Báo cáo khoa học: Towards understanding the functional role of the glycosyltransferases involved in the biosynthesis of Moraxella catarrhalis lipooligosaccharide ppt

Furthermore, it is concluded that Lgt4 functions as an N-acetylglucosylamine transferase responsible for the addi-tion of an a-d-GlcNAc 1fi 2 glycosidic linkage to the 1 fi 4 branch, and a

glycosyltransferases involved in the biosynthesis of

Moraxella catarrhalis lipooligosaccharide

Ian R Peak1, I D Grice1, Isabelle Faglin1, Zoran Klipic1, Patrick M Collins1,

Lucien van Schendel1, Paul G Hitchen2, Howard R Morris3, Anne Dell2and Jennifer C Wilson1

1 Institute for Glycomics, Griffith University, Gold Coast Campus, Queensland, Australia

2 Division of Molecular Biosciences, Faculty of Natural Sciences, Imperial College London, UK

3 M-SCAN Mass Spectrometry Research and Training Centre, Silwood Park, Ascot, UK

Glycosyltransferases are enzymes that synthesize the

carbohydrate structures of the lipooligosaccharides

(LOSs) that are abundant on the surface of

Gram-negative bacteria These carbohydrate-rich structures

have been implicated in the pathogenic mechanisms of

many bacteria Generally, each glycosyltransferase is

exquisitely unique, in that it has its own donor,

accep-tor and linkage specificity, and a vast array of these enzymes are required to assemble these complex struc-tures [1–3]

Recently, there has been some progress towards identifying the genes expressing the glycosyltrans-ferase enzymes involved in LOS biosynthesis in Moraxella catarrhalis, a human upper respiratory

Keywords

glycosyltransferase; lipooligosaccharide

biosynthesis; Moraxella catarrhalis; MS;

NMR spectroscopy

Correspondence

J C Wilson, Institute for Glycomics, Griffith

University, Gold Coast Campus, PMB 50,

QLD 4215, Australia

Fax: +61 7 555 28908

Tel: +61 7 555 28077

E-mail: Jennifer.wilson@griffith.edu.au

Database

Sequences have been deposited under the

accession numbers DQ071425 (2951 locus),

DQ071426 (3292 locus) and DQ071427–

DQ071431 (lgt2 alleles)

(Received 24 August 2006, revised 29

January 2007, accepted 16 February 2007)

doi:10.1111/j.1742-4658.2007.05746.x

The glycosyltransferase enzymes (Lgts) responsible for the biosynthesis of the lipooligosaccharide-derived oligosaccharide structures from Moraxella catarrhalis have been investigated This upper respiratory tract pathogen is responsible for a spectrum of illnesses, including otitis media (middle ear infection) in children, and contributes to exacerbations of chronic obstruct-ive pulmonary disease in elderly patients To investigate the function of the glycosyltransferase enzymes involved in the biosynthesis of lipooligosaccha-ride of M catarrhalis and to gain some insight into the mechanism of sero-type specificity for this microorganism, mutant strains of M catarrhalis were produced Examination by NMR and MS of the oligosaccharide structures produced by double-mutant strains (2951lgt1⁄ 4D and 2951lgt5⁄ 4D) and a single-mutant strain (2951lgt2D) of the bacterium has allowed us to propose a model for the serotype-specific expression of lipo-oligosaccharide in M catarrhalis According to this model, the presence⁄ absence of Lgt4 and the Lgt2 allele determines the lipooligosaccharide structure produced by a strain Furthermore, it is concluded that Lgt4 functions as an N-acetylglucosylamine transferase responsible for the addi-tion of an a-d-GlcNAc (1fi 2) glycosidic linkage to the (1 fi 4) branch, and also that there is competition between the glycosyltransferases Lgt1 and Lgt4 That is, in the presence of an active Lgt4, GlcNAc is preferen-tially added to the (1fi 4) chain of the growing oligosaccharide, instead of Glc In serotype B strains, which lack Lgt4, Lgt1 adds a Glc at this posi-tion This implies that active Lgt4 has a much higher affinity⁄ specificity for the b-(1fi 4)-linked Glc on the (1 fi 4) branch than does Lgt1

Abbreviations

APT, attached proton test; BHI, Brain Heart Infusion; DSS, 2,2-dimethylsilapentane-S-sulphonic acid; LOS, lipooligosaccharide;

OS, oligosaccharide; TMS, trimethylsilyl.

tract pathogen [4–6] Along with Haemophilus influenzae

and Streptococcus pneumoniae, this microorganism

is responsible for acute otitis media (middle ear

infection) in infants [7] M catarrhalis also contributes

to a spectrum of respiratory tract conditions occurring

in adult patients and causing or exacerbating

sinus-itis, pneumonia and chronic obstructive pulmonary

disease [7–9]

There are three major LOS serotypes of M

catarr-halis, A, B and C, which differ in the carbohydrate

content of the oligosaccharide component of their

LOS These serotypes represent 61%, 28.8% and

5.3% of clinical isolates in one study [10] Structural

analysis has revealed the oligosaccharide structure of

each of the serotypes [11–14] Several

glycosyltrans-ferase-encoding genes have been identified for

sero-types A and B, but the exact function of some of the

glycosyltransferase enzymes remain unclear [4–6]

Fur-thermore, the mechanism of LOS biosynthesis with

regard to serotype specificity remains to be elucidated

for this microorganism Herein are described two

double-mutant strains of M catarrhalis 2951, namely,

2951lgt1⁄ 4D and 2951lgt5 ⁄ 4D In addition, a

single-mutant strain 2951lgt2D is also described The

oligo-saccharide structures produced by these mutant

strains have provided significant insights into the

sequential addition of carbohydrate moieties to the

final LOS structure Furthermore, examination of

these oligosaccharide structures has revealed the

fol-lowing: (a) lgt4 encodes an N-acetylglucosamine

transferase; and (b) removal of the lgt4 gene leads to

replacement of a GlcNAc with a Glc These findings

have prompted us to conclude that there is

competi-tion between glycosyltransferases Lgt1 and Lgt4

Results

In order to investigate the function of the

glycosyl-transferase enzymes involved in the biosynthesis of

LOS of M catarrhalis, and to gain some insights into

the mechanism of serotype specificity for this

micro-organism, mutant strains of M catarrhalis were

produced Mutation of genes encoding the

glycosyl-transferase enzymes that assemble the LOS of

M catarrhalis leads to mutant bacteria that produce

truncated oligosaccharide structures Examination of

the truncated oligosaccharide structures has allowed

us to infer a function for the role of the

glycosyl-transferase enzymes in LOS biosynthesis The

oligo-saccharide structures isolated from these mutant

bacteria are designated 2951lgt2D, 2951lgt1⁄ 4D, and

2951lgt5⁄ 4D

Analysis of the genes encoding the glycosyltransferase enzymes responsible for LOS biosynthesis for serotype A M catarrhalis Table 1 gives the strains and plasmids utilized in this study Table 2 summarizes the oligonucleotide primers used to amplify or sequence the glycosyltransferase genes

lgt2 Sequence analysis of lgt2 from CCUG 3292 (serotype B) revealed that it is 765 bp in length, and is identical

to that of the reported serotype B strain 7169 (des-cribed as lgt2B⁄ C by Edwards et al [5]) The corres-ponding gene from serotype A strain 2951 (lgt2A) is also 765 bp long, but differs significantly from lgt2B⁄ C: these alleles differ by only one nucleotide in the first

287 nucleotides, whereas the remainder of the gene is only 52% identical, giving an overall 70% identity The lgt2 gene was amplified and sequenced from sev-eral strains of different serotypes (described in Table 1), using primers UORF2:2205 (within the con-served 5¢ region of lgt2) and DORF3:3434 (within lgt1) Our results confirm a previous report [5] that all serotype B and C strains contain the lgt2B⁄ C allele, whereas all serotype A strains contain the lgt2Aallele The function of the lgt2Aallele has not been previously described

lgt1, lgt4, and lgt5 Amplification and sequencing of lgt1 from CCUG

3292 (using primers DORF3:2768 and UORF3:4093) revealed an identical sequence to that of lgt1 of strains 7169 (serotype B [4]), and ATCC 43617 (sero-type B [5]) Amplification using these primers from strain 2951 produced a molecule approximately 1 kbp larger than that from the serotype B strain CCUG

3292 Sequence analysis of this larger fragment revealed that it contains an lgt1 allele (984 bp, 95% identical to lgt1 of ATCC 43617 and 7169), and an additional ORF of 996 bp with similarity to glycosyl-transferase-encoding genes The presence of this addi-tional gene, lgt4, was previously reported in strains of serotypes A and C [5], and our PCR and sequence analysis results confirm that this gene is restricted to strains of serotypes A and C, although its function has not been described to date We have previously described an additional gene, lgt5, present in all sero-types, that encodes an a-(1fi 4)-galactosyltrans-ferase [6]

Table 1 Strains used in this study, and plasmids used for mutagenesis.

Moraxella catarrhalis

Escherichia coli

DH5a /80 dLacZDM15 recA1 endA1 gyrA96 thi-1 hsdR17 (r k ,m k+) supE44 relA1

deoRD(lacZYA-argF)U169

Invitrogen Plasmid ⁄ vector

For mutation of lgt2

For mutation of lgt1 ⁄ 4

For mutation of lgt4 ⁄ 5

Table 2 Oligonucleotide sequence used to amplify and ⁄ or sequence the glycosyltransferase genes.

Oligonucleotide sequence

Forward or reverse with respect to orientation of submitted sequences Primers used to amplify and ⁄ or sequence lgt2

Primers used to amplify and ⁄ or sequence lgt1

Primers used to amplify and ⁄ or sequence lgt4 + lgt5

Structural analysis of the oligosaccharide

derived from single-mutant (2951lgt2D) and

double-mutant (2951lgt1⁄ 4D and 2951lgt5 ⁄ 4D)

strains of serotype A M catarrhalis

The genes lgt1, lgt2 and lgt5 were cloned from strain

CCUG 3292, and disrupted by insertion of Kanr into

convenient restriction sites within each ORF (Fig 1)

Linearized plasmid containing the mutant allele was

transferred into strain 2951 by natural transformation

Allelic replacement was confirmed by PCR

2951lgt2D

MS analysis of the LOS oligosaccharide showed that

mutation of the gene encoding Lgt2 results in a

trun-cated oligosaccharide as compared to the

oligosaccha-ride produced by the wild-type bacteria (Fig 2)

Negative-ion ESI-MS spectra for the 2951lgt2D

oligo-saccharide gave a molecular ion signal at m⁄ z 1251,

consistent with the calculated molecular mass for an

oligosaccharide of composition Hex5ÆHexÆNAcÆKdo

(1252 atomic mass unit) MALDI-MS analysis of the

methylated oligosaccharide (75% acetonitrile fraction

from Sep-pak C18 purification) yielded a molecular

ion signal at 1611 m⁄ z [M + Na]+, supporting this

assignment GC-MS sugar analysis [trimethylsilyl

(TMS) derivative] confirmed the presence of Glc,

Glc-NAc and Kdo, whereas GC-MS linkage analysis of the

permethylated sample identified terminal Glc, terminal

GlcNAc, 2-linked Glc, and 3,4,6-linked Glc (Table 3)

For each of the oligosaccharides studied, NMR

spectral assignment was aided by a combination of

one-dimensional and two-dimensional experiments,

including 1H, 13C-attached proton test (APT), COSY,

1H-13C-HSQC and 1H-13C-HSQC-TOCSY and edited

versions of these experiments Chemical shift

assign-ments for 2951lgt2D are given in Table 4 The

sequence of the sugar residues was confirmed by

exam-ination of 400 ms NOESY and1H-13C-HSQC-NOESY

experimental results For the 2951lgt2D

oligosaccha-ride, complete 1H chemical shift assignment of the

highly branched a-d-glucose residue (residue C) was

made possible by examination of the COSY spectra, as

many of the ring protons for this residue lie outside

the crowded 3.2–4.1 p.p.m region of the spectra For

the other hexose residues, the anomeric and H2

pro-tons could also be assigned using the COSY spectra

1H and 13C chemical shift assignments for the other

ring protons and carbons were possible using the

1H-13C-HSQC-TOCSY spectra in combination with

the13C-APT and1H-13C-HSQC spectra The anomeric

configuration for three of the six hexoses (residues A,

B, and C) could be confirmed by 3J1,2 coupling con-stants from a 1H-NMR spectrum However, 3J1,2 coupling constants could not be determined for the anomeric protons of residues D and E, due to overlap,

or for the anomeric proton of residue G, which over-laps with the signal from H5 of residue C Complete assignment of the Kdo residue was possible by exam-ination of TOCSY correlations with the well-dispersed H3 and H8 methylene protons of Kdo Glycosidic linkages for each sugar residue in the oligosaccharide were confirmed by examination of 400 ms NOESY and1H-13C-HSQC-NOESY spectra

2951lgt1⁄ 4D

An lgt1⁄ 4 double mutant was constructed by trans-forming the 2951 strain with the lgt13292::KAN con-struct As 3292 does not contain lgt4, this construct recombined in lgt1 and lgt5 (confirmed by PCR and sequencing), inactivating lgt1 as a result of the Kanr insertion, and also deleting lgt4 (Fig 1B): it was con-firmed that the construct had not illegitimately recom-bined in lgt2, as lgt2 was amplified and sequenced from strain 2951lgt1⁄ 4D using primers UORF2:2040 and DORF2:3120 and found to be identical to lgt22951 Mutation of the genes encoding Lgt1 and Lgt4 results in a very truncated oligosaccharide (Fig 2) as compared to the oligosaccharide produced by the wild-type bacteria Negative-ion ESI-MS spectra for the 2951lgt1⁄ 4D oligosaccharide gave a molecular ion at

m⁄ z 886 consistent with the composition Hex4ÆKdo MALDI-MS analysis of the methylated oligosac-charide (75% acetonitrile fraction from Sep-pak C18 purification) gave a molecular ion at m⁄ z 1161 [M + Na]+, supporting this assignment GC-MS sugar analysis confirmed the presence of Glc and Kdo The sample failed to give GC-MS linkage data; how-ever, the NMR data described below are unequivocal

1H and13C assignments for the 2951lgt1⁄ 4D oligosac-charide are given in Table 5 Chemical shift assignment for this oligosaccharide was relatively straightforward, due to the excellent signal dispersion and the reduced number of sugar residues, as shown in Fig 3 The chemical shift for each of the anomeric signals was sig-nificantly different from those chemical shifts for the same residue in the less truncated oligosaccharides 2951lgt2D and 2951lgt5⁄ 4D This has previously been noted for the synthetically prepared analog of the 2951lgt1⁄ 4D oligosaccharide [15]

Complete assignment of residue C, the central a-d-Glc residue linked to the Kdo and three b-d-Glc residues via (1fi 6), (1 fi 4) and (1 fi 3) glycosidic linkages, was achieved by examination of a COSY

spectrum, as shown in Fig 4 The anomeric

configur-ation of residue C was confirmed by the3J1,2coupling

constant of 3.95 Hz The sequential assignment of

resi-dues D, B and G, the three b-d-Glc resiresi-dues (each with

a 3J1,2 coupling constant of  8 Hz), was significantly aided by 1H,13C-HSQC-TOCSY and one-dimensional

Fig 1 Schematic representation of LOS biosynthetic locus of strains 3292 (serotype B) and 2951 (serotype A), including constructs for mut-agenesis Large open arrows represent ORFs that are highly similar between strains Filled regions of arrows represent sequence diversity between strains Small arrows represent oligonucleotide primers used for amplification (A) Construction of plasmid for mutation of lgt2: lgt2B⁄ Cand flanking sequence were amplified from strain 3292 using primers UORF2:2040 and DORF2:3120, and disrupted by insertion of kan r into the PstI site Transformation of strain 2951 resulted in allelic replacement (B) Construction of plasmid for mutation of lgt1 and dele-tion of lgt4: lgt1 and flanking sequence were amplified from strain 3292 using primers UORF3:4093 and DORF3:2768, and disrupted by insertion of kan r into the StyI site Transformation of strain 2951 resulted in recombination within lgt5 and lgt1, resulting in deletion of lgt4 and disruption of lgt1 (C) Construction of plasmid for mutation of lgt5 and deletion of lgt4: lgt5 and flanking sequence were amplified from strain 3292 using primers DORF4:5047 and UORF4:3684, and disrupted by insertion of kanrinto the NsiI site Transformation of strain 2951 resulted in recombination within lgt5 and lgt1, resulting in deletion of lgt4 and disruption of lgt5.

selective TOCSY experiments Although the anomeric

signals of each of the b-d-Glc residues were well

resolved, it was necessary to resort to one-dimensional

selective TOCSY experiments to assign the remaining

ring protons of these residues (D, B and G), which are

overlapped in both the 1H and 13C dimensions

Com-plete assignment of the Kdo residue was possible from

examination of TOCSY correlations with the

well-dispersed H3 and H8 methylene protons of Kdo An

APT experiment was used to obtain the 13C shift of the C1 and C2 resonances of Kdo Examination of the

400 ms NOESY spectrum confirmed each of the gly-cosidic linkages for the 2951lgt1⁄ 4D oligosaccharide and the terminal location of each of the b-d-Glc resi-dues D, B, and G As lgt1 has previously been shown

to encode an a-(1fi 2)-glucosyltransferase that adds residue A [4], we concluded that lgt4 encodes the a-(1fi 2)-N-acetylglycosyltransferase

Fig 2 Structures of wild-type serotype A

oligosaccharide (strain 2951) and 2951lgt2D,

2951lgt1 ⁄ 4D and 2951lgt5 ⁄ 4D mutant

oligo-saccharides Letters refer to designated

sugar residues.

2951lgt5⁄ 4D

In order to further investigate the function of lgt4, an

lgt4⁄ 5 double mutant was also produced To delete

lgt4 from strain 2951, lgt1 and lgt5 were amplified

from strain 3292, and Kanr was inserted into lgt5 As

strain 3292 does not contain lgt4, this construct

recom-bined in lgt1 and lgt5, inactivating lgt5 as a result of

the Kanrinsertion, and also deleting lgt4 (Fig 1C)

Mutation of the genes encoding Lgt4 and Lgt5

results in a truncated oligosaccharide, as shown in

Fig 2 Compared to the wild-type 2951

charide, or the recently published 2951lgt5D

oligosac-charide [6], it was immediately obvious from the

2951lgt5⁄ 4D oligosaccharide NMR spectra that the

a-d-GlcNAc residue on the (1fi 4) branch was missing,

due to the lack of peaks indicative of an N-acetamido

peak at 2.19 p.p.m (1H) and 25.4⁄ 177.0 p.p.m (13C)

MS sugar analysis supported this, indicating the

pres-ence of Glc, Gal and Kdo (in the approximate ratio

6 : 1 : 1) As expected, mutation of the lgt5 gene

resul-ted in a truncation of the (1fi 6) branch, and

conse-quently the b-d-Gal was found in the terminal position

because the a-d-Gal found in the wild-type 2951

oligo-saccharide was missing What was most surprising,

however, was that mutation of the lgt4 and lgt5

genes resulted in an oligosaccharide in which the

a-d-GlcNAc residue on the (1fi 4) branch was

replaced by an a-d-Glc residue

MALDI-MS analysis of the methylated

oligosaccha-ride gave a molecular ion at m⁄ z 1773.8 [M + Na]+

(75% acetonitrile fraction following Sep-pak C18

puri-fication), consistent with the composition Hex7ÆKdo

GC-MS sugar analysis (TMS derivative) indicated

an oligosaccharide composed of Glc, Gal, and Kdo

GC-MS linkage analysis of the permethylated sample

Table 3 GC-MS analysis of partially methylated alditol acetates

obtained from 2951lgt2D oligosaccharide (OS) and 2951lgt5 ⁄ 4D OS

from serotype A M catarrhalis following Sep-pak C18 purification

(75% acetonitrile fractions).

Sample

Elution

time

(min)

Characteristic fragment ions Assignment 2951lgt2D OS 18.52 118, 129, 145, 205 Terminal Glcp

19.68 129, 130, 161, 190 2-linked Glcp

21.98 118, 333 3,4,6-linked Glcp

22.37 117, 159, 205 Terminal GlcpÆNAc

2951lgt5 ⁄ 4D OS 18.52 118, 129, 145, 205 Terminal Glcp

18.82 118, 129, 145, 205 Terminal Galp

19.68 129, 130, 161, 190 2-linked Glcp

19.88 113, 118, 162, 233 4-linked Glcp

21.98 118, 333 3,4,6-linked Glcp

1 H

D2

1 H

3 J1,2

3 J1,2

3 J1,2

identified terminal Glc, terminal Gal, 2-linked Glc,

4-linked Glc, and 3,4,6-linked Glc (Table 3)

1H and 13C assignments for the 2951lgt5⁄ 4D

oligo-saccharide are given in Table 6 The anomeric

confi-guration for each sugar residue was obtained from

the 3J1,2 coupling constants, as shown in Table 6

Chemical shift assignment for this oligosaccharide

was more challenging than for the other

oligosaccha-rides, due to the poor dispersion of signals in the

3.2–4.1 p.p.m region of the spectra Although the 1H

signals of the anomeric protons of the 2951lgt5⁄ 4D

oligosaccharide were well dispersed, the 13C signals

were not, as can be seen in the anomeric region of

the 1H-13C-HSQC spectrum shown in Fig 5 In fact,

the 13C anomeric signal for residues D, G and H overlapped, as did those of E and B For this rea-son, it was necessary to perform a series of selective one-dimensional TOCSY experiments, irradiating each of the anomeric signals in turn to obtain the

1H chemicals shifts of the corresponding remaining ring protons Again, assignment of the Kdo residue was possible from examination of TOCSY correla-tions in the 120 ms 1H-13C-HSQC-TOCSY spectrum from the dispersed H3 and H8 methylene protons of Kdo, and an APT experiment was used to elucidate the 13C shift of the C1 and C2 resonance of Kdo For the highly branched, central a-d-Glc residue C, the locations of the H3⁄ C3 and H5 ⁄ C5 correlations were conspicuous in the 1H-13C-HSQC spectrum Examination of the anomeric 13C line of residue C

in the 120 ms 1H-13C-HSQC-TOCSY spectrum revealed the location of the remaining ring protons

Fig 3 Anomeric region of the 1H-NMR spectrum (600 MHz,

298 K, D2O) 2951lgt1 ⁄ 4D mutant OS Letters refer to designated

sugar residues as shown for 2951lgt1 ⁄ 4D in Fig 2.

C1-C2

C2-C3 C3-C4

C4-C5 H5-H6R/S H6R-H6S

C1

Fig 4. 1H,1H-COSY NMR spectrum (600 MHz, 298 K, D 2 O) 2951lgt1 ⁄ 4D mutant oligosaccharide.

Table 5 1 H and 13 C chemical shifts (p.p.m.) for oligosaccharides isolated from the M catarrhalis 2951lgt1 ⁄ 4D mutant in D 2 O referenced to DSS (0.0 p.p.m.), at 298 K, on a Bruker Avance spectrometer operating at 600 and 150 MHz, respectively.

Sugar residue

1

H Chemical shift (d, p.p.m.) 13C Chemical shift (d, p.p.m.)

(B) b- D -Glcp-(1 fi 4.68 3.33 3.49 3.38 3.44 3.90 3.71 7.9 103.8 75.8 78.4 72.2 78.4 63.5

(C) fi 3,4,6)-a- D -Glcp- 5.13 3.80 4.26 3.95 4.41 4.17 4.02 3.9 102.2 75.0 79.2 75.6 72.6 70.0

(D) b- D -Glcp-(1 fi 4.93 3.36 3.49 3.38 3.43 3.90 3.71 8.0 104.1 76.0 78.6 72.2 78.6 63.5

(G) b- D -Glcp-(1 fi 4.48 3.30 3.49 3.38 3.43 3.90 3.71 8.0 105.0 75.8 78.5 72.4 78.5 63.5

For each sugar residue, the 1H anomeric chemical

shift gave TOCSY correlations in the 120 ms 1H-13

C-HSQC-TOCSY spectrum with the 13C positions of

each of the ring protons, including C6 This

informa-tion, coupled with the selective one-dimensional

TOCSY experiments, was sufficient to structurally

assign the 2951lgt5⁄ 4D oligosaccharide The only residue

for which assignments (H5⁄ C5 and H6 ⁄ C6) remained

outstanding was the terminal b-d-gal, H Even from

the selective one-dimensional TOCSY experiment, it

was not possible to ascertain their assignment All

other assignments for the 2951lgt5⁄ 4D oligosaccharide

were in agreement with literature values (where

appro-priate) for the wild-type 2951 oligosaccharide [11–14]

Discussion Functional analysis of glycosyltransferase enzymes

In serotype B strains (3292) of M catarrhalis, Lgt1 catalyzes the addition of the a-d-Glc-(1 fi 2) glyco-sidic linkage to both the (1fi 6) and (1 fi 4) branches

of the growing oligosaccharide chain [4] This serotype lacks the lgt4 glycosyltransferase gene present in sero-type A and C strains The presence of the lgt4 gene

on the lgt locus of serotype A and C strains of

M catarrhalishas been noted [5], however, its function has not been determined Serotype A and C strains have the lgt4 gene and express GlcNAc on their LOS structures Serotype B strains lack this gene, and do not have GlcNAc as part of their LOS It was there-fore of interest to determine whether the lgt4 gene encoded an N-acetylglucosamine transferase In order

to ascribe a function to the product of this gene, we endeavored to produce mutant strains of M catarr-halis 2951 lacking the lgt4 gene, in order to ascertain from the degree of truncation the function of the Lgt4 glycosyltransferase Unfortunately, all attempts to pro-duce bacteria expressing lgt4D mutant oligosaccharide were unsuccessful

In an alternative approach to studying the function

of the Lgt4 glycosyltransferase, 2951lgt1⁄ 4D double-mutant bacteria were produced The mutational strat-egy for this double mutation took advantage of the absence of lgt4 in serotype B strains, in that the mutant alleles were constructed using 3292 (serotype B)-derived alleles that would delete lgt4 when introduced into a serotype A strain (Fig 1B) This strategy was also suc-cessfully employed to make another double mutation, 2951lgt5⁄ 4D (Fig 1C, and see below) The LOS from mutant bacteria was harvested, and the truncated oligo-saccharide examined by NMR and MS analysis The

Table 6 1 H and 13 C chemical shifts (p.p.m.) for oligosaccharides isolated from the M catarrhalis 2951lgt5⁄ 4D mutant in D 2 O referenced to DSS (0 p.p.m.), at 298 K, on a Bruker Avance spectrometer operating at 600 and 150 MHz, respectively ND, not determined.

Sugar residue

1

H Chemical shift (d, p.p.m.) 13C Chemical shift (d, p.p.m.)

(A) fi 4)-a- D -Glcp-(1 fi 5.42 3.63 3.87 3.66 4.10 3.76 3.99 4.2 99.7 74.0 75.7 81.0 73.2 62.9

(B) fi 2)-b- D -Glcp-(1 fi 5.06 3.44 3.57 3.41 3.60 3.70 3.82 7.2 101.3 81.8 77.9 72.5 77.9 63.2

(C) fi 3,4,6)-a- D -Glcp- 5.11 3.87 4.38 4.01 4.56 4.05 4.16 < 1 102.4 75.5 78.4 76.2 72.8 70.6

(D) b- D -Glcp-(1 fi 4.90 3.39 3.50 3.38 3.51 3.73 3.94 7.5 105.3 76.1 78.6 72.3 78.6 63.7

(G) fi 2)-b- D -Glcp-(1 fi 4.60 3.46 3.54 3.41 3.50 3.72 3.92 7.7 105.7 79.1 77.4 72.5 79.1 63.6

A

E

B

C

D

Fig 5 Anomeric region of the 1H,13C-HSQC NMR spectrum

(600 MHz, 298 K, D2O) 2951lgt5 ⁄ 4D mutant oligosaccharide Refer

to Fig 2 for letter designations.

2951lgt1⁄ 4D double-mutant oligosaccharide was

com-posed of a central a-d-Glc residue (1fi 6)-, (1 fi

4)-and (1fi 3)-linked to three b-d-Glc residues, as shown

in Fig 2 As mentioned previously, the presence of the

glucosyltransferase Lgt1 could account for the addition

of an a-d-Glc (1fi 2) glycosidic linkage to the (1 fi 6)

branch; however, in serotype A and C strains, there is

an a-d-GlcNAc residue with a (1fi 2) glycosidic

link-age on the (1fi 4) chain To further investigate the

function of Lgt4 and to explore the interrelationship

between the activity of the Lgt1 and Lgt4

glycosyl-transferase enzymes, a 2951Lgt5⁄ 4D serotype A mutant

was produced Lgt5 is the galactosyltransferase

respon-sible for the addition of a terminal a-d-Galp (1fi 4) to

the (1fi 6) branch of serotype A strain 2951 [6]

Dis-ruption of the genes that encode Lgt4 and Lgt5 would

ensure retention of Lgt1 glycosyltransferase activity,

and potentially produce truncated oligosaccharides

lacking the terminal a-d-Galp (1fi 4) on the (1 fi 6)

branch and the a-d-GlcNAc (1fi 2) glycosidic linkage

on the (1fi 4) branch i.e., an oligosaccharide

compri-sing six sugar units (not including Kdo) Fascinatingly,

NMR and MS examination of the oligosaccharide

iso-lated from the mutant M catarrhalis 2951lgt5⁄ 4D

revealed that an oligosaccharide containing seven

hexose sugar units was produced by these mutant

bac-teria This is clearly evident from the anomeric region

of the1H,13C-HSQC spectrum, as shown in Fig 5, and

the MALDI-MS methylation data Moreover, this

1H,13C-HSQC spectrum differed from that of the

oligo-saccharide produced by Lgt5D mutant M catarrhalis,

because the spectrum lacked an N-acetamido methyl

peak at 2 p.p.m that would have been indicative of a

GlcNAc being retained at the terminal position of the

(1fi 4) chain Additionally, the MS sugar analysis

clearly indicated the absence of the GlcNAc residue

Instead, in the absence of Lgt4 and Lgt5, a glucose

resi-due was added to the terminal position of the (1fi 4)

chain by Lgt1 This finding demonstrates that, in the

absence of a functional Lgt4, Lgt1 is able to add an

a-d-Glc (1fi 2) glycosidic linkage to the (1 fi 4) branch

Serotype B strains lack lgt4, and therefore have a

Glc at this position We and others [4–6] have

observed the presence of two different alleles of lgt2

In serotype A strains of M catarrhalis, Lgt2 (Lgt2A)

adds a b-d-Gal (1fi 4) to the a-d-Glc (1 fi 2)

glycosi-dic linkage added by Lgt1 (see Fig 2) to the (1fi 6)

branch In serotype B strains, however, Lgt2 (Lgt2B ⁄ C)

adds a b-d-Gal (1fi 4) to both the (1 fi 4) and

(1fi 6) branches This allele (Lgt2B ⁄ C) of lgt2, present

in serotype B and C strains of M catarrhalis,

corre-lates with the extension to the (1fi 4) branch

regard-less of whether the acceptor molecule is a terminal

glucose or N-acetylgalactosamine Our results suggest that the serotype A allele (Lgt2A) is unable to extend either the terminal glucose or N-acetylgalactosamine onto the (1fi 4) branch This observation implies that the serotype A Lgt2 has a higher acceptor specificity than that found in serotype B and C strains

Furthermore, 2951lgt1D M catarrhalis bacteria pro-duced LOS-derived oligosaccharide with the same degree of truncation as the 2951lgt1⁄ 4D double-mutant bacteria as determined by tricine SDS⁄ PAGE (data not shown) A possible explanation for this observa-tion is that it is necessary for Lgt1 to catalyze the addition of the a-d-Glc (1fi 2) glycosidic linkage to the (1fi 6) branch before Lgt4 can act by adding the a-d-GlcNAc (1fi 2) glycosidic linkage to the (1 fi 4) branch Such a requirement for addition of a hexose

to one chain before an enzyme can add to another has been reported for biosynthesis of LOS⁄ lipopoly-saccharide in other organisms; for example, Lic2C catalyzes the addition of glucose to the core HepII of

H influenzae, but requires that LgtF has added glucose

to HepI first [16]

Accordingly, we propose that Lgt4 is a N-acetyl-glucosylamine transferase responsible for the addition

of an a-d-GlcNAc (1 fi 2) glycosidic linkage to the (1fi 4) branch In the presence of an active Lgt4 (lgt4

is present only in serotype A and C strains), GlcNAc

is preferentially added to the (1fi 4) chain This implies that active Lgt4 has a much higher affin-ity⁄ specificity for the b-(1 fi 4)-linked Glc than does Lgt1 Competitive addition of hexoses has previously been reported in LOS⁄ lipopolysaccharide biosynthesis; for example, in pathogenic Neisseria strains, lgtA and lgtC encode an N-acetylglucosylamine transferase and

a galactosyltransferase, respectively In the presence of active LgtA and LgtC, GlcNAc is added by LgtA Only in the absence of LgtA can LgtC add Gal [17] From our experimental data and those of others, the role of Lgt1–5 in serotypes A and B has been con-firmed (Fig 6): the presence of Lgt3 is required for the addition of three glucosyl residues to the core GlcÆKdo, as mutation of lgt3 results in production of Kdo with only a single Glc residue [4] Lgt1 then adds a-(1fi 2)-Glc to the (1 fi 4) chain and also to the (1fi 6) chain in serotype B strains Lgt4 (present only

in serotypes A and C), then adds an a-(1fi 2)-GlcNAc to the (1fi 4) chain instead of Glc Lgt2A (in serotype A strains) and Lgt2B ⁄ C (prototypic isoform,

in serotype B and C strains) then adds b-(1fi 4)-Gal

to the (1 fi 6) chain In serotype B strains, the proto-typic Lgt2 also adds b-(1fi 4)-Gal to the (1 fi 4) chain Finally, Lgt5 adds an a-(1fi 4)-Gal to terminal Gal residues, when present This model is consistent