Substrate specificity of Saccharomyces cerevisiae myristoyl-CoA: protein N-myristoyltransferase. Analysis of fatty acid analogs containing carbonyl groups, nitrogen heteroatoms, and nitrogen heterocycles in an in vitro enzyme assay and subsequent identification of inhibitors of human immunodeficienc

Covalent attachment of myristic acid (C14:0) to the amino-terminal glycine residue of a variety of eukaryotic cellular and viral proteins can have a profound influence on their biological properties. The enzyme that catalyzes this modification, myristoyl-CoA-protein N-myristoyltransferase (NMT), has been identified as a potential target for antiviral and antifungal therapy. Its reaction mechanism is ordered Bi Bi with myristoyl-CoA binding occurring before binding of peptide and CoA release preceding release of myristoylpeptide. Perturbations in the binding of its acyl-CoA substrate would therefore be expected to have an important influence on catalysis. We have synthesized 56 analogs of myristic acid (C14:0) to further characterize the acyl-CoA binding site of Saccharomyces cerevisiae NMT. The activity of fatty acid analogs was assessed using a coupled in vitro assay system that employed the reportedly nonspecific Pseudomonas acyl-CoA synthetase, purified S. cerevisiae NMT, and octapeptide substrates derived from residues 2-9 of the catalytic subunit of cyclic AMP-dependent protein kinase and the Pr55gag polyprotein precursor of human immunodeficiency virus I (HIV-I). Analysis of ketocarbonyl-, ester-, and amide-containing myristic acid analogs (the latter in two isomeric arrangements, the acylamino acid (-CO-NH-) and the amide (-NH-CO)) indicated that the enzyme's binding site is able to accommodate a dipolar protrusion from C4 through C13. This includes the region of the acyl chain occurring near C5-C6 (numbered from carboxyl) that appears to be bound in a bent conformation of 140-150 degrees. The activities of NMT's acyl-CoA substrates decrease with increasing polarity. This relationship was particularly apparent from an analysis of a series of analogs in which the hydrocarbon chain was terminated by (i) an azido group or (ii) one of three nitrogen heterocycles (imidazole, triazole, and tetrazole) alkylated at either nitrogen or carbon. This inverse relationship between polarity and activity was confirmed after comparison of the activities of the closely related ester- or amide-containing tetradecanoyl-CoA derivatives. Members from all of the analog series were surveyed to determine whether they could inhibit replication of human immunodeficiency virus I (HIV-I), a retrovirus that depends upon N-myristoylation of its Pr55gag for propagation. 12-Azidododecanoic acid was the most active analog tested, producing a 60-90% inhibition of viral production in both acutely and chronically infected T-lymphocyte cell lines at a concentration of 10-50 microM without associated cellular toxicity.

Covalent attachment of myristic acid (C14:O) to the amino-terminal glycine residue of a variety of eukaryotic cellular and viral proteins can have a profound influence on their biological properties. The enzyme that catalyzes this modification, myristoyl-CoA-protein N-myristoyltransferase (NMT), has been identified as a potential target for antiviral and antifungal therapy. Its reaction mechanism is ordered Bi Bi with myristoyl-CoA binding occurring before binding of peptide and CoA release preceding release of myristoylpeptide. Perturbations in the binding of its acyl-CoA substrate would therefore be expected to have an important influence on catalysis. We have synthesized 66 analogs of myristic acid (C14:O) to further characterize the acyl-CoA binding site of Saccharomyces ccrevisiae NMT. The activity of fatty acid analogs was assessed using a coupled in vitro assay system that employed the reportedly nonspecific Pseudomonas acyl-CoA synthetase, purified S. cerevisiae NMT, and octapeptide substrates derived from residues 2-9 of the catalytic subunit of cyclic AMP-dependent protein kinase and the Pr65gag polyprotein precursor of human immunodeficiency virus I (HIV-I). Analysis of ketocarbonyl-, ester-, and amide-containing myristic acid analogs (the latter in two isomeric arrangements, the acylamino acid (-CO-NH-) and the amide (-NH-CO)) indicated that the enzyme's binding site is able to accommodate a dipolar protrusion from C4 through C13. This includes the region of the acyl chain occurring near C6-C6 (numbered from carboxyl) that appears to be bound in a bent conformation of 140-160°. The activities of NMT's acyl-CoA substrates decrease with increasing polarity. This relationship was particularly apparent from an analysis of a series of analogs in which the hydrocarbon chain was terminated by (i) an azido group or (ii) one of three nitrogen heterocycles (imidazole, triazole, and tetrazole) alkylated at either nitrogen or carbon. This inverse relationship between * This work was supported by Grants AI-27179 and AI-30188 from the National Institutes of Health and Monsanto Company. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
( 1 To whom correspondence should be addressed: Dept. of Molecular Biology and Pharmacology, Box 8103, Washington University School of Medicine, 660 So. Euclid Ave., St. Louis, polarity and activity was confirmed after comparison of the activities of the closely related ester-or amidecontaining tetradecanoyl-CoA derivatives. Members from all of the analog series were surveyed to determine whether they could inhibit replication of human immunodeficiency virus I (HIV-I), a retrovirus that depends upon N-myristoylation of its Pr6SgW for propagation. 12-Azidododecanoic acid was the most active analog tested, producing a 60-90% inhibition of viral production in both acutely and chronically infected T-lymphocyte cell lines at a concentration of 10-50 PM without associated cellular toxicity.
The best characterized NMT is from Saccharomyces cereuisiae. The monomeric 53-kDa enzyme has been purified to The abbreviations used are: NMT, myristoyl-CoA:protein Nmyristoyltransferase; HIV-I, human immunodeficiency virus I; HPLC, high-performance liquid chromatography; CPE, cytopathic effect; EGTA, [ethylenebis(oxyethylenenitrilo)]tetraacetic acid; TCID50, dose of virus that infects 50% of cells in culture; TDsI, dose of analog that kills 50% of noninfected cells in culture; AZT, 3'azido-3'-deoxythymidine. homogeneity (Towler et al., 1987b). The single copy, haploid essential, S. cerevisiae NMTl gene has been meiotically mapped and isolated (Duronio et al., , 1991b. NMTl specifies a 455-residue protein that has no obvious primary structural similarity to any entry in current databases. The Saccharomyces enzyme has been efficiently expressed in Escherichia coli, allowing isolation of large quantities of active NMT (Duronio et al., 1990a(Duronio et al., , 1990bRudnick et al., 1990). Detailed in vitro kinetic analyses  indicate that its mechanism of catalysis is ordered Bi Bi with myristoyl-CoA binding to NMT occurring prior to binding of peptide and CoA release taking place prior to release of its acylpeptide product as follows. Protein N-myristoylation appears to be an irreversible covalent modification (Olson, 1988;James and Olson, 1989), although exceptions may exist (da Silva and Klein, 1990). Interestingly, the enzyme is able to catalyze the reverse reaction in vitro as follows, Myristoylpeptide + CoA + NMT + NMT-myristoyl-CoA + peptide albeit a t a rate that is significantly slower than that of the forward reaction . This deacylation, together with the results of biophysical studies (Rudnick et al., 1990;1991), provide additional evidence for a high affinity myristoyl-CoA-NMT intermediate.
Little is known about how this enzyme recognizes the coenzyme A moiety of its myristoyl-CoA substrate. Limited studies with 3'-dephospho-CoA suggest that the 3"phosphate is not necessary for peptide acylation (or deacylation of the myristoylpeptide) but may facilitate the formation of the acyl-CoA-NMT intermediate .
Considerably more is known about structure/activity relationships in the fatty acid chain of myristoyl-CoA. I n vitro kinetic studies of >90 fatty acid analogs with one or more oxygen or sulfur atoms, trans (E) or cis (Z) double bonds, triple bonds, or aromatic systems have helped define the basis for the enzyme's acyl-CoA selectivity (Heuckeroth et al., 1988Kishore et al., 1991;Gokel et al., 1992). These analyses indicate that hydrophobicity is less important than chain length. The acyl chain of myristoyl-CoA substrate appears to be present in a bent conformation with the principal bend (approximately 140-150") occurring near C5-C6 (numbered from carboxyl). The binding pocket seems to have three reference points: (i) at the carboxyl terminus (S-CoA); (ii) at the omega methyl group, and (iii) the bend which, in turn, affects the distance measurement in both directions from the bend. The acyl-CoA binding pocket possesses a sensor that detects the terminal methyl group. The apparent conical shape of this sensor also permits it to monitor bulk .
We have now synthesized 56 myristic acid analogs containing ketocarbonyl, ester, amide, acylamino acid, alkyl azido, and a variety of heterocycles including imidazole, triazole, and tetrazole. This broad family of compounds permitted us to explore the effects of polarity, basicity, and the related issues of branching and bulk on the physicochemical basis of the host-guest relationships. Biological studies of members of the panel of analogs revealed that 12-azidododecanoic acid was a potent inhibitor of HIV-I replication in both acutely and chronically infected T-lymphocytes at concentrations that were not associated with cellular toxicity.

EXPERIMENTAL PROCEDURES'
Compound Structures and Abbreviations-The single number following the abbreviation CO indicates the methylene group of myristic acid replaced by the carbonyl group. The two numbers that follow the abbreviations EST, AMD, and AAA indicate the two methylene groups which have been replaced. The numbers preceding IM, TRI, and TET indicate positional isomers. Sample compound structures and abbreviations are shown in Tables I, IV, and V.
Syntheses of Myristic Acid Analogs-Each of the ketocarbonyl compounds used in this study was prepared by one of four basic methods. These include the malonic ester acylation-decarboxylation sequence (CO-3); the acetoacetic ester synthesis ((20-4, CO-5, CO-13); enamine acylation followed by hydroxide-induced ring cleavage (CO-6, CO-7); and coupling of a Grignard-derived cadmium salt to an acyl chloride (CO-9, CO-10, . The ester analogs were prepared by mono-esterification of the appropriate diacid. Likewise, the amide derivatives were obtained by formation of the monoamide of a diacid. Acylamino acid containing analogs were obtained by acylation of an w-amino acid derivative. w-Azido substituted alkanoic acids were prepared by azide displacement on the corresponding w-iodoalkanoic acid. The thia-and oxa-azido alkanoic acids were prepared in a similar fashion from the appropriate oxa-or thia-wiodocarboxylic acid. The imidazole, triazole, and tetrazole derivatives were all synthesized by alkylation of the heterocycle using appropriate protection when necessary. Further details of the preparation, purification, and characterization of the analogs prepared for this study may he found in the miniprint. Synthesis of Radiolabeled O~tapeptides-Gly-Ala-Arg-[~H]Ala-Ser-Val-Leu-Ser-NH2 (GAR[3H]ASVLS-NH2) representing residues 2-9 of the HIV-I Pr55g"K was synthesized and purified as described in Kishore et al. (1991). Its specific activity was 1.2 Ci/mmol. Radiochemical and chemical purity was >95%. Synthesis and purification of Gly-Ser-['H]Ala-Ala-Ser-Ala-Arg-Arg-NHz (GS['H]AASARR-NH2; specific activity, 1.1 Ci/mmol) is detailed in Heuckeroth et al. (1990).
In Vitro Pseudomonas Acyl-CoA Synthetase Assay-Conditions used to monitor the efficiency of conversion of analogs to their CoA thioesters are described in Kishore et al. (1991). The 100-pl reaction mixture contained fatty acid (final concentration, 160 p~) , unlabeled CoA (1 mM), [3H]CoA (1 pCi, 1.2 Ci/mmol, see Kishore et al., 1991), dithiothreitol (3 mM), Triton X-100 (0.05%), Tris (5 mM, pH 7.4), MgClz (2.5 mM), EGTA (50 p~) , and Pseudomonas aeruginosa acyl-CoA synthetase (0.3 units/ml; Sigma). Following a 25-min incubation at 30 "C, an equal volume of ethanokacetic acid (1:l) was added. The mixture was cooled on ice for 10 min and then fractionated by ClS reversed-phase HPLC using gradient conditions detailed in Kishore et al. (1991). The amount of 'H-labeled analog-CoA produced was quantitated using an in-line detector  and compared to the amount of myristoyl-CoA produced in a parallel reaction. All assays were performed at least in duplicate.
In Vitro Discontinuous NMTAssay-A comprehensive description of this assay and the concentrations of each reactant is provided in earlier publications (Towler and Glaser, 1986;Duronio et al., 199Ob;Kishore et al., 1991). Briefly, an initial, single point assay was performed as follows: myristic acid or its analog (final concentration, 15 MM) was converted to its CoA thioester using 0.3 units/ml of Pseudomonas acyl-CoA synthetase. Following a 25-min incubation at 30 "C, E. coli-derived, purified S. cerevisiue NMT (final concentration, 0.5 pg/ml) was added together with either one of the two tritiated peptides (final concentration, 25 p~) .
After an additional 10-min incubation at 30 "C, the enzymatically generated acylpeptide was purified using Cla reversed-phase HPLC and quantitated using an inline scintillation counter. A reference control reaction was always run using myristic acid so that the amount of analog peptide produced could be expressed as a percentage of myristoylpeptide . Assays were run in duplicate or triplicate for each experiment.
Selected analogs were characterized further by more detailed kinetic studies. The apparent K, and V,,, were first determined using saturating concentrations of analog. Apparent acyl-CoA K, and V, were then calculated using the ['Hloctapeptide at its K,. Myristic acid was used as a reference control in each experiment. All V, data were expressed as a percentage of the V, obtained with myristoyl-CoA. All assays were performed in duplicate for each experiment. All experiments were repeated at least twice. Inhibition Assays-Several of the acylamino acid and amide containing myristic acid analogs were also tested as inhibitors of E. coliderived S. cerevisiue NMT using the in vitro discontinuous assay. These compounds were tested either directly (1-100 pM; AMD 617,516,415;AAA 617,516,415) or after conversion to the corresponding CoA derivatives (10 p~; AAA 13/12, 10/9,9/8, AMD 6/7, 516, 4/51, in the presence of 0.23 p~ ['Hlmyristoyl-CoA and 180 p~ unlabeled GNAAAARR-NHz (Towler et al., 1987a). Acute Virus Replication Assays-The effect of each analog on virus replication was assessed in two acute infectivity assays. The first assay was adapted from Pauwels et al. (1988). Assays were performed in 96-well tissue culture plates. CEM cells were grown in RPMI-1640 medium (GIBCO) supplemented with 10% fetal calf serum. Following treatment of the plate with polybrene (2 pg/ml), an 80-pl aliquot containing 1 X lo' cells was dispensed into each well of the plate.
One hundred p1 of RPMI-1640 plus analog (dissolved in dimethyl sulfoxide) was then added and the cells were incubated at 37 "C for 1 h. (Note that the final concentration of dimethyl sulfoxide did not exceed 1.5%). A frozen stock of HIV-I was thawed and diluted in culture medium to a concentration of 5 X lo' TCIDW/ml (TCIDM equals the dose of virus that infects 50% of cells in tissue culture). Twenty pl of the virus sample (containing 1000 TCIDW of virus) were added to wells containing analog and to wells containing only medium plus dimethyl sulfoxide (infected control cells). This results in a multiplicity of infection of 0.1 (MOI = number of infectious units/ number of cells in culture). Several wells received culture medium without virus (uninfected control cells). Likewise, the intrinsic toxicity of the test compound was determined by adding medium without virus to several wells containing analog.
Following addition of virus, cells were incubated at 37 "C in a humidified atmosphere containing 95% air/5% COZ for 7 days. Additional aliquots of test compounds were added on days 2 and 5. On day 7, the cells in each well were resuspended and a 100-pl aliquot of each suspension was removed for assay. Twenty pl of a 5 mg/ml solution of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide was added to each 100 pl of suspended cells. The mixture was incubated for 4 h at 37 'C in 95% air/5% COz. During this incubation, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide is metabolically reduced by living cells resulting in the production of an intracellular colored formazan product. One hundred pl of 10% sodium dodecyl sulfate in 0.01 N HCl was added to lyse the cells. Following an overnight incubation, the absorbance of each sample was determined at 590 nm.
The percentage reduction of the virus-induced cytopathic effect (CPE) by analog was determined using the following formula.
Absorbance of analog-treated infected sampleabsorbance of virus control Absorbance of cell control x 100 absorbance of virus control The dose that inhibits 50% of the cytopathic effect is referred to as IDso.
The second assay has been described in detail elsewhere (Bryant et al., 1989). It involves (i) infection of CD4+ H9 cells with HIV-I; (ii) addition of analog to serum-containing RPMI 1640 medium within 1 h after exposure of cells to the virus; (iii) replacement of the medium with fresh medium plus analog every 48 h; and (iv) measurement of virus production in days 8-10 medium by monitoring reverse transcriptase (RT) levels, and by determining p24 antigen concentrations (Bryant and Ratner, 1990). Data were referenced to three negative controls: cells treated with no additions, cells treated with ethanol alone (0.1%, the vehicle used to prepare stock solutions of the fatty acids for these experiments), or cells treated with a naturally occurring saturated fatty acid whose hydrophobicity is equivalent to that of the analog (Le. C100 or C120) but which is known not to be an active NMT substrate (Heuckeroth et al., 1988). Analog toxicity was assessed by (i) pulse labeling an aliquot of analog-treated and control cells on the last day of the experiment with ['Hlleucine or ['HI thymidine for 2-4 h and measuring incorporation into protein and DNA, respectively, and (ii) measuring viable cell number by trypan blue exclusion (Bryant et al., 1989).
Chronic Virus Replication Assay-Details of this assay are provided in Bryant et al. (1991). H9IIIB cells which chronically produce HIV-I were incubated with RPMI 1640 medium with or without analog (cell density, 2 X 106/ml in each well of a 48-well culture plate). After 2 days, the medium was removed and replaced with fresh medium. At the end of the fourth day of the experiment, medium was harvested, filtered (0.22 pm, Millipore) and assayed for reverse transcriptase and p24 virus antigen. Analog toxicity was assessed at the end of the treatment period by metabolic labeling with ['Hlleucine and ['HI thymidine (see above) and by measuring cell viability by trypan blue exclusion. Negative controls included treatment with medium alone plus or minus ethanol (0.1%) or C100. 13-Oxatetradecanoic acid was included as a positive control (Bryant et al., 1991).

RESULTS AND DISCUSSION
Our understanding of the interactions of purified S. cerevisiue NMT with myristoyl-CoA has resulted in part from in vitro studies of various structural analogs. As noted in the Introduction, these included compounds longer and shorter than myristate, compounds with double and triple bonds, and various aromatic residues. Certain of the aromatic subunits contained oxygen or sulfur. They were explored as were ether and thioether analogs (Heuckeroth et al., 1988Kishore et al., 1991). To extend these analyses, we decided to incorporate oxygen in its carbonyl form (ketones, esters, amides). The amides also contained nitrogen which prompted a study of the effects of this atom in a variety of forms.
Myristic Acid Analogs with Ketocarbonyl for Methylene Substitutions-A panel of 11 analogs was prepared in which each -CH2group of C14:O (from C3-Cl3) was replaced by the carbonyl function (see "Experimental Procedures"). The ketocarbonyl analogs differ from previously reported oxatetradecanoic acid analogs (Heuckeroth et aL, 1988;Kishore et al., 1991) by the location and oxidation state of the oxygen atom. The carbonyl oxygen protrudes from the main chain at a distance of -1.3 A in contrast to ether oxygen which is integral to the chain. The carbonyl group is a polar cylinder with electronegative oxygen at one terminus and electronrich T bonds connecting it to the main hydrocarbon chain. This group also exhibits fixed directionality, i.e. the oxygen of the carbonyl is coplanar with the attached carbon and its adjacent carbons. The sulfoxide, 11-thiatetradecanoic acid Soxide, is the only previously reported analog whose structure resembles that of these ketocarbonyl compounds. However, it is not a substrate for S. cerevisiae NMT in vitro. By contrast, 11-thiatetradecanoic acid and 11-oxatetradecanoic acid have activities which are, within experimental error, identical to that of myristate . The different activities of these three compounds could reflect differences in their hydrophobicities, polarities, and/or bulk. Replacement of carbonyl functions at C3 through C12 of myristic acid provided a panel of "probes" of defined directionality, bulk, and polarity which allowed us to "sweep" the acyl-CoA binding site of NMT and thereby infer its geometric, steric, and polar properties.
The coupled, discontinuous in vitro NMT assay requires an initial conversion of the analog to its acyl-CoA derivative by the "nonspecific" Pseudomonas acyl-CoA synthetase (Shimizu et al., 1980;Kishore et al., 1991). We, therefore, first measured the efficiency of this reaction using radiolabeled CoA and the bacterial enzyme. The extent of CoA acylation for 10 of the 11 compounds was similar to that of myristate: 88-139% (Fig. 1A). Only 3-oxotetradecanoic acid had a conversion efficiency less than 50%. Next, NMT-dependent transfer of unlabeled analog from CoA to a radiolabeled octapeptide substrate (GAR[3H]ASVLS-NH,) representing residues 2-9 of the gag polyprotein precursor encoded by HIV-I, was assessed using the single point assay described under "Experimental Procedures." The results are presented in Fig.  2. Fig. 2A compares the amount of analogpeptide produced with the amount of myristoylpeptide generated in a parallel incubation that contained identical concentrations of reactants. The data indicate that the carbonyl group is accommodated by the binding pocket at all positions surveyed. However, activity was position-dependent. Location of the ketocarbonyl at position 6 (near the presumed bend) produced the most active analog (CO-6) while displacement of the C=O by one carbon in either direction diminished activity (e.g. CO-5 = 175% of myristate; CO-6 = 250%; CO-7 = 150%). Activity generally diminished as the carbonyl was placed closer to the C14 position (CH3) of myristate. Similar results were obtained using a different radiolabeled octapeptide substrate (GS[3H] AASARR-NHJ, and selected members of this series (Fig. An unexpected observation was made with 4-oxotetradecanoic acid. Two acylpeptide species were noted after this analog was introduced into the coupled in vitro NMT assay. This occurred with either octapeptide substrate. By contrast, all other oxotetradecanoic acids surveyed produced a single peak of radioactivity (see Fig. 2 4 ) . NMT was required for generation of the two peaks: addition of 4-oxotetradecanoic acid to Pseudomonas acyl-CoA synthetase yielded a single acyl-CoA derivative (see Fig. 3).3

).
Since CO-4 contains two carbonyl groups (a ketone and a thioester), the possibility of base-catalyzed, intramolecular, aldol-type reactions exists. It, therefore, seems possible that the two peaks observed for CO-4 actually reflect the analog and a self-condensation product or even two self-condensation products. Thus far, we have been unable to confirm this possibility, although numerous plausible condensation products have been considered. It is also possible that CO-5 could undergo similar self-condensation reactions but no evidence of such a process has been observed. More detailed kinetic analyses (Table 11) revealed that the apparent peptide (GARASVLS-NH2) K, in the presence of saturating amounts of CO-6 was indistinguishable from that obtained with myristoyl-CoA, although the peptide V,,, was -3-fold greater with the analog. The apparent K , and V, for the analog-CoA were remarkably similar to myristoyl-CoA. Analyses of 5-oxo-and 7-oxotetradecanoic acid (CO-5 and CO-7) indicated that they had comparable kinetic properties: their K , for the enzyme was similar to myristoyl-CoA and their effects on peptide (GAR[3H]ASVLS-NH2) catalytic efficiency ( V,/K,) were essentially equivalent (see Table 11).
As was the case for 6-oxotetradecanoic acid (C0-6), 6oxatetradecanoic acid (0-6) showed the greatest activity in the analogous panel of ether derivatives. While the activities of ketocarbonyl analogs diminished as the functional group approached either the methyl or carboxyl terminus, the effect of moving the ether (C-0-C) unit was different when the ether linkage approached the carboxyl group a diminution in activity was observed as in the carbonyl series but no progressive decline was seen in the 0-6-0-13 analogs (compare A and C in Fig. 2).

Myristic Acid Analogs
Containing Ester Functional Groups-The ester functional group was incorporated into the myristic acid backbone by esterifying a series ofdicarbox- Screening assay to determine the substrate properties of myristic acid analogs with keto, ester, and ether functional groups. All assays were performed in duplicate or triplicate using [3H]octapeptides representing residues 2-9 of the HIV-I Pr55gag (GARASVLS-NH,, Ratner et al., 1985) or a derivative of residues 2-9 of the catalytic subunit of mouse PK-A (GNAAAARR-NH, Towler and Glaser, 1986). The concentrations of reactants in the single point discontinuous in vitro NMT assay are described under "Experimental Procedures." The amount of analogpeptide produced is expressed as a percentage of myristoylpeptide generated in a reference incubation conducted in parallel. The interassay variation in the amount of radiolabeled myristoyl-GARASVLS and myristoyl-GNAAAARR produced in these reference incubations was <30% . ylic acids. Thus, the orientation of the ester functional group (-OC-OR) is opposite to that of the myristic acid carboxyl (-CO-OH). The ester functional group is, in a sense, a composite of the carbonyl and ether. Like the carbonyl group, there is a polar oxygen protruding at a distance of -1.3 A from the main hydrocarbon chain. The cylindrical bulk of the carbonyl portion of the ester group is identical to that of an isolated carbonyl. The adjacent oxygen is less basic and less flexible than the corresponding ether oxygens in the oxatetradecanoic acid series because of resonance between it and the carbonyl group. Thus, several comparisons are possible: the ether with the ester and the ester with the carbonyl compounds. Such comparisons allow one to assess the effects of hydrophobicity on substrate activity, ester being the most hydrophilic of the three followed by carbonyl and ether. The effects of rigidity can also be assessed the ester functional group constrains the positions of five contiguous atoms, the carbonyl four, and the ether three. Finally, since the polarity of these three functional groups varies, an additional comparison is possible. The ether functional group is the least polar, but the relative polarity of ester versus carbonyl will depend upon the context: the ester carbonyl is less reactive toward nucleophiles than is an isolated carbonyl but is a stronger Lewis base donor at the oxygen end.
The efficiency of conversion of each of the 10 ester analogs to its CoA thioester is at least as good as that of myristate (with the exception of EST3/4, Fig. 1B). Fig. 2B compares the activities of the esters in the single point NMT assay using the HIV-I-derived octapeptide substrate. Interestingly, monoheptyl adipate (EST6/7) was the most active ester analog. Moving the ester functional group closer to the carboxyl end of the fatty acid produced a more dramatic reduction in activity than was observed with either the ether or the carbonyl group (which are the elements that comprise the ester). The activity profile of the position 6 to position 13 ester analogs resembled the ether analogs more closely than the Comparison of the leftand right-hand columns demonstrates that the two species obtained with 4-oxotetradecanoic acid were not generated by the acyl-CoA synthetase but rather were dependent upon addition of purified E. coli-derived S. cereuisiue NMT to the reaction mixture. Note also that the appearance of the two species was a phenomenon unique to this oxotetradecanoic acid analog: e.g. unique analogpeptide peaks are seen with 5-OXO-, 6-oxo-, and 7-oxotetradecanoic acid derivatives as well as all the other ketocarbonyl compounds analyzed (data not shown). carbonyl derivatives (Fig. 2
Taken together, these results suggest that the enzyme can tolerate relatively nonpolar functional groups, even protruding ones, throughout most of the acyl chain. It should be noted that the position of the carbonyl group relative to the carboxyl will determine whether these two carbonyl groups are pointed in the same, or opposite, directions (assuming a fully extended conformation of the myristic acid chain). Thus,  a carbonyl group at position 5 will be oriented with a dihedral angle of zero (synclinal) with respect to the Cl carbonyl and at C6 the carbonyl angle will be 180" (antiperiplanar) in the absence of other forces. There may, however, be severe constraints placed upon the conformation by the residues which form the acyl-CoA binding site of S. cerevisioe NMT.
Our analog panel revealed one other positional effect (Fig.  4). Hydrophobicity, as judged by acylpeptide retention time on the CIS reversed-phase column, varied as a function of distance between the carboxyl and functional group. Thus, when the ester group in the analogGARASVLS was moved farther from the amide linkage, retention time decreased in a monotonic fashion. For example, EST3/4-peptide behaved on this chromatographic matrix as if it were a dodecanoylpeptide, while the ESTll/lZ-peptide resembled a decanoylpeptide. Comparable differences in hydrophobicities were apparent with the corresponding oxa-and thia-tetradecanoylpeptides (e.g. see Kishore et al., 1991 for an analysis of 3-oxatetradecanoyl-through 13-oxatetradecanoylGARASVLS-NHJ. It is unclear at the present time why such dramatic differences in retention time would be caused by "simple" displacement of functional groups in an acylpeptide that has high overall polarity. The demonstration of such sensitivity may provide a powerful model system for (i) operationally defining the conformation of acyl chains in acylpeptides; (ii) exploring the contribution of acyl chains to the interactions of acylpeptides and acylproteins with model or naturally occurring membranes; and (iii) describing the environment "experienced" by the acyl chains within these membrane systems.

Analyses of the Peptide Functional Group in Two Isomeric
Arrangements-The ability of the enzyme to accommodate the carbonyl group either in its "simple" form or as part of the ester functional group raised the question of how varied the carbonyl-containing functional group could be. In addition, we wondered whether or not the enzyme would be "confused" by the presence of an amide functional group. Although the enzyme possesses two functionally distinguishable binding sites (Towler et al., 1987a(Towler et al., , 1987bHeuckeroth et al., 1988Heuckeroth and Gordon, 1989;Rudnick et al., 1990Rudnick et al., ,1991Kishore et al., 1991), their structural relationships have not been well characterized. The introduction of a peptide functional group (i.e. amide) in the fatty acid chain, challenges the system to distinguish whether it is a component of a peptide or a myristic acid analog. Therefore, two series of analogs were synthesized (i) the amides (AMD) -NH-CO-and (ii) the acylamino acids (AAA) in which the amide functional group is oriented -CO-NH-. The size and shape of these amide derivatives are expected to be similar to the esters while their polarity and rigidity (due to resonance) are greater than the esters, irrespective of the amide's isomeric arrangement. All nine of the amide derivatives surveyed were substrates for the acyl-CoA synthetase (Fig. IC). However, as with EST3/ 4, AMD3/4 was barely accommodated. Unlike EST4/5 and EST5/6, the corresponding amide analogs (AMD4/5 and AMD5/6) had activities of <20% of myristate. Any positioning of the amide group closer to the w terminus yielded analogs that were considerably more active substrates for the synthetase (activity, 55 & 12% of C14:O). Despite the relatively efficient conversion of AMD6/7-AMD12/13, none of their CoA derivatives yielded detectable levels of analogpeptide after incubation with NMT (limits of detection 25% of myristoylpeptide). AMD3/4-516 appeared to be poor substrates for both acyl-CoA synthetase and NMT (i.e. no analogpeptide production was detected in the single point assay).
Of the eight acylamino acid analogs tested, only three were converted to their CoA derivatives at measurable levels. The three analogs in which the functional group was most remote from carboxyl were the only ones thioacylated. Fig. 1D illustrates the sharpness of the positional boundary between active and inactive substrates (compare the remarkable and dramatic difference in activities between AAA7/8 and AAA8/9).
The CoA thioesters of AAA8/9-AAA12/13 were further characterized in the discontinuous in vitro NMT assay. None yielded detectable levels of analogpeptide (data not shown).
Together, these results allowed us to draw two conclusions. First, irrespective of the orientation of the isomeric amide groups, the CoA derivatives are not substrates for S. cerevisiae NMT. Second, the amides were generally thioacylated whereas the acylamino acids generally were not. Given the ordered Bi Bi reaction mechanism of NMT, failure of these compounds to yield analogpeptide could reflect their inability to bind to the enzyme's acyl-CoA binding site or their failure to be transferred to bound peptide. We examined these possibilities as follows: the fatty acid or the enzymatically generated analog-CoA was tested for its ability to inhibit transfer of [3H]myristate from myristoyl-CoA to a well characterized (Towler et al., 1987a(Towler et al., , 1987b) octapeptide substrate derived from residues 2-9 of the catalytic subunit of CAMP-dependent protein kinase (GNAAAARR-NHp). A 4-400-fold molar excess of AMD4/5, AMD5/6 or AMD6/7 (or AAA4/5, AAA5/6, or AAA6/7) over myristoyl-CoA, did not reduce the amount of [3H]myristoylpeptide generated compared to control incubations containing no analog (n = 3 independent experiments with each assay done in duplicate). In addition, no inhibition of NMT activity was observed with a 40-fold molar excess (relative to myristoyl-CoA) of the CoA derivatives of AAA8/ 9-12/13 or AMD4/5-6/7 (n = 2 experiments). Thus, these compounds fail to bind to purified, E. coli-derived, S. cerevisiae NMT in vitro.
Myristic Acid Analogs Containing Terminal Azido Groups-The azide function is unique in organic chemistry. The three nitrogen atoms, N-=N=N'-are a less reactive linear array than such linear functional groups as -N=C=O or -N=C=S. The -NB functional group is a ?r electron-rich cylinder although it is attached at an angle to carbon. Azide is relatively nonpolar: the dipole moment for phenylazide (C6HB-NB) is -1.440 (Debye) compared to C,H,-C=N, where g = -4.050. The molecular volume of azide is also relatively small. The conformational preference (-AGO) for the equatorial position of cyclohexane is 1.70 kcal/mol for -CHs and 1.75 kcal/mole for -CH,CH3. Azide has the cylindrical symmetry of cyano (NsC-) or acetylene (HC=C-) which show conformational preferences of less than 0.5 kcal/mol.
Compared to a simple hydrocarbon (i.e. (CH2),CH3), azide is more reactive, more polar, and smaller, but the polarity and reactivity of the system are not dramatic. Azide occupies less volume than propyl but is similar in overall length. It is a poor H-bond acceptor although superior to a simple hydrocarbon.
Like the acetylene linkage that has previously been explored , azide, when attached to carbon, is a linear array of four atoms. Unlike the acetylene group, the azide function must be at the end of the alkyl chain rather than within it. Azide was included at the w terminus of four N3-(CH),-COOH analogs (where n = 8, 10, 11, and 12). In contrast to the acylamino acid derivatives, each of these compounds was an excellent substrate for the Pseudomonas acyl-CoA synthetase (percentage of conversion to analog-CoA = 150-200% that of myristate, see Table 111). 12-Azidododecanoic acid (CIPN3) has the same overall length as pentadecanoic acid and was the most active substrate in the single point assay with either of the two octapeptides (Fig. 5). Detailed kinetic studies (Table 111) confirmed these observations. The relative activities of the azide derivatives were similar to that observed with saturated fatty acids of equivalent chain length (i.e. C12:0, C14:0, C15:0, and C160, see Kishore et al. (1991)). These data indicate that the nonbulky, nonpolar, and nonhydrogen bonding properties of the azide function emulate the methyl terminus of a saturated fatty acid of comparable chain length.
Nitrogen Heterocycle Containing Myristic Acid Analogs-Previous studies of 10-(2-thienyl)decanoic acid and GSAA-SARR-NHz or GARASVLS-NH, indicated that it was more  5. Comparison of the activities of azido-fatty acid analogs having different chain lengths. The numbers inparentheses refers to the retention time (in min) of the analogpeptide on the CIS reversed-phase column relative to the retention time of the corresponding myristoylpeptide. The negative numbers refer to the fact that the analogpeptide was retained for a shorter period of time using the HPLC conditions described in Kishore et al. (1991). See Fig. 4 for the relative retention time of heptanoyl-through hexatetradecanoylpeptides.
active than myristate, whereas 12-(2-thienyl)dodecanoic acid was considerably less active . In contrast, 12-(2-furyl)dodecanoic acid showed almost identical activity to myristate (Table IV and Kishore et al., 1991). A final group of analogs was therefore synthesized in which the hydrocarbon chain was terminated by a nitrogen heterocycle. The three heterocycles, imidazole, triazole, and tetrazole, are all five-membered rings and contain 2, 3, or 4 nitrogen atoms, respectively. The heterocycles were alkylated at either nitrogen or carbon. The overall shape is similar for 0-, s-, and Ncontaining heterocycles.
We anticipated that since C-alkylation, rather than Nalkylation, of the nitrogen heterocycles results in a free NH residue, hydrophilicity and polarity would be greater than for the N-alkylated isomers. The results of the analog surveys described above and in earlier publications (Heuckeroth et al., 1988Kishore et al., 1991) have suggested an inverse correlation between the overall polarity of the analog (exclusive of protein) and its activity in the in vitro NMT assay. This led us to predict that the N-alkylated isomers would be less active than the C-alkylated compounds. All of these compounds were substrates for the Pseudomonas acyl-CoA synthetase (activity varied between 20 and 60% of myristate, Table IV). Whenever a free NH was present, the activity of the analog CoA in the presence of NMT and either GSAA-SARR-NH2 or GARASVLS-NH2 was far lower than the Nalkylated isomer. For example, imidazole derivatives of undecanoic and dodecanoic acid were more active when Nalkylated (compare analogs 2-IM-11 and 1-IM-11 with 2-1" 12 and 1-IM-12 in Table IV). The tetrazole derivatives of dodecanoic acid (compounds 5-TET-12 and 1-TET-12) provide a dramatic illustration of this difference: the activities of these two compounds differ by an order of magnitude in the presence of either GSAASARR-NH, or GARASVLS-NH, even though they have comparable reactivity in the acyl-CoA synthetase assay (31-49% that of myristate, see Table IV).
Conclusions from the Coupled in Vitro NMT Assay-These studies indicate that the acyl-CoA binding site of S. cerevisiae NMT has a remarkable spectrum of sensitivities. Analyses of the ketocarbonyl-, ester-, and amide-containing analogs indicate that the binding site is able to accommodate a dipolar protrusion from C4 through C13 of myristoyl-CoA analogs. This protrusion is tolerated without compromising activity on either side of the main chain and even where the chain is bent in the vicinity of C5-C6. In a qualitative sense, the  activities of NMT's acyl-CoA substrates diminish with increasing polarity. This relationship was apparent in analogs containing nitrogen heterocycles in which Cand N-alkylated isomers could be compared. Comparison of the closely related ester-or amide-containing tetradecanoyl-CoAs further confirm this inverse relationship between polarity and activity.
The ketocarbonyl-, ester-, azido-, and nitrogen heterocyclecontaining analogs were all substrates for the reportedly nonspecific Pseudomonas acyl-CoA synthetase. However, analyses of isomeric amide-containing tetradecanoic acid analogs revealed a striking sensitivity of the synthetase for the orientation of the carboxyl group in a fatty acid. This sensitivity extends from C3 to C8 but is lost thereafter. The fact that none of the amide (AMD, AAA) CoA derivatives can bind to NMT, in contrast to the corresponding EST-CoA derivatives, suggests a fundamental difference in the way each enzyme senses its substrate.
Biological Activities of Myristic Acid Analogs-Recent metabolic labeling studies using tritiated oxatetradecanoic acids with oxygen substitutions at C6, C11, and C13 and several cultured mammalian cell lines indicated that these low molecular weight (Mr 230-250) compounds readily traverse the cell membrane and are substrates for both acyl-CoA synthetase and NMT Johnson et al., 1990). Analogs are selectively incorporated into distinct yet overlapping subsets of cellular N-myristoylproteins. This selective, analog-specific incorporation presumably arises because of the cooperative interactions that occur between NMT's acyl-CoA and peptide binding sites and the enzyme's ordered Bi Bi reaction mechanism (i.e. different analog CoAs Substrate Specificity of N-Myristoyltransferase Triazolyl and tetrazolyl-substituted 1-TRI-12 1-TRI-12A 1-TRI-13 2-TRI-13 TET-12 TET-13 1-TET-12 Analogs See "Experimental Procedures" for details of this assay. *Dose of analog that produces a 50% reduction in the number of viable CME lymphocytes (cells were not e T, toxic; defined as producing a >50% killing of uninfected cells at this concentration. infected with HIV-I when determining the TDm of analog).
-, not determined. produce different effects on the peptide binding site). For (selective incorporation and selective perturbation of funcsome cellular N-myristoylproteins, incorporation leads to an-tion) probably accounts for their lack of cellular toxicity. alog-specific and -dependent redistribution from membrane Some, but not all, oxatetradecanoic acid analogs can inhibit to cytosolic fractions (i.e. a given protein may undergo redis-replication of HIV-I in acutely and chronically infected CD4+ tribution with one, but not another, analog depending upon H9 cells (Bryant et al., 1989(Bryant et al., ,1991. 13-Oxatetradecanoic acid the location of the heteroatom for -CH2substitution, com-is the most potent member of this group of compounds. Its pare with Johnson et al. (1990)). This dual level of selectivity maximum inhibitory effect was observed at 20-40 p~ and was not accompanied by cellular toxicity. Its mechanism of action involves, at least in part, incorporation into the Pr55gag and nef proteins of HIV-I (Bryant et al., 1991). The 13-oxatetradecanoyl-containing gag polyprotein precursor undergoes redistribution from membrane to cytosolic fractions and a reduction in its proteolytic processing which is possibly due to an effect of the analog on dimerization of gag-pol and subsequent release of viral protease (Naria et al., 1989;Bryant et al., 1991).
These results suggested that cellular acyl-CoA synthetase and NMT activities could be exploited to deliver these analogs to target proteins and indicated that HIV-I replication was a sensitive and useful model system for assaying such events. We, therefore, screened members of the panel of 56 analogs described above in an acute viral replication assay to infer whether they may function as alternative substrates for these two enzymes and whether their incorporation into viral/ cellular proteins could interfere with viral assembly. Since Pr55gag and nef are both substrates for S. cereuisiae NMT (Jacobs et Bryant et al., 1991), we also tried to ascertain whether the activity of an analog in the coupled in uitro enzyme assay containing Pseudomonas acyl-CoA synthetase, S. cereuisiae NMT, and GARASVLS-NH2 could accurately forecast its antiviral activity.
The results of this antiviral screen are presented in Table  V. Compounds were tested at concentrations of 1, 3, and 10 pg/ml (approximately 4-40 p~) using acutely infected CEM cells and an assay adapted from Pauwles et al. (1988, see "Experimental Procedures"). Inhibition of viral replication was quantitated by observing the percentage of reduction of virus-induced cytopathic effect (CPE). Several of the ketocar-bony1 containing analogs that were good substrates for Pseudomonas acyl-CoA synthetase and S. cereuisiae NMT (C013, C08, C06) produced 30-40% reductions in virus-induced cytopathicity at a concentration (40 JLM) that was 2-4-fold lower than the dose that killed 50% of noninfected cells (defined as the TD50). 3-Oxotetradecanoic acid (C03) which is the least active NMT substrate in the series (Fig. 2 A ) was the least active in the acute viral replication assay ( 4 0 % reduction in CPE. at 40 pM, a dose close to its TD50, see Table V). The most active ester analog in the coupled in uitro NMT assay, monoheptyl adipate (EST6/7, see Fig. 2B) had no antiviral activity, as was the case for most of the 10 analogs in this series. It is interesting to note that of the two EST compounds with modest, specific inhibitory effects (ESTlZ/ 13 and EST3/4, Table V), the latter was a very poor substrate for both the prokaryotic acyl-CoA synthetase and the yeast acyltransferase (see Figs. 1B and 2B). As expected from the in uitro assay results, analogs containing the peptide functional groups in either of the two isomeric arrangements were generally inactive ( 4 0 % inhibition of CPE at concentrations up to 40 p~) .
The only exception appeared to be AAA6/7 which produced a 24% reduction in CPE at a dose that was 10-fold below its TDso. The triazoles tested were quite toxic at low concentrations (the TD50 ranged from <4-100 p~) . The analogs surveyed with tetrazolyl or imidazoyl groups were inactive or toxic, except for 12(N-l-methyl-2-imidazoyl) dodecanoic acid (cf . Table IV) which produced a modest 31% reduction in CPE at 40 JLM, a concentration which was 6-fold lower than its TD5o (Table V).
The most active compound of the 45 surveyed in the acute viral propagation assay was 12-azidododecanoic acid (12AZ): at 4 pM it produced an -80% reduction in cytopathic effect. The IDSO of 12AZ was <4 JLM while its TDso was 184 p~. These later values were comparable to those of 13-oxatetradecanoic acid (IDGo (4 p~, TDSo = 90 PM). The antiviral

FIG. 6. Effect of myristic acid analogs containing terminal azido groups on replication of HIV-I in H9 cells. A, H9 cells were acutely infected with HIV-I (see Bryant et a!., 1989) and treated
with myristic acid analogs, decanoic acid (ClO:O, hydrophobicity similar to that of 11-azidoundecanoic acid (11-AZ) and 12-azidododecanoic acid (12-AZ), see Fig. 4), or AZT at the indicated concentrations in serum-containing medium for 10 days. Medium was changed every 2 days. The antiviral effect of these compounds was determined by measuring reverse transcriptase activity (Poiesz et al., 1980) in cell culture supernatants prepared on the last day of the experiment (ie. containing any virus produced between days 8-10). AZT was used as a positive control. No additions or 0.1% ethanol (its concentration in media containing 100 p M analog) were employed as negative controls (earlier experiments indicated ethanol had no effect on virus production; cf. Bryant et al., 1989 and data not shown). The mean plus S.E. are shown; triplicate samples were assayed, all assays were performed in triplicate. B, the same media collected for reverse transcriptase assays were also used for measurement of the p24 viral antigen by enzyme-linked immunosorbent assay (Bryant et al., 1989). Note that the concentration of p24 antigen in cultures incubated with serum-containing media alone or with medium plus 0.1% ethanol was 2 pg/ml. C and D, the toxicity of 11-azidoundecanoic acid and 12azidododecanoic acid was assessed by measuring viable cell number on the last day of the experiment (viability defined by trypan blue exclusion, control = cells incubated with medium alone, C ) and by incorporation of [3H]leucine into cellular protein after a 4-h pulse labeling (results expressed as a percentage of the incorporation observed in untreated cells, D). activity of 12-azidododecanoic acid contrasted with the lack of activity of 13-azidotridecanoic acid and the modest reductions in CPE produced by 11-azidoundecanoic acid (16% at 4 p~, 35% at 10 PM, TDGo = 100 KM, see Table V). The great differences in biological activities of these analogs was not anticipated from their relative activities in the coupled in vitro NMT assay system (Fig. 5 ) .
To verify this finding, a second assay for acute viral infection was employed that used a different T-lymphocyte cell line (H9) and different parameters for measuring viral replication (reverse transcriptase activity and p24 antigen accumulation in culture supernatants). The results (Fig. 6) confirmed that 12-azidododecanoic acid inhibited viral replication in acutely infected cells more effectively than 11- an analog previously shown to inhibit viral propagation in this assay system (Bryant et al., 1991), decanoic acid (ClO:O, 50 pM), AZT (5 p M ) , or ethanol (0.1%). Chronically infected cells were treated with the compounds over a 12-day period using a protocol described in an earlier publication [Bryant et al., 1991). Medium containing serum and the compound was replaced every other day. On day 12, medium was harvested and assayed for reverse transcriptase activity ( A ) and p24 antigen ( B ) . Cell viability was assessed by trypan blue exclusion ( C ) and by incorporation of [3H]leucine (D). Control = H9IIIB cells that had been incubated with serum-containing medium alone. Reverse transcriptase and p24 assays were performed on triplicate samples and the results expressed as the mean plus S.E. Measurements of cell viability were done on duplicate samples and the results were averaged. azidoundecanoic acid (60% uers'sus <lo% at 10 p~) and with a dose-response (from 1 to 100 p~) that was comparable to that of 13-oxatetradecanoic acid (see Fig. 6, A and B ) . The antiviral effect of 12-azidodecanoic acid observed at the concentrations tested was not accompanied by significant cellular toxicity, as judged by the number of viable cells remaining after 10 days of treatment (no detectable differences from control, Fig. 6D) or by the incorporation of tritiated leucine into newly synthesized proteins (no difference at 10 pM, 20-30% inhibition at 50-100 p~, Fig. 6C).
We do not have any labeled 12-azidododecanoic acid to perform metabolic labeling studies that might define its mechanism of action. However, we were able to confirm that this analog, like 13-oxatetradecanoic acid, can inhibit virus production by 60-70% (at 50 p~) in chronically infected H9IIIB cells without appreciable associated cellular toxicity (less than 20% reductions in viable cell count or tritiated leucine incor-poration, Fig. 7). This contrasts with the inability of AZT, an inhibitor of viral reverse transcriptase, to produce reductions in virus production in this model assay system (see Bryant et al., 1991).
stirred at -40 OC for 15 min and the temperature was raised to -25 OC. After sumng for 1.5 The reaction mlxture was cooled to -40 "C and added cold 2N HCI 14.5 mL) and then paured onto a mlxture of crushed Ice. water ( I O mL) and EtOAc I25 mLI. After repeated extracUOnS with EtOAc. the pH of the aqueous phase was brought to 3.5 with IN NaOH. when a solld began to separate. It was filtered. washed thoroughly with water and acetonluile and finally