Cloning, Sequencing, Expression, and Insertional Inactivation of the Gene for the Large Subunit of the Coenzyme B12-dependent Isobutyryl-CoA Mutase fromStreptomyces cinnamonensis *

Purification of the coenzyme B12-dependent isobutyryl-CoA mutase (ICM) from Streptomyces cinnamonensis gave a protein of ∼65 kDa by SDS-polyacrylamide gel electrophoresis, whose gene icmAwas cloned using sequences derived from tryptic peptide fragments. The gene encodes a protein of 566 residues (62,487 Da), with 43–44% sequence identity to the large subunit of methylmalonyl-CoA mutase (MCM) from S. cinnamonensis and Propionibacterium shermanii. Targeted disruption of the icmA gene yielded an S. cinnamonensis mutant devoid of ICM activity. The IcmA protein is ∼160 residues shorter than the large subunit of the bacterial MCMs, corresponding to a loss of the entire C-terminal coenzyme B12 binding domain. The sequence of the (β/α)8-barrel comprising residues A1–A400 in P. shermanii MCM is highly conserved in IcmA. The protein was produced in Streptomyces lividans and Escherichia coli with an N-terminal His6 tag (His6-IcmA), but after purification His6-IcmA showed no ICM activity. In the presence of coenzyme B12, protein from S. lividans and S. cinnamonensisof ∼17 kDa by SDS-polyacrylamide gel electrophoresis could be selectively eluted with His6-IcmA from a Ni2+affinity column. After purification, this small subunit showed no ICM activity but gave active enzyme when recombined with coenzyme B12 and IcmA or His6-IcmA.

monensis, which have led to the cloning and sequencing of a gene encoding its large subunit, denoted here icmA. This gene was used to produce a recombinant protein in Streptomyces lividans and Escherichia coli with a His 6 tag fused to the N terminus. We also show that this recombinant protein can be used to isolate an additional small subunit of the enzyme present in S. lividans and S. cinnamonensis. This work has also allowed a comparison of the primary sequences of ICM and MCM large subunits, with unexpected implications regarding the mode of coenzyme B 12 binding to ICM.

Assays
The assay used for ICM is essentially that described previously (6). Protein concentration was determined by Bradford assay (16).

Fermentation
S. cinnamonensis A3823.5 (a high yield monensin-producing strain kindly made available by Lilly (17)) was grown in 15-liter batch fermentations using a procedure described earlier (18). Cell paste (500 -600 g per fermentation) could be stored at Ϫ70°C over several weeks without substantial loss of ICM activity.
Cell Disruption and Ammonium Sulfate Fractionation-Ultrasonic disruption of cell paste (Ϸ500 g) in buffer B (750 ml) was carried out over 15 min at 4 -10°C, and then solids were removed by centrifugation (27,500 ϫ g for 45 min). Ammonium sulfate was added to the supernatant to 35% saturation at 4°C and pH 7.5. After centrifugation (47,800 ϫ g, 1 h) and filtration through glass wool, ammonium sulfate was added to 75% saturation. Centrifugation (47,800 ϫ g for 1 h) afforded a protein pellet (Table I), which could be stored at Ϫ80°C over several months.
Chromatography on DEAE-Sepharose-Protein from above (Ϸ5 g wet weight) was applied to DEAE-Sepharose (2.6 ϫ 45 cm, Pharmacia) equilibrated with buffer A and eluted with a gradient (0 -100% buffer C) over 500 ml at a flow rate of 3 ml/min. ICM appeared in the eluate at Ϸ 92-100% buffer C. The active fractions containing Ϸ300 mg of protein were concentrated by ultrafiltration (Centriprep-10, Amicon).
Preparative Gel Electrophoresis-Continuous preparative native gel electrophoresis was performed with a model 491 Prep-Cell (Bio-Rad). Protein from the preceding step (20 mg) was applied in buffer E (3 ml) to a gel comprising a stacking layer (5.25% acrylamide) followed by the fractionating gel (8% acrylamide). Electrophoresis (at 40 mA) was performed using buffer F as running buffer at 4°C. Proteins eluted from the gel were diverted to a fraction collector.
Affinity Chromatography-Protein from the previous step (2.5 mg) was applied to affinity resin (1 ml, see above) in buffer A. The column was then eluted with a KCl gradient (10 -100% buffer C). ICM appeared from the column at Ϸ0.2 M KCl. This fraction showed a major protein band on SDS-PAGE (Fig. 2), with apparent mass of Ϸ65 kDa and several minor components of lower mass.

Peptide Sequencing
The ICM-containing protein from above (Ϸ200 g) was electrophoresed by SDS-PAGE (12%, 10 ϫ 10-cm gel), electroblotted onto a cationic polyvinylidene difluoride membrane (Immobilon CD, Millipore), and visualized by negative staining (Quickstain, Zoion Research). The membrane spot containing the adsorbed protein (of Ϸ65 kDa) was cut out and incubated in buffer G (10 l) with trypsin (0.3 g, Promega) for 15 h at 37°C. Free peptides were washed from the membrane with 10% aqueous trifluoroacetic acid (1 l) and 10% aqueous MeCN with 0.1% trifluoroacetic acid (10 l). The peptides were analyzed by microbore high pressure liquid chromatography (C 8 column, 300 Ϸ pore size, 1 ϫ 250 mm, Vydac) eluting with a gradient of 2-80% v/v MeCN in water,  Table I with 0.05% trifluoroacetic acid. The eluate was split; 90% was collected, and the remainder was analyzed by electrospray-mass spectrometry. Selected fractions were subjected to sequence analysis using the automated Edman method (Applied Biosystems 477A sequencer) (see legend to Fig. 4).

Gene Cloning and Sequencing
General DNA manipulation was performed in E. coli (23,24) and in Streptomyces according to Ref. 25. S. cinnamonensis genomic DNA was partially digested with Sau3A, fractionated by sucrose density centrifugation, and fragments of Ϸ15 kb were ligated between the BamHI/ SalI sites in EMBL4 DNA (26). Recombinant EMBL4 clones were isolated from this library by plaque hybridization using the Ϸ310-bp PCR product as probe DNA. A 7.6-kb BamHI fragment containing the icmA gene was isolated from one clone and ligated into BamHIdigested pUC18, to afford pOCI602 (Fig. 3). A 3.75-kb BamHI/BglII fragment from this region was ligated in both orientations into BamHIcut pUC18 to afford pOCI609 and pOCI613. Similarly, a 4.1-kb BamHI/ SphI fragment from pOCI602, after end-filling with PolIk, was ligated into SmaI-digested pUC18 to afford pOCI611 and pOCI612 (Fig. 3). The plasmids pOCI609, pOCI611, pOCI612, and pOCI613 were used to determine the 4.3-kb nt sequence shown in Fig. 4, on both DNA strands, by the dideoxy method (27) using dye terminator chemistry (Perkin-Elmer). Sequence information has been submitted to the EMBL-Gen-Bank TM data base, accession number U67612.

Insertional Inactivation of IcmA
A BstEII/BamHI fragment containing the icmA gene was isolated from pOCI611, filled-in with PolIk, and cloned into the SmaI site of a pUC18 derivative lacking an SacI site, to give pOCI641. The pUC18 derivative was made by digesting pUC18 with SacI, digesting with PolI, and re-ligating. In this way, a unique SacI site is available close to the center of icmA (cf. Fig. 3 and Fig. 6). A BglII/BamHI fragment including the hygB gene was isolated from pIJ963 (28) and, after filling ends with PolIk, was inserted into the PolI-digested SacI site in pOCI641 to give pOCI642. In this way, hygB is inserted into the unique SacI site in icmA. The disrupted icmA gene was then recovered by partial digestion with EcoRI/HindIII, end-filled with PolIk, and cloned into the PolIkfilled BamHI site in pGM160 (29), to give pOCI643. However, the resulting plasmid could not be introduced into S. cinnamonensis. To improve stability of the construct in Streptomyces, the entire E. coli sequences were deleted by digesting pOCI643 with EcoRI/HindIII, endfilling with PolIk, and religation. The resulting plasmid was passaged through S. lividans 1326, then denatured using the procedure of Oh and Chater (30), and used to transform S. cinnamonensis to Hm R and Ts R by selection on R5 agar plates (25) at 30°C. The transformant was then grown in liquid YEME medium (25) with Hm at 39°C and then plated onto R5 agar medium with Hm at 39°C. After sporulation, relicaplating separately to R5 with Ts, and R5 with Hm, gave several Ts S Hm R transformants, one of which was selected for further investigation. Southern blotting (Fig. 6) demonstrated that the icmA gene in this transformant had been inactivated by insertion of the hygB gene. No ICM activity was observed in cell extracts of the mutant grown in the usual way.

Production of IcmA and His 6 -IcmA
The icmA gene was amplified by PCR using the following primers (NdeI and BamHI sites are underlined): KB3, 5Ј-CCATGGATCCTCA- A PCR was performed with primers KB3 and KB4 (1.0 M), dNTPs (440 M), and denatured template DNA (pOCI612, 0.5 g), using the conditions described above. The PCR product was gel-purified, digested with NdeI and BamHI, and cloned between the NdeI/BamHI sites in the plasmid pET3a (31) to give plasmid pOCI614. The correctness of the icmA nt sequence in this plasmid was proven by DNA sequencing. The gene was then excised by digestion with NdeI and BamHI and cloned between the NdeI/BamHI sites in the plasmids pIJ4123 (32) and pET14b (Novagen) to afford plasmids pOCI633 and pOCI640, respectively.
Plasmid pOCI614 was introduced into E. coli BL21(DE3)pLysS (31). After growth in LB medium (200 ml) with ampicillin and chloramphenicol at 30°C and induction at A 600 ϭ 0.7 with 0.4 mM IPTG, cells were shaken for a further 4 h at 30°C, then collected, washed, and sonicated in buffer A. After removal of cell debris by centrifugation, IcmA was purified by chromatography on Q-Sepharose (Pharmacia) with buffer A and a gradient of 0 -100% KCl (0.5 M). After dialysis against buffer A, gel filtration (TSK G3000SW) with buffer A afforded IcmA (4.3 mg) which was homogeneous by SDS-PAGE and gave the expected N-terminal amino acid sequence.
Plasmid pOCI640 was introduced into E. coli BL21(DE3)pLysS. After growth in LB medium (4 liters) with ampicillin and chloramphenicol at 30°C and induction at A 600 ϭ 0.6 with IPTG (0.1 mM), cells were shaken for a further 4 h at 30°C, then collected, washed, and sonicated in buffer K. After removal of cell debris by centrifugation, the protein . This construct was made by cloning icmA between the NdeI and BamHI sites in pET14b (Novagen). Shown is the N-terminal tag sequence including His 6 , the thrombin cleavage site, and the first residue (Met) of icmA.
was applied in portions to Ni 2ϩ -NTA resin (3 ϫ 1.5 cm, Pharmacia) and washed with buffer K, and His 6 -IcmA was eluted with buffer L and then dialyzed against buffer M. Chromatography on MonoQ with buffer M and a gradient of 0 -100% KCl (1 M) afforded His 6 -IcmA (30 mg) which was homogeneous by SDS-PAGE (Fig. 7).
Plasmid pOCI633 was introduced into S. lividans 1362 (25). After growth in YEME (5 liters) with kanamycin (5 g/ml) at 30°C to an A 600 of 0.7-1.0, the cultures were induced with Ts (5 g/ml). After a further 12-15 h the cells were collected and sonicated in buffer K; cell debris was removed by centrifugation, and His 6 -IcmA was purified as above (yield 16 mg).
Purification of IcmB from S. lividans S. lividans 1362[pOCI633] was grown in YEME (5 liters) and induced with Ts, as described above. The cells were sonicated in buffer K; cell debris was removed; coenzyme B 12 (10 M) was added, and the protein (ϳ1.6 g) was chromatographed in portions on Ni 2ϩ -NTA resin (3 ϫ 1.5 cm) in the dark. The resin was washed with buffer K containing coenzyme B 12 (10 M), before eluting with buffer L and dialyzing against buffer M (yield 24 mg). The sample was applied to MonoQ with buffer M containing coenzyme B 12 (10 M), washed with buffer M (no coenzyme B 12 ), and eluted with a gradient of 0 -100% KCl (1 M) to give a fraction containing mainly His 6 -IcmA (16 mg, eluting at 270 mM KCl), followed by a pink colored protein (2.5 mg, eluting at 370 mM KCl) with high ICM activity and a UV spectrum with an absorption maximum at 525 nm. The pink colored protein was dialyzed against buffer M, applied to MonoQ, and eluted in buffer M with a gradient of 0 -100% KCl (1 M). The small subunit IcmB (ϳ4 g) eluted as a sharp peak at 200 mM KCl and was homogeneous by SDS-PAGE (see Fig. 7).
FIG. 4. The nt sequence of the cloned DNA from S. cinnamonensis. The location of orfs predicted from a FRAME analysis are shown. The orf4 is encoded on the opposite strand (not shown), and the predicted aa sequence is shown above the nt sequence. The other orfs are encoded on the DNA strand shown, and the aa sequence is given below. The aa sequences of seven IcmA tryptic peptides, and the N-terminal sequence, determined by protein sequencing are given in italics and are underlined. The predicted start codons (determined for orf1) are shown in bold. An inverted repeat between orf1 and orf2 is underlined. The nt sequences at the 5Ј and 3Ј ends of icmA corresponding to the target sites of PCR primers KB3 and KB4 are also underlined.
Purification of IcmB from S. cinnamonensis S. cinnamonensis was grown for 3-4 days at 30°C in YEME (5 liters) supplemented with valine (6.6 g/liter). The cells were collected and sonicated in buffer K, and cell debris was removed by centrifugation (yield 1.8 g of protein). To this was added recombinant His 6 -IcmA (29 mg), prepared as described above, and coenzyme B 12 (10 M). The protein was chromatographed in portions on Ni 2ϩ -NTA resin (3 ϫ 1.5 cm) in the dark. The resin was washed with buffer K containing coenzyme B 12 (10 M), before eluting with buffer L and dialyzing against buffer M (yield 22 mg). The sample was chromatographed on MonoQ, as described above for IcmB from S. lividans. The protein fraction eluting at ϳ300 mM KCl (15 mg) contained mainly His 6 -IcmA, whereas fractions containing the holoenzyme (4.5 mg) eluted at ϳ350 mM KCl and were dialyzed against buffer M. The protein was then applied to a gel filtration column (Superose 12, Pharmacia) and eluted at a flow rate of 0.2 ml/min with buffer N. A peak containing IcmB (ϳ2 g) eluted with an apparent mass of ϳ16 -18 kDa (Fig. 7). This protein showed no ICM activity until both IcmA and coenzyme B 12 were added (Fig. 8).

RESULTS
Enzyme Assay-The ICM assay involves hydrolyzing CoA thioesters at the end of the reaction, extraction of n-and isobutyric acids into ethyl acetate, and quantification by GC. Typical GC chromatograms from assays performed with recombinant ICM (see above) are shown in Fig. 8. To aid in the quantification of isobutyrate formed, a known amount of valeric acid was added to each assay as an internal standard.
Enzyme Purification and Peptide Sequencing-The ICM was purified as outlined in Table I. The MCM and ICM activities were separated by Q-Sepharose ion-exchange chromatography, with ICM eluting in the middle and MCM at the end of the salt gradient. The protein finally obtained was shown by SDS-PAGE to contain a major component with apparent mass Ϸ65 kDa, together with several minor components of lower mass (Fig. 2). Attempts to further purify the Ϸ65-kDa protein led to large losses in ICM activity. The Ϸ65-kDa protein was isolated by SDS-PAGE and subjected to N-terminal amino acid sequence analysis (see Fig. 4). The Ϸ65-kDa protein was also digested with trypsin, and the resulting peptides were analyzed by high pressure liquid chromatography and electrospray-mass spectrometry. The masses of tryptic fragments were compared with the MOWSE peptide mass fingerprint data base (33), but no similar entries were found. Several tryptic fragments were sequenced by Edman degradation (Fig. 4), revealing up to 75% sequence identity to segments of the MCM large subunit from S. cinnamonensis (6) and P. shermanii (7). Gene Cloning and Sequencing-Two tryptic peptide sequences were used to design oligonucleotides for PCR. The PCR afforded a Ϸ310-bp DNA fragment, which was found to be 70% identical in DNA sequence and 55% in translated protein sequence to the MCM large subunit from S. cinnamonensis. This PCR product was used as a probe to isolate hybridizing clones from a genomic DNA library prepared in EMBL4. From one clone, the region encoding the putative icmA gene was isolated and sequenced on both strands by the dideoxy method (see Figs. 3 and 4).
Sequence Analysis-The 4362-bp DNA segment sequenced showed a total G/C content of 71%. A frame analysis (Fig. 5) was performed using CODONPREFERENCE in the GCG software (The Genetics Computer Group, Madison, WI, version 8.1-UNIX (34)). This revealed three complete orfs (orf1, orf2, and orf3 in Figs. 3 and 4), each with a G/C content of Ϸ75, Ϸ50, and Ϸ95% at the first, second, and third positions of each codon, respectively, which is highly characteristic of protein coding regions in Streptomyces DNA (35). Downstream of the presumptive stop codon of orf1, the G/C distribution changes (Fig. 5), strongly suggesting that the stop codon has been correctly identified. Two incomplete orfs (orf4 and orf5) were also predicted, extending outward from each end of the region sequenced. The incomplete orf4 (nt 1-583) shared over the available protein sequence a similarity of Ϸ32% to endoglucanases in the EMBL/SWISSPROT data base. Comparisons of orf2 (nt 1393-1725), -3 (nt 794 -1303), and -5 (nt 3938 -4326) with the data base failed to identify proteins with significant sequence similarities, so their functions are presently uncertain. The orf1 was identified as the putative icmA gene due to its high sequence similarity, at the DNA and protein levels, to the large  a Amount of protein available following each step, starting from ϳ500 g (wet weight) cell paste.
b The concentration of isobutyrate attained at the end of the assay is calculated by comparison to the amount of n-valeric acid in the EtOAc extract. In each assay, n-butyryl-CoA (250 M) in buffer (200 l) and coenzyme B 12 were incubated with protein for 30 min at 30°C (see "Experimental Procedures").
c The ratio of n-to isobutyrate peak volumes detected by GC is shown. This indicates how far the interconversion has proceeded in each assay. No correction for hydrolysis of substrates during the course of the assay is included. d Activity indicates amount of isobutyrate formed (see Footnote b) in mol/min/mg protein.
subunits of S. cinnamonensis and P. shermanii MCM, as well as to the human and mouse MCMs (Table II). The 3Ј-untranslated region downstream of orf1 shows no homology at the nt or aa levels to MCMs, again consistent with the correct identification of the stop codon of orf1.
The N-terminal amino acid sequence determined for IcmA agrees with that predicted by the DNA sequence, starting at nt 1800 with an ATG codon. Termination occurs at nt 3500 with a TGA codon, corresponding to a protein with 566 aa, and a mass of 62,487 Da, which agrees well with the mass of ϳ65 kDa estimated by SDS-PAGE. The peptide sequences determined from tryptic fragments are encoded at the expected locations in the icmA gene sequence (Fig. 4). A comparison of the IcmA protein sequence, with those of the homodimeric and heterodimeric MCM large subunits from various organisms, was performed with PILEUP in the GCG software (Fig. 9). A DOT-PLOT comparison between ICM and the MCM large subunit from P. shermanii is shown in Fig. 10.
Disruption of the S. cinnamonensis icmA Gene-A targeted insertional inactivation of the icmA gene in S. cinnamonensis was achieved by first inserting a cassette containing a functional Hm resistance gene (hygB) into the unique SacI site within the cloned S. cinnamonensis icmA gene (Figs. 3 and 6). The icmA containing hygB was cloned into the vector pGM160 (29) to give plasmid pOCI643 which, however, could not be introduced into S. cinnamonensis, possibly due to instability of the plasmid under the growth conditions. Subsequently, by removing the entire E. coli sequences from pOCI643, and introducing a plasmid denaturation step (30), S. cinnamonensis Ts R Hm R transformants were isolated, which after further growth at 39°C yielded Ts S Hm R colonies. A Southern blot hybridization analysis of genomic DNA isolated from one of these clones confirmed that the icmA gene had been inactivated, consistent with a double crossover event (Fig. 6). Extracts of the S. cinnamonensis icmA::hygB mutant were devoid of ICM activity.
Expression of the icmA Gene-The icmA gene was amplified by PCR using oligonucleotide primers incorporating NdeI and BamHI sites, such that the NdeI site incorporates an ATG start codon. The PCR product was cloned after digestion with NdeI and BamHI between the NdeI/BamHI sites in pET3a to afford pOCI614. After introduction into E. coli BL21(DE3)pLysS and induction with IPTG at 30°C, large amounts of soluble protein were isolated, with the correct apparent mass on SDS-PAGE, and the correct N-terminal amino acid sequence. This protein, however, was devoid of ICM activity.
To produce IcmA in S. lividans 1326, the gene was cloned on the NdeI/BamHI fragment into the high copy number expression vector pIJ4123 to afford pOCI633. Only a very low ICM activity was found in cell extracts of S. lividans 1326[pIJ4123] grown in YEME. However, cell extracts from S. lividans 1326[pOCI633] after induction with Ts showed high levels of ICM activity, typically about 5-10 ϫ higher than seen in ex-tracts of S. cinnamonensis. The His 6 -IcmA was purified to homogeneity by Ni 2ϩ -chelate affinity and gel filtration chromatography but showed no mutase activity (Fig. 8). The same His 6 -IcmA was also produced in E. coli using the vector pET14b (Novagen) but again showed no ICM activity.
Purification of an ICM Small Subunit-Cell extracts from S. lividans 1326[pOCI633] were fractionated by metal-chelate affinity chromatography in the presence of coenzyme B 12 , to recover His 6 -IcmA and its associated subunit. Subsequent ionexchange chromatography on MonoQ (Pharmacia) in the absence of coenzyme B 12 gave a protein similar to 17 kDa by SDS-PAGE (Fig. 7), which by itself was devoid of ICM activity, but gave highly active ICM after incubation with His 6 -IcmA (or IcmA) and coenzyme B 12 . The intact holoenzyme showed a UV-visible absorption spectrum with a maximum at 525 nm typical of protein-bound adenosyl cobalamin (data not shown).  6. A, the genomic region of wild type (wt) S. cinnamonensis encoding IcmA, as deduced from the sequenced DNA (see Figs. 3 and 4). The restriction sites (BglI, BglII, SalI, and SacI) and their positions found within the 4362 bp sequenced are indicated. Below is a map showing the expected organization of orfs, and positions of restriction sites, after a double crossover from the hygB-containing icmA gene in the vector and selection for Ts S and Hm R . B, hybridization analysis of a Southern blot of DNA from the Ts S Hm R transformant (denoted ko) and the wild type strain, digested with BglI, BglII, or SalI. The probe was the PCR fragment amplified by primers KB3 and KB4 (see "Experimental Procedures") and digested with NdeI/BamHI. This probe includes the entire orf1. The positions of size markers are shown to the left.
The His 6 -IcmA alone showed no UV-visible absorption maximum at 525 nm and does not bind coenzyme B 12 under these conditions.
The ICM small subunit was isolated from wild type S. cinnamonensis, and from the icmA::hygB mutant (see above), in a similar way by addition of His 6 -IcmA to cell extracts, followed by metal-chelate affinity chromatography in the presence of coenzyme B 12 , ion-exchange chromatography on MonoQ, and gel filtration. The yield of the small subunit was lower, but SDS-PAGE again revealed a protein of ϳ17 kDa (Fig. 7), which by itself was inactive but yielded highly active ICM upon incubation with both coenzyme B 12 and His 6 -IcmA (Table III and Fig. 8), and afforded a holoenzyme with a UV-visible maximum at 525 nm.

DISCUSSION
Crucial to any enzyme purification is an assay that allows detection and quantification of catalytic activity. The assay for ICM used here is sensitive but ill-suited for accurate quantification of specific activity, especially when limited amounts of protein are available. For a typical assay during the purification of ICM, sufficient protein (ϳ50 -200 g) was taken to afford between a Ϸ10:1 to 2:1 ratio (as determined by GC) of nto isobutyrates in a single 30-min incubation at 30°C, with n-butyryl-CoA as substrate. The amount of isobutyrate formed per min per mg of protein was then estimated, based on this single time point in the reaction. This gives an estimate of the mutase activity at each stage of the purification (Table I) but clearly does not correspond to the specific activity of the enzyme.
The enzyme is present in low amounts in cell extracts of S. cinnamonensis but is stable at room temperature over several hours. A variety of chromatographic methods failed to yield a significant improvement in purity without incurring major losses of ICM activity. With hindsight, it seems likely that these losses were due to the separation of subunits of the enzyme. This was not anticipated, since we had succeeded in purifying the heterodimeric MCM from S. cinnamonensis without major difficulties. 2 However, a significant gain in ICM purity was achieved by incorporating vitamin B 12 affinity chromatography late in the purification scheme (Table I).
After six purification steps the ICM contained a major component with apparent mass Ϸ65 kDa on SDS-PAGE, along with several proteins of lower mass (Fig. 2). No protein of 2 A. Leiser, unpublished work.   Fig. 7) in the presence of coenzyme B 12 (20 M) Without added coenzyme B 12 the enzyme is completely inactive. The protein concentrations were determined by the Bradford assay, assuming a mass of 17 kDa for IcmB. The total amount of each subunit added to each assay in picomoles is shown. The substrate was n-butyryl-CoA (280 M) in potassium phosphate buffer (50 mM, pH 7.4) with EDTA (5 mM) and glycerol (5% v/v), at 30°C. The i/n ratio gives the ratio of isoto n-butyrate determined by GC after 30 min incubation (compare Fig.  8), from which the amount of isobutyryl-CoA formed and hence the enzyme activity were determined (see "Discussion"). Compare with Table I 8. Gas chromatograms from assays of ICM activity performed with purified His 6 -IcmA (ϳ1 g) and small subunit from S. cinnamonensis (ϳ0.1 g) with added coenzyme B 12 (A); only the small subunit (ϳ0.1 g) with added coenzyme B 12 (B); both proteins as in A, but without added coenzyme B 12 (C). Only in A is the formation of isobutyrate (i) from n-butyrate (n) apparent. The valeric acid added as a standard prior to extraction of the n and isobutyrate is denoted by V. The IcmA protein alone shows no activity and gives assay results exactly comparable to those shown in B and C. higher mass was apparent by SDS-PAGE. Tryptic peptide fragments isolated from the Ϸ65-kDa protein showed high sequence identities (25-75%) to portions of the large subunits of MCM from both S. cinnamonensis (6) and P. shermanii (7), consistent with this being a subunit of a closely related enzyme.

FIG. 7. Above is shown a Coomassie Blue-stained SDS-PAGE (8 -25% gradient) gel of purified His 6 -IcmA from E. coli (lane A) and IcmB isolated from S. lividans (lane B). Below is shown a
A PCR-based reverse genetic approach then allowed the cloning and sequencing of the icmA gene (denoted orf1 in Fig. 3). The translated sequence of 566 aa (M r 62,487) shows a high similarity across almost its entire length to the large subunit of MCM from microbial sources (Fig. 10), as well as to the homodimeric human and mouse MCMs (Table II). It is noteworthy that an orf similar in size and sequence to that of the small subunit of MCMs from S. cinnamonensis and P. shermanii was not found directly adjacent to this icmA gene (Fig. 3). In contrast, the orfs for the large and small subunits of S. cinnamonensis and P. shermanii MCM possess overlapping stop and start codons, a device which is thought to lead to translational coupling and hence to the production of stoichiometric amounts of the two polypeptides.
Proof that orf1 is necessary for ICM activity was obtained by disruption of the gene in S. cinnamonensis. By targeted insertional inactivation, the chromosomal icmA gene was replaced in a double crossover with a copy containing a functional Hm resistance gene inserted into its unique SacI site (Fig. 6), using the vector pGM160. This vector contains a temperature-sensitive Streptomyces origin of replication, resulting in its loss from host cells grown at the non-permissive temperature of 39°C (29). The vector has been used previously for gene disruptions in S. cinnamonensis (36,37). The resulting S. cinnamonensis icmA::hygB mutant was devoid of ICM activity under the usual assay conditions, providing direct evidence for a functional role of icmA in ICM activity. This mutant will be of value to study the influence of ICM on polyketide antibiotic production in this strain, since the enzyme has been implicated in an important biochemical pathway, furnishing methylmalonyl-CoA from isobutyryl-CoA and n-butyryl-CoA (1).
A catalytic function for the IcmA protein was sought by expression in a heterologous host. In a first attempt, the protein was made in the cytoplasm of E. coli, by placing the icmA gene under the transcriptional control of a T7 RNA polymerase promoter in the plasmid pET3a (31). Although large amounts of soluble IcmA could be made in this way, it was devoid of ICM activity. The reason for the lack of mutase activity became clear after IcmA had been produced in S. lividans.
A second attempt to produce IcmA was made using a high copy number expression vector (pIJ4123) suitable for Streptomyces spp. (32). The vector contains the thiostrepton-inducible promoter and ribosome-binding site of the tipA gene. Immediately downstream is a translational start codon (ATG) followed by a sequence encoding a 20-residue N-terminal peptide including a His 6 tag and a thrombin recognition sequence, followed by a unique NdeI site allowing fusion of the peptide leader to the protein of interest. After subcloning the icmA gene into this vector, and introduction into S. lividans 1326, substantially higher ICM activity was detected in cell-free extracts than seen in S. lividans 1326 with pIJ4123. The His 6 -IcmA was readily purified by Ni 2ϩ -chelate affinity chromatography and gel filtration but then showed no ICM activity. Another protein fraction was detected, however, eluting from the gel filtration column after His 6 -IcmA, which showed high ICM activity. This fraction contained several proteins in the size range 12-50 kDa, as well as small residual amounts of His 6 -IcmA. As expected, the activity was dependent upon added coenzyme B 12 . This suggested that at least one additional smaller subunit is necessary to complement the IcmA large subunit and afford active mutase in vitro. Indeed, it is notable that active mutase can be reconstituted with His 6 -IcmA derived from S. cinnamonensis and small subunit(s) endogenous to the wild type S. lividans in which the large subunit had been produced.
The ICM small subunit was purified from S. lividans and subsequently also from S. cinnamonensis, by relying on its association with His 6 -IcmA in the presence of coenzyme B 12 , and exploiting the convenient His 6 -affinity handle. The combination of metal-chelate affinity chromatography in the presence of coenzyme B 12 , and subsequent ion-exchange and/or gel filtration chromatography in the absence of the coenzyme, gave a protein of about 17 kDa by SDS-PAGE (Fig. 7), which was inactive alone but afforded highly active ICM after addition of His 6 -IcmA (or IcmA) and coenzyme B 12 (Fig. 8). The activity of the reconstituted mutase was estimated to be approximately 1.0 mol/min/mg, as shown in Table III. However, we note again here that the assay, as described above, is not well suited for determining specific activities. Nevertheless, this value can be compared with the activity determined for ICM isolated from S. cinnamonensis, as outlined in Table I. From this comparison it is clear that the mutase reconstituted from recombinant His 6 -IcmA and IcmB from S. cinnamonensis (Table III and Fig. 7) has a higher activity than that found for the wild type enzyme at the end of the purification (Table I and Fig. 2).
A comparison of the IcmA protein sequence with those of the human and mouse MCMs and the large subunits from P. gingivalis, S. cinnamonensis, and P. shermanii MCMs performed using PILEUP in the GCG software (34) is shown in Fig. 9. A DOTPLOT comparison of IcmA and the MCM large subunit from P. shermanii is shown in Fig. 10. The DOTPLOT comparison reveals that the region of highest sequence similarity extends approximately over residues 60 -400 in both proteins. The sequence identity in this region is about 50%. The most striking difference, however, is the significant truncation of IcmA in comparison to all MCMs (Fig. 9), corresponding to the loss of the C-terminal Ϸ160 amino acid residues from MCM. A second significant difference is a 16-residue insertion in IcmA (residues 424 -439), which is absent in all the MCM sequences reported to date.
The crystal structure of the P. shermanii MCM reported recently (15) revealed an N-terminal (␤/␣) 8 -barrel domain in the large subunit, from residues A1-A400. The high sequence identity (Ϸ50%) of this region to residues 1-392 in IcmA suggests that the (␤/␣) 8 -barrel is conserved in the structure of IcmA. Residues A401-A559 in the P. shermanii MCM correspond to a largely helical linker, which connects the (␤/␣) 8barrel with the C-terminal, so-called coenzyme B 12 binding, flavodoxin-like domain (A560 -A728). The linker residues A401-A559 in this MCM correspond in the sequence comparison to residues 393-560 in IcmA (Fig. 9), although the sequence identity is only Ϸ18% in this region (Fig. 10). But after just 6 more residues IcmA terminates.
A striking aspect of the recently determined crystal structures of MCM is the replacement of the dimethylbenzimidazole group of coenzyme B 12 as an axial Co 3ϩ ligand by the imidazole of a histidine situated in the C-terminal coenzyme B 12 binding domain of the large subunit. This imidazole is linked through a hydrogen-bonded network to the side chains of two other residues forming a ligand triad (38), which in MCM is His 610 -Asp 608 -Lys 604 . The nucleotide tail of the cofactor fills a cavity in this domain, which places the dimethylbenzimidazole group in a tight hydrophobic pocket. These intimate interactions between coenzyme B 12 and protein suggest a key role for this domain in modulating the reactivity of MCM.
In the case of ICM, the large subunit contains no contiguous coenzyme B 12 binding domain but requires a separate small subunit (IcmB) of ϳ17 kDa to bind coenzyme B 12 and afford active mutase. This indicates that the IcmB small subunit has assumed the role of a coenzyme B 12 binding domain in ICM and will most likely be homologous to the corresponding region of the MCM large subunit. In support of this conclusion, preliminary results from ongoing work have shown that the IcmB from S. lividans and S. cinnamonensis have N-terminal protein sequences that are about 70% identical to the coenzyme B 12 binding domain in MCM (data not shown). In addition, a thorough sequence comparison has shown that the IcmB N-terminal protein sequence is not encoded in the genomic DNA shown in Fig. 4. Presently, we must conclude that the icmB gene is not encoded by one of the small orfs found adjacent to icmA in this work. Future work will focus on cloning icmB from S. cinnamonensis, the quaternary structure of the holoenzyme, and the determination of the kinetic and thermodynamic parameters of this mutase reaction.