Chemical structure and translation inhibition studies of the antibiotic microcin C7.

Escherichia coli microcin C7 (MccC7) is an antibiotic that inhibits protein synthesis in vivo. It is a heptapeptide containing unknown modifications at the N and C termini (García-Bustos, J. F., Pezzi, N., and Méndez, E. (1985) Antimicrob. Agents Chemoth. 27, 791-797). The chemical structure of MccC7 has been characterized by use of 1H homonuclear and heteronuclear (C, N, P) nuclear magnetic resonance spectroscopy as well as mass spectrometry (1177 ± 1 Da). The heptapeptide Met-Arg-Thr-Gly-Asn-Ala-Asp is substituted at the N terminus by a N-formyl group. The C-terminal substituent consists of the phosphodiester of 5′-adenylic acid and n-aminopropanol (AMPap), which is linked via the phosphorus atom to an amide group, thus forming a phosphoramide. The main chain carbonyl of the C-terminal aspartic acid residue is connected via this amide bond to the modified nucleotide unit. MccC7 and the peptide unit inhibit protein translation in vitro while a synthetic analog of the AMPap substituent is not active. Neither the peptide nor the AMPap molecule has an effect on the growth of MccC7-sensible cells. Our results strongly suggest that the peptide is responsible for MccC7 antibiotic activity while the C-terminal substituent is needed for MccC7 transport. Implications of the structure determined in this work for MccC7 synthesis and mode of action are discussed.

Microcin C7 (MccC7) 1 is a small peptide antibiotic produced by Escherichia coli during the stationary phase of growth (1). The spectrum of activity of this microcin includes several members of the Enterobacteriaceae family (2). MccC7 exerts its bacteriostatic action by blocking protein synthesis. Like for other microcins, the bacterial strains that produce MccC7 are immune (resistant) to this microcin (3). The heptapeptide of MccC7 is synthesized in ribosomes (4) and undergoes posttranslational modifications to yield the mature molecule.
The genetic determinants for MccC7 synthesis, export, and immunity have been cloned from the 43 kilobases E. coli pM-ccC7 plasmid into multicopy plasmids that overproduce MccC7 and express MccC7 immunity (5). These determinants lie on a 6.2-kilobase region of pMccC7, which has been entirely sequenced. 2 Different complementary approaches, such as physical and phenotypical characterization of insertion mutations and complementation studies, have shown that this region contains six genes (mccABCDEF) (5). 2 Genes A, B, C, D, and E are involved in the production of mature extracellular microcin. Genes C, E, and F code for self-immunity bestowing products. Genes mccA-D are directly involved in the synthesis and export of MccC7 and constitute an operon transcribed from a promoter (mccp), located upstream of mccA (4,6). Expression of these genes is regulated by the cAMP-cAMP regulatory protein complex and by the stationary phase factor RpoS (also called AppR) (6,7). 3 mccA codes for the unmodified peptide of MccC7, MRTGNAN (MccA) (4). The predicted gene polypeptide product of mccB (350 residues) is strikingly homologous to a 81-residue fragment of the ubiquitin-activating enzyme from different eucaryotic species (UBA1) (8), ThiF (9) and ChlN (10) from E. coli which participate, respectively, in the biosynthesis of thiamine pyrophosphate and of molibdopterin, and HesA, an enzyme required by Anabaena for nitrogen fixation (11). The predicted mccC product (404 residues) contains 11 potential transmembrane domains and displays significant similarity with stretches of transport proteins, 2 suggesting that MccC is responsible for MccC7 export and explaining why it also confers resistance to exogenous MccC7. The carboxyl end of the expected MccE 521-residue long polypeptide is highly similar to RimL, an enzyme that acetylates the ribosomal protein L12 from E. coli (12). Principally on the basis of this homology with an acetylating protein, and because target alteration is a common mechanism of antibiotic resistance (13), it has been proposed that MccE might confer MccC7 immunity to producing cells by acetylating the target of microcin C7. 2 No similarity was found for the predicted MccD (267 residues) and MccF (334 residues) polypeptides.
The knowledge of the chemical structure of MccC7 is neces-* This investigation was supported by funds from the Institut Pasteur, the Centre National de la Recherche Scientifique (URA 1129 and URA 489), and the European Union (Grant CI1*-CT92-0011 to F. M.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ Fellow of the Universidad Nacional Autónoma de México and latter of the Pasteur-Weizmann Association.
ʈ Fellow of the Dirección General de Investigación Científica y Técnica of Spain.
§ § To whom correspondence should be addressed. Tel.: 331-45688871; Fax: 331-45688885; E-mail: murield@pasteur.fr. 1 The abbreviations used are: MccC7, microcin C7; AMPap, 5Ј-adenylic acid and n-aminopropanol phosphodiester; COSY, correlation spectroscopy; Me 2 SO-d 6 , dimethyl sulfoxide 100% deuterated; f-M and f-Met, N-formylmethionine; HMBC and HSQC, respectively, heteronuclear multiple bond and single quantum correlation spectroscopies; MccC51, microcin C51; ROESY, nuclear Overhauser effect spectroscopy in the rotating frame; RP-HPLC, reverse-phase high performance liquid chromatography; TOCSY, total correlation spectroscopy. sary to understand its mode of action, biosynthesis, export out of the producing cell and intake into the target cell, and the self-immunity mechanisms. For these reasons, we undertook the elucidation of the microcin C7 chemical structure, one of the best genetically and functionally characterized microcins.
Amino acid analysis and sequencing of peptides derived from MccC7 tryptic digestions showed that the sequence of the MccC7 peptide is MRTGNAD (2). The N and C termini are both modified, and it has been suggested that the N-terminal might be acetylated, while the C-terminal modification remains unknown (2).
Here, we report on the chemical structure of microcin C7 ( Fig. 1) based on complete NMR assignment of 1 H, 13 C, 15 N, and 31 P resonances at natural abundance, as well as on the determination of the molecular mass of the molecule (1177 Ϯ 1 Da). We also show that MccC7 inhibits protein synthesis in a cell-free coupled transcription-translation system. Furthermore, we demonstrate that the peptide of MccC7 is responsible for protein synthesis blocking in vitro, while the C-and Nterminal substituents do not have any effect on this function. Finally, some implications of the chemical structure of MccC7 relative to its biosynthesis and mode of action are discussed.

MATERIALS AND METHODS
Microcin Production and Purification-Microcin was prepared from E. coli MC4100 strain containing the multicopy PMM550 plasmid coding for MccC7 (5). After 28 h of incubation at 37°C in M63 minimal medium, the growth culture was centrifuged. Supernatant was passed through a Sep-Pak C18 (Waters) column previously washed with 3 ml of ethanol and 30 ml of deionized water. Once the supernatant had passed through the silica bed, the column was washed with 3 ml of H 2 O and MccC7 activity was eluted with 2 ml of methanol. A 4:1 (v/v) mixture of chloroform and 50 mM KCl was added to the methanolic solution (2.5 milliliters of mixture per milliliter of methanol). The aqueous phase was then passed through a Sephadex G-25 column equilibrated with a 50 mM KCl aqueous solution. Antibiotic activity and the absorbance at 259 nm were used throughout the purification to identify the microcin containing fractions. Active fractions were pooled, lyophilized, dissolved in H 2 O, and further purified by RP-HPLC on a Delta-Pak C18 column (Waters). The column was eluted with a 10 -40% isocratic gradient of CH 3 CN, 0.1% trifluoroacetic acid. Fractions containing MccC7 were pooled and then freeze-dried. Typically, 1 mg of microcin was purified from 2 liters of culture.
Synthetic Molecules-Peptides f-MRTGNAD, f-MRTGNAN, and MRTGNAD were synthesized by the solid phase method of Merrifield (14) using an Applied Biosystems 430A synthesizer. P. methylbenzhydrylamine resins were used to obtain C terminus-amidated peptides. After HF cleavage, the crude peptides were purified by C18 reversephase medium pressure liquid chromatography followed by semi-preparative C18 RP-HPLC using linear gradients of CH 3 CN in 0.1% trifluoroacetic acid.
The phosphodiester of 5Ј-adenylic acid and n-aminopropanol (AM-Pap) was obtained by reacting overnight the tetrabutyl ammonium salt of 5Ј-adenylic acid with 1-bromo-3-aminopropane in dry CH 3 CN at 80°C. After evaporation, the reaction mixture was dissolved in H 2 O and desalted on a G-25 column. The first eluted fractions that were ninhydrin-positive and contained the AMPap molecule, were pooled and lyophilized. AMPap was then purified by C18 RP-HPLC using a linear gradient of CH 3 CN in 0.1% trifluoroacetic acid.
Final purity (Ͼ95%) of synthetic molecules was verified by analytical RP-HPLC. The molecular mass of the formylated peptides was determined by electrospray mass spectrometry and was in every case in accordance with the expected mass. NMR spectra analysis of the peptides and the AMPap molecule was also in agreement with the expected formulas.
NMR Spectroscopy-Dried powder of purified microcin was dissolved either in 700 l of H 2 O (10% D 2 O), D 2 O (100%) (SDS, France), or Me 2 SO-d 6 (100%) (Commissariat à l'Energie Atomique, France). Sample concentration ranged from 0.9 to 9 mM. When needed, the pH was adjusted by adding DCl/HCl or NaOD/NaOH solutions (0.33 N) to D 2 O/H 2 O microcin samples. The acidity/basicity of MccC7 samples in Me 2 SO was also changed with these solutions. pH and pD values are reported without correction for isotopic effects. All experiments were run at 27°C. NMR experiments were performed on a Varian Unity 500 MHz spectrometer equipped with a Sun Sparc 2 station. An indirect detection probe or a multinuclei broad band direct probe were used. Spectra were processed on a Sparc 2 station with Varian's Vnmr software (Varian, Inc.). Sweep widths were for 1 H, 5200 Hz in H 2 O or D 2 O, and 7200 and 5200 Hz in Me 2 SO; for 13 C, 32,000 Hz; for 15 N, 4000 Hz; and for 31 P, 7500 (direct detection) and 600 Hz (indirect detection). Chemical shifts were referenced to: 3-(trimethylsilyl)-propane sulfonic acid, sodium salt, for 1 H and 13 C, 15 NH 4 Cl for 15 N, and 85% H 3 PO 4 for 31 P, used as external references. All heteronuclear experiments were carried out at isotope natural abundance. When needed, the water peak was eliminated by presaturation during the appropriate delays.
Phase-sensitive COSY (15,16), and TOCSY (17) using the MLEV17 sequence for spin-locking (18) and mixing times of 80 or 100 ms were performed to identify short and long range scalar couplings. NOESY experiments (19) were run with a 500-ms mixing time. A pulse at variable times was used during the mixing time to reduce zero quantum contributions. 4 ROESY experiments were carried out with a mixing time of 500 ms and a spin lock field of 2.5 kHz to reduce Hartmann Hahn transfers (20). Two-dimensional spectra were obtained with quadrature detection in both dimensions using the hypercomplex method in the f 1 dimension (19). Usually, 2048 points were acquired in the f 2 dimension, and 400 or 512 complex points in the f 1 dimension. For each complex data point in the first dimension 8, 16, or 24 free induction decays were accumulated with a relaxation delay of at least 2.4 s. All spectra were apodized with a shifted sine-bell function in both dimension or with a sine-bell function for the COSY experiments.
One bond, 1 H-13 C, 1 H-15 N, and one to three bonds, 1 H-31 P, correlation maps were obtained from 1 H detected HSQC experiments (21,22). For 1 H-13 C HSQC experiments, a BIRD pulse sequence followed by a 0.9-s delay was used to increase the cancellation of signals of protons attached to 12 C (23). 1 H-15 N spectra in dimethyl sulfoxide were acquired without water suppression to observe the maximum of 15 N nuclei bearing a labile hydrogen. 1 H-31 P HSQC spectra were acquired in one and two dimensions for various conditions of solvent and pH. Multiple bond 1 H-13 C interactions were obtained from HMBC experiments (22,24). The delay preceding the 13 C pulse for the creation of multiple quanta coherences through several bonds was set to 60 ms. Heteronuclear coupling constant values used in the HSQC and HMBC experiments to establish the delays needed to select the protons coupled to the heteronuclei were the following: 150 Hz ( 1 H-13 C), 90 Hz ( 1 H-15 N), and 15 Hz ( 1 H-31 P). All reverse detection spectra were obtained with the WALTZ-16 pulse scheme for heteronuclear decoupling (25). Heteronuclear two-dimensional spectra were obtained with quadrature detection in both dimensions using the hypercomplex method (19). An exponential function in the t 2 dimension and a shifted sine bell in the t 1 dimension were used to Fourier transform the heteronuclear spectra.
One-dimensional 13 C direct detection spectra were obtained in D 2 O at different pD values (pD 3.4, 4.3, and 7.3) with microcin sample concentrations higher than 7 mM. A long delay, 16 s, as well as a /3 pulse, were used to allow an efficient relaxation of the quaternary carbons, thus permitting a reliable integration of the signals. Spectra were recorded with 5,800 (pD 3.4), 8,944 (pD 4.3), and 20,032 (pD 7.3) scans over 64 K data points using broad band proton decoupling. Onedimensional 31 P spectra were acquired in different conditions of pH and solvent, always with a 3.5-s relaxation delay.
Mass Spectrometry-Electrospray mass spectrometry was performed at the Laboratoire de Spectrométrie de Masse Bio-Organique (Strasbourg) and at the Université de Paris VII (Paris). In order to avoid fragmentation of labile bonds, MccC7 mass spectra were obtained in conditions of solvent similar to those used during purification, that is a mixture of 1:1 H 2 O/CH 3 CN (v/v), 0.1% trifluoroacetic acid. The observed mass for microcin was 1177 Ϯ 1 Da. The experiment, repeated under different conditions (50% H 2 O, 25% CH 3 OH, and 25% CH 3 COOH) gave the same result.
In Vitro Coupled Transcription-Translation Assays-S30 extracts prepared from E. coli (strain MRE600, RNase I Ϫ ), and the reaction mixture for coupled transcription-translation were purchased from Amersham (Procaryotic DNA-directed translation kit N.380Z). Circular plasmid DNA template was purified by gradient centrifugation in CsCl, and was added at 4 g/50 l of reaction mixture. MccC7 and peptides were added to the concentrations indicated in Fig. 11. Reactions were allowed to proceed for 30 min at 37°C in the presence of L-[ 35 S]methionine (15 Ci, 1000 Ci/mmol). Proteins were then precipitated and the incorporation of radioactivity was determined by liquid scintillation counting.
MccC7 concentration was determined by measuring the absorbance at 259 nm, using an extinction coefficient of 16,500 M Ϫ1 cm Ϫ1 , determined through the quantitative amino acid analysis of a MccC7 solution of known absorbance. The concentration of peptides was determined with the BCA protein assay reagent (Pierce), using a solution of MccC7 of known concentration as standard.

1
H homonuclear shift correlation spectra COSY and TOCSY allowed the identification of three distinct units in the molecule: the peptide chain modified in its N-terminal (X-peptide), a sugar ring, and an aliphatic chain. In addition, there are two singlets in the low field region of the spectrum resonating at 8.41 and 8.35 ppm, and integrating for one proton each (Fig. 2). We will describe these different units in the following sections.
X-peptide Structure-All proton resonances in the peptide chain were assigned by analysis of COSY, TOCSY, and ROESY spectra following the method proposed by Wü thrich (26). Proton chemical shifts in H 2 O are reported in Table I. The sequential assignments found agree with the previously published sequence MRTGNA (2). The C-terminal residue is an AMX spin system in agreement with amino acid composition analysis (N or D) (1), peptide microsequencing (D) (2), and the gene sequence (N) (4). As there are only two amide signals (7.55 and 6.89 ppm) corresponding to an asparagine residue (Fig. 2a) there must be only one asparagine in MccC7. The distinction between an asparagine and an aspartic acid residue was further done from the study of 1 H chemical shift variation with pH, as asparagine is neutral while aspartic acid is not. Proton chemical shift variation of the C-terminal residue as a function of pD in D 2 O or as a function of NaOD added in Me 2 SO-d 6 showed that this residue has an ionizable group. The observed pK a of ϳ4.2 in D 2 O is in agreement with an aspartic acid at this position.
One-bond 1 H-13 C correlation maps (HSQC) and two-or three-bond 1 H-13 C heteronuclear couplings (HMBC) obtained in D 2 O allowed the assignment of 13 C signals including those of the quaternary carbons. Chemical shifts are reported in Table II.
In previous studies (1), it was not possible to perform the N-terminal sequence of MccC7 by the Edman degradation method, suggesting the presence of an NH 2 blocking group at the N terminus of the peptide. The singlet observed at 8.12 ppm in the one-dimensional proton spectrum is in fact scalar coupled (J ϭ 1.5 Hz) to the methionine amide proton resonating at 8.45 ppm as observed in COSY spectra. In addition, these two protons are in dipolar contact (rOe). The 8.12 ppm proton is borne by a carbon resonating at 166.8 ppm as shown in a 1 H-13 C HSQC experiment run in D 2 O (pD 4.3). This position is compatible with a carbonyl group such as that of formamide (27). From the HMBC experiment, it was possible to (i) determine a 1 H-13 C coupling constant of 198 Hz, in agreement with an aldehyde such as formamide whose coupling constant is 188 Hz (28) and (ii) to link the proton resonating at 8.12 ppm with the methionine ␣-carbon at 54.1 ppm. Taken together, these data are in favor of a N-formyl-Met concatenation. This is further confirmed by analysis of the NMR spectra of the synthetic peptide f-MRTGNAD. Indeed, the one-dimensional proton spectrum of this synthetic peptide also contains a singlet at 8.12 ppm which shows the same coupling pattern (COSY, TOCSY, and ROESY) as the analogous signal in the MccC7 spectra.
In addition to the N-and C-terminal substitutions, the peptide of MccC7 might be modified at other positions. It is thus necessary to identify all the peptide signals, especially those of exchangeable protons. It must be mentioned though, that the 1 H and 13 C chemical shifts are in agreement with values expected for nonsubstituted amino acids in unstructured peptides (26,29). As rapidly exchanging proton signals might not be observed in aqueous solvents, proton as well as nitrogen 15 ( 1 H-15 N, HSQC) assignments were completed in dimethyl sulfoxide (Me 2 SO-d 6 ). For this purpose, MccC7 in aqueous solution at pH 3.2 was freeze-dried and dissolved in Me 2 SO-d 6 . Assignments are reported in Table III. All the expected exchangeable protons could be identified. Indeed, the threonine hydroxyl proton (4.95 ppm), the arginine NH⑀ (7.56 ppm), and guanidinium group (7.15 ppm) integrating, respectively, for one and four protons, and the two asparagine side chain amide protons (7.52 and 7.03 ppm) were assigned via chemical shift analysis, through-bond (COSY, TOCSY) and through-space (NOESY, a Main and side chain carbonyl carbons of Asn and Asp residues were, respectively, assigned from the H ␣ -C et H ␤␤ -C correlations peaks in the HMBC spectrum. In agreement with this, the main chain carbonyl carbon signal of Asp is coupled to the phosphorus nucleus of the Z substituent while the side chain carbonyl signal is not. ROESY) interactions (Fig. 3a). The broad signal at 4.95 ppm corresponding to the threonine hydroxyl proton, sharpened upon addition of a small amount of NaOH and became a doublet with a coupling constant of 4.2 Hz, as a result of throughbond coupling with the threonine H␤ proton.
All these signals are also present in the spectra of the synthetic peptide f-MRTGNAD (data not shown). Hence, the peptide is not substituted and the remaining spin systems identified in the proton homonuclear experiments must belong to the C-terminal modification.
C-terminal Modification, Z-In addition to the peptide chain, a sugar ring, an aliphatic chain, and two low field singlets were identified. All these spin systems must belong to the C-terminal modification and have to be linked together to form the Z substituent. Hence, in this section we describe (i) the sugar ring (and the associated base, see below), (ii) the aliphatic chain, and (iii) the structure of the Z substituent.
(i) The six non-exchangeable sugar ring protons were assigned mainly from COSY and TOCSY spectra analysis and their chemical shifts are reported in Tables I (H  . The sugar corresponds to a ribose ring. The chemical shift values obtained for the H5Ј5Љ protons, downfield shifted by ϳ0.7 ppm relative to ribose, suggest that the corresponding carbon is substituted (26). For the same reason, the C2Ј and C3Ј carbons do not seem to be substituted. This is validated by the presence of two broad signals of exchangeable protons at 5.58 and 5.38 ppm (Fig. 3b), chemical shifts that are in agreement with that of hydroxyl protons. These two broad signals sharpened upon addition of NaOH. Using a short mixing time of 10 ms in TOCSY spectra to limit the transfer to two or three bonds, it was possible to observe cross-peaks from the signal at 5.58 ppm to the H2Ј sugar proton at 4.57 ppm, and from the signal at 5.38 ppm to the H3Ј sugar proton at 4.19 ppm.
Having established the proton network in the 1 H spectra, the sugar ring carbon signals were identified from a 1 H-detected 1 H-13 C HSQC experiment carried out in D 2 O (Table II). The chemical shifts values obtained are in accordance with a C5Ј substituted ribose bearing a N-glycosyl bond in C1Ј. The H1Ј ribose proton (6.13 ppm) gives an interaction with a quaternary carbon at 151.6 ppm and with a signal at 142.7 ppm (HMBC, Fig. 4a) to which the low field proton resonating at 8.37 ppm is attached (HSQC). In addition, in ROESY experiments, a dipolar interaction between the ribose H1Ј sugar proton and the singlet at 8.37 ppm is observed, suggesting that the latter could correspond to a purine proton. Furthermore, this low field proton resonating as a singlet at 8.37 ppm showed an interaction in the HMBC spectrum with two quaternary carbons resonating at 121.3 and 151.6 ppm (Fig. 4b). Finally, the remaining low field proton resonating as a singlet at 8.29 ppm and attached to the carbon resonating at 154.4 ppm (HSQC) is linked via scalar interactions to two quaternary carbons resonating at 157.4 and 151.6 ppm (Fig. 4b). Thus, the proton singlets at 8.37 and 8.29 ppm are both coupled to a carbon resonating at 151.6 ppm and belong to the same coupling network. The coupling pattern observed for the base and the sugar carbons and protons, the proton and carbon chemical shifts (27,30), and the number of protons and carbons for this unit are characteristic of an adenine or a hypoxanthine linked to a ribose via a N-glycosyl bond. The fact that the MccC7 UV absorbance spectrum presents a maximum at 259 nm (H 2 O, pH 7) is in favor of an adenine rather than hypoxanthine. Indeed, adenosine presents a UV absorbance maximum at 260 nm at neutral pH, while hypoxanthosine, under the same conditions, absorbs maximally at 249.5 nm (31). These results are strengthened by the identification of the two adenine amino protons (Fig. 3a) as well as of the nitrogen bearing those protons in an 1 H-15 N HSQC experiment in dimethyl sulfoxide (Fig.  5). Under more basic conditions, the signal corresponding to the two amino protons at 7.82 ppm (Fig. 3a) sharpened up to become a singlet integrating for two protons while moving upfield at the same time that the H2 and H8 singlets titrated. Under these conditions, a through-bond interaction (TOCSY) with the H2 base proton is observed. Titration of the base amino, H2 and H8 protons is surely caused by the imino nitrogen deprotonation. (ii) The aliphatic chain contains three methylene groups coupled one to another, as shown by integration curves and COSY spectra. In Me 2 SO-d 6 , these methylene proton signals were assigned at 4.07, 1.88, and 2.89 ppm (Fig. 3b). The methylene group at 2.89 ppm is scalar coupled to three exchangeable protons resonating at 7.78 ppm (COSY), and which are attached to a nitrogen resonating at 88.4 ppm ( 1 H-15 N HSQC, see Fig. 5). Thus, this group is a n-aminopropane chain (-CH 2 -CH 2 -CH 2 -NH 3 ϩ ). (iii) For the Z substituent, taking into account the chemical shift values observed for the ribose H5Ј5Љ and the C5Ј signals, as well as the presence of heteronuclear couplings in the onedimensional proton-decoupled carbon spectrum for the C5Ј and C4Ј signals (Fig. 6), it was important to check if the molecule contained any phosphorus atom. One signal is indeed observed, at Ϫ2.36 ppm in the one-dimensional phosphorus spectrum obtained in Me 2 SO. From a 1 H-31 P HSQC experiment, it was possible to link the phosphorus nucleus to the ribose H5Ј5Љ protons (4.12 and 4.26 ppm), to the aliphatic chain (CH 2 )␣ protons (4.07 ppm) as well as to a signal at 9.46 ppm (Fig. 7). The intensity of the signal at 9.46 ppm corresponds to that of one proton. Its coupling constant of 10.5 Hz (Fig. 3a) disappears in proton one-dimensional spectra acquired with phosphorus broad band decoupling, indicating that this signal is coupled to the phosphorus nucleus but not to any other proton. A 1 H-15 N HSQC shows that this proton is attached to a nitrogen resonating at 92.8 ppm (Fig. 5). Thus, MccC7 contains a P-NH group. In addition, the C5Ј ribose and the C␣ aliphatic chain chemical shifts (69.3 and 68.5 ppm, respectively) in D 2 O (pD 4.3), together with the phosphorus chemical shift (Ϫ1.63 ppm) are in favor of P-O-C bonds rather than of phosphine bonds (P-C) (27). The 13 C-31 P coupling constants C5Ј-P (4.7 Hz) and C␣-P (5.5 Hz) are also in agreement with two-bond C-O-P coupling constants (27). This holds true for the three-bond coupling constants C4Ј-P (7.9 Hz) and C␤-P (7.6 Hz). Hence, the MccC7 phosphorus group is a phosphoramide. Together, these results lead to the formula shown in Fig. 1 for the Z substituent.
The comparative analysis of the chemical shifts (and coupling patterns) of the proton signals of the MccC7 Z substituent and of the analogous molecule AMPap (phosphodiester of 5Јadenylic acid and n-aminopropanol) is in agreement with this chemical structure. Indeed, in D 2 O and Me 2 SO, chemical shifts are similar for the corresponding proton signals in both molecules, with the exception of those protons that are closer to the differently substituted phosphorus atom: the ribose H5Ј5Љ protons and the (CH 2 )␣ aliphatic protons.
We will now describe how the Z substituent is linked to the peptide.
The Microcin C7 Structure-The phosphoramide NH signal (NH p ), which could only be seen in Me 2 SO, is not scalar coupled to other protons in the molecule. Since all the exchangeable proton signals of the Z substituent were identified, the only possibility to append the X-peptide is via this NH group. On the other hand, microcin C7 is not degraded by carboxypeptidases A, B, and Y, suggesting that the C-terminal residue is cyclized, modified, or bears an end blocking group (2). There are two possible positions in the C-terminal residue for branching of the Z substituent: the main chain or the side chain carboxylic groups. The existence of a 10.6 Hz coupling constant between the 31 P and the 13 C␣ of the aspartic residue is in favor of a main chain branching of the Z substituent (Fig. 6).
The proton one-dimensional spectrum in Me 2 SO displays a broad signal at 12.5 ppm, a typical position for a carboxylic proton (not shown). However, this signal is too broad to produce scalar or dipolar interactions. The chemical shift variation as a function of pD in D 2 O, or of NaOD added in Me 2 SO, indicates that the aspartic residue contains a ionizable group. In addition, in the one-dimensional 13 C spectrum (D 2 O), nine carbonyl signals are observed (Fig. 8): six belong to the peptide amide groups f-M, R, T, G, N, and A, and one to the asparagine side chain. Thus, the two remaining carbons must belong to the aspartic residue. In agreement with this hypothesis, the 1 H-13 C HMBC experiment shows an interaction between the Asp ␣ proton and a carbon signal at 177.5 ppm as well as an interaction between the Asp ␤ protons and a carbon signal at 177.8 ppm. 5 It should be mentioned though, that the digital resolution in the carbon dimension was quite low (2.65 ppm/point) and it is not possible to know if this correlation peaks correspond to one or two carbons. Finally, one of these two carbons is scalar coupled to the phosphorus (J13 C-31 P ϭ 3.2 Hz, see Fig.  8). Hence, we can conclude that the aspartic residue contains a carboxylic and a carbonyl group.
Taking into account the MccC7 molecular weight, the number of carbons determined from the one-dimensional 13 C spectrum and the assignments already established, there is only one solution for the Z substituent branching: an amide peptide bond to the phosphoramide NH as shown in Fig. 1.
Several experimental evidences support the proposed branching and lead to the conclusion that the side chain carboxylic group is free. Indeed, (a) the Asp C␣ is scalar coupled to the phosphorus nucleus (10.6 Hz) while the C␤ is not. This coupling constant value is of the same order of magnitude as the three-bond 13 C-31 P coupling constant observed between the phosphorus and the sugar ring C4Ј or the aliphatic chain C␤. In agreement with this observation, the sugar ring C3Ј and the aliphatic chain C␥, separated by four bonds from the phosphorus atom, are not scalar coupled to it. Furthermore, in D 2 O, the Asp C␤ signal broadened at pD 4.3, that is, at a pD value close to the carboxylic group pK a value, whereas the C␣ signal remained sharp. In addition, the Asp C␣ signal broadened (23 Hz at half-height) at pD 7.3 in the phosphoramide pK a value range.
(b) In Me 2 SO, a larger chemical shift variation is observed upon the carboxylic group ionization for the Asp H␤ protons (Ϫ0.26 ppm and Ϫ0.13 ppm) than for the ␣ proton (Ϫ0.10 ppm). It is well known that for protons, the inductive effect decreases as the number of bond increases. This observation is also valid in aqueous solution but the interpretation is not straightforward, as in this environment the aspartic residue and the adenine base titrated with identical pK a values.
(c) In D 2 O, the Asp ␤ proton chemical shifts remained constant when the pD was changed in the phosphoramide group pK a range (Fig. 9), while the Asp H␣ chemical shift varied (Ϫ0.09 ppm) between pD 7.2 and 9. Hence, the Asp ␤ protons are not in the direct environment of the phosphoramide group. Moreover, the Asp ␣ proton displayed the same chemical shift variation pattern as a function of pD than the H5Ј5Љ protons and the aliphatic chain ␣ protons which are at a two-bond distance from the phosphorus (Fig. 9).
(d) There is a dipolar coupling between the Asp ␣ proton and the phosphoramide NH p proton in Me 2 SO (Fig. 10). This interaction is analogous to the interaction observed in the peptide of MccC7 between the H␣ proton of residue i and the amide proton of residue iϩ1.
It was previously reported that MccC7 does not react with ninhydrin (2), a compound that is commonly used to reveal aliphatic amines. However, in the MccC7 structure determined here, there is a primary amine group that should react with ninhydrin. We thus performed the ninhydrin test on thin layers and found that MccC7 is ninhydrin-positive.
Chemical Shift Variations with pD-There are three distinct pD regions for which chemical shifts vary: 3 to 5, 7 to 9, and 10 to 12 (Fig. 9). The transition observed for the aliphatic chain ␤ and ␥ protons between ϳ10 and ϳ12, centered at 10.7 corresponds to a typical pK a value for primary amines or for a lysine side chain, for instance. The same pK a value was observed for the amine group in the analog molecule AMPap. The transition observed between ϳ3 and ϳ5 involves all the protons close to the adenine imino group or to the ␥ carboxyl of the aspartic residue: the purine H2 and H8 protons, the aliphatic chain methylene protons (CH 2 )␣ and (CH 2 )␤, the ribose ring H5Ј5Љ protons, and the aspartic H␣ and H␤ protons. The transition between ϳ7 and ϳ9 and centered at pD 7.8 involves protons close to the phosphorus atom. This transition must be the result of the phosphoramide nitrogen deprotonation. Indeed, the magnitude of the induced shift upon titration (13.3 ppm) as well as the direction of it (downfield shift upon ionization) are in accordance with nitrogen deprotonation (33).
It is interesting to note that the phosphorus signal splits into two broad signals when the pD value is near 6. Close to pD 7.8, the phosphorus signal spreads over several units of ppm. At pD values higher than 9, it becomes a single sharp peak once again. These line width changes and splitting are reproducible and reversible, indicating that these are not due to hydrolysis phenomena or other possible modifications of the molecule. The splitting of the phosphorus signal also occurs in Me 2 SO. However, in this solvent, both phosphorus signals remain sharp, and the splitting is observed after DCl addition to Me 2 SO samples prepared from MccC7 "acid" samples (see "Materials and Methods"). Although we cannot explain this phenomenon, it is likely to be due to a pH-dependent conformational equilibrium. This is substantiated by the lack of scalar coupling between one of the phosphorus signals when splitting occurs, and the phosphoramide proton NH p . That MccC7 was not modified by the exposure to different conditions of pH and solvent is further corroborated by the fact that, after NMR experiments (in H 2 O, D 2 O, and Me 2 SO), MccC7 samples conserved the same antibiotic activity than fresh samples.
MccC7 and Its Peptide Chain Inhibit in Vitro Protein Synthesis-Previous in vivo studies showed that MccC7 blocks protein translation without affecting either DNA replication or RNA transcription (2). It has also been shown that cellular growth is inhibited in cells producing the peptide MccA in the absence of other mcc genes, suggesting that MccA may also block protein synthesis. 2 To know if these molecules are also active in vitro, we examined protein translation in the coupled transcription-translation system described under "Materials and Methods." As shown in Fig. 11 We also checked these peptides and AMPap for cell growth inhibition. Typically, the growth of E. coli MccC7-susceptible cells is inhibited at a 20 M concentration of MccC7. In contrast, AMPap and the peptides did not show any effect on growth when used at a 10-fold higher concentration. These results suggest that the unmodified peptide of MccC7 is not able to reach the MccC7 intracellular target, and that the AMPap group may be required for microcin transport through the cell envelope. DISCUSSION Results described here allow us to conclude on the structure of microcin C7 as presented in Fig. 1. The interpretation of the MccC7 NMR spectra is reinforced through comparative analysis with NMR spectra of the synthetic molecules f-MRTGNAD and AMPap, analogs of the MccC7 formylated peptide and of the C-terminal substituent Z, respectively.
Implication of Our Results for Microcin C7 Biosynthesis-The polypeptide predicted from the gene mccB, which is known to be directly involved in MccC7 synthesis, is highly similar to a central segment of 158 residues 2 of the enzyme UBA1 (8). The identity is as high as 40% for a sequence of 76 residues that includes the motif Gly-X-Gly-X-X-Gly, a hallmark of nucleotide binding domains. If conservative substitutions are allowed, the similarity increases to 60%. The ubiquitin-activating enzyme UBA1 (also called E1) catalyzes two reactions: first, the adenylation of the ubiquitin C-terminal glycine, and then the formation of a thioester bond between the adenylated glycine and a cysteine residue of the enzyme, thus forming a ubiquitin-UBA1 covalent complex (34). In the adenylated ubiquitin that is formed after the first reaction, the C-terminal carboxyl of ubiquitin is linked via an acyl-phosphate anhydride to the 5Ј-adenylic acid (34,35). The homology between this peptidoacyl adenylate and the structure of MccC7, together with the homology found between MccB and UBA1, and the fact that the gene mccB is essential for MccC7 synthesis (but does not participate in immunity), supports very strongly that MccB is responsible for the adenylation of the peptide of MccC7. It is attractive to think that, as in the case of ubiquitin, ATP may act as an AMP donor. We do not know, however, if the n-aminopropane group is incorporated before, after, or simultaneously with the AMP group.
That the C-terminal residue of mature MccC7 is an aspartic acid while the gene mccA codes for an asparagine at this position is not surprising to us: chemical deamidation of Asn to produce Asp has been recognized in proteins and peptides for many years (36 -38). In principle, the side chain carbonyl carbon of the C-terminal asparagine should be susceptible to nucleophilic attack by the phosphoramide nitrogen. Hence, chemical deamidation, which is not usually observed on peptide C-terminal asparagines lacking a neighboring main chain amide nitrogen, is conceivable for MccC7. However, we do not know whether this deamidation occurs enzymatically or chemically, in vivo or during MccC7 purification, whether it is important for peptide modification or not, or whether it is concomitant with peptide modification (adenylation) or not.
Microcin C7 Mode of Action-MccC7 blocks protein synthesis in vivo without affecting the incorporation of precursors into DNA, RNA, and the cell wall, as shown in previous work (2). Although this result does not permit us to exclude that MccC7 might alter other cellular processes, it does indicate that the antibiotic activity of MccC7 is due, at least in part, to its translation inhibition activity. Hence, that the synthetic peptide f-MRTGNAD inhibits protein translation in vitro with the same efficiency than MccC7, implies that the antibiotic activity of MccC7 is due to its peptide unit. That AMPap, an analog of the modified nucleotide unit of MccC7, does not inhibit protein synthesis in vitro (and in vivo) is in agreement with this proposal. This is further corroborated by the fact that cells carrying the gene mccA coding for the peptide, but lacking the other genes necessary for the synthesis of mature extracellular MccC7 and for microcin immunity, show a decreased growth rate and enter the stationary phase before nutrient depletion. 2 These two observations strongly suggest that the unsubsti- tuted peptide is toxic when localized in the cell. As the synthetic peptides f-MRTGNAD and f-MRTGNAN do not influence cellular growth when added to sensible cell cultures, it might be inferred that the peptides cannot enter into the cell. Hence, the modified nucleotide substituent seems to be necessary for the import of MccC7. However, our results do not rule out the possibility that the peptides might be degraded by extracellular proteases. If this were true, the C-terminal substituent would play a protective role against proteases. It should be mentioned that it is also conceivable that the modified nucleotide, and/or the region linking the peptide and the nucleotide might inhibit another cellular function.
The peptide coded by the gene mccA, f-MRTGNAN, inhibits protein synthesis in vitro like the peptide found in mature microcin: f-MRTGNAD. Hence, the deamidation of the asparagine residue is not important for translation inhibition. However, as already noted, this deamidation might be necessary for (or parallel to) the modification of the peptide.
The peptide of MccC7 (f-MRTGNAD) is similar to the leader peptide (MSTSKNAD) involved in the regulation by translation attenuation of the gene cmlA from E. coli, which codes for non-enzymatic resistance to chloramphenicol (39). This peptide inhibits the peptidyl transferase activity of the ribosomal 50 S subunit, and binds the peptidyl transferase center of the 23 S rRNA. Given that the peptide of MccC7 inhibits protein synthesis, that both peptides have five residues in common, that the side chain of residue 2 in f-MRTGNAD is not important for protein synthesis inhibition, 6 and that the unformylated peptide (MRTGNAD) has the same in vitro activity that the formylated peptide, it is tempting to think that the target of MccC7 might be the same as that of the cmlA leader peptide. Further 6 J. E. Gonzá lez-Pastor, unpublished data. experimental work is needed to test this hypothesis.
Comparison of Microcins C7 and C51-It seems important to us to discuss here about microcin C7 and microcin C51 (MccC51) structures, because these independently isolated antibiotics are very similar, and might in fact be the same. The resemblance of the physical and genetic maps of MccC7 (5) and MccC51 (40) is striking. In addition, E. coli strains producing MccC7 and MccC51 exhibit cross-immunity (40). The antibacterial spectrum of activity of both molecules also appears to be similar. However, as different bacterial strains and testing conditions were used in the studies of MccC7 (2) and MccC51 (40), it is not possible to assess if the spectra of action are actually the same or not. Some functional differences might exist: while microcin C7 has been described as a bacteriostatic agent that inhibits protein synthesis (Ref. 2, and this article), microcin C51 has been reported to be a bactericidal agent that inhibits incorporation of precursors into DNA, RNA, and proteins (40). To our knowledge, as of this writing, the announced results of MccC51 inhibition of DNA, RNA, and protein synthesis have not been published. Although, these functional differences might result in being significant, at present it is difficult to ascertain if this is the case: classification of an antibiotic as a bactericidal or bacteriostatic agent depends on experimental conditions such as drug concentration, susceptible strains used, time of incubation in the presence of the drug, and culture conditions. These experimental conditions also have an important influence on the outcome of incorporation of radiolabeled precursors into macromolecules experiments.
Principally based on the molecular mass of MccC51 (1177.5 Da), analysis of MccC51 NMR spectra obtained in D 2 O at 40°C, and amino acid sequencing (and mass determination) of MccC51 proteolytic peptides (MRTGNAX, where X represents N or D), Metlitskaya and co-workers (41) have proposed a chemical structure for MccC51. MccC51 would consist of the peptide f-MRTGNAN, a three-methylene chain, and the nucleotide nebularine 5Ј-monophosphate. The side chain of the Nterminal asparagine would be linked via an amide bond to the three-methylene chain. The aliphatic chain would be linked to the nucleotide phosphate. Finally, the hydroxyl oxygen of threonine would be involved in an hydroxylamine group (-O-NH 2 ). However, we have several arguments to conclude that this proposed structure is not correct. Indeed, the one-dimensional proton spectra of MccC7 (40°C, H 2 O, pH 3.2, not shown) and of MccC51 (40°C, D 2 O, pD not specified) are very similar. The differences can be accounted for by impurities, labile proton signals which are not present in D 2 O spectra, line shape of proton signals coupled to exchangeable protons, and a systematic ϳ0.05 ppm shift for all proton signals (which is probably linked to the different references used in both studies).
In the region between 8.0 and 8.5 ppm, the proton spectrum of MccC51 contains three singlets that integrate for one proton each (8.44, 8.40, and 8.18 ppm). These signals were assigned to the H2, H6, and H8 protons of the nucleoside nebularine, without any experimental proof that these belong to the same system of connectivities. It is astonishing to note that the chemical shift of these resonances corresponds, respectively, to the adenine H8, H2 protons, and to the formyl proton of f-Met in the MccC7 spectrum. A resonance at 12.20 ppm was attributed to the formyl proton of f-Met, arguing that 12.20 ppm is a chemical shift compatible with a formyl signal. No information of scalar coupling with other protons or carbons of the methionine residue is mentioned. However, for MccC7, we were able to demonstrate unambiguously that the signal at 8.12 ppm corresponds to the formyl proton. This assignment is grounded on the scalar (and dipolar) connectivities observed for this proton signal and the methionine protons (and even the ␣ carbon), its intensity and the chemical shift of the carbon attached to it. Furthermore, this signal is also present in spectra of the synthetic peptide f-MRTGNAD and shows the same pattern of connectivities than that seen in MccC7 spectra. That the chemical shift of the formyl proton (and of the rest of the methionine residue protons) is the same in the synthetic peptide and in MccC7, indicates that the C-terminal substituent has no influence on it. Given the similarity of MccC7 and MccC51 spectra (and molecules) it is very unlikely that the chemical shift of the formyl proton could shift as much as ϳ4 The proton chemical shifts of the ribose and the three-methylene chain are identical in MccC7 and MccC51 spectra, taking into account the ϳ0.05 ppm correction. Moreover, the phosphorus signal is coupled to the same proton signals in both molecules spectra (H5Ј5Љ at 4.49 -4.43 ppm and (CH 2 )␣ at 4.23 ppm for MccC51). A comparison of the phosphorus chemical shift is not possible because Metlitskaya and co-workers (41) did not specify the phosphorus reference used in their study. The ribose and three-methylene chain carbon chemical shifts are also very similar for both molecules. 7 The only significant difference is found for the methylene carbon that bears the protons resonating at 3.13 ppm (MccC51 chemical shift), which resonates at 54.5 ppm for MccC51 and at 39.1 ppm for MccC7. Such a big difference in this carbon chemical shift while all the surrounding nuclei resonate at the same position seems surprising to us. In the proposed structure of MccC51, the phosphorus atom is implicated in a disubstituted phosphate. However, the H5Ј5Љ (4.49 -4.43 ppm) and the phosphate-bound methylene (4.23 ppm) proton chemical shifts are in disagreement with this structure. Indeed, in the phosphodiester of 5Ј-adenylic acid and n-aminopropanol (AMPap), which contains this atom arrangement, the H5Ј5Љ and methylene protons are observed, respectively, at 4.15 and 3.94 ppm (27°C, D 2 O, pD 3.3).
With respect to the branching of the peptide and the Cterminal substituent of MccC51, no evidence is given either to support the decision to link them together via the side chain of the C-terminal residue rather than via the peptide main chain or to back the bridging of the peptide and the nucleotide via the three-methylene chain. As the chemical shifts of the C-terminal residue and the methylene protons are the same in the spectrum of MccC7 and MccC51, it is very likely that the peptide and the substituent are linked together in the same way in both molecules.
Finally, it was proposed that the threonine residue of MccC51 contains a hydroxylamine group. Three arguments were put forward to support this conclusion: (i) the observation of "some deviation from normal threonine mobility" for the product of the first cycle of the Edman degradation of the proteolytic peptide "Thr-Gly-Asn-Ala-Asx." (ii) "The mass spectrometry of threonine-containing proteolytic fragments" which "revealed the molecular mass increasing by 15" (which fragments? what proteases were used? Did the proteolytic fragments contain the C-terminal substituent?). (iii) NMR data, which was not shown nor specified. In the present work, we showed unambiguously that the threonine residue of MccC7 is not modified (see "Results"). It is thus surprising to note that the chemical shift of the ␤ proton signal of Thr, as well as that of its ␣ and ␥ protons are identical for MccC7 and MccC51: the replacement of a hydrogen by an amine group having an electronegative nitrogen should influence the chemical shift of the surrounding protons. However, this might not be the case and MccC7 and MccC51 might be different at this position. Nevertheless, it should be pointed out that small variations in elution time are often observed for amino acid phenylthiohydantoin derivatives relative to reference compounds during the sequencing of peptides. In addition, there is a priori no reason to expect that the mobility of a threonine-phenylthiohydantoin derivative might be similar to that of a phenylthiohydantoin derivative of a hydroxylamine-substituted threonine.
In summary, although at present we cannot know if MccC7 and MccC51 are the same molecule or not, they are very similar. The comparative analysis of the results presented in this work (obtained in different conditions of solvent and pH, with many NMR experiences that produced complementary data, and with synthetic analogs of the peptide and the C-terminal substituent of MccC7), with those reported for MccC51 (41) (only in D 2 O, at a single pD value and with relatively few NMR data), permits us to conclude that the chemical structure proposed for MccC51 is not correct. It is noteworthy that it has been suggested that the antibiotic activity of MccC51 might be due, at least in part, to its nebularine unit (41), nebularine being a well known antibiotic (42). However, as stated by Metlitskaya et al. (41), MccC51 and nebularine have different antibiotic activity spectra. This apparent contradiction might be easily explained, if as argued above, MccC51 does not contain a nebularine unit.