Structural, Enzymatic, and Genetic Studies of jl-Ketoacyl-Acyl Carrier Protein Synthases I and I1 of Escherichia coli *

fl-KetoacyI-acyl carrier protein synthases 1 and LI of Escherichia coli were purified and characterized. Synthase I was shown to have a molecular weight of 80,000 f 5,000 and to be composed of two similarly sized subunits. Synthase II had a molecular weight of 85,000 f 5,000 and also was apparently homodimeric. Gel electrophoresis of partial proteolytic digests demon- strated that synthases I and II share few if any common peptides. Synthases I and I3 also were shown to be unrelated by immunological criteria. An improved assay for 8-ketoacyl-acyl carrier protein synthase activity gave kinetic parameters for synthases I and I1 at both 27OC and 37°C using five long chain acyl-acyl carrier protein substrates. The properties of synthase 11 are consistent with the proposed role of this enzyme in the modulation of fatty acid synthesis by tempera- ture. fubF mutants of E. coli lack synthase 11. The fabF locus was mapped at min 24.5 of the E. coli genetic map and the clockwise map order was found to be pyrC, fabD, fabe purB.

I The abbreviations used are: ACP, acyl carrier protein; tep, tetracycline-resistant; SDS, sodium dodecyl sulfate. ular weight (9). We have recently shown that fabFmutants of E. coli, which are deficient both in the temperature regulation of fatty acid synthesis and in the elongation of palmitoleic acid to cis-vaccenic acid (IO), lack P-ketoacyl synthase I1 (11). D'Agnolo and co-workers (9) had previously reported that a class of mutants (fabB), deficient in overall unsaturated fatty acid synthesis, lack P-ketoacyl-ACP synthase I. We further demonstrated that the fabB locus is the structural gene for P-ketoacyl-ACP synthase I (11). Since fa@ mutants possess synthase I activity, fabB mutants possess synthase I1 activity, and fabF fabB double mutants lack all fatty acid elongation activity (ll), it was considered likely that synthases I and I1 are distinct enzymes and the products of different structural genes (11). However, the two enzyme forms co-purify through several protein fractionation steps (2) and have similar properties. Thus, it seemed possible that synthase I1 is a modified form of synthase I. This putative modification could be involved in the temperature control of fatty acid composition of the membrane phospholipids of E. coli, since synthase I1 has a key role in temperature control (11). Therefore, we have purified the two synthases and compared their properties.
In this paper, we report conclusive evidence that /3-ketoacyl synthases I and 11 have different primary structures. An improved assay for P-ketoacyl-ACP synthase activity is reported, and the substrate specificities of synthases I and I1 were analyzed at two different temperatures with five long chain acyl-ACP substrates. The relevance of these data to the regulation of fatty acid synthesis i s discussed. The E. coli genetic map location of the fabF gene, the presumptive structurd gene for P-ketoacyl-ACP synthase 11, has also been determined.

Molecular Characterization of Synthase I-We purifed p-
ketoacyl-ACP synthase I by a minor modification of the scheme reported by D'Agnolo et al. (9) and obtained a preparation having a specific activity of 5.5 units/mg of protein.
All synthase I preparations gave a single stained protein band upon polyacrylamide gel electrophoresis in the presence of sodium dodecyl sulfate (SDS) (Fig. 5)  and contained 10% acrylamide cross-linked with 0.27% bisacrylamide. Staining and destaining were done as previously described (54). the molecular weight of synthase I under denaturing conditions are somewhat greater than the values previously reported by Prescott and Vagelos (32). Those workers had obtained molecular weights of 35,000 and 37,000 by SDS-gel electrophoresis and by gel fitration in the presence of guanidine HCl, respectively (32). All of these values are incompatible with the native molecular weight of 66,000 reported by Greenspan and Vagelos (22), and thus, we determined the molecular weight of synthase I by the sedimentation equilibrium method of Bothwell et al. (53). The distribution of synthase I as determined by enzymatic activity gave a molecular weight for the active enzyme of 80,000 f 5,000 (Fig. 6). This value is in good agreement with the average (39,000) of the various determinations of the subunit molecular weight.
The amino acid composition we obtained for synthase I ( Table   3) agrees well with that previously reported by Greenspan and Vagelos (22) for the enzyme from E. coli B.
/3-Ketoacyl-ACP synthase I1 was also purified to homogeneity. Our best preparation had a specific activity of 6.3 units/ mg protein and gave a single protein band on SDS-gels (Fig.   5). The apparent molecular weight of synthase I1 was 44,000 to 45,000, a value slightly (although significantly) larger than that of synthase I. We also determined the molecular weight of the native molecule by sedimentation equilibrium and obtained a value of 85, OOO * 5,000 (Fig. 6). This value indicates that synthase I1 like synthase I is composed of two subunits of similar or identical molecular weights. By gel filtration an apparent molecular weight for synthase I1 of 76,500 was obtained by D'Agnolo et al. (9). The similarity of this value to that obtained by sedimentation equilibrium argues that synthase I1 is a globular protein. The amino acid composition of synthase I1 was similar but not identical to that of synthase I ( Table 3).

Comparison of the Primary Structures of Synthases
Z and ZZ-We tested the relationship between synthases I and I1 by peptide mapping using the method of Cleveland et al. (54). Homogeneous samples of synthases I and I1 were digested with a protease in the presence of SDS. The digestions were run in parallel and the resulting peptides were separated by polyacrylamide gel electrophoresis in the presence of SDS and urea (Fig. 7).
The peptide maps of synthases I and I1 obtained with Staphylococcus V8 protease, chymotrypsin, and papain were strikingly different. Furthermore, s-ynthases I and I1 differed greatly in their sensitivity to both Staphylococcus V8 protease and papain (Fig. 6). Peptide maps of synthases I and I1 cleaved with CNBr also differed markedly, but the Coomassie blue staining was too faint for adequate photographic reproduction (data not shown). We conclude that synthases I and I1 share few if any amino acid sequences.
We have also tested the immunological relationship between synthases I and I1 (Fig. 8). A purified IgG fraction was obtained from the serum of a rabbit injected with homogeneous synthase I. The anti-synthase I IgG preparation gave a readily detectable precipitin line with partially purified /3ketoacyl-ACP synthase I, but no precipitin line was detected when equivalent activities (and thus equivalent masses of protein (9)) of synthase I1 preparations were exposed to the antibody. We conclude that synthases I and I1 have few antigenic determinants in common. Substrate Specificities of Synthases Z a n d ZZ-Extensive data on the substrate specificity of /3-ketoacyl-ACP synthase I have been reported. However, the kinetic constants for the various substrates differ greatly among the various reports. As discussed in more detail in the miniprint section, we attribute this variability to two defects in the assay method: the specificity of the enzyme catalyzing the coupled reaction that allows the activity to be monitored spectrophotometrically; and the chemical preparation of the acyl-ACP substrates. We have developed a radiochemical assay to avoid the first problem (see "Experimental Procedures" in miniprint section) and use enzymatically synthesized acyl-ACP substrates to avoid the artifacts of chemical synthesis.
The relative activities of synthases I and I1 can be greatly altered by the assay conditions used (9), and thus we have normalized our maximal velocity data to that obtained with . The enzyme preparations used were purified by the standard procedure except that the final gel filtration step was omitted.

Kinetic constants of P-ketoacyl-ACP synthases I and II
The Vr,.l values are apparent Vn,ax values expressed relative to the apparent Vmax of tetradecanoyl (C14:O)-ACP at 37OC. The V.,.. and K,, values were obtained from Lineweaver-Burk (44) plots (Fig.   4). The plots had four or five different concentrations of acyl-ACP evenly distributed in inverse substrate concentration over a 4-fold concentration range. The correlation coefficients (least squares regression) for each data set were >0.98. The Vm,ax values with tetradecanoyl-ACP a t 37OC were 2.9 and 1.1 units/mg of protein for synthases I and 11, respectively. The synthase preparations were purified through the hydroxylapatite step from the E. coli K-12 strain, AB2829 pyrC.
Synthase I1 did not follow Michaelis-Menten kinetics with C12:O-ACP (Fig. 4). tetradecanoyl-ACP as the substrate (Table  IV). We chose tetradecanoyl-ACP because it is an excellent substrate for both synthases in vitro and in vivo (as argued from genetic evidence (1 1)).

Acyl-ACP substrate
Both /3-ketoacyl-ACP synthases I and I1 were essentially inactive with cis-vaccenoyl-ACP and palmitoyl-ACP ( 4 % of the activity with C14:O-ACP). This is consistent with the finding that E. coli contains only traces (if any) of the final elongation products (cis-11 eicosenoic acid and stearic acid, respectively). Both synthases funtioned with all of the other substrates tested although there was one striking difference between the two enzymes in that palmitoleoyl-ACP was an excellent substrate for synthase I1 but a poor substrate for fi-Ketoacyl-ACP Synthases of E. coli synthase I (Table IV). This difference was primarily due to the slow rate at which synthase I elongated this substrate, as the Michaelis constants of the two enzymes for palmitoleoyl-ACP were similar.
We have proposed that P-ketoacyl-ACP synthase I1 is intimately involved in the temperature regulation of fatty acid composition in E. coli (ll), and thus we tested the effect of a decreased assay temperature on the kinetic constants of both synthases. As expected, at 27°C with palmitoleoyl-ACP as the substrate, the difference between synthase I and I1 was greater than that found at 37°C (Table IV). Although both enzymes had lower K,,, values for palmitoleoyl-ACP at the lower temperature, the relative velocity of the synthase I1 reaction was disproportionately greater.
Genetic Analysis of a Synthase 11 Mutant-Although fabF mutants lack ,f3-ketoacyl-ACP synthase 11, these strains grow normally (10). However, if a temperature-sensitive fabB mutation (fabBt") is introduced into a fabF strain, these double mutants are unable to grow on media supplemented with oleate at 42°C (11) (fabBt" mutants grow well at 42°C if supplemented with oleate). This growth phenotype was used to locate the fabF locus on the genetic map of E. coli.
Interrupted matings of a fabF, fabBb strain, CY216, with several different fab+ Hfr strains were carried out by the method of Zipkas and Riley (55). These experiments indicated that the fabF gene was located near min 24 of the current genetic map of E. coli (56). Finer mapping was carried out by transduction with phage P1 (Table V). The fabF locus was co-transduced almost equally (22 to 27%) with two markers in this region, p y r e and purB (Table V). The p y r e and purB loci are only a few per cent co-transduced (45,56), and thus, the fabFgene must be located approximately midway between these two genes. Another lesion in fatty acid synthesis, the fabD gene that codes for malonyl transacylase, has been mapped between the pyrC and purB loci by Semple and Silbert (45). We mapped the fabF locus in relation to the fabD locus using phage P1 stocks grown on a fabF strain to transduce a fabD strain to temperature resistance. The phospholipid fatty acid compositions of the fabD+ recombinants were then analyzed by thin layer chromatography. fabF mutants are sufficiently deficient in cis-vaccenate synthesis that this deficiency is readily scored by visual inspection of autoradiograms of the thin layer chromatograms (6,10). These experiments demonstrated that the fabF and fabD genes are tightly (89%) linked ( Table V).
The order of the fabF and fabD genes on the genetic map was determined in relation to the purB locus. fabD mutants have the same growth phenotype as fabF, f a b P strains (45), and since the fabF growth phenotype depends on having a fabBt" lesion in the same strain (ll), elaborate conventional strain construction would have been needed to establish the map order. To simplify the strain construction and analysis, a strain carrying a TnlO transposon integrated very close to the purB locus was used. This strain was isolated by selecting simultaneously for purine-independent and tetracycline-resistant (tep) recombinants of the purB strain, PCO540, with P1 phage grown on a pool of random TnlO insertions (38,39). The TnlO insertion used was >99% linked to the purB locus (Table V).
Strains were constructed carrying the TnlO insertion and either fabD or fabF. P1 phage grown on these strains were used to infect either a fabF or a fabD strain. All recipient and donor strains carried a fabB'" mutation so that the fabF genotype could be scored by its growth phenotype. Equal volumes of each transduction mixture were plated on two plates containing tetracycline. One plate was incubated at 30°C to select for tetracycline resistance (tep) and the other was incubated at 42°C to select for tep, fabD+fabF+. The results of these crosses (Table V) show that if fadD was carried by the donor and fabF by the recipient, 22% of the tetR recombinants were fabD+fabF+, whereas in the reverse cross VabF in the donor), <0.3% of the t e p recombinants were fadD'fabli". The latter result is that expected for a four cross-over class of recombinants whereas the former result is that expected for a two-cross-over class. These data are only consistent with the order fabD, fabF, TnlO. Since the TnlO insertion used was very tightly linked to purB (Table V), the clockwise map order must be pyre, fabD, fabF, purB.

DISCUSSION
P-Ketoacyl-ACP synthases I and I1 of E. coli are two distinct proteins. Synthase I is coded by the fabB gene (11)

TABLE V
Transductional mapping of the fabF gene All strains except WNI, LA2-89, PC0254, and MA1008 also carry then one of the two plates was shifted to 42OC. The 30°C plate gave a fabB' lesion (fabB2I). The two factor crosses were performed and the number of tet' recombinants whereas only tet' fabD+F+ recomscored by standard procedures except Cross 3 in which fabF was binants grew on the 42°C plate. The same procedure was used for scored by fatty acid analysis (see text). In Crosses 8 and 9, equal Crosses 10 and 11. The medium used in Crosses 8 to 11 was R broth volumes of a single transduction mixture were plated on two plates containing Medium E and 10 p g / d of tetracycline-HCl. and is a dimer of molecular weight 80,000 ( Fig. 6) with two similar, probably identical (32) subunits (Fig. 1). Synthase I readily catalyzes all the condensation reactions of long chain fatty acid synthesis except the elongation of palmitoleoyl-ACP (Table I). Previous workers had reported that synthase I has a molecular weight of 66,000 by sedimentation equilibrium (22) whereas the apparent subunit molecular weight was 35,000 to 37,000 (32). Our subunit molecular weight (44,000 to 45,000) was obtained by SDS-polyacrylamide gel electrophoresis on slab gels, a method more reliable than the early version of the technique used previously (32). A larger discrepancy occurs between our value for the native molecular weight, 80,000 (Fig. 6), and the previous value (22) of 66,000.
Although both values were obtained by sedimentation equilibrium, we used the method of Bothwell et ul. (53) and determined the distribution of the protein by enzymatic activity, whereas the previous workers (22) assayed the total protein distribution by ultraviolet scanning. Heterogeneity was evident in the untraviolet scan for protein reported (22). Since ultraviolet scanning is an insensitive assay for heterogeneity (57), considerable heterogeneity may have been present. The conditions used in the sedimentation experiment of Greenspan and Vagelos (22) were later shown (32) to result in structural changes in the protein (probably dissociation into monomers).
It should be noted that our molecular weight estimate for synthase I is compatible with previous sedimentation velocity (22) and gel fitration (9) data and together with these data indicate a globular shape for 6-ketoacyl-ACP synthase I. P-Ketoacyl-ACP synthase I1 has a molecular weight of approximately 85,000 (Fig. 6) and is composed of two similarly sued subunits (Fig. 5). The subunits are probably identical, on the basis of data reported by Prescott and Vagelos (32). Tryptic digestion of synthase I gave a peptide map with 23 strongly staining spots and a similar number of lightly staining spots (32). Because the synthase I was purified by batch elution from hydroxylapatite rather than by gradient elution, contamination with synthase I1 was likely.* The two synthases have similar lysine plus arginine contents (Table 3) and thus it seems probable that the lightly staining peptides were derived from synthase 11. The number of these peptides is only consistent with a homo-dimeric native structure. The simplicity of our partial peptide digestions (Fig. 7) is also consistent with homodimeric structures for both P-ketoacyl-ACP synthases. The similarity of the molecular weight values obtained by sedimentation equilibrium (Fig. 6) to that previously inferred from gel filtration (9) indicates the synthase I1 is a globular protein.
The data presented in this paper further support our hypothesis that P-ketoacyl-ACP synthase I1 plays a major role in the thermal regulation of fatty acid synthesis. Physiological studies indicated that an increase in the rate of cis-vaccenate synthesis is the primary response of fatty acid synthesis to a decrease in temperature (4-6). This change is accomplished by changes in the activity of a pre-existing enzyme(s) (3). fubF mutants, which do not elongate palmitoleate and do not thermoregulate their fatty acid composition, lack synthase 11 (11). Revertants of fubF simultaneously exhibit normalization of fatty acid composition, thermoregulation, and P-ketoacyl-ACP synthase I1 activity (11). We report here that at 37"C, synthase I1 elongates palmitoleoyl-ACP with a relative velocity 12-fold more rapid than synthase I (Table IV). At 27"C, the differential is more than 30-fold. In addition, the apparent K,,, at 27°C is significantly lower than at 37°C. The changes in kinetic parameters for synthase I1 are not only consistent with a major role in temperature regulation. They are also consistent with our finding that changes of intrinsic enzyme activity, J. L. Ganvin, unpublished data. rather than de novo synthesis or enzyme modification, are the basis for the temperature regulation of fatty acid synthesis in E. coli (3). P-Ketoacyl-ACP synthase I is essential for unsaturated fatty acid biosynthesis (9) and thus synthase I catalyzes a reaction in unsaturated fatty acid synthesis that synthase I1 cannot. The identity of this reaction remains unknown. The most probable site for the unique role of synthase I in unsaturated fatty acid synthesis is the elongation of cis-3-decenoyl-ACP. We have argued that synthase I should be very active on this substrate, whereas synthase I1 should be inactive (11). Unfortunately, we have been unable to synthesize significant amounts of cis-3-decenoyl-ACP using either acyl-ACP synthetase (28) or the transacylation activity of synthase I (58). It has been reported that synthases I and 11 both catalyze the elongation of cis-3-decenoyl-ACP samples synthesized by chemical means (9). However, these substrates lack native structure (28, 29) and thus a definitive test of our hypothesis must await the synthesis of native cis-3-decenoyl-ACP.
We have shown that the fubF locus is very closely linked to the fubD locus, the structural gene for malonyl transacylase (Table V). The linkage is sufficiently close that fubF and fubD could be neighboring genes (59) and thus coordinately controlled. If, as seems likely, the fubF locus is the structural gene for P-ketoacyl-ACP synthase 11, coordinate synthesis of malonyl transacylase and synthase I1 may regulate the relative rates of two consecutive steps of fatty acid synthesis.