Báo cáo khoa học: Epitope mapping of the O-chain polysaccharide of Legionella pneumophila serogroup 1 lipopolysaccharide by saturation-transfer-difference NMR spectroscopy pot

Epitope mapping of the O-chain polysaccharide of Legionella pneumophila serogroup 1 lipopolysaccharide by saturation-transfer-difference NMR spectroscopy Oliver Kooistra 1 , Lars Herfurth 2 , Edeltraud LuÈ neberg 3 , Matthias Frosch 3 , Thomas Peters 2 and Ulrich ZaÈ hringer 1 1 Research Center Borstel, Center for Medicine and Biosciences, Germany; 2 Institute for Chemistry, Medical University of Lu È beck, Germany; 3 Institute for Hygiene and Microbiology, University of Wu È rzburg, Germany Two modi®cations of 5-acetimidoylamino-7-acetamido- 3,5,7,9-tetradeoxy- D -glycero- D -galacto-non-2-ulosonic acid (5-N-acetimidoyl-7-N-acetyllegionaminic acid) in the O-chain polysaccharide (OPS) of the Legionella pneumophila serogroup 1 lipopolysaccharide (LPS) concern N-methyla- tion of the 5-N-acetimidoyl group in legionaminic acid. Both N-methylated substituents, the (N,N-dimethylacetimidoyl) amino and acetimidoyl(N-methyl)amino group, could be allocated to one single legionaminic acid residue in the long- and middle-chain OPS, respectively. Using mutants devoid of N-methylated legionaminic acid derivatives, it could be shown that N-methylation of legionaminic acid correlated with the expression of the mAb 2625 epitope. In the present study we investigated the binding of the LPS-speci®c mon- oclonal antibody mAb 2625 to isolated OPS with surface- plasmon-resonance biomolecular interaction analysis and saturation-transfer-dierence (STD) NMR spec troscopy in order to map the mAb 2625 epitope on a molecular level. It could be demonstrated that the b inding anity of the N-methylated legionaminic acid derivatives was indepen- dent from the size of the isolated OPS molecular species. In addition, STD NMR spectroscopic studies with polysac- charide ligands with an average molecular mass of up to 14 kDa revealed that binding was mainly mediated via the N-methylated acetimidoylamino group and via the closely located 7-N-acetyl group of the respective legionaminic acid residue, t hus indicating these derivatives to represent t he major epitope of mAb 2625. Keywords: lipopolysaccharide; Legionella pneumophila; bioanity studies; NMR. Legionella pneumophila is a facultative intracellular Gram- negative bacterium and the cause of legionellosis, a pneu- monia with a sometimes fatal progression [1]. The reservoirs of legion ellae a re natural or man-made water systems and their natural hosts are various amoebae species [2]. In the human lung L. pneumophila invades and replicates within alveolar macrophages [3]. The serogroup-speci®c antigens of the Gram-negative legionellae reside in the lipopolysacchar- ide (LPS) of the outer membrane [4,5]. The O-chain polysaccharide (OPS) of serogroup (Sg) 1 LPS is a homopolymer of the 5-N-acetimidoyl-7-N-acetyl derivative of 3,5,7,9-tetradeoxy- D -glycero- D -galacto-non-2- ulosonic acid, termed legionaminic acid (Fig. 1, structure 1) [6,7], wh ich i s quantitatively 8-O-acetylated in strains belonging to the Pontiac group [5,6,8], but only partially in other Sg 1 strains of the non-Pontiac group [5,9]. In L. pneumophila Sg 1 LPS the OPS is linked to a terminal nonreducing L -rhamnose (Rha II ) of the core oligosaccharide [10,11]. The core of the LPS lacks heptose and phosphate, contains abundant 6-deoxy sugars and N- acetylated amino sugars, and is highly O-acetylated [9±12]. Recently, a phase-variable expression of a virulence- associated LPS epitope of L. pneumophila has been described previously [13]. Chromosomal insertion and excision of a 29-kb unstable genetic element, possibly o f phage origin, was identi®ed as the molecular mechanism for phase variation [ 14]. The altered LPS phenotype of the spontaneous phase variant could be distinguished with the aid of the LPS-speci®c mAb 2625. The reactivity of mAb 2625 w as related to the presence of N-methyl groups at the 5-N-acetimidoyl group of legionaminic acid, a modi®cation of bacterial polysaccharid es, which is described for the ®rst time in the accompanying paper [ 15]. The c omponents identi®ed were the 5-N-(N,N-dimethylacetimidoyl)-7-N- acetyl- and 5-N-acetimidoyl-5-N-methyl-7-N-acetyl- deriv- atives of legionaminic acid (Fig. 1, structures 2 and 3, respectively) probably located proximal to the core oligo- saccharide of long and middle O-chain LPS from wild-type RC1 [15]. Although serological data strongly indicate that the N-methylated derivatives of legionaminic acid are located close to the outer region of the core oligosaccharide, their precise position could, unfortunately, not be deter- mined [15]. N-Methylation was limited to one single Correspondence to U. Za È hringer, Forschungszentrum Borstel, Zentrum fu È r Medizin und Biowissenschaften, Parkallee 22, D-23845 Borstel, Germany. Fax: + 49 4537 188612, Tel.: + 49 4537 188462 , E-mail: uzaehr@fz-borstel.de Abbreviations: LPS, lipopolysaccharide; O PS, O-cha in polysaccharide; PS, polysaccharide; Sg, serogroup; GPC, gel-permeation chromato- graphy; Kdo, 3-deoxy- D -manno-oct-2-ulosonicacid;Rha, L -rhamnose; SPR, surface-plasmon-resonance; STD, saturation-transfer-dier- ence; EXCY, exchange spectroscopy; FID, free induction decay. (Received 8 August 2001, revised 13 November 2001, accepted 16 November 2001) Eur. J. Biochem. 269, 573±582 (2002) Ó FEBS 2002 legionaminic acid residue of each polysaccharide chain above a certain length, and was absent from short O-chain LPSs of wild-type RC1, from the LPS of a s pontaneous phase variant (strain 811), and an isogenic mutant ( strain 5215) [15]. In the present study we investigated the binding of the antibody to the isolated OPS with surface-plasmon- resonance (SPR) biomolecular interaction analyses and saturation-transfer-difference (STD) NMR spectroscopy in order to determine binding af®nity and the binding epitope of the mAb 2625. Because it has not been so far possible to depolymerize the polylegionaminic acid OPS [6] or to deconvolute the polymers, the various legionaminic acid derivatives could not be isolated as monomers or as homogeneous polymers, respectively, f or separate investi- gations. But with the aid of STD NMR spectrosc opy, a new method for c haracterization of ligand binding [16], it could be shown that mAb 2625 binds directly to the N-methylated structures in the polymer. This is the ®rst description of antibody-LPS binding examined by STD NMR spectroscopy and shows the advantages of this direct approach for the purpose of relatively quick and direct epitope mapping. MATERIALS AND METHODS Bacterial strains, cultivation, and extraction of LPS L. pneumophila virulent wild-type strain RC1 (Sg 1, sub- group OLDA) is a clinical isolate described previously [13]. Avirulent strain 811 is a spontaneous phase variant derived from wild-type RC1 [13]. Mutant strain 5215 was con- structed by deletion of the Orf 8±12 operon required for the biosynthesis of the mAb 2625 epitope from wild-type RC1 as described in the accompanying paper [15]. All s trains were grown on buffered charcoal yeast extract agar with a-growth supplement (Merck). LPS was extracted from enzyme-digested cells by a m odi®ed pheno l/chloroform/ petroleum ether procedure as described previously [6,17]. Preparation, modi®cation, and fractionation of PS and PS NH4OH/HF LPS each of wild-type RC1, mutant 5215, and phase variant 811 was degraded at 100 °Cfor2.5hwith0.1 M NaOAc/ HOAc buffer (pH 4.4, 10 mgámL )1 LPS), and the resultant lipid A was removed by centrifugation (5000 g,30min). The supernatant was lyophilized and fractionated by gel- permeation chromatography (GPC) on a column (2.5 ´ 120 cm; Bio-Rad) of Sephadex G-50 (S) (Pharmacia) using 50 m M pyridinium/acetate buffer (pH 4.3) and mon- itoring with a differential refractometer (Knauer). Fractions corresponding to long- a nd short-chain polysaccharide (PS, i.e. OPS linked to the core oligosaccharide), core oligosac- charide, and m ono- and disaccharides contaminated with salt were pooled and lyophilized. The PS portion was de-O-acetylated (20%, v/v, aqueous NH 4 OH, 37 °C, 16 h) and treated with 48% (v/v) aqueous hydro¯uoric acid (HF, 4 °C, 168 h) in order to selectively cleave the glycosidic linkage of 6-deoxy sugars [18] to obtain PS NH 4 OH/HF as described in the accompanying paper [15]. PS NH 4 OH/HF was fractionated by tandem GPC to long-, middle-, and short-chain molecular species [15]. Preparation of mAb 2625 For production of mAb 2625, the hybridoma cell line was propagated in Dulbecco's minimal essential cell culture medium (Biochrom) supplemented with 10% heat-inacti- vated fetal bovine serum (Biochrom). The culture s uperna- tant was tested for the presence of mAb 2625 in a colony blot assay, before antibody puri®cation was carried out. Anti- body puri®cation was performed using a HiTrap protein G column (Pharmacia) with a GradiFrac system device (Pharmacia). Antibodies were eluted from the protein G resin w ith 0.1 M glycine and eluted fractions were neutral- ized with 1 M Tris/HCl buffer (pH 9). Fractions were pooled and dialysed against phosphate buffer (137.9 m M NaCl, 2.7 m M KCl, 8.1 m M Na 2 HPO 4 ,1.5m M KH 2 PO 4 ,pH 7.4). Fig. 1. Proposed structure of Legionella pneu- mophila PS NH 4 OH/HF from wild-type RC1. 1,5-N-acetimidoyl-7-N-acetyllegionaminic acid; 2-E,5-N-(N,N-dim ethylacetimidoyl)- 7-N-acetylaminolegionaminic acid ( the descriptors cis and trans designate the posi- tions of the N-methyl groups relative to N 2 ); 3-E and 3-Z, stereoisomers of 5-N-acetimi- doyl-7-N-acetyl-5-N-methyllegionaminic acid. The reducin g Rha I residue is only present in 70% with a-andb-con®guration in a ratio of approximately 5 : 1, which is also the case for free Rha II in the other 30% of the molecules. The anomeric con®guration of the ketosidic linkage of the legionaminic acid residue attached to Rha II may be dierent and the position of the N-methylated legionaminic acid derivatives have not been con®rmed. n is 40 on average for long-chain PS NH 4 OH/HF and 18 on average for middle-chain PS NH 4 OH/HF . 574 O. Kooistra et al. (Eur. J. Biochem. 269) Ó FEBS 2002 The protein concentration was determined using the bicinchoninic acid protein assay reagent kit (Pierce). NMR spectroscopy 1D 1 H NMR and STD spectra were recorded with a Bruker Avance DRX-600 or DRX-500 spectrometer. Standard Bruker software was used to acquire and process the NMR data. Polysaccharide samples were lyophilized three times from 2 H 2 O and measured in 2 H 2 O( 2 H, 99.996%; Cambridge Isotope Laboratories) at 27 °C. Chemical shifts were re fer- enced to external acetone (d H 2.225 p.p.m.; d C 31.45 p.p.m.). For analysis of temperature and pH dependence of the N-methyl signals long-chain PS NH 4 OH/HF from wild-type RC1 was dissolved in 10% deuterated water, the pH was adjusted within a range of pH 2 to pH 11 with 1 M HCl or 1 M NaOH and recording 1D 1 H NMR spectra at constant temperature ( 275 K). At pH 7.5 1D 1 H NMR spectra were recorded at temperatures between 283 and 323 K in 10-K intervals. MAb 2625 was ultra®ltrated 10 times using a 6-mL 10-kDa molecular mass cut-off Vivaspin centrifugal concen- trator device (Sartorius) with deuterated phosphate buffer composed as described above. The NMR samples were adjusted to a mAb 2625 concentration of 16.2 l M based on the U V absorption at 280 nm. A 20-fold ligand excess (640 l M ) over binding sites was used throughout the studies. The time dependence of the saturation transfer was investigated by recording STD spectra with 1 k scans and saturation times from 0.25 s to 5 s. Relative STD values were calculated by dividing STD signal intensities by the inten- sities of the corresponding signals in a 1D 1 HNMRreference spectrum of the same sample reco rded with 512 scans. STD NMR spectra for epitope mapping were acquired using a series of equally spaced 50 ms Gaussian shaped pulses for saturation, with 1 ms delay betw een the pulses, and a total saturation time of approximately 3 s. The frequency of the protein (on-resonance) irradiation was set to the maximum of the broad hump of overlapping protein 1 H NMR signals in the a romatic region 7 .2 p.p.m. It was t ested t hat no a mido- and amidino-protons (6.51 and 7.88 p.p.m., respectively, as measured in 10% deuterated water) of the ligand were irradiated by this setting of the on-resonance frequency. The off-resonance irradiation frequency was set at 33.0 p.p.m. Free induction decay values (FIDs) with on- and off- resonance protein saturation were recorded in an alternating fashion. Subtraction was achieved via phase cycling. A total relaxation delay of 4.3 s and 128 dummy scans were employed to reduce subtraction artefacts. The overall measurement time using 6 k scans was approximately 12 h. Protein resonances were suppressed by application of a 15-ms low power spin-lock pulse prior to acquisition. Residual 1 H 2 HO was not suppressed. STD NMR spectra were recorded at 300 K. In order to assess the temperature dependence, STD NMR spectra were also recorded a t 293 an d 315 K . Surface-plasmon-resonance (SPR) biomolecular interaction analyses SPR analyses were carried out using an automated BIAcore 3000 biosensor instrument (BIAcore). mAb 2625, an IgG, was immobilized on a research grade CM5 senso r chip in 10 m M sodium acetate ( pH 4.5) using the amine coupling kit supplied by the manufacturer (BIAcore). Unreacted moieties were blocked with ethanolamine. A control surface with an anti-myoglobin IgG (BIAcore) was prepared in the same manner. All measurements were performed in 10 m M Hepes buffer (pH 7.4) containing 150 m M NaCl and 0.005% (v/v) polysorbate 20 (BIAcore) at a ¯ow rate of 10 lLámin )1 . Surfaces were regenerated by normal dissoci- ation or with distilled water. Sensorgram data were analysed using the BIAevaluation 3.0.2 s oftware (BIAcore). Binding af®nity (K d ) was determined by steady state af®nity line- ®tting based on end point values at equilibrium binding of a series of sensorgrams generated with at least seven ligand concentrations ranging from 1.5 l M to 875 l M and with each concentration measured at least twice. Alternatively, K d values were determined by linear regression of Scatchard plots. RESULTS Preparation and characterization of ligands LPS of L. pneumophila wild-type RC1 subjected to mild acid hydrolysis is cleaved at the ketosidic linkage of 3-deoxy- D -manno-oct-2-ulosonic acid residues (Kdo I and Kdo II )and in some molecules at the ketosidic linkage between the legionaminic acid of the OPS a nd Rha of the core oligosaccharide, to release lipid A, a lateral a- D -man- nose II -(1 ® 8)-Kdo II disaccharide, a major heptasaccharide core fragment, OPS, and PS (i.e. O PS linked t o the core heptasaccharide), respectively [9±11]. In the majority o f the molecules, OPS was attached to the core heptasaccharide. The PS was fractionated by G PC to long- and short-chain molecular species, the latter containing also middle-chain molecular species. Part of the long-chain PS was used without further treatment for STD NMR spectroscopy experiments (see below). The rest of the PS was de-O- acetylated to remove abundant O-acetyl groups in the linkage region between the core o ligosaccharide and the OPS [9,10], subsequently subjected to HF-treatment to cleave the glycosidic linkage of the 6-deoxy sugars, e.g. L -rhamnose, between the core oligosaccharide and the OPS [15,18], and fractionated by tandem GPC. By this procedure long-chain, a low amount of middle-chain, and short-chain PS NH 4 OH/HF devoid of most core sugars were isolated [15]. LPS from mutant 5215 and phase variant 811 was degraded by the same procedure. As described, the isolated PS NH 4 OH/HF contained only Rha linked as ® 3)-a- L -Rha II - (1 ® 3)- L -Rha I disaccharide (70%) or as ® 3)- L -Rha II monosaccharide (30%) to polylegionaminic acid [15] (Fig. 1 ). Only the long- a nd middle-chain PS (as well as PS NH 4 OH/HF ) from wild-type RC1 contained legionaminic acid derivatives N-methylated at the 5-acetimidoylamino group, which were absent from short-chain OPS of wild- type RC1, the entire OPS of mutant 5215, and only found in traces in the OPS of phase variant 811. Long-, middle-, a nd short-chain PS NH 4 OH/HF from wild- type RC1 were investigated by 1D 1 H NMR spectroscopy and signal integration was performed to calculate the average chain-length of the PS NH 4 OH/HF and the distribution of N-methylated legionaminic acid derivatives [15]. Integra- tion of the signals of 1D 1 H NMR spectra indicated that the Ó FEBS 2002 Epitope mapping with mAb 2625 (Eur. J. Biochem. 269) 575 average chain-length of long-, middle-, and short-chain PS NH 4 OH/HF is about 40, 18, and 10 legionaminic acid res- idues [15], resulting in a calculated average molecular mass of approximately 12.9 kDa, 6.0 kDa, an d 3.4 kDa, respec- tively. The ratio of the 5-N-(N,N-dimethylacetimidoyl)-7-N- acetyl and 5-N-ace timidoyl-5-N-methyl-7-N-acetyl deriva- tives of legionaminic acid was 1 : 1 in long-chain and 1 : 2 in middle-chain PS NH 4 OH/HF , respectively. Based on the rela- tive intensities of the proton signals it was concluded that only one legionaminic acid residue is N-methylated in each polysaccharide chain above a speci®c length. The proposed structure of PS NH 4 OH/HF from wild -type RC1, which was used for SPR analyses and STD NMR spectroscopy experiments i s presented in Fig. 1. The PS NH 4 OH/HF from mutant 5215 and phase variant 811 had the same chain- length as that from wild-type RC1 [15]. SPR studies with immobilized mAb 2625 In order to investigate the binding behaviour of mAb 2625, binding af®nity (K d ) was determined by SPR for the binding to immobilize d mAb 2625 of isolated long- and middle- chain PS NH 4 OH/HF from L. pneumophila wild-type RC1, mutant 5215, and phase variant 811. The PS NH 4 OH/HF from wild-type RC1 bound to mAb 2625 with a rapid association and dissociation to and from the antibody, typical for low± af®nity interaction like antibody-carbohydrate binding [19]. The K d value for middle-chain PS NH 4 OH/HF from wild-type RC1 determined at equilibrium binding was 26 l M (Fig. 2 , top panel), determination by linear regression analysis of Scatchard plot gave a value of 21 l M (Fig. 2, bottom panel). Direct comparison of long- and middle-chain PS NH 4 OH/HF from wild-type RC1, obtained with less data points, generated K d values in the same range: 43 l M (Scatchard: 43 l M ) for long-chain PS NH 4 OH/HF and 30 l M (Scatchard: 31 l M ) for middle-chain PS NH 4 OH/HF . The sensorgrams for long-chain PS NH 4 OH/HF had a similar square pulse form as that for short-chain PS NH 4 OH/HF . With the long- and middle-chain PS NH 4 OH/HF from mutant 5215 and phase variant 811, n o measurable af®nity could be determined. Resonance units were not higher than those obtained with buffer as control experiments. STD NMR experiments of middle-chain PS NH 4 OH/HF from wild-type RC1 in the presence of mAb 2625 To describe the epitope responsible for the binding inter- action of the polysaccharide chain with mAb 2625 at atomic resolution, both PS NH 4 OH/HF and PS were investigated by STD NMR experiments. Because of the complexity of the ligand molecules, initial investigations were done using the smaller, apparently less complex middle-chain PS NH 4 OH/HF and were completed with long-chain PS (see below) in order to study the in¯uence of the carbohydrate polymer chain on binding. Four samples were prepared, two contained t he binding middle-chain PS NH 4 OH/HF from wild-type RC1 wit h and without mAb 2625 and two contained the middle-chain PS NH 4 OH/HF from mutant 5215 lacking the N-methyl groups also with and without antibody. The latter PS NH 4 OH/HF did not show binding activity in SPR experiments. Optimization of the experimental set-up for STD NMR spectroscopy was achieved using samples without any mAb present. In that case, STD spectra did not contain ligand signals, because saturation transfer does not occur without the protein (data not shown). Investigation of the time dependence of the saturation transfer with saturation times from 0.25 s to 5 s showed that 3 s was suf®cient for ef®cient transfer of saturation from the protein to the ligand protons (Fig. 3). The signals of all N- and C-linked methyl groups present in STD spectra showed similar behaviour. Only the sample containing mAb 2625 and middle-chain PS NH 4 OH/HF from wild-type RC1 showed signi®cant sat- uration transfer from the protein to the ligand in the STD spectra (Fig. 4B). Comparison of the STD spectrum with the corresponding 1D 1 H N MR spectrum (Fig. 4A) c learly demonstrated the involvement of the N-methyl groups of the N-methylated legionaminic acid derivatives 2, 3-E and 3-Z in binding. Investigation of the time dependence revealed that saturation transfer to the two N-methyl groups in 2 was identical and reached a maximum STD of  15%. The maximum values for 3-E and 3-Z were considerably lower,  7 and 10%, respectively (Fig. 3A). Therefore, the N-methyl groups in 2 showed a twofold more effective saturation transfer compared to the ones in 3-E and 3- Z. Similar effects were observed for 1 HNMRsignalsofthe C-methyl groups of the N-acetimidoyl a nd N-acetyl groups in 2, 3-E and 3-Z (Fig. 5). The signals were partially superimposed by the i ntense resonances of the correspond- ing methyl groups of the major component in the mixture, legionaminic acid 1. Signals for the C-methyl group of the N-acetimidoyl group in 2, 3-E and 3-Z reached a maximum Fig. 2. Surface-plasmon-reson ance a na lysis o f middle-chain PS NH 4 OH/HF from wild-type RC1 with immobilized mAb 2625. Steady-state anity line-®tting based on end point values at equilibrium binding obtained with 14 ligand concentrations between 1.5 l M and 292 l M (A). Scat- chard analysis based on the same d ata (B). 576 O. Kooistra et al. (Eur. J. Biochem. 269) Ó FEBS 2002 STD effect o f  9, 6.6, and 7.5%, respectively (Fig. 3B). Signals for the C-methyl group of the N-acetyl group reached a maximum STD effect of  9% in 2,and6%in 3-E (Fig. 3C). The assignment of the N-acetyl group of 2 was solely based on the STD NMR experiments. Further- more, one signal of the N-acetyl group of 3-Z could not be identi®ed unambiguously, either in the 1D 1 HNMR spectrum or in t he STD NMR spectrum, and one signal (d H 2.22) in the STD NMR spectrum showing signi®cant saturation transfer ( 6% maximum STD effect) could not be assigned at all. Proton signals from t he pyranose ring or the side chain of the N-methylated legionaminic acid derivatives could not be assigned unequivocally due to noise. For the major nonmethylated legionaminic acid (1), only a signal for H9 and for the other two groups with the most intense signals in the 1D 1 H NMR spectrum, i.e. the C-methyl groups of the N-acetimidoyl group and the N-acetyl group, respectively, were observed. However, maximum STD e ffects for these signals were rather lo w ( 3 and 2%; Fig. 3 B,C) and, furthermore, these signals were also de tected as the only signals in the STD spectrum of the middle-chain PS NH 4 OH/HF from mutant 5215 (Fig. 4 D). Most probably, the STD NMR signals of the C-methyl groups of the N-acetimidoyl group and the N-acetyl group of 1 were due to relaxation artefacts. Signals of the Rha protons were not present in the STD spectra, which is most obvious for the signals of the anomeric proton s and the methyl protons of the 6-deoxy groups because these signals are well separated in the corresponding 1D 1 H NMR spectra. Therefore, partici- pation in binding of these residues located at the reducing end of PS NH 4 OH/HF could not be con®rmed by our experiments. The 1D 1 H NMR spectra of mAb 2625 together with middle-chain PS NH 4 OH/HF from both strains (Fig. 4A,C) showed signals belonging most likely to glycerol. They were not present in the corresponding STD spectra (Fig. 4B,D) because glycerol does not bind to the mAb. Glycerol probably originated from the membrane of the centrifugal concentrator device or from ®lters used during the prepa- ration of the samples or mAb 2625. Temperature and pH dependence of 1D 1 H NMR signals of N-methyl groups To measure the temperature and pH dependence of the signals o f the N-methyl groups in 2 and 3 due to chemical exchange [20], 1D 1 H NMR spectra of long-chain PS NH 4 OH/HF from strain RC1 were recorded under various conditions. It was observed that both changes of the pH at constant temperature and changes o f the temperature at appropriate constant pH in¯uenced the form of the signals in a similar manner. Lowering the pH had a similar effect as a decrease in temperature and vice versa, although the latter couldbebettermonitoredinsmallsteps. At constant temperature (275 K), the four separated N-methylsignalscouldbeobserveduptopH 7and beginning with pH  8 the lower-®eld pair of signals of 2-E [d H 3.30 (trans)andd H 3.19 (cis)] broadened and began to coalesce, so that from pH  9 only one sharp signal was detected (Fig. 6A±F). The higher-®eld pair o f N-methyl signals [d H 3.03 (3-Z)andd H 2.95 (3-E)] remained unchanged at high pH, although at l ow pH (pH  2) it seemed that the ratio of the signals, balanced at neutral pH, was slightly changed towards the 3-E isomer. On the other hand, at constant pH 7.5 the increase of the temperature from 283 K in 1 0-K steps to 323 K showed that the two separated signals for the N-me thyl groups of 2 broadened, coalesced, and ®nally were observed as one sharp signal with an average chemical shift ( Fig. 6G±M). The two separated signals of the N -methyl group of 3-E and 3-Z did not signi®cantly change within this range (Fig. 6 G±M). T he N-methyl signals of 3 did not change even under the drastic conditions pH  11 and 323 K, a pH Fig. 3. Time dependence of magnetization transfer for selected saturated signals of methyl groups of the legionaminic acid derivatives. Th e time dependence for the N-methyl groups (A) and the C-linked methyl groups of the a cetimidoylamin o (B) and the acetamido (C) substitu- ents, respectively, of 2 (s), 3-E (n), 3-Z (e), and 1 (h) are shown. The two signals of the N-methyl groups of 2 showed identical behaviour. A signal of the C-methyl group of the acetamido group in 3-Z could not be identi®ed unambiguously, and one signal (´) could not be assigned to any proton. Magnetization transfer for the C-methyl groups of 1 probably accounts for relaxation artefacts. Ó FEBS 2002 Epitope mapping with mAb 2625 (Eur. J. Biochem. 269) 577 at which t he N-methyl signals of 2 already coalesced at low temperature (283 K; Fig. 6F). The signals of the major nonmethylated legionaminic acid (1) were not signi®cantly changed apart from better resolution at low pH or high temperature. STD NMR experiments with mAb 2625 together with long-chain PS from wild-type RC1 at different temperatures STD NMR spectroscopy experiments with t he long-chain PS from strain RC1 were performed for several reasons. After mild acid hydrolysis of the LPS without further degradation, it is dif®cult to obtain middle-chain PS, which can only be i solated as a mixture with short-chain PS [9], which in contrast t o long- and middle-chain PS, does not contain N-methylated l egionaminic acid derivatives [15]. Long-chain PS on the other hand, which quantitatively contains one N-methylated legionaminic acid derivative, could be isolated as a well-separated fraction. Furthermore, Fig. 4. 1D 1 HNMR(AandC)andSTD NMR (B and D) spectra of middle-chain PS NH 4 OH/HF from wild-type RC1 (A an B) and mutant 5215 (C and D) in the presence of mAb 2625. Low intensity signals in the spectrum in (D) are probably due to subtraction artefacts of the originally most intense proton signals of the N-acetimidoyl and N-acetyl grou ps in 1, respectively. Spectra were recorded at 300 K. Bold numbers refer to structures shown in Fig. 1. NMe cis and NMe trans ,N-methyl groups of 2-E;NMe,N-methylgroupofthe isomers of 3; NAm CH 3 and NAc CH 3 ,C-methyl group of the acetimidoylamino and acetamido substituents, respectively. Fig. 5. Detail of the STD NMR spectrum of the middle-chain PS NH 4 OH/HF from wild-type RC1 in the presence of mAb 2625 showing the resonance region of the C-linked methyl groups. Signals of 1 are probably d ue to subtraction artefacts of the originally most intense proton signals of the N-acetimidoyl and the N-acetyl groups, respect- ively. The sign al marke d by ´ co uld not be assigned to any proton. Bold numbers refer to structures sh own in Fig. 1. F or abb reviatio ns see legend to Fig. 4. 578 O. Kooistra et al. (Eur. J. Biochem. 269) Ó FEBS 2002 it was the aim to investigate the native PS molecule, i.e. with the complete O-acetylated core heptasaccharide, and also to measure ligands with a considerably high molecular mass (11±17 kDa). The molecular mass of the ligand is a sensitive factor in STD NMR spectroscopy, becau se the higher the mass of the ligand, the slower its motion, and the more effective i s spin diffusion. STD spectra with temperature variations were recorded to investigate if the method is applicable to epitopes such as 2, which are subjected to chemical exchange (see above). In the STD spectrum of the sample containing mAb 2625 together with the l ong-chain PS from strain RC1 with an average molecular mass of 14 kDa, signi®cant saturation transfer f rom the protein to the liga n d p rotons a t 2 93 K c ould be detected mainly for t he sign als o f the N- methyl groups in 2 (Fig. 7B) as was observed with PS NH 4 OH/HF (see above). The STD spectrum of the long-chain PS from strain RC1 containing no protein (Fig. 7C) was performed as reference experiment and showed that direct irradiation of ligand resonances could not be avoided under these experimental conditions, d espite a r elaxation delay of 4.3 s and a satu ration time of 3 s. Nevertheless, saturation transfer to the N-methyl groups was not observed under these conditions. A satura- tion transfer was observed for the polysaccharide. The experiment also shows that the molecular mass of a ligand in STD NMR experiments may well exceed a few kDa. Interestingly, saturation transfer could also be detected under conditions, where the two signals o f the N-methyl groups in 2 were coalesced, i.e. at elevated temperatures (315 K; Fig. 8D). Although there is probably no chemical exchange in the bound state, only the single broad proton signal arising from chemical exchange in the free state was observed. DISCUSSION Structural studies aiming at an exact description of the epitope of monoclonal antibodies are time-intensive and laborious. F or example, the epitopes of two anti-L. pneu- mophila LPS antibodies have been described by a series of Fig. 6. Dependence of proton signals of the N-methyl groups in 2 and 3 from pH and tem- perature. 1D 1 H NMR spectra of the long- chain PS NH 4 OH/HF from wild-type RC1 were recorded in 10% deuterated water at constant temperature (275 K) with p 1 H/ 2 H values of  2,  7,  8,  9, and  11 (A±F), and with constant pH (p 1 H/ 2 H7.5)attemperatures between 323 and 283 K raised in 10-K inter- vals (G±M), respectively. Only the resonance region of the N-methyl groups (2.8± 3.4 p.p.m.) is shown. Ó FEBS 2002 Epitope mapping with mAb 2625 (Eur. J. Biochem. 269) 579 extensive e xperiments; the epitope of mAb 3/1 is associated with quantitative 8-O-acetylation of polylegionaminic acid [9,21,22] and mAb LPS-1 recognizes the highly O-acetylated region interve ning t he core oligosaccharide and the OPS of Sg 1 strains [9,13,23]. Investigations of crystal structures of monoclonal anti- bodies in complex with carbohydrate antigens have shown that a small antigenic determinant can dictate a highly speci®c immune response [24]. The OPS of the LPS from the two Vibrio cholerae serotypes Inaba and Ogawa is a homo- polymer o f a-(1 ® 2)-linked N-(3-deoxy- L -glycero-tetronyl)- D -perosamine [25,26] differing only by the presence of a single residue of 2-O-methy l-N-(3-deoxy- L -glycero-tetronyl)- D -perosamine as nonreducing terminal unit in the OPS of serotype Ogawa [27,28]. The crystal structure of a F ab fragment from mAb S-20-4 in complex with synthetic OPS fragments as antigen showed that the t erminal 2-O-met hyl- N-(3-deoxy- L -glycero-tetronyl)- D -perosamine r esidue is the primary antigenic determinant [24]. STD NMR spectroscopy [16] offers an ef®cient alterna- tive approach to identify the residues or substructures involved in binding to monoclonal antibodies or other receptor proteins. A p rerequisite for S TD NMR s pectro- scopy is that the ligand is reversibly bound to the protein. Fig. 7. 1D 1 H NMR (A) and STD NMR (B and C) spectra of long-chain PS from strain RC1inthepresence(AandB)andabsence(C) of mAb 2625. Rather strong unspeci®c irradi- ation of ligand signals in the spectrum in (C) is observed despite the absence o f protein, but not of the signals of the N-methyl groups as in the spectrum in (B) reco rd ed in the pres ence of mAb 2625. Spectra were recorded at 300 K. Bold numbers refer to structures shown in Fig. 1. For abbreviations see legend to Fig. 4. Fig. 8. 1D 1 HNMR(AandC)andSTD NMR (B and D) spectra of long-chain PS from strain RC1 in the p resence of mAb 2625 recorded with p 2 H 7.4 at 293 K (A and B) and 315 K (C and D). Only the resonance region of the N-methyl groups ( 2.8±3.4 p.p.m.) is shown. 580 O. Kooistra et al. (Eur. J. Biochem. 269) Ó FEBS 2002 The binding af®nity (K d ) should b e in the range of 1 m M to  10 n M . Stronger binding often suffers from off-rates being too low. This usually prevents suf®cient amounts of saturated ligand that can be detected in the unbound state. The bound ligand cannot be detected because line widths are far too large for a complex of this size. The binding af®nity can be measured by surface-plasmon-resonance biomolecular interaction analyses [29]. The 5-N-(N,N-dimethylacetimidoyl)-7-N-acetyl (2)and 5-N-acetimidoyl-5-N-methyl-7-N-acetyl (3) derivatives of legionaminic acid were identi®ed as being responsible for phase variation of the epitope of the LPS-speci®c mAb 2625 [15]. In order to determine t he binding af®nity of isolated PS NH 4 OH/HF molecular s pecies of different size, SPR ana- lyses were performed with immobilized mAb 2625. It could be shown that wild-type but not mutant middle- and long- chain PS NH 4 OH/HF bound with signi®can t af®nity, which proved the epitope still to be present in the degraded PS NH 4 OH/HF . The binding af®nity was low, in the range of approximately 30 l M , which could also be seen from the rapid association and dissociation to and from the antibody, typically observed for low-af®nity interaction [19]. Mixtures of PS NH 4 OH/HF molecular species of different size containing different ratios of the N-methylated legionaminic acid derivates bound with similar af®nities. Nevertheless, the observed low af®nity allowed to perform STD NMR spectroscopy with the aim of more precisely describing its epitope. STD spectra unequivocally demonstrated both types of N-methyl groups (2, 3-E and 3-Z) of one single legionaminic acid derivative in the polymer to be involved in binding. Only appropriate material from wild-type RC1 interacted with mAb 2625 and material from mutant 5215 was not interacting with mAb 2625. Although middle-chain PS NH 4 OH/HF from wild-type RC1 that was used for STD experiments was a heterogeneous mixture with respect to chain-length, number of Rha residues at the reducing end, and the content of different derivatives of N-methylated legionaminic acid, the method could be used to show a preference for binding of mAb 2625 to only these N-methylated legionaminic acid derivatives in the polymer. Moreover, i t could be shown that not only the N-methyl groups of the respective N-acetimidoyl groups but also other groups in close proximity were involved in binding. The signal of the C-methyl group of the same N-acetimidoyl groups was observed as was the signal of the C-methyl group of the N-acetyl group linked to C7 in the side chain of the respective legionaminic acid derivative. The N-methyl groups in 2 showed a twofold more effective saturation transfer compared to those in 3-E and 3-Z. If the explanation for more intensive saturation transfer is a larger fraction bound due to stronger binding (i.e. higher af®nity) or shorter distances between protons of ligand and protein cannot be distinguished at the moment. However, b oth cases would suggest 5-N-(N,N-dimethylace- timidoyl)-7-N-acetyllegionaminic acid (2) to be a preferred epitope of mAb 2625. The lower saturation transfer to the N-methyl group in 3-E or 3-Z on the other hand, might be explained by a lower off-rate of the ligand resulting in a lower amount of free but saturated ligand, which would i n turn indicate a higher af®nity. The question of whether the N-methylated legionaminic acid derivative binds with different af®nity to mAb 2625 can only be answered by experiments with homogeneous and structurally de®ned ligands, which to date are not available. However, as the three different N-methylated acetimidoylamino groups (i.e. in 2, 3-E,and3-Z) share great structural similarities it is still possible that they bind with equal af®nity to mAb 2625 (data not shown). The proton signals of the N-methyl groups of 5-N- (N,N-dimethylacetimidoyl)-7-N-acetyllegionaminic acid (2) showed a temperature- and pH-dependent behaviour typical for a rotatio n process a t a partial double bond [20]. The proton sign als o f t he N -methyl group of the 5 -N-acetimidoyl- 5-N-methyl-7-N-acetyllegionaminic acid (3) d id not show such a behaviour, although a partial double bond character was observed; no chemical exchange could be observed under the conditions applied. From these data it is conclud ed that under certain conditions it is possible to observe the interconversion of the cis and trans N-methyl groups in 2 at the partial double bond between the dimethylated n itrogen and the nonprotonated carbon, i.e. the chemical exchange process. Rotation is fast on the NMR timescale o r, more precisely, more frequent [20] and, thus, only one proton signal for both N-methyl groups with the average chemical shift can be observed. The reason could either be deproto- nation of the group at high pH,which destabilize s thedouble- bonded transition s tate, or due to a lowered activation energy of the r otational barrier at high temperature [20], or a combination of both. In contrast, the isomers 3-E and 3-Z were not observed to undergo chemical exchange. This can probably be ascribed mainly to steric hindrance of the bulky group of the polymer -linked derivatives of legionaminic a cid, so that chemical exchange is slow on the NMR timescale, thus indicating that it rarely occurs. Nevertheless, at the moment both isomers o f 3 were present in approximately equimolar ratio. However, at low pH, protonation of both nitrogens of the acetimidoyl(N-methyl)amino group could be the reason for preponderance of one of the isomers (3-E ). The observed chemical exchange for 2, but not for 3,was con®rmed by 2D EXSY experiments [30] at 300 K and 343 K of samples at neutral pH, where o nly cross-correlations for the proton signals of the exchanging N-methyl groups in 2 could be detected (data not shown). Interestingly, recording of STD spectra under conditions were the (N,N-dimethylacetimidoyl)amino group is under- going chemical exc hange, i.e. a t elevated tempe ratures was also possible. Although there is probably no chemical exchange in the bound state, only the single (coalesced) proton signal arising from chemical e xchange in t he free state is observed. Despite the high average molecular mass of the ligand (11±17 kDa) and the epitope being just a minor modi®cation of the OPS, a suf®cient magnetization transfer was observed, showing that in speci®c cases the molecular mass limit of the ligand for STD NMR spectroscopy can be extended. This is the ® rst d escription of an application o f STD NMR spectroscopy to identify the LPS epitope of a monoclonal antibody showing the advantages of this direct approach for the purpose of r elatively quick and direct epitope determi- nation with relatively small amounts of protein and ligands, which do not need to be puri®ed to absolute homogeneity. ACKNOWLEDGEMENTS We thank Dr C. Roll for help with temperature dependence NMR spectroscopy expe riments, and Dr T. W eimar for help with SPR Ó FEBS 2002 Epitope mapping with mAb 2625 (Eur. J. Biochem. 269) 581 analysis. T his work was ®nancially s upported by grants from the Deutsche Forschungsgemeinschaft, LU 514/2-2 (E. L. and M. F.) and ZA 149/3-2 (U. Z.). REFERENCES 1. Winn, W.C. Jr (1988) Legionnaires' disease: historical perspective. Clin. Microbiol. Rev. 1, 60±81. 2. Fields, B.S. (1996) The molecular ecology of legionellae. Trends Microbiol. 4, 286±290. 3. Shuman, H.A. & Horwitz, M.A. (1996) Legionella pneumophila invasion of mononuc lear phagocytes. Curr. Top. Microbiol. Immunol. 209, 99±112. 4. Ciesielski, C.A., Blaser, M.J. & Wang, W.L. (1986) Serogroup speci®city of Legionella pneumophila is related to lipopolysacchar- ide characteristics. Infect. Immun. 51, 397±404. 5. Helbig,J.H.,Kurtz,J.B.,Pastoris,M.C.,Pelaz,C.&Lu È ck, P.C. (1997) Antigenic lipopolysaccharide components of Legionella pneumophila recognized by monoclonal antibodies: possibilities and limitations for division of the species into serogroups. J. Clin. Microbiol. 35, 2841±2845. 6. Knirel, Y.A., Rietschel, E.T., Marre, R. & Za È hringer, U. (1994) The structu re o f the O-speci®c chain of Legionella pneumophila serogroup 1 lipopolysaccharide. Eur. J. Biochem. 221, 239±245. 7. Tsvetkov, Y.E., Shashkov, A.S., Knirel, Y.A. & Za È hringer, U. (2001) Synthesis and identi®cation i n bacterial lipopolysacchar- ides of 5,7-diacetamido-3,5,7,9-tetradeoxy- D -gly cero- D -galacto and - D -glycero- D -talo-nonulo sonic acids. Carbohydr. Res. 331, 233±237. 8. Joly, J.R., McKinney, R.M., Tobin, J.O., Bibb, W.F., Watkins, I.D. & Ramsay, D. (1986) Development of a standardized sub- grouping scheme for Legionella pneumophila serogroup 1 using monoclonal antibodies. J. Clin. Microbiol. 23, 768±771. 9. Kooistra , O., Lu È neberg, E., Lindner, B., Knirel, Y.A., Frosch, M. &Za È hringer, U. (2001) Complex O-acetylation in Legionella pneumophila serogroup 1 lipopolysaccharide. Evidence for t wo genes involved in 8-O-acetylation of legionaminic acid. Biochem- istry 40, 7630±7640. 10. Knirel, Y.A., Moll, H. & Za È hringer, U. (1996) Structural study of ahighlyO-acetylated core of Legionella pneumophila serogroup 1 lipopolysaccharide. Carbohydr. Res. 293, 223±234. 11. Moll, H., Knirel, Y.A., Helbig, J.H. & Za È hringer, U. (199 7) Identi®catio n of an a- D -Manp-(1 ® 8)-K do disacch aride in the inner core region and the structure of the complete core region of the Legionella pneumophila s erogr oup 1 lipopolysac charide . Car- bohydr. Res. 304, 91±95. 12. Za È hringer, U., Knirel, Y.A., L indner, B., Helbig, J.H., Sonesson, A., Marre, R. & Rietschel, E.T. (1995) The lipopolysaccharide of Legionella pneumophila serogroup 1 (strain Philadelphia 1): chemical structure and b iological signi®can ce. Prog. Clin. Biol. Res. 392, 113±139. 13. Lu È neberg, E., Za È hringer,U.,Knirel,Y.A.,Steinmann,D., Hartmann,M.,Steinmetz,I.,Rohde,M.,Kohl,J.&Frosch,M. (1998) Phase-variable expression of lipopolysaccharide contributes to the virulence of Legionella pneumophila. J. Exp. Med. 188, 49±60. 14. Lu È neberg, E., Mayer, B., Daryab, N., Kooistra, O., Za È hringer, U., Rohde,M.,Swanson,J.&Frosch,M.(2001)Chromosomal insertion a nd excision of a 30 k b instable genetic element is responsible for phase variation o f lipopolysaccharide and other virulence determinants in Legionella pneumophila. Mol. Microbiol. 39, 1259±1271. 15. Kooist ra , O., L u È neberg, E., Knirel, Y.A., Frosch, M. & Za È hrin- ger, U. (2001) N-Methylation in polylegionaminic acid is associ- ated with the phase-variable epitope of Legionella pneumophila serogroup 1 lipopolysaccharide. Identi®cation of 5-(N,N-dime- thylacetimidoyl) amino- and 5-acetimidoyl (N-methyl) amino- 7-acetamido-3,5,7,9-tetradeoxynon-2-ulosonic acid in the O-chain polysaccharide. Eur. J. Biochem. 269, 560±572. 16. Mayer, M. & Meyer, B. (1999) Characterization of ligand binding by saturation transfer dierence NMR spectroscopy. Angew. Chem. International Ed. 38 , 1784±1788. 17.Moll,H.,Sonesson,A.,Jantzen,E.,Marre,R.&Za È hringer, U. (1992) Identi®cation of 27-oxo-octacosanoic acid and hepta- cosane-1,27-dioic acid in Legionella pneumophila. FEMS Micro- biol. Lett. 92, 1±6. 18. Helander, I.M. & Kitunen, V. (1989) Cleavage of the O-antigen 4, 5, 12 of Salmonella typhimurium by hydro¯uoric acid. FEBS Lett. 250, 565±569. 19. Ohlson, S., Strandh, M. & Nilshans, H . (1997) Dete ction and characterization of we ak anity antibody an tigen recognition with biomolecular interaction analysis. J. Mol. Recognit. 10, 135±138. 20. Kessler, H. (1970) Nachweis gehinderter Rotationen und Inver- sionen durch NMR-Spektroskopie. Angew. Chem. 82, 237±253. 21. Helbig, J.H., Lu È ck, P.C., Knirel, Y.A., Witzleb, W. & Za È hringer, U. (1995) Molecular characterization of a virulence-associated epitope on the lipopolysac charid e o f Legionella pneumophila serogroup 1. Epidemiol. Infect. 115, 71±78. 22. Zou, C.H., Knirel, Y.A., Helbig, J.H., Za È hringer, U. & Mintz, C.S. (1999) Molecular cloning and characterization of a locus responsible for O-acetylatio n of the O-polysaccharide of Legionella pneumophila serogroup 1 lipopolysaccharide. J. Bacteriol. 181, 4137±4141. 23. Sethi, K.K. & Brandis, H. (1983) Establishment of hybridoma cell lines secreting an ti-Legionella pneumophila serogroup 1 m ono- clonal antibodies with immunodiagnostic potential. Zentralbl. Bakteriol. Mikrobiol. Hyg. [A] 255, 294±298. 24. Villeneuve, S., Soucho n, H., Rio ttot, M.M., M azie  , J.C., Lei, P., Glaudemans, C.P., Kova  c, P., Fournier, J.M. & Alzari, P.M. (2000) Crystal structure of an anti-carbohydrate antibody directed against Vibrio cholerae O1 in complex with antigen: Molecular basis for serotype speci®city. Proc. Natl. Acad. Sci. USA 97, 8433± 8438. 25. Redmond, J.W. (1979) The structure of the O-antigenic side chain of the lipopolysaccharide of Vibrio cholerae 569B (Inaba). Bio- chim. Biophys. Acta 584, 346±352. 26. Kenne, L., Lin dberg, B., Unger, P., Gustafsson, B. & Holme, T. (1982) Structural studies of the Vibrio cholerae O-antigen. Car- bohydr. Res. 100, 341±349. 27. Hisatsune, K., Kondo, S., Isshiki, Y., Iguchi, T. & Haishima, Y. (1993) Occurrence of 2-O-methyl-N-(3-deoxy- L -glycero-tetronyl)- D -perosamine (4-amino-4,6-dideoxy- D -manno-pyranose) in lipo- polysaccharide from Ogawa but not from In aba O forms of O1 Vibrio cholerae. Biochem. Biophys. Res. Commun. 190, 302±307. 28. Ito, T., Higuchi, T., Hirobe, M., Hiramatsu, K. & Yokota, T. (1994) Identi®cation o f a novel sugar, 4-amino-4,6-dideoxy-2- O-methylmannose in the lipopolysaccharide of Vibrio cholerae O1 serotype Ogawa. Carbohydr. Res. 256, 113±128. 29. Jo È nsson, U., Fa È gerstam, L., Ivarsson, B., Johnsson, B.,Karlsson , R., Lundh, K., Lo È fa Ê s, S., Persson, B., Roos, H., Ro È nnberg, I., Sjo È lander, S ., Stenberg, E ., Sta Ê hlberg, R ., Urbaniczky, S ., O È stlin, H . & Malmqvist, M. (1991) Real±time biospeci®c interaction analysis using surface plasmon resonance and a sensor chip techn ology. Biotechniques 11, 620±627. 30. Perrin, C.L. & Dwyer, T.J. (1990) Application of two-dimensional NMR to kinetics of chemical exchange. Chem. Rev. 90, 935±967. SUPPLEMENTARY MATERIAL The following material is available from http://www. blackwell-science.com/products/journals/suppmat/ejb/ ejb2684/ejb2684sm.htm 582 O. Kooistra et al. (Eur. J. Biochem. 269) Ó FEBS 2002 . Epitope mapping of the O-chain polysaccharide of Legionella pneumophila serogroup 1 lipopolysaccharide by saturation-transfer-difference NMR spectroscopy Oliver. acid (5-N-acetimidoyl-7-N-acetyllegionaminic acid) in the O-chain polysaccharide (OPS) of the Legionella pneumophila serogroup 1 lipopolysaccharide (LPS) concern N-methyla- tion of the 5-N-acetimidoyl