The enigmatic acyl carrier protein phosphodiesterase of Escherichia coli: genetic and enzymological characterization.

The acyl carrier proteins (ACPs) of fatty acid synthesis are functional only when modified by attachment of the prosthetic group, 4'-phosphopantetheine (4'-PP), which is transferred from CoA to the hydroxyl group of a specific serine residue. Almost 40 years ago Vagelos and Larrabee reported an enzyme from Escherichia coli that removed the prosthetic group. We report that this enzyme, called ACP hydrolyase or ACP phosphodiesterase, is encoded by a gene (yajB) of previously unknown function that we have renamed acpH. A mutant E. coli strain having a total deletion of the acpH gene has been constructed that grows normally, showing that phosphodiesterase activity is not essential for growth, although it is required for turnover of the ACP prosthetic group in vivo. ACP phosphodiesterase (AcpH) has been purified to homogeneity for the first time and is a soluble protein that very readily aggregates upon overexpression in vivo or concentration in vitro. The purified enzyme has been shown to cleave acyl-ACP species with acyl chains of 6-16 carbon atoms and is active on some, but not all, non-native ACP species tested. Possible physiological roles for AcpH are discussed.

Acyl carrier proteins (ACPs) 2 are small (Ͻ80 residues), highly acidic proteins that are modified by covalent attachment of 4Ј-phosphopantetheine (4Ј-PP) that is transferred from CoA and attached via a phosphodiester linkage to the hydroxyl group of a specific serine residue located in the midst of the protein sequence (1). The paradigm ACPs are those of bacterial fatty acid synthesis where the sulfhydryl group of the 4Ј-PP moiety carries the growing fatty acid chain (1). In addition to bacteria, ACPs are found in plant plastids (2), mitochondria (3), and the apicoplasts of apicomplexan parasites (4). The 4Ј-PP moiety is attached to the apo form of ACP by enzymes called 4Ј-phosphopantetheinyl transferases (5). Escherichia coli AcpS (ACP synthase) was the first such enzyme described (6) and is the only 4Ј-PP transferase studied physiologically (7,8).
Proteins similar to the ACPs of fatty acid synthesis are found in polyketide and nonribosomal polypeptide synthesis where these proteins (called PKS ACPs and PCPs, respectively) perform analogous carrier functions (5). Indeed, these ACP-like proteins have the same general four-helix bundle structure predicted (9) and subsequently demonstrated (10 -12) for E. coli ACP. Soon after the identification of the prosthetic group of ACP as 4Ј-PP and of 4Ј-PP transferase activity, Vagelos and Larrabee (13) reported an activity that removed the 4Ј-PP moiety of ACP in E. coli. The enzyme (called ACP hydrolyase and now called ACP phosphodiesterase) was partially purified and shown to require a divalent ion and produce apo-ACP and 4Ј-PP (13). The enzyme was inactive on peptide fragments of ACP suggesting that it recognized the folded structure of ACP (13). Although crude preparations were occasionally used to convert holo-ACP to apo-ACP, this paper remained the only study of this enzyme for 23 years until Fischl and Kennedy (14) reported its purification to apparent homogeneity. These workers showed that ACP phosphodiesterase was an unusually stable protein of M r 25,000 and reported an N-terminal sequence attributed to the enzyme protein. Unfortunately, the N-terminal sequence was found to be that of a flavin-containing protein that lacked phosphodiesterase activity 3,4 and which has recently been shown to be an azoreductase (15). It seems that Fischl and Kennedy (14) had determined the sequence of a major contaminating protein rather than that of the phosphodiesterase. Regrettably based on this erroneous sequence attribution, many genes (often called acpD) have been annotated in bacterial genomes as encoding ACP phosphodiesterase rather than azoreductase.
The discovery of ACP phosphodiesterase activity logically raised the question of whether or not the prosthetic group of ACP turns over independently of the protein moiety. This was soon shown to be the case (16), although interpretation of these data were complicated by the large pools of CoA and its thioesters. Elegant deuterium labeling experiments subsequently showed that the rate of ACP prosthetic group turnover depended on the intracellular concentration of CoA (17). At low CoA levels, prosthetic group turnover was four times faster than the rate of new synthesis of the protein moiety, but at the higher coenzyme A concentrations characteristic of logarithmic growth, turnover was an order of magnitude slower, amounting to ϳ25% of the ACP pool per generation (17).
We report the isolation of the gene encoding the ACP phosphodiesterase of E. coli. We have named the gene acpH based on the original enzyme name given by Vagelos and Larrabee (13) to prevent overlap with the mistaken AcpD annotations. The protein has been purified to homogeneity and its substrate specificity studied. The acpH gene has also been deleted from the E. coli chromosome and the consequences of loss of the enzyme studied.  (TABLE ONE). Strain SJ16 and its derivatives were grown on minimal E medium (18) with 0.4% glucose or glycerol, as carbon source, 0.001% thiamine, and 0.1% casamino acids (Difco). When added, ampicillin was used at 100 g/ml, chloramphenicol at 25 g/ml, spectinomycin at 50 g/ml, and kanamycin at 30 g/ml.

Materials-Restriction
Construction and Screening of the Cosmid Library-An expression library was constructed essentially according to the method described previously (19). Strain MG1655 DNA purified using the Wizard genomic DNA purification kit (Promega) was sheared to ϳ30 -50-kb fragments by repeated pipetting. Sheared DNA was then end repaired to generate blunt ends using T4 DNA polymerase and run on a 1% low melting point agarose gel. Fragments ϳ40 kb in size were excised, DNA was recovered from the gel and ligated to Eco72I-digested pCC1FOS.
The ligation mixture was size-selected by packaging into phage particles and used to infect EPI300 to chloramphenicol resistance. Transformants were restreaked and separated into sets of 50 based on colony size. Individual cultures of each set of colonies were grown overnight in LB/chloramphenicol, pooled together, and diluted to A 600 of 0.2 in 100 ml of the same medium. Each pooled culture was grown to A 600 of 0.6 and induced with 0.02% arabinose for a further 4 h. The cells were washed with an equal volume of AcpH storage buffer (50 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 , 2 mM MnCl 2 , 5 mM DTT, 10% glycerol, and 0.2 mM phenylmethylsulfonyl fluoride), resuspended in 1 ml of the same buffer, and lysed by sonication. Cell-free extracts were assayed for AcpH activity as described above. The pool whose extract showed a level of AcpH activity above that of the vector control was further subdivided into 10 sets of five colonies each and pooled cultures were grown, induced, and lysed as before. In the final screen of the library extracts were made and assayed from cultures grown from single colonies. Cosmid DNA was isolated from an induced culture of the clone (pJT1) expressing AcpH (Qiagen Large Construct kit) and used to transform strain EPI300. Upon confirmation of increased activity, the encoding gene was subcloned from cosmid pJT1 as follows. The cosmid DNA was diluted to 20 g/ml in 10 mM Tris-HCl containing 1 mM EDTA (pH 8.0) buffer and sonicated for 4 s at the lowest power setting. The sheared DNA was end-repaired as above and fractionated on a 1% agarose gel. Fragments of 2-3 kb were extracted from the gel and ligated to Eco721digested pCC1FOS. The ligation mixture was used to transform EPI300, and 200 clones were screened as before. Cosmid DNA was purified from clones that showed increased AcpH activity and sequenced with primers CC1Fos-F and CC1Fos-R. DNA Manipulations-The sequences of the primers used are given in TABLE ONE. The acpH gene was PCR amplified from pJT1 using primers yajB-pET101-F and yajB-pET101-R. The resulting PCR product was cloned into the TOPO vector pET101D to give plasmid pJT4 in which a C-terminal His-tagged AcpH was expressed from a T7 promoter. The acpH sequence was also PCR-amplified using primers yajB-pBAD-F and yajB-pBAD-R and the KpnI-and PstI-digested PCR products were cloned into pBAD322 digested with the same enzymes to give plasmid pJT5. Strain JT1 was constructed using the Red-mediated recombination system (20). The chloramphenicol acetyltransferase (cat) gene from plasmid pKD3 was amplified using primers yajB-KCm-F and yajB-KCm-R and electroporated into strain MC1061 carrying the helper plasmid pKD46 resulting in replacement of the entire coding sequence of acpH with the cat gene. Chloramphenicol-resistant colonies were verified by colony-PCR using primers yajb-chk-F and yajBchk-R. Strain JT2 was constructed by using a P1 lysate made on JT1 to transduce SJ16 to chloramphenicol resistance. The acpH Ϫ genotype was verified by assaying crude extracts of both strains for AcpH activity as described below.
Purification of ACP Species-Holo-ACP was purified using a modified procedure based on the methods described previously (21,22). Strain DK574 (which is SJ16 carrying plasmids pMS421 and pMR19) was grown to an A 600 of 0.8 and induced with 15 M isopropyl ␤-Dthiogalactopyranoside for a further 4 h. The cells were pelleted, washed in an equal volume of 50 mM sodium MES, pH 6.1, and concentrated 20-fold in the same buffer. Crude extracts were prepared by sonication and an equal volume of ice-cold isopropyl alcohol was added to the extract and incubated with stirring at 4°C for 1 h. The suspension was centrifuged, and the supernatant was applied to a Vivaspin D column equilibrated to pH 6.1. The column was washed with 25 mM sodium MES, pH 6.1, containing 0.25 M LiCl, and holo-ACP was eluted with 0.5 M LiCl. The protein was concentrated a further 50-fold by precipitation with 0.02% sodium deoxycholate and 5% trichloroacetic acid followed by resuspension in 0.5 M Tris-HCl, pH 8.0, and dialysis against 10 mM Tris-HCl, pH 8.0, containing 1 mM DTT overnight. Holo-ACP labeled in the 4Ј-PP group was purified in the same manner with the following exceptions; strain DK574 was grown overnight in medium containing 0.5 M ␤-alanine (a condition that results in growth-limiting coenzyme A levels) followed by subculture into medium containing 8 M ␤-[3-3 H]alanine (0.125 Ci/mol) and growth for a further 5 h. [ 3 H]ACP was purified from these cultures as described above. The apo form of ACP was obtained as a mixture of the apo and holo forms as above except that strain DK574 was induced with 50 M isopropyl-␤-D-thiogalactopyranoside, which resulted in a greater fraction of apo-ACP (21).
AcpS Assay-AcpS was used to phosphopantetheinylate apo-ACP using the assay conditions described previously (23). The reactions contained 1 mM CoA, 10 -40 M apo-ACP, 10 mM MgCl 2 , 50 mM Tris-HCl, pH 8.8, 1 mM DTT, and 1 M AcpS in a volume of 0.1 ml and were incubated at 37°C for 4 h. The conversion to holo-ACP was determined by conformationally sensitive gel electrophoresis (24) on 20% non-denaturing polyacrylamide gels containing 0.5 M urea (25) and visualized by staining with Coomassie Blue.
Acyl-ACP Synthesis-Acyl-ACPs were synthesized according to the method of Shen et al. (26). Reactions contained 100 mM Tris-HCl, pH 7.8, 10 mM MgCl 2 , 10 mM ATP, 1 mM DTT, 50 M holo-ACP, 80 M concentrations of each fatty acid, and 50 nM purified Vibrio harveyi acyl-ACP synthetase 5 in a 50-l volume and were incubated at 37°C for 4 h. Essentially complete conversion to the acylated forms was shown by gel electrophoresis.
In Vitro Assay of AcpH Activity-AcpH activity was assayed essentially by the method of Fischl and Kennedy (14). The assay was performed at 25°C and contained 50 mM Tris-HCl, pH 8.6, 0.02 mM MnCl 2 , 25 mM MgCl 2 , 1 mM DTT, and 10 -50 M acyl carrier protein substrate in a final volume of 50 l. Reactions were initiated by the addition of 20 g of crude extract protein or 1-200 nM purified AcpH and terminated after 60 min by the addition of trichloroacetic acid to 6%. The precipitate was washed in 6% trichloroacetic acid and resuspended in 50 l of 0.5 M Tris-HCl, pH 6.8. 10 l of this was analyzed by conformationally sensitive gel electrophoresis with detection by staining with Coomassie Brilliant Blue R-250. When analyzed by mass spectrometry, the reaction mixtures were first dialyzed against 2 mM ammonium acetate with two buffer changes and then dried under vacuum. Radioactive AcpH assays were performed under the same conditions except that 15-40 M [ 3 H]holo-ACP or 3-15 g of protein from an extract of ␤-alanine-labeled cells was used as substrate. The reactions were either analyzed by fluorography as described below or by release of the 3 H label from the protein quantitated as follows: 30 l of the trichloroacetic acid supernatant was treated with 200 l of water-saturated diethyl ether to remove the acid and 20 l of the aqueous phase was then mixed with 5 ml of scintillation fluid and counted in a Beckman LS6500 scintillation counter. For fluorography following electrophoresis, the acrylamide gels were fixed in 50% methanol and 10% acetic acid for 30 min then incubated in Amplify solution (Amersham Biosciences) for a further 30 min and dried at 40°C. The dried gels were exposed to preflashed X-AR Bio-Max films (Kodak) at Ϫ80°C.
Holo-ACP Turnover-Strain SJ16 derivatives were grown in minimal medium containing 0.5 M ␤-alanine overnight to reduce the coenzyme A pools. Overnight cultures were subcultured in minimal medium with 0.5 M [ 3 H]␤-alanine (1 Ci/mol) until they reached an A 600 of 1.0. Cells were pelleted and washed twice in minimal medium and then diluted 5-fold in the same medium containing 8 M unlabeled ␤-alanine. 0.5-ml samples were withdrawn at various time points, and cell extracts were made in 25 mM sodium MES, pH 6.1. Ten l of each extract was run on 20% polyacrylamide gels and analyzed by fluorography as given above.
Expression and Purification of C-terminal Hexahistidine-tagged AcpH-Strain BL21 DE3/pJT4 was grown in LB-ampicillin to A 600 0.8 and induced with 1 mM isopropyl ␤-D-thiogalactopyranoside for a further 4 h. The cells were washed in an equal volume of lysis buffer (100 mM NaH 2 PO 4 , 10 mM Tris-HCl, pH 8.0) and resuspended in one-tenth volume of the same buffer containing 8 M urea. The cell suspension was lysed by shaking at room temperature for 1 h and then centrifuged. AcpH was purified from the supernatant on Ni-NTA-agarose under denaturing conditions as follows. The supernatant was applied to a column of Ni-NTA-agarose resin, and the column was washed and eluted with buffers of the same composition but decreasing pH: wash buffer 1, pH 6.3; wash buffer 2, pH 5.9; and elution buffer, pH 4.5. Fractions were analyzed on a 12% SDS-polyacrylamide gel, purified AcpH was diluted to a concentration of 10 g/ml and dialyzed against AcpH storage buffer mentioned above with the addition of 8 M urea and 50 mM each of L-arginine and L-glutamate to prevent aggregation (27). The molar con-5 Y. Jiang and J. E. Cronan, manuscript in preparation. centration of urea in the arginine-glutamate buffer was decreased stepwise in successive dialysis steps (8, 4, 2, 1, 0.5, and 0) for 2 h each followed by overnight dialysis (with three changes of buffer) against the final buffer.
Gel Filtration Chromatography-A 5-ml culture of BL21 DE3 transformed with plasmid pJT4 was grown to mid-log phase, treated with 200 g/ml rifampicin and 42.5 nM L-[ 35 S]methionine (250 Ci) (28). The culture was grown for a further 1 h, concentrated 10-fold, and crude extracts were made in column running buffer (50 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 , 0.15 M NaCl, 5 mM DTT, 50 mM L-glutamate, and 50 mM L-arginine). The extract was dialyzed against the same buffer with two changes to remove soluble label and 0.1 ml of the dialyzed extract was applied to a Superdex 200 column (1 ϫ 44 cm). 0.5-ml fractions were collected at a flow rate of 65 l/min and 0.1 ml of each fraction was mixed with 5 ml of scintillation fluid and analyzed for radioactivity in a Beckman LS6500 scintillation counter. The standard proteins were run in the same manner except that 20 g of each purified protein was applied to the column and the fractions were analyzed by the Bradford assay (29).

RESULTS
Expression Cloning of ACP Phosphodiesterase-Because prior attempts at identification of the gene encoding ACP phosphodiesterase by N-terminal sequencing of the enzyme purified Ͼ3000-fold from E. coli cell extracts (14) gave the sequence of an unrelated contaminating protein (AzoR) (15), it seemed that the phosphodiesterase is a very non-abundant protein. We therefore took a different approach, that of screening a clone bank for plasmid clones that gave increased phosphodiesterase activity in vitro. We had previously used this approach to identify the gene encoding Enterococcus faecalis lipoamidase (19). As in that study toxicity of the expressed gene seemed likely to be a problem, as high levels of phosphodiesterase activity would be expected to inhibit growth by depletion of holo-ACP pools required for fatty acid synthesis and by increasing the relative amount of apo-ACP, a potent inhibitor of cell growth (21). Therefore, we used the same vector system (30) as in the lipoamidase work that has the following features. The vector is a cosmid that carries an F-factor origin of replication (ori2) and partitioning function that maintains the cosmid clones in single copy under uninduced conditions. This alleviates the problem of under-representation of clones carrying genes that are toxic in multiple copies. A second origin of replication, oriV, is also present on the cosmid that requires the TrfA protein for function. The host strain carries a copy-up allele of the trfA gene (30) integrated into the chromosome under control of a pBAD promoter. Induction with arabinose causes expression of TrfA allowing initiation of replication from the oriV origin to give a 40 -80-fold increase in the cosmid copy number. This inducible increase in copy number is useful both to increase yields during isolation of cosmid DNA and more importantly, to increase expression of genes carried on the insert for easier expression-based cloning. In addition, the cosmid contains cos sites that allow it to be packaged into particles in vitro or in vivo. The dependence of in vitro packaging on the size of the packaging substrate gives a size selection ensuring that the library comprises only clones containing inserts of ϳ40 kb.
Sheared E. coli MG1655 DNA was end repaired and ligated into Eco72I-digested pCC1FOS. The ligation mixture was then packaged into phage particles and used to infect strain EPI300 giving 3 ϫ 10 4 chloramphenicol-resistant colonies per ml of packaged ligation mixture. The colonies were grown for 48 h at 37°C to allow slower growing colonies to reach visible size. We picked a library of 500 colonies because to obtain a given E. coli K12 genome sequence with 99% prob-ability, 461 clones of a 40-kb insert genomic library are required (31). The library colonies were divided into 10 sets of 50 colonies each: four sets of small colonies, three sets each of medium and large sized colonies. Individual cultures of each set of colonies were grown, pooled, induced, and extracts were made from the collected cells, and the extracts were assayed for AcpH activity by conformationally sensitive gel electrophoretic assay. We expected that a clone encoding ACP phosphodiesterase would give a 40 -80-fold increase in activity in the cells that carried that clone. Although this increased activity would be diluted by the other clones of the pool (which were unlikely to carry the gene), we expected that, given the size of our pools, we should be able to observe the presence of a single overproducing clone. The protein concentrations of the extracts of the pools were determined, and the extracts were diluted such that in the absence of overproduction, ACP phosphodiesterase activity could barely be detected in a wild type extract. (Extracts of EPI300 transformed with the vector pCC1FOS were made for each test and used as a negative control). We also thought it possible that library clones that carried the acpP gene might produce high levels of apo-ACP because of titration of the acyl carrier protein synthase activity (21) and thereby give false positives. Therefore, the extracts were also incubated without addition of the holo-ACP substrate such that the cellular ACP could be observed. In the first round of screening a pool made from small colonies showed a significant increase in AcpH activity (Fig. 1A). This pool was then subdivided into 10 subpools of five colonies each. One subpool showed increased activity and when individual colonies were analyzed clone SA1 (Fig. 1B) was identified as a cosmid that gave overexpression of ACP phosphodiesterase activity.
Cosmid DNA was isolated from clone SA1, transformed into EPI300, and gave increased phosphodiesterase activity. To locate the gene further, the cosmid (named pJT1) DNA was randomly sheared by sonication into 2-3-kb fragments that were then ligated into pCC1FOS. Two hundred subclones were streaked from the resulting transformants and divided into sets of 10, the pooled cultures were induced and crude extracts were screened as before. After two rounds of screening, a single clone JT1S25 was identified as carrying a small cosmid (designated pJT2) that gave ACP phosphodiesterase expression (Fig. 1C).
yajB Encodes ACP Phosphodiesterase-The pJT2 plasmid was sequenced using the primers CC1Fos-F and CC1Fos-R. BLAST searches with the resulting sequence data showed that the insert  5 and 6). A mixture of apo-and holo-ACP was run as a marker (Stds). As negative controls, assays containing only holo-ACP were performed (Holo-ACP) spanned a region from nucleotide 422807 (within the malZ gene) to 425159 (within the queA gene). The insert contained 749 bp of malZ, 924 bp of queA, and the complete yajB open reading frame (582 bp) that is currently annotated in GenBank TM as encoding a putative glycoprotein of uncharacterized function. The calculated molecular weight of 22,961 for the encoded protein agreed well with the findings of Fischl and Kennedy (14) who recovered phosphodiesterase activity from the 25,000 region of SDS gels (14). The original name given to ACP phosphodiesterase was ACP hydrolyase (13) and thus we have renamed yajB as acpH to acknowledge this precedent. Protein-protein BLAST searches in GenBank revealed that AcpH has homologues only in Gram-negative bacteria and mainly among the ␥-proteobacteria (Fig.  2), none of which has a biochemically characterized function.
Strain JT1, in which the entire coding sequence for acpH was replaced by the cat gene, was constructed by homologous recombination as described under "Experimental Procedures. " This strain demonstrated no noticeable growth phenotype when grown in rich or minimal media. Crude extracts of JT1 were prepared from log phase cultures and assayed for phosphodiesterase activity that was undetectable both by the gel shift (Fig. 3) and [ 3 H]4Ј-PP release (Ͻ1% of the activity of the wild type extract) assays. Strain MC1061 was also transformed with plasmid pJT5 carrying acpH under pBAD control. Crude extracts made from an arabinose-induced culture of this strain were assayed for AcpH activity (Fig. 3). As expected, this strain exhibited about 2-fold greater levels of AcpH activity than the vector control strain (higher levels of expression were precluded by aggregation of the protein).
Expression and Purification of AcpH-The acpH gene was PCR amplified from pJT2 and cloned into the TOPO expression vector pET101D (Invitrogen) to encode a protein that had a C-terminal hexahistidine tag and was expressed under control of a T7 promoter. This plasmid (called pJT4) was then transformed into strain BL21(DE3). Although crude extracts of the transformants showed overexpressed AcpH activity both under induced and non-induced conditions (data not shown), the bulk of the protein was insoluble. Fischl and Kennedy (14) had a similar aggregation problem during their protein purification. Altering isopropyl ␤-D-thiogalactopyranoside concentrations, growth temperatures, or use of N-terminal fusion tags (thioredoxin and maltose binding protein) previously reported to increase solubility of expressed proteins (32,33) failed to alleviate the aggregation problem.
The hexahistidine-tagged AcpH was readily purified from inclusion bodies under denaturing conditions (Fig. 4) but the protein aggregated during refolding from urea. However, when 50 mM L-arginine and 50 mM L-glutamate were included in the refolding buffers as suggested by the work of Golovanov et al. (27), AcpH remained soluble, but only at low protein concentrations (maximum of 10 g/ml). Attempts to concentrate the protein further led to aggregation, but this concentration was sufficient for enzyme assays. We performed gel exclusion chromatography to determine the native form of AcpH in solution. The active protein refolded from the inclusion bodies was eluted in the void volumes of both Superose 6 and Superdex 200 columns suggesting that the enzyme, although active, remains in an aggregated form.
Because the purified protein remained soluble only at low concentrations, we examined dialyzed crude extracts in which AcpH expressed from a T7 promoter was specifically labeled with [ 35 S]L-methionine in the presence of rifampicin (28,34). The major labeled peaks in the included volume of the column eluted in essentially the same fractions as did chymotrypsinogen and bovine serum albumin (Fig. 5), suggesting that monomeric and trimeric species are present. A major fraction of the 35 S label eluted in the void volume of the column, presumably because of aggregation of AcpH during the dialysis needed to remove unincorporated [ 35 S]methionine. We attempted to decrease aggregation by lowing expression levels, but the void volume material appeared even when only the basal level (uninduced) of expression was used for specific labeling of AcpH (Fig. 5). Moreover, use of the glutamate/arginine mixture during both dialysis and chromatography failed to prevent the appearance of labeled protein in the void volume.
The Protein Product of AcpH Is Apo-ACP-To verify that AcpH encodes a phosphodiesterase and not, for example, a protease giving products that have an altered mobility on non-denaturing gels, purified holo-ACP (Fig. 6A) was treated with purified AcpH and the products were analyzed by electrospray ionization mass spectrometry (Fig. 6B). The observed mass of the reaction product, 8508.4 Da, was in excellent agreement with the calculated mass of apo-ACP (8508.33 Da). The reaction product was also a substrate for phosphopantetheinyl transfer from CoA catalyzed by AcpS (Fig. 6C), thereby confirming the product to be apo-ACP. Prior workers have demonstrated the other product to be 4Ј-PP (13).
AcpH Is Responsible for ACP Prosthetic Group Turnover in Vivo-We then sought to clarify the role of AcpH in ACP prosthetic group turnover in vivo by performing turnover experiments with derivatives of strain SJ16, which carries a panD2 lesion and requires externally supplied ␤-alanine for growth on minimal medium (35), allowing radiolabeling of the 4Ј-phosphopantetheine group of CoA and holo-ACP pools by growth on [ 3 H]␤-alanine. Strains SJ16 and JT2 (SJ16 acpH::cat) were depleted of CoA by starvation for ␤-alanine (see "Experimental Procedures") then subcultured in minimal medium containing 0.5 M [ 3 H]␤alanine (1 Ci/mol) to early stationary phase. The cells were then collected, washed, and shifted to medium containing unlabeled ␤-alanine (8 M). Samples (0.5 ml) of each culture were withdrawn at regular time intervals. Crude extracts were then made and analyzed by gel electrophoresis followed by fluorography. In agreement with a previous study (17), turnover in strain SJ16 was most rapid within the first 10 min of recovery from ␤-alanine starvation (Fig. 7A). After this time however, the amount of label failed to decrease, probably because the CoA pool had not been sufficiently depleted for efficient chase with unlabeled ␤-alanine. In contrast, the amount of 3 H label bound to ACP in strain JT2 remained relatively constant (Fig. 7B) over the time course of the experiment. These results indicate that deletion of acpH abolishes ACP prosthetic group turnover in vivo. Jackowski and Rock (35,36) have reported that the 4Ј-PP from ACP prosthetic group turnover is excreted to the culture medium and cannot be recovered by the cells. We therefore followed the excretion of ␤-[3-3 H]alanine-labeled metabolites during the chase period (Fig. 7C) and found, as expected, that labeled metabolite excretion was significantly decreased in the mutant strain. Vallari and Jackowski (37) have reported that 4Ј-PP can be directly derived from degradation of CoA and have argued that this pathway does not involve ACP phosphodiesterase. We have assayed the labeled material excreted by strain JT2 (⌬acpH) for 4Ј-PP by thin layer chromatography and our preliminary results indicate that a major fraction of FIGURE 5. Gel filtration analysis of native AcpH. A crude extract containing [ 35 S]methionine-labeled AcpH was analyzed on a Superdex 200 column (see "Experimental Procedures"). Bovine serum albumin (67 kDa), ovalbumin (43 kDa), and chymotrypsinogen A (25 kDa) were used as standards that (assuming that AcpH is a globular protein) should elute at approximately the same positions as the trimer, dimer, and monomer forms of AcpH, respectively. Fractions of 0.5 ml were taken at a flow rate of 65 l/min.

FIGURE 4. Purification of hexahistidine-tagged
AcpH under denaturing conditions. Low range molecular weight markers were run in the lanes marked M. Soluble and insoluble fractions were loaded in lanes SF and IF. W represents the wash fraction. The proteins were eluted from the Ni-NTA column with buffers of decreasing pH values: fractions E1-E4, pH 5.9, and fractions E5-E8, pH 4.5. Fractions E6, E7, and E8 were pooled, diluted to 10 g/ml, and refolded as described under "Experimental Procedures." the excreted material has the chromatographic properties of 4Ј-PP consistent with the report of Vallari and Jackowski (37).
Prosthetic group turnover was also examined in a derivative of strain SJ16 that overexpressed AcpH from plasmid pJT5. However, in this case, the turnover profile resembled that of the vector control (data not shown). This is presumably because an increase in prosthetic group turnover rate does not increase excretion of the label and reattachment of 4Ј-PP by AcpS maintains the level of labeled ACP. However, it should be noted that we could only very modestly overproduce AcpH because higher level production resulted in intracellular aggregation of the protein.
Substrate Specificity of AcpH-Vagelos and Larrabee (13) reported that partially purified AcpH cleaves acetyl-ACP at essentially the same rate as ACP. To determine whether or not AcpH was also capable of hydrolyzing ACP species having acyl groups of greater chain length, acyl-ACP substrates having fatty acyl chain lengths from C-6 to C-16 were synthesized from the free fatty acids, ACP, and purified V. harveyi acyl-ACP synthetase (26). 5 The acyl-ACPs were treated with AcpH as described, and reaction products were analyzed on 20% polyacrylamide gels containing 2.5 M urea (Fig. 8). AcpH hydrolyzed all of the acyl-ACP thioesters tested. Note that no holo-ACP was formed in these reactions, indicating that the acyl thioester linkages stayed intact during the assay. Although the results are not quantitative, the rate of hydrolysis appears to decrease with increasing chain lengths.
The enzyme therefore appears to demonstrate a preference for the unacylated ACP and short chain fatty acyl-ACPs over the medium and long chain species.
The ability of AcpH to cleave 4Ј-PP from the ACPs of other species was also examined. As described under "Experimental Procedures," strain SJ16 transformed with plasmids carrying genes encoding putative fatty acid synthesis acyl carrier proteins from Aquifex aeolicus, Bacillus subtilis, Lactococcus lactis, and the mitochondrial ACP of Bos taurus 6 under control of the E. coli PBAD promoter were depleted of CoA, then grown in medium containing 1.0 M [ 3 H]␤-alanine (1 Ci/mol). Expression of the ACPs was induced with 0.2% arabinose. Of the proteins tested only L. lactis ACP was not phosphopantetheinylated by E. coli AcpS (data not shown). Therefore the strain carrying this gene was additionally transformed with pNRD136, which expresses Sfp, a phosphopantetheinyl transferase of broad substrate specificity (38). The 3 H-labeled crude extracts of these cultures were treated with AcpH and analyzed by gel electrophoresis and fluorography (Fig. 9A). AcpH 6 N. R. DeLay and J. E. Cronan, manuscript in preparation. FIGURE 6. The protein product of the AcpH reaction. Panel A, electrospray mass spectral analysis of holo-ACP, purified, and prepared as described under "Experimental Procedures. " The peak of mass 8849.3 Da corresponds to holo-ACP. Panel B, the holo-ACP in A treated with purified AcpH. The minor peaks of greater mass represent sodium, potassium, and ammonium adducts of ACP. Panel C, phosphopantetheinylation of the AcpH reaction product by AcpS. Holo-ACP was either treated with 40 nM AcpH or left untreated. The AcpH-treated sample was trichloroacetic acid precipitated, dissolved in 1 M Tris base, and equilibrated with 10 mM Tris-HCl, pH 8.0. One-half of the reaction mixture was then treated with AcpS and CoA (see "Experimental Procedures") and the other half was left untreated. The products were analyzed by electrophoresis on a 20% non-denaturing gel. hydrolyzed the 4Ј-PP prosthetic group from the ACPs of A. aeolicus and B. subtilis but not those of L. lactis or B. taurus (note that all lanes contain the host ACP band). Sequence alignments of the regions flanking the phosphopantetheinylated serine are shown in Fig. 9B. These regions, in particular helix 2 of ACP, have been implicated as the primary determinants for substrate specificity of AcpS for carrier proteins (39,43,44). It is interesting to note that although the ACPs of B. subtilis and A. aeolicus are structurally similar to E. coli ACP and are substrates of E. coli AcpH, our searches failed to detect acpH homologs in these species.

DISCUSSION
Although ACP phosphodiesterase was discovered almost 40 years ago (13), the enzyme had not been purified to homogeneity and there was no valid information on the gene(s) that encode the enzymatic activity. We have accomplished both of these aims. The encoding gene formerly called yajB and now renamed acpH lies between two genes that encode enzymes unrelated to lipid metabolism. The upstream gene queA encodes an S-adenosylmethionine:tRNA ribosyltransferaseisomerase, whereas the downstream gene, malZ, encodes maltodextrin glucosidase. The acpH gene must be independently transcribed because both adjacent genes are transcribed from the other DNA strand.
We anticipated that multiple copies of acpH would be toxic to E. coli because apo-ACP would accumulate to high levels and block cell growth. However, this prediction was incorrect. Strains overproducing AcpH grow only slightly more slowly than wild type strains and this modest effect on growth could be as readily because of the accumulation of protein aggregates as to a direct effect of phosphodiesterase activity. Indeed, we found no detectable increase in the rate of ACP prosthetic group turnover upon AcpH overexpression (data not shown), which, however, is limited to modest levels of overexpression because of the aggregation of the protein upon overproduction. However, even modest overproduction might be inhibitory because AcpS, the enzyme responsible for attaching the 4Ј-PP prosthetic group to apo-ACP, is strongly inhibited by apo-ACP in vitro (8). Therefore, a plausible scenario would be that increased levels of apo-ACP engendered by AcpH overproduction would inhibit AcpS and thereby prevent reattachment of the prosthetic group. However, AcpH activity in crude extracts has been reported to be inhibited by physiological concentrations of CoA and acetyl-CoA both of which are substrates of AcpS (40). E. coli has two known 4Ј-PP transferases active on ACP, AcpS and AcpT (6,8). AcpS is thought to be the enzyme of physiological importance because it is an essential gene (41) and mutants deficient in AcpS activity accumulate apo-ACP (8,42). If in vivo E. coli AcpS is strongly inhibited by an AcpH substrate such as holo-ACP or an acyl-ACP, then the lack of toxicity of AcpH upon overexpression can be rationalized. In this scenario cleavage of the inhibitor would result in a large increase in AcpS activity that would counteract AcpH action and thereby give the constant holo-ACP pool we observe. Therefore, AcpH and AcpS may be prevented from forming a complete futile cycle by inhibition of one enzyme by the substrates of the other. We are currently investigating this possibility. It seems likely that AcpH activity is somehow modulated in vivo because based on the specific activity of purified AcpH and the cellular level of the protein measured by use of a strain carrying an epitope-tagged chromosomal acpH gene 7 that the cellular content of AcpH activity is suffi-7 J. Thomas and J. E. Cronan, unpublished data. FIGURE 8. AcpH utilizes acyl-ACP substrates. The acyl-ACPs were synthesized using free fatty acids and V. harveyi acyl-ACP synthetase as described under "Experimental Procedures." One-half of the acyl-ACP was treated with AcpH and the other half was left untreated. The AcpH reaction mixtures contained 5 M acyl-ACP and 33 nM AcpH. Holo-ACP was also treated with AcpH as a positive control. The strain carrying L. lactis ACP also carried plasmid pDHK29 encoding Sfp, which was induced with 10 M isopropyl ␤-D-thiogalactopyranoside. Crude extracts were made of which one-half was treated with AcpH and the other left untreated, and AcpH activity was followed by loss of the labeled bands. The reaction products were analyzed by gel electrophoresis followed by fluorography (see "Experimental Procedures"). Each well received 3 g of total protein. Strain SJ16 transformed with pBAD322 was used as a vector control and as a reference for E. coli ACP expressed from the chromosome in all extracts (lanes marked Control). Panel B, sequence alignments of conserved regions surrounding the phosphopantetheinylated serine of the acyl carrier proteins in panel A. The alignments were generated using ClustalW. Sequence identities are shaded in black and similarities in gray. The asterisk denotes the phosphopantetheinylated serine and the bracket denotes the helix 2 region. cient to hydrolyze all of the cellular ACP in Ͻ1 min. Hence there is clearly sufficient AcpH activity to account for the rate of prosthetic group turnover seen in vivo.
It seems interesting that thus far only those ACPs that are good substrates for AcpS in their apo forms are substrates for AcpH in their modified forms, suggesting that the two enzymes recognize the same or very similar ACP features. From the crystal structure of the B. subtilis AcpS-ACP cocrystals (12) and helix modification experiments (39), it is clear that the important AcpS recognition sites reside in helix 2 of ACP and hence it seems likely that AcpH also recognizes helix 2. Indeed all of the ACPs we tested that were not AcpH substrates have helix 2 sequences that differ markedly from that of E. coli (Fig. 9B).
Despite the progress we report, the physiological role of AcpH remains enigmatic. AcpH seems limited to Gram-negative organisms and we have shown that the enzyme is not essential for growth of E. coli. It also seems that AcpH is not required for growth in the natural environment because two organisms very closely related to E. coli, Shigella dysenteriae, and two strains of Shigella flexneri have an 8-bp deletion (probably because of recombination between two copies of a short directly repeated sequence) within the 5Ј-end of the gene expected to result in a truncated protein of 50 residues (Fig. 2). Surprisingly, although it should not be expressed because of the shift in reading frame caused by the deletion, translation of the DNA sequence downstream of the deletion gives a protein that is 96% identical to E. coli AcpH (Fig. 2). The same is true of the N-terminal fragment (Fig. 2). The fact that these sequences have not appreciably drifted away from those of E. coli argues that the deletion event was a recent occurrence or that the protein fragments perform some function that provides a selection for their maintenance. Note that Shigella sonnei, the only sequenced Shigella that has an intact acpH gene, is the most closely related of these bacteria to E. coli K12, the organism we have studied (45,46). It is interesting that when compared with E. coli the Shigellae seem prone to lose catabolic genes (46) because AcpH can be considered to be an enzyme involved in the catabolism of CoA (by the combined action of AcpS and AcpH CoA is converted to 4Ј-PP and 3Ј, 5Ј-ADP).
Given that AcpH is not an essential protein, what is its physiological role? The fact that the protein is found only in Gram-negative organisms suggests that it plays a role in some aspect of lipid metabolism that is unique to these organisms, the most obvious of which is biosynthesis of lipid A. Because AcpH is a hydrolyase, the most straightforward role would be to act as an editing enzyme that intercepts acyl-ACPs that would give an inappropriate lipid A structure if used as acyl donors. Recent work has shown that the composition of lipid A is not static (as long supposed), but changes markedly in response to environmental conditions (47,48). There may be environmental conditions that arise, perhaps during pathogenesis, in which an inappropriate acyl-ACP must be eliminated.