Sequence and Inactivation of the pss Gene of Escherichia coli

Vol. 266, No. 8, Issue of March 15, pp. 5323-5332,1991 Printed in U S .A.

THEJOURNAL OF BIOLOGICAL CHEMISTRY

0 1991 by The American Society for Biochemistry and Molecular Biology, Inc.

Sequence and Inactivationof the pss Gene of Escherichia coli PHOSPHATIDYLETHANOLAMINE MAY NOTBEESSENTIAL

FOR CELL VIABILITY* (Received for publication, September 19, 1990)

Anne DeChavigny, Philip N. Heacock, and William DowhanS From the Department of Biochemistry and Molecular Biology, University of Texas Medical School, Houston, Texas 77225

Phosphatidylethanolamine is the only zwitterionic phospholipid in Escherichia coli and accounts for 7080%of the total glycerophospholipidsof this organism. To investigate the function of phosphatidylethanolamine in E. coli, we constructed an inactivated allele (pss93::kan)of the gene encoding the phosphatidylserine synthase which catalyzes the committed step to the synthesis of phosphatidylethanolamine. Growth of this mutant was dependent on a plasmid-borne copy of the wild type gene. After curing the mutant of the wild type gene, growth stopped when the content of phosphatidylethanolamine reached 30%of the total phospholipid. Divalent metal ions at millimolar concentrations suppressed the growth phenotype of the mutant in the following order of efficiency: Ca2+ > Mg2+ > Sr2+. Although phosphatidylserine synthase activity was not detectable, phosphatidylethanolaminewas still present at 0.007% of the total phospholipid after growth for many generations in rich medium containing 20 mM Mgz+. The remainder of the phospholipid was primarily phosphatidylglycerol and cardiolipin with no other unique phosphate-containing chloroform-soluble material present. The phospholipid to protein ratio and the fatty acid composition were very similar to the parental strain. The broad divalent metal ion auxotrophy brought about bythe lack of phosphatidylethanolamine suggests a primarily structural role for this phospholipid in E. coli.

thesis. For instance, phosphatidylglycerol supplies components of the membrane-derived oligosaccharides (Jackson and Kennedy, 1983) and several outer membranelipoproteins (Yu et al., 1986; Wu, 1987; Takase et al., 1987). Parts of phosphatidylethanolamine end up in lipopolysaccharide (Hasin and Kennedy, 1982), membrane-derived oligosaccharides (Miller and Kennedy, 1987),and the major outer membrane lipoprotein (Jackowski and Rock, 1986). It is not clear whether all of these precursor functions are absolutely essential for cell viability, but these functions are important to normal cell physiology. Genetic and biochemical evidence has establishedthe major biosynthetic pathway and the gene products responsible for the synthesis of the phospholipids of E . coli (Raetz and Dowhan, 1990). Some information is available as to which phospholipids are absolutely essential, but themolecular basis for their requirementhas not been defined. Many of the existing point mutationsin genes encoding phospholipid biosynthetic enzymes are leaky or conditions can be found which partially relieve the growth phenotype of these mutants. Since an interruptedallele of the pgsA gene (encoding the phosphatidylglycerophosphate synthase) cannot support growth, phosphatidylglycerol is necessary to E. coli (Heacock and Dowhan, 1987); this phospholipid plays a crucial role in the movement of some proteins through the inner membrane (de Vrije et al., 1988; Lill et al., 1990) and possibly in the initiation of DNA replication (Yung and Kornberg, 1988). Although the cls gene product (cardiolipin synthase) is not required (Nishijima et al., 19881, cardiolipin may yet prove to be essential There is general agreement that phospholipids are essential since it can still be detected in cells carrying an interrupted for cell function, are responsible for membrane integrity, and allele of the cls gene. Both the pss and psd gene products, serve as thematrix for membrane-associated activities. What which are required for phosphatidylethanolamine biosynis less clear is the function of a particular phospholipid or thesis, are essential under normal growth conditions (Ohta class of phospholipids. Some distinct functions have been and Shibuya, 1977; Raetz et al., 1979; Hawrot and Kennedy, assigned to some of the minor phospholipids such as phos- 1978). However, the growth phenotype of existing mutations phatidylinositol (Rando, 1988) and platelet activating factor in both genes can be partially suppressed by addition of (Snyder, 1987). In Escherichia coli phospholipids are both the magnesium ions to the growth medium. Under these condiend point in metabolism, supplying the components of the tions the cells still contain significant levels of phosphatidylmembrane matrix, and precursors, supplying components for ethanolamine either because the mutations are “leaky” or several metabolic pathways involved in macromolecular syn- there is an alternate pathway for the synthesis of phosphatidylethanolamine. Is the lethal event under normal growth * This work was supported by United States Public Health Service conditions the lack of phosphatidylethanolamine itself, the Grant GM20487 from the National Institutes of General Medical increase in the proportion of acidic phospholipids, or the Sciences. The costs of publication of this article were defrayed in part decrease in membrane stability? Can cells survive and grow by the payment of page charges. This article musttherefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 without phosphatidylethanolamine? These questions have, as yet, not been rigorously answered. solely to indicate this fact. The nucleotide sequencefs) reported in thispaper has been submitted In this article we report the construction of an interrupted to the GenBankTM/EMBLDataBankwith accession number(s) allele of the pss gene of E. coli. The pss gene product (phosM58699. $ To whom all correspondence should be addressed Dept. of Bio- phatidylserine synthase) catalyzes the committed stepto biosynthesis from precursors chemistry and Molecular Biology, University of Texas Medical phosphatidylethanolamine which should not accumulate to high levels in the absence of School, P. 0.Box20708, Houston, T X 77225.Tel.:713-792-5600; Fax: 713-794-4150. this enzymatic activity. Using strains carrying thisinter-

5323

5324

Role of Phosphatidylethanolamine in E.coli TABLEI

rupted allele we have investigated the nature of the requirement of E. coli for zwitterionic phospholipid. EXPERIMENTALPROCEDURES

Plasmids and strains strain

Or

Relevant characteristics"

Source or reference

Materials-All chemicals were reagent gradeor better. RadiochemPlasmids icals and scintillationsupplies were obtained from Amersham/Searle Ohta et al., 1981 pPS4017 pss+tet amp Bethesda Research and Du Pont-New EnglandNuclear. Restriction endonucleases, DNA M13mpll 1aCI'OPZ' Laboratories molecular weight standards, T4 DNA ligase, and T4DNA polymerase This work were purchased from International Biotechnologies, Inc., Bethesda mPSlOOO pss+(derivative of M13mpll) Research Laboratories, and New England Biolabs,Inc. The BCA This work mPS2000 pss+ (derivative of protein assay kit was purchased from Pierce Chemical Co. Bacterial M13mpll) growth media were obtainedfrom Difco, andGeneScreen Plus" transfer membranes were supplied by Schleicher & Schuell. AMPpMTlOl cam tet (temperature sensi- P. Schimmel SCREEN" was purchasedfromBethesdaResearchLaboratories. tive for replication) Ohta et al., 1981 Precoated silica gel thin-layer chromatography plates were obtained pBR322 amp kan This work pss+cam (derivative of from E. Merck. The Genius@DNA detection kit was purchased from pDD72 Boehringer Mannheim. The Sequenasem DNA sequencing kit was pMT101) This work pss+tet (derivative of obtained from United States Biochemical Corp. Short oligonucleopDD88 pBR322) tides used as primers for sequencing were synthesized by Genetic pss93::kun tet (derivative of This work Designs, Inc. The kanamycin resistance gene (1.3 kb)' derived from pADlOl pDD88) Tn903, which is flanked by endonuclease restriction sites including E. coli PstI, was purchased from Pharmacia. Bacterial fatty acid methyl Raetz et al., 1979 pss21 thi RA2021 ester standards were purchased from Matreya, Inc. G. Weinstock Growth of Bacteria-Rich medium forgrowth of cells in either G122 pol At.? F+ nadB7 supE CGSCb liquid culture or on agar plates was LB medium (10 g of BactoW3899 F- X- IN(rrnD-rrnE)l CGSCb tryptone, 5 g of Bacto-yeast extract, and 5 g of NaCl/liter). Media W3110 E. P. Kennedy mdoA::TnlOpyrC were supplemented as indicated with ampicillin (100 pg/ml), tetraNFB2OO G. Weinstock recA sr1::TnlO GE2046 cycline (20 pg/ml), chloramphenicol (20 pg/ml)or, kanamycin (50 This work pss93::kan nadB+ (P1 depg/ml) to selectforcells carrying these drug markers. Sensitivity AD90' rivative of W3899) (recA) or resistance (recA+) to UV light was determined by survival This work pss93::kan recA sr1::TnlO AD93' after irradiation at 0.6 Joule/s m2 as a function of time as previously nadB+ (P1 derivative of described (Sharma etal., 1982). Ampicillin-sensitive colonies on agar AD90) were distinguished from resistant coloniesby the secretion of ppss93::kan (P1 derivative of This work AH930' lactamase from the latter colonies detected using AMPSCREEN9 W3110) Nucleic Acid Manipulations-Methods for DNA isolation, restricThis work pss93::kan mdoA::TnlO ADN1' tion enzymedigestion,DNAligation, andseparatingDNAusing (P1 derivative of W3110) agarose gel electrophoresis have been described previously (Gopalakrishnan et al., 1986; Maniatis etal., 1982; Davis etal., 1986). Bacterial a Partial list of genetic markers. transformations were carried out as described by Chung etal., (1989). Coli Genetic Stock Center. Southern hybridization analysis (Maniatis et al., 1982) wasperformed Dependent on pDD72 for growth. on DNA fragments which had been treated with base prior to transfer from agarose gels to GeneScreen Plus"; transfer was effected using buffer flow from the gel to GeneScreen Plus@ as suggested by the manufacturer. DNA probes for Southern hybridization analysiswere labeled by randomprimedincorporation of nucleotide analogues containing the steroid haptendigoxigenin. The probes were detected by a hapten-specific antibody conjugated with alkaline phosphatase using theGenius@DNA detection kitaccording to the manufacturer's suggestions. The sequence of the DNA fragment carrying thepss gene was determined using the two single-stranded templates, mPSlOOO and mPS2000. The sequencing was done using the dideoxy chain termination method (Sanger etal., 1977) with the aidof the Sequenase@kitas describedby the manufacturer; DNA sequencing was initiated using either the universal primer or synthetic oligonucleotide Hlnd Ill primers as necessary. Structure of Plasmids-The genetic characteristics of all the plasmids used in thiswork are summarized in TableI; Fig.1 diagrams the structures of these plasmids. The E. coli DNA used in this work was derived from a 3.2-kb ClaI fragment originally carried in plasmid pPS4015 (Ohta et al., 1981) and subcloned into the AccI site of the multicloning region of plasmid M13mpll toyield plasmids mPSlOOO and mPS2000. Plasmid pMTlOl contains a temperature-sensitive replicon and was derived from plasmid pSClOl (9.8 kb); therefore, this plasmid can be cured fromcells by growtha t 43 "C (Heacock and Dowhan, 1987) and is compatible with plasmids dependent on the ColEl replicon for replication (Cabello et al., 1976). Plasmid pDD72 pDDW wasconstructed by ligation of a HindIII-BamHIfragmentfrom PAD101 6.6 kb 7 8 kb mPS2000 between the Hind111 and the BamHI sites of the tetracycline resistance gene of plasmid pMT101; cells containing plasmid pDD72 were selected by their ability to grow on chloramphenicol a t 30 "C and screened for theirsensitivitytotetracycline.Plasmid PStl pDD72 directed the expected overproduction of phosphatidylserine synthase activity, contained the predicted insert into the tetracycline FIG.1. Structures of the plasmids used in this work. The gene, and allowed growth of cells on chloramphenicol a t 30 "C but solid regions in plasmids mPSlOOO and mPS2000 denote the DNA derived from E. coli. For the remaining plasmids the solid regions The abbreviations used are: kb, kilobase pair; MOPS, 4-morpho- denote the domainof the pss gene. The arrows indicate thedirection of transcription of the respective genes. linepropanesulfonic acid.

Role of Phosphatidylethanolamine in E,coli not at 43 "C. Therefore, this plasmid could be used as the source of functional pss gene product a t 30 "C but not at43 "C. Plasmid pDD88 was made by inserting an EcoRI-PstIfragment from plasmid mPSlOOO between the same sites of plasmid pBR322 thus inactivating the ampicillin resistance gene. This plasmid, when transformed into strain RA2021 (see Table I), corrected the temperature-sensitive growth phenotype of this strain (due to the pss2lallele) and resulted in the overproduction of phosphatidylserine synthase activity. Finally, the pss gene of plasmid pDD88 was inactivated by insertion of a kanamycin resistance gene cassette ( k n )into the MluI site of the pss gene (see Figs. 2 and 3) using the aid of synthetic oligonucleotide linkers designed to join the PstI sites at the ends of the insert with the MluI sites of the open vector resulting in plasmid pAD101. This plasmid could not correct the temperature-sensitive growth phenotype of strain RA2021 and did not result in overproduction of the enzymatic activity. Since this plasmid has the ColEl origin of replication, it requires a functional pol4 gene product for autonomous replication (Greener and Hill, 1980; Russel and Model, 1984). Construction of Bacterial Strains-Table I summarizes the genetic characteristics and sources of the bacterial strains used in this work. For strains constructed using phage Pluir transduction (Miller, 1972) only the characteristics known to be transferred to the recipient strains are shown. The construction of strain AD93 (containing a pss gene interrupted by the kanamycin resistance marker) was carried out by a modification of a procedure used to inactivate the pgsA gene of E. coli (Heacock and Dowhan, 1987). Single colonies of strain G122 (polA,.) containing plasmid pADlOl were isolated on agar plates in the presence of kanamycin at 43 "C in order to obtain cells in which the plasmid hadintegratedinto the chromosome by homologous recombination at thepss locus; such a strain would have both the pss andpss::kn genes in close proximity in the chromosome. A phage P1 lysate was made from a culture of a pool of the above kanamycinresistant cells grown at 30 "C and used to replace the chromosomal copy of the pss gene of strain W3899/pDD72 with an interrupted copy (pss::kan) of the gene. Strain W3899 is po!A+ (which prevents plasmid pADlOl from being stably maintained in the chromosome) and contains a d B - marker closely linked (Raetz, 1976) to thepss locus; selection for both kanamycin resistance and d B + was used to obtain those derivatives which carried a chromosomal copy of the interruptedpss gene rather than only plasmid pAD101. Plasmid pDD72 assured the presence of a functional copy of the pss gene in the transductants. A single PI lysate was made from a pooled culture of 16 kanamycin-resistant d B + colonies and used to transduce strain W3899/pDD72 once again. Of 442 kanamycin-resistant colonies (carrying the interrupted pss gene), 5 were tetracycline-sensitive (missing the vector portion of plasmid pAD101); four of the five colonies were nadB+ (transfer occurred by integration into the chromosome) and one was nadB-. To minimize loss of the interrupted gene through recombination with the functional gene on plasmid pDD72, one of the mdB+ strains (AD90)was made recA (strain AD93) by transduction to tetracycline resistant using a P1 lysate of strain GE2046 (recAand sr1::TnlO are closely linked). Strain AH930/ pDD72 was constructed by P1 transduction of strain W3110/pDD72 to kanamycin resistance using strain AD90/pDD72 as donor. The mdoA::TnlO locus of strain NFB2OO was introduced by P1 transduction intostrain W3110/pDD72 prior to introduction of the pss93::kan locus for the construction of strain ADNl/pDD72. All strains carrying the pss93::kan gene werecured of plasmid pDD72 by growth at 43 "C on LB agar plates containing 20 mM MgCl,. Strain AD93/pDD72 was found to be UV light-sensitive (defective in DNA repair) and tetracycline resistant which is consistent with it being recA (recombination negative) since it was constructed by P1 transduction using a donor strain in which the recA gene is closely linked to a TnlO transposon. The P1 transduction which lead to the kanamycin resistance of strain AD90 also resulted in correction of the d B locus (closely linked to pss)of the recipient strain W3899/ pDD72 without transfer of the tetracycline resistance encoded by plasmid pADlOl supporting replacement of the chromosomal copy of the pss gene with the pss93::kan allele. Strain AD93/pDD72was temperature sensitive for growth at 43 "C when plated on LBmedium without drug supplementation consistent with the mutation of an essential chromosomal gene (pss) which was corrected by the only E. coli-derived gene (pss) on plasmid pDD72 (not maintained in cells grown at 43 "C). The occasional colony which formed under these conditions was chloramphenicol resistant (a plasmid revertant) and had the expected enzyme level of a strain retaining plasmid pDD72. PhospholipidAnalysis-Steady state phospholipid composition and the phospholipid to protein ratio were determined in cells uniformly

5325

labeled in LB medium with 32P04 prior to sampling and isolating the labeled phospholipid fraction as previously described (Heacock and Dowhan, 1987).Phospholipids were quantified by scraping and counting radiolabeled spotsafterseparation on boric acid-impregnated thin-layer silica gel plates (Fine and Sprecher, 1982) in one dimension using ch1oroform:methanol:water:ammonium hydroxide (120:75:62) or on H P thin-layer silica gel chromatography plates in two dimensions as described by Nishijima and Raetz (1979). Phospholipids were extracted from thin-layer platesfor further chromatographic analysis using the same conditions as for extracting phospholipids from cells. The mol % of each phospholipid was calculated as a percent of the total counts recovered and adjusted by the number of phosphate residues in each phospholipid. The fatty acid composition of the total phospholipids extracted from cells was determined after conversion to their methyl esters (Silbertet al., 1968). Chromatography on a fused silica capillary column was used to identify and quantify the fatty acids versus a standard mixture of fatty acid methyl esters. Miscellaneous Procedures-Phosphatidylserine synthase activity was determined by 30 "C as previously described by Larson and Dowhan (1976). Protein was determined by the BCA assay method according to themanufacturer's suggestions. The NHn-terminal protein sequence of phosphatidylserine synthase, purified as described by Ohta etal. (1981), was determined as previously described (Li and Dowhan, 1988). RESULTS

DNA Sequence of the pss Gene-The strategy used to determine the DNA sequence of the pss gene is diagramed in Fig. 2. Approximately 2000 base pairs (both strands and all overlaps) of the E. coli fragment subcloned from plasmid pPS4017 were sequenced. An open reading frame (Fig. 3) of 1356 base pairs, which could encode a protein of 452 amino acids, begins about 300 base pairs past the Sal1 site nearest the EcoRI site shown in Fig. 2. A strong potential ribosomal binding site and a weak potential RNA polymerase binding site (57%homology (Mulligan et al., 1984)with the consensus sequence) end 32 and 62 base pairs from the AUG initiation codon of the open reading frame, respectively. Protein Sequence of Phosphatidylserine Synthase-Aminoterminal protein sequencing of the phosphatidylserine synthase confirmed a free amino-terminal methionine followed by the next 11amino acids as predicted by the DNA sequence which supports the predicted start site for translationas shown in Fig. 3. The predicted molecular mass of the enzyme (52,817 Da) is in good agreement with the molecular mass (54,000 Da) estimated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (Larsonand Dowhan, 1976). A hydrophobicity plot (Fig. 4) using the method of Kyte and Doolittle (1982) shows two extended hydrophobic domains (residues 120-155 and 240-285) with the remainder of the protein being primarily hydrophilic. The most unusual feature of the protein is the high concentration of basic amino acids at both ends of the protein; 5 out of 9 and 10 out of 22 of the amino-terminal and carboxyl-terminal amino acids, respectively, carry a positive charge. There areno negatively charged amino acids among these amino acids at the amino-terminal end and only two negatively charged amino acids within the

1

kb

1

1

1

1

1

0.5

1

1

1

1

1

1.0

1

1

1

1

1

1.5

1

1

1

I

I

2.0

FIG. 2. Strategy used in sequencing thepss gene. The strategy used to determine the sequence of both strands of DNA is diagrammed. The unique restriction endonuclease restriction sites within this sequence are listed including the MluI site used to interrupt thepss gene.

Role of Phosphatidylethanolaminein E.coli

5326

-312 AT GATANXTGT CMACATGAC AATTCGTAGT ATACCWICGA -330 GGACTACMCGTCTWULTGG TATTACGATG C W A A T C -220

TCATGATTTT CMAPACGGT CAGGTTGTCG ACGCTTAATG OARMGTTGA TTACCTATAC GCCAGATUC

GTGGCATGGT GGCGCTAGCC OClVLGCTrGG GTITMCGTT GATATCUGC T C D M W G UTCGGTTGG CGTTACGCCA MTCTTGCCC AGCGCGAGGA ATUTWLGTA -110 AGGT-CTTAATCGGACT-CC GCTTATGTCG TCTGCATTGC AC -C CTTCMTCAR UcllGMGM AVXACTGIG -1

90 30 110 60 210 90 3 60 120

4 50 150

FIG. 3. Sequence of the pss gene. The determined DNA sequence along with the deduced protein sequence is shown. The potential ribosomal binding site (AGGA) and RNA polymerase recognition site (-35 and -10 regions) are underlined. Base pair and amino acid numberingappear at the end of each line. The MluI site used to interrupt the pss gene is at codon 39.

540 180

630 210 120 240

810 270 900 300

990 330 LOBO

3 60

1170 390 1260 420 1350 450

ax:

CTG TAAT CAUVLCCCCG TCCTGTACGG GGTPTGTTTT TTGCJLGGCCA CGTTTTCCTA TTCTTTTTGT CTGl'TCGCTA TTGTTACCTT CTGGATGCAG 1450 11e leu CCATATGGCT AACGATAGCT CGAGCGGGCA GULTAMGCT CMCACTTTA TCGCCTCGGC GATGCTTTCC GCCGCCGGAA ATGAATATTC ACAGCATCAG GaATGACXG 1560 GGATCGCRGT GCATGTTTGG ATTWLTGTCT CTGTCAGTTG OOCCT 1605

TABLE I1 Predicted amino acid composition Amino acid Residues/52,800 Da

21 AA#

100

200

300

400

FIG. 4. Hydropathy plot. The hydropathy index (positive values indicate hydrophobic regions) as a function of the protein sequence ( A A # )was calculated as described byKyte and Doolittle (1982) using an 11-amino acid sliding average.

group at the carboxyl-terminal end.37Both acidic and basic amino acids are more or less evenly distributed throughout the remainder of the protein. Table I1 summarizes the amino acid composition of the protein as predicted by the DNA sequence. Computer-assisted analysis of the protein sequence revealed little or no homology with the amino acid sequences of other CDP-diacylglycerol-dependent phosphatidyltransferases. Southern Hybridization Analysis of Strain AD93"Based on the rationale used to construct strain AD93 and its genetic and phenotypic properties (see "Experimental Procedures"), this strain should contain a single chromosomal copy of the pss gene interrupted by the kan gene and a plasmid borne copy of the wild type pss gene. The interrupted copy of the pss gene used to replace the wild type chromosomal copy of the gene was constructed by insertion of a 1.3-kb kanamycin resistance gene into a MluI site (see Figs. 2 and 3) which lies within the coding sequence of the pss gene and between the 2 SalI sites shown in Fig. 2. Therefore, the SalI fragment containing the pss gene should be increased from approximately 600 base pairs to approximatley 1.9 kb if it contains the kanamycin insert. WhenDNA isolated from strain AD931 pDD72 grown at 30 "C was digested with SalI and probed

ASP Asn Thr Ser Glu Gln Pro G1Y Ala CYS Val Met Ile Leu TYr Phe His LYS Trp A%

39 26 14 17 21 17 20 27 4 30 8

51 19

15

13 26 7 40

with the labeled 564-base pair SalI fragment, both a 600-base pair and 1.9-kb fragment were detected (Fig. 5), consistent with the strain carrying both forms of the pss gene; strain W3899 showed only the uninterrupted chromosomal copy of the gene. DNA isolated from strain AD93 lacking plasmid pDD72 (due to isolation after growth at 43 "C) was completely lacking the 600-base pair fragment characteristic of the uninterrupted gene, but contained the expected 1.9-kb fragment. Probing the digested DNA with the labeled 1.3-kb kanamycin fragment established that the 1.9-kb fragment characteristic of the interrupted gene did indeed contain the kanamycin fragment. Growth Characteristics of Strain AD93lpDD72"Strain AD93 was constructed to establish whether or not phospha-

Role of Phosphatidylethanolamine in E.coli 1

2

3

4

5

6

kb

5327

loo000 10000

FIG.5. Southern hybridization analysis. DNA isolated from the indicated strains was digested with SalI, separated by agarose gel electrophoresis relative to known standards (indicated in the center column), and subjected to Southern hybridization analysis as described under "Experimental Procedures." The labeled DNA probe used was either the 564-base pair SalI fragment containing part of the pss gene (lanes 1-3) or the 1.3-kb fragmentcontaining the kanamycin resistance gene (lanes 4-6). DNA was isolated from strain W3899 (lanes I and 4 ) , strain AD93/pDD72 grown a t 43 "C (lanes 2 and 5 ) , or strain AD93/pDD72 grown a t 30 "C (lanes 3 and 6 ) . tidylserine synthaseand ultimately phosphatidylethanolamine are essential for cell viability. Since plasmid pDD72 is temperature sensitive for replication, the plasmid will cease to replicate in cells grown at 43 "C resulting in the propagation of daughter cells (AD93 alone) lacking the plasmid and its copy of the functional pss gene. If the pss gene product is essential for growth, these daughter cells will stop growing when the level of existing gene product is diluted to below a critical level. This level should be determined by the amount of phosphatidylserine and further metabolites, such as phosphatidylethanolamine, which are essential for cell function. Fig. 6A shows a time course of cell growth, as measured by increase in absorbance, for cultures of strains W3899 and AD93/pDD72 after being shifted from 30 to 43 "C. Strain AD93 reproducibly stopped growing after about3.5-4 h at the restrictive temperature for plasmid replication while the wild type parent continued to grow normally. Viability of the cells in the culture was also analyzed for the formation of single colonies after plating serial dilutions on LB medium at 30 "C in theabsence of drug selection (data not shown). Within an hour after the temperature shift,the number of viable cells in the culture of strain AD93 increased more slowly than in the culture of strain W3899. The viable cell count ceased to increase in the former culture after about 2.5-3 h at 42 "C, but continued to increase in the latter culture. All the viable cells in the culture of strain AD93 still contained plasmid pDD72 as evidenced by their resistance to chloramphenicol. This is consistent with the multicopy number plasmid requiring several generations to dilute to anaverage of one copy per cell (constant number of viable cells) followed by the formation of daughter cells lacking the plasmid (increase in absorbance without an increase in the viable cell count). Enzymatic Activity as a Function of Growth TemperaturePhosphatidylserine synthase activity declined rapidly in cultures of strain AD93/pDD72 (Fig. 6B)after shifting cells to growth at 43 "C in parallel with the loss of viability. The

0

60 120 180 240 300

360 420

Minutes at 43oC

FIG.6. Effect of growth temperature on the properties of

strain AD93/pDD72. Thedata reported are from one of two experiments with essentially identical results. Cultures of strains W3899 (open symbols) and AD93/pDD72 (closed symbols) were grown overnight a t 30 "Cin LB medium. Cells were diluted into fresh medium and incubated a t 30 "C until the resumption of logarithmic growth was observed. A, both cultures were then shifted to 43 "C and monitored for growth by the increase in turbidity a t 550 nm as a function of time. The cultures were diluted periodically with fresh media to maintain logarithmic growth. B, at the indicated times the growing cultures were sampled, and cells harvested, and assayed for phosphatidylserine synthase specific activity as outlined under "Experimental Procedures." C, parallel cultures were grown in an identicalmanner as described above exceptin the presence of 32P04 beginning with the overnight culturesso that thephosphate containing compounds were uniformly labeled a t all times. Samples were removed at the indicated times and analyzed for their phospholipid composition using two-dimensional thin-layer chromatography as discussed under "Experimental Procedures." Only data, expressed as mol ?6 of each phospholipid, for strain AD93 is shown. Phosphatidylethanolamine (A);phosphatidylglycerol (W); cardiolipin (+). The phospholipid composition of strain W3899 remained similar throughout theexperiment to the zero time values of strain AD93/pDD72.

mutant strain began with a specific activity for the enzyme about %fold higher (multiple copies of thepss gene) than that for the wild type strain. The specific activity measurements indicate that after about 2.5 h (the point at which the viable cell count no longer increased) the cells isolated from the culture of strain AD93 had a specific activity for the phosphatidylserine synthase equivalent to that of strain W3899 (one copy of the gene). Subsequent increase in cell mass did not result in an increase in total enzymatic activity, hence the existing enzyme was diluted out reducing the specific activity. The specific activity of the mutant culture at the point of cell arrest was about 10-20% of the specific activity of the wild type culture, indicatingthat the phosphatidylserine

5328

Role of Phosphutidylethunolaminein E.coli

synthase is in a 5- to 10-fold excess over the minimal level needed for cell growth. Effect of Enzyme Level on Phospholipid Composition-The phospholipid composition of both strainswas nearly identical for the first 2 h after the temperature shift (see Fig. 6C for strain AD93; data for strain W3899 not shown). For strain AD93 the most dramatic reduction in the ratio of zwitterionic phospholipid to acidic phospholipid occurred between 2 and 3 h after the temperature shift, which was the point in time when there was a significant decline below wild type levels in the specific activity of the phosphatidylserine synthase (Fig. 6B).Total acidic phospholipid continued to increase until cell arrest at 4 h, after which time the major change was the continual synthesis of cardiolipin, presumably at theexpense of existing phosphatidylglycerol. The limiting level of phosphatidylethanolamine was about 30% of the total phospholipid at the point of cell arrest as compared to about 70% at the time of the temperature shift. Fig. 6C only showsthe results for the threemajor phospholipids of E. coli. Phosphate positive material remained at the origin in the chromatographic system used in this experiment, after two aqueous washes of the chloroform phase, and increased from 1.5% (similar to thevalue at all times for strain W3899) to 3.5% for strain AD93 during the course of the samplings; in the solvent system employed CDP-diacylglycerol remains at the origin. A spot with the mobility of phosphatidic acid fluctuated between 2 and 6% for both strains over the course of the experiment. Finally, material with the mobility of lysophosphatidylethanolamine remained under 3% for both strains except for the last time point for strain AD93 where the level increased to 8%. Suppression of the Growth Phenotype by Divalent Metal Ions-Since partial phenotypic suppression of strains carrying existing mutation of either the pss (Ohta and Shibuya, 1977; Raetz et al., 1979) or psd (Hawrot and Kennedy, 1978) genes can be brought about by addition of magnesium and/or sucrose to thegrowth medium, the possibility existed that the pss gene product is not absolutely essential under certain growth conditions, particularly those in which the cell membrane may be stabilized. To investigate this possibility, a constant number of cells (between 100 and 500) of strain AD93/pDD72 or AH930/pDD72 were plated at both 30 and 43 "C on LB medium supplemented either with or without 20 mM MgC1,; this level ofMgCl, is similar tothat which phenotypically suppresses the pssl allele (Ohta and Shibuya, 1977). As expected, the plating efficiency at 30 "C was independent of MgC12, while at 43 "C the formation of colonies (at the same efficiency as at 30 "C) was absolutely dependent on MgC1,. All colonies which formed under these conditions at 43 "C werekanamycin resistant and chloramphenicol-sensitive, consistent with the absence of a wild type copy of the pss gene either in the chromosome or carried on an extrachromosomal plasmid. Resistance to kanamycin was reduced to 10 pg/ml, consistent with the previous observation that pss- strains are hypersensitive to water-soluble antibiotics (Raetz and Foulds, 1977). Southern hybridization analysis, using a pss-specific probe, of DNA isolated from strain AD93 grown in the presence of MgC1, (now lacking plasmid pDD72) verified the absence of a wild type copy of the pss gene, but did show a fragment of the size expected for the interrupted gene (datanot shown); the hybridization experiment also verified that the isolates were a derivative of E. coli and not contaminants. Properties of Strains Lacking a Functional pss GeneGrowth of strain AH930 on agar plates containingMgC1, was facilitated in the presence of 400 mM sucrose but was inde-

pendent of NaCl up to 500 mM; in all cases growthwas dependent on the presence of MgCl,. Dilution (100-fold) of a liquid culture grown in the presence of 20 mM MgC12 into liquid LB medium between 1 and 50 mM in various divalent metal ions resulted in detectable growth after 72 h in the presence of a minimum of 5 mM CaC12, 5 mM MgC12, or 7.5 mM SrC12. After an additional 24 h, detectable growth was seen at 2 mM CaCl, and 5 mM SrC1,. No growth was detected in the presence ofBaC1,. Theseresults suggest a general divalent metal ion requirement for this strain rather than a requirement for a specific ion, ionic strength, or osmolarity. These results also support the following hierarchy of effectiveness in substituting for the lack of phosphatidylethanolamine: Ca2+> M$+ > Sr2+.Divalent cobalt, nickle, or manganese did not supportgrowth but was also toxic to theparent strain at thelevels used. Spermidine at 20 mM did not sustain growth. The strains formed red colonies on MacConkey agar (with lactose as thecarbon source) supplemented with 20 mM MgC12, indicating thatthe cells could transport lactose (Miller, 1972) and could grow in the presence of ionic detergent. Individual cells tended to form long filaments which increased in number and size asthe culture reached the stationary phase of growth, similar to that reported for pss (Raetz, 1976) and psd (Hawrot and Kennedy, 1978) temperature-sensitive mutants grown under restrictive conditions. Supplementation of LB medium with glucose as a carbon source or 100 mM MOPS (pH 7.2) for buffering did not noticeably affect the growth phenotype. Conditions could not be found to maintain growth in minimal defined medium either with or without supplementation with caseine-derived amino acids. The generation time of strain AH930 at 37 "C in LB medium was dependent on the MgC1, concentration as follows: 150 min at 7.5 mM, 75 min at 20 mM, and 45 min at 50 mM. The parent strain W3110 had a doubling time of 25 min over the same concentration range of MgCl2. Strain ADNl (pss93::kan rndoA::TnlO) had a doubling time of 30 min at 50 mM MgClz suggesting that the turnover of phosphatidylglycerol to membrane-derived oligosaccharide which is dependent on the mdoA locus (Kennedy and Rumley, 1988; Bohin and Kennedy, 1984) may be growth limiting for strain AH930.However, supplementation of the growth medium with 400 mM sucrose, which suppresses the synthesis of membrane-derived oligosaccharide (Bohinand Kennedy, 1984), had no effect on the growth rate of strain AH930. The growth properties of strain AD93 wereindistinguishable from strain AH930. Phospholipid Content of Strains AD93 and AH930 Lacking a Functional pss Gene-The phospholipid composition of strain AD93 growingexponentially in 20 mMMgC12 was 63% phosphatidylglycerol, 13% cardiolipin, 8%phosphatidic acid, and 14% other phosphate-containingchloroform-soluble material (among these were probably CDP-diacylglycerol and lysophosphatidic acid) which ran closer to theorigin than the identified phospholipids in the two-dimensional thin-layer chromatography system. Unlabeled carrier phosphatidylethanolamine was included to detect the position of this phospholipid on the thin-layer chromatography plates; no significant radiolabel was detected in this region. Strain AH930 has a very similar phospholipid composition to strain AD93 under these conditions. To determine the relative phospholipid content of strains AH930 and W3110, cells were uniformly labeled by growth of the same specific overnight in LB medium containing 32P04 activity and supplemented with 50 mM MgC12. Cellswere harvested by centrifugation, suspended in 10 mM Tris-HC1 (pH 7.0), and disrupted by sonication; whole cells were re-

5329

Role of Phosphatidylethanolamine in E. coli moved by centrifugation at 5,000 x gav.The membrane fraction was isolated from a portion of each of the sonicates by centrifugation a t 200,000 x g,, for 30 min. The radiolabel incorporated into phospholipid present in the membrane fraction, suspended in 10 mM Tris-HC1 (pH 7.0), and the whole sonicate was determined after extractionof the phospholipids as described under “Experimental Procedures.” The phospholipid content was normalized to the protein content(disintegrations/min/mg of protein) for each fraction. The ratio of the phospholipid content of strain AH930 to strain W3110 for the whole sonicate and themembrane fraction was 0.84 0.09 and 0.97 k 0.08, respectively (average of four determinations). Therefore, there is no significant difference in the phospholipid content of the membranes of the mutant strain relative to thewild type strain with a possible small decrease in the totalphospholipid content when comparing the whole cells. The absence of zwitterionic phospholipid is therefore compensated by an increase in almost exclusively phosphatidylglycerol and cardiolipin (as shown below) resulting in maintenance of the normal phospholipid to protein ratio in the cell membrane. Fig. 7 shows the distribution of the phospholipids in the cell extracts which were used to determine thephospholipid content of strains AH930 and W3110. The short-term exposures (panels A and B ) of the chromatography plates show the expected absence of phosphatidylethanolamine and its lyso-derivative in extracts from strain AH930. Extracts from strain AH930 contained increased levels of three spots(phosphatidic acid, lysophosphatidic acid, and probably CDP-diacylglycerol) relative tostrain W3110 extracts,butasis apparent from the long-term exposures (panels C and D )of the chromatography plates, these components were found in trace amounts in the extracts of strain W3110. The major difference between the long-term exposures is the absence of many minor spots in the extract from strain AH930 which suggests the absence of several minor components derived from phosphatidylethanolamine and/or phosphatidylserine (panel C ) . As previously recognized by Raetz (1976), long exposures of this typereveal many previously uncharacterized chloroform-soluble phosphate-containing metabolites.

*

On closer examination strain AH930 appears tostill contain trace amounts of phosphatidylethanolamine. Strain AH930 was grown to stationary phaseof growth at 37 “C in 2 ml of LB medium containing 50 mM MgC12 supplemented with 1.6 mCi of :”PO4.The phospholipid was extracted in the presence of 25 pg of exogenously added unlabeled phosphatidylethanolamine and the chloroform phase neutralized using Tris free base and made 60% with respectto methanol. The sample was passed through a 3.5-ml DEAE-cellulose column equilibrated with ch1oroform:methanol (2:3, v/v) which retained over 99% of the radiolabel (the acidic phospholipids), but allowed 70% of the phosphatidylethanolamine carrier to pass through the column (determined on a separate column). The run-through fraction was subjected to two-dimensional thinlayer chromatography and the region corresponding to the carrier phosphatidylethanolamine was identified by ninhydrin spray reagent. Autoradiography revealed detectable radiolabel associatedwith the phosphatidylethanolamine which was quantified using direct counting of the material in the presence of water without scintillation fluid; this material, after extraction from the silica gel, still migrated with the phosphatidylethanolamine carrierin the ammonia-containing solvent system described under “Experimental Procedures.” After correction for yield and decay, the radiolabeled material, which co-chromatographedin three solventsystemswith phosphatidylethanolamine, accounted for 0.007% of the total radiolabel originally extracted. Since strain AH930 appears to have a similar phospholipid content as wild type cells, strain AH930 may contain approximately 1500 molecules of phosphatidylethanolamine, assuming approximately2.2 X 10’ phospholipid molecules/cell as calculated for E. coli strain B/ r (Neidhardt, 1987). Fatty Acid Compositionof Strains Lacking Phsphatidylethanolamine-Cultures of strains W3110 and AH930 were grown to mid-log phase of growth a t 37 “C in LB medium containing 50 mM MgC12 and the phospholipids extracted from the whole cells and analyzed for total fattyacid composition as described under“Experimental Procedures.” As shown in Table 111, the fattyacid composition of both strains was remarkably similar exceptfor the increase in the content of 17:A at the apparentexpense of 16:l in strain AH930. DISCUSSION

There were no particularly unusual features revealed from the DNA sequence of the pss gene. The lack of a strong consensuspromoter is not surprising since the enzyme is expressed constitutively inlow amounts (Larson and Dowhan, 1976). The predicted protein sequence contained two hydrophobic domains. Although the enzyme appears not to be an integral membrane protein (Louie and Dowhan, 1980), the association of the enzyme with detergent micelles and memTABLE111 Fatty acid composition of total cellular phospholipid fraction Mol% Fatty acid

FIG. 7. Two-dimensionalthin-layerchromatograms of chloroform extracts of strains W3110 and AH930. Cells were uniformly labeled with ?‘PO, in LB medium containing 50 mM MgC12 and harvested in the stationary phase of growth. Radiolabeled phospholipids were extracted as described in the text and subjected to thin-layer chromatography. Panels A and C are extracts from strain W3110 and panels R and D are extracts from strain AH930. The thin layer plates were exposed to photographic film for either 5 ( A and R ) or 72 h ( C and D); PE, phosphatidylethanolamine; PG, phosphatidylglycerol; CL, cardiolipin; PA, phosphatidic acid; PS, phosphatidylserine; LyPE, lysophosphatidylethanolamine; LyPA, lysophosphatidic acid; CDP-DG, CDP-diacylglycerol.

42

5

12:o 14:O 16:O 16:l 17:A” 180 181 19:Ah Other

W3110

c1 1.7 45 15 1.7 20 14