Nucleotide Sequence of the Escherichia coli r f e Gene Involved in the Synthesis of Enterobacterial Common Antigen MOLECULAR CLONING OF THE rfe-rff GENE CLUSTER*

The genetic determinants of enterobacterial common antigen (ECA) include the rfe and rff genes located between ilv and cya near min 86 on the Escherichia coli chromosome. The rfe-rff gene cluster of E. coli K-12 was cloned in the cosmid pHC79. The cosmid clone complemented mutants defective in the synthesis of ECA due to lesions in the rfe, rffE, rffD, rffA, rffC, r f f T , and rffM genes. Restriction endonuclease mapping combined with complementation studies of the original cosmid clone and six subclones revealed the order of genes in this region to be rfe-rffDlrffE-rffAl rffC-rfff-rffM. The rfe gene was localized to a 2.54-kilobase CZuI fragment of DNA, and the complete nucleotide se- quence of this fragment was determined. The nucleotide sequencing data revealed two open reading frames, ORF-1 and ORF-2, located on the same strand of DNA. The putative initiation codon of ORF-1 was found to be 670 nucleotides downstream from the termination codon of rho. ORF-1 and ORF-2 specify pu- tative proteins of 267 and 348 amino acids with calculated M, values of 29,010 and 39,771,

The early steps involved in synthesis of ECA polysaccharide chains have been determined in considerable detail (8-10). The initial trisaccharide repeat unit is assembled as a lipidlinked intermediate (lipid 111) in a sequential series of reactions that involve participation of the polyisoprenoid lipid, undecaprenyl monophosphate (Fig. 1). Subsequent steps utilize lipid I11 as a substrate for polysaccharide chain elongation (8). However, the mechanism of chain elongation remains to be established.
The genetic determinants of ECA synthesis include the rferff gene cluster located between ilv and uvrD at min 85 of the Escherichia coli chromosome (11,12). The rffE and rffD genes are involved in the synthesis of UDP-ManNAcA, and they code for the enzymes UDP-GlcNAc-2-epimerase and UDP-ManNAc dehydrogenase, respectively (11). The isolation and characterization of TnlO-insertion mutants of E. coli defective in ECA synthesis have recently resulted in the identification of additional genes involved in the ECA biosynthetic pathway ( Fig. 1) (13). The rffA gene codes for the transaminase that catalyzes the conversion of TDP-4-keto-6-deoxy-~-glucose to TDP-D-fucosamine. The rffM and rffT genes specify the transferases responsible for the transfer of ManNAcA from UDP-ManNAcA to lipid I and the transfer of Fuc4NAc from TDP-Fuc4NAc to lipid 11, respectively. The rffC locus appears to be involved in the elongation of ECA chains. In addition, his-linked rfb genes of Salmonella typhimurium have been shown to be required for synthesis of the Fuc4NAc residues of ECA (14). Accordingly, rfbA is the structural gene for TDPglucose pyrophosphorylase and the rfbB gene codes for TDPglucose oxidoreductase.
The rfe gene is essential for ECA synthesis, but the function of rfe has not been determined. The rfe gene is also required for synthesis of lipopolysaccharide 0-side chains in Salmonella montevideo, Salmonella minnesota, and E. coli 08, 09, The present study describes the isolation of a cosmid clone that contains all of the rff and rfe genes known to be involved in ECA synthesis. Restriction enzyme mapping of the cosmid clone and subclones combined with complementation studies has allowed the construction of a partial map of the rfe-rff region. We also report the cloning of the rfe gene on a 2.54kb ClaI fragment of E. coli DNA. An open reading frame within the 2.54-kb fragment was identified that complements defects in the synthesis of both ECA and 08-side chains in rfe::TnlO-insertion mutants of E. coli, and the nucleotide sequence of the putative rfe gene is presented. Data are also presented which suggest the possibility that the rfe gene encodes the UDP-G1cNAc:undecaprenylphosphate GlcNAc-1-phosphate transferase responsible for lipid I synthesis.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Media-Bacterial strains used in this study are listed in Table I DNA Methods-The DNA methods used are those described by Sambrook et ai. (19). These include the isolation of both chromosomal and plasmid DNA and the analysis of restriction endonucleasetreated plasmid DNA by agarose gel electrophoresis. Restriction endonuclease digestions and DNA ligations were carried out as recommended by the manufacturer.
Construction of Cosmid Clone pCAll and Subclones-The cosmid cloning of E. coli DNA was carried out using total chromosomal DNA from E. coli strain AN2618 that had been partially cut with the enzyme Sau3A. DNA fragments (20 pg) within the 34-49-kb size range were ligated to the cosmid vector pHC79 (10 pg) following digestion of the vector with BamHI and treatment with alkaline phosphatase. A portion of the ligated DNA (7.5 pg) was packaged BHB2690 and BHB2688 (19). The ECA-negative mutant strains into X-phage heads using a packaging extract prepared from strains 21546, 21548, and 21550 were each transduced with 1.5 pg of the packaged DNA and ampicillin-resistant transductants were selected. ECA synthesis was rescued in one of the transductants derived from strain 21546 as determined by passive hemagglutination and immunoblot assays; cosmid pCAll was isolated from this clone. Subclones of pCAll were constructed, and they were used to transform ECA-negative strains possessing various rjf and rje mutations. All of the transformants were selected as ampicillin-resistant colonies due to the presence of the ampicillin resistance determinant in the respective plasmid vector sequences. The ability of subclones to complement the various rjj and rje mutations was determined by screening the transformants for their ability to synthesize ECA as determined by colony immunoblot and passive hemagglutination assays. Plasmid pCA25 contains the entire pHC79 vector sequence, and it was obtained by digestion of pCA11 with XhoI followed by religation. Plasmid pCA81 contains only a portion of the pHC79 vector sequence, and it was contructed by digestion of pCAll with BglII followed by religation. Plasmid pCA32 was obtained by digestion of pCAll with HindIII followed by ligation of the resulting HindIII fragments into the single HindIII site of pBR322. Plasmid pCA53 was constructed by digestion of pCAll with Cia1 and subsequent ligation of the ClaI fragments into the single ClaI site of pBR322. Plasmid pCA62 was constructed by digestion of pCA32 with XhoI and HindIII followed by ligation of the resulting 4.9-kb fragment into the Sal1 and HindIII sites of pBR322. Plasmid pRL105 was obtained by digestion of pCA25 with BamHI and XhoI followed by ligation of the resulting 7.5-kb fragment into the corresponding sites in the polylinker region of pBluescript KS.
Strategy of DNA Sequencing-The nucleotide sequence of the 2.54kb ClaI fragment containing the rfe gene was obtained by the dideoxy chain termination method of Sanger et al. (20) using Sequenase. Double-stranded plasmid templates included the intact ClaI fragment in pBluescript SK and pBluescript KS as well as various subclones obtained by restriction fragment subcloning into pBluescript SK. Sequence data were obtained using either T3 or T7 primers or synthetic oligonucleotides as primers. The sequences of both strands of the 2.54-kb ClaI fragment were determined. All sequence data were obtained by the United States Biochemical Corporation Custom Sequencing Service.
Genetic Procedures and the Determination of ECA and 08-Antigen-The procedure used for P1-transduction has been described previously (12). Transformation was carried out as described by Maniatis et al. (17). The occurrence of nonradiolabeled ECA was determined by colony immunoblot (21) and passive hemagglutination (12) assays. The presence of Ob-side chains was determined by sensitivity to phage Q8 (12).
In Vivo Incorporation of fH]GlcNAc into ECA and Analysis of Radioactive ECA by SDS-PAGE and Fluorography-The in vivo incorporation of exogenously supplied [3H]GlcNAc into ECA in the presence of various concentrations of tunicamycin was analyzed by SDS-PAGE and fluorography as described previously (10). Briefly, cultures of strain CAlO and transformants of strain CAlO containing pRLlOO were grown with vigorous aeration at 37 "C in medium A containing various concentrations of tunicamycin as indicated (Fig.  5). Cultures were grown to an absorbance (600 nm) of 0.6. [3H] GlcNAc was then added to portions of each culture (10 ml) and shaking was continued for an additional 30 min. Labeling was terminated by pouring the cultures into a flask containing 10 g of crushed ice and 1 ml each of 0.2% chloramphenicol, 2% sodium azide, and 0.1 M 2,4-dinitrophenol. The cells were then harvested by centrifugation (10 min, 12,000 X g, 4 "C), and cell envelopes were isolated as previously described (9). The cell envelopes were subsequently analyzed by SDS-PAGE and fluorography (9). Electrophoresis was carried out using 12% gels according to Laemmli (22).
Assay Procedure for the in Vitro Synthesis of Lipid I-Assay procedures for the in vitro synthesis of [3H]GlcNAc-labeled lipid I were as previously described (9). The amount of [3H]GlcNAc-labeled lipid I in reaction mixtures was determined by ascending paper chromatography using Whatman SG-81 paper. SG-81 papers were pretreated by dipping them in EDTA (0.068 M, pH 7.7) and allowing them to dry at room temperature. The EDTA-treated papers were incubated at 110 "C for 10 min immediately prior to use. Chromatograms were developed with chloroform/methanol/water (65:25:4, v/ v/v). Protein was determined by the method of Lowry et al. (23).
Computer Methods-Hydrophilicity and secondary structure analyses were analyzed according to the methods Kyte and Doolittle (24) and Chou and Fasman (25), respectively, using the Protylze Protein Structure Predictor software package (Scientific and Educational Software). Primary sequence comparisons were analyzed by the Bestfit program from the Genetics Computer Group sequence analysis package using a gap weight of 3.000 and a length weight of 0.1000.

Isolation of a Cosmid Clone Containing the E. coli rfe-rff
Gene Cluster-E.
coli strains 21546 (rffC:TnIO), 21548 (rfe::TnIO), and 21550 (rffA::TnIO) were used as recipients in transduction with an E. coli cosmid bank prepared with vector pHC79. A total of 658 ampicillin-resistant clones were screened for their ability to synthesize ECA using colony immunoblot and passive hemagglutination assays. Of these, 177 were derived from strain 21546, 190 were derived from strain 21548, and 291 were derived from strain 21550. One of the clones derived from strain 21546 was found to give a strong positive reaction for the presence of ECA as determined by both of the above assays. The cosmid pCAll was isolated from this clone, and it was found to complement the defects in ECA synthesis in all of the ECA-negative TnIO-insertion mutants (rfe and rffA, C,D, E,M) derived from strain AB1133 (Table I)   b2 red Eam Sam/h)) uncC543 argH pyrE entA 08:K27-his pro pyr met rha as AB1133, but rffD::TnlO as AB1133, but rfe::TnlO-48 as AB1133, but rffA::TnlO as AB1133, but rffE::TnlO as AB1133, but rffA::TnlO as AB1133, but rffM::TnlO as AB1133, but rffC::TnlO as AB1133, but rfe::TnIO-88 as AB1133, but rfe::Tnl0-91 as AB1133, but rfe::TnIO-92 as AB1133, but rffA::TnlO as AB1133, but rffC::TnlO as 2443, but rfe::TnlO-48 as 2443, but rfe::TnlO-88 as 2443, but rfe::Tnl0-91 as 2443, but rfe::Tn10-92 as AB1133, but zie2::TnlO Prior to determining the function of the rffT gene, this locus was originally defined by a mutation designated rff-726 (12). Subclones Containing rfe and rff Genes-Subclones of pCAll were constructed as described under "Experimental Procedures," and they were used to transform ECA-negative strains possessing various rff and rfe mutations. Partial restriction maps of the individual subclones are shown in Fig.  2, and their complementation profiles are presented in Table  11. Plasmid pCA25 complemented mutations in all of the rff genes except those in the rffE and rffD loci. However, none of the mutations in rfe genes were complemented by pCA25. Only pCA53, which contained a 2.54-kb ClaI fragment, was able to complement mutations in the rfe genes. Indeed, pCA53 was able to complement the defects in all of the rfe mutants, but it was unable to complement any of the mutations in rff genes. Plasmid pCA81 only complemented the mutation in the rffM gene whereas pCA32 complemented mutations in the rffA, rffC, rffE, and rffD genes. Neither of these subclones were able to complement the mutation in the rffT gene even though this defect was complemented by pCA25. However, the mutation in the rffT gene was complemented by pRL105 which possesses a 7.5-kb BamHI-XhoI fragment that includes the region of chromosomal DNA located between the chromosomal fragments in pCA81 and pCA32. Plasmid pRL105 also complemented mutations in rffM, rffA, and rffC genes.   Complementation of 08-Side C h i n Synthesis in rfe::TnlO-Insertion Mutants by pCA53-E.
coli strains 21608 (rfbos rfe::TnlO-48) and 21609 (rfbos rfe::Tn10-88) are unable to synthesize either ECA or 08-side chains due to the rfe defect. Both strains were transformed with plasmid pCA53, and ampicillin-resistant transformants were subsequently analyzed for their ability to synthesize ECA and 08-side chains. ECA synthesis was determined by colony immunoblot and passive hemagglutination assays, and 08-side chain synthesis was evaluated by determining the sensitivity of the transformants to the 08-specific phage, 08. Plasmid pCA53 complemented the defects in both ECA and 08-side chain synthesis in transformants derived from both strains (data not shown). Identification and Nucleotide Sequence of the rfe Gene-The nucleotide sequence of the 2.54-kb ClaI fragment of pCA53 was determined as an initial step toward the identification of the rfe gene(s) responsible for both the synthesis of ECA and OB-side chain synthesis. A partial restriction map of the fragment, and diagrams of the extent and direction of nucleotide sequencing are shown in Fig. 3. The entire nucleotide sequence of the 2.54-kb fragment was determined in both directions, and the sequence is shown in Fig. 4. The sequence data revealed two open reading frames, ORF-1 and ORF-2, located on the same strand. The termination codon of ORF-1 and the putative initiation codon for ORF-2 are separated by only 11 nucleotides. The sequence data also revealed that the two open reading frames do not occur in the same translational reading frame. Open reading frames 1 and 2 specify proteins of 257 and 348 amino acids with calculated M, values of 29,010 and 39,771, respectively; the putative initiation codon of ORF-1 is located 570 nucleotides downstream from the termination codon of rho (26).
Experiments were carried out to identify which of the potential open reading frames corresponded to the rfe gene. The 2.54-kb ClaI fragment possesses a unique XmaIII site located 37 nucleotides downstream from the termination codon of ORF-1 and 25 nucleotides downstream from the initiation codon of ORF-2. Accordingly, a 1.46-kb ClaI-XmaIII fragment containing ORF-1 was subcloned into compatible sites of pBR322. The resulting construct, designated pRL100, was used to transform various rfe mutants of E. coli, and the ability of the transformants to synthesize ECA and 08-side chains was determined (Table 111). E. coli strains 21548, 21588, 21591, and 21592 are rfe::TnlO-insertion mutants derived from the E. coli K-12 strain AB1133 (13). These mutants are only defective in ECA synthesis since they lack the rfb genes that specify 08-side chains. The defect in ECA synthesis in these mutants was complemented by the 1.46-kb ClaI-XmaIII fragment as determined by passive hemagglutination were deduced from nucleotide sequencing data (see Fig. 4). The stippled bar designates the 3' terminus of the rho gene. The solid arrows indicate the direction and lengths of the regions sequenced. The complete overlapping sequence was determined by the dideoxy chain termination method of Sanger et al. (20)  assay. Strains 21608, 21609, 21610, and 21611 are rfe::TnlOinsertion mutants derived from E. coli 2443; strain 2443 is an E. coli K-12 strain that possesses the rfb genes of E. coli 08 (21). Accordingly, the rfe lesion in these mutants renders them defective in the synthesis of both ECA and 08-side chains. However, following the transformation of these mutants with pRL100, their ability to synthesize both ECA and 08-side chains was restored as determined by passive hemagglutination assay and sensitivity to phage 8, respectively. The rfe-dependent synthesis of both ECA and 08-side chains was also restored in transformants of E. coli F1469 (rfe) which belongs to the 08-serogroup (12) ( Table 111). Transformation of the above mutants with pBR322 alone failed to complement the defects in either ECA or OS-side chain synthesis (data not shown). These data strongly support the conclusion that ORF-1 specifies the rfe gene.
Relationship between the rfe Gene and UDP-GlcNAc: Undecaprenylphosphate GkNAc-1-phosphate Transferase Activity-Previous investigations revealed that ECA-negative rfe::TnlO-insertion mutants of E. coli are defective in lipid I synthesis (13). These observations suggest the possibility that the rfe gene specifies the UDP-GlcNAc: undecaprenylphosphate GlcNAc-1-phosphate transferase responsible for lipid I synthesis. Accordingly, experiments were carried out to determine if the 1.46-kb ClaI-XmaIII fragment containing ORF-1 restored the ability of rfe::TnlO-insertion mutants to synthesize lipid I in vitro.
Significant synthesis of radioactive lipid I was observed when cell envelope membranes obtained from rfe::TnlO-insertion mutants transformed with pRLlOO were incubated with UDP-[3H]GlcNAc (Table IV). Indeed, the amount of in vitro lipid I synthesis using cell envelope membranes from the transformants was markedly elevated in comparison to that observed using membranes from the parental strain, E. coli AB1133. Essentially no radioactive lipid I synthesis was detected when reaction mixtures contained membranes obtained from the rfe::TnlO-insertion mutants transformed with the vector pBR322.
Sensitivity of ECA Synthesis to Tunicamycin in rfe Mutants Transformed with pRL100"The above data suggest that ORF-1 is the structural gene for the UDP-G1cNAc:undecapre n y 1 p h o s p h a t e GlcNAc-1-phosphate transferase. Thus, additional experiments were conducted to examine this possibility. The synthesis of ECA is inhibited by the antibiotic tunicamycin due to the sensitivity of the GlcNAc-1-phosphate transferase to the drug (9, 10). Therefore, synthesis of ECA by cells harboring a cloned gene encoding the GlcNAc-1phosphate transferase should exhibit increased resistance to tunicamycin in comparison to wild-type cells. Accordingly, the in vivo synthesis of ECA was abolished when tunicamycin was added to cultures to give a final concentration of 3 pg/ml (Fig. 5A). These results are in agreement with previous observations regarding the in vivo sensitivity of ECA synthesis to tunicamycin (10). In contrast, the in vivo synthesis of ECA by transformants that possessed pRLlOO was unaffected by tunicamycin up to a final concentration of 20 pg/ml which was the highest concentration of the drug administered (Fig.  5 B ) .
It is interesting to note that apparent degrees of polymerization of ECA chains synthesized by the nontransformed strain were increased at very low concentrations of tunicamycin where inhibition was incomplete (Fig. 5A). However, the significance of this observation is not understood.
Structure of the Putative Rfe Protein-The nucleotide sequence of ORF-1 indicates the Rfe protein has a predicted M , of 29,010. Based on these data, a hydropathic plot of Rfe was constructed according to the methods of Kyte and Doolittle (24) (Fig. 6). The hydropathic profile of Rfe is characterized by several alternating hydrophobic-hydrophilic domains. Indeed, the Rfe protein was found to have a mean hydropathy index of +0.94 indicating that the overall hydrophobicity of the protein is very high. The distribution of charged amino acids among hydrophilic domains is generally such that hydrophobic domains are flanked by hydrophilic domains possessing net negative and net positive charges. This charge distribution is characteristic of intergral membrane proteins in which positively and negatively charged hydrophilic domains are localized on the cytoplasmic face and periplasmic face of the membrane, respectively; adjacent hydrophobic domains function as transmembrane regions.
The secondary structure of Rfe was predicted according to Chou and Fasman (25). These analyses indicated that Rfe contains relatively few a-helical regions; but rather, it is comprised primarily of @-sheet and @-turn regions. Regions of /3-sheet secondary structure were particularly associated with the putative transmembrane domains. More detailed studies are clearly required in order to adequately define the topology and structure of Rfe.
A computer search of nucleotide and amino acid sequences contained in both Genbank and EMBL failed to identify any genes with significant homology to rfe. Nevertheless, a bestfit comparison of the amino acid sequence of Rfe with the UDP-G1cNAc:dolichylphosphate GlcNAc-1-phosphate transferases of yeast (27) and Chinese hamster ovary cells (28,29) revealed overall general similarities (data not shown). Indeed, the percents similarity and identity of the Rfe protein with the primary structure of the GlcNAc-1-phosphate transferase from Chinese hamster ovary cells were found to be 57.5 and 23.8%, respectively. However, comparisons with the yeast and Chinese hamster ovary cell enzymes did not reveal any conserved regions with high peptide homology.

DISCUSSION
Analyses of cloned DNA fragments containing rff and rfe genes by both restriction mapping and complementation studies have allowed the construction of a partial map of the rferff region of E. coli K-12 (Fig. 7). The order of the genes in this region proceeding from ilv toward cya was found to be  insertions in these strains have been characterized as mutaof DNA. The data presented here clearly indicate that the rfe tions that prevent elongation of ECA heteropolysaccharide gene is specified by ORF-1 since ORF-1 alone was able to chains (13). However, the mechanism of chain elongation, as complement defects in ECA synthesis and 08-side chain syn-   thesis in rfe mutants of E. coli. In addition, the available evidence suggests that the rfe gene may be the structural gene for the UDP-G1cNAc:undecaprenylphosphate GlcNAc-lphosphate transferase responsible for lipid I synthesis. Thus, the synthesis of lipid I was restored in rfe mutants following their transformation with a pBR322 derivative containing only ORF-1 (pRL100). In addition, the transformants possessed increased amounts of the GlcNAc-1-phosphate transferase in comparison with wild-type parental strains as indicated by (i) significantly elevated levels of GlcNAc-l-phosphate transferase activity in membrane preparations, and (ii) a pronounced decrease in the sensitivity of ECA synthesis to tunicamycin in uiuo. These observations most likely reflect the occurrence of multiple copies of the structural gene for the GlcNAc-1-phosphate transferase in the transformants due to the copy number of the pBR322 derivative, pRLlOO (32).

ATG ATG GTG TTC GGC MG CTT TAT CTC AGT AGC CTG OGT TAT
The primary sequence of the Rfe protein was found to be somewhat similar to that of the UDP-G1cNAc:dolichylphosphate GlcNAc-1-phosphate transferases found in yeast cells (27) and Chinese hamster ovary cells (28,29). However, Rfe does not contain a sequence that conforms to the putative consensus sequence for dolichol recognition found in these enzymes and other enzymes that catalyze the transfer of sugars from nucleotide sugar donors to dolichol phosphate derivatives (33). The dolichols possess a saturated a-isoprene unit, and it appears that this is a critical structural feature that is recognized by sugar transferases that utilize dolichol phosphates as substrates (33). In contrast, undecaprenols possess an unsaturated a-isoprene unit. Accordingly, the lack of a sequence in Rfe that conforms to the putative consensus sequence for dolichol recognition may reflect this difference. The possibility also exists that the rfe gene is not the structural gene for the GlcNAc-1-phosphate transferase, but rather that it encodes a protein which affects activity of the transferase in some as yet unrecognized manner. Additional studies are clearly required to unequivocally establish the Rfe protein as the GlcNAc-1-phosphate transferase.
The expression of rfe in mutants transformed with pRLlOO clearly indicates the presence of promoter sequences of the cloned fragment located upstream from the putative initiation codon. However, these sequences have not yet been defined. In addition, the function of ORF-2 has not been established. Indeed, it is possible that its function is unrelated to either the synthesis of ECA or the assembly of ECA into the outer membrane.
The role of Rfe in 0-side chain synthesis is not explained by the tentative identification of this protein as the GlcNAc-1-phosphate transferase involved in lipid I synthesis. The repeat units of rfe-dependent 0-side chains either lack GlcNAc completely, or they contain GlcNAc in a position that does not suggest the involvement of lipid I in their assembly. For example, the 0-side chains of E. coli 08 and 09 are reported to be homopolymers of mannose that are comprised of a-(1+2) and a-(1+3)-linked mannosyl residues (34). The repeating unit of the 08-side chain is +Q)-a-D-Man-(1 +-2)-a-D-Man-(1+2)-a-D-Man-(l+ (37) whereas the repeating unit of the 09-side chain is +3)-a-~-Man-(1+2)-a-

~-Man-(1+2)-a-D-Man-(1+2)-a-D-Man-(1+3)-a-D-Man-
(1+ (38). The 0-side chain of E. coli 020 also lacks GlcNAc, and this polymer is comprised of the disaccharide repeat unit +4)-a-~-Ga1,-(1+2)-a-~-Ribr(l+ (34). The 0-side chain of S. minnesota is comprised of a pentasaccharide repeat unit that contains GlcNAc, GalNAc, and galactose. However, the GlcNAc occurs as a branch residue (35). Accordingly, it is possible that lipid I functions in the synthesis of the above 0-side chains in a previously unrecognized manner. Alternatively, since the function of the Rfe protein as the GlcNAc 1phosphate has not yet been unequivocally established, the possibility still exists that Rfe possesses a hitherto unrecognized function that affects the synthesis of both ECA and rfedependent 0-side chains.