A new, robust, and nonradioactive approach for exploring N-myristoylation.

Myristoyl-CoA (CoA):protein N-myristoyltransferase (NMT) catalyzes protein modification through covalent attachment of a C14 fatty acid (myristic acid) to the N-terminal glycine of proteins, thus promoting protein-protein and protein-membrane interactions. NMT is essential for the viability of numerous human pathogens and is also up-regulated in several tumors. Here we describe a new, nonradioactive, ELISA-based method for measuring NMT activity. After the NMT-catalyzed reaction between a FLAG-tagged peptide and azido-dodecanoyl-CoA (analog of myristoyl-CoA), the resulting azido-dodecanoyl-peptide-FLAG was coupled to phosphine-biotin by Staudinger ligation, captured by plate-bound anti-FLAG antibodies and detected by streptavidin-peroxidase. The assay was validated with negative controls (including inhibitors), corroborated by HPLC analysis, and demonstrated to function with fresh or frozen tissues. Recombinant murine NMT1 and NMT2 were characterized using this new method. This versatile assay is applicable for exploring recombinant NMTs with regard to their activity, substrate specificity, and possible inhibitors as well as for measuring NMT-activity in tissues.

Hilden, Germany) was used for transformation and synthesis of the recombinant proteins according to manufacturer's ins tructions. Recombinant Nmt1 and Nmt2 were purifi ed under native conditions using Ni-NTA agarose columns according to the specifi cations of the producer (Qiagen). To detect the fractions containing the enzyme, protein samples were separated by a dis continuous SDS-PAGE. The gels were either stained with Coo massie Brilliant Blue (Merck, Darmstadt, Germany) or blotted onto nitrocellulose membranes (GE Healthcare, Freiburg, Germany). The membranes were incubated with anti-(His) 6 -tag antibody (kindly provided by Hanswalter Zentgraf) followed by incubation with horseradish peroxidase conjugated anti-mouse secondary anti body (1: 10,000 dilution; Santa Cruz, Heidelberg, Germany). For development of the blots, an en hanced chemiluminescence system (GE Healthcare) was used according to manu-facturer´s instructions.

Preparation of mouse tissue extracts
C57BL/6 male mice were euthanized by cervical dislocation. Tissues were homogenized in digitonin lysis buffer (20 mM HEPES-NaOH buffer [pH 7.4] containing 25 mM KCl, 250 mM sucrose, 2 mM MgCl 2 , with freshly added 1% digitonin [Sigma], complete protease inhibitors [Roche], and 0.5 mM DTT) using Ultra Turrax T25 (Ika, Koenigswinter, Germany). The homogenate was centrifuged at 2,000 g for 20 min at 4°C, and the supernatant was further centrifuged at 39,000 g for 1 h at 4°C. The cytosolic fraction (supernatant) obtained was used immediately. Protein concentrations were determined by Bradford assay (Invitrogen, Karlsruhe, Germany) using BSA as protein standards. Animal procedures were approved by local authorities.
The vital dependency on protein myristoylation in the aforementioned human pathogens and tumors emphasizes the importance of further insight into myristoylationrelated processes and makes NMT a potential target for antimicrobial and antineoplastic chemotherapy ( 23,25,34 ).
A central tool for accomplishing these goals is a robust, reliable, and user-friendly assay to measure NMT activity. Methods for detecting NMT activity have involved radioactive and nonradioactive assays, such as use of reversephase HPLC, and the so-called "continuous assays" in which enzymatic activity is detected while the reaction is progressing. Radioactive methods are based on the quantifi cation of [ 3 H]myristate incorporation into peptides, whereas continuous assays depend on the spectrophotometric or fl uorometric detection of products released during myristoylation (e.g., HS-CoA) ( 1,4,(35)(36)(37)(38)(39)(40). A nonradioactive assay based on ELISA has been described using an antibody against N-myristoyl-glycine ( 41 ). Continuous and ELISAbased assays are more suitable for inhibitor screening. In contrast, radioactive and HPLC-based approaches are distinguished by a higher sensitivity and substrate specifi city; however, they are expensive and time consuming and require complex technical skills.
Here we describe a new, versatile, nonradioactive ELISAbased method for measuring N-myristoyltransferase activity in vitro. We validate this assay using various negative controls and HPLC analysis. We demonstrate that it is applicable to exploring recombinant NMTs with regard to their activity, substrate specifi city, and possible inhibitors. In addition, we show that the assay is suitable for measuring endogenous NMT activity in fresh or frozen tissues.

HPLC
Analytical HPLC-ESI-MS was conducted on a Hewlett-Packard 1090 liquid chromatograph (Agilent Technologies, Waldbronn, Germany) fi tted with a reverse-phase C18 Gemini column (250 mm × 4 mm i.d., 5 µm; (Phenomenex Ltd, Aschaffenberg, Germany). The mobile phase consisted of 15% acetonitrile in double distilled water (solvent A) and 10% double distilled water in acetonitrile (solvent B) with the following gradient profi le: 100% A for 5 min; reduced to 50% A over 14 min; 0% A over 1 min, continuing at 0% A until completion of the run. The fl ow rate of the mobile phase was constantly 1.0 ml/min. The injection volume was 1.0 µl and myristoylated compounds in the eluent were detected at 214 nm with a diode-array UV detector (Hewlett-Packard 1040M).
Mass spectra in the negative-ion mode were generated under the following conditions: fragmenter voltage = 100 V, capillary voltage = 2500 V, nebulizer pressure = 30 psi, drying gas temperature = 350°C, m/z = 100-1,500. Instrument control and data handling was by means of a Hewlett-Packard Chemstation operating in the Microsoft Windows software environment.

The principle of the NMT-azido-ELISA assay for measurement of NMT activity
Our aim was to develop an assay that would allow measurement of NMT activity in tissues and cell lysates as well as in experiments dealing with recombinant NMTs, enzyme kinetics, substrate specifi city, and inhibitors. The assay should be easy to establish and suitable for high-throughput applications. To address these issues, an ELISA-like approach ( Fig. 1 ) was established. As acyl donor, azido-dodecanoyl-CoA, which represents a broadly applied bio-orthogonal analog of myristoyl-CoA, was used ( 45 ). Three paradigmatic nonapeptides were selected as Nmt substrates. They represent the N-termini of proteins previously demonstrated to be myristoylated and of therapeutic interest. The peptides were derived from three different species: Lck from mouse, (DYKDDDDK) were provided by the Peptide Synthesis Unit at DKFZ Heidelberg, Germany. As negative controls, mutated Lck ( A CVCSSNPE-DYKDDDDK) or azido-tetradecanoyl-CoA (Invitrogen) instead of azido-dodecanoyl-CoA were used. Reactions (50 µl) containing the purifi ed enzyme or 50 µg of cytosolic fraction obtained from organ homogenates, azido-dodecanoyl-CoA, and peptide-FLAG at different concentrations in 50 mM HEPES (pH 7.4) were allowed to proceed at 37°C for 1 h. In case of kinetic studies, the reaction was allowed to proceed for 2 min and terminated by incubation at 95°C for 5 min.

NMT-azido-ELISA
Ninety-six-well plates (Nunc, Langenselbold, Germany) were coated overnight with 100 µl of anti-FLAG antibody ( were placed into the wells and diluted with 50 µl of washing buffer. Plates were incubated for 2 h at RT. The wells were fi lled with 100 µl of phosphine-biotin (50 µM in DMSO; Fisher Scientifi c, Schwerte, Germany), and the Staudinger ligation was allowed to proceed at 4°C overnight. Wells were washed three times, followed by the addition of 100 µl streptavidin-HRP (1:10,000 in blocking buffer; BD Biosciences, Heidelberg, Germany) and incubation for 1 h at RT. Bound streptavidin-HRP was detected by adding 100 µl of the substrate tetramethylbenzidine (TMB) (BD Bioscience) for 20-30 min in the dark at RT. The reaction was terminated by the addition of 50 µl of 2 M H 2 SO 4 , and the absorbance was measured at 450 nm using ELISA reader (Tecan, Mainz, Germany). For the inhibitor study, Tris(dibenzylideneacetone)dipalladium (Tris DBA) (Sigma) and 5-(3-carboxy-4-nitrophenyl)disulfanyl-2-nitrobenzoic acid (DTNB) (Sigma) were used at concentrations indicated in the Results section. Absorbance values were defi ned as the absorbance of each sample well minus the mean absorbance of background controls in wells containing anti-FLAG antibody and phosphine biotin. A standard curve was created using biotin-FLAG (Peptide Synthesis Unit, DKFZ) as standard. Calculations in kinetic and inhibitor studies were performed using GraphPad Prism 5 (GraphPad Software, La Jolla, CA). concentration of Nmt1 or Nmt2 used to perform each experiment, except where otherwise specifi ed, was 30 nM.

Assay validation
Numerous controls were used to validate the assay. Signals from both enzymes were observed exclusively in the presence of all assay components ( Fig. 2a, b ). In contrast, omitting each of the assay components interfered with the development of a signal ( Fig. 2a, b ).
Myristoylation can occur only at the N-terminal glycine ( 1 ). To demonstrate that the reaction assayed here was specifi c for the N-terminal glycine, Lck-FLAG was replaced by a mutated peptide with N-terminal glycine to alanine substitution (Lck-G1A-FLAG). As expected, this mutated peptide elicited no signals for either Nmt1 or Nmt2 ( Fig. 2c, d ).
It has been reported that NMT can use, albeit with a lower affi nity, palmitoyl-CoA as a substrate ( 4 ). To test this in our assay, azido-dodecanoyl-CoA was substituted by azidotetradecanoyl-CoA, an azido-analog of palmitoyl-CoA, which led to low but detectible signals for both enzymes ( Fig. 2c, d ).
Nef from HIV-1, and Arf from Trypanosoma brucei ( 34,46,47 ). The peptides were each C-terminally linked to a FLAG-tag and are referred to as Lck-FLAG, Nef-FLAG, and Arf-FLAG.
After the Nmt reaction ( Fig. 1a ), the azido-dodecanoylated nonapeptide-FLAG was coupled to phosphine-biotin via Staudinger ligation ( Fig. 1b ). The products of the reaction, coupled to biotin, could be detected and quantifi ed by an ELISA, in which it was captured on an immunoplate precoated with anti-FLAG antibody and detected by the addition of streptavidin-peroxidase ( Fig. 1c ).

Cloning and expression of murine Nmt1 and Nmt2
Murine Nmt1 and Nmt 2 were cloned from the brain of a C57BL/6 male mouse and inserted into the pQC-His express ion vector. Insert sequencing revealed the canonical form of Nmt1 . In the case of Nmt2 , repetitive sequencing of different clones demonstrated a previously unknown splice variant of Nmt2 lacking exons 3 and 4 ( Nmt2 _ OTTMUST00000025234 ). Both recombinant (His) 6 -tagged enzymes were expressed in E. coli and purifi ed under native conditions. The To this end, Lck-FLAG was replaced by a peptide with N-terminal glycine to alanine substitution (Lck-G1A-FLAG), which was not recognized by the enzyme. Replacing azido-dodecanoyl-CoA for a two-carbon longer azido-tetradecanoyl-CoA diminished the signals. Values shown are the averages of two independent experiments, which were performed in duplicates. Error bars represent SD. ( Fig. 5c ). The calculated half-maximal inhibition IC 50 was 0.5 ± 0.1 µM for Nmt1 and 1.3 ± 0.1 µM for Nmt2.

Repression of cellular NMT1 and NMT2 using siRNA
All aforementioned experiments were performed with recombinant enzymes. To assess that myristoylation activity can also be measured in cell lysates and that the detected signals refl ect NMT activity, we selectively repressed NMT1 and/or NMT2 in the human ovarian adenocarcinoma cell line SK-OV-3 using previously described NMT1 and NMT2 siRNAs ( 11 ). The repression of NMT1 and NMT2 was quantifi ed by qPCR. As compared with control siRNA, the treatment with NMT1 or NMT2 siRNA decreased their mRNA levels by 83% and 76%, respectively. The simultaneous treatment with both NMT siRNAs reduced the NMT1 and NMT2 mRNA levels by 81% and 70%, respectively ( Fig. 6a ).
To confi rm that the Nmt reactions yielded azido-dodecanoyl-Lck-FLAG, the reaction products for both enzymes were separated by HPLC. As compared with a negative control that lacked enzyme ( Fig. 3a ), the addition of Nmt1 or Nmt2 to the substrates led to the emergence of a new peak ( Fig. 3b, c, gray line ).
The sensitivity of the NMT-azido-ELISA, defi ned as the slope of the calibration curve, ( 48 ) was 0.112 liters/µg. The limit of detection (LOD), calculated as x bl + 3s bl , where x bl is the mean of the blank measures and s bl is the standard deviation of the blank measures ( 48 ), had a value of 10 ng/l. Linear and dynamic ranges were two and three orders of magnitude, respectively. Signal to noise ratio, calculated as ( x signal Ϫ x bl )/ s bl , where x signal is the mean signal and x bl and s bl are as defi ned above ( 49 ), had a value of 385 for recombinant Nmt1 (applied at a concentration of 0.9 µg/l), 146 for brain (using 50 µg of lysate), and 15 for liver (using 50 µg of lysate).

NMT-azido-ELISA can be used for comparative substrate studies
We sought to develop an assay that would allow testing a variety of substrates without changing the setting of the assay. To validate this, Lck-FLAG was tested in parallel with other paradigmatic peptides (Nef-FLAG and Arf-FLAG). Nmt1 and Nmt2 generated different signals according to the specifi c substrates-FLAG, demonstrating that each isoform has different peptide specifi cities ( Fig. 4 ). Finally, as described above for azido-dodecanoyl-Lck-FLAG, the formation of azido-dodecanoylated Nef-FLAG and azidododecanoylated Arf-FLAG was confi rmed by HPLC analysis (data not shown).

Kinetic and inhibitor studies using NMT-azido-ELISA
Before implementing the NMT-azido-ELISA in the investigation of enzyme kinetics, we generated a standard curve that allowed correlating the absorbance values to the numbers of molecules captured on the plate. To this end, the assay was performed with increasing concentrations of biotinylated FLAG (data not shown).
Using the NMT-azido-ELISA, K m values of the azidododecanoyl-CoA and of one of the paradigmatic nonapeptides (Lck-FLAG) were determined for Nmt1 and Nmt2. After verifying that the rate of myristoylation remained linear for the fi rst 2 min, the initial velocity (Vi) was calculated over the fi rst 2 min of the reaction. The K m value of azido-dodecanoyl-CoA was determined using increasing concentrations of azido-dodecanoyl-CoA, and a saturating concentration (200 µM) of Lck-FLAG . Similarly, the K m value of Lck-FLAG was calculated using increasing concentrations of Lck-FLAG, and a saturating concentration (200 µM) of azido-dodecanoyl-CoA. Azido-dodecanoyl-CoA showed a K m of 14 ± 2 (SD) µM for Nmt1 and 9 ± 3 µM for Nmt2 ( Fig. 5a ). Lck-FLAG had a K m of 26 ± 5 µM and 17 ± 2 µM for Nmt1 and Nmt2, respectively ( Fig. 5b ).
For the investigation of enzyme inhibition, a previously described NMT1 inhibitor Tris DBA was used ( 50 ). Nmt1 and Nmt2 were inhibited by increasing Tris DBA concentrations Fig. 3. Formation of the azido-dodecanoyl-Lck-FLAG was confi rmed via reverse-phase HPLC. Chromatograms of reactants without addition of the enzyme (a) and after the addition of Nmt1 (b) or Nmt2 (c). As compared with the analysis of reactants alone (a), the addition of the enzyme (b and c) led to the emergence of a new peak (gray line), which corresponds to azido-dodecanoyl-Lck-FLAG.
in embryogenesis, neoplasia, and plant biology and in the propagation of protozoal, fungal, and viral infections. It is anticipated that future research will focus on the development of small molecule inhibitors of NMT with the aim of suppressing infectious and malignant diseases. This goal requires a robust and easy to use assay for the measurement of NMT activity. Most of the methods in use are radioactive ( 1,4,36,37,40 ). Nonradioactive HPLC-based methods represent a sensitive alternative to radioactive assays but are not suitable for high-throughput analysis ( 41 ). Several continuous spectrophotometric or fl uorescent procedures have been developed ( 35,38,39 ). Although they are relatively easy to perform and applicable for high throughput analysis, continuous assays depend strongly on the milieu in which the reaction occurs. The purity of The diminished expression of NMT1 and/or NMT2 was paralleled by a diminished myristoylation activity, which decreased by 80% when the expression of NMT1 and NMT2 was repressed ( Fig. 6b ). This implicates that the signal generated by the NMT-azido-ELISA specifi cally refl ects the NMT activity in cell lysates.

NMT-azido-ELISA is suitable for the measurement of myristoylation activity in tissues
To document that NMT-azido-ELISA is suitable for the measurement of NMT activity in tissues, different murine organs were assayed for Nmt activity. Organs from adult C57BL/6 male mice were lysed, and cytosolic proteins were enriched. To demonstrate that the signals detected were a result of enzymatic activity, an aliquot of the preparation was thermally inactivated and treated under the same conditions as detailed above. Moreover, in parallel, two Nmt inhibitors were used: Tris DBA and DTNB, each of them at two different concentrations ( Fig. 7a ) ( 50, 51 ). Murine brain was found to contain the highest activity of Nmt, followed by thymus, heart, and liver ( Fig. 7a ). Heat inactivation of the organ extracts abolished the signals. In samples treated with Nmt inhibitors Tris DBA or DTNB, the myristoylation activity decreased in a dosedependent manner ( Fig. 7a ). The presence of the azidododecanoyl-Lck-FLAG in organ lysates was confi rmed by HPLC (supplemental Fig. I and data not shown). In addition, the liver preparation that showed the lowest Nmt activity ( Fig. 7a ) was spiked with increasing concentrations of recombinant Nmt1, which led to a linear increase of the signal ( Fig. 7b ).
To test whether this assay could be applied to frozen tissues (e.g., samples from a tissue bank), activity measurement was performed in murine brain, which had been stored at Ϫ 80°C for 1 week. In parallel, a freshly isolated brain was analyzed. The results demonstrated that Nmt activity could be assayed in frozen tissues without a pronounced decline in activity ( Fig. 7c ).

DISCUSSION
Protein myristoylation has been attracting increasing attention because it has been demonstrated to be of importance  The K m of azido-dodecanoyl-CoA was 14 ± 2 (SD) µM for Nmt1 and 9 ± 3 µM for Nmt2, whereas Lck-FLAG had a K m of 26 ± 5 µM for Nmt1 and 17 ± 2 µM for Nmt2. c: Inhibition of murine Nmt1 and Nmt2 by Tris DBA. Enzymes were incubated with increasing concentrations of Tris DBA using azido-dodecanoyl-CoA and Lck-FLAG as substrates. The calculated values of IC 50 were 0.5 ± 0.1 µM for Nmt1 and 1.3 ± 0.1 µM for Nmt2. Values shown are the average of duplicate measurements. a potential method in which antibodies directed against myristoyl-glycine were used in an ELISA. This approach has not been used according to published reports since its description.
The assay described here represents a new method for myristoylation measurement that uses affordable, commercially enzymes, or the presence of particular contaminants in the samples (e.g., derived from a tissue preparation), may infl uence the assay outcome ( 35 ). Similarly, certain groups of substances (e.g., strong nucleophilic inhibitors) in these assays are excluded from testing because they interfere with the signal detection used ( 38 ). Takamune et al. (41) described   7. Detection of NMT activity in murine organs. a: Tissue enzyme activity was determined using NMTazido-ELISA and Lck-FLAG as acyl acceptor. To demonstrate that the detected signal was a result of enzyme activity, aliquots of organ preparation were thermally inactivated at 95°C for 10 min or treated with two inhibitors (Tris DBA and DTNB) at different concentrations as indicated. All samples were processed in parallel. b: The liver lysate was spiked with increased amounts of recombinant Nmt1, and the Nmt activity was measured. Enzyme activity was normalized for the activity in the liver lysate without enzyme addition. The x axis has a log 2 scale. AU, arbitrary unit. c: Measurement of enzymatic activity in brain lysates prepared freshly in comparison to frozen brains stored at Ϫ 80°C. Values shown are the average of duplicate measurements of two independent experiments. Error bars are SD. strated that the NMT-azido-ELISA is in principle also suitable for this application ( Figs. 5c and 7a ). Moreover, IC 50 values for the inhibitor Tris DBA calculated by this approach for recombinant Nmt1 (0.5 ± 0.1 µM) and Nmt2 (1.3 ± 0.1 µM) were comparable to IC 50 values previously published for this inhibitor (1.0 ± 0.26 µM for NMT1) ( 50 ).
Measurement of NMT activity is of interest in cancer research. An important aim of the present study was to demonstrate that the assay could be used without further adaptations in the measurement of enzymatic acti vity in organ lysates. To this end, Nmt activity was mea sured in different murine organs: brain, thymus, heart, and liver ( Fig. 7a, b ). The Nmt activity has not been previously reported in these murine organs; however, its distribution among these tissues is consistent with previous observations in rat and cattle ( 4,14 ). The possibility of measuring Nmt activity in frozen organs ( Fig. 7c ) constitutes a signi fi cant advantage in clinical research and therapy trials.
We are convinced that a major advantage of this assay is its versatility. We show that it can be applied for investigations on recombinant Nmt enzymes and their substrates and inhibitors as well as for assessing Nmt activity in fresh or frozen tissues. The assay is user friendly, does not require complex equipment such as HPLC or radioactive material, and can be adopted for high-throughput analysis. Moreover, with appropriate modifi cations, other enzymes involved in protein acylation (e.g., palmitoylation) might be tested using this system. available substances in combination with an ELISA-like system. It has the advantage of using the myristoylation reaction that occurs between a FLAG-tagged peptide and azido-dodecanoyl-CoA, a bio-orthogonal analog of myristoyl-CoA ( 45 ). After subsequent coupling to phosphine-biotin via Staudinger ligation, the product is detected by ELISA using plate-bound anti-FLAG antibodies and avidinconjugated HRP ( Fig. 1 ). The combination of Staudinger ligation and ELISA has also been described to be useful for the detection of glycosyltransferase activity, although the assay arrangement was different ( 52 ). Furthermore, it has been shown that the azido-myristoylation can be used to tag proteins for proteomic analysis ( 53 ).
The presented assay was validated using recombinant murine Nmt1 and Nmt2, azido-dodecanoyl-CoA, and three paradigmatic FLAG-tagged substrate peptides (Lck, Nef, and Arf). Successive omission of each single component of the reaction and of the detection system abolished the signals, implicating a high specifi city of this approach ( Fig. 2a, b ). Similarly, substitution of the indispensable N-terminal glycine for alanine interfered with substrate recognition ( Fig. 2c, d ). The identity of the reaction pro ducts was verifi ed by HPLC-MS when using recombinant Nmt1, Nmt2, or tissue lysates as the source of Nmt activity ( Fig. 3 , supplemental Fig. I, and data not shown). Offering a substrate with a two-carbon longer acyl chain (i.e., azido-tetradecanoyl-CoA), low signals were detected ( Fig. 2c, d ); this is in line with a previous observation that NMT can also use palmitoyl-CoA, albeit ineffi ciently ( 4 ).
The presence of certain substances can potentially modify the assay readout. In particular, reducing thiol compounds such as DTT (400 mM) and 2-mercaptoethanol (2 mM) can enhance NMT activity up to 6-fold ( 51 ). Similarly, various types of detergents (Triton 770 and deoxycholate) can increase NMT activity ( 54 ). Conversely, certain oxidizing agents can have an inhibitory effect on NMT ( 51 ). Reducing reagents might interfere with the stability of the azidogroup, and oxidation of the phosphine can hinder the Staudinger ligation ( 55 ).
A general myristoylation consensus sequence is already known, but the substrate specifi city of NMT differs among species and among different isoenzymes ( 15,56,57 ). For this reason, myristoylation predictions based on this consensus sequence are not fail safe ( 58 ). We believe that the method presented here would allow easy screening of peptide libraries (e.g., based on certain protein motifs or on the proteome of a particular cell of interest) for potential myristoylation. We tested this possibility using three different peptides ( Fig. 4 ).
In addition, the assay can be used for enzyme kinetic studies ( Fig. 5a, b ). We have now obtained kinetic parameters of murine recombinant NMT. Comparison of our data to the human K m for myristoyl-CoA showed similar values; it has been given to be 8.24 µM for NMT1 and 7.24 µM for NMT2 ( 38 ), as compared with our results of 14 µM and 9 µM for murine NMT1 and NMT2, respectively.
Inhibitor studies are of particular interest because several pathogens encode for their own NMT ( 25,26 ). We demon-