S. Müller-Loennies, Borstel Research Center, Parkallee 22, 23845 Borstel, Germany. Fax: + 49 4537 188 419, Tel.: + 49 4537 188 467, E-mail: firstname.lastname@example.org
Lipopolysaccharide (LPS) of Escherichia coli strain 2513 (R4 core-type) yielded after alkaline deacylation one major oligosaccharide by high-performance anion-exchange chromatography (HPAEC) which had a molecular mass of 2486.59 Da as determined by electrospray ionization mass spectrometry. This was in accordance with the calculated molecular mass of a tetraphosphorylated dodecasaccharide of the composition shown below. NMR-analyses identified the chemical structure as
where l-α-d-Hep is l-glycero-α-d-manno-heptopyranose and Kdo is 3-deoxy-α-d-manno-oct-2-ulopyranosylonic acid and all hexoses are present as d-pyranoses.
We have also isolated the complete core-oligosaccharides of E. coli F653 LPS for which only preliminary data were available and investigated the deacylated LPS by NMR and MS. The proposed structure determined previously by methylation analysis was confirmed and is shown below.
In addition we have quantified the side-chain heptose substitution of the inner core with GlcpN (≈ 30%) and confirmed that this sugar is only present when the phosphate at the second l,d-Hepp residue is absent.
Distortionless enhancement by polarization transfer
high-performance anion exchange chromatography
Lipopolysaccharide (LPS) is the major component of the outer leaflet of the outer membrane of Gram-negative bacteria . LPS of enterobacteria consist of three domains, namely lipid A, core-region and O-antigen . Due to its exposed location, it is the major target of the humoral immune response in mammals and the lipid A moiety is responsible for many of the pathological effects seen in septic shock patients. Whereas the chemical structure of the O-antigen is highly variable, the core-region and lipid A show only a limited structural variability within the same species. This prompted many investigators to attempt the isolation of antibodies directed against the conserved regions of LPS, i.e. lipid A and core-region (reviewed in ). It was assumed that these antibodies would be both cross-reactive and cross-protective against different Gram-negative pathogens. Whereas a cross-protective effect was described for a polyclonal antiserum as early as in 1966 , all subsequently isolated monoclonal antibodies failed to show cross-reactivity in vitro and cross-protectivity in vivo, except one reported by DiPadova et al. (mAb WN1 222-5). This mAb recognized LPS from all tested clinical isolates of E. coli, Salmonella, and Shigella in Western-blots and showed cross-protective effects in vivo against endotoxic activities of LPS . The cross-reactivity was attributed to a common epitope located in the inner core-region of these LPS. In order to verify this assumption, we have determined the as yet unknown chemical structures of those LPS that reacted with this mAb.
The chemical structures of four E. coli core-oligosaccharides (R1, R2, R3, and K-12) and two core-oligosaccharides of S. enterica are known. The chemical structure of the E. coli R3 core-type was determined by methylation analysis [6,7]. Complete core-oligosaccharides were isolated and NMR chemical shift data were determined for core-oligosaccharides of the R1 and R2 core-types  whereas the chemical structure of the inner core-region of the E. coli R4 core-type was hitherto unknown. The chemical structure of the outer core region of the latter was determined by methylation analysis . We have now isolated the complete core-oligosaccharides and investigated the chemical structure of this LPS in detail to understand the cross-reactivity of WN1 222-5. Since these data are a prerequisite for NMR based conformational analysis of the inner core region of enterobacterial LPS and epitope mapping of WN1 222-5 we have in addition isolated the complete core-oligosaccharides of E. coli F653 (R3-core) and determined NMR chemical shift values.
Materials and methods
Bacteria and bacterial LPS
E. coli strains 2513 and F653 were cultivated and used for the isolation of LPS by phenol/chloroform/petrolether-extraction as reported .
Neutral sugars, GlcN, Kdo and bound organic phosphate were determined as described .
Preparation of deacylated LPS of E. coli 2513
LPS (5 g) was de-O-acylated by mild hydrazinolysis  (yield 3.84 g) and 400 mg of the latter were subjected to alkaline de-N-acylation as described . After neutralization by addition of ion exchanger Amberlite IRA120 H+ (Serva), 160 mg of the deacylated oligosaccharide fraction (yield 217 mg) was subjected to high-performance anion-exchange chromatography (HPAEC; eight runs of 20 mg each) using a semipreparative CarboPak PA100 column (9 × 250 mm) and a DX300 chromatography system (Dionex, Germany). The main (fraction 2; oligosaccharide 1, yield 31.44 mg) and the minor oligosaccharide (fraction 1; oligosaccharide 2, yield 10.96 mg) were collected, neutralized and desalted as described above by addition of ion-exchanger followed by lyophilization. Conditions for semipreparative and analytical HPAEC were as described previously .
Preparation of deacylated LPS of E. coli F653
LPS (2.11 g) was de-O-acylated by mild hydrazinolysis (yield 1.425 g) and 902.5 mg were further subjected to alkaline de-N-acylation as above. The solution was neutralized by addition of 8 m HCl and extracted three times with dichloromethane. Subsequent desalting was achieved by gelfiltration on Sephadex G10 (2.5 × 65 cm) in 10 mm ammonium carbonate (yield 420 mg). A portion (417 mg) of the desalted oligosaccharide mixture was subjected to semipreparative HPAEC as described above. The sample was redissolved in water at a concentration of 90 mg·mL−1 and 450 µL per run loaded onto the HPAEC column. Elution and separation was achieved by a linear gradient of 2–600 mm NaOAc over a time of 70 min. Fractions were analyzed by analytical HPAEC and appropriately combined. Desalting was achieved by gelfiltration as described above. Two pure oligosaccharides were obtained (fraction 1; oligosaccharide 3, 145.22 mg; fraction 2; oligosaccharide 4, 70.7 mg).
NMR-spectra were recorded of samples of deacylated LPS (11 mg each of R4 oligosaccharides 1 and 2 and 10 mg each of R3 oligosaccharides 3 and 4) in 0.5 mL solutions in D2O using a Bruker DRX 600 Avance spectrometer equipped with a multinuclear probehead with z-gradient. Acetone served as a reference 2.225 p.p.m. (1H) and 31.5 p.p.m. (13C). All spectra were run at a temperature of 300 K.
NMR of oligosaccharide 1 (R4 core). Two-dimensional homonuclear 1H,1H-COSY was performed with a double quantum filter and time-proportional phase incrementation (TPPI) (DQF-COSY). The Bruker cosydftp pulseprogram was modified to allow water suppression with 10 Gaussian shaped pulses of 100 ms defined by 1024 points during the relaxation delay. Five-hundred and twelve experiments of 4096 data points each were recorded over a spectral width of 6.5 p.p.m. in each dimension. Prior to Fourier transformation F1 was zero-filled to 1024 data points.
TOCSY was performed at a spinlock field strength of 8 kHz for 75.15 ms using the Bruker mlevprtp pulse program and the same experimental parameters that were used for TOCSY-ROESY (TORO).
A TORO-spectrum [14–17] was recorded as a two-dimensional experiment using a fixed delay as the second mixing time (ROESY-step). The spectrum was recorded phase-sensitive by applying TPPI. Four-thousand and ninety-six data points were recorded over 512 experiments consisting of 40 scans each over a spectral width of 8 p.p.m. in each dimension. Water suppression was achieved by presaturation on resonance during the relaxation delay. Prior to FT the data were multiplied by a shifted sine bell function and zero-filled in F1, 1024 data points.
NOESY was recorded phase-sensitive using the Bruker noesyprtp pulse program. Four-thousand and ninety-six data points in F2 and 512 experiments in F1 were recorded over a spectral width of 10 p.p.m. in both dimensions. Prior to Fourier transformation, the FID was multiplied with a shifted sine bell window function and zero-filling was applied in F1, 1024 points. The mixing time was 200 ms.
For heteronuclear 1H,13C-NMR correlation spectroscopy the Bruker standard pulse programs inv4prst (HMQC), inv4ndtp (HMQC without decoupling during acquisition), inv4mlprtp (HMQC-TOCSY), indecobimltppr (DEPT-HMQC-TOCSY), and inv4lrndpr (HMBC) were used. These spectra were recorded with 4096 data points in F2 and 512 experiments in F1 over spectral widths of 10 and 120 p.p.m, respectively. Zero-filling was applied to 1024 data points in F1. For TOCSY a spinlock period of 81 ms was applied at a field strength of 8.3 kHz. For DEPT-HMQC-TOCSY the sweep width was reduced to 15 p.p.m. in F1 and 3.5 p.p.m. in F2. Two-hundred and fifty-six experiments were recorded at 2048 data points per increment and a TOCSY mixing time of 67 ms. For HMBC, F1 was enlarged to 180 p.p.m. and the delay for the evolution of long-range couplings was set to 50 ms.
31P spectroscopy was performed after addition of NaOD (Sigma) until all signals appeared as sharp singuletts. The pD was then approximated using pH paper and found to be pD 9.
31P,1H-HMQC was performed using a modified Bruker pulse program (inviprtp) which was using continuous wave instead of composite pulse decoupling during acquisition. The spectrum consisted of 128 experiments of 2048 data points covering 10 p.p.m. in both dimensions. The delay for evolution of couplings was adjusted for a 3JP,H of 10 Hz (d4 = 25 ms).
NMR of oligosaccharides 3 and 4 (R3 core). DQF-COSY and NOESY were recorded as described above. In COSY the spectral width was reduced to 5.5 p.p.m. in each dimension. TOCSY was performed using the DIPSI-2 composite pulse for spin-lock at a field strength of 7.8 kHz and a spectral width of 5.5 p.p.m. in each dimension. Five-hundred and twelve experiments were recorded consisting of 4096 data points each. Water presaturation was achieved by a shaped pulse as described for DQF-COSY.
ROESY-TOCSY (ROTO) was performed as TORO (see above) as a two-dimensional experiment but using a fixed delay as the TOCSY mixing time. The spectrum was recorded phase-sensitive by applying TPPI. Two-thousand and forty-eight data points were recorded of 256 experiments consisting of 32 scans each over a spectral width of 10 p.p.m. in each dimension. Water suppression was achieved by presaturation on resonance during the relaxation delay. Prior to Fourier transformation, the data were multiplied by a shifted sine bell function and zero-filled in F1 and F2, 4096 and 512 data points, respectively. The spin-lock for ROE of 5.6 kHz was applied for 150 ms and TOCSY was performed with a mixing time of 77 ms. The field strength of the TOCSY spin-lock in this experiment was 9.4 kHz.
HMQC, HMBC and HMQC-TOCSY were recorded as z-gradient experiments using standard Bruker software. The experimental setup was otherwise identical to the same spectra recorded of oligosaccharide 1. DEPT-HMQC-TOCSY was run as described above but the spectral width was reduced to 30 p.p.m. in F1 and 5 p.p.m. in F2. One-hundred and twenty-eight experiments of 64 scans with 2048 data points were recorded. The TOCSY spinlock was applied for 90 ms.
Mass spectra were recorded in the negative ion mode of the mixture of oligosaccharides prior to HPAEC, of the isolated main oligosaccharide of deacylated LPS and of acylated purified LPS from E. coli F2513 (R4 core-type). In addition, the deacylated minor core-oligosaccharide of E. coli F653 (R3 core-type) was analyzed. Negative ion electrospray ionization mass spectra were recorded on a Fourier Transform Ion Cyclotron Resonance FT-ICR mass spectrometer (APEX II, Bruker Daltonics, Billerica, USA) equipped with a 7 Tesla actively shielded magnet and an Apollo ion source. Samples were dissolved at a concentration of ≈ 10 ng·µL−1 in a 50 : 50 : 0.001 (v/v/v) mixture of 2-propanol, water, and triethylamine and sprayed at a flow rate of 2 µL·min−1. For straightforward interpretation the spectra were charge-deconvoluted.
Structural analysis of E. coli 2513 core-oligosaccharide (R4 core)
Compositional analysis of LPS identified Gal, Glc, GlcN, Kdo, l,d-Hep, and -P in a molar ratio given in Table 1. In addition 3OH-C14:0, C12, and C14 were found in accordance with the common acylation pattern of E. coli lipid A .
Table 1. Composition of E. coli R4 LPS. KdoAcP, Kdo determination after hydrolysis in acetate buffer pH 4.5; KdoHCl, Kdo determination after hydrolysis in 0.1 m HCl.
nmol component per mg
a Relative to GlcN = 2.0.
C12 : 0
C14 : 0
3OH-C14 : 0
Analytical HPAEC revealed that the deacylated LPS fraction contained one major oligosaccharide isolated by semipreparative HPAEC. The charge deconvoluted negative ion-mode ESI-FT-ICR mass spectrum of the major oligosaccharide fraction (Fig. 1) obtained by deacylation of LPS revealed a prominent ion with a mass of 2486.59 m/z. This was indicative of a composition of two HexN-, two Kdo-, three heptose-, five hexose-residues, and four phosphates (theoretical mass [M-H]– 2485.577). Thus, this oligosaccharide was a dodecasaccharide carrying four phosphate substituents, in agreement with the compositional analysis. Further sodium and potassium ion-clusters were observed that indicated the loss of one phosphate group (m/z 80 lower mass). A 31P,1H-COSY NMR-spectrum revealed that the phosphate group at the 4′ position of lipid A was missing (data not shown).
In addition ESI-FT-ICR contained the signals of a molecular ion with m/z 2566.56, which was 80 Da higher than the main fraction and was indicative of an additional phosphate residue. Due to its low abundance this fraction could not be isolated and subjected to a more detailed analysis. Further molecular ions with a mass of 18 m/z lower were observed which were not present in the same spectrum of purified LPS prior to deacylation (not shown).
NMR-spectroscopy of the main oligosaccharide confirmed the compositional and mass spectrometrical analyses. The 1H-NMR spectrum (Fig. 2) contained 10 signals of anomeric protons (Table 2). In addition, two pairs of signals originating from 3-deoxy protons of Kdo-residues were present. Full assignment of proton and carbon chemical shifts and determination of 3JH,H-coupling constant values identified two pyranosidic Kdo-residues. Their α-configuration was evident from the resonance frequencies of the deoxy-protons (equatorial H3 > 2.4 p.p.m. for β-Kdop) and the chemical shift values of the H-5 protons . All sugars were present as pyranoses which was deduced from their C-4 carbon chemical shifts (above 80 p.p.m. for furanoses, Table 3). Correlation signals from anomeric protons to intraresidue C-5 in HMBC corroborated the pyranose-configuration of sugar residues. All other sugars except two were also α-configurated which was determined by the analysis of JC-1,H-1-coupling constants (> 172 Hz) from a HMQC spectrum recorded without decoupling during acquisition. Signals of anomeric carbons at 99.75 p.p.m. and 103.20 p.p.m. were assigned to a β-GlcpN (164 Hz, residue B) and β-Galp (164 Hz, residue M), respectively. Their β-configuration was confirmed by their 3JH,H-coupling constants (≈ 8 Hz) and their intraresidual NOE connectivities between H-1, H-3 and H-5. Three signals of anomeric protons showed 3JH,H-coupling constant values of less than 1 Hz and thus the H-2 in these residues was in equatorial position as in manno-configurated sugars (residues E, F, H). The determination of the spin system and coupling constants revealed that they belonged to l,d-Hepp-residues. The analysis of a CH2-edited DEPT-HMQC spectrum and a DEPT-HMQC-TOCSY spectrum allowed the assignment of H-7a/b and C-7 of l,d-Hepp residues F and H. The chemical shift of H-6 of residue F could also be assigned in the latter spectrum. The chemical shift of the H-6 proton of residue H, however, could not be identified in this spectrum. Analysis of the 1H,13C correlation spectrum indicated its chemical shift at 4.025 p.p.m., in agreement with the chemical shift of the same proton in previously analyzed oligosaccharides from E. coli J-5 . Further residues were identified as α-Glcp (residues G and I), α-Galp (residues K and L), and α-GlcpN (A).
Table 2. . 1H-NMR chemical shifts (p.p.m.) of deacylated LPS of E. coli strain 2513 (R4 core-type, 1) and F653 (R3 core-type, 3 major and 4 minor).
The analysis of an HMBC spectrum showing intraresidual cross-correlation signals from anomeric protons to carbons C-3 and C-5 was important for the assignment of spin-systems and chemical shifts of carbon. Additionally, long-range correlations between protons of adjacent sugar residues across the glycosidic bond established their sequence; this was confirmed by the analysis of a NOESY spectrum and a 2D-TOCSY-ROESY (TORO) [14–17] spectrum (Fig. 3). This latter experiment facilitated the assignment because all protons connected by scalar couplings and part of a spin system detectable by TOCSY show connectivities to protons close in space to any of these protons. Therefore, more correlation signals are observed in the region of anomeric protons that resolves signal overlap, the identification of ROE signals is simplified and corroborated by further correlation signals within the adjacent residue. Furthermore, due to the asymmetry of the experiment with respect to the magnetization transfer mechanism the pulse sequence generates signals in the vertical plane only for protons within the preceeding residue and in the horizontal plane only for protons within an attached residue (most importantly the anomeric proton, if present). The same result is obtained for a ROESY-TOCSY (ROTO) experiment where a mirror image of the TORO spectrum is obtained. For example, the well resolved anomeric proton E1 shows cross-correlation signals not only to E2 but also interresidual connectivities to protons C5, C6, C7, C8a and C8b. The latter cross-correlation signals appeared in opposite phase in the spectrum. The correlation signal between I1 and G1 is only seen in the vertical plane, proving that residue I is attached to residue G. Due to the opposite phases, cancellation or diminished signal intensity may occur and signals from direct ROE effects are still present. For some overlapping signals arising from the two different pathways, mixed phase signals instead of complete cancellation was observed. However, even if some signals may have cancelled out, this possible disadvantage is more than compensated for by the additional information provided. The results of these experiments are summarized in Table 5. Residue B was thus connected to residue A in β1→6 linkage and these represented the lipid A backbone. Characteristic NOEs between protons H-3a (weak) and H-3e (strong) of α-Kdop (residue C) and H-6 of α-Kdop (residue D) confirmed their 2→4-linkage . The heptose-region was composed of three heptose residues of the sequence l-α-d-Hepp-(1→7)-l-α-d-Hepp-(1→3)-l-α-d-Hepp which was connected to the inner Kdo (residue C) in position 5. The residues G, I, K, L, and M were those of the outer core and the NMR analyses confirmed the results obtained by methylation analysis . Long-range NOEs were observed between H-3a (strong) and H-3e (very weak) of α-Kdop (residue C) and H-1 of l,d-Hepp (residue H) and between the anomeric proton of the inner l,d-Hepp residue (E) and the equatorial H-3 of the side-chain α-Kdop (residue D).
Four phosphate residues were identified (Fig. 4A, Table 4) that were shown by HMQC to be linked to protons A1 (α-GlcpN), B4 (β-GlcpN), E4 (l,d-Hepp I) and F4 (l,d-Hepp II). The substitution with phosphate led to significant downfield shifts of protons and carbons at these positions. The additional scalar coupling led to splitting of the corresponding signals in 1H-, and 13C-NMR spectroscopy.
In accordance with all experimental data the chemical structure of deacylated E. coli 2513 LPS (R4 core-type) is as depicted in Fig. 5.
Structural analysis of E. coli F653 core-oligosaccharides (R3 core)
We have subjected purified LPS from E. coli F653 to alkaline deacylation by hot alkali and separated the oligosaccharide mixture obtained by HPAEC. Two main fractions were obtained. The 1H-NMR spectrum of the major oligosaccharide (Fig. 6A) contained 10 signals of anomeric protons and two pairs of signals originating from 3-deoxy protons characteristic of αKdop. The assignment of 1H and 13C-resonances (Tables 2 and 3) and determination of vicinal coupling constants (3JHn, Hn+1) revealed that it was composed of three GlcpN (residues A, B, M, Fig. 8), two Kdop (C, D), three L,d-Hepp (E, F, H), three Glcp (G, K, L) residues and one Galp (I) residue. All sugars except one were present as α-pyranoses as evident from 3JH-1,H-2 and JC-1,H-1 coupling constants and 13C chemical shifts. The chemical shift and the 3JH-1,H-2 coupling constant of the anomeric proton (8.5 Hz) of one GlcpN (residue B) as well as NOE correlation signals between H-1, -3 and -5 confirmed its β-configuration. In addition, four phosphate residues were present (Fig. 4B, Table 4) which were located at positions 1 and 4′ of the lipid A backbone (GlcpN-disaccharide at the reducing end, residues A and B) and at positions 4 of both l,d-Hepp-residues E and F as determined by 31P,1H-COSY, as well as 1H- and 13C-chemical shift analysis. The sequence of residues was determined by NOESY, ROTO, and HMBC. In NOESY the anomeric protons of residues B to G showed NOEs in agreement with the chemical structure present in the R4-core (see above) and in the corresponding partial structure isolated previously from the Rc-mutant strain E. coli J-5 . In addition, interresidual NOEs were observed from I1 to G3, M1 to I3, K1 to I1 and I2, and L1 to K1 and K2 establishing the sequence of these residues as α-d-Glcp1→2-α-d-Glcp1→2-[α-d-GlcpN1→3]-α-d-Galp1→3-α-d-Glcp. These residues thus represented the outer core and confirmed the results previously obtained by methylation analyses [6,7,9,21]. The sequence deduced from observed NOEs was corroborated by cross-correlation signals in HMBC across the glycosidic bonds.
Table 3. 13C-NMR chemical shifts (p.p.m.) of deacylated LPS of E. coli strain 2513 (R4 core, 1) and F653 (R3 core, 3 major and 4 minor).
In comparison to the major oligosaccharide 2, 1H- (Fig. 6B) and 13C-NMR-spectra of the minor core oligosaccharide 3 contained an additional set of signals originating from an α-d-GlcpN-residue which was therefore a tridecasaccharide. NOE NMR-spectra, ROTO and HMBC confirmed that this residue (N, Fig. 7) was located at position 7 of the side chain l,d-Hepp (residue H) leading to downfield shifts of protons F7a and F7b as well as F6. NOE correlations were also observed between proton H1 of this GlcpN-residue (residue N) to protons H-6, H-7a and H-7b of residue H. Simultaneous upfield shifts of proton and carbon 4 of residue F (second l,d-Hepp) indicated that this position was not substituted with phosphate in contrast to the major oligosaccharide. This was corroborated by only three phosphorus resonances in 31P-spectra (Fig. 4C, Table 4). A 1H,31P-NMR correlation spectrum confirmed that apart from two phosphates in the lipid A (positions 1 and 4′) only position 4 of residue E was phosphorylated.
ESI-FT-MS of this oligosaccharide showed prominent ions with a mass of 2566.70 m/z(Fig. 8) and was consistent with the composition deduced from NMR.
LPS of E. coli strain 2513 (R4 core-type) contains the major oligosaccharide shown in Fig. 5. ESI-FT-MS also revealed the presence of a minor fraction devoid of one phosphate residue and NMR-analysis showed that the 4′-phosphate group was missing. A molecular ion with a mass difference of an additional 80 Da indicated the presence of a fifth phosphate residue. This compound, however, was present only in minute amounts and was not seen in HPAEC. It is known that in enterobacterial LPS the first heptose (residue E) is substituted in position 4 with 2-aminoethanol diphosphate instead of a monophosphate . The origin of this oligosaccharide can thus be explained by the possibility that the strong alkaline treatment did not hydrolyze the 2-aminoethanol diphosphodiester between phosphate groups but to a small extent between the 2-aminoethanol and phosphate leading to a diphosphate at this position. A mass spectrum of purified LPS (not shown) recorded to explain the origin of this molecular ion and of additional molecular ions with a lower mass of m/z 18, which were present in the spectra of alkali treated LPS (Fig. 1), only contained molecular ions with a mass difference of m/z 123 (2-aminoethanol monophosphate) but not with additional m/z 80 (phosphate). Therefore an additional monophosphate substitution at a different location may be excluded. Furthermore, the spectrum did not contain molecular ions with a lower mass of Δm/z 18 which thus were artefacts. Similar artefacts have been described previously by Olsthoorn et al. .
There was no heterogeneity with respect to the substitution by additional sugar residues attached to the side chain-heptose as observed in other LPS-core structures (E. coli F470, E. coli F653, Shigella sonnei, Shigella flexneri, Erwinia carotovora FERM P-7576, Proteus mirabilis R110/1959, Citrobacter freundii O23) . Notably, the phosphate substitution at position 4 of the second heptose (residue F) is quantitative and no smaller molecule is present which lacks the side chain heptose (residue H) as found in an Rc-mutant strain of E. coli F653 (E. coli J-5, R3 core-type) .
The NMR analyses of the oligosaccharides derived from E. coli F653 (R3 core) confirmed the results of methylation analyses published earlier [6,7]. The relative amounts of GlcpN in the preparation were approximately 30%. Important from a biosynthetic point of view, we observed a correlation of side-chain heptose substitution with GlcpN and the lack of phosphate substitution at the second heptose of the inner core. This has been observed the first time in the rough mutant strain E. coli J-5 . We have therefore reinvestigated the minor core-oligosaccharide of E. coli F470  which also contains this substitution by 1H,1H-DQF-COSY. The correlation signals of H-4 of the second heptose, which in case of phosphorylation are shifted downfield (≈ 4.5 p.p.m), were not present at this frequency but had moved upfield. Therefore, we concluded that also in this case the phosphate group at this position is missing and this correlation seems to be true in general, at least for all E. coli LPS. Obviously, this phosphate residue must be removed either prior, during or after the transfer of this GlcpN. The enzymatic activity and the corresponding gene locus have not been identified so far. It is tempting to speculate that this GlcpN-transferase may be able to act as a phosphorylase and that both reactions occur simultaneously since no molecules are detected which contain both, GlcpN and phosphate, or no GlcpN and no phosphate at this position. The presence of phosphate at this position in oligosaccharides, which possess a side-chain heptose residue is explained by the fact that the activity of heptosytransferase III (WaaQ) is dependent on the phosphate at this position . Thus, it may be that the reaction of this GlcpN-transferase is energetically driven by the removal of the phosphate. It may also be that two different enzymatic activities are active and the lack of the side-chain GlcpN substitution in the R4 core oligosaccharide can be either due to the lack of the responsible GlcpN-transferase or by the lack of a phosphatase if the phosphate has to be removed prior to glycosylation.
The reactivity of E. coli 2513 LPS with mAb WN1 222-5 can be explained by an inner-core structure identical to those found in LPS from all core-types of E. coli and Salmonella. Investigation of the inter-residual NOE connectivities revealed that no long range NOEs indicative of a backfolding of the outer core were found. Therefore, it is apparent that the outer core and the inner core form two structural confined domains, which may be a prerequisite for the accessibility of the inner core sugars for the binding by mAb WN1 222-5.
Previously, an NOE between the anomeric proton of the side-chain heptose (residue H) and the axial H-3 of the first αKdop (residue C) has been observed in the analysis of the R1 and R2 core oligosaccharides  and a computational calculation of a partial oligosaccharide excluding phosphate substitution performed in this study revealed a distance of 5–6 Å of these protons, inconsistent with reported results  and interpreted as spin diffusion. Also these NOEs were not observed in NOESY spectra of smaller oligosaccharides from the Rc-mutant E. coli strain J-5 . The same long-range interactions between the side-chain l,d-Hepp (residue H) and the inner α-Kdop (C), and between the adjacent l,d-Hepp (residue F) and the side-chain α-Kdop (residue D) were observed in the analysis of the R3 and R4 core oligosaccharides. To provide further evidence we have therefore conducted a series of NOESY experiments with varying mixing times that showed that these were present also at mixing times as short as 50 ms. They may therefore indicate a conformational change of the inner core with respect to molecules that do not have an outer core rather than the result of spin diffusion. In order to arrive at meaningful calculations, extensive NMR measurements will have to be performed and pH and the presence of divalent cations should be taken into account.
With respect to the recognition of these structures by antibodies and the structural characterization of epitopes leading to cross-reactivity, the full assignment of carbon and, in particular, proton resonances now provides the basis for detailed NMR-based conformational analysis of the inner core region of enterobacterial LPS. Furthermore, it will allow the interpretation of saturation transfer difference measurements  aiming at further characterization of the epitope recognized by the cross-reactive mAb WN1 222-5.
We greatfully acknowledge the technical assistance of V. Susott, and Helga Lüthje as well as the kind gift of the minor core oligosaccharide from E. coli rough mutant F470 by Drs O. Holst and E. Vinogradov. This research was financially supported by the Deutsche Forschungsgemeinschaft Grants DFG L1-448 (BL).