Holo-(Acyl Carrier Protein) Synthase and Phosphopantetheinyl Transfer in Escherichia coli *

Holo-(acyl carrier protein) synthase (AcpS) post-translationally modifies apoacyl carrier protein (apoACP) via transfer of 4′-phosphopantetheine from coenzyme A (CoA) to the conserved serine 36 γ-OH of apoACP. The resulting holo-acyl carrier protein (holo-ACP) is then active as the central coenzyme of fatty acid biosynthesis. The acpS gene has previously been identified and shown to be essential for Escherichia coli growth. Earlier mutagenic studies isolated the E. coli MP4 strain, whose elevated growth requirement for CoA was ascribed to a deficiency in holoACP synthesis. Sequencing of the acpS gene from theE. coli MP4 strain (denoted acpS1) showed that the AcpS1 protein contains a G4D mutation. AcpS1 exhibited a ∼5-fold reduction in its catalytic efficiency when compared with wild type AcpS, accounting for the E. coli MP4 strain phenotype. It is shown that a conditional acpS mutant accumulates apoACPin vivo under nonpermissive conditions in a manner similar to the E. coli MP4 strain. In addition, it is demonstrated that the gene product, YhhU, of a previously identified E. coli open reading frame can completely suppress theacpS conditional, lethal phenotype upon overexpression of the protein, suggesting that YhhU may be involved in an alternative pathway for phosphopantetheinyl transfer and holoACP synthesis inE. coli.

Escherichia coli utilizes a repeated cycle of condensation, reduction, dehydration, and isomerization reactions to produce saturated and unsaturated fatty acids (1,2). This biosynthetic mechanism employs multiple enzymatic activities conglomerately termed the fatty acid synthase. The central coenzyme in all fatty acid synthases is the holo form of acyl carrier protein.
ACPs, 1 as free proteins or as domains of large multifunctional proteins, function in a variety of synthases as hubs to which growing acyl intermediates and nascent product molecules are covalently tethered during the elongation and modi-fication steps required to produce the final compound. The apo to holo conversion of ACPs by post-translational modification with the 4Ј-phosphopantetheine (4Ј-PP) prosthetic group is essential to their function, since all acyl chain molecules undergoing biosynthesis by these systems are attached to the terminal cysteamine thiol of 4Ј-PP via a thioester bond.
The E. coli fatty acid synthase has served as a model system for understanding the 4Ј-PP post-translational modification, structure, and function of ACPs. The E. coli ACP is a highly acidic protein of 77 residues. It is composed primarily of a 3-helix bundle (3,4), and recent structural studies reinforce the notion that many ACPs from different synthases and/or organisms are of similar size and structure to E. coli ACP (5)(6)(7).
E. coli ACP is post-translationally modified by holo-(acyl carrier protein) synthase (EC 2.7.8.7) in a Mg 2ϩ -dependent reaction, where the 4Ј-PP moiety is transferred from CoA to the conserved serine 36 ␥-OH of ACP (8). AcpS therefore converts inactive apoACP to its activated holo form. HoloACP is then capable of being acetylated or malonylated, initiating the biosynthesis of a fatty acid (1,2). Upon the completion of the synthesis of the tethered fatty acyl chain, the acyl holoACP is a substrate for acyltransferase activities, which link fatty acid synthesis to phospholipid and lipid A synthesis (1,2).
A previous study of in vivo ACP phosphopantetheinylation (9) employed ethyl methanesulfonate mutagenesis of the pantothenate auxotrophic E. coli MP3 strain followed by screening for an elevated pantothenate growth requirement to identify mutants with altered AcpS activity. This approach was designed to yield mutants having an increased K m for CoA, since the intracellular concentration of CoA can be manipulated in strains auxotrophic for pantothenate (a CoA precursor) (9). This work resulted in the identification of the E. coli MP4 strain, which has a deficiency in AcpS activity and elevated levels of intracellular apoACP (9). Further study of this mutant by traditional genetic approaches (such as cloning by complementation) was hindered by the pleiotropic nature of the mutation and the rapid accumulation of revertants and/or suppressors.
Takiff et al. (10) first identified the E. coli open reading frame containing the gene that encodes AcpS. At the time, this gene was of unknown function and was named dpj. It was identified as an essential gene through the use of the conditional mutant E. coli HT253 strain, generated by a mini-Tn10 transposable element deletion inserted upstream to dpj. This transposon blocks transcription from the upstream promoter, resulting in a polar effect on downstream genes (10). Tetracycline induces the divergent promoters within the mini-Tn10 transposon and partially restores transcription of downstream genes (10). This analysis and that of Lam et al. (11) show dpj to be essential for the growth of E. coli.
Subsequent identification of dpj as the gene encoding the E. coli AcpS activity resulted in the renaming of dpj to acpS (12). This was the first gene known to encode a phosphopantetheinyltransferase, and its discovery renewed general interest in PPTases and the role of the 4Ј-PP post-translational modification in activating the synthases that produce a myriad of secondary metabolites (13). The E. coli acpS sequence led to the identification and subsequent confirmation of other PPTases in a variety of organisms (14). This includes two additional PPTases in E. coli: EntD, which phosphopantetheinylates the EntB and EntF members of the Ent biosynthetic gene cluster responsible for producing the enterobactin siderophore (15), and YhhU, a previously uncharacterized open reading frame of 195 amino acids (called o195) (14). Both the entD and yhhU gene products were demonstrated to modify apoACP in vitro at a very low rate (14). EntD has been shown to efficiently modify EntF in vitro, which validates an in vivo role of EntD in activating EntF consistent with the co-induction of the two genes and their presence in the same operon. The very low in vitro activity of the YhhU protein in modifying EntF and ACP led to the conclusion that the physiological substrate was an undiscovered acceptor protein (14).

EXPERIMENTAL PROCEDURES
Bacterial Strains and Plasmids-Strains and plasmids that were used are listed in Table I. All bacterial strains are derivatives of E. coli K-12. Strains carrying plasmids were constructed by electroporation or by 50 mM CaCl 2 treatment followed by heat shock with the respective plasmids (16). Strain RF104 was constructed by transduction of strain MP4 with a P1vir phage lysate grown on strain HT210 with selection for kanamycin resistance.
Genetic Techniques-P1 transduction was done according to Miller (17). Arabinose-inducible plasmids were induced by additions of 0.4% arabinose, alternatively 0.4% glucose was added to decrease basal expression from the arabinose promoter, or no sugars were added to give basal level transcription.
Recombinant DNA Manipulations-General DNA manipulations and standard molecular biology procedures were carried out as described in Sambrook et al. (16). Chromosomal DNA was prepared using the Genome DNA Kit from Bio 101, Inc. Plasmid DNA was purified by Plasmid Mini Kit from Qiagen, Inc. Southern blot analyses were performed with the Genius System from Roche Molecular Biochemicals. Restriction endonucleases were obtained from New England BioLabs, Inc. or Life Technologies, Inc.
Identification of the Mutation in the acpS Gene of E. coli MP4 -In the Boston laboratory, four independent PCR amplifications of the acpS gene from the E. coli MP4 strain were performed using the forward primer 5Ј-CGCCATCCCTGAGATGCATGA-3Ј and the reverse primer 5Ј-ACACTGATCCATAATCGCTGT-3Ј. These primers anneal to the E. coli chromosome at approximately 100 bases upstream and downstream from the acpS start and stop codons, respectively. PCR products were purified by agarose gel electrophoresis, and sequencing revealed a mutation in the acpS of the E. coli MP4 strain (denoted acpS1) containing a single G to A transition in the fifth codon (underlined), 5Ј-ATG-GCAATATTAGATX n . This mutation corresponds to a G4D substitution in the expressed protein (denoted AcpS1). This was mutation was confirmed in the Urbana laboratory upon sequencing of the PCR product obtained using another set of primers.
Cloning of acpS1 from E. coli MP4 for in Vitro Kinetic Studies-acpS1 was PCR-amplified using a forward primer that incorporated an NdeI restriction site (underlined) and the start codon 5Ј-TGTACCTC-AGACCATATGGCAATATTAGATTTAGGCACGG-3Ј. The reverse primer incorporated a HindIII restriction site (underlined) after the stop codon 5Ј-TGATGTCAGTCAAGCTTAACTTTCAATAATTACCGT-GGCA-3Ј. The resulting PCR product was digested with NdeI and HindIII, then subcloned into the corresponding restriction sites in the pET22b expression plasmid from Novagen, Inc. This construct was designated pRL102 and subsequently transformed into BL21(DE3), creating the strain RL102 for overexpression of the AcpS1 protein. rnc and pdxJ operons inserted into pBR322 10 pTX290 Ap r 11 pK18 Kn r 18 pRL101 Ap r acpS under T7 promoter in pET22b 12 pRL102 Ap r apcS1 under T7 promoter in pET22b This work a Antibiotic resistance to ampicillin, chloramphenicol, kanamycin, and tetracycline is indicated by Ap r , Cm r , Kn r , and Tc r , respectively. Tc s indicates sensitivity to tetracycline.
Cloning of acpS for in Vivo Complementation Studies-The acpS plasmid pCY314 was made by ligating the SspI-HindIII yhhU fragment of pTX290 (11) to the annealed product of two phosphorylated oligonucleotides, 5Ј-GTACCCATGGCTAT-3Ј and 5Ј-ATAGCCATGG-3Ј. The ligation product was then ligated to pK18 (18) digested with Acc65I and HindIII to give an acpS gene having a NcoI site (underlined) overlapping the acpS initiation codon. The NcoI-HindIII fragment that encodes a wild type AcpS protein was then ligated to NcoI plus the HindIII digest of pBAD22. Finally the ClaI-ScaI fragment of this construct was replaced with the ClaI-HincII fragment of pACYC184 to give pCY314.
Cloning of acpS1 from E. coli MP4 for in Vivo Complementation Studies-The acpS1 gene was amplified by PCR from the E. coli MP4 strain in the following manner. An NcoI site (underlined) was engineered at the start codon using primers 5Ј-CGTTTCCCATGGCAATAT-TAGAT-3Ј and 5Ј-GCATGCTGCACAAAATTAAAG-3Ј. The PCR product was blunt-ended with mung bean nuclease from New England BioLabs, Inc. and subsequently digested with NcoI. The NcoI/blunt product was ligated into NcoI/SmaI double-digested pBAD22 (19) to produce pYON107. To reduce the copy number, the ClaI/ScaI fragment of pYON107 carrying acpS1 was ligated into HincII/ClaI-digested pA-CYC184. The resulting plasmid was named pYON108.
Cloning of yhhU from E. coli W3110 for in Vivo Complementation Studies-The YhhU overexpression plasmid pYON113 was made using a PCR-amplified product from the wild type strain W3110 using an amino-terminal primer containing an introduced AflIII site (underlined) 5Ј-GGCCCTGACATGTATCGGATAGTTCTGGGG-3Ј and a carboxyl-terminal primer containing an introduced HindIII site (underlined) 5Ј-GTCGGGTAAGCTTATCAGTTAACTGAATCG-3Ј. The PCR products were digested with AflIII plus HindIII and ligated into pBAD22 digested with NcoI and HindIII.
Purification of Holo-(Acyl Carrier Protein) Synthases-AcpS was overexpressed from strain RL101 and purified according to a previously developed method (12). An additional purification step for the protein was appended to this method employing Sephacryl S-300 size exclusion chromatography from Amersham Pharmacia Biotech to remove a high molecular weight contaminant. AcpS1 from the E. coli MP4 strain was overexpressed from strain RL102. Purification of the AcpS1 was performed in a manner identical to wild type AcpS. Amino-terminal protein sequencing of the purified AcpS1 confirmed the presence of the G4D mutation.
Purification of Apoacyl Carrier Proteins-The E. coli strains DK554, DK675, and DK700 were used to overexpress the wild type, S36A, and S36T E. coli ACPs, respectively (20). Purification of apoACP was performed using a scheme involving a freeze/thaw osmotic shock release of ACP from cells followed by anion exchange chromatography with Q Sepharose from Amersham Pharmacia Biotech, Inc. The ACP containing eluant was passed through a Sephacryl S-100 size exclusion chromatography column from Amersham Pharmacia Biotech to remove a high molecular weight contaminant and to exchange the sample into a nonreducing buffer. The ACP pool was then slowly filtered through Thiopropyl Sepharose 6B from Amersham Pharmacia Biotech to separate apoACP from the contaminant holoACP, which covalently binds to the resin. The apoACP-containing flow-through was then subjected to Sephacryl S-100 size exclusion chromatography a second time to remove the trace disulfide-linked holoACP dimer, based on its 2-fold molecular weight difference. Analysis of the relative apo to holo content of the final purified protein using a previously developed HPLC method (21,22) indicated a Ͼ99% purity for apoACP. 2 The S36A and S36T apoACPs were purified in an analogous manner. Cells overexpressing each protein were subjected to the freeze/thaw, Q Sepharose, and Sephacryl S-100 purification steps. Subsequent filtration through Thiopropyl Sepharose 6B and Sephacryl S-100 were not performed, since it was expected that overexpression of the S36A and S36T ACPs would produce primarily the apo form, based on previous studies (20). Characterization of the purified S36A and S36T apoACPs by HPLC (21,22)  H]CoA, 10 mM MgCl 2 , 2.5 mM dithiotheitol, and 75 mM Tris base, pH 8.8, in a 100 l volume. After incubation at 37°C for 25 min, 900 l of 10% trichloroacetic acid was added to quench the reaction. The radiolabeled holo-ACP was precipitated and isolated by adding 500 g of bovine serum albumin as a carrier and centrifugation at 10,000 ϫ g for 5 min. Unreacted [ 3 H]CoA and 3Ј,5Ј-[ 3 H]ADP product were removed by decanting the supernatant and rinsing the protein pellet three times with 1 ml of 10% trichloroacetic acid. Pellets were dissolved in 150 l of 1.0 M Tris base (pH untitrated) and transferred to scintillation vials containing 2.5 ml of scintillation fluid from Packard, Inc. The 3 H-labeled holoACP was quantified with liquid scintillation counting. Reagents incorporated in the reaction mixtures varied depending upon the study being performed. Such exceptions to protocol are noted in relevant results sections and figure legends.
Analysis of in Vivo ACP Pools-ApoACP was analyzed by urea-PAGE as described previously (23) with the following modifications. Overnight cultures were diluted into fresh medium subcultured several generations to mid-log phase and then treated with 5% trichloroacetic acid. After centrifugation, the protein pellets were washed twice (at 4°C) with 1% trichloroacetic acid and resuspended in 1 M urea. These samples were then subjected to a 13% PAGE with 1 M urea. The separated proteins were subsequently visualized by staining with Coomassie Blue.

RESULTS
Kinetic Characterization of AcpS-Since the acpS1 mutation was suppressible by high intracellular CoA levels, the phenotype of the E. coli MP4 mutant strain seemed likely to be explained by kinetics of the mutant enzyme. However, the kinetic parameters of the wild type enzyme had to first be determined. Studies of the 4Ј-PP post-translational modification catalyzed by the wild type AcpS were performed with the 4Ј-[ 3 H]PP transfer assay. This assay allows for direct measurement of apoACP to holoACP conversion via trichloroacetic acid precipitation of the 3 H-labeled holo product and subsequent quantification by liquid scintillation counting. The predominant kinetic feature of AcpS was the severe substrate inhibition of the enzyme by apoACP at concentrations Ͼ5 M (Fig. 1,  panel A). AcpS had a K m of 1.5 M for apoACP and a peak 4Ј-PP transfer activity of k cat of 65 min Ϫ1 , occurring at an apoACP concentration of 5 M. The kinetic profile of AcpS as a function of apoACP concentration and all kinetic data displaying substrate inhibition have been fit by linear regression with the general substrate inhibition equation of Cleland (24). AcpS exhibits Michaelis-Menten kinetics and a K m of 50 M with respect to CoA as the variable substrate ( Fig. 1, panel B).
NaCl Effects on AcpS Kinetics-One explanation for the observed substrate inhibition was an electrostatic interaction between the basic AcpS enzyme (pI 9.6) and the acidic apoACP substrate (pI 4.1), which should be suppressed by high salt concentrations. Indeed, at an NaCl concentration of 250 mM, a moderate relaxation in the substrate inhibition of AcpS by apoACP was observed (Fig. 2). This relaxation was characterized by a slight increase in the K m for apoACP and an increase in the k cat of the enzyme at saturating apoACP concentration. The maximum observed activity for AcpS in the presence of 250 mM NaCl is 62 min Ϫ1 . At 500 and 750 mM NaCl, substrate inhibition of AcpS was dramatically reduced, and the enzyme exhibited near Michaelis-Menten behavior (Fig. 2). The enzyme continued to maintain a peak activity of ϳ62 min Ϫ1 . The apoACP concentration at which peak activity occurred increased ϳ10-fold as a function of the NaCl concentration, from 4 to 40 M.
Inhibitors of AcpS-Prior in vivo work (20) showed that a mutant ACP in which the target serine 36 had been mutated to a threonine residue was an inactive substrate for phosphopantetheinylation. This protein was tested to determine if is was an inhibitor of AcpS as well as an inactive substrate. The in vivo results were first confirmed in vitro. S36T apoACP was overexpressed and purified to Ͼ99% homogeneity. In addition, S36A apoACP was prepared in an identical manner as a nonphosphopantetheinylatable control. Incubations of AcpS with [ 3 H]CoA and either apoACP, S36A apoACP, or S36T apoACP were conducted for periods up to 5 h. The results of these assays clearly indicated the absence of any 4Ј-[ 3 H]PP labeling of S36T apoACP or the S36A apoACP control, 2 although the wild type ACP was readily labeled. The apparent inability of AcpS to phosphopantetheinylate S36T apoACP was also confirmed through separation of the apo and holo forms of ACPs after 1-h incubations with AcpS and CoA, 2 using a previously developed HPLC method (21,22). S36A and S36T apoACPs were then characterized as inhibitors of AcpS using the 4Ј-[ 3 H]PP transfer assay. Several incubations of AcpS with 10 M apoACP were performed with successively increased concentrations of either S36A apoACP or S36T apoACP in the absence of NaCl 2 as well as in the presence of 500 mM NaCl (Fig. 3). An IC 50 of 10 M was obtained for S36A apoACP (Fig. 3, panel A). This value remained nearly the same, regardless of the presence of NaCl in the reaction mixtures. An IC 50 of 7 M was obtained for S36T apoACP in the presence of 500 mM NaCl (Fig. 3, panel B). An IC 50 value for S36T apoACP in reactions conducted in the absence of NaCl could not be obtained, because it was observed that adding successively greater concentrations of S36T apoACP to these incubations resulted in a slight increase in AcpS activity. 2 Ability of S36T apoACP to Relieve AcpS Substrate Inhibition-In response to the observation that successively increasing amounts of S36T apoACP increased AcpS catalysis of 4Ј-PP transfer to apoACP in the absence of NaCl, a kinetic study of the S36T apoACP activation of AcpS was performed. 100 M S36T apoACP was added to a series of 4Ј-[ 3 H]PP transfer assays investigating AcpS catalytic activity with respect to the apoACP substrate concentration (Fig. 4). Comparison of the results of this study with the previous characterization of AcpS kinetics performed in the absence of S36T apoACP clearly indicated that S36T apoACP activates AcpS through relief of substrate inhibition by apoACP. In the presence of 100 M S36T apoACP, AcpS had Michaelis-Menten kinetics with a k cat of 80 min Ϫ1 and a K m of 18 M for apoACP.
Identification of G4D Mutation in AcpS1-The E. coli MP4 phenotype had been attributed to a deficiency of AcpS activity due to expression of a mutant allele (denoted acpS1) encoding a protein having a postulated elevated K m for CoA (9). PCR amplification and subsequent DNA sequencing of acpS1 from E. coli MP4 revealed that the fifth codon contained a single G to A transition. This base change resulted in a glycine to aspartate substitution at the fourth residue in the final acpS1 gene product. Recombinant E. coli acpS1 was expressed and purified in a manner identical to that used for AcpS. The G4D mutation in the purified AcpS1 protein was confirmed by amino-terminal protein sequencing.
Kinetic Characterization of AcpS1-Kinetic characterization of AcpS1 was conducted in a manner analogous to that of AcpS. The AcpS1 enzyme showed a ϳ5-fold reduction in its catalytic activity as a result of the G4D mutation, with a k cat of 7.5 min Ϫ1 at saturating apoACP concentration and a peak activity of k cat ϭ 10 min Ϫ1 at an apoACP concentration of 5 M (Fig. 1,   panel A). Substrate inhibition of AcpS1 by apoACP was significantly less severe than that which was observed for AcpS. The AcpS1 K m for apoACP was 2.5 M, only marginally greater than that observed for AcpS. AcpS1 had Michaelis-Menten kinetics with respect to the CoA substrate with a K m of 75 M (Fig. 1,  panel B).
acpS Mutants Accumulate ApoACP in Vivo-Previous studies of acpS activity in vivo utilized the E. coli MP4 acpS1 mutant strain, which has a complex phenotype (9). Since a characterization of the AcpS1 protein in vivo was desired, the more straightforward mutant E. coli HT253 strain of Takiff et al. (10) seemed ideal because under nonpermissive conditions this strain should produce no AcpS protein. This lack of protein production would preclude formation of mixed AcpS dimers (having a mutant subunit and a wild type subunit) which could complicate the phenotypic analysis. However, since strain HT253 had not been examined biochemically, the ACP species present in vivo were examined under both the permissive and nonpermissive conditions. It was expected that if AcpS is the primary or only enzyme capable of carrying out the 4Ј-PP modification of ACP in E. coli, then a mutation that affects the transcription of acpS should hinder modification of apoACP and result in considerable accumulation of intracellular apoACP. The acpS tetracycline-dependent, conditional mutant E. coli HT253 strain was grown overnight under permissive conditions (the presence of tetracycline to induce the tet promoter), then subcultured under either permissive or nonpermissive (absence of tetracycline) conditions. The ACP species present were subsequently identified by their migration patterns on urea-PAGE (Fig. 5). Strain HT253 accumulated detectable quantities of apoACP even in the presence of tetracycline, indicating that transcription off the mini-Tn10 element inserted between acpS and its promoter did not result in normal levels of AcpS production and holoACP synthesis (Fig. 5,  lanes 1 and 2). In the absence of tetracycline the apoACP level increased further, and the holoACP decreased (Fig. 5, lane 3). It should be noted that strain HT253 continued growth for a few hours after tetracycline removal, an interval that probably depleted the pool of ACP. The acpS1 Allele of MP4 Is Responsible for the Mutant Phenotype-Experiments were undertaken to demonstrate that the acpS1 of the E. coli MP4 strain is responsible for its mutant phenotype, where high intracellular CoA concentrations are required for growth. The acpS1 gene of the E. coli MP4 strain was cloned behind an inducible araBAD promoter, and the ability to complement the tetracycline-dependent acpS condi-tional mutant E. coli HT253 strain was tested (Fig. 6). Unlike the wild type acpS plasmid pCY314, which complemented the mutant without induction of the arabinose promoter, the plasmid pYON108 carrying the acpS1 mutant allele complemented the mutation only upon induction by addition of 0.4% arabinose (Fig. 6, plate 3). This experiment indicated that the acpS1 allele encoded a less active enzyme and that it is this mutant gene which is responsible for the phenotype of the E. coli MP4 strain, requiring elevated CoA levels for growth.
Additional studies of acpS1 examined phenotypes of strains using the procedure of Polacco and Cronan (9), which was originally employed in the identification of the E. coli MP4 strain. This assay examined the growth of strains carrying a panB6 lesion in minimal media supplemented with pantothenate in order to distinguish among strains requiring different levels of intracellular CoA for viability.
Strain RF103 was generated by transforming the E. coli MP4 strain with the complementing plasmid pDLC140, which carries the wild type acpS under the control of its native promoter. Growth of this strain in minimal media supplemented with 0.25 M pantothenate and in the presence of ampicillin was monitored by light absorbance at wavelength 600 nm. Comparison of the growth phenotype of the strain RF103 with strains RF101 (MP3 carrying pBR322) and RF102 (MP4 carrying pBR322) indicates a partial recovery of the mutant phenotype through direct complementation of acpS1 (Fig. 7, panel A).
replaced with wild type acpS. Phenotyping of this strain using the Polacco and Cronan (9) assay also revealed partial recovery of the MP4 phenotype when compared with strains MP3 and MP4 (Fig. 7, panel B).
Overproduction of YhhU Suppressed the acpS Mutation of HT253-The exact role of the third E. coli PPTase YhhU remains unclear (14). As with EntD, studies revealed that YhhU modified apoACP in vitro with very low efficiency (14). However, there were two caveats to these findings. First, the protein was produced as inclusion bodies, and refolding to the native structure was assumed but not demonstrated. Second, the protein produced contained an extra glycine residue adjacent to the initiation methionine. Since the mutation in acpS1 of the E. coli MP4 strain indicated that the amino terminus played an important role in enyzme activity, this extra residue might have compromised the activity of the protein.
The ability of the YhhU product to modify apoACP in vivo was tested. The yhhU gene was placed behind the araBAD promoter to give pYON113, in which the gene was expressed from the plasmid promoter and ribosome binding site. When transformed into the tetracycline conditional E. coli HT253 strain and induced with 0.4% arabinose, YhhU fully complemented the tetracycline-dependent growth phenotype of the strain (Fig. 8, plate 3) and blocked the accumulation of apoACP (Fig. 9, lane 4). However, when expression off the arabinose promoter was not induced or was decreased by the addition of glucose to the medium, growth of the strain was tetracyclinedependent (Fig. 8, plate 2 and Fig. 9, lane 3). These results confirm the ability of the yhhU gene product to functionally replace AcpS in vivo but only when expressed at high levels. It is proposed that this gene be named acpT to reflect this property.

DISCUSSION
Phosphopantetheinyl transfer by AcpS was studied in vitro, establishing a context for understanding the physiology of holo-ACP formation in E. coli. The predominant feature of AcpS steady state kinetics is the unusually severe substrate inhibition of the enzyme by apoACP (Fig. 1). Although this effect crippled AcpS catalytic activity by nearly 50% at apoACP concentrations above 10 M, there appeared to be no physiological basis for the accumulation of apoACP and substrate inhibition of the enzyme in vivo. Previous studies of in vivo ACP pools reveal that the apo from of the protein is not detectable in wild type E. coli (25), implying a rapid and efficient conversion of the protein to the holo form by AcpS. Indeed, the overexpression and accumulation of apoACP in E. coli has been shown to be toxic (20).
Factors that relieve AcpS substrate inhibition in vitro were identified, providing clues about a possible biophysical mechanism(s) responsible for this effect. High concentrations of NaCl in the 250 to 750 mM range elicited a dramatic relief of the substrate inhibition of AcpS by apoACP (Fig. 2). This effect is attributed specifically to the Na ϩ cation rather than the Cl Ϫ anion, since other monovalent and divalent chloride salts have different impacts on AcpS kinetics. 2 Monovalent and divalent cations are known to activate specific component enzymes in E. coli fatty acid biosynthesis (26) and can influence ACP conformational stability and structure (27). It should be noted that the severe substrate inhibition of AcpS observed in the absence of NaCl is not likely to represent the true intracellular kinetic profile of the enzyme, given the ionic content of cells.
The presence of S36T apoACP was also shown to relieve substrate inhibition of AcpS by apoACP in the absence of NaCl (Fig. 4), even though S36T apoACP itself cannot be phosphopantetheinylated. 2 The introduction of the ␥-CH 3 in S36T apoACP blocked the conversion of this protein to the holo form, but the binding of S36T apoACP by AcpS may induce an allosteric change in the homodimeric enzyme that relieves substrate inhibition by apoACP. Based on these findings, it is believed that AcpS substrate inhibition may arise from the presence of multiple apoACP conformers, improper binding of apoACP to active site(s) on AcpS, and/or negative cooperativity between apoACP binding sites on the AcpS homodimer. Additional structural information about AcpS will be required to fully understand the nature of this complex kinetic behavior.
Having obtained a reasonable kinetic profile for phosphopantetheinyl transfer by AcpS, the physiology of the E. coli MP4 strain and the basis for its deficiency in holoACP synthesis were investigated. The E. coli MP4 strain was phenotypically selected as a strain requiring growth medium supple- mented with an unusually high 25 M pantothenate concentration (a CoA precursor) for viability. At 0.25 M pantothenate, a level sufficient to support growth of strains having a wild type acpS gene, E. coli MP4 cultures ceased growth and had barely detectable holoACP levels. When supplemented with 25 M pantothenate, E. coli MP4 cultures grew and con-tained a mixture of both apo and holo forms of ACP. The rationale of this selection was that the addition of high levels of pantothenate to the growth medium would increase the intracellular CoA concentration and allow function of a mutant AcpS activity having an elevated K m for CoA (9).
Sequencing of the acpS gene from the E. coli MP4 strain (denoted acpS1) identified a G4D mutation located immediately amino-terminal to the (V/I)G(V/I)D motif conserved among currently known PPTases (14). The kinetic profile of the AcpS1 protein in vitro did not reveal a significantly elevated K m for CoA. Instead, it is believed to be the ϳ5-fold reduction in catalytic efficiency k cat /K m of AcpS1 that results in phenotypic 2-to 10-fold lower in vivo holoACP content of the E. coli MP4 strain.
Although it had been shown that a point mutation existed in the acpS1 gene isolated from the E. coli MP4 strain, the possibility that a second mutation elsewhere in the chromosome might be responsible for the elevated CoA growth phenotype of this mutant had not been ruled out. Therefore, genetic analysis of the acpS1 allele of the E. coli MP4 strain was undertaken. Complementation of the tetracylcine-dependent acpS strain HT253 by overproducing the acpS1 allele carrying the MP4 point mutation was successful (Fig. 6). In addition, complementation and chromosomal replacement of the acpS1 with wild type acpS in strains RF103 and RF104, respectively, resulted in partial recovery of the MP4 growth phenotype (Fig. 7). These results suggest that the mutation in the acpS1 gene is largely responsible for the high in vivo CoA concentrations required by the E. coli MP4 strain for growth. The original mapping of the MP4 mutation to a different locus on the E. coli chromosome (9) could have been due to a lack of genetic markers then available within the region, resulting in inaccurate positioning of the mutation. Another possibility may be that the pleiotropic nature of the mutation and the rapid accumulation of revertants or suppressors led to mapping of a suppressor locus.
Since these in vivo results indicate that the acpS1 mutation is responsible for the phenotype of MP4, the kinetic properties of the expressed AcpS1 protein must now be reconciled with the physiology of the E. coli MP4 strain. The ability of such a modest mutational alteration of AcpS catalytic efficiency to produce such large physiological effects needs to be resolved.
E. coli AcpS is normally made in little functional excess over the enzymatic level required to convert all of the apoACP to the holo form. This was shown by Keating et al. (20), who report that even a 2-fold overproduction of apoACP resulted in detectable accumulation of the unmodified form in vivo. This unusually close matching of the synthetic capacity of an enzyme to the level needed for sufficient synthesis of the product means that only a small loss of catalytic efficiency is required to elicit a physiological effect. Therefore, the 5-fold reduction in catalytic efficiency of AcpS1 when compared with wild type AcpS can readily explain the accumulation of apoACP observed in the E. coli MP4 strain. However, the suppression of the physiological defect by the elevated CoA levels resulting from pantothenate supplementation remains to be explained.
It is believed that the mechanism of suppression of the effects of the acpS1 mutation by elevated CoA levels also involves AcpS synthetic capacity. It is clear that the wild type AcpS does not function at its maximal synthetic capacity in vivo. Keating et al. (20) show that expansion of the intracellular CoA pools in cells overproducing apoACP (through pantothenate supplementation together with a mutation giving a feedback-insensitive pantothenate kinase) markedly increased the conversion of apoACP to the holo form. Therefore, in wild type E. coli the activity of AcpS is limited by the CoA supply, and additional holoACP synthetic capacity resulted from increased CoA pools. This view is consistent with the K m for CoA that was obtained for AcpS and the intracellular concentration of CoA (see below).
It is expected that CoA pool expansion similarly increases the catalytic capacity of the AcpS1 protein, thus allowing sufficient holoACP synthesis for growth of the E. coli MP4 strain. The magnitude of the increased synthetic capacity seems likely to be about 2-to 3-fold, since the intracellular concentration of CoA in glucose-grown E. coli strains is about 50 M, whereas the K m for CoA of AcpS1 is 70 M (28). Since intracellular CoA levels reach Ͼ500 M upon the addition of 25 M pantothenate (9), the AcpS1 enzyme should operate at its maximal synthetic capacity under these conditions to give a holoACP pool about 20 -30% that of a wild type strain, a level that allows cell growth (9).
Previous studies also have demonstrated that disruptions of acpS are inviable but could survive in the presence of suppressor mutations (11). However, these disruptions were not confirmed by Southern analysis and may have resulted in the formation of tandem duplications on the chromosome that might have led to residual AcpS activity. To rule out such trivial modes of suppression, a Southern analysis was performed on TX2004 (11), an acpS::MudJ mutant. Using the entire acpS gene as the probe, the MudJ insertion into the acpS gene was confirmed. 3 Because suppressor mutants of E. coli can be isolated despite the lack of functional AcpS, alternative post-translational modification enzyme(s) for apoACP must exist in the organism.
There is evidence that the two other E. coli PPTases, EntD and YhhU, are capable of inefficiently modifying apoACP in vitro (14). Therefore, the ability of EntD and YhhU to modify apoACP was tested in vivo. Expression of EntD and other genes involved in enterobactin biosynthesis is under the regulation of Fur, a transcriptional regulator that responds to intracellular iron concentrations (29). The presence of intracellular iron is sensed by Fur, and Fur correspondingly acts as a transcriptional repressor for iron-regulated genes. Regulation of AcpS has not been observed to be repressed by Fur, and it is unlikely that residual expression of EntD under normal, iron-rich conditions would provide sufficient activity to modify apoACP in an acpS Ϫ strain. Nonetheless, acpS suppression by derepression of entD by iron starvation was examined, and no suppression of acpS was found under these conditions. 3 Cloning of entD under an inducible promoter and testing its ability to suppress an acpS Ϫ phenotype gave a negative result, 3 despite the fact that the EntD that was expressed was capable of phosphopantetheinylating EntF in crude cell extracts. This provides further evidence that entD is not able to suppress acpS and modify apoACP in vivo.
Up to this time, not much was known about YhhU other than the fact that it has sequence homology with other PPTases. Similar to entD above, the yhhU gene was cloned under an inducible promoter to test for the ability to suppress an acpS Ϫ phenotype. Induction of yhhU allowed the HT253 conditional strain to grow in nonpermissive media lacking tetracycline, as well as completely modifying all of the apoACP pool (Figs. 8  and 9). The fact that the in vitro analysis of YhhU demonstrated such low activity on apoACP (14) does not necessarily contradict this result. One of the problems with the in vitro analysis of YhhU was that the initial attempts in cloning and purification of YhhU led to an additional glycine after the start site. Another problem was that the purification process led to an inclusion-body form of YhhU that might have not been properly refolded (14). Improper folding of YhhU could have been responsible for the low in vitro activity. This work demonstrated that YhhU is able to modify apoACP when overexpressed in conjunction with an acpS mutation.
One possible explanation of acpS suppression by excess YhhU is that E. coli evolved two enzymes that are capable of modifying apoACP due to its crucial role in cell survival. In a normal cell, AcpS could be the primary enzyme with more efficient activity. The secondary enzyme, YhhU, is less active and thus plays no major role in 4Ј-PP modification of apoACP. Mutations in acpS could possibly allow selection for higher expression of the secondary enzyme to provide apoACP modification for survival.
Previous physiological studies originally revealed the ability of the E. coli AcpS to maintain complete holoACP pools in vivo under normal conditions. The results presented in this work indicate that AcpS also exists at levels minimally required and operates with an efficiency minimally necessary to ensure the complete conversion of apoACP to holoACP. A disruption of AcpS production or a small reduction of AcpS catalytic efficiency are capable of reducing phosphopantetheinyl transfer to the point where significant (and possibly toxic) levels of apoACP accumulate in vivo. Under circumstances where AcpS activity is compromised, possible mechanisms for physiological recovery by E. coli have been identified and involve elevations in intracellular CoA pools or increased expression of the complementing PPTase YhhU such that in vivo holoACP is maintained at levels sufficient for cellular viability.