The gene encoding Escherichia coli acyl carrier protein lies within a cluster of fatty acid biosynthetic genes.

The gene encoding Escherichia coli acyl carrier protein (ACP) has been isolated and sequenced. The ACP gene (called acpP) was located on the genetic map between fabF and fabD which encode two fatty acid biosynthetic enzymes, 3-ketoacyl-ACP synthase II and malonyl CoA-ACP transacylase, respectively. An open reading frame between acpP and fabD encodes a 26.5-kDa protein that has significant sequence identity (greater than 40%) with two acetoacetyl-CoA reductases and thus is believed to encode a 3-ketoacyl-ACP reductase. This gene (called fabG) is cotranscribed with acpP. Thus, the gene encoding ACP, the key carrier protein of fatty acid synthesis, is located within a cluster of fatty acid biosynthetic genes.

ACP has also been shown to function in three unexpected areas of metabolism: (i) as a cofactor in the synthesis of the membrane-derived oligosaccharides found in the periplasm of E. coli (9); (ii) as an essential component in the induction of nitrogen-fixing nodules by Rhizobia (where ACP appears involved in the synthesis of acylated oligosaccharides (10); and (iii) most recently as a subunit of mitochondrial NADHubiquinone oxidoreductases (11, 12).
E. coli ACP is the paradigm of this class of proteins. E. coli ACP was the first such protein isolated (13), the first in which the primary sequence was determined (14), and is the only ACP of known solution structure (15,16). Despite the continuing interest in E. coli ACP, the gene encoding this protein had not been isolated, and no mutants are available. We report the isolation of the ACP-encoding gene and its location within a cluster of genes encoding known enzymes of fatty acid synthesis.

EXPERIMENTAL PROCEDURES
All bacterial strains used in this study were derivatives of E. coli K-12. Strains JM103 (17) and F-M15A (18) have been described elsewhere. Strain DB6430 (F-argE A(lac-pro) rif, nal') was used as a source of chromosomal DNA. Strains L48 (fabD89), DM57 (fabBZ0 zfa::TnlO), and DM83 (fabF3 fabB20) used to map the kanamycin resistance (KanR) determinant were described previously (19). The growth media and genetic methods were as described (19). Plasmid pMR23 was constructed by ligation of a 0.9-kbp PstI-ScaI fragment from Tn9 into pACYC177 (20) digested with the same two enzymes. Plasmid pMR24 was constructed by ligating the 2.6-kbp EcoRI-PstI chromosomal fragment ( Fig. 1) from the M13 mp18 clone containing the acpP region to pMR23 digested with the same enzymes. Plasmid pMR33 was constructed by ligating the 1.5-kbp EcoRI-PuuII fragment of pMR24 into pTZ19R (21) digested with EcoRI and HincII. Plasmid pMR36 is a derivative of pMR24 having the 2.1-kbp KpnI fragment in the opposite orientation. Plasmid pMR39 was constructed by inserting the KanR gene excised from plasmid pUC4K (Pharmacia LKB Biotechnology Inc.) with HincII into the NruI site of pMR33 and was used to introduce the KanR determinant into the chromosome by homologous recombination (19). Plasmid pMR48 was derived from the 1.1-kbp PstI-PuuII chromosomal fragment modified to include a second flanking PstI site. This PstI fragment was ligated to PstI-digested pTZ19R (21) such that the fabG gene was in the orientation opposite that of the vector lac gene. Plasmid pMR62 was constructed by digestion of pMR24 with EcoRI and Sam, filling of the recessed ends by DNA polymerase I, and followed by religation.
The acpP gene was isolated from a library of 2-3-kbp SalI-BglII fragments of strain DB6430 genomic DNA ligated into M13 mp18 RF digested with Sal1 and BamHI. Strain JM103 was transformed with the recombinant DNA and the resultant plaques transferred in situ to nitrocellulose membranes such that single-stranded DNA was selectively retained (22). The plaques were screened (17) with the [a-32P]ATP-labeled synthetic ACP gene (18) that encodes the entire protein sequence and has 88% DNA sequence identity with the acpP gene.

RESULTS AND DISCUSSION
Our previous attempts to isolate the ACP gene were unsuccessful despite application of several different cloning and detection strategies. It, therefore, seemed possible that DNA segments encoding ACP were somehow toxic to E. coli. To investigate this possibility, we assembled a synthetic gene encoding ACP (18) and, indeed, found that high level production of ACP was lethal to E. coli (39).' In light of this finding, ' M. Rawlings and J. E. Cronan, Jr., manuscript in preparation. 5751 we sought the ACP gene using the synthetic sequence as a hybridization probe and maintained cloned DNA segments in a low copy number vector to limit ACP expression.
The synthetic ACP gene proved a stringent hybridization probe in Southern blot analysis of E. coli chromosomal DNA fragments (data not shown). A size-selected mini-library of genomic fragments was constructed in a phage M13 vector, and recombinant phage plaques were screened with the 32Plabeled synthetic gene. Screening was done under conditions allowing hybridization only to single-stranded DNA bound to the nitrocellulose filters, thus avoiding the background due to homologous sequences present in the chromosomal DNA (22). Consistent with the toxicity of the synthetic gene, the high (albeit variable) copy number of M13 clones carrying the natural gene (called acpP) gave spontaneously deleted variants at very high frequency, thus necessitating transfer of acpP to a low copy number vector to give plasmid pMR24. Even in such a plasmid, the presence of the acpP gene resulted in a decreased cellular growth rate.
The nucleotide sequence of the acpP gene region ( Fig. 1) showed an open reading frame (ORF) that agreed with the ACP amino acid sequence. Note that the two published ACP amino acid sequences conflict at two positions. Vanaman et al. (14) reported residues 24 and 43 as Asp and Val, respectively, whereas Jackowski and Rock (23) reported residues 24 " " " " " " and 43 as Asn and Ile, respectively. We find residue 24 to be Asn and residue 43 to be Val. Our results are consistent with assignments from nuclear magnetic resonance analysis (15). The observed post-translational removal of the N-terminal Met is consistent with the known specificity of the aminopeptidase (24). The codon preference (25) (27)). This location is very close to those we previously assigned to genes encoding two other fatty acid biosynthetic proteins, fabD and fabF, which encode malonyl-CoA-ACP transacylase and 3-ketoacyl-ACP synthase 11, respectively (1,19). It therefore, seemed probable that acpP was linked to these genes. Indeed, when phage P1 grown on the KanR insertion strain was used to transduce a fubD strain to KanR, 98% of the transductants were fab+. The KanR insertion of strain MR52 could not be mapped in relation to fabF because the insertion of the KanR element resulted in a strain having a fabF phenotype. fabF mutants have no growth phenotype unless the strain also has a lesion in the fabB gene that encodes 3-ketoacyl-ACP synthase I (1,19). Mutants with a temperature-sensitive lesion in fabB (fabBt") fail to grow at 42 "C on the usual media but grow well if the medium is supplemented with oleate (or other appropriate unsaturated fatty acids). fabFfabBta double mutants fail to grow at 42 "C even when supplemented with oleate (due to defective synthesis of saturated fatty acids). We found that P1 cotransduction of a fabBt" lesion into strain MR52 gave a fabBtsfabF phenotype. Strain MR52 also showed other aspects of the fabF phenotype (19): (i) an increased level of palmitoleic acid and a decreased level of cis-vaccenic acid compared with the parental strain lacking the insertion; (ii) defective thermal regulation of fatty acid composition; (iii) lack of the 3-ketoacyl-ACP synthase 11-ACP mixed disulfide in cell extracts. Thus, the KanR insertion of strain MR52 is either in fabF or is polar on fabF expression. We favor the former explanation since strain MR52 was unable to donate a functional fabF gene to a fabBt8fabF strain via P1 transduction. The segment of DNA containing acpP was also located on the physical map of E. coli. The acpP DNA segment hybridized to phages 235 and 236 of the ordered miniset bank of Kohara and co-workers (28). Comparison of our restriction map to the physical map (28) placed the acpP gene at 1170 kbp, a site fully consistent with the genetic map location.
Given the close genetic linkage of acpP, fabD, and fabF, we examined the proteins encoded by the aepP plasmid to see if proteins the size of the fabD and fabF gene products (35 and 43 kDa, respectively) were encoded by the chromosomal insert of the plasmid. Analysis of the products of a maxicell (29) labeling procedure (which gives specific labeling of plasmidencoded proteins) by sodium dodecyl sulfate-polyacrylamide gel electrophoresis showed that the chromosomal insert encoded proteins of 20, 26.5, and 43 kDa in addition to the proteins encoded by the vector sequences (Fig. 2). The 20-kDa protein is ACP which migrates aberrantly (as though a s7.53  (20); lane 4, pMR24, which carries the entire 2.6-kbp chromosomal insert in pMR62 (Fig. 1); lane li, pMR48, which carries the left-hand 1.1-kbp PstI-PuuII fragment (Fig. 1); and lane 6, pMR33 which contains the right-hand EcoRI-PuuII fragment (Fig. 1). pMR48 and pMR33 are derived from vector pTZ19R (20). The 20-and 26.5-kDa proteins were the products of the acpP and fabC genes, respectively. kan, aminoglycoside phosphotransferase; pRla, the precursor of 8-lactamase; Rla, 8-lactamase. CHIC PAT . . .

now K A L~Q R A~A D L C C V D I L V N N A G I T K D G L~L H~A D~D * D I V L u w K T A~D K V K S C V G~V D V L I N W A~I T~D V V~R K~T R A D N D A V I
n n n n r-

G~A D R O~G R I V N I S S V N G Q K C Q I
n n n n n

O K C K I Y V A I P I H R I G T G T C V A I V K A I R O D V L D K I V A T I P V X R L C L P C E I A S
n n n nn  2 ) or pMR36 (lnnr 5 ) were prohed with the synthetic ACP gene. Plasmid pMR24 contains the intact acpP-fabC region whereas in pMR36 the acpl' gene is inverted. Imnes 3 and 4 are the same RNA samples of lanes i and 2, respectively, hut the probe was the 0.4-khp Kpnl-Stul fragment (Fig. 1 ) specific for the fabC gene. Plasmid DNA (which is isolated with the RNA in the procedure used (38)). the cross-hyhridizing rRNAs. and the size (in kb) of the two major transcripts are indicated at the l~f t margin.
much larger protein) due to lower sodium dodecyl sulfate binding than the marker proteins (30). Deletion of the ACP sequence resulted in loss of the 20-kDa protein. Maxicell analysis of various subclones (Fig. 2) showed that the 43-kDa protein was encoded by a DNA segment located downstream of acpP, and hence, this protein may well represent a slightly truncated fabF gene product, 3-ketoacyl-ACP synthase 11, a protein of 44 kDa (31). (The Kan" insertion of strain MR52 would interrupt the synthesis of this protein.) The 26.5-kDa protein is encoded upstream of acpP but is not the fahD gene product since malonyl CoA-ACP transacylase has a molecular mass of 36.5 kDa (32). We have sequenced the DNA segment upstream of acpP and find an ORF that encodes a protein of 244 residues having a calculated molecular weight of 25,549 in good agreement with the maxicell results (Fig. 2). Comparison of the derived amino acid sequence of this ORF with those of GenBank showed a number of proteins with strong similarity to the ORF. The most definitive of these similarities was with two enzymes involved in poly-3-hydroxybutyrate synthesis in bacteria (33, 34). These are acetoacetyl-CoA reductases which reduce acetoacetyl-CoA (formed by condensation of two acetyl-coA units) to the 3-hydroxybutryl-CoA used in polymer synthesis (33, 34). The ORF upstream of acpP showed 43 and 41% amino acid identity with the NADPH-specific acetoacetyl-CoA reductases of Zmgloea ramigeria (33) and Alcaligenes eutrophrcs (34), respectively (Fig.  3). We also found significant similarities to a segment of the large polyfunctional fatty acid synthase proteins of rat ( 5 ) and chicken (35). Strong similarities (40-S3% amino acid identities) were also found to genes involved in polyketide synthesis in various Streptomyces (7) and to the nodG protein of Rhizobium meliloti (36) which may be involved in synthesis of acylated polysaccharides (10,37). These relationships (Fig.  3) together with the presence of a plausible NADPH binding site in the upstream ORF and cotranscription of the ORF with acpP (see below) lead us to believe this ORF encodes a 3-ketoacyl-ACP reductase of fatty acid biosynthesis (13), a gene we term fabC.
The close juxtaposition of these coding sequences suggested possible cotranscription of these genes. The maxicell results suggested promoters were present just upstream of both the acpP and fabG coding sequences, and this was confirmed by Northern blot analyses. Two chromosomal transcripts of about 0.3 and 1.1 kb were detected using the synthetic ACP gene as a probe (Fig. 4). A strain carrying pMR24 showed increased levels of both mRNA species. The 0.3-kb transcript has the capacity to encode ACP whereas the 1.1-kb transcript could encode both acpP and fabG. Only the 1.1-kb transcript was observed when a probe containing fabG sequences alone was used (Fig. 4). The relationship between the 0.3-and 1.1kb transcripts was examined by inverting the acpP coding sequence within pMR24. When probed with the synthetic ACP gene, the 0.3-kb mRNA was the dominant transcript ( Fig. 4). Therefore, the 0.3-kb mRNA seems a primary transcript and not a degradation product of the 1.1-kb mRNA. Thus, acpP is transcribed from two promoters. A strong promoter is located just upstream of the coding sequence, and a second is located upstream of the fabG sequence.

CONCLUSIONS
ACP is encoded by the acpP gene. Genes encoding other fatty acid biosynthetic genes lie both upstream and downstream of acpP. One downstream gene is fabF, encoding 3ketoacyl-ACP synthase 11, and several genes are located upstream. We have shown that fabG (which almost certainly encodes a 3-ketoacyl-ACP reductase) lies just upstream of acpP and is cotranscribed with acpP. In other work (41) we find that fabD encoding malonyl CoA-ACP transacylase is located just upstream of fabG. Indeed the first 63 bp of Fig. 1 encode the last 21 amino acids of fabD (41). Upstream of fabD lies another ORF (called fabH) that encodes 3-ketoacyl-ACP synthase I11 (40). Thus, the genes encoding several enzymes of fatty acid synthesis are clustered around the acpP gene and the toxicity of increased expression of acpP probably explains prior difficulties in cloning these genes. Our current map of this region (clockwise on the physical/genetic map) is fabH-fabD-fabG-acpP-fabF. The gene order has no obvious relationship to the order of protein domains in the polyfunctional fatty acid synthases of mammals or fungi.