Characterization of Cross-linking of Cell Walls of Bacillus subtilis by a Combination of Magic-angle Spinning NMR and Gas Chromatography-Mass Spectrometry of Both Intact and Hydrolyzed l3C= and “N-Labeled Cell-wall Peptidoglycan*

Cross-polarization magic-angle spinning 13C and 16N NMR, rotational-echo double resonance 13C NMR, and delays alternating with nutation for tailored excita-tion-difference “C NMR spectra have been obtained from lyophilized cell walls of Bacillus subtilis grown on a synthetic medium containing D,L-[~-’~C,’~N]~S-partate and D-[l-’3C]alanine. Label from aspartate is incorporated into D-glutamic acid and m-diaminopi-melic acid of cell-wall peptidoglycan, while label from alanine appears in the C-1 positions of both D- and L-alanyl residues. The cross-link index (the fraction of peptide stems joined by an isopeptide covalent bond) is obtained directly from analysis of the results of the 13C NMR experiments. However, specific isotopic enrichments of cell-wall components cannot be obtained from NMR data alone. The latter are determined either from a gas chromatographic-mass spectrometric analysis of the amino acids derived from hydrolysis of cell-wall peptidoglycan, or from a combination of NMR and gas chromatographic-mass spectrometric results. The combined analysis is overdetermined and so involves the least error for evaluations of both

The structural component of bacterial cell walls is a crosslinked polysaccharide known as peptidoglycan (1). The peptidoglycan polysaccharide is derived from a monomeric unit of P-1,4-(N-acetylmuramyl N-acekylglucosamine) and is cross-linked through peptide stems pendant on the muramyl carboxyl group. In Bacillus subtilis polysaccharide the chain length was found to be about 168 dimer units (2). The most common type of cross-linking occurs as an t-amide or a peptide bridge between the C-terminal D-alanine of one stem and a diamino acid of a neighboring peptide stem. The peptide units may be classified as dimers, trimers, etc., depending upon the number of stems linked together. The peptide stem is biosynthesized as a pentapeptide comprising four alternating L-and D-amino acids (one of which is a diamino acid) plus a terminal D-alanine. The diamino acid is almost always * This work was supported by National Science Foundation Grants DIR-8714035 (to J.S.), DIR-8720089 (to J.S.), and DMB-8803687 (to G.E.W.). 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.
L-lysine or m-diaminopimelic acid (Dprn).' In some cases the peptidoglycan is covalently linked to polyribitol or polygly-cero1 phosphate polymers known as teichoic acids.
For different bacteria the sugar dimer is conserved, but'the peptide stems and cross-links provide a source of diversity and a basis for classification (3,4). The peptidoglycan of B. subtilis, classified as Aly, is derived from the pentapeptide L-alanyl-D-isoglutaminyl-m-diaminopimoyl-D-alanyl-D-alanine. It is directly cross-linked by an amide bond between a D-alanine in the fourth position of one peptide stem and Dpm of another (5, 6). Additionally, the cell wall of B. subtilis contains a ribitol teichoic acid esterified in part by D-alanine (7). By determination of Dpm and dinitrophenyl-Dpm from hydrolysates of dinitrophenylated cell walls and by careful chromatography of autolyzed cell walls Warth and Strominger (5) determined that B. subtilis was cross-linked to the extent of 41% with 72% dimers and 7.2% trimers.
In previous work, we developed solids NMR methods to characterize the peptidoglycan of Aerococcus uiridans (8)(9)(10) and the effects of penicillin on its structure (11). The solids NMR methods avoid the loss-prone solubilization and manipulations of cell walls required by other methods of studying peptidoglycan. The peptidoglycan of A. uiridans has a direct cross-link between D-alanine and L-lysine, two amino acids easily labeled by direct incorporation with little scrambling. Cross-polarization magic-angle spinning 15N NMR was used to characterize cross-linking of the peptidoglycan by measuring the ratio of the intensity of the 15N signal from the eamino of uncross-linked L-lysine to the intensity of the camide of cross-linked L-lysine. The peptidoglycan was found to be 49% cross-linked, a value reduced to about 34% when the bacteria were grown in the presence of 0.2 pg/liter penicillin (1 1).
In this paper we report the results of experiments to extend solids NMR to the characterization of cross-links in B. subtilis 6633. A complication in the labeling of B. subtilis 6633 is that the organism does not transport Dpm. Aspartate, a precursor of Dpm, is transported into the cell, but both carbon and nitrogen labels of aspartate are metabolized into purines and the amino acids of the aspartate family as well as Dpm. Aspartate is also a substrate for a transaminase which distributes the nitrogen to amino acids outside the aspartate family.

24485
These complications were overcome by isotopic labeling of both D-alanine and Dpm and by using GCjMS to assess sitespecific activities of labeled cell-wall amino acids.

ATER RIALS AND M E~~O D S
CPMAS 13C and "N NMR-Cross-polarization transfers from protons to either I3C or I 5 N were made under matched spin-lock conditions at 38 kHz with magic-angle spinning at 3.205 kHz (12). Transfer times of 2 ms were used for all the spectra displayed in the figures. Peak intensities were corrected for proton rotating-frame relaxation by systematic variation of transfer times (13). Dipolar decoupling was performed at 95 kHz. Powdered samples with weights between 20 and 100 mg were contained in double-bearing zirconia rotors fitted with Kel-F end and drive caps.
REDOR I3C NMR-The pulse sequence used for REDOR experiments is described in the supplemental material. The single I3C T pulse in the middle of the REDOR carbon-magnetization dephasing period refocuses all isotropic chemical shifts at the start of data acquisition (14). Application of I5N T pulses every half rotor cycle causes a net dephasing of the transverse m a~e t i~t i o n of those carbons dipolar coupled to 15N (15). Weak REDOR difference signals (the difference between 13C rotational-echo intensities with and without dephasing I5N ?r pulses) can be obtained reliably because the operating conditions of the observation channel do not change from scan to scan. REDOR dephasing was summed over four rotor cycles with magic-angle spinning at 3.205 kHz. A four rotor-cycle dephasing period is optimal for detection of directly bonded '3C-15N pairs, for which the REDOR dephasing is appro xi mat el^ 80% of the full-echo signal (16). For I3C and 15N separated by 4 A, the four-rotor cycle dephasing decreases to 0.4% of the full-echo signal. Residual spinning sidebands are not suppressed in REDOR experiments.
CPMAS DANTE I3C NMR-Centerband and sideband families can be inverted by asynchronous DANTE pulse trains whose irradiation sidebands are outside the spectrum of interest (17). Polarization imbalances created by selective DANTE irradiation are equilibrated during the delay period, T , by spin exchange between 13C-'3C dipolar- HzO, 10 mg of FeS04.7H20, 0.2 g of MgSOd. 7Hz0, 11.92 g of KCl, 10.0 g of D-glucose, 1.05 g of (NH4)&04, 40 pg of the antibiotic, cephalothin, and 0.1 g of all 20 common amino acids. The pH was adjusted to 6.1, and the medium was then filter sterilized by passage through a 0.2-pm membrane filter. One-liter growths were performed in an Applikon 2-liter benchtop fermentor (Cole-Parmer Instrument Co., Chicago, IL) operated at 30 "C with 5 p.s.i. air pressure and stirring at 650 rpm. Bacterial growth was initiated by asceptic transfer of cells from a starter culture to an absorbance of 0.04-0.06 at 660 nm. B. subtilis were harvested at mid-log phase, absorbance approximately 3.0. For metabolism studies, bacteria were grown on BSSM in which the natural abundance amino acid was replaced with the labeled amino acid.

B. subtilis Growth on Spizizen Salts Minimal Media-B. subtilis
were grown on Spizizen Salts (19) which contained the following on a per liter basis: 18.34 g of K,HPO,. 3H20, 6.0 g of KH2P04, 0.2 g of MgS04. 7H20, 1.14 g of sodium citrate dihydrate, 1.0 g of NH&l, 5.0 g of D-glucose. The pH was adjusted to a range of 7.0-7.4.
For uniformly I5N-labeled cells the ammonium chloride was replaced hy "NH4C1.
Cell Harvesting and Peptidogly~nn Isolation-Bacteria were harvested by centrifugation at 10,000 X g for 10 min at 4 "C and washed once by resuspension in cold, sterile 0.025 M potassium phosphate buffer, pH 7.0. Washed cells were pelleted by centrifugation at 10,000 X g for 15 min at 4 "C and resuspended in 50 ml of cold, sterile 0.025 M potassium phosphate buffer containing 5 mg of DNase, pH 7.0, and disrupted in the 60-ml chamber of a Bead-Beater (Biospec Products, Bartlesville, OK) one-third full of 0.5-mm diameter glass beads at 0 "C, Glass beads were removed using a coarse sintered glass funnel and washed with 1 liter of a solution containing 0.1 mM EDTA in sterile 0.025 M potassium phosphate buffer, pH 7.0, 4 "C. Centrif-ugation of the filtrate at 17,700 X g for 30 min at 4 "C provided crude cell walls. A suspension of the crude cell-wall pellet in sterile water was added dropwise with stirring to 100 ml of boiling 4% sodium dodecyl sulfate. The suspension was allowed to cool with stirring for 2 h, after which it was allowed to stand unstirred overnight at room temperature, then sedimented by centrifugation at 78,000 X g for 20 min at 20 "C, and washed at least three times with sterile reagent grade water. The pellet was incubated at 37 "C with stirring for 16 h in 50 ml of 0.01 M Tris buffer, pH 8,2, containing 15 mg of trypsin, 15 mg of chymotrypsin, and 5 mg of DNase, then sedimented by centrifugation at 100,000 X g for 1 h at 20 'C and washed three times with water.
Peptidoglycan Hydrolys~-T~ically, a weighed 10.0-to 20.0-mg sample of clean peptidoglycan in a capped 5-ml microproduct vial was hydrolyzed in 6 M HCl at 110 "C for 48 h under nitrogen. The sample was dried in uucuo and rinsed three times with reagent grade water. L-Pipecolic acid, the internal standard, was added to the dry sample, and the entire sample was then derivatized for GC/MS analysis by the method used for amino acid derivatization described below.
ami^ Acid Der~vat~atwn-N-Trifluoroacetyl n-propyl esters of both amino acid standards and amino acids resulting from the complete hydrolysis of B. Deerfield, IL), the D-isomers eluting first. Helium, at a flow rate of 1.5 mllmin, was used as the carrier gas. The injection port was maintained at 250 "C, and the injector was operated in split mode at a ratio of 701. Both the N-trifluoroacetyl and N-acetyl n-propyl ester derivatives of diaminopimelic acid standards and bacterial samples were run isothermally at 200 "C. Typically, 1-p1 injections were made using an autosampler.
Quantification of Alanine, Diaminopimelic Acid, and Glutamic Acid-Cell-wall amino acid compositions were obtained by mass spectroscopic monitoring of the gas chromatographic separations of the N-t~fluoroacetyl n-propyl ester derivatives of hydrolyzed bacterial cell walls (21). The G1u:D-A1a:Dpm ratios were obtained from the total ion current by comparisons to mixtures of known composition. L-Alanine coeluted with an unknown material which accounted for about 10% of the total ion current of the peak. Therefore the D-Ala:L-Ala ratio was obtained by selective ion monitoring of the peak at m/e 140, a peak absent from the spectrum of the contaminant. The specific activities of D-and ~-[l-'~C]alanine and ~-['~N]glutamate were determined using multiple line pair analyses as previously described (22).
Isotopomer Analysis of Diaminopimelic Acid by GCIMS--Ion clusters for fragments containing all labeled carbon and nitrogen atoms were found at mlz 406, 379, 353, and 320; for fragments containing both a-carbon atoms and a single nitrogen atom, at mlz 266, 224, 206, and 178; and for fragments containing one of the a-carbon atoms and one of the nitrogen atoms, at m/z 152. Intensities of the p, p+l, and p+2 peaks obtained in scan mode, after application of a natural abundance correction, could be solved for the isotopomeric specific activities E-I, Table I.  1, top). The CPMAS 13C cell-wall spectrum obtained from B.    subtilis grown on medium containing ~~-[2-'~C,''N]aspartate and D-[l-13C]alanine shows increased intensity only a t two frequencies: 6~ 60 and 175 (Fig. 1, bottom), corresponding to incorporation of label in a-carbons (from aspartate) and C-1 carbons (from alanine) of peptidoglycan. There is no indication of scrambling of 13C label (see Miniprint). The selectivity of routing of 15N label from aspartate to specific sites in nucleic acid bases and selected amino acids (see Miniprint) is consistent with the simplicity of the CPMAS I' N NMR spectra of cell walls grown on ["Nlaspartate (Fig. 2). The spectrum has only two peaks, arising from amine and amide nitrogens, with resonances at bN 18 and 100, respectively. Virtually all of the observed intensity arises from label. REDOR 13C NMR-Only two REDOR difference peaks (& 60 and 175) are observed in the spectrum of cell walls from bacteria grown on media containing a combination of D,L-[2-13C,'5N]a~partate and D-[l-'3C]alanine (Fig. 3). The other minor peaks in the REDOR difference spectrum arise from spinning sidebands.
DANTE-Difference 13C NMR-The carbonyl-carbon peak and its spinning sidebands are inverted (Fig. 4, bottom)  ence spectrum ( T = 50 ps minus 'T = 30 ms) has a negativegoing peak a t & 175 and a positive-going peak a t 6~ 60 ( Fig.   4, top). This indicates a transfer of polarization from acarbons to carbonyl carbons that are separated by two bonds (22). Because there was no DANTE difference observed for 'T = 50 ps minus T = 3 ms (data not shown), one-bond 13C-13C separations are not present. Resonances from isolated "Cs have disappeared in the difference spectrum of Fig. 4.
Concentrations of Labels-For the purpose of NMR analysis we describe the distribution of peptidoglycan labels in terms of the concentrations of 13C and "N per peptidoglycan peptide stem (Fig. 5). This description is independent of the sugar composition of the isolated cell-wall material and is also independent of the number of terminal D-alanines of peptide stems. Neither of these quantities is known with certainty. peptide stem that are l3C-1abeled; Clz is the number of C-l carbons of L-alanine that are %-labeled; Nlz is the number of a-nitrogens of D-glutamic acid that are "N-labeled; C , is the number of a-carbons of Dpm that are 13C-laheied and are also directly bonded to a Dpm I6N; NZl (which equals Czl) is the corresponding number of a-nitrogens of Dpm that are "N-labeled and are also bonded to a Dpm I3C; C,, is the number of a-carbons of Dpm that are '%-labeled but are not bonded to a Dpm I5N; and NZ2 is the number of a-nitrogens of Dpm that are I5N-labeled but are not bonded to a Dpm I3C. The cross-link index is p. on a per stem basis, Clz is the number of C-1 carbons of Lalanine that are 13CC-labeled; Nlz is the number of a-nitrogens of D-glutamic acid that are 15N-labeled; Czl is the number of a-carbons of Dpm that are 13C-labeled and are also directly bonded to a Dpm 15N; Nzl (which equals Czl) is the corresponding number of a-nitrogens of Dpm that are "N-labeled and are also bonded to a Dpm 13C; Cz2 is the number of acarbons of Dpm that are 13C-labeled but are not bonded to a Dpm 15N; and N22 is the number of a-nitrogens of Dpm that are "N-labeled but are not bonded to a Dpm 13C. The crosslink index is p.

~M R I n t e~i t~
Ratios-We describe intensity ratios observed in the 13C and 15N NMR experiments in terms of ratios of Cij values.
3. The ratio of the REDOR I3C difference intensity to fullecho intensity at 6c 175 (Fig. 3) gives the ratio of 13C-15N double-labeled peptide bonds to 13C single-labeled peptide bonds, R3 = K~C I Z N I~ where fA is the isotopic specific activity of ~-[l-'~C]Ala in peptidoglycan. The second term in the numerator for the expression for R3 contains fA rather than Cl1 because crosslinks do not form to D-Ala units in the fifth position of pentapeptide stems.
4. The ratio of the REDOR 13C difference intensity to fullecho intensity at SC 60 ( Fig. 3) gives the ratio of 13C-"N double-labeled a-carbons to 13C single-labeled a-carbons,

5.
The ratio of the DANTE 13C difference intensity at SC 60 ( Fig. 4) to the carbonyl-carbon intensity at 6,175 gives the ratio of I3C-labeled a-carbons with "C-labeled carbonyl carbons that are two bonds away to the total I3C-labeled carbonyl carbons, RS = O.b%(cz~ + c,zt/'(C~, + CIII

(5)
In writing these ratios, we assume that all natural abundance contributions to integrated intensities have been removed by subtraction of spectra of unlabeled cell-wall sam- Determination of the Cross-link Index-The cross-link index, p, can be determined directly by NMR data using the product, R1. R5. Thus, from the data of Figs. 1 and 4, we obtain p = 0.33 for cell walls of B. subtilis grown in the presence of cephalothin. (In a subsequent paper we will describe the effect on B. subtilis peptidoglycan of growth in media with and without added cephalothin).
The distribution of label represented by Fig. 5 involves seven parameters, while only five parameters can be evaluated from NMR data. Thus, although the cross-link index can be determined from NMR data alone, no tests of internal selfconsistency are available. A more reliable procedure is to combine the results of NMR and GC/MS experiments. For example, if we use the isotopic specific activities of D-and Lalanine of Table I, then ci1 = Cafa and c1z = CBfB, (6) where the ci values are relative molar concentrations of amino acids from hydrolyzed peptidoglycan determined by GC, and the f; values are isotopic specific activities determined by MS of the A, B, . . ., J fragments identified in Table I Three internal self-consistency checks are now available. From the NMR-determined Cij and NG values and the GCdetermined ci values, we can predict the isotopic specific activities ( f , values) of all MS-observed fragments (except those of D-and L-alanine which were used to evaluate the R3 NMR ratio). The three observed and predicted isotopic specific activities are in agreement (Table 11).  the cross-link index, p, with either RZ, R3, or R5, and the appropriate ci and f i values. For example, using R z and the GC/MS data of Table I pertaining to D-glUtamiC acid and Dpm, we obtain p = 0.30. We believe that this is the most reliable combination of data to use because it involves the least complicated NMR experiment, and only MS fragments (or combinations of fragments) that produce high ion currents. Using R5 (a ratio which requires the most complicated correction factor, K5) we also obtain p = 0.30. More reliable evaluations of R5 may soon be possible using new magic-angle spinning 13C NMR experiments to measure the concentration of isolated homonuclear pairs of spins (23). These experiments rely on dephasing rather than polarization transfer.

Strategies for
The carbonyl-carbon REDOR ratio, R3, leads to p = 0.12. This ratio is dominated by the many 13C-15N bonds formed between L-alanine and D-glutamic acid (dark lines, Fig. 5) and so is the least sensitive of the three intensity ratios that depend on the cross-link index, p . A cross-link index as small as 0.12 seems implausible in view of the other values and the measurements of Warth and Strominger (7,8).
If we assume the NMR-determined p of 0.33 (from the product of R1 and R2), we can use all the GC/MS data to predict the observed intensity ratios from the five types of NMR experiments. The observed and predicted values are in reasonable agreement (Table 111). We take the degree of selfconsistency of Table I11 as the measure of NMR and GC/MS experimental accuracy. The cross-link index is therefore determined to k0.03.
In conclusion, the most reliable determination of the crosslink index for cell walls of B. subtilis combines simple CPMAS 15N NMR data (the amide to amine intensity ratio, Rz) with GC/MS data on the relative concentrations and isotopic enrichments of D-glutamic acid and diaminopimelic acid. Results from experiments involving either single-or doublelabeled aspartate can be compared directly. Such experiments can be performed in any laboratory equipped with conventional solid state NMR and GC/MS instruments. r. which is followsd by data acquisition with 'H-"C dipolar decaupling.