Báo cáo khoa học: Structural characterization of a novel branching pattern in the lipopolysaccharide from nontypeable Haemophilus influenzae pot

Fax: + 46 8585 838 20, Tel.: + 46 8585 838 23, E-mail: elke.schweda@kfc.ki.se Abbreviations: CE, capillary electrophoresis; PCho, phosphocholine; PEtn, phosphoethanolamine; PPEtn, pyroph

Structural characterization of a novel branching pattern in the

Martin Ma˚nsson1, Derek W Hood2, E Richard Moxon2and Elke K H Schweda1

1

Clinical Research Centre, Karolinska Institutet and University College of South Stockholm, Novum, Huddinge, Sweden;

2

Molecular Infectious Diseases Group and Department of Paediatrics, Weatherall Institute of Molecular Medicine,

John Radcliffe Hospital, Oxford, UK

Structural analysis of the lipopolysaccharide (LPS) from

nontypeable Haemophilus influenzae strain 981 has been

achieved using NMRspectroscopy and ESI-MS on

O-deacylated LPS and core oligosaccharide (OS) material as

well as by ESI-MSn on permethylated dephosphorylated

OS A heterogeneous glycoform population was identified,

resulting from the variable length of the OS branches

attached to the glucose residue in the common structural

element of H influenzae LPS, L-a-D

-Hepp-(1fi2)-[PEtnfi6]-L-a-D-Hepp-(1fi3)-[b-D-Glcp-(1fi4)]-L-a-D

-Hepp-(1fi5)-[PPEtnfi4]-a-Kdop-(2fi6)-Lipid A Notably, the

O-6 position of the b-D-Glcp residue was either substituted

by PCho or the disaccharide branch b-D

-Galp-(1fi4)-D-a-D-Hepp, while the O-4 position was substituted by the globotetraose unit, b-D-GalpNAc-(1fi3)-a-D -Galp-(1fi4)-b-D-Galp-(1fi4)-b-D-Glcp, or sequentially truncated ver-sions thereof This is the first time a branching sugar residue has been reported in the outer-core region of H influenzae LPS Additionally, a PEtn group was identified at O-3 of the distal heptose residue in the inner-core

Keywords: Haemophilus; lipopolysaccharide; phosphocho-line; structural analysis; ESI-MSn

Haemophilus influenzaeis a Gram-negative pathogen that

routinely colonizes the human upper respiratory tract and

which can be found both in encapsulated (types a–f) and

unencapsulated (nontypeable) forms While the incidence

of disease caused by H influenzae type b (invasive diseases,

including meningitis and pneumonia) has been greatly

reduced in recent years as a result of the development of

conjugate vaccines, there exists no vaccine against

nontype-able H influenzae (NTHi) NTHi strains are a common

cause of otitis media and respiratory tract infections [1] and

its lipopolysaccharide (LPS) molecule has been shown to be

important for colonization and bacterial persistence during

infection H influenzae LPS is composed of a

membrane-anchoring lipid A moiety linked by a single phosphorylated

3-deoxy-D-manno-oct-2-ulosonic acid (Kdo) residue to a

variable core oligosaccharide (OS) portion The carbo-hydrate regions provide targets for recognition by host immune responses and some of these OS epitopes mimic human antigens, suggesting that the bacteria may use these structures to evade the host immune system [2,3] Moreover, the OS portion of H influenzae LPS has a propensity for reversible switching of terminal epitopes (phase variation), which is a major factor in the generation of the vast intrastrain LPS heterogeneity usually found [4] This heterogeneity is thought to be an advantage to the bacteria, allowing them to better confront different host compart-ments and microenvironcompart-ments and to survive the host immune response [5] The availability of the complete genome sequence of H influenzae strain Rd [6] has facili-tated a comprehensive study of LPS biosynthetic loci in the homologous strain RM118 [7] and in the type b strains Eagan (RM153) and RM7004 [8] Gene functions have been identified that are responsible for most of the steps in the biosynthesis of the OS portion of their LPS molecules Molecular structural studies of LPS from H influenzae strains [9–23] have resulted in a structural model consisting

of a conserved phosphoethanolamine (PEtn)-substituted triheptosyl inner-core moiety (labelled HepI–HepIII) in which each of the heptose residues can provide a point for attachment of OS chains or noncarbohydrate substituents (Fig 1) Notably, HepIII and the b-D-Glcp residue that is linked to HepI (labelled GlcI) have an especially wide range

of alternatives in the substitution pattern HepIII has been found to be substituted by a b-D-Glcp residue (either at O-2 [13] or O-3 [16]) or a b-D-Galp residue (either at O-2 [10] or O-3 [19]) Analysis of LPS from lpsA mutants established in

a number of strain backgrounds supports a role for LpsA in each of the alternative glycose substitutions of HepIII HepIII has also been found to be substituted by the

Correspondence to E Schweda, University College of South

Stockholm, Clinical Research Centre, NOVUM, S-141 86 Huddinge,

Sweden Fax: + 46 8585 838 20, Tel.: + 46 8585 838 23,

E-mail: elke.schweda@kfc.ki.se

Abbreviations: CE, capillary electrophoresis; PCho, phosphocholine;

PEtn, phosphoethanolamine; PPEtn, pyrophosphoethanolamine;

Hep, heptose; D , D -Hep, D -glycero- D -manno-heptose; L , D -Hep,

L -glycero- D -manno-heptose; Hex, hexose; HexNAc,

N-acetylhexos-amine; HMBC, heteronuclear multiple-bond correlation; lipid A-OH,

O-deacylated lipid A; Kdo, 3-deoxy- D -manno-oct-2-ulosonic acid;

LPS, lipopolysaccharide; LPS-OH, O-deacylated LPS; MSn, multiple

step tandem mass spectrometry; Neu5Ac, N-acetylneuraminic acid;

NTHi, nontypeable Haemophilus influenzae; OS, oligosaccharide;

AnKdo-ol, reduced anhydro Kdo.

(Received 27 February 2003, revised 13 May 2003,

accepted 16 May 2003)

noncarbohydrate substituents Ac (either at O-2 [16] or O-3

[15]), Gly [16,24], P (at O-4 [11]) and PEtn [9] For GlcI, the

O-4 position has been found to be substituted by a b-D-Galp

residue [9] or a b-D-Glcp residue [10], while in other strains,

O-6 was substituted by phosphocholine (PCho) [13],

D-glycero-D-manno-heptose (D,D-Hep) [14] or L-glycero-D

-manno-heptose (L,D-Hep) [21] In several strains, a

disubsti-tution-pattern of GlcI has been observed, including b-D

-Glcp (at O-4)/PCho (at O-6) [17], b-D-Galp (at O-4)/PCho

(at O-6) [20], Ac (at O-4)/PCho (at O-6) [18] and Ac (at

O-4)/L,D-Hep (at O-6) [21] In each strain analysed, PCho

addition has been shown to be directed by the products of

the lic1 locus Our recent studies have focussed on the

structural diversity of LPS expression, and the genetic basis

for that diversity, in a representative set consisting of 25

NTHi clinical isolates obtained from otitis media patients

[25] In the present investigation we report on the structural

analysis of LPS from one of these isolates (NTHi strain

981), which provides evidence for a novel branching pattern

at GlcI

Experimental procedures

Bacterial culture and preparation of LPS

NTHi strain 981 was obtained from the Finnish Otitis

Media Study Group and is an isolate obtained from the

middle ear [25] Bacteria were grown in brain-heart

infusion broth supplemented with haemin (10 lgÆmL)1),

NAD (2 lgÆmL)1) and N-acetylneuraminic acid (Neu5Ac)

(25 lgÆmL)1) LPS was extracted by the

phenol/chloro-form/light petroleum method, as described previously

[16]

Chromatography

Gel filtration chromatography and GLC were carried out as

described previously [16]

Preparation of OS material O-Deacylation of LPS O-Deacylation of LPS was achieved with anhydrous hydrazine, as described previously [16,26]

Mild acid hydrolysis of LPS Reduced core OS material was obtained after mild acid hydrolysis (1% aqueous acetic acid, pH 3.1, 100C, 2 h) and simultaneous reduction (borane-N-methylmorpholine complex, 20 mg) of LPS (120 mg) The insoluble lipid A (62 mg) was separated

by centrifugation and the water-soluble part was repeatedly chromatographed on a P-4 column, giving a major (OS-1, 8.5 mg) and a minor (OS-2, 3.7 mg) OS-containing fraction

Dephosphorylation of OS Dephosphorylation of OS material was performed with 48% aqueous HF, as described previously [18]

Mass spectrometry GLC-MS and ESI-MS were performed as described previously [16,21] ESI-MSn on permethylated dephos-phorylated OS was performed using a Finnigan-MAT LCQ ion trap mass spectrometer (Finnigan-MAT; San Jose, CA, USA) in the positive ion mode The samples were dissolved in methanol/water (7 : 3) containing 1 mM NaOAc to a concentration of about 1 mgÆmL)1, and were injected into a running solvent of identical composition at

10 lLÆmin)1 NMR spectroscopy NMRspectra were obtained at 25C (OS) or 20 C [O-deacylated LPS (LPS-OH)] either on a Varian UNITY

600 MHz spectrometer or on a JEOL JNM-ECP500 spectrometer, using the previously described experiments [16,18,21]

Analytical methods Sugars were identified as their alditol acetates, as previously described [27] Methylation analysis was performed as described previously [16] The relative proportions of the various alditol acetates and partially methylated alditol acetates obtained in sugar- and methylation analyses correspond to the detector response of the GLC-MS Permethylation of dephosphorylated OS was performed in the same way as in the methylation analyses [16], but without the prior acetylation step The absolute configura-tions of the hexoses were determined by the method devised

by Gerwig et al [28] The total content of fatty acids was analysed as described previously [29]

Results

Characterization of LPS NTHi strain 981 was cultured in liquid media and the LPS was extracted using the phenol/chloroform/light pet-roleum method Compositional sugar analysis of the LPS

Fig 1.

5 Structural model of the lipopolysaccharide from Haemophilus

influenzae R 1 ¼ H, phosphocholine (PCho), D -glycero- D

-manno-heptose ( D , D -Hep) or L -glycero- D -manno-heptose ( L , D -Hep); R 2 , R 4 ,

R 5 ¼ H, glucose (Glc), galactose (Gal) or acetate (Ac)

Y ¼ Gly, P or phosphoethanolamine (PEtn) Note that substitution

with Gly has also been reported at HepI, HepII and Kdo [24].

sample indicated D-glucose (Glc), D-galactose (Gal),

2-amino-2-deoxy-D-glucose (GlcN),

2-amino-2-deoxy-D-galactose (GalN),D-glycero-D-manno-heptose (D,D-Hep)

and L-glycero-D-manno-heptose (L,D-Hep) in the ratio

25 : 26 : 18 : 1 : 5 : 25, as identified by GLC-MS of their

corresponding alditol acetate and 2-butyl glycoside

deriva-tives [28] In previous investigations the LPS was found

to contain ester-linked glycine [24] and a low level of

Neu5Ac [30]

On treatment of the LPS with anhydrous hydrazine

under mild conditions, water-soluble O-deacylated LPS

(LPS-OH) was obtained ESI-MS data (Table 1) indicated a

heterogeneous mixture of glycoforms consisting of two

subpopulations: a major subpopulation in which the

glycoform compositions comprised three heptoses and, to

a great extent, PCho (Hep3-glycoforms); and a minor

subpopulation with compositions comprising four heptoses

but lacking PCho (Hep4-glycoforms) Quadruply charged

ions were observed at m/z 640.5/671.3 (minor), 680.8/

711.5 (major), 721.4/751.7 and 772.3/802.9, corresponding

to Hep3-glycoforms with the compositions PChoÆHex2Æ

Hep3ÆPEtn2)3ÆP1ÆKdoÆLipid A-OH, PChoÆHex3ÆHep3Æ

PEtn2)3ÆP1ÆKdoÆLipid A-OH, PChoÆHex4ÆHep3ÆPEtn2)3ÆP1Æ

KdoÆLipid A-OH and PChoÆHexNAc1ÆHex4ÆHep3ÆPEtn2)3Æ

P1ÆKdoÆLipid A-OH, respectively Minor quadruply charged

ions were also found at m/z 639.8/670.3, consistent with

Hep3-glycoforms that were lacking PCho with the

compo-sitions Hex3ÆHep3ÆPEtn2)3ÆP1ÆKdoÆLipid A-OH Quadruply

charged ions at m/z 646.9/677.8, 687.7/718.4 (minor) and

728.1/758.8 (major) corresponded to Hep4-glycoforms

with the compositions Hex2ÆHep4ÆPEtn2)3ÆP1ÆKdoÆLipid

A-OH, Hex3ÆHep4ÆPEtn2)3ÆP1ÆKdoÆLipid A-OH and Hex4Æ

Hep4ÆPEtn2)3ÆP1ÆKdoÆLipid A-OH, respectively

Characterization of OS Core OS material was obtained after partial acid hydrolysis

of LPS with dilute aqueous acetic acid The OS material was repeatedly chromatographed on a P-4 column with inter-mediate selection for the Hep3- and Hep4-glycoforms (by ESI-MS) This procedure resulted in a major fraction (OS-1, 8.5 mg) in which the Hep3-glycoforms accounted for about 95%, and a minor fraction (OS-2, 3.7 mg) in which the Hep4-glycoforms accounted for about 80% (Table 2) In the ESI-MS spectrum of OS-1 (Fig 2A), doubly charged ions at m/z 684.9 (minor), 766.0 (minor), 847.2 (major) and 928.3, corresponded to the respective compositions PChoÆ Hex1)4ÆHep3ÆPEtn2ÆAnKdo-ol, while ions at m/z 949.1 (very minor) and 1029.9 were consistent with the composi-tions PChoÆHexNAc1ÆHex3)4ÆHep3ÆPEtn2ÆAnKdo-ol, res-pectively In the ESI-MS spectrum of OS-2 (Fig 2B), doubly charged ions at m/z 698.4 (minor), 779.6 (major), 860.6, 941.8 (major) and 1022.5 (very minor) were consistent with the respective compositions Hex1)5ÆHep4ÆPEtn2 ÆAnK-do-ol In both OS fractions, minor peaks were observed corresponding to the above-mentioned compositions but additionally containing glycine or a phosphate group Minor peaks could also be observed corresponding to glycoforms containing only one PEtn group or containing one PEtn group and a phosphate group

In order to obtain sequence and branching information, the OS fractions were dephosphorylated and permethylated and subjected to ESI-MSn [21,31] Sodiated adduct ions were observed in the MS spectra (positive mode) corres-ponding to the compositions Hex1)4ÆHep3ÆAnKdo-ol, Hex-NAc1ÆHex3)4ÆHep3ÆAnKdo-ol, Hex1)5ÆHep4ÆAnKdo-ol and HexNAc1ÆHex5ÆHep4ÆAnKdo-ol (Tables 3 and 4) The

Table 1 Negative ion ESI-MS data and proposed compositions for O-deacylated lipopolysaccharide (LPS-OH) of nontypeable Haemophilus influ-enzae (NTHi) strain 981 Average mass units were used for calculation of molecular mass values based on proposed compositions, as follows: Hex (hexose), 162.14; HexNAc (N-acetylhexosamine), 203.19; Hep (heptose), 192.17; Kdo (3-deoxy- D -manno-oct-2-ulosonic acid), 220.18; P (phos-phate), 79.98; PEtn (phosphoethanolamine), 123.05; PCho (phosphocholine), 165.13 and lipid A-OH (O-deacylated lipid A), 953.02 Relative abundance was estimated from the area of molecular ion peak relative to the total area (expressed as percentage) Peaks representing less than 3% of the base peak were not included in the table.

Observed ions (m/z) Molecular mass (Da)

Relative abundance (%) Proposed composition

monosaccharide sequence and branching for the different

glycoforms were obtained following collision-induced

dis-sociation (CID) of the glycosidic bonds [21,31] Through the

ion mass distinction between reducing, nonreducing and

internal fragments resulting from the bond ruptures [21,31],

the topology could be determined for nearly all

composi-tions found in the MS profiling spectra (Tables 3 and 4) In

those cases, where the same glycoform was found in both

OS-1 and OS-2 (Table 2), experiments showed that identical

structures were present in both fractions

For the major Hep3-glycoform with the composition

Hex3ÆHep3ÆAnKdo-ol ([M + Na]+1671.8 Da), ion

selec-tion and collisional activaselec-tion of the precursor ion at m/z

1671.8 provided the MS2 spectrum shown in Fig 3 The

fragmentation pattern was consistent with two nonreducing

terminal branches attached to a disubstituted heptose

(HepI) which is linked to the terminal AnKdo-ol residue

The ions at m/z 1409.7 and 1161.6 corresponded to losses of

terminal-Hep (t-Hep) and t-Hep–Hep, respectively, which

indicated the presence of a HepIII–HepII-disaccharide

branch Furthermore, ions at m/z 1453.7, 1249.6 and

1045.5, corresponding to the respective loss of t-Hex,

t-Hex–Hex and t-Hex–Hex–Hex, indicated the presence of

a Hex–Hex–Hex-trisaccharide branch Loss of the terminal

AnKdo-ol residue was also observed from the ion at m/z

1393.6 The assigned structure was confirmed by MS3

experiments on selected product ions (data not shown) The

structures of the other Hep3-glycoforms were obtained in an

analogous manner (data not shown) and for all glycoforms

the hexoses were found to be members of a linear chain

attached to HepI (Table 3)

For the major Hep4-glycoform with the composition

Hex4ÆHep4ÆAnKdo-ol ([M + Na]+2124.0 Da), ions in the

MS2 spectrum (precursor ion [M + 2Na]2+ 1073.5 Da)

(Fig 4A) at m/z 1861.8 (loss of t-Hep) and 1613.8 (loss of

t-Hep–Hep) indicated the presence of a HepIII–HepII-branch attached to HepI, as was also found in the Hep3-glycoforms (Tables 3 and 4) An OS extension consisting

of four hexose residues and one heptose residue was indicated to be linked to HepI from the occurrence of an ion at m/z 1045.6 (counterpart at m/z 1101.6) correspond-ing to loss of the entire Hex4Hep moiety The fragmen-tation pattern clearly indicated this OS extension not to be arranged in a linear chain as fragment ions were found at m/z1905.8 (loss of t-Hex), 1701.7 (loss of t-Hex–Hex) and 1657.7 (either loss of t-Hex–Hep or t-Hep–Hex), consis-tent with the presence of a Hex–Hex-disaccharide branch and a second disaccharide branch that either could be Hex–Hep- or Hep–Hex- (the former alternative was shown, see below) Ions resulting from double glycosidic bond cleavage could also be observed at m/z 1395.7 (e.g loss of t-Hep–Hep/t-Hex) and at m/z 1191.5 (loss of t-Hep–Hep/t-Hex–Hex) An MS3 experiment on m/z 1613.8 (Fig 4B) gave fragment ions at m/z 1101.6 (coun-terpart at m/z 535.3), as a result of glycosidic cleavage between the Hex4Hep moiety and HepI Ions at m/z 1395.7, 1191.6 and 1147.4, corresponding to the respective loss of t-Hex, t-Hex–Hex and t-Hex–Hep, indicated that the Hex4Hep moiety consisted of two nonreducing ter-minal disaccharide branches (Hex–Hex- and Hex–Hep-) attached to a disubstituted hexose To confirm the assigned structure, an MS4 experiment on m/z 1191.6 was per-formed (Fig 4C), which showed the expected ions at m/z 973.5 (loss of t-Hex) and 725.3 (loss of t-Hex–Hep) The topology of the other Hep4-glycoforms were determined in

a similar manner (data not shown) For the higher molecular mass Hep4-glycoforms (Hex3–Hex5 and Hex-NAc1Hex5), the hexose residue attached to HepI was found to be disubstituted with a Hex–Hep-branch and a second branch containing hexoses (Table 4)

Table 2 Negative ion ESI-MS data and proposed compositions for oligosaccharide (OS)-1 and (OS)-2 derived from the lipopolysaccharide (LPS) of nontypeable Haemophilus influenzae (NTHi) strain 981 Average mass units were used for calculation of molecular mass values based on proposed compositions as follows: Hex (hexose), 162.14; HexNAc (N-acetylhexosamine), 203.19; Hep (heptose), 192.17; AnKdo-ol (reduced anhydro Kdo), 222.20; PEtn (phosphoethanolamine), 123.05; and PCho (phosphocholine), 165.13 Relative abundance was estimated from the area of molecular ion peak relative to the total area (expressed as a percentage) Minor peaks were observed corresponding to the proposed compositions but additionally containing glycine or a phosphate group Minor peaks could also be observed corresponding to glycoforms containing only one PEtn group or containing one PEtn group and a phosphate group ND, not detected.

Observed ions

(m/z) (M-2H)2)

Proposed composition

a

Trace amounts, defined as peaks representing less than 3% of the base peak.

Methylation analysis of dephosphorylated OS-1

indi-cated terminal-Gal (t-Gal), 4-substituted-Gal (4-Gal),

4-Glc, 3-Gal, t-L,D-Hep, 4,6-disubstituted-Glc (4,6-Glc),

4-D,D-Hep, 2-L,D-Hep and 3,4-L,D-Hep in the relative proportions 24 : 5 : 31 : 2 : 9 : 3 : 3 : 10 : 13 together with trace amounts of t-Glc and t-GalN The methylation analysis of intact OS-1 showed significantly decreasing amounts of 2-L,D-Hep and t-L,D-Hep, which could be derived from the presence of PEtn substituents at HepII and HepIII, respectively (see below) Methylation analysis of dephosphorylated OS-2 showed t-Gal, 4-Glc, 6-Glc, t-L,D -Hep, 4,6-Glc, 4-D,D-Hep, 2-L,D-Hep and 3,4-L,D-Hep in the ratio 32 : 15 : 2 : 10 : 5 : 15 : 10 : 11 together with trace amounts of t-Glc, 4-Gal, 3-Gal and t-D,D-Hep The increas-ing amounts (compared to dephosphorylated OS-1) of t-Gal, 4-D,D-Hep, 6-Glc and 4,6-Glc indicated that the structural element Hex–Hep–Hex–(–) in the Hep4-glyco-forms was probably Gal-(1fi4)-D,D-Hep-(1fi4,6)-Glc, and this was confirmed and detailed by NMRspectroscopy (see below)

Characterization of OS fractions and LPS-OH by NMR The 1H NMRresonances of OS fractions and LPS-OH were assigned by1H–1H chemical shift correlation experi-ments (DQF-COSY and TOCSY) Subspectra correspond-ing to the individual glycosyl residues were identified on the basis of spin-connectivity pathways delineated in the 1H chemical shift correlation maps, the chemical shift values, and the vicinal coupling constants From the glycoform compositions of the different OS fractions indicated by ESI-MS (Table 2), the spin-systems could more easily be identified as originating from either the Hep3- or the Hep4-glycoform population The 13C NMRresonances of OS material and LPS-OH were assigned by heteronuclear

1H–13C chemical shift correlation in the1H detected mode (HSQC) The chemical shift data obtained for the Hep3-and Hep4-glycoforms are presented in Tables 5 Hep3-and 6, respectively, and are consistent with each D-sugar residue being present in the pyranosyl ring form Further evidence for this conclusion was obtained from NOE data, which also served to confirm the anomeric configurations of the linkages and, together with a heteronuclear multiple-bond correlation (HMBC) experiment on OS-2, determined the monosaccharide sequence and branching pattern

Characterization of the Kdo-lipid A-OH region ESI-MS data (Table 1), fatty acid compositional analysis (yielding 3-hydroxytetradecanoic acid) and NMRexperiments on LPS-OH (data not shown, giving similar results as for NTHi strains 486 [16] and 1003 [18]), indicated the presence

of the common Kdo-(2fi6)-lipid A-OH element in NTHi strain 981 As observed previously [16,18], two spin-systems

Table 3 Structures of the Hep3-glycoforms of nontypeable Haemophilus influenzae (NTHi) strain 981, as indicated by ESI-MS n on permethylated dephosphorylated oligosaccharide (OS) ND, not determined.

8

Fig 2 Negative ion ESI-MS spectra of oligosaccharide (OS)-1 (A) and

OS-2 (B) derived from the lipopolysaccharide (LPS) of nontypeable

Haemophilus influenzae (NTHi) strain 981 showing doubly charged ions.

(A) The peak at m/z 684.9 corresponds to a glycoform with the

com-position, phosphocholine (PCho)ÆHex 1 ÆHep 3 Æphosphoethanolamine

(PEtn) 2 ÆAnKdo-ol (B) The peak at m/z 698.4 corresponds to a

gly-coform with the composition Hex 1 ÆHep 4 ÆPEtn 2 ÆAnKdo-ol Minor

peaks were observed corresponding to the compositions proposed in

Table 2 but additionally containing glycine (indicated by · ) or a

phosphate group (indicated by ) Minor peaks could also be observed

corresponding to glycoforms containing only one PEtn group

(indi-cated by e) or containing one PEtn group and a phosphate group

(indicated by +).

could be traced for the single a-linked Kdo residue, probably owing to the partial occurrence of PEtn attached

to the phosphate group at O-4 of Kdo [11,13,16]

Structure of the core region of the Hep3-glycoforms In the 1H NMRspectrum of OS-1 (Fig 5A), anomeric resonances of HepI–HepIII were observed at d 5.89–5.82 (1H, not resolved), 5.24–5.22 (1H, not resolved) and 5.15– 5.01 (1H, not resolved) The Hep ring systems were identified on the basis of the observed small J1,2 values, and their a-configurations were confirmed by the occurrence

of single intraresidue NOE between the respective H-1 and H-2 resonances, as observed previously [16,18] An intense signal from methyl protons of an N-acetyl group was obser-ved at d 2.04, which correlated to a13C signal at d 22.7 in the HSQC spectrum A crosspeak from the ester-linked glycine substituent was observed at d 4.00/40.7 (in the HSQC spectrum) as a result of correlation between the methylene proton and its carbon Several signals for methylene protons

of AnKdo-ol were observed in the DQF-COSY and TOCSY spectra of OS-1 in the region d 2.26–1.67, as observed and explained previously [10,16,21] The mono-saccharide sequence within the inner-core region, as indicated by ESI-MSn (described above), was confirmed and detailed from transglycosidic NOE connectivities between the proton pairs HepIII H-1/HepII H-2, HepII H-1/HepI H-3 (LPS-OH and OS-1) and HepI H-1/Kdo H-5 and H-7 (LPS-OH), evidencing the sequence as L-a-D -Hepp-(1fi2)-L-a-D-Hepp-(1fi3)-L-a-D-Hepp-(1fi5)-a-Kdop Relatively large J1,2values ( 7.8 Hz) of the anomeric resonances observed at d 4.64, 4.62, 4.55, 4.52 and 4.46, indicated that each of the corresponding residues had the b-anomeric configuration The residues with anomeric signals at d 4.94 (J 4.0 Hz) and 4.91 (J 4.0 Hz) were identified as having the a-anomeric configuration The assignments of the anomeric configurations were also supported by the occurrence of intraresidue NOE between the respective H-1, H-3 and H-5 resonances (b-configur-ation) or between H-1 and H-2 (a-configur(b-configur-ation) On the basis of the chemical shift data and the large J2,3, J3,4and

n on

a Majo

2 exper

b M

Fig 3 ESI-MS2analysis of permethylated dephosphorylated oligosac-charide (OS) derived from the lipopolysacoligosac-charide (LPS) of nontypeable Haemophilus influenzae (NTHi) strain 981 The product ion spectrum

is shown of [M + Na]+m/z 1671.8, corresponding to a glycoform with the composition Hex 3 ÆHep 3 ÆAnKdo-ol The proposed structure is shown in the inset.

J4,5values ( 9 Hz), the residues with anomeric shifts of d

4.55 and 4.64 could be attributed to the 4-Glc (GlcI and

GlcII) identified by methylation analysis On the basis of

low J3,4and J4,5values (<4 Hz) and chemical shift data, the

residues with anomeric resonances at d 4.46, 4.94, 4.52, 4.91

and 4.62 were attributed to the t-Gal (GalI and GalII),

4-Gal (GalI*), 3-Gal (GalII*) and t-GalNAc (GalNAc)

identified by linkage analysis

Signals for the methyl protons of PCho were observed at

d 3.23 and characteristic spin-systems for ethylene protons

from this residue and from the two PEtn residues were

found at d 4.41/3.69 (PCho), 4.15/3.29 (PEtnI) and 4.17/

3.29 (PEtnII) 1H–31P NMRcorrelation studies (Fig 6)

demonstrated PCho to be located at GlcI and the two PEtn

residues to be situated at HepII and HepIII, respectively

Intense31P resonances from phosphodiesters were observed

at d 0.33,)0.50 and )0.62 Correlations between the former

signal and the signals from H-6 of HepII (d 4.55) and the methylene proton pair of PEtnI (d 4.15) in the 1H–31P HMQC experiment evidenced substitution by PEtn at O-6

of HepII The second PEtn residue was demonstrated to be linked to O-3 of HepIII as the 31P resonance at d)0.50 correlated to the signals from H-3 of HepIII (d 4.33) and the methylene proton pair of PEtnII (d 4.17) Correlations between the signal at d)0.62 and the signals from the H-6 protons of GlcI (d 4.30) and the methylene protons of PCho (d 4.41) established the PCho substituent to be located at O-6 of this residue

The occurrence of interresidue NOE connectivities between the proton pairs GalI H-1/GlcII H-4, GlcII H-1/ GlcI H-4 and GlcI H-1/HepI H-4 and H-6 established the presence of the tetrasaccharide unit b-D -Galp-(1fi4)-b-D-Glcp-(1fi4)-[PChofi6]-b-D-Glcp-(1fi4)-L-a-D -Hepp-(1fi) This element was shown to be further chain elongated

by an a-D-Galp residue because transglycosidic NOE between GalII H-1/GalI* H-4 and GalI* H-1/GlcII H-4 were observed Further chain extension by a b-D-GalpNAc residue was evidenced from the occurrence of interresidue NOE between GalNAc H-1/GalII* H-3 and GalII* H-1/ GalI* H-4 From the combined data, the structure in Fig 7

is proposed for the fully extended globotetraose-containing Hep3-glycoform (HexNAc1Hex4) of NTHi strain 981 It was thus established that the Hex4 and Hex3 glycoforms are sequential truncations of this structure It is also likely that the Hex2 and Hex1 glycoforms are further sequential truncations, which is consistent with the occurrence of small amounts of t-Glc in the methylation analysis Indeed, two weak crosspeaks are observed in the COSY spectrum at d 4.57/3.50 and 4.56/3.38, which possibly arise from the two b-D-Glcp residues appearing as terminal residues; however, this could not be confirmed because of a significant overlap

of signals

Structure of the core region of the Hep4-glycoforms In the 1H NMRspectrum of OS-2 (Fig 5B), anomeric resonances of HepI–HepIII were observed at d 6.01–5.87 (HepII, not resolved), 5.17–5.03 (HepI, not resolved) and 5.14/5.10 (HepIII, not resolved) The chemical shift values for the other resonances of HepI–HepIII (data not shown) were similar to the corresponding values in the Hep3-glycoforms Transglycosidic NOE connectivities (data simi-lar to that of the Hep3-glycoforms) confirmed the sequence

of the characteristic triheptosyl unit Anomeric signals of the fourth heptose were identified at d 5.06/4.94 (HepIV, not resolved) and single intraresidue NOE between H-1 and H-2 confirmed its a-configuration This residue could be attributed to the 4-D,D-Hep identified in the methylation analysis on the basis of chemical shift data (Table 6)

1H–31P NMRcorrelation studies (data similar to that of the Hep3-glycoforms) demonstrated the PEtn substituents to be located at O-6 of HepII and at O-3 of HepIII, as previously observed (see above)

Relatively large J1,2values ( 7.8 Hz) of the anomeric resonances observed at d 4.57, 4.54, 4.53 (very weak), 4.51, 4.50, 4.49 and 4.48, indicated each of the corresponding residues to have the b-anomeric configuration The residue with an anomeric signal at d 4.95 (very weak, J 4.0 Hz) was identified as having the a-anomeric configuration The anomeric configurations were also confirmed from

Fig 4 ESI-MSnanalysis of permethylated dephosphorylated

oligosac-charide (OS) derived from the lipopolysacoligosac-charide (LPS) of nontypeable

Haemophilus influenzae (NTHi) strain 981 (A) MS 2 spectrum of

[M + 2Na]2+m/z 1073.5, corresponding to a glycoform with the

composition Hex 4 ÆHep 4 ÆAnKdo-ol The proposed structure is shown

in the inset in B (B) MS 3 spectrum of the fragment ion at m/z 1613.8.

(C) MS4spectrum of the fragment ion at m/z 1191.6

intraresidue NOE, as described earlier (see above) The

residues with anomeric shifts of d 4.50, 4.54 and 4.57 could

be attributed to the 6-Glc (GlcI), 4,6-Glc (GlcI*) and 4-Glc

(GlcII) identified by methylation analysis, on the basis of the

chemical shift data and the large J2,3, J3,4and J4,5values

( 9 Hz) On the basis of low J3,4and J4,5values (<4 Hz)

and chemical shift data, the residues with anomeric

resonances at d 4.48, 4.95, 4.51/4.49 and 4.53 were attributed

to the t-Gal (GalI, GalII and GalIII) and 4-Gal (GalI*)

identified by linkage analysis

Interresidue NOE were observed between the proton

pairs GalIII H-1 (d 4.51)/HepIV H-4 (d 3.92), HepIV H-1

(d 4.94)/GlcI H-6A, H-6B and GlcI H-1/HepI H-4 and H-6,

which established the tetrasaccharide unit b-D

-Galp-(1fi4)-D-a-D-Hepp-(1fi6)-b-D-Glcp-(1fi4)-L-a-D-Hepp-(1fi of the Hex2 glycoform The structure of the Hex4 glycoform was concluded from interresidue NOE between the proton pairs GalIII H-1 (d 4.49)/HepIV H-4 (d 3.90), HepIV H-1 (d 5.06)/GlcI* H-6A, H-6B, GlcI* H-1/HepI H-4 and H-6 and between the proton pairs GalI H-1/GlcII H-4, GlcII H-1/ GlcI* H-4, which evidenced the hexasaccharide unit b-D -Galp-(1fi4)-b-D-Glcp-(1fi4)-[b-D-Galp-(1fi4)-D-a-D-Hep p-(1fi6)]-b-D-Glcp-(1fi4)-L-a-D-Hepp-(1fi This monosac-charide connectivity was also confirmed by transglycosidic correlations in an HMBC experiment, where correlations were seen between GalIII H-1/HepIV C-4, HepIV

Table 5. 1H and13C NMR chemical shifts for Hep3-glycoforms of oligosaccharide (OS)-1 derived from the lipopolysaccharide (LPS) of nontypeable Haemophilus influenzae (NTHi) strain 981 Data were recorded in D 2 O at 25 C Signals corresponding to N-acetyl- D -galactosamine (GalNAc) and phosphocholine (PCho) methyl protons and carbons occurred at 2.04/22.7 and 3.23/54.5 p.p.m., respectively Pairs of deoxyprotons of AnKdo-ol (reduced anhydro 3-deoxy- D -manno-oct-2-ulosonic acid) were identified in the DQF-COSY spectrum at 2.26–1.67 p.p.m.

6

› PEtn

3

›

PEtn

6

› PCho

GalII* e

a Several signals were observed for HepI, HepII and HepIII owing to heterogeneity in the AnKdo moiety b H-4/H-6 of HepI were identified

at d 4.28/4.09–4.11 by NOE from GlcI c –, not obtained owing to the complexity of the spectrum d Tentative assignment from NOE data.

e

Residue marked with * denotes a further substituted analogue of the corresponding residue without *.

H-1/GlcI* 6, GalI H-1/GlcII 4 and GlcII H-1/GlcI*

C-4 This element was shown to be further chain extended by

an a-D-Galp residue, providing the Hex5 glycoform, as transglycosidic NOE between GalII H-1/GalI* H-4 and GalI* H-1/GlcII H-4 were observed From the combined data, the structure in Fig 8 is proposed for the Hex5 glycoform in the Hep4-glycoform population of NTHi strain 981 The Hex3 and Hex1 glycoforms are probably sequential truncations of the established structures, which is consistent with the low amounts of t-Glc and t-D,D-Hep found in the methylation analysis A weak crosspeak in the COSY spectrum at d 4.55/3.36 may arise from the terminal

Table 6.1H and13C NMR chemical shifts for Hep4-glycoforms of oligosaccharide (OS)-2 derived from the lipopolysaccharide (LPS) of nontypeable Haemophilus influenzae (NTHi) strain 981 Data were recorded in D 2 O at 25 C Chemical shift values for resonances of HepI-HepIII, PEtn and Gly were similar to the corresponding values in the Hep3-glycoforms (see Table 5).

GlcI* d

GalI* d

a

Two spin-systems were found for the residue probably as a result of differences in the chemical environments of the various glycoforms The first spin-system is attributed to the Hex2 glycoform, while the second spin-system is attributed to the Hex4 and Hex5 glycoforms.b–, not obtained owing to the complexity of the spectrum c Tentative assignment from NOE data d Residue marked with * denotes a further substituted analogue of the corresponding residue without *.

Fig 5 The 600-MHz1H NMR spectra of oligosaccharide (OS)-1 (A)

and OS-2 (B) derived from the lipopolysaccharide (LPS) of nontypeable

Haemophilus influenzae (NTHi) strain 981 showing the anomeric

regions (A) Anomeric resonances that are characteristic for the

Hep3-glycoforms are labelled Also indicated is an ethylene proton signal

from phosphocholine (PCho) at d 4.41 (B) Anomeric resonances that

are characteristic for the Hep4-glycoforms are labelled.

Fig 6 Part of the 500-MHz heteronuclear1H–31P HMQC spectrum of oligosaccharide (OS)-1 derived from the lipopolysaccharide (LPS) of nontypeable Haemophilus influenzae (NTHi) strain 981 Assignments are labelled.

b-D-Glcp residue in the major isomer of the Hex3 glycoform

(Table 4), but, as a result of overlapping signals, this could

not be confirmed Given the structure of the HexNAc1Hex5

glycoform by ESI-MSn(Table 4), a terminal b-D-GalpNAc

residue is assumed in analogy of the Hep3-glycoforms

Discussion

This investigation has shown that LPS from NTHi strain

981 contain either three heptoses (Hep3-glycoforms) or four

heptoses (Hep4-glycoforms) In both of these glycoform

subpopulations, either the fully assembled globotetraose

unit [b-D-GalpNAc-(1fi3)-a-D-Galp-(1fi4)-b-D

-Galp-(1fi4)-b-D-Glcp] or sequentially truncated versions thereof

were found to be linked to O-4 of the b-D-Glcp residue

(labelled GlcI) that is attached to HepI of the conserved

triheptosyl inner-core moiety This is the first example of an

H influenzaestrain expressing globotetraose in this

mole-cular environment However, recently the H influenzae

type b strain, RM7004, was reported to express a

trun-cated epitope, the globoside trisaccharide (globotriose)

[a-D-Galp-(1fi4)-b-D-Galp-(1fi4)-b-D-Glcp] in the same

molecular environment [22] Additionally, in that strain,

the globotriose epitope was expressed as a terminal unit of a

tetrasaccharide extension from HepII, the same

tetrasac-charide that previously had been found in the type b strain,

RM153, at that location [11] A third molecular environ-ment for the globotriose epitope has been reported for

H influenzae strain RM118 (Rd) [13] and several NTHi isolates [21], where globotriose/globotetraose are present as terminal OS extensions at HepIII The globotriose epitope is found on many human cells (pkblood group antigen), and

in H influenzae is thought to be important for virulence by acting as a mimic of these host structures, thus allowing the organism to evade some element of the host immune response [5,32] For strain RM118, the glycosyltransferases involved in the assembly of its globotetraose side-chain were recently identified [7] The lpsA gene was shown to be involved in the addition of a b-D-Glcp residue in a 1,2-linkage to initiate chain extension from HepIII Furthermore, the lic2A, lgtC and lgtD genes were shown

to encode glycosyltransferases involved in sequential addi-tion of b-1,4-linked Galp (Lic2A), a-1,4-linked Galp (LgtC) and b-1,3-linked GalpNAc (LgtD), resulting in the fully assembled globotetraose structure Both lic2A and lgtC are phase-variable genes [32,33] as is lex2 [34], which is the gene responsible for the addition of a b-D-Glcp residue in a 1,4-linkage to GlcI in strain RM7004 (R Aubrey, A D Cox, K Makepeace, J C Richards, E R Moxon & D W Hood, unpublished results) Genetic analysis has indicated the presence of lex2, lic2A, lgtC and lgtD gene homologues

in NTHi strain 981 (data not shown), and although it would

Fig 7 Structure proposed for the fully extended Hep3-glycoform (HexNAc1Hex4) of nontypeable Haemophilus influenzae (NTHi) strain 981.

Fig 8 Structure proposed for a Hep4-glycoform (Hex5) of nontypeable Haemophilus influenzae (NTHi) strain 981.