Evolution of Coenzyme BI2 Synthesis Among Enteric Bacteria

Home , Glycerol dehydratase, John Roth (geneticist), Propanediol dehydratase

Evolution of Coenzyme BI2Synthesis Among Enteric Bacteria: Evidence for Loss and Reacquisition of a Multigene Complex

Jeffrey G. Lawrence and John R. Roth

Department of Biology, University of Utah, Salt Lake City, Utah 84112 Manuscript received June 16, 1995 Accepted for publication October 4, 1995

ABSTRACT Wehave examined the distribution of cobalamin (coenzyme BI2) synthetic ability and cobalamindependent metabolism among entericbacteria. Most species of enteric bacteria tested synthesize cobalamin under both aerobic and anaerobic conditions and ferment glycerol in a cobalamindependent fashion. The group of species including Escha'chia coli and Salmonella typhimurium cannot ferment glycerol. E. coli strains cannot synthesize cobalamin de novo, and Salmonella spp. synthesize cobalamin only under anaerobic conditions. In addition, the cobalamin synthetic genes of Salmonella spp. (cob) show a regulatory pattern different from that of other enteric taxa tested. We propose that the cobalamin synthetic genes, as well asgenes providing cobalamindependent diol dehydratase, were lostby a common ancestor of E. coli and Salmonella spp. and were reintroduced as a single fragment into the Salmonella lineage from an exogenous source. Consistent with this hypothesis, the S. typhimurium cob genes do not hybridize with the genomes of other enteric species. The Salmonella cob operon may represent a class of genes characterized by periodic loss and reacquisition by host genomes. This process may be an important aspect of bacterial population genetics and evolution.

OBALAMIN (coenzyme BIZ) is a large evolution- The cobalamin biosynthetic genes have been charac- C arily ancient molecule ( GEORGOPAPADAKOUand terized in S. typhimurium; most genes are located in a SCOTT 1977; ESCHENMOSER1988; SCOTT 1990, 1993) large 20-gene cluster at minute 41 on the genetic map that participates as a cofactor in both prokaryotic and ( JETER and ROTH 1987; ROTHet al. 1993). Within the eukaryotic metabolism (SCHNEIDERand STRO~NSKI 1987; cluster, genes encoding the enzymes for the threeparts STRO~NSKI 1987).Cobalamin is derived from uroporph- of the pathway are clustered (Figure 1).Mutations con- yrinogen I11 (Uro 111), a precursor in the synthesis of ferring CobI-, CobIII-, and CobII- phenotypes are lo- heme, siroheme, cobamides and chlorophylls (in pho- calized, in that order,within the cob operon ( JETER and tosynthetic organisms). As detailed in Figure 1, the con- ROTH1987; ESCALANTE-SEMERENAet al. 1992; O'TOOLE version of Uro I11 to the complex coenzyme BI2 inter- et al. 1993; ROTHet al. 1993). Additional genes outside mediate cobinamide is referred to as part I of the this cluster encode both biosynthetic enzymes for corri- biosynthetic pathway (hereafter denoted CobI) . Part noid adenosylation (ESCALANTE-SEMERENAet al. 1990) I1 of the pathway (CobII) entails the biosynthesis of and synthesis of aminopropanol, thelinker between the dimethylbenzimidazole (DMB) from probable flavin corrin ring and the lower ligand (GRABAUand ROTH precursors (HORIGand RENZ 1978;JOHNSON and ESCA- 1992); transportof cobalamin is also encoded by genes LANTE-SEMERENA1992; LINCENSet al. 1992; CHENet al. unlinked to the cob operon (RIOUX and KADNER 1989; 1995). The covalent joining of cobinamide, DMB, and RIOUX et al. 1990; BRADBEER1991). Despite this large a phosphoribosyl group donated by nicotinate mono- genetic investment in cobalamin synthesis and acquisi- nucleotide is achieved by part I11 of the pathway (Cob tion, entailing nearly 1% of the S. typhimurium chromo- 111). Among the enteric bacteria, both Salmonella typhi- some, only four known, and one likely, cobalaminde- murium ( JETER et al. 1984) and Klebsiella pneumonia pendent reactions have been characterized among the (ALBERT et al. 1980) are known to synthesize cobalamin enteric bacteria: de novo under anaerobic conditions. These organisms 1. Homocysteine methyltransferase catalyzes the final can perform all three parts of the biosynthetic pathway. step in methionine biosynthesis. The metE and metH Escherichiacoli synthesizes cobalamin only when pro- genes encode cobalamin-independent and cobalamin- vided with cobinamide (VOLCANIet al. 1961); therefore, dependent methyltransferases, respectively (SMITHand E. coli performs only parts I1 and I11 of the cobalamin CHILDS 1966;CHILDS and SMITH1969), in S. typhimu- biosynthetic pathway. rium. These genes are also found in E. coli (DAVISand MINGIOLI1950). Ethanolamine-ammonia lyase converts ethanol- Corresponding authm:John Roth, Department of Biology, University 2. of Utah, Salt Lake City, UT 84112. amineinto acetylaldehyde and ammoniaand is en- E-mail: [email protected] coded by the eutBC genes in s. typhimurium; the enzyme

Genetics 142 11-24 (January, 1996) 12 J. G. Lawrence and J. R. Roth

FIGURE1.-The metabolism of tetrapyr- roles in Salmonella typhimunum. Heavy arrows Mmute 41 Miule 34 denote reactions specific for cobalamin biosynthesis. Gray arrows denote regulatory in- pdu operon 11 IlpocR I I cbiARCDEFFGHJKLM~QP cobUS cob7 teractions. Abbreviations are as follows: Ado, (cobalamin-dependent) adenosyl; Cbi, cobinamide; CN, cyano; DMB, dimethylbenzimidazole; FMN,flavin mono- L nucleotide; Uro 111, uroporphyrinogen 111. CobI CobIII CobII The CobI, CobII, and CobIII pathways are ~RNAC‘” ,- UroIII -b t2: 7DMB +FMN discussed in the text. The hagenes are de- I scribed by Xu et al. (1992); the cbi and cob hemEFGIjN 4ACNCobalamin genes are described by ROTH et al. (1993) and CHENet al. (1995).

is also present in strains of E. coli and K. aerogenes minsynthesis andcobalamindependent metabolism (CHANCand CHANC1975; SCARLETT and TURNER1976; among enteric bacteria. In this manner, the selective ROOF and ROTH 1988, 1989). influences operating on cobalamin synthetic genes may 3. Propanediol dehydratase catalyzes the conversion be elucidated. We suggest that the Salmonella cob and of 1,2-propanediol to propionaldehyde and is encoded pdu operons were acquired by horizontal transfer after in the pdu operon in S. typhimurium ( JETER 1990; BOBIK the loss ofthese genes by a common ancestor of both E. et al. 1992); the pdu operon maps adjacent to the cob coli and Salmonella spp. The evolution of the cobalamin operon described here. In this paper, the generalterm synthesis and utilization may provide a general model propanediol will always refer to the 1,2 isomer. Cobala- for understanding the evolution of functions and phe- rnindependent diol dehydratases, activewith 1,2 notypes under selection that is temporally or spatially ethanediol and propanediol, have also been described heterogeneous. in Citrobacter and Klebsiella species (ABELES and LEE 1961; FORAGEand FOSTER1979; TORAYAet al. 1979, MATERIALSAND METHODS 1980; OBRADORSet al. 1988). 4. Glycerol dehydratase is an enzyme distinct from Bacterial strains and culture conditions: Strains of enteric diol dehydratase (TORAYAand FUKUI 1977),allowing bacteria employed in this study are listed in Table 1. Strains of the SARA collection (BELTRANet al. 1991) of natural iso- anaerobic fermentation of glycerol in species of Kleb lates of Salmonella spp. were kindly provided by K. SANDERSON siella and Citrobacter (SCHNEIDERet al. 1970; FORAGE and H. OCHMAN.Strains of the ECOR (OCHMANand SEL and FOSTER 1979,1982); this enzyme has not been de- ANDER 1984) and SARB (BOYD et al. 1993) collections of natu- scribed among strains of Salmonella or E. coli. ral isolates of E. coli and Salmonella spp, were kindly provided 5. Queuosine synthetase catalyzes the final step in the by H. OCHMAN.An E. coli strain harboring a mtE:: Tn IOinser- tion was provided by G. STAUFFER;the wild-type K. aerogenes synthesis of queuosine, a hypermodified base present strain M5al was provided by G. ROBERTS.Additional species in four tRNA andhas been reportedto be acobalamin- of Salmonella and other species of enteric bacteria were ob- dependent enzyme in E. coli (NOGUCHIet al. 1982; FREY tained from laboratory collections. The rich medium used was et al. 1988). LB defined media included E medium (VOGELand BONNER The known cobalamindependent reactions in S. typh- 1956) as well as its derivative NCE, which lacks carbon compounds. MacConkey indicator agarwas used with the addition imunum do not obviously justify this organism’s large of specific carbon sources to 1% concentrations. For anaero- genetic investment in cobalamin biosynthesis and trans- bic growth on solid medium, cultures were grown under an port. The cobalamindependentmethionine synthetase atmosphere of 89% NP, 6% COP,and 5% HP. Liquid culture is redundant, and the queuosine synthetase is appar- media were prepared, degassed for 24 hr, inoculated, and ently nonessential under laboratory conditions. Pro- sealed in tubes with aluminum caps under the anoxic atmosphere described. Headspace gas was exchanged three times panediol utilization appears to be the primary use for with 100% N2 under a pressure of 2 atmospheres. Loss of cobalamin in S. typhimun’um since the adjacent cobala- pressure in a sealed tube during incubation indicated a possi- min biosynthetic genes (cob) and the propanediol deg- ble decay of anoxic conditions and the experiment was re- radative operon (pdu) are coregulated by the same pro- peated. tein (PocR) and are both induced in the presence of Cobalamin synthesis bioassay: To assay ability to synthesize cobalamin, strains were grown to stationary phase on a de- propanediol (BOBIK et al. 1992; CHENet al. 1994). A fined medium under appropriate conditions. Cells were re- paradox is evident in that propanediol serves as a sole covered from a volume of 1.5 ml (aerobic) or 4 ml (anaero- carbon and energy source only in the presence of oxy- bic) of growth medium by centrifugation, were rinsed with a gen while cobalamin is synthesized & nmo only under solution of 50 mM Tris pH 8.0 and 100 mM NaCI, recovered anaerobic growth conditions. by centrifugation, and resuspended in the same solution at 5 pl/mg cell weight. The cells were lysedby boiling for To understand the large investment made by Salmo- 15 min; cell debris was removed by centrifugation. Sterile nella spp. in cobalamin metabolism despite its nonobvi- filter disks were soaked with 10 pl aliquots of the appropriate ous utility, we have examined thedistribution of cobala- dilutions of this supernatant and were placed on solid mini- Evolution of Cobalamin Synthesis 13

TABLE 1 strain list

Species Strain (s) Phenotype Citrobacterfreundii TR7173 Natural isolate Escherichia blattae ATCC 29907, 33429, 33430 E. coli ECOR 1-72 Natural isolates* TR7177 (W3100; K-12) Wild-type laboratory strain IT17196 pe3 Alac-6 supE44 xyM hG218 ilvC7 metBl rpsLlO9 eutA403meE 163:Tn10 E. fmgusonii ATCC 35469-35473 Clinical isolate E. hannii ATCC 33650-33652 Clinical isolate E. vulneris ATCC 29943, 33821, 33832 Clinical isolate Enterobacter cloacae TR7174 (E482) Natural isolate Klebsiella aerogenes TR7176 (M5al) Laboratory strain Klebsiella pneumonia TR7175 (LD119) Natural isolate Salmonella typhimurium IT10858 metE205 are9 cob66::MudJ TTl 1855 metE205 are9 DEL299 (his-, cob-) TR7 172 metA53 IT14305 metE2113::MudJ trplOl Salmonella sp. SARA 1-72 Natural isolates‘ SARB 1-72 Natural isolatesd S. seminole TR6332 Natural isolate Serratia jicari ATCC 33105 S. fonticola ATCC 29844 S. grimesii ATCC 14460 S. liquefaciena ATCC 27592 S. marcesens ATCC 13880 S. odmyera ATCC 33077 S. plymuthica ATCC 183 S. p-oteamaculans ATCC 19323 Shigella dysenteriae ATCC 13313 S. jlexnerii ATCC 29508 S. sonnei ATCC 29930 “All laboratory S. typhimurium strains are derived from strain LT2. Sources of non-Salmonella species are described in LAWRENCE et al. (1991). Alternate strain names are in parentheses. ’The ECOR collection is described by OCHMANand SELANDER(1984). The SARA collection is described by BELTRANet al. (1991). dThe SARB collection is described by Born et al. (1993). mal medium seeded with lo6cells of S. typhimurium indicator minimal glycerol medium when provided cobalt or cobala- strain TT10858 or TT11855. A mutation in the metE gene of min; strains unable to ferment glycerol required fumarate or these strains requires them to utilize the cobalamindepen- nitrate as an electron acceptor. Minimal medium and Mac- dent MetH protein for methionine synthesis, and the Cob111 Conkey indicator agar provide insufficient cobalt to allow de mutations present in these indicator strains preclude cobala- novo synthesis ofcobalamin under aerobic conditions by most min biosynthesis. Growth of the indicator strain on minimal of the species of bacteria tested. Therefore, when a positive medium requires an exogenous source of either methionine result on either medium (growth on minimal medium, color or cobalamin. Growth of the indicator strain around the filter change on MacConkey)was dependent on either cobalt or disk indicated that the cell lysate contained either cobalamin cobalamin, that result was judged cobalamin-dependent. All or methionine. None of the cell lysates tested provided suffi- strains showing a positive result upon cobalt addition also cient free methionine to allow growth of the metA defective showed a positive result upon cobalamin addition. Therefore, strain TR7172, indicating that all positive responses indicated it is unlikely that cobaltdependence reflected a cobaltdepen- the presence of cobalamin in the lysate. The radius of the dent enzyme. Onlyone such enzyme, methionine aminopepti- growth ring for various dilutions of the lysate were compared dase, is known among enteric bacteria (RODERICKand MAT- to a standard curve. Positive results are seen with >0.5 ng THEWS 1993). In all strains that showed a cobaltdependent cobalamin present in the aliquot applied to the disc. growth response, the cobaltdependent synthesis of cobalamin Assays of cobalamin usage: To determine if propanediol was verified by the bioassay described above. and ethanolamine were degraded in a cobalamindependent DNA hybridization: DNA fragments for use as radioactively fashion, MacConkey plates were supplemented with either labeled probes in Southern hybridizationswere prepared carbon source to 1% final concentration. Acid excretion only from the S. typhimuriumpocR regulon as follows. The Cob1 upon the addition of cobalt or cobalamin is recorded as a gene fragments were isolated as the following restriction frag- cobalamindependent change in color. Strains capable of co- ments from plasmids p41-1, p53-3, pJEl, and pJE2 (plasmid balamindependent fermentation of glycerol (as a sole carbon descriptions and sequence coordinates from ROTH et al. and energy source) were able to grow anaerobically on NCE 1993): (1) a 2490-bp MluI fragment (bp 1218-3705) carrying 14 J. G. Lawrence and J. R. Roth

portions of the cbiA, cbiB, and cbiC genes, (2) a 900-bp MluI &,SptheSiS Utilization fragment (bp 3705-4607) carrying portions of the cbiC and Dendrogram SIT& Species I (a2)I (-09 It I11 Gly PDiol EA cbiD genes, (3) a 4360-bp HpaI fragment(bp 2115-6476) b"29 S. saintpnul - + + + - + + carrying portions of the cbiA,cbiB, cbiC, cbiD, cbiE, and cbiT 70S.m&chen - + + + - + + genes, (4) a 920-bp HpaI fragment (bp 6476-7399) carrying 48 Sparalyphi - + + + - + + I 56 S.paralyphi - + + + - + + portions of the cbiT, cbs, and cbiG genes, (5)a 4490-bp HpuI 62 S, paratyphi - + + + - + + fragment (bp7793- 12080) carrying portions of the cbiG, cbiH, 69 S.muenchen - + + + - + + cbiJ, cbiK,cbiL, and cbiM genes, (6) a 780-bp HpuI fragment . 61 S. paratyphi - + + + - + + 71 S.muenchen - + + + - + + (bp 12080-12857) carrying portions of the cbiM,cbiN, and RS.muenchen - + + + - + + cbiQ genes, and (7) a 2150-bp HpuI fragment (bp 12857- -67 S. muenchen - + + + - + + 15006) carrying portions of the cbiQcbi0, and cbiP genes. 1 ~66S.muenchen - + + + - + + The CobIII/II gene fragment was isolated as a 2180-bp CluI/ "65S.muenChen - + + + - + + 64 S.muenchen - + + + - + + BstEII restriction fragment from plasmid pJE2 bearing por- 63 S.muenchen - - - - - ++ tions of the cbiP, cobU, cobs, and cobTgenes. Additional CobIII I 68S.muenchen - + + + - + + DNA was amplified via PCR (SAIKIet al. 1985, 1988). The pocR 49 S. paratyphi - + + + - + + 47 S. pratyphi - + + + - + + gene fragment was isolated as a 1005-bp (bp 213-1218) MluI _I -46 S.paratyphi - + + + - + + fragment from plasmid p51-3. -45 S. pratyphi - + + + - + + Chromosomal DNA was isolated from bacterial strains and 44S.paratyphi ndnd nd nd nd nd quantified by spectrophotometry as described (SAMBROOKet - - 43S.pmatyphi - + + + - + + al. 1989). Plasmid DNA was prepared from Qlagen columns 42 S. paratyphi - + + + - + + 41 S. paratyphi nd nd nd nd nd nd nd according to the manufacturer's instructions. DNA was di- 50 Sparatyphi - + + + - + + gested by restriction endonucleases, and DNA fragments were 51 S.paratyphi - + + + - + + separated by electrophoresis,transferred to nylon mem- 152S. paratyphi - + + + - + + '53 S. paratyphi - - branes, andbound byUV irradiation using conventional - + + + + + 44S.paratyphi - + + + - + + methods (SAMBROOKet al. 1989). Two membranes were pre- -55 S. paratyphi - + + + - + + pared from the same DNA samples by a bidirectional transfer 57 S. paratyphi - + + + - + + method. Following electrophoresis and denaturation of the 58 S. paratyphi - + + + - + + - 59 S. paratyphi - + + + - + + DNA within the agarose gel, nylon membranes were placed 60 S. paratyphi - + + + - + + on both sides of the gel. Both membranes were covered with 30 S.heidelberg - + + + - + + Whattman 1MM filter paper, and the gel/membrane assem- 31 S.heidelberg - + + + - + + bly was placed between two stacks of paper toweling. In this - 32 S.heidelberg nd nd ndnd nd nd nd '33s.heidelberg - + + + - + + manner, equal amountsof DNA were transferred to each of r ~34S.heidelber.g - + + + - + + the two membranes. One membrane from each membrane- 4oS.heidelberg - + + + - + + pair was consistently used as a DNA hybridization control. 38 S.heidelberg - + + + - + + The DNA fragments described above were labeled to high 39 S.heidelberg - + + + - + + 35 S.heidelberg - + + + - + + specific activity with "'P-dATP and hybridized to the nylon- 36 S.heidelberg - + + + - + + bound DNA as described (FEINBERG andVOGFLSTEIN 1983). 37 S.heidelberg - + + + - + + - 11 S. typhimurium - + + + - + + Genetic methods: Transduction crosses mediated by bacte- - + + + - + + riophage P22 were performed as described (DAVISet al. 1980) €6 23 S. saintpaultyphimurium - + + + - - + using the high transducing mutant P22 HT int-205. Bacterio- 24 S. saintpaul - + + + - + + phage P1 lysates were prepared using the high transducing 25 S. saintpaul - + + + - + + " mutant P1 vir, kindly provided by P. HIGGINS.Fresh overnight 26 S.saintpau1 - + + + - + + '27 S. saintpaul - + + + - + + cultures of E. coli W3110 were diluted to 1 X lo7 cfu/ml in '28 S. saintpaul - + + + - + + - - LB, 5 mM CaCI,, 0.2% glucose and grown for 30 min at 37". b - 10 S. typhimurium + + + + - Bacteriophage P1 was added to a final concentration of 1 X 22 S. saintpaul nd nd nd nd nd nd 8 S. typhimurium - + + + - + - 10' pfu/ml, and the cultures were grown for an additional 4 7 S. typhimurium - + + + - + + hr, or until the culture cleared. Cell debris was removed by 9 S. typhimurium - + + + - + - centrifugation and lysates were stored over three drops of 14 S. typhimurium nd nd nd nd nd chloroform at 4". For bacteriophage PI-mediated transduc- 1 S. typhimurium - + + + - + + 2 S. typhimurium - + + + - + + tional crosses, fresh overnight cultures of E. coli W3110 were 3 S.typhimurium - + + + - + + diluted to 1 X lo7 cfu/ml in LB and grown for 2 hr at 37". 4 S. typhimurium - + + + - + + Cells were recovered by centrifugation and resuspended in 5 S. typhimurium nd nd nd nd nd nd nd an equal volume of 5 mM CaCI,, 10 mM MgS04.A volume 18 S. typhimurium - + + + - + + 15 S. typhimurium - + + + - + + of 100 pl cells was added to an equalvolume of bacteriophage 16 S. typhimurium - + + + - + + P1 lysate and incubated for 30 min at 30". The sample was 17 S. typhimurium - + + + - + + added to 300 pl 1 M sodium citrate and plated on the appro- 19 S. typhimurium - + + + - + + 20 S. typhimurium - + + i - + + priate medium. 13 S. typhimurium - + + + - + + 21 S. typhimurium - + + + - + + 12 S. - + + + - + - RESULTS typhimurium FIC:URE2.-Cobalamin phenotypes among strains of the Distribution of cobalamin synthesis among Salmonella Salmonella SARA collection (BELTRANet al. 1991). For coen- spp.: The laboratory strain S. typhimuriurn LT2 synthe- zyme B,* synthesis, I, 11, and I11 refer to the CobI, CobII, and sizes cobalamin de novo under anaerobic conditions,but CobIII portions of the biosynthetic pathway (see text). CobI the distribution of this function among other Salmo- function was assayed under aerobic (denoted +O,) and anaerobic (denoted -0,) conditions. The abilities to ferment nellaspecies was unclear. A cobalaminbioassay (de- glycerol (Gly), utilize propanediol (PD), and utilize ethanol- scribed in MATERIALS AND METHODS) was employed to amine (EA) in cobalamin-dependent fashions were tested as ascertain the distribution of cobalamin synthesis among described in the text. nd, not done. lamin Synthesis 15 Synthesis Evolution of Coballamin the Salmonellae. Natural isolates of certain Salmonella ~,Synthesi Utilization Dendrogram Strain Species I (-,)I (-03 11 111 Gly PDiol subspecies collected by R. K SELANDERand colleagues FA 14 S. dublin -+++-++ (BELTRANet al. 1991; BOYD et al. 1993), as well as ad- 22 s. haqa -+++-++ ditional subspecies (see Table1) were tested for cobal- 55 S. saintpaul -+++-++ 65 S. typhimurium -+++-++ amin production and use (Figures 2 and 3). Nearly 66 S. typhimurium - + ++- + - 67 - + ++-++ all of the Salmonella strains tested were capable of S. typhimurium 68 S. typhimurium -+++-++ cobalamin synthesis. In all cases, de novo cobalamin 46 S. paratyphi B - + ++-++ 45 S. puratyphi B - + ++-++ synthesis occurred only under anaerobic growth con- 34 S. muenchen - + ++-++ ditions. Among the 138SARA and SARB strains tested, 32 S. muenchen -+++-++ 43 S. parahphi B -+++-++ fourstrains (2.9%) did not synthesize cobalamin. 44 S. paratyphi B - + ++-++ Since these four strains are not phylogenetically re- 42 S. paratyphi A - + ++- -+ 58 S. di - + ++-+- lated (Figures 2 and 3),these biosynthetic deficiencies 18 s. entm'tidis - + ++-++ are likely to represent four independent losses of co- -21 S. gallinarum - - ""- 51 S. pullorum - + ++- - - balamin synthetic capacity. 52 S. pullarm - + ++- - - The cobalamin biosynthetic genes of S. typhimum'um 62 S. thompson - + ++-++ 35 S. muenchen - + ++-++ LT2 are induced by propanediol and high levels of the 37 s. newport - + ++-++ Crp/cAMP complex via the PocR regulatory protein 38 s. newport - + ++-++ 56 S. saintpaul - + ++-++ (ESCALANTE-SEMERENAand ROTH 1987; ANDERSON and k 15 S. duisburg -+++-++ ROTH 1989a,b; BOBIKet al. 1992; RONDONand ESCA- 10 s. derby -+++-++ 2 S.anatum - + ++-+- LANTE-SEMERENA1992; AILION et al. 1993). To deter- 23 s. heidelberg -+++-++ mine thegenerality of this regulatory model, cobalamin 24 S. heidelberg -+++-++ 60 S. stanley -+++-++ production was quantitatedamong even-numbered 61 S. stanleyville - + ++-++ SARB strains grown on a variety ofsubstrates. All strains 71 S. wien -+++-++ 72 s. wien -+++-++ produced the greatest amount of cobalamin (3-15 ng/ 7 S. cholermuis -+++-++ pg wet weight cells) during anaerobic respiration on 20 S. onek -+++-++ 25 S. indiana - - -+-++ pyruvate/propanediol/fumarate. A reduced amountof -€19 S. entm'tidis - + ++-++ cobalamin was produced during growth on glucose/ 16 S. entm'tidis - + ++-++ L 13 S. dublin -+++-++ fumarate (<0.1 ng/pg), pyruvate/fumarate (<0.1 ng/ € 12 S. dublin -+++-++ pg), or glucose/fumarate/propanediol (<0.1 ng/pg). 11 s. derby -+++-++ 69 S. typhisius - + ++- - + These results show that the pattern of regulation seen 49 S. parahphi C - + ++" - 48 S. paratyphi C - + ++" - for S. typhimurium LT2, requiring both propanedioland 4 s.cholermuis - -++-+- high levels of the Crp/cAMP complex for induction, 6 S. cho~~is - + ++-++ appears to be generalized among Salmonella species. 5 S. cholermuis - + ++-++ 3"c=:26 S. mfintis - + ++-++ Distribution of cobalamin-dependent metabolism 1 S.agona - + ++-++ 36 S. navwrt - + ++-++ among Salmonella spp.: Like S. typhimurium LT2, nu- 9 S.der& - + ++-++ merous natural isolates of Salmonella employed co- 63 S. typhi - + ++- - - 64 S. typhi - + ++- - - balamin for the degradationof propanediol andetha- 59 s. mflenberg - + ++- ++ nolamine; no strain was found todegrade either 3 s.brandenburg - + ++-++ 17 S. enteritidis -+++-++ compoundin cobalamin-independenta fashion. 41 S. panam + ++ + + Among the 138 isolates tested, 11 strains of Salmo- 28 S. miami + ++ + + 29 S. miami + ++ + + nella had lost the ability to degrade ethanolamine. 39 s. pama + ++ + + Inspection of the phylogeny reveals that these strains 40 S. panama + ++ + + 50 S. paratyphi C + ++ + + are likely to representat least seven independent 54 S. rubislaw + ++ + + losses of this ability. Nine strains have lost the ability 31 S. monteveoideo + ++ + + 30 S. montevideo + ++ + + to degrade propanediol; these strains arelikely to rep- 27 S. mfantis + ++ + + resent at least four losses of this ability. No strain of 53 S. rending + ++ + + 57 S. sfhwartzengrund + ++ + + Salmonella was found to ferment glycerol, that is, 8 S. deentur + ++ + + grow anaerobically on minimal glycerol medium with- 70 S. hmhisius + ++ + + 33 s. muenchen - + ++-++ out an electron acceptor.For these assays, cobalamin -47 S. paratyphi B - + ++-++ was added to the growth medium. These results show that the patternsof cobalamin utilization found in the FIGURE3.-Cobalamin phenotypes among strains of the SalmonellaSARB collection (BOYDetal. 1993).See the legend laboratory strain S. typhimum'um LT2 can be general- for Figure 2 for abbreviations for phenotypic characters. ized to include strains representing the diversity of the Salmonellae. Moreover, although the sample size of the other or with loss of cobalamin synthetic ca- is small, ethanolamine degradation was found to be pacity. lost nearly twice as often as propanediol degradation. Distribution of cobalamin synthesis among E. coli: In The loss of neither phenotype was correlated with loss contrast to Salmonella isolates, no strain of E. coli could 16 J. G. Lawrence and J. R. Roth

cobalamin, all strains are phenotypically CobI-. Unlike the characterized laboratory strains, 38% of natural iso- 10 - - -+-- + lates of E. coli require exogenous DMB in addition to 11 - - "" - 25 - - ++" + cobinamide for cobalamin biosynthesis. Therefore, not all strains bear CobII gene functions. We cannot formally eliminate the possibility that E. coli strains harbor Cob1 and CobII gene functions that remain inactive under test conditions. The 27 strains lacking CobII 7 ++ - + + function do not form a monophyletic group (Figure 14 -+ - + 4);rather, they are likely to represent 5 15 independent 20 ++ - + 21 ++ - + losses of this pathway among E. coli lineages. Current 13 ++ - + 18 -+ - + evidence suggests that the CobII function is provided 19 ++ - + ++ - + by a single gene, cobT (CHENet al. 1995). o+ - Unlike the regulated coenzyme BI2biosynthesis ob- -+ - + -+ - + served in Salmonella species, the levels of cobalamin ++ - + ++ - + produced from cobinamide in E. coli strains were inde- ++ - + pendent of growth conditions, including carbonsource ++ - + + 30 - - ++- + + (Crp/cAMP levels) and the presence of propanediol; 32 - - -+-- + 33 - - -+-- + similar amounts of cobalamin weresynthesized in - ++" + strains grown in minimal glucose medium and in succi- - ++- - + 68 - - -+" + nate/propanediolmedium (-5ng/pg cells). These 42 - - ++" + 67 - - ++-- + data suggest that regulation of the few cobalamin syn- ++ + + ++ + + thetic genes of E. coli differs from that of the S. typhimu- -+ + rium cob operon. 51 ++ 54 -+ Distribution of cobalamindependent metabolism 55 -+ + 56 ++ + among E. coli: When assayed for cobalamin-dependent 52 ++ functions, only ethanolamine degradation was found 57 -+ 63 ++ among tested E. coli isolates. No strain of E. coli was 61 ++ + 53 ++ found to degrade either glycerol or propanediol in a 60 ++ cobalamin-dependent fashion under either aerobic or 59 " + -+ anaerobic growth conditions. Although most E. coli iso- " + " + lates could degrade ethanolamine in a cobalamin-de- ++ + pendent fashion, 13 strains could not. These strains are " ++ + -+ + likely to represent at least eight independent losses of 26 - ++- + + this ability. Among strains unable to degrade ethanol- 27 - ++- + + 69 - - ++" + amine, the majority (9/12) had lost CobII function as I' 48 - - ++" + well, twice as many(75%) as expected (38%).Although 35 - - ++-++ 36 - - ++-++ the sample size is small, these data suggest that the loss 38 - - -+-++ 40 - - -+" + of eut function may influence the maintenance of the 41 - - -+-- + 39 - - -+-++ CobII gene functions amongE. coli strains. These results 46 - - ++" + show that the patternsof both cobalamin synthesis and 44 - - ++" + 47 - - ++- + + use differ substantially between strains of E. coli and 49 - - ++- + + 50 - - ++-+- Salmonella spp. 37 - - ++- + + Although some 20% of E. coli isolates could degrade FIGURE4.-Cobalamin phenotypes among E. coli ECOR propanediol, they did so only anaerobically and in a strains. Strain numbers refer to the ECOR collection (OCH- cobalamin-independent fashion (Table 2). Lactalde- MAN and SELANDER1984); dendrogram is after SELANDERet hyde can be converted to lactate in E. coli by the ald al. (1987). See the legend for Figure 2 for abbreviations for gene product. Since propanediol is converted to lactal- phenotypic characters.0, no growth. Propanediol is utilized dehyde anaerobically by the fucO gene product, consti- in a cobalamin-independent fashion among thesestrains (see text). tutive expression of the fucO gene could give the observed phenotype. Such phenotypes have been synthesize cobalamin de novo (Figure 4).Yet most strains observed among mutant isolates of laboratory strains retained thecapacity to synthesize cobalamin when pro- (SRIDHARA et al. 1969; HACKING et al. 1978; OBRADORS vided with complex precursors; that is, most retain Cob- et al. 1988). In contrast, no Salmonella isolate degraded I11 gene functions (Figure 4, see also Figure 1). Since propanediolin a cobalamin-independent fashion. S. all strains of E. coli require cobinamide to synthesize typhimurium lacks the ald gene and cannot convert lac- E volution of Cobalamin Synthesis 17 Synthesis Cobalamin of Evolution

TABLE 2 Phenotypes of selected enteric bacteria on MacConkey-propanediol indicator agar

Aerobic additionb additionAnaerobic

Species Phenotype" Strain - co2+ B12 - co2+ B12 E. coli TR7177 Wild type W W W W W W E. coli ECOR 36 FucO' W W W R R R K. pneumoniae TR7176 Wild type W R R W R R S. typhimurium LT2 Wild Type W W R W R R S. typhimurium TT10858 Cob- W W R W W R S. typhimurium TT11855 Cob-Pdu- W W W W W W Relevant phenotype of strain; full genotypes are presented in Table 1. The FucO' phenotype of E. coli strain ECOR 36 was inferred by analogy to the behaviour of E. coli K12 mutants having the same phenotype (see text). Additions to MacConkey-propanediol agar medium were made as described.Colors of colonies were either red (R) or white (W). Red colony color indicated that propanediol was consumed and acid was excreted as a consequence. taldehyde into lactate (A. LIMONand J. AGUILAR,per- However, we cannot formally excludethe possibility sonal communication). that our failure to observe cobalamin synthesis in some Distribution of cobalamin synthesis among enteric species was due to inappropriate growth conditions. bacteria: Biosynthesis of cobalamin is evident among Unlike strains of Salmonella, however, cobalamin numerous species of enteric bacteria (Figure 5). K. production among other entericspecies was unaffected pneumonia, K. aerogenes, C. freundii, and E. vulnmk, E. by growth medium and oxygen levels. Strains grown on hermannii, E. blattae and Enterobacter cloacae all synthesize glucose or succinate/propanediol synthesized compa- cobalamin de novo under anaerobic growth conditions. rable amounts of cobalamin (3-15 ng per pg cells). Unlike Salmonella spp., cobalamin synthesis under aero- Since neither Crp/cAMP levels nor thepresence of pro- bic conditions in these taxa was facilitated by the addi- panediol affects cobalamin synthesis among non-Salmo- tion of excess cobalt (10 pM). Strains were grown in a nella enteric species, these data suggest a modeof regu- variety of different media; growth conditions were var- lation of cobalamin synthesis among non-Salmonella ied with respect to the presence of ethanolamine, glu- enteric bacteria that differs fundamentally from that cose, glycerol, oxygen, propanediol, pyruvate, and suc- observed in Salmonella. The only factor strongly influ- cinate. These growth conditions included those known encing the productionof cobalamin among non-Salmo- to induce or repress cobalamin synthesis in S. typhimu- nella species was the addition of exogenous cobalt. It rium (BOBIKet al. 1992). Comparablelevels of synthesis was inferred that the levels of cobalt in the local water of coenzyme BI2were detected among all non-Salmo- supply typically supplied insufficient cobalt to allow de nella species regardless of growth conditions. In most novo cobalamin synthesis under aerobic conditions for cases, multiple strains of each species were tested (Table non-Salmonella species. However, the cobalt levels in 1); results were congruent among conspecific strains. water were typicallyadequate forde novo cobalamin synthesis under anaerobic conditions forall species. From

Dendrogram Species these data, we inferred that the expression of cobalt

Salmonella spp. transport mechanisms may be induced under anaerobic Salmonella seminole growth conditions. Escherichia fergusonii Species of Serratia, including S. jicari, S. grimesii, S. - liquefaciena, S. marcesens, S. odm@-a,S. plymuthica, S. pro- scherichia coli - ~ +I-+ - - + teamaculans, were found to lack cobalamin synthetic ca- -Citrobacterfreundii + + +++ + + pacity. One species, s. fonticola, showed Cob11 and Cob- - Escherichia hennannii + + +++ + - Escherichia mrlnm's + + +++ + + I11 gene functions (Figure 5). These taxa are among - Enterobacter cloacae + + ++- + - the entericspecies tested most distantly related to Salmo- Klebsiella spp. + + +++ + + nella spp. and are generally considered to be soil-resi- Escherichia oulMlis + + +++ + + dent bacteria. The relationship between the lack of co- - Escherichia blattae + + +++ - - balamin synthesis and the ecology of Serratia species is Serratia spp. - - - - - unclear. Serratia jonticola " ++- - - Morganella morganii Distribution of cobalamin-dependentmetabolism among enteric bacteria: Cobalamin-dependent use of FrCURE 5."Coba1amin phenotypesamong enteric bacteria. propanediol was widespread among enterics, as was use Dendrogram is after LAWRENCE et al. (1991). The species of ShiEella, Salmonella,Klebsiella, andSerratia tested listedare Of Both functions were absent from in Table 1 and Figures 2 and 3, See the legend for Figure 2 strains Of E. blattae; E. cloacae also failed to Utilize etha- for abbreviations for phenotypic characters. nolamine in a cobalamin-dependent fashion. Cobala- 18 Lawrence J. G. Rothand J. R.

TABLE 3 Phenotypes of selected entericbacteria on defined glycerol medium

Anaerobic addition" COX+ Species Strain Species - B12 Fumarate Glucose Aerobic' E. coli TR7177 - - - + + + K. pnmmoniae TR7176 - + + + + + S. tvphimurium LT2 - - - + + + "Additions to NCE glycerol medium were made as described. +, growth on the supplemented NCE glycerol medium under anaerobic conditions. " +, indicate growth on NCE glycerol medium under aerobic conditions. mindependent fermentation of glycerol was found in bacteria, cleavedwith restriction endonucleases, size most species of enteric bacteria tested, including E. blat- fractionated by agarose gel electrophoresis, and trans- tae (see also Table 3). The clade comprising strains of ferred to nylon membranes. The DNA fragments de- Salmonella spp., E. coli, Shigella spp., and E. fhgusonii did tailed above were employed to detect potential homo- not ferment glycerol even when provided with coen- logues among the entericbacterial chromosomes. zyme BI2.Therefore, the capacity to ferment glycerol Figure 6 shows the hybridization of a DNA fragments correlated best with a strain'sability to synthesize cobal- bearing S. typhimurium cbiD and cbiE genes, encoding amin. All strains that synthesized cobalamin d? novoaer- Cob1 enzymes; the cobU, cobs, and cobTgenes, encoding obically and anaerobically also fermented glycerol an- Cob111 and Cob11 enzymes; and the cobA gene. The cobA aerobically. Species of Serratia, which failed to gene is unlinked to the S. typhimurium cob operon and synthesize cobalamin, did not participate in any of the is homologous to the E. coli btuR gene (Figure 1). The cobalamindependent metabolism tested. CobA and BtuR proteins adenosylate corrins during Verification of bioassays: To verify that the bioassay both biosynthesis and transport of cobalamin. The S. was providing an accurate assessment of a strain's ability typhimurium cobA gene hybridizes to S. arizonae, S. semi- to synthesize cobalamin, mtE mutations were trans- nok, E. coli, and K. pneumonia. For the cbiDE and cobUST duced via bacteriophage P22 (among Salmonella spp.) probes strong hybridization is shown by S. arizonae. S. or via bacteriophage P1 (among E. coli strains)into arizonae is distantly related to S. typhimurium and is otherwise wild-type backgrounds. Introduction of these placed in a separate genus by some taxonomists. The mutations resulted in strains dependent on cobalamin cobA, cbiDE and cobUST probes hybridize strongly to all for methioninesynthesis; inthis manner, thecobalamin cobalamin-synthesizing Salmonella (data not shown). synthetic capacityof the strain was tested directly. The cbiDE and cobUST gene probes do not hybridize to Twelve strains of Salmonella from the SARJ3 and laboratory collections and 12 strains of E. coli from the ECOR collection were tested. In all cases, strains determined toproduce cobalamin anaerobically by the bioassay were found to produce cobalamin by the genetic test in that they could grow without exogenous methionine under anaerobic conditions, butnot underaerobic conditions (data not shown).Growth under aerobic conditions was restored by the addition of exogenous cobalamin. Those strains shown not to produce cobalamin by the bioassay required exogenous cobalamin or methionine for growth on defined medium under both aerobic and anaerobic conditions. We conclude that the m bioassay is an accurate measure of a strain's ability to synthesis cobalamin and is applicable in organisms lacking suitable tools for genetic testing. Distribution of sequences homologousto the S. Whi- cbiDE cobUST cobA murium cob operon among enteric bacteria: To assay FIGURE6.-DNA hybridization using probes from the S. the distributionof homologues of the S. typhimurium cob typhimurium cbD and cbiEgenes; S. typhimurium cobU, cobs, and operon among related bacteria, DNA fragments were cobT genes; and S. typhimurium cobA gene. Abbreviations are as follows: StyLT2, S. typhimurium LT2; StyDcob, S. typhimurium isolated from the pocR regulatory gene andfrom various TT11855, which contains a deletion of the cob operon; Saz, S. parts of the cob operon as described in MATERlALS AND arizonae TR6331; Sse, S. seminok TR6332; Eco, E. coli TR7177; METHODS. Chromosomal DNA was isolated from enteric Kpn, K. pneumonia TR7176. Evolution of Cobalamin Synthesis 19

S. Seminole, which fails to synthesize cobalamin under ditions) and the Crp/cAMP (growth on poor carbon laboratory conditions. More importantly, the cbiDE and sources) complexes are required for induction of both cobUST genes do nothybridize with E. coli or K. pneumo- operons (BOBIKet al. 1992; AILION et al. 1993). As de- nia. Yet both taxa contain cobUST gene functions and scribed, thecobalamin biosynthetic genes amongother K. pneumonia also shows cbiDE gene functions. Similar Salmonella isolates appear to beregulated similarly. results were obtained for all DNA fragments isolated While it is known that the diol and glycerol dehydra- from the cob operon and the PocR regulatory gene; in tases of Klebsiella spp. are regulated by Crp/cAMP and no case did DNA from the S. typhimurium cob operon propanediol levels (RUCHet al. 1974; TORAYAet al. 1978; hybridize with a non-Salmonella strain. Control hybrid- FORAGEand FOSTER1982), cobalamin synthesis in this izations with DNA fragments from the S. typhimurium andother species of enteric bacteria remains unaf- cobA, gap, ompA, and trp genes all detected homologues fected by propanediol or by CRP/cAMP levels. In addi- as strongly hybridizing bands among all strains tested tion, the compoundspyrroloquinoline quinone (PQQ (data for the cobA gene are shown in Figure 6). and aspartic acid have been shown to stimulate cobala- These results indicate that the genes encodingcobal- min synthesis in Klebsiella (OHSUGIet al. 1989);no amin synthetic enzymes of Salmonella spp. are notclosely stimulation of cobalamin synthesis by these compounds related to the analogousof other entericspecies tested. was observed in Salmonella (data not shown).This is a The DNA hybridization conditions employed detect second major difference in the manner by which cobal- DNA sequences >70% identical to theDNA sequences. amin synthesis is regulated in Salmonella species as Typical chromosomal loci shared among the enteric compared with other species of enteric bacteria. bacteria listed in Table 1 are>70% identical (LAW- Cobalamin-dependent utilization of ethanolamine, RENCE et al. 1991). As expected, hybridizations with the propanediol, and glycerol is widespread among enterics Salmonella cobA, gap, ompA, and trp loci detected homo- (Figure 5). However,glycerol fermentation is absent logues amongenteric genomes. Therefore, we con- from the clade including E. coli, Shigella spp., E.fergusonii clude that the cob operon of Salmonella spp. differs sig- and Salmonella spp. (Figures 2-5). In viewing these phe- nificantly atthe nucleotide level from thegenes notypes parsimoniously, one may speculate thatthe encoding cobalamin synthetic genes among other en- ability to ferment glycerol was lost by the common an- teric bacteria. cestor of all these taxa. These phenotypic data strongly suggest that the co- DISCUSSION balamin synthetic genes in Salmonella taxa differ substantially from those present in other enteric bacteria. Although most species of enteric bacteria synthesize When tested by DNA hybridization, it was found that and use cobalamin to some extent, it is clear that species the S. typhimurium cbiDE and cobUST genes did not hy- of Salmonella differ from other enterics in several re- bridize with the genomes of closely related bacteria ex- spects. First, Salmonella species synthesize cobalamin cept those of other Salmonella species (Figure 6). Un- de novo only under anaerobic conditions. In contrast, der identical conditions, the unlinked S. typhimurium all other species of enteric bacteria capable of de novo cobA gene hybridized strongly with homologues in these synthesis do so under both anaerobic and aerobic con- same taxa (Figure 6). Yet all of these genes participate ditions; aerobic synthesis of cobalamin by these organ- in the production of the coenzyme adenosylcobalamin isms invariably required exogenous cobalt (see Table and should be subject to similar selective constraints. 2). Supplementation with exogenous cobalt did not From these results, we conclude that the genes of the allow aerobic cobalamin biosynthesis in Salmonella spe- cob operon of Salmonella spp. differs genotypically from cies. Therefore, synthesis of cobalamin in Salmonella spp. the analogous genes in other enteric bacteria. resembles cobalamin biosynthesis in Propionibacterium Two possible explanations for these data arise: (1) spp. in that pathways in both organisms function most The S. typhimurium cob operon has undergonerapid efficiently under anaerobic growth conditions (MENON evolutionary change, resulting in both an altered mode and SHEMIN1967). In contrast, the cobalamin biosyn- of expression and a significantly altered DNA sequence. thetic pathway among other entericsresembles the Pseu- (2) The coboperon has been introduced into theSalmo- domonas denitniJicans pathway in that these pathways nella genome from an exogenous source as a single function well under aerobic growth conditions (LAGO fragment that included thepdu operon and the shared and DEMAIN1969). For reviews of cobalamin biosynthe- regulatory mechanism. The case of rapid evolution sis, see BATTERSBY(1994) and SCOTT(1990, 1993). seems unlikely for several reasons, aside from the ad In S. typhimurium the cobalamin biosynthetic operon, hoc necessity for locally increased substitution rates. cob, and the propanediol utilization operon, pdu, are 1. DNA fragments from theS. typhimurium cob operon immediately adjacent on thegenetic map and are coor- hybridize as strongly with the genomes of sister Salmo- dinately controlled by regulatory proteins. The PocR nella species as do DNA fragments from other chromo- protein induces thecob and pdu operons in the presence somal genes. If the rate of nucleotide substitution were of propanediol; both theArcAl3 (anaerobic growth con- accelerated forthe S. typhimurium cob operon,one 20 J. G. Lawrence and J. R. Roth Horizontal EvolutionTransferRapid Model Model Functions Regained Functions Modified by Salmonella spp. -//=- Accelerated Mutation + ~~/~//~/~//~~//////////~,. Salmonella Seminole

Escherichiafmgusonii Escherichia coli Diol dehydratase Shigella spp. Function Lost Glycerol dehydratase

Citrobucterfreundii I Escherichia hennnnnii I Eschm'chia vulneris I r Enterobucter clwcue KZebsieIlu spp. Escherichia vulnerii Functions Absent Functions Escherichia bluttue Glycerol dehydratase Cobalaminsthesis Serrutiu spp. c-Morgunellu morgunii FIGURE7.-Models for the evolution of cobalamin synthesis and use among enteric bacteria. Solid bars denote species capable of both glycerol fermentation and aerobic cobalamin synthesis. Open bars denote the lack of these phenotypes. Shaded bars denote species with anaerobic cobalamin synthesis and diol dehydratase phenotypes. Striped bars denote species with modified cobinamide synthetic genes. would have expected divergence and weak hybridiza- in the E. coli/Salmonella ancestor coincides with the tion, among at least the most divergent of the Salmo- loss of the glycerol dehydratase and oxygen-sensitive nella sublineages. diol dehydratases present in allother cobalamin-synthe- 2. The S. typhimurium cob operon bears numerous sizing enteric species. aberrantproperties atypical of other chromosomal Model for the evolution of cobalamin synthesis and genes, including a high %G + C content ("59%) and use among enteric bacteria: We propose the following an unusual codon bias (ROTH et al. 1993; LAWRENCE model for the evolution of cobalamin synthesis and use and ROTH 1995). Unusual codon biases are unlikely to among entericbacteria (Figure 7). Cobalamin synthesis represent accelerated rates of nucleotide substitution, is maintained among most enteric bacteria for use in but rather are likely to represent the codon bias of a glycerol dehydratase. The ancestor of E. coli and Salme foreign donor genome. The unusual features of the cob nella spp. lost the glycerol dehydratase and diol dehydra- operon sequences are shared by known genes of the tase functions and subsequently lost de novo cobalamin pdu operon (T. BOBIK,personal communication). synthetic capacity. Salmonella spp. reacquired the adja- 3. The genes encoding the Cob11 and Cob111 genes cent pdu and cob operons, which provide cobalamin of E. coli have been cloned, and their nucleotide se- synthetic capacity and propanediol degradation func- quences have been determined (LAWRENCE and ROTH tions. These genes were receivedfrom an organism that 1995). Comparisons of the DNA sequences of homolo- synthesized BI2 only during anaerobic growth. Salme gous E. coli and S. typhimurium cob genes show numbers nella spp. maintain cobalamin synthetic capacity via se- of synonymoussubstitutions in excess of those expected lection for the degradationof propanediol. This model even if nucleotide substitutions had accumulated with- is consistent with the coregulation of the S. typhimunum out selection (LAWRENCE and ROTH 1995). These data cob and pdu operons by the PocR protein (BOBIKet al. suggest that the divergence of these gene sets predated 1992; AILION et al. 1993). the divergence of the E. coli and S. typhimunum ge- While this model is consistent with our data, the dif- nomes. ferences between the S. typhimunum cob operon and the 4. A comparison of the DNA sequences of the cobU, cobalamin synthetic genes of other enteric species may cobs, and cobTgenes of S. typhimunum and their homo- reflect internal evolutionary processes. If so, the genes logues in E. coli reveals a bias toward higher %C + C encoding Cob1 and diol dehydratase functions may inexcess of 12% at synonymoussites among the S. have been lost from the ancestor of E. coli and E. fmgu- typhimunum cob genes (LAWRENCE and ROTH 1995). It sonnii (Figure 7). The cob genes of extant Salmonella spP. is unlikely that an accelerated rate of evolution would must have evolved to be functional only under anaero- lead to a direction bias in genic G + C content. bic conditions, regulated by Crp/cAMP, induced by 5. The proposed loss of cobalamin synthetic capacity propanediol, high in %G + C, and twice as divergent Evolution of Cobalamin Synthesis 21

S. Srphimurium - -* FIGURE 8.-Metabolism of methyl- DihydroxyacetoneP pentoses in E. coli and typhimurium. / s. Genes encodingrelevant enzymes are + heavy me- ald noted. The arrows represent Central Lactate Lactaldehyde Propanediol '(KGZ) tabolism unique to typhimun'um. The Metabolism + fuco- 1,2 + S. 0 - shadedarrows represent metabolism unique to E. coli.

Propionaldehyde from their E. coli homologues as typical chromosomal pentoses rhamnose and fucose, whose degradation can genes (LAWRENCE and ROTH 1995). As is the case for yield the diol as a byproduct (Figure 8) (for a review all examples of horizontal transfer, we cannot formally see LIN 1987). Both sugars are catabolized by E. coli exclude this possibility, but we find the necessary as- (EAGON1961) and by S. typhimun'um (OLD andMORT- sumptions to be highly unlikely. LOCK 1977; AKHYet al. 1984) to produce dihydroxyace- Intrinsic to these analyses is the use of the phylogeny tone phosphate (DI") and lactaldehyde (Figure 8). shown in Figure 7. The conclusions discussed here rely The DI" enters central metabolism as a glycolytic upon the statistical confidence of the E. coli/E. fugu- intermediate in both taxa, but the fate of the lactalde- sonii/Salmonella clade. Overwhelming genetic and mo- hyde differs between the two species (Figure 8). lecular data divide E. coli and Salmonella spp. into two In E. coli lactaldehyde can be converted to lactate distinct groups. Although the relationships of these spe- aerobically by the ald gene product (CABALLERO et al. cies to other enteric bacteria are diagrammed precisely 1983). Under anaerobic conditions, E. coli converts lac- in Figures 5 and 7, the confidence of every node of taldehyde into propanediol andexcretes it into theme- the dendrogram has not been determined. Rather, the dium. Mutant strains of E. coli K12 (SRIDWet al. clade comprising E. coli, E. ferff21sonii,and S. typhimun'um 1969; COCKS et al. 1974; HACKING et al. 1978; OBRADORS was shown to be a statistically significant monophyletic et al. 1988) and 20% of natural isolates (Figure 4) can group (LAWRENCE et al. 1991). K. pneumonia, C. freundii, metabolize propanediol anaerobically without coen- E. hennannii, E. vulneris, E. blattae, E. cloacae, and Serratia zyme BIZ. It is deduced that propanediol induces the spp. were shown to lie outside this group. The models FucO enzyme, which converts propanediol into lactal- presented here rely onlyupon the established statistical dehyde, which, in turn, is metabolized by the Ald en- confidence of the distinction between the Salmonella zyme. In this manner, propanediol is metabolized in a and E. coli groups and the distinction of their common cobalamin-independent fashion. This function was ob- clade from all other enteric species tested. served for several other species of enteric bacteria, in- Cobalamin synthesis and use is notably absent among cluding E. vulnais and s. liquefaciena (data not shown). enteric taxa distantly related to Salmonella spp., such Salmonella isolates, in contrast, lack the ald gene Serratia. Species of Serratia differ from the other en- product (A. LIMONand J. AGUILAR,personal communi- teric species analyzed here in that they are isolated pri- cation). As expected, no strain of Salmonella tested in marily from soil and growwell at low temperatures this study was able to utilize propanediol in a cobala- (26"). In contrast, the other enterictaxa are found pri- min-independent fashion. It is unclear, however, marily as gut inhabitants of mammals, avians, reptiles, whether the ald gene function seen in E. coli has been and insects. The significance of these environmental lost from the Salmonella lineage or was acquired by differences and their potential contributionto the the E. coli lineage. Rather, Salmonella spp. employ the maintenance of cobalamin synthesis and use remain cobalamindependent enzyme encoded by the pdu op- unclear. Serratia species may have lost the ability to eron ( JETER 1990) to convert propanediol to propionyl- synthesize cobalamin that was present in the ancestor CoA (Figure 8). Propionyl-CoA can be disproportion- of all the enteric bacteria discussed here. Alternatively, ated by pgencoded enzymes to yield either propanol the ancestor of cobalamin-synthesizingenteric bacteria and ATP, or pyruvate (J. TITTENSOR,personal commu- may have gained this ability after the divergence of the nication). E. coli species lack the pdu operon. lineage containing Serratia species. We view the Salmonella pdu operon and the E. coli Propanediolmetabolism in Salmonella e.and E, ald gene as alternative pathways for lactaldehyde dissim- coli: Consistent with the horizontal transfer model are ilation during the oxidation of methylpentoses. E. coli the substantial differences in the ways Salmonella spe- employs the aZd gene product directly during aerobic cies and E. coli metabolize propanediol. Likely sources growth to convert lactaldehyde to lactate and support of propanediol for enteric bacteria include the methyl- growth (Figure 8). During anaerobiosis, propanediol is 22 J. G. Lawrence and J. R. Roth

created and excreted. Among Salmonella spp., no ald to harbor numerous nontraditional cobamides whose genefunction exists and all lactaldehyde produced a-ligands comprise compounds such as 5-hydroxyben- must be converted topropanediol. To use the pro- zimidazole, 5-methylbenzimidazole, adenine,naph- panediol, Salmonella employs the cobalamin-depen- thimidazole, Z-methyladenine, and 2-methylmercap- dent propanediol dehydratase in the pdu operon (Fig- toadenine (FRIEDRICH1975). When tested in various ure 8). Since Salmonella spp. synthesize coenzyme B12 bioassays, these compounds provided measurable func- only under anaerobic conditions, we presume that it tion as coenzymes (FRIEDRICH 1975).While the func- catabolizes propanediol mainly under these conditions. tions of these compounds in E. coli are unclear; they The large genetic investment that Salmonella commits may serveas coenzymes in thecobamide-dependent to this function may indicate that propanediolis unusu- methionine synthase MetH (DAVISand MINGIOLI1950) ally important in Salmonella biology. A recent method or in the ethanolamine ammonialyase encoded by the for the discrimination of Salmonellae from other en- eut operon (S~ARLETTand TURNER 1976),or they may teric bacteria exploits Salmonella's ability both to syn- play roles in an as-yet-undiscovered cobalamin-depen- thesize cobalamin and to degrade propanediol (RA" dent metabolism. BACH 1990). We have provided evidence that the cobalamin syn- Use of dimethylbenzimidazole in E. coli: Also of in- thetic pathway of S. typhimurium represents a case of terest is the curious lack of CobII gene function among the evolutionary loss of a biological function and the numerous isolates of E. coli (Figure 4). Assuming that reacquisition of thatfunction by horizontal transfer the strain ancestral to the ECOR strains bore CobII from a foreign donor. We suggest that this process may function, inspection of Figure 4 reveals that these func- frequently affect functions under weak selection. The tions have been lost no fewer than 15 times from the resulting turnover of genes within genomes may play set of 72 strains analyzed here. The absence of Cob1 an important role in bacterial evolution. gene functionfrom CobIII' E. coli isolates is not surpris- We thank P. HIGGINS,H. OCHMAN,G. ROBERTS,K SANDERSON, ing. This function requires 17 genes in S. typhimurium and G. STAUFFERfor strains, J. TITTENSORfor sharing unpublished and therefore represents a large target for mutation. data, M. AILION,N. BENSON,T. BOBIK,P. CHEN,D. DYKHUIZEN,T. The complex precursor of coenzyme BIZ,cobinamide, GALITSKI, D.GUTTMAN, K HAACK,P. HIGGINS,E. KOFOID,L. MIESEI., is transported well (58% therate of vitamin B12uptake) H. OCHMAN,E. PRESLEY,and D. WALTERfor helpful and enlightening discussions, and R. JENSEN,E. JONES,and ananonymous reviewer for in E. coliK-12 (BRADBEERet al. 1978; KENLEY et al. 1978); critical readings of the manuscript. Thiswork was supported by grants cobinamide transport was curiously lacking, however, GM-15868 (J.G.L.) and GM-34804 (J.R.R.) from the National Insti- in E. coli B (TAYLORet al. 1972). Itis reasonable to posit tutes of Health. that the uptake of complex corrinoid precursors from the growth environment allows cobalamin synthesis LITERATURE CITED from this complex intermediate and provides selection for the observed maintenance of CobIII functions and ABELES,R. H., and H. A. LEE,JR., 1961 An intramolecular oxidation- reduction requiring a cobamide coenzyme. J. Biol. Chem. 236: cobalamindependentethanolamine degradation. In 2347-2350. contrast, the lack of CobII functions in strains of E. coli AILION, M., T. A. BOBIKand J. R. ROTH,1993 Two global regulatory maintaining CobIII functions is enigmatic. systems (Crp andArc) control the cobalamin/propanediolregu- Ion of Salmonella typhimurium. J. Bacteriol. 175: 7200-7208. From a naive standpoint,one may conclude that AKHY, M. T., C. M. BROWNand D. C. OLD,1984 LRhamnose utiliza- these strains employ the CobIII pathway to synthesize tion in Salmonella typhimurium. J. Appl. Bacteriol. 56: 269-274. cobalamin only when both exogenous cobinamide and ALBERT, M. J., V. I. MATHANand S.J. BAKER,1980 Vitamin BIZsynthe- sis by human small intestinal bacteria. Nature 283: 781-782. DMB are present. We find it unlikely that these condi- ANDERSSON,D. I., and J. R. ROTH,1989a Mutations affecting regula- tions arise with sufficient frequency in natural environ- tion of cobinamide synthesis in Salmonella typhimun'um.J. Bacte- ments to maintain CobIII gene functions in these rial. 171: 6726-6733. ANDERSON, D. I., and J. R. ROTH, 1989b Redox regulation of the strains. Rather, it seems more likely that E. coli employs genes for cobinamide biosynthesis in Salmonella typhimurium. J. alternative a-ligands to DMB in the synthesis ofcobam- Bacteriol. 171: 6734-6739. ides or uses exogenous non-DMB bearing cobamides BATTERSBY,A. R., 1994 How nature builds the pigments of life: the conquest of vitamin BIZ.Science 264: 1551-1557. directly. BELTRAN,P., S. A. PI.OCK,N. H. SMITH,T. S. WHITTAM,D. C. OLD Strains of E. coli with functional CobII and CobIII et al., 1991 Reference collection of strains of the Salmonella pathways, e.g., E. coli 113-3, will readily synthesize alter- typhimurium complex from natural populations. J. Gen. Micro- biol. 137: 601-606. nate cobamides when provided with cobinamide plus BOBIK,T. A,, M. AKION and J. R. ROTH, 1992 A single regulatory excess exogenous compounds such as benzimidazole, gene integrates control of vitamin BIZsynthesis and propanediol 5-methylbenzimidazole, aminobenzene, and adenine degradation. J. Bacteriol. 174 2253-2266. BOYD, E. F., F.S. WANG,P. BELTRAN,S. A. PLOCK,K. NEISON et al., (FORDet al. 1955). These compoundsresult in the for- 1993 Salmonella reference collection B (SARB): strains of 37 mation of cobamides that differ structurally from coen- serovars of subspecies I. J. Gen. Microbiol. 139: 1125-1132. zyme B12. Therefore, E. coli is able to synthesis non- BRADBEER,C., 1991 Cobalamin transport in Escherichia coli. Biofac- ton 3: 11-19. DMB-bearing cobamides under laboratory conditions. BRADBEER,C., J. S. KENIXY, D. R. DI MASI and M. LEIGHTON, 1978 Moreover, natural isolates of E. coli have been shown Transport of vitamin BI2in Eschm'chia coli. Corrinoid specificities Evolution of Cobalamin Synthesis 23

of the periplasmic BITbinding protein andof energydependent thetic genes of Salmonella typhimurium. J. Bacteriol. 169: 3189- BI2 transport.J. Biol. Chem. 253 1347-1352. 3198. CABALLERO,E., L.BALDOMA, J. Ros, A. BORONATand J. AGUILAR, JETER,R. M., B. M. OLNERAand J. R. ROTH,1984 Salmonella typhimu- 1983 Identification of lactaldehyde dehydrogenase and glycol- rium synthesizes cobalamin (vitamin BIZ)de novo under anaero- aldehydedehydrogenase as functions of the same protein in bic growth conditions. J. Bacteriol. 159: 206-213. Escherichia coli. J. Biol. Chem. 258 7788-7792. JOHNSON,M. G., andJ. C. ESCALANTE-SEMERENA,1992 Identification CHANG,G. W., and J. T. CWG, 1975 Evidence for the Bldepen- of 5,6-dimethyl-benzimidazoleas the Coa ligand of the cobamide dent enzyme ethanolamine deaminase in Salmonella. Nature 254. synthesized by Salmonella typhimunum.J. Biol. Chem. 267: 13302- 150-151. 13305. CHEN,P., D. I. ANDERSON and J. R. ROTH, 1994 The control region KENLEY, J. S., M. LEICHTONand C. BRADBEER,1978 Transport of of the cob/pdu regulon in Salmonella typhimunum. J. Bacteriol. vitamin BIZin Escherichia coli. Corrinoid specificity of the outer 176: 5474-5482. membrane receptor. J. Biol. Chem. 253: 1341-1346. CHEN,P., M. AILION,N. WEVLANDand J. R. ROTH, 1995 The end LAGO, B. D., and A. L.DEMAIN, 1969 Alternate requirementfor of the cob operon: evidence that the last gene (cob7) catalyzes vitamin BI2 or methionineof Pseudomonas denitnjicans, a vitamin the synthesis of the lower ligand of vitamin BI2,dimethylbenzimi- BITproducing bacterium. J. Bacteriol. 99 347-349. dazole. J. Bacteriol. 177: 1461-1469. LAWRENCE,J. G., and J. R. ROTH, 1995 The cobalamin (coenzyme CHILDS,J. D., and D.A. SMITH,1969 New methionine structural BI2)biosynthetic genes of Escherichia coli.J. Bacteriol. 177: 6371- gene in Salmonella typhimurium. J. Bacteriol. 100: 377-382. 6380. COCKS,G. T., J. AGUILARand E. C. C. LIN, 1974 Evolution of L LAWRENCE,J. G., H. OCHMANand D. L. HARTL,1991 Molecular 1,2 propanediol catabolism in Escherichia coli by recruitment of and evolutionary relationships among enteric bacteria. J. Gen. enzymes for Lfucose and Llactatemetabolism. J. Bacteriol. 118: Microbiol. 137: 1911-1921. 83-88. LIN, E.C. C., 1987 Dissimilatory pathways for sugars, polyols, and DAVIS,B. D., and E. MINGIOLI,1950 Mutants of Escherichia colirequir- carboxylates, pp. 244-284 in Escherichia coli and Salmonella typht ing methionine or vitamin BIZ.J. Bacteriol. 60: 17-28. murium: Cellular and MokcularBiology, edited by F. D. NIENHARDT, DAVIS,R. W., D. BOTSTEINand J. R. ROTH, 1980 A manual for ge- J. L. INGRAHAM,K. B. Low, B. MAGASANIK, M. SCHAECHTERand netic engineering: advanced bacterial genetics. Cold Spring Har- H. E. UMBARGER.American Society for Microbiology, Washing- bor Laboratory, Cold Spring Harbor, NY. ton, DC. EAGON, R. G., 1961 Bacterial dissimilation of L-fucose and L-rham- LINGENS,B., T. A. SCHILD,B. VOCLERand P. RENZ, 1992 Biosynthesis nose. J. Bacteriol. 82: 548-550. of vitamin BIZ: transformationof riboflavin ‘H-labeled in the 1’R ESCAIANTE-SEMERENA,J. C., and J. R. ROTH, 1987 Regulation of position or 1‘s position into 5,&dimethylbenzimidazole. Eur. J. cobalamin biosynthetic operons inSalmonella typhimurium. J. Bac- Biochem. 207: 981-985. teriol. 169: 2251-2258. MENON,I. A,, and D. SHEMIN,1967 Concurrent decrease of enzy- ESCAIANTE~EMERENA,C.,J. S.-J. SUH and J. R. ROTH, 1990 cobA matic activities concerned with the synthesis of coenzyme BIZ function is required for bothde novo cobalamin biosynthesis and and of propionic acid in Propionibacteria. Arch. Biochem. Biophys. assimilation of exogenous corrinoids in Salmonella typhimurium. 121: 304-310. J. Bacteriol. 172 273-280. NOGUCHI,S.,Y. NISHIMURA,~.HIROTA and S. NISHIMURA,1982 Isola- ESWTE-SEMERENA,J. C., M. G. JOHNSON and J. R. ROTH, 1992 tion and characterizationof an E. coli mutant lacking tRNA- The Cob11 and Cob111 regions of the cobalamin (vitamin BIZ) guanine transglycosylase. Function and biosynthesis of queuo- biosynthetic operon of Salmonella typhimurium. J. Bacteriol. 174: sine tRNA. J. Biol. Chem. 257: 6544-6550. 24-29. OBRADORS,N., J. BAD^, L. BALDOMAand J. AGUILAR,1988 Anaero- ESCHENMOSER,A., 1988 Vitamin BIZ:experiments concerning the bic metabolism of the L-rhamnose fermentation product 1,2- origin of its molecular structure. Angew. Chem. Int. Ed. Engl. propanediol in Salmonella typhimurium. J. Bacteriol. 170: 2159- 27: 5-39. 2162. FEINBERG,A. P., and B. VOGELSTEIN,1983 A technique for radiola- OCHMAN,H., and R. IC SELANDER,1984 A reference collection of beling DNA restriction endonuclease fragments to high specific natural isolates of Escherichia coli. J. Bacteriol. 157: 517-524. activity. Anal. Biochem. 132: 6-13. OHSUGI,M., H. NODA,K. MURO,A. ISHIBA,Y. KONDOet al., 1989 FORAGE,R. G., and M. A. FOSTER,1979 Resolution of the coenzyme- Effects of yeast extract components pyrroloquinoline quinone B+lependent dehydratases of Klebsiella sp. and Citrobacterfreun- and aspartic acid on vitamin BIZproduction in Klebsiella pneume dii. Biochim. et Biophys. Acta 569 249-258. niaeIFO 13541. J. Nutr. Sci. Vitaminol. 35: 661-665. FORAGE,R. G., and M.A. FOSTER,1982 Glycerol fermentationin OLD,D. C., and R. P. MORTLOCK,1977 The metabolism of D-arabi- Klebsiella pneumonia: functions of the coenzyme Blpdependent nose by Salmonella typhimunum.J. Gen. Microbiol. 101: 341-344. glycerol and diol dehydratases. J. Bacteriol. 149: 413-419. O’TOOLE,G. A,, M. R. RONDONand J. C. ESCALANTE-SEMERENA,1993 FORD,J. E., E. S. HOLDSWORTHand S. K. KON,1955 The biosynthesis Analysis of mutants of Salmonella typhimurium defective in the of vitamin BI2-likecompounds. Biochem. J. 59: 86-93. synthesis of the nucleotide loop of cobalamin. J. Bacteriol. 175: FREY,B., J. MCCLOSKEY,W. KERSTEN and H. KERSTEN, 1988 New 3317-3326. function of vitamin BIZ:cobamidedependent reduction of ep RAMBACH,A. 1990 New plate medium for facilitated differentiation oxyqueosine in tRNAs of Escherichia coli and Salmonella typhimu- of Salmonella spp. from Proteus spp. and other enteric bacteria. rium. J. Bacteriol. 170: 2078-2082. Appl. Env. Microbiol. 56: 301-303. FRIEDRICH,W., 1975 Fermente, hormone, vitamine, band 111/2, pp. moux, C. R., and R. J. KADNER, 1989 Two outer membrane trans- 181 - 186 in Vitamin und vmndte corrinoide, edited byR. A” port systems for vitamin BIZ inSalmonella tyfhimurium. J. Bacte- MON and W. DIRSCHERL.Verlag, Stuttgart. riol. 171: 2986-2993. GEORGOPAPADAKOU,N. H., and A. I. SCO~,1977 On BIZbiosynthe- R~OUX,C. R., M. J. FRIEDRICHand R. J. KADNER,1990 Genes on the sis and evolution. J. Theor. Biol. 69: 381-384 90-kilobase plasmid of Salmonella typhimurium confer low-affinity GRABAU,C., and J. R. ROTH, 1992 A Salmonella typhimunum cobala- cobalamin transport: relationship to fimbria biosynthesis genes. mindeficient mutant blocked in 1-amino-2-propanol synthesis. J. Bacteriol. 172: 6217-6222. J. Bacteriol. 174: 2138-2144. RODERICK,S. L., and B. W. MATTHEWS,1993 Structure of the cobalt- HACKING,A. J.,J. AGUILARand E. C. C. LIN, 1978 Evolution of dependent methionine aminopeptidase from Escherichia coli: a propanediol utilization in Escherichia coli. Mutants with improved new type of proteolytic enzyme. Biochemistry 32 3907-3912. substrate-scavenging power. J. Bacteriol. 136: 522-530. RONDON,M. R., and J. C. ESCALANTE-SEMERENA,1992 The poc locus HORIG, J. A., and P. RENZ, 1978 [5-15N]Riboflavinas precursor in is required for 1,2-propanedioldependenttranscription of the the biosynthesis of the 5,6dimethylbenzimidazolemoiety of vita- cobalamin biosynthetic (cob) and propanediol utilization (pdu) min BIZ.J. Biol. Chem. 253: 7410-7414. genes of Salmonella typhimurium. J. Bacteriol. 174: 2267-2272. JETER, R. M., 1990 Cobalamin dependent 1,2-propanediol utiliza- ROOF, D. M., and J. R. ROTH, 1988 Ethanolamine utilization in tion by Salmonella typhimurium. J. Gen. Microbiol. 136: 887-896. Salmonella typhimunum. J. Bacteriol. 170: 3855-3863. JETER,R. M., and J. R. ROTH, 1987 Cobalamin (vitamin BI2)biosyn- ROOF, D. M., and J. R. ROTH,1989 Functions required for vitamin 24 J. G. Lawrence and J. R. Roth

Bidependent ethanolamine utilization in Salmonella typhimu- SMITH,D. A,, and J. D. CHILDS, 1966Methionine genes and enzymes rium. J. Bacteriol. 171: 3316-3323. of Salmonella typhimurium. Heredity 21: 265-86. ROTH,J. R.,J. G. LAWRENCE, M. RUBENFIELD,S. KIEFFER-HIGGINS and SRIDHARA,S., T. T. Wu, T. M. CHUSEDand E. C. C. LIN, 1969 Fer- G. M. CHURCH,1993 Characterization of the cobalamin (vita- rous-activated nicotinamide adenine dinucleotide-linked dehy- min Biz) biosynthetic genes of Salmonella typhimurium. J. Bacte- drogenase from a mutant of Escherichia coli capable of growth on ri01. 175: 3303-3316. 1,2-propanediol.J. Bacteriol. 98: 87-95. RUCH,F. E.,J. LENCELER and E. C. C. LIN, 1974 Regulation of glycerol STRO~N~KI, A,,1987 Adenosylcobamidedependent reactions, pp. catabolism in Klebsiella nerogenes. J. Bacteriol. 119 50-56. 226-259 in Comprehensive B,2, edited by Z. SCHNEIDER andA. SAIKI, R. K., S. SC-, F. FALOONA,K. B. MULLIS,G. T. HORNet al., STROINSKI.de Gruyter, Berlin. 1985Enzymatic amplification of beta-globin genomic se- TAYLOR,R. T., M. P.NEVINS and M.L. HANNA,1972 Uptake of quences and restriction site analysis for diagnosis of sickle cell cyanocobalamin by Escherichia coli B: corrinoid specificity and the anemia. Science 230: 1350-1354. relationship of a binder. Arch. Biochem. Biophys.149: 232-243. SAIKI,R. IC, D. H. GELFAND,S. STOFFEL,S. J. SCHARF,R. HIGUCHIet TORAYA,T., and S. FUKUI,1977 Immunochemical evidence for the al., 1988 Primerdirected enzymatic amplification of DNA with difference between coenzyme-B1,dependent diol dehydratase a thermostable DNA polymerase. Science 239: 487-491. and glycerol dehydratase. Eur. J. Biochem. 76: 285-289. SAMBROOK,J., E. F. FRITSCH and T. MANIATIS,1989 Molecular cloning. TORAYA,T., S. HONDA,S. KUNO and S. FUKUI,1978 Coenzyme B12- A laboratory manual. ColdSpring Harbor Press, Plainview, NY. dependent diol dehydratase: regulation of apoenzyme synthesis SCARLETT,F. A,, and J. M. TURNER,1976 Microbial metabolism of in Klebsiella pneumonia (Aerobacter aerogenes) ATCC 8724.J. Bacte- amino alcohols. Ethanolamine catabolism mediated by coen- riol. 135: 7216-729. zyme B,dependent ethanolamine ammonia-lyase in Escherichia TORAYA,T., S. HONDAand S. FUKUI,1979 Fermentation of 1,2- coli and Klebsiella amogenes. J. Gen. Microbiol. 95: 173-176. propanediol and 1,Zethanediol by some genera of Enterobac- SCHNEIDER,Z., and A. STRO~NSKI,1987 Methylcobamide-dependent teriaceae, involving coenzyme B12dependentdiol dehydratase. reactions, pp. 259-266 in Comprehensive B,2. edited by Z. SCHNEI- J. Bacteriol. 139: 39-47. DER and A. STROfNSKI. de Gruyter, Berlin. TORAYA,T., S. KUNO and S. FUKUI,1980 Distribution of coenzyme- SCHNEIDER,Z., E. G. LARSEN, G. JACOBSON,B.C. JOHNSON and J. B12-dependentdiol dehydratase and glycerol dehydratase in se- PAWELKIEWICZ,1970 Purification and properties of glycerol de- lected genera of Enterobacteriaceae and Propionibacteriaceae. hydratase. J. Biol. Chem. 245: 3388-3396. J. Bacteriol. 141: 1439-1442. SCOTT,A. I., 1990 Mechanistic and evolutionary aspects of vitamin VOGEL, H.J., and D. M. BONNER,1956 Acetylornithase ofEscherichia B,, biosynthesis. ACC.Chem. Res. 23: 308-317. coli: partial purification and some properties. J. Biol. Chem. 218: SCOTT,A. I., 1993 How nature synthesizes vitamin BI2-a survey of 97-106. the last four billion years. Angew.Chem. Int. Ed. Engl.32: 1223- VOLCANI,B. E., J. I. TOOHEYand H. A. BARKER,1961 Detection 1243. of cobamide coenzymes in microorganisms by the ionophoretic SELANDER,R. K., D. A. CAUCANT and T. S. WHITTAM,1987 Genetic bioautographic method. Arch. Biochem. Biophys. 92: 381-391. structure and variation in natural populationsof Escherichia coli, Xu, K., J. DELLINCand T. ELLIOT,1992 The genes required for pp. 1625-1648 in Escherichia coli and Salmonella typhimurium: heme synthesis inSalmonella typhimurium include those encoding Cellular and Molecular Biology, edited by F. D. NIENHARDT,J. L. alternate functions for aerobic and anaerobic coporphyrinogen INGRAHAM,K. B. Low, B. MAGASANIK,M. SCHAECHTERand H. E. oxidation. J. Bacteriol. 174 3953-3963. UMBARGER.American Society for Microbiology, Washington, DC. Communicating editor: E. JONES