq mmol^-1); all assays were performed at least in duplicate. Quantification was performed after TLC (see below) using a PhosphorImager (Molecular Dynamics). Background was subtracted using controls that lacked phosphoribosyl or ADPribosyl moiety donors. Kinetic constants were derived from non-linear regression analysis of the NAD⁺ and NaMN rate data using GraphPad Prism software (Intuitive Software for Science).

Nucleotide pyrophosphatase assays.
A 50 µl reaction mixture containing 100 mM MOPS buffer pH 7, 2·5 mM MgCl₂, 0·05 U snake venom nucleotide pyrophosphatase (Sigma) and purified α-DAD (200 µM) was incubated for 4 h at 37 °C. A reaction mixture without enzyme was used as negative control. Reaction mixtures were analysed by ion-exchange HPLC (see below).

In vitro NLA assays.
NLA assays were performed at the 500 µl scale as described (Maggio-Hall & Escalante-Semerena, 1999) except that cell-free extracts were added simultaneously instead of sequentially. Where indicated, α-DAD (120 µM) substituted for NaMN and DMB in the reaction mixture. Plasmids pNLA1 (cobUST⁺; Maggio-Hall & Escalante-Semerena, 1999), pJO46 (cobC⁺; O'Toole et al., 1994) or pT7-5 (expression vector; Tabor, 1990) were introduced into strain JE6200 [metE205 ara-9 Δ299(hisGcobT) cobC1175 : : Tn10Δ16Δ17 pnuE : : MudQ]. Strain JE6200 was constructed by P22-mediated transduction of the pnuE : : MudQ allele from strain SF456 (pnuE : : MudQ) into strain JE2197 [metE205 ara-9 Δ299(hisGcobT) cobC1175 : : Tn10d(Tc)] using previously described methods (Chan et al., 1972; Davis et al., 1980). Reactions (final vol. 500 µl) were stopped by the addition of 100 µl 100 mM KCN followed by heating at 80 °C for 10 min to convert adenosylcorrinoids to cyanocorrinoids. Reaction mixtures were passed over C₁₈ SepPak columns (Waters), vacuumed to dryness in a SpeedVac concentrator (Savant Instruments), resuspended in 200 µl double-distilled water and analysed by reverse-phase HPLC (RP-HPLC) (see below).

Synthesis and purification of [2-¹⁴C]DMB.
Radiolabelled [2-¹⁴C]DMB was synthesized as described (Trzebiatowski & Escalante-Semerena, 1997).

Purification of NAD⁺.
NAD⁺ was resolved from ADPribose and NMN using a fast-flow DEAE anion-exchange resin (2·5x40 cm; Toyopearl; Rohm & Haas) previously equilibrated with 10 mM NaCl at a flow rate of 250 ml h^-1; a linear gradient from 10300 mM NaCl resolved NAD⁺ from the above-mentioned contaminants (Dickinson & Engel, 1977). Fractions containing NAD⁺ were identified by UV-visible spectroscopy, were concentrated under vacuum and applied onto a reverse-phase C₁₈ HPLC column to desalt NAD⁺ (see below). The concentration of the purified product was determined using the molar extinction coefficient at 260 nm (₂₆₀) of 17 600 M^-1 (Dalziel & Dickinson, 1966).

Synthesis and purification of α-DAD.
A 5 ml CobT reaction mixture (pH 10) containing NaAD and 55 µg CobT protein was used to isolate microgram amounts of α-DAD. The incubation time was extended to 3 h, and α-DAD was isolated from the mixture by RP-HPLC followed by ion-exchange HPLC (see below); α-DAD was desalted by RP-HPLC and dried under vacuum.

Chromatographic techniques
TLC.
Reagents and products of the CobT reaction were resolved using TLC as described (Maggio-Hall & Escalante-Semerena, 1999).

HPLC.
α-DAD was isolated using a previously described RP-HPLC protocol (Maggio-Hall & Escalante-Semerena, 1999). This procedure was also used to desalt α-DAD (retention time, 12·5 min) and NAD⁺ (retention time, 10 min). Ion-exchange HPLC was performed on a Spheroclone SAX column (4·6x250 mm; Phenomenex). The mobile phase was a 26 min gradient of potassium phosphate, pH 5·5 (40500 mM). The column was developed at a flow rate of 1 ml min^-1. Under these conditions the retention time for α-DAD was 21 min. Cyanocorrinoids were resolved using a previously published RP-HPLC protocol (Blanche et al., 1990; O'Toole et al., 1993). Samples (200 µl) were injected onto a Prodigy C₁₈ column (Phenomenex) equilibrated with a mobile phase containing 98 % solvent A (100 mM potassium phosphate buffer pH 8, containing 10 mM KCN), 1 % solvent B (100 mM potassium phosphate buffer pH 6·8, containing 10 mM KCN) and 1 % solvent C (acetonitrile). The column was developed with a 45 min linear gradient that changed the composition of the mobile phase to a 1 : 1 ratio of solvent B : solvent C. The column was equilibrated and developed at a rate of 1 ml min^-1.

Mass spectrometry.
α-DAD and cyanocobalamin isolated from NLA reactions were subjected to electrospray ionization mass spectrometry (ESIMS) analysis at the Biotechnology Center at the University of Wisconsin-Madison.

NAD⁺ or NaAD serve as substrate for the CobT enzyme
In vitro CobT activity assay mixtures containing DMB (80 µM) were performed using either NAD⁺ or NaAD as co-substrates. TLC analysis of the reaction mixtures revealed a DMB derivative with a relative mobility (R_F 0·44) that was clearly distinct from that of α-ribazole-5'-P (R_F 0·33). The specific activity of CobT with NaAD was 10-fold higher than the specific activity of the enzyme when NAD⁺ was the substrate [35 vs 2·5 nmol min^-1 (mg protein)^-1, respectively]. α-Ribazole-5'-P was observed when either NaAD or NAD⁺ was used as substrate due to small contaminating amounts of mononucleotide in the dinucleotide stocks. This contaminant was subsequently removed from the NAD⁺ stock as described in Methods before determining the initial velocity measurements presented below. Fig. 3 shows the results of CobT reactions performed with NaMN (Fig. 3, lane 1) or NaAD (Fig. 3, lane 2). Under the conditions used, in both reactions nearly half of the substrate (∼48 %) was converted to product.

(61K): Fig. 3. Resolution of products and reactants of the CobT phosphoribosyltransferase and ADP-ribosyltransferase reactions by TLC. Shown are reactions containing DMB and either NaMN (lane 1) or NaAD (lane 2). The RF value of each labelled compound is indicated in parentheses. No labelled compounds migrated beyond α-DAD; the solvent front is not shown.

The product of the CobT-catalysed reaction between DMB and NaAD is α-DAD
A 5 ml reaction mixture containing NaAD and DMB was used to obtain product for spectroscopic analysis. After a 3 h incubation, the reaction mixture was analysed by HPLC (data not shown). Unreacted NaAD and DMB eluted at 8 min and 41 min, respectively. Nicotinate eluted with the void volume. A third compound eluted 13 min after injection, with a spectrum similar to that of NaAD, but with a slight shoulder at 280 nm (spectrum not shown). The product was collected and subjected to an additional purification step using ion-exchange HPLC to eliminate minor contaminants. The compound was dried down and subjected to ESIMS (Fig. 4). The mass obtained, 686 atomic mass units (a.m.u.; M-1), agreed with the predicted mass of α-DAD (687 a.m.u.).

(17K): Fig. 4. Identification of α-DAD by ESIMS. Shown is the negative ion mass spectrum. Signals at 686·2 (M-1) and 342·6 a.m.u. correspond to the predicted mass of α-DAD at z=-1 and z=-2, respectively. The peak at 708·2 a.m.u. is the sodium salt of α-DAD (M-1+Na).

α-DAD can serve as the source of α-ribazole-5'-P during in vitro assembly of the nucleotide loop of AdoCbl
The nucleotide loop from AdoCbl was assembled in vitro using the protocol described in Methods. Functions of the CobU, CobS, CobT and CobC enzymes were confirmed by the results of a control experiment that used NaMN and DMB, AdoCbi and GTP as substrates (Fig. 5c). HPLC and mass spectrometry data confirmed that when α-DAD substituted for DMB and NaMN in a reaction mixture containing CobU, CobT, CobS and CobC enzymes, the resulting adenosylcobamide was AdoCbl (Fig. 5a; HPLC, UV-visible spectroscopy, MS data not shown). When a similar experiment was performed in the absence of the CobC enzyme, AdoCbl-P was obtained (Fig. 5b). The latter results suggested that a dinucleotide pyrophosphatase enzyme present in the extract cleaved α-DAD into α-ribazole-5'-P and AMP, with the subsequent incorporation of α-ribazole-5'-P into AdoCbl-P (Fig. 1). It should be noted that the strains used to generate the CobUST- and CobC-enriched cell-free extracts were strains deficient in the periplasmic PnuE nucleotide pyrophosphatase. The absence of PnuE in the extract greatly reduced the background α-DAD cleaving activity (data not shown).

(14K): Fig. 5. Incorporation of α-DAD into AdoCbl by cell-free extracts of S. typhimurium. Shown are chromatograms of the RP-HPLC analysis of cyanocorrinoids. The degree of cyanation was not established, but the derivatization reaction was performed with excess cyanide over corrinoid to ensure full derivatization (Maggio-Hall & Escalante-Semerena, 1999). In (a) and (b), α-DAD (120 µM) was added to the reaction mixture as the sole source of the α-ribazole moiety. (c) shows the analysis of a reaction mixture that contained NaMN and DMB in lieu of α-DAD (positive control). Reaction mixtures for the experiments shown in (a) and (c) contained cell-free extracts enriched for CobU, CobS, CobT and CobC proteins, so the anticipated reaction product was AdoCbl. The reaction mixture for the experiment shown in (b) contained cell-free extracts enriched for CobU, CobS and CobT, but not CobC. The accumulation of Cbl-P suggested α-DAD was cleaved by a dinucleotide pyrophosphatase. Peaks were identified by the retention time of standards and their UV-visible spectra. Cbi, dicyanocobinamide; Cbi-GDP, dicyanocobinamide-GDP; Cbl-P, cyanocobalamin-5'-phosphate; Cbl, cyanocobalamin.

Specific activities of CobT with mononucleotide (NaMN, NMN) or dinucleotide (NaAD, NAD⁺) substrates
Initial velocity measurements were performed to quantitatively compare the reactivity of CobT with NaAD and NAD⁺ with those of the mononucleotides NaMN and NMN. CobT had higher specific activities with NaMN or NMN than with NaAD or NAD⁺, and in both cases the nicotinate forms were preferred over the nicotinamide derivative (data not shown). The apparent K_m of the enzyme for NAD⁺ was calculated to be 9·0 mM, the specific activity was 8·3 pmol min^-1 (mg protein)^-1 and the k_cat was 0·62 min^-1. The kinetic parameters for NaMN were also determined at pH 7 to compare them with the NAD⁺ parameters. For NaMN the apparent K_m at pH 7 was found to be 0·51 mM, the specific activity was 2 083 pmol min^-1 (mg protein)^-1 and the k_cat was 149 min^-1. These data (Table 1) show that under the conditions tested NaMN was the preferred substrate for the CobT enzyme.

Table 1. Kinetic parameters for CobT with NAD or NaMN as substrates at pH 7

CobT has ADPribosyltransferase activity
The data presented here show that the CobT enzyme of S. typhimurium has ADPribosyltransferase activity in vitro, and that the product of this activity is a new dinucleotide, namely α-5,6-dimethylbenzimidazole adenine dinucleotide (α-DAD). The presence of the α-N-glycosidic bond in α-DAD was inferred from knowledge of the mechanism of the phosphoribosyltransferase reaction catalysed by CobT (Cheong et al., 1999, 2001, 2002) and from the results of the in vitro synthesis of AdoCbl-P by cell-free extracts lacking CobC phosphatase activity. Synthesis of AdoCbl-P from α-DAD and AdoCbi-GDP implies that an α-N-glycosidic bond was present in the ribazole-5'-P derived from DAD, so the 3' OH of the ribosyl moiety of the ribazole-5'-P could be coupled to AdoCbi-GDP by the CobS enzyme to yield the observed AdoCbl-P (Fig. 1). Although it is clear that NaMN is the preferred substrate for the CobT enzyme, it cannot be ignored that the CobT enzyme can use NAD⁺ as substrate (this work), that exponential-phase wild-type cells of S. typhimurium contain undetectable levels of NaMN, NMN and NaAD, and that the concentration of NAD⁺ in exponential-phase cells is almost three orders of magnitude higher than any of these precursors (Bochner & Ames, 1982).

The ability of the S. typhimurium CobT enzyme to use NAD⁺ and its precursors NaMN, NMN and NaAD as substrates provides flexibility to the synthesis of AdoCbl under diverse physiological conditions
As mentioned above, the NAD⁺ concentration in exponential-phase cells of S. typhimurium is several orders of magnitude higher than any of the above-mentioned NAD⁺ precursors (790 µM vs undetectable) (Bochner & Ames, 1982). This difference in the levels of NAD⁺ and its above-mentioned precursors would be enough to compensate for the 18-fold higher K_m of CobT for NAD⁺ than NaMN (0·51 mM) under physiological conditions of active growth. However, because the concentrations of NAD⁺ precursors under other physiological conditions have not been measured, it is possible that the concentration of these precursors may rise under specific growth conditions. The ability of the CobT enzyme to use NaMN, NMN, NaAD or NAD⁺ as substrates would allow S. typhimurium to synthesize α-ribazole-5'-P for the assembly of AdoCbl under any growth conditions. We propose that both enzymic activities of the S. typhimurium CobT enzyme (phosphoribosyltransferase, ADPribosyltransferase) are physiologically relevant.

If NAD⁺ is a physiological substrate for CobT in vivo, why does the CobT enzyme of S. typhimurium have a lower K_m for NaMN?
The answer to this question may lie in the origin of the cob operon. It has been postulated that the entire cob operon (including cobT) was inherited by S. typhimurium from an unknown donor (Lawrence & Roth, 1996). It is reasonable to speculate that the extant CobT enzyme might have evolved in a prokaryote whose intracellular level of NaMN was substantially higher than that found in S. typhimurium. Support for this idea can be found in studies of the CobT homologue of Propionibacterium fruendenreichii subsp. shermanii, which was reported to be unable to use NAD⁺ as substrate (Friedmann, 1965). The ability of the S. typhimurium CobT enzyme to use other NAD⁺ precursors such as NMN as substrate is not shared by all CobT homologues, as reported for the CobT enzyme activity of Clostridium sticklandii which failed to use NMN as substrate (Fyfe & Friedmann, 1969). The ability of the S. typhimurium CobT enzyme to use NAD⁺ or its mononucleotide and dinucleotide precursors may be interpreted to mean that the physiological levels of these precursors may vary in this bacterium. Low levels of NAD⁺ precursors may be the selective pressure for the evolution of CobT enzymes able to use NAD⁺ as substrate.

Supporting in vivo evidence that NAD⁺ is a substrate for the CobT enzyme in vivo
The ADPribosyltransferase activity of CobT is not likely to be an in vitro artefact. Recall that the CobB enzyme uses NAD⁺ as substrate, that CobB has NAD⁺-dependent ADPribosyltransferase activity, that AdoCbl biosynthesis in cobT mutant strains of S. typhimurium is restored by the addition of DMB to the medium and that the response of cobT mutants to DMB depends on a functional CobB enzyme. Together, these facts strongly suggest that CobB transfers the ADPribose moiety of NAD⁺ to DMB in vivo, resulting in the synthesis of α-DAD, effectively compensating for the lack of CobT enzyme (Fig. 6). It is important, however, to keep in mind that the CobB enzyme is not part of the AdoCbl biosynthetic pathway; instead CobB function is critical for the post-translational regulation of acyl-coenzyme A synthetase activities and probably of other members of the AMP-forming family of enzymes (Starai et al., 2002, 2003). Under growth conditions that require low levels of AdoCbl (e.g. methionine synthesis), the ADPribosyltransferase activity of CobB fully compensates for the lack of CobT (Trzebiatowski et al., 1994). However, under physiological conditions that require a higher level of the coenzyme (e.g. growth on ethanolamine or 1,2-propanediol as carbon and energy source), CobB does not compensate for the lack of CobT, suggesting that the synthesis of α-DAD by CobB is an inefficient side reaction of this enzyme with limited physiological significance to the cell (J. C. Escalante-Semerena, unpublished results).

(20K): Fig. 6. Modified NLA pathway. In this model, the CobT enzyme acts as a NaMN-dependent phosphoribosyltransferase or a NAD+-dependent ADPribosyltransferase to yield α-ribazole-5'-P. CobB is the S. typhimurium homologue of the eukaryotic Sir2 enzyme. Sir2-like enzymes, including CobB, have protein deacetylase and ADPribosyltransferase activities. CobB is proposed to catalyse the inefficient formation of α-DAD from NAD+ and DMB. Cleavage of α-DAD by an as-yet unidentified glycohydrolase enzyme would generate α-ribazole, bypassing the need for CobC activity. Evidence reported in this paper supports the existence of a dinucleotide pyrophosphatase activity capable of hydrolysing α-DAD into AMP and α-ribazole-5'-P. At present, no evidence for an α-ribazole-forming glycohydrolase activity has been obtained. Question marks represent unidentified enzymes. Genetic and biochemical evidence support the reactions represented by solid arrows. Only in vitro biochemical evidence supports the reactions represented by broken arrows.

A dinucleotide pyrophosphatase enzyme is likely involved in AdoCbl biosynthesis
Results obtained from in vitro NLA assays using α-DAD in lieu of DMB and NaMN suggest that a dinucleotide pyrophosphatase in S. typhimurium cleaves α-DAD into α-ribazole-5'-P and AMP (Fig. 6). This conclusion is supported by the detection of AdoCbl-P amongst the products of the in vitro conversion of AdoCbi and α-DAD to AdoCbl when the reaction was performed with cell-free extracts that lacked CobC phosphatase activity. The identity of the dinucleotide pyrophosphatase enzyme that cleaves α-DAD remains unclear. However, because α-DAD was also cleaved by snake venom nucleotide pyrophosphatase (data not shown), cleavage of α-DAD to α-ribazole-5'-P and AMP may be performed by an enzyme that is not dedicated to AdoCbl biosynthesis. This work was supported by NIH grant GM40313 to J. C. E.-S. L. A. M.-H. was supported by a NSF predoctoral fellowship, a UW-Madison WARF Fellowship and by NIH Biotechnology Training grant GM08349. The authors thank John Foster for strains.