Proc. Nati. Acad. Sci. USAVol. 87, pp. 1506-1510, February 1990Biochemistry

Protein N-myristoylation in Escherichia coli: Reconstitution of aeukaryotic protein modification in bacteria

(myristoyl-CoA:protein N-myristoyltransferase/E. cofi expression vectors/fatty add analogs/protein kinase A)

ROBERT J. DuRoNIo*, EMILY JACKSON-MACHELSKI*, ROBERT 0. HEUCKEROTH*, PETER 0. OLINSt,CATHERINE S. DEVINEt, WES YONEMOTOt, LEE W. SLICEt, SUSAN S. TAYLORt, AND JEFFREY I. GoRDON*§¶Departments of *Biochemistry and Molecular Biophysics and WMedicine, Washington University School of Medicine, St. Louis, MO 63110; tMonsantoCompany, St. Louis, MO 63198; and *Department of Chemistry, University of California, San Diego, La Jolla, CA 92093

Communicated by Carl Frieden, November 27, 1989 (received for review October 9, 1989)

ABSTRACT Protein N-myristoylation refers to the cova-lent attachment of a myristoyl group (C14:0), via amidelinkage, to the NH2-terminal glycine residue of certain cellularand viral proteins. Myristoyl-CoA:protein N-myristoyltrans-ferase (NMT) catalyzes this cotranslational modification. Wehave developed a system for studying the substrate require-ments and biological effects of protein N-myristoylation as wellas NMT structure-activit relationships. Expression of theyeast NMT1 gene in Escherchia cofi, a bacterium that has noendogenous NMT activity, results in production of the intact53-kDa NMT polypeptide as well as a truncated polypeptidederived from proteolytic removal of its NH2-terminal 39 aminoacids. Each E. coli-synthesized NMT species has fatty acid andpeptide substrate specificities that are indistinguishable fromthose of NMT recovered from Saccharomyces cerevisiae, sug-gesting that the NH2-terminal domain of this enzyme is notrequired for its catalytic activity. By using a dual plasmidsystem, N-myristoylation of a mammalian protein was recon-stituted inE. coliby simultaneous expression ofthe yeastNMT1gene and a murine cDNA encoding the catalytic (C) subunit ofcAMP-dependent protein kinase (PK-A). The fatty acid spec-ificity of N-myristoylation was preserved in this system:[9,10(n)-3H]myristate but not [9,10(n)3H~palmitate was effi-ciently linked to Gly-1 of the C subunit. [13,14(n)-3HJ10-Prop-oxydecanoic acid, a heteroatom-containing analog of myristicacid with reduced hydrophobicity but similar chain length, wasan effective alternative substrate for NMT that also could beincorporated into the C subunit of PK-A. Such analogs haverecently been shown to inhibit replication of certain retrovirusesthat depend upon linkage of a myristoyl group to their gagpolyprotein precursors (e.g., the Pr55P9 of human immunode-ficiency virus type 1). A major advantage of the bacterial systemover eukaryotic systems is the absence of endogenous NMT andsubstrates, providing a more straightforward way of preparingmyristoylated, analog-substituted, and nonmyristoylated formsof a given protein for comparison of their structural andfunctional properties. The system should facilitate screening ofenzyme inhibitors as well as alternative NMoT fatty add sub-strates for their ability to be incorporated into a specific targetprotein. Our experimental system may prove useful for reca-pitulating other eukaryotic protein modifications in E. coil sothat structure-activity relationships of modifying enzymes andtheir substrates can be more readily assessed.

Cotranslational (1) covalent attachment of myristic acid(C14:0) to the NH2-terminal glycine residue of a variety ofcellular and viral proteins is, in many instances, required forfull expression of their biological activity (reviewed in refs. 2and 3). Current approaches to understanding the contribution

of N-myristoylation to protein structure and function haveinvolved site-directed mutagenesis of the NH2-terminal gly-cine to prevent acylation or the incorporation of heteroatom-containing analogs of myristic acid with reduced hydropho-bicity into N-myristoylated proteins in vivo (4). For example,abolishing myristoylation of the tyrosine kinase p6Ov-src bydeletion of its Gly-1 residue or by Gly-1 -+ Ala substitutionrevealed (5, 6) that the C14:0 fatty acid is required for theprotein's stable association with the plasma membrane (prob-ably through interaction with a high-affinity myristoyl-srcreceptor; refs. 7 and 8) and its ability to transform cells.Analogous mutagenesis of the Gly-1 residues of the Moloneymurine leukemia virus Pr65g (9), the Mason-Pfizer monkeyvirus Pr7859 (10), and the Pr55n of human immunodefi-ciency virus I (11) blocks viral replication. X-ray crystallo-graphic studies (12) and site-directed mutagenesis (13) of theN-myristoylated poliovirus capsid protein VP4 have alsoindicated that myristic acid is involved in protein-proteininteractions and in viral capsid protein assembly.We have used an in vitro assay system to study the

substrate specificity of myristoyl-CoA:protein N-myristoyl-transferase (NMT; E.C. 2.3.1.97) from yeast and mammaliancells (14-16). Detailed analyses ofSaccharomyces cerevisiaeNMT using octapeptide substrates and naturally occurringacyl-CoAs of different chain lengths revealed that the highdegree of selectivity of the enzyme for myristoyl-CoA arisesfrom an apparent cooperativity between its acyl-CoA andpeptide binding sites. When acyl-CoAs of "incorrect" chainlength are bound by NMT, dramatic reductions occur in theenzyme's affinity for peptide substrates, thus preventinggeneration of acyl-peptides (16). These studies also showedthat the acyl-CoA and peptide substrate specificities ofNMTs from species as diverse as yeast and mammals arehighly conserved (15).

Substitution of an oxygen or sulfur for a methylene groupin the hydrocarbon chain of myristic acid produces substan-tial reductions in hydrophobicity (equivalent to the loss oftwo to four methylene groups) without affecting chain length(16). In vitro studies (16) indicated that the CoA thioesters ofthese analogs are good substrates for NMT, although theapparent cooperativity between the acyl-CoA and peptidebinding sites produces sequence-dependent differences in theefficiency oftheir transfer to octapeptides. Similar selectivitywas observed in vivo when one such analog, 10-[3H]propoxy-decanoate (4) was incorporated into a subset of cellularN-myristoylated proteins synthesized in a mouse myocyte

Abbreviations: NMT, myristoyl-CoA:protein N-myristoyltrans-ferase; glO-L, gene 10 leader of phage T7; PK-A, cAMP-depen-dent protein kinase; C-subunit, catalytic subunit of PK-A.$To whom reprint requests should be addressed at: Department ofBiochemistry and Molecular Biophysics, Washington UniversitySchool ofMedicine, 660 South Euclid Avenue, Box 8231, St. Louis,MO 63110.

1506

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Dow

nloa

ded

by g

uest

on

Mar

ch 6

, 202

1

Proc. Natl. Acad. Sci. USA 87 (1990) 1507

cell line (BC3H1). In addition, only a subset of analog-substituted proteins exhibited redistribution from the mem-brane to cytosolic fractions. The selective incorporation ofheteroatom-containing analogs into myristoylated proteinsand their selective sequence-dependent effects on membranetargeting and/or association may account for their relativelack of toxicity to mammalian cells and also for their antiviralproperties. We have recently demonstrated that replicationof HIV-I and Moloney murine leukemia virus is inhibited bydifferent heteroatom-containing analogs of myristic acid (17).To further define NMT structure-activity relationships and

to understand the role of covalently bound myristic acid inprotein function and structure, we have developed a noveldual plasmid expression system that allows us to recreate thiseukaryotic protein modification in Escherichia coli.

MATERIALS AND METHODS

Expression of NMTJ in E. coil. A 780-base-pair (bp) region(nucleotides 213-993) ofthe 2.1-kilobase (kb) BamHI-HindIIIS. cerevisiae genomic NMT1 fragment (18) was amplified byusing the polymerase chain reaction (19) and a mutagenicoligonucleotide (5'-CGGTAGTAAACGATCC-ATACCATGGCAGAAGAGGATAAAGCGAAAAAAT-3 ').This allowed us to introduce an Nco I restriction enzyme siteat the initiator ATG codon of NMTJ. The new Nco I site alsochanged codon two of NMTJ from a serine- to an alanine-encoding codon. Amplification products were subcloned backinto pBB105 (18) to generate the alteredNMT1 allele. The NcoI site allowed us to link the NMTJ gene to the E. coli recApromoter and a translational enhancer element (gene 10 leader)from phage T7 (gJO-L in Fig. la). This was accomplished byligating the newly generated 1.9-kb Nco I-HindIII fragmentinto Nco I/HindIII-digested pMON5840, a derivative ofpMON5515 (20). The resulting plasmid (pBB125) was used totransform E. coli strain JM101. Transformants were shaken at

37°C in LB broth containing 100 ,ug of ampicillin per ml to an

OD6w of 1.0. The recA promoter was induced by addingnalidixic acid to a final concentration of 50 ,ug/ml (21). Aftera 15- to 20-min incubation at 37°C, cells were harvested bycentrifugation and broken under pressure [2000 pounds persquare inch (psi); 1 psi = 6.89 kPa] with a french press. NMTspecies were purified as described in the text.Plasmids for Coexpression in E. coil of NMT1 and cDNA

Encoding the Catalytic (C) Subunit Type a of PK-A. pBB131was constructed by cloning the 1.9-kb Nco I-HindIIINMTIfragment into pMON5839 (P.O.O. and C.S.D., unpublisheddata), a derivative of pACYC177 (22). Ca cDNAs of PK-A(23) were cloned as 1.8-kb Nde I-Kpn I fragments intopMON2670 (24). pBB132 contains the wild-type Ca cDNA(kindly supplied by G. Stanley McKnight), which specifies aglycine at position 2 of its primary translation product (theinitiator methionine occupies position 1). A mutant CacDNAwith an alanine at position 2 was made. Uracil-containingsingle-stranded template was prepared from the phagemidpUC119/Ca in RZ1032 cells, and Gly-2 Ala mutagenesiswas performed with the oligomer 5'-CATATGGCCAACGC-CGCC-3' (25). The mutation was confirmed by dideoxynu-cleotide sequencing. pBB131 (NMTI) and pBB132 (encodingGly-2-containing Ca) or pBB133 (encoding Ala-2-containingCa) were used to transform E. coli strain JM101. Restrictionanalysis of plasmid DNA confirmed the identity of constructswithin ampicillin/kanamycin-resistant double transformants.

Tritiated Fatty Acid Labeling of PK-A C Subunit Producedin E. coli. Four-milliliter cultures of the double transformantswere shaken at 370C to an OD6N of0.5 in LB broth containing100 ,ug of ampicillin and 100 pugof kanamycin sulfate per ml.

Isopropyl-8-D-thiogalactopyranoside was then added to afinal concentration of 1 mM to induce NMT production (seeResults). When the cultures reached OD600= 1.0, nalidixic

acid was added to a final concentration of50 Ag/ml to induceC-subunit production (see Results). [9,10(n)-3H]Myristate[New England Nuclear; 39.3 Ci/mmol (1 Ci = 37 GBq), 113puCi per ml of culture], [9,10(n)-3H]palmitate (New EnglandNuclear; 30 Ci/mmol, 143 ,Ci/ml), or [13,14(n)-3H] 10-propoxydecanoate (ref. 4; 31.7 Ci/mmol, 800 LCi/ml) wasadded simultaneously with the nalidixic acid. Cultures wereshaken for an additional 20 min at 370C, and the cells wereharvested by centrifugation. Lysates were prepared by boil-ing E. coli contained in the pellet for 10 min in 40 ,ul of 125mM Tris, pH 8.0/4% sodium dodecyl sulfate (SDS)/20%(vol/vol) glycerol/10% 2-mercaptoethanol/0.2M dithiothrei-tol. Cell debris was removed by centrifugation, and 15-1ilaliquots of the supernatant were subjected to SDS/PAGE(26) and subsequent fluorography with EN3HANCE (NewEngland Nuclear). For immunoblot (Western blot) analyses,50 ,ug of reduced and denatured lysate proteins, preparedfrom unlabeled E. coli producing Gly-2- or Ala-2-containingC subunit, were separated by SDS/PAGE and electroblottedonto nitrocellulose. The filters were probed with polyclonal,monospecific rabbit antisera raised against purified mouse Csubunit. Antigen-antibody complexes were visualized with125I-labeled protein A.In Vitro Assay System for NMT Activity. To assess the

peptide and acyl-CoA substrate specificities of E. coli-derived S. cerevisiae NMT, crude lysates or partially purifiedenzyme preparations were added to a coupled in vitro assaysystem (16).

RESULTS AND DISCUSSIONS. cerevisiae NMT Produced in E. coli Has a Substrate

Specificity Similar to That of the Authentic Yeast Enzyme. TheS. cerevisiae NMTI gene encodes a protein of 455 aminoacids with a calculated Mr of 52,837 that is essential forvegetative cell growth (18). A six-step 11,000-fold purifica-tion involving the use of four different chromatographicmatrices was required to obtain an apparently homogeneouspreparation of enzyme from this yeast (14). E. coli lysatescontain no detectable NMT activity as judged by a sensitivein vitro assay (16) for the enzyme (data not shown). Thus,expression of yeast NMT in the prokaryote offers an oppor-tunity to obtain large quantities of wild-type (or mutant)protein whose activity can be measured in the absence ofanyendogenous myristoyltransferases.The expression of S. cerevisiae NMT in E. coli was

achieved by using pMON plasmid vectors. These vectorscontain inducible promotors fused to a translational "enhanc-er" derived from the gene 10 leader region (glO-L) ofbacteriophage T7 (20). The yeast NMTJ gene was clonedimmediately downstream of glO-L, and its transcription wascontrolled with the E. coli recA promoter located 5' to gJO-L.E. coli strain JM101 carrying this recombinant plasmid wasgrown to mid-logarithmic phase and then treated with nali-dixic acid to induce the recA promoter. NMT was subse-quently purified -750-fold by sequential ammonium sulfatefractionation, DEAE-Sepharose CL-6B, and CoA-agaroseaffinity column chromatography of induced cell lysates (14).The partially purified E. coli-derived yeast NMT displayedKm and Vmax values for a variety of octapeptide substratesthat were nearly identical to those measured with a partiallypurified NMT preparation from S. cerevisiae (Table 1). Forexample, introduction of a serine residue at position 5 of a"parental" octapeptide, GNAAAARR-NH2, obtained fromthe NH2-terminal sequence of the C subunit ofPK-A reducesits apparent Kmby a factor of >100 for both NMT prepara-tions. An NH2-terminal glycine is absolutely required. Sub-stitution ofan Ala-1for the Gly-1 residue converts the peptideinto an inactive substrate. Addition of an NH2-terminalmethionine residue also generates an inactive peptide, indi-

Biochemistry: Duronio et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 6

, 202

1

1508 Biochemistry: Duronio et al.

cating that yeast NMT partially purified from E. coli, likeNMT isolated from S. cerevisiae (14), has no associatedmethionylaminopeptidase activity.To verify that E. coli was producing an intact yeast NMT,

proteins eluted from the CoA-agarose column with 100 mMKCl were separated by SDS/PAGE and transferred to a

polyvinylidene difluoride membrane. An =53-kDa polypep-tide, which corresponds to the mass of the 455-residue yeastNMT (18), was excised from this membrane and subjected toEdman degradation. The NH2-terminal sequence indicatedthat the 53-kDa polypeptide represented intact yeast NMT(data not shown).To obtain a homogenous preparation of the enzyme, E.

coli-produced NMT was further purified by Mono S fastprotein liquid chromatography (FPLC). Coomassie bluestaining of SDS/PAGE gels of the 250 mM NaCl eluate froma Mono S column (14) revealed an additional band of %'45kDa, which coeluted with NMT catalytic activity. Edmandegradation of the 45-kDa polypeptide revealed that it wasmissing the NH2-terminal 39 residues of NMT, suggestingthat portions of the polypeptide chain, such as the bondbetween Lys-39 and Phe-40, are susceptible to proteolysisand rapidly lost either during purification or shortly aftersynthesis in E. coli.The Mono S-purified 45-kDa NMT species retained the

ability to readily distinguish between myristoyl-CoA andpalmitoyl-CoA and displayed the reduction by 100 in apparentKm for Ser-5-substituted GNAAAARR-NH2 (Table 1). There-fore, the 45-kDa proteolytic fragment retains the core catalyticdomain. The role of the missing 39 amino acids remainsunknown, but they may be needed for (essential) interactionsofNMT with additional factors within yeast or for its properintracellular targeting. Determining whether a genetically en-gineered 45-kDa NMT could rescue the inviable Nmt- phe-notype of S. cerevisiae (18) should permit us to address thesequestions. Since E. coli-derived NMT has kinetic properties

similar to yeast-derived NMT, we can conclude that theenzyme's peptide and fatty acyl-CoA substrate specificitiesare not dependent upon either a eukaryotic protein modifica-tion or additional yeast specific factors.

Reconstitution of Protein N-myristolation in E. coli. Thedata in Table 1 indicated that expression of yeast NMT in E.coli yielded an enzyme that was properly folded in that itssubstrate specificities were largely indistinguishable fromthose of NMT isolated from S. cerevisiae. Since there is noendogenous NMT activity in E. coli, our results raised thepossibility that coexpression of yeast NMT and a eukaryoticprotein substrate in this bacterium would permit us to repro-

duce in a prokaryote a protein modification that is apparentlyexclusively eukaryotic.PK-A was one of the earliest protein kinases to be discov-

ered and is also one ofthe best understood biochemically (27).The kinase is involved in the regulation of cell growth andmetabolism, and its C subunit was the first protein shown tobe N-myristoylated (28). Expression of Ca cDNA (23) in E.coli led to the isolation of a soluble and active form of theprotein (29) that lacked myristate at the NH2-terminal glycine.Having shown that an octapeptide derived from the NH2terminus of the murine C subunit was a good substrate for theintact (and truncated) E. coli-derived yeast NMT in vitro(Table 1), we decided to use the C subunit as a model proteinfor our in vivo reconstitution experiments. The dual plasmidsystem outlined in Fig. 1A was utilized to coexpress the yeastNMTJ gene and the C, cDNA. The vectors were designed soeach (i) could be simultaneously maintained as a stable epi-somal plasmid and (ii) could support independent induction oftranscription of their foreign DNA sequences. Expression ofNMTJ was placed under the control of the isopropyl-f3-D-thiogalactopyranoside-inducible tac (30) promoter and thegJO-L ribosome binding site (20) contained in a plasmid basedon pACYC177 (22). This plasmid includes the pi5A origin ofreplication and a kanamycin-resistance gene. Expression of

Table 1. Peptide and acyl-CoA substrate specificities of E. coli-derived S. cerevisiae NMTE. coli NMT

Total 45 kDa Yeast NMT

Substrate Kin, AM Vmax, % Kmi ALM Vmax, % Kmi, LM Vmaxs %PeptideGNAAAARR 65.8 100 62.5 100 60 100GNAASARR 0.103 3.6 0.38 4.9 0.1 3GSAAAARR 697 27.6 ND ND 1700 50GSAASARR 2.1 29 ND ND 3.0 34GSSKSKPK 48.5 33 66 52 40 43GPAAAARR 7100 21.6 <2 <2GNAADARR 503 4.9 <2 <2GDAAAARR 645 4.2 <2 <2ANAAAARR Not a substrate Not a substrate Not a substrateMGNAAAARR Not a substrate Not a substrate Not a substrateGYAAAARR <2 <2 Not a substrateASSKSKPK Not a substrate Not a substrate Not a substrate

Fatty acidMyristoyl-CoA ND ND 0.36 100 0.675 100Palmitoyl-CoA ND ND 0.96 22.4 0.700 2111-Oxamyristoyl-CoA ND ND 1.2 450 6.1 270

The peptide Km values and relative Vmax values shown in this Table are averages obtained from four-or five-point Lineweaver-Burk plots. These plots were produced by using data generated in at leastthree independent in vitro NMT assays. Various concentrations of peptides were assayed with 15 ,uMmyristate to determine peptide Km and Vmax values. Fatty acyl-CoA Km and Vma values were obtainedin a similar way, except that various concentrations of fatty acid were used with 60 ,tM GNAAAARR(i.e., peptide at its Km concentration). Yeast NMT was from a 570-fold purified preparation (14).Synthetic peptide substrates are represented by the one-letter amino acid code. Vmax values arepercentages of those obtained when GNAAAARR and myristoyl-CoA were used as the peptide andacyl-CoA substrates, respectively; "-" indicates that the quantity of enzyme needed to accuratelydetermine the Km was prohibitively large. Peptides labeled "not a substrate" had Vmax < 1%. ND =not determined. 11-Oxamyristate = 10-propoxydecanoate.

Proc. Natl. Acad. Sci. USA 87 (1990)

Dow

nloa

ded

by g

uest

on

Mar

ch 6

, 202

1

Proc. Natl. Acad. Sci. USA 87 (1990) 1509

a

pBB131

pBB132pBB133(G2-.A2)

b

NcolPfc 910-LtSsss .......

NdeIPrecA g1O-L|

C

HindIIINMT1 J KAN R ori-p

KpnICa I

d

)15A

AMP R ori-ColEl

e 1 2 3

12 3 4 5 6 7 2 3 4 5 6 7 1 2 3 4 5 6 7

97-

68 -

97-

68-

43 -

43- __

25-

25-18-

FIG. 1. N-myristoylation of the murine PK-A C subunit synthesized in E. coli. (a) Schematic representation of plasmid constructs used toexpress NMTJ and Cc, cDNA in E. coli. KANR, gene for kanamycin resistance; AMPR, gene for ampicillin resistance; Ca, Ca cDNA(crosshatched area); G2 -* A2, Gly-2 -- Ala. (b-d) Lysates were prepared from E. coli transformants containing different combinations ofplasmids after labeling with exogenously added [3H]myristate (b), [3H]palmitate (c) or [3H]10-propoxydecanoate ([3H]11-oxamyristate) (d).Tritiated lysate proteins were then subjected to SDS/PAGE and fluorography. Sizes are shown in kDa. The arrow indicates the position ofmigration of purified mouse C subunit. Note that the fluorographic exposure time for the gels shown in b and c was 4 days, whereas the gelshown in d was exposed for 15 days. Lanes: 1, E. coli strain JM101 without plasmids; 2, JM101 with parental vectors; 3, JM101 with recombinantNMTI- and Ca cDNA-containing plasmids, without induction; 4, JM101 with NMTJ- and Ca cDNA-containing plasmids after induction; 5,JM101 with NMTJ- and mutant Ala-2 Ca cDNA-containing plasmids after induction; 6, JM101 with NMTJ and parental vector lacking the CacDNA insert after induction; 7, JM101 with Ca cDNA and parental vector lacking the NMTI insert after induction. (e) Western blot analysisof E. coli strain JM101 lysate (lane 1) containing wild-type (Gly-2) C subunit (lane 2) or mutant (Ala-2) C subunit (lane 3). The blots were probedwith a rabbit anti-mouse C-subunit antiserum.

two Ca cDNAs was placed under the control of the recApromoter and glO-L present in a plasmid containing theampicillin-resistance gene and ColEl origin ofreplication. Oneof these cDNAs encoded the wild-type 40-kDa C subunit ofPK-A containing Gly-2, while the other produced a variant thathad an Ala-2 -* Gly substitution. Note that this mutant Csubunit should not be a substrate for NMT (Table 1).

Pairwise combinations of the parental vectors and theirNMTI- and Ca cDNA-containing recombinant derivativeswere cotransfected into E. coli strain JM101, where they weremaintained by ampicillin and kanamycin selection.[3H]Myristate was used to label cultures of logarithmicallygrowing cells during sequential expression of yeast NMTIfollowed by Ca cDNA. Lysates prepared from the cultureswere subjected to SDS/PAGE and fluorography to examineradiolabeled fatty acid incorporation into protein. When theNMTJ and wild type Ca cDNA sequences were coexpressedin E. coli, a 40-kDa protein was metabolically labeled afteraddition of [3H]myristic acid to the culture medium (lane 4 ofFig. lb). This protein comigrated with purified C-subunitstandards. Labeling of the 40-kDa protein was absolutelydependent upon the presence of both NMTJ and wild type CacDNA. E. coli that expressed NMTJ but lacked C, cDNA andE. coli that lacked NMTJ but expressed Ca cDNA each failedto label the 40-kDa protein with tritiated fatty acid (lanes 6 and7 of Fig. lb, respectively). Moreover, the 40-kDa protein wasnot labeled in cells expressing NMTJ and the mutant Ca cDNAencoding the Ala-2-substituted C subunit (lane 5 of Fig. lb).Western blot analysis with rabbit polyclonal antisera raisedagainst the C subunit of mouse PK-A confirmed the presenceof equivalent amounts of the Gly-2- and Ala-2-containing40-kDa proteins in lysates prepared from E. coli strain JM101

with the wild-type and mutant Ca recombinant plasmids,respectively (Fig. le). The level of production of the two Csubunits was estimated to be 0.1% oftotal E. coli protein basedon the signal intensities of purified C-subunit standards in-cluded in the Western blot. NMT represented -0.2% ofE. coliproteins after induction. This value was calculated from theNMT activities in crude lysates and the specific activity ofpurified yeast NMT (14). Coexpression of NMTI and CacDNA had no deleterious effects on E. coli growth kineticsduring the induction period (data not shown).We excised the [3H]myristate-labeled 40-kDa protein from

an SDS/polyacrylamide gel and digested it with Pronase E toinvestigate the nature of the fatty acyl-protein linkage. Alabeled product was produced that comigrated with ourchemically synthesized [3H]myristoylglycine standard (31)on C18 reverse-phase HPLC (4). Together these data allowedus to conclude (i) that the C subunit of PK-A can bemyristoylated in a Gly-2-dependent manner only in E. colicells producing yeast NMT and (ii) that the endogenousmethionylaminopeptidase activity of E. coli (32) can removethe initiator methionine of the C subunit, thereby exposing itsGly-2 residue for NMT-catalyzed transfer of myristate. Thislatter result confirms earlier results identifying the NH2-terminal sequence of the C subunit synthesized in E. coli asGly-Asn-Ala-Ala ... (29).We estimated that the overall efficiency ofNMT-catalyzed

linkage of the myristoyl moiety to the C subunit in E. coli wasvirtually 100% by measuring the following three parameters:(i) the concentration of C subunit in E. coli lysates fromWestern blot hybridization analysis; (ii) the amount of[3H]myristate incorporated into the protein after excisingbands from SDS/polyacrylamide gels containing a known

i

7: :: ::..::

Biochemistry: Duronio et al.

_- af

Dow

nloa

ded

by g

uest

on

Mar

ch 6

, 202

1

Proc. Natl. Acad. Sci. USA 87 (1990)

amount of E. coli lysate proteins; and (iii) the final specificactivity of [3H]myristic acid in E. colill after labeling (13,uCi/nmol) (33).Reconstitution of Protein N-myristoylation in E. con Is Spe-

cific for C14 Fatty Acids. 10-Propoxydecanoic acid (11-oxamyristic acid) is an analog of myristic acid that has a similarchain length but a reduced hydrophobicity (comparable todecanoic acid) because of the substitution of an oxygen atomfor a methylene group at position 11 of the hydrocarbon chain(16). When lO-propoxydecanoic acid is incorporated intop60vsrc in vivo, it causes a significant redistribution of theprotein from membrane to cytosolic fractions (4). Metaboliclabeling experiments analogous to those described above with[3H]myristic acid indicated that the C subunit could also belabeled in a Gly-2-dependent manner when exogenous 10-[3H]propoxydecanoic acid or [3H]palmitate was added toNMT-producing E. coli cells (Fig. 1 d and c, respectively).Previous studies have suggested that palmitate must be met-abolically converted to myristate before incorporation intoN-myristoylated proteins (16, 31, 34). Pronase E digestion ofthe [3H]palmitate-labeled C subunit yielded a product thatcomigrated on C18 reverse-phase HPLC with [3H]myristoyl-glycine. Thus, the coupled E. coli expression recapitulates theremarkable specificity for fatty acyl-CoA chain length ob-served in S. cerevisiae and mammalian cells.At least two proteins of -45 and -55 kDa incorporated all

three exogenously added tritiated fatty acids in E. coli strainsthat expressed yeast NMT (Fig. 1 b-d, lanes 3-6). Cellswithout plasmids (Fig. 1 b-d, lanes 1) or cells carrying theparental vector lacking NMTJ (Fig. 1 b-d, lanes 2) did notincorporate label into these proteins. Since the apparentmolecular mass of these proteins is conspicuously similar tothe two forms of NMT produced in E. coli, we investigatedthe nature of the protein-tritiated fatty acid association.Pronase E digestion of these proteins labeled with either[3H]myristate or [3H]palmitate yielded a product that comi-grated with [3H]myristoylglycine on C18 reverse-phaseHPLC. The fact that myristoylglycine was detected, togetherwith the observation that neither intact yeast NMT nor itsproteolytically processed 45-kDa form contains a glycine atits NH2 terminus, allowed us to conclude that these tritiatedproteins do not arise from N-myristoylation ofNMT itself butrather from N-myristoylation of endogenous E. coli proteins.However, we cannot eliminate the possibility that a portionof the band intensities arises from a tight noncovalent orcovalent association of the tritiated fatty acids with NMTspecies. A search (15) of the National Biomedical ResearchFoundation Protein Sequence Database (Release 19.0) for E.coli protein sequences that begin with Met-Gly- and thereforemight be acylated by NMT did not reveal any of the appro-priate molecular mass. Even with these two "endogenous"protein substrates, a major advantage of the bacterial systemover eukaryotic systems is the absence of both endogenousNMT activity and substrates.

In summary, coexpression of NMT and its protein sub-strates in E. coli should facilitate analysis ofNMT structure-activity relationships, help to identify structural features ofits protein substrates that are necessary for N-myristoyla-tion, and provide insights about the role of the myristoylmoiety in the function of individual N-myristoylated pro-teins. The effect of modifying specific N-myristoylated pro-teins with heteroatom-containing analogs of myristate cannow be directly assessed by comparative studies of E.coli-synthesized nonacylated, myristoylated, and analog sub-stituted species. Mutant strains of E. coli deficient in the

metabolism of fatty acids may prove particularly useful inthese studies by improving our ability to deliver exogenousfatty acids and their analogs to the acylation apparatus.Finally, given the observation (16) that acyl-CoA binding toNMT affects the enzyme's affinity for various peptide sub-strates, coexpression ofNMT and its protein substrates in E.coli could provide a good functional assay for screening therelative efficiency of incorporation of different analogs into agiven target protein. In this sense, the N-myristoylationsystem described here may aid in the identification of usefulantineoplastic or antiviral (17) drugs.

This work was supported in part by grants from the MonsantoCompany and the National Institutes of Health (AI27179) to J.I.G.and by grants from the American Cancer Society (CD255) and fromthe Lucille P. Markey Foundation to S.S.T.; R.J.D. is the recipientof a Josiah P. Macy, Jr. Predoctoral Fellowship. R.O.H. was fundedby Medical Scientist Training Program Grant GM07200. W.Y. isfunded by a fellowship from the California Division of the AmericanCancer Society. L.W.S. was supported in part by U.S. Public HealthService Training Grant GM07313. J.I.G. is an Established Investi-gator of the American Heart Association.1. Wilcox, C., Hu, J.-S. & Olson, E. N. (1987) Science 238, 1275-1278.2. Towler, D. A., Gordon, J. I., Adams, S. P. & Glaser, L. (1988) Annu.

Rev. Biochem. 57, 69-99.3. Schultz, A. M., Henderson, L. E. & Oroszlan, S. (1988) Annu. Rev. Cell

Biol. 4, 611-647.4. Heuckeroth, R. 0. & Gordon, J. I. (1989) Proc. Natl. Acad. Sci. USA 86,

5262-5266.5. Cross, F. R., Garber, E. A., Pellman, D. & Hanafusa, H. (1984) Mol.

Cell. Biol. 4, 1834-1842.6. Kamps, M. P., Buss, J. E. & Sefton, B. M. (1985) Proc. Natl. Acad. Sci.

USA 82, 4625-4628.7. Resh, M. D. (1989) Cell 58, 281-286.8. Goddard, C., Arnold, S. T. & Felsted, R. L. (1989) J. Biol. Chem. 264,

15173-15176.9. Rein, A., McClure, M. R., Rice, N. R., Luftig, R. B. & Schultz, A. M.

(1986) Proc. Natl. Acad. Sci. USA 83, 7246-7250.10. Rhee, S. S. & Hunter, E. (1987) J. Virol. 61, 1045-1053.11. Gottlinger, H. G., Sodroski, J. G. & Haseltine, W. A. (1989) Proc. NatI.

Acad. Sci. USA 86, 5781-5785.12. Chow, M., Newman, J. F. E., Filman, D., Hogle, J. M., Rowlands, D. J.

& Brown, F. (1987) Nature (London) 327, 482-486.13. Marc, D., Drugeon, G., Haenni, A.-L., Girard, M. & van der Werf, S.

(1989) EMBO J. 8, 2661-2668.14. Towler, D. A., Adams, S. P., Eubanks, S. R., Towery, D. S., Jackson-

Machelski, E., Glaser, L. & Gordon, J. I. (1987) Proc. Natl. Acad. Sci.USA 84, 2708-2712.

15. Towler, D. A., Adams, S. P., Eubanks, S. R., Towery, D. S., Jackson-Machelski, E., Glaser, L. & Gordon, J. I. (1988) J. Biol. Chem. 263,1784-1790.

16. Heuckeroth, R. O., Glaser, L. & Gordon, J. 1. (1988) Proc. Natl. Acad.Sci. USA 85, 8795-8799.

17. Bryant, M. L., Heuckeroth, R. O., Kimata, J. T., Ratner, L. & Gordon,J. I. (1989) Proc. Natl. Acad. Sci. USA 86, 8655-8659.

18. Duronio, R. J., Towler, D. A., Heuckeroth, R. 0. & Gordon, J. 1. (1989)Science 243, 7%-800.

19. Saiki, R. K., Gelfand, D. H., Stoffel, S., Scharf, S. J., Higuchi, R.,Horn, G. T., Mullis, K. B. & Erlich, H. A. (1988) Science 239,487-491.

20. Olins, P. O., Devine, C. S., Rangwala, S. H. & Kavka, K. S. (1988)Gene 73, 227-235.

21. Feinstein, S. I., Chernajovsky, Y., Chen, L., Maroteaux, L. & Mory, Y.(1983) Nucleic Acids Res. 11, 2927-2941.

22. Chang, A. C. Y. & Cohen, S. N. (1978) J. Bacteriol. 134, 1141-1156.23. Uhler, M. D., Carmichael, D. F., Lee, D. C., Chrivia, J. C., Krebs,

E. G. & McKnight, G. S. (1986) Proc. Natl. Acad. Sci. USA 83,1300-1304.

24. Li, E., Locke, B., Yang, N. C., Ong, D. E. & Gordon, J. I. (1987) J. Biol.Chem. 262, 13773-13779.

25. Kunkel, T. A. (1985) Proc. Natl. Acad. Sci. USA 32, 488-492.26. Laemmli, U. K. (1970) Nature (London) 227, 680-685.27. Taylor, S. S. (1989) J. Biol. Chem. 264, 8445-8446.28. Carr, S. A., Biemann, K., Shoji, S., Parmelee, D. C. & Titani, K. (1982)

Proc. Natl. Acad. Sci. USA 79, 6128-6131.29. Slice, L. W. & Taylor, S. S. (1989) J. Biol. Chem. 264, 20940-20946.30. De Boer, H. A., Hui, A., Comstock, L. J., Wong, E. & Vasser, M. (1983)

DNA 2, 231-235.31. Towler, D. & Glaser, L. (1986) Biochemistry 25, 878-884.32. Sherman, F., Stewart, J. W. & Tsunasawa, S. (1985) BioEssays 3, 27-31.33. Silbert, D. F., Ulbright, T. M. & Honegger, J. L. (1973) Biochemistry 12,

164-171.34. Olson, E. N., Towler, D. A. & Glaser, L. (1985) J. Biol. Chem. 260,

3784-3790.

IINote that this latter measurement includes any myristic acid that isincorporated into phospholipid. Thus, our calculation of the effi-ciency of N-myristoylation represents a first approximation.

1510 Biochemistry: Duronio et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 6

, 202

1