Báo cáo Y học: The opgGIH and opgC genes of Rhodobacter sphaeroides form an operon that controls backbone synthesis and succinylation of osmoregulated periplasmic glucans pot

The opgGIH and opgC genes of Rhodobacter sphaeroides forman operon that controls backbone synthesis and succinylation of osmoregulated periplasmic glucans Virginie Cogez1, Evgueni Gak2,

The opgGIH and opgC genes of Rhodobacter sphaeroides form

an operon that controls backbone synthesis and succinylation

of osmoregulated periplasmic glucans

Virginie Cogez1, Evgueni Gak2, Agnes Puskas2, Samuel Kaplan2and Jean-Pierre Bohin1

1

Unite´ de Glycobiologie Structurale et Fonctionnelle, CNRS UMR8576, Universite´ des Sciences et Technologies de Lille, France;

2 Department of Microbiology and Molecular Genetics, University of Texas HealthScience Center, Houston, TX, USA

Osmoregulated periplasmic glucans (OPGs) of Rhodobacter

sphaeroidesare anionic cyclic molecules that accumulate in

large amounts in the periplasmic space in response to low

osmolarity of the medium Their anionic character is

pro-vided by the substitution of the glucosidic backbone by

succinyl residues A wild-type strain was subject to

trans-poson mutagenesis, and putative mutant clones were

screened for changes in OPGs by thin layer

chromatogra-phy One mutant deficient in succinyl substitution of the

OPGs was obtained and the gene inactivated in this mutant

was characterized and named opgC opgC is located

downstream of three ORFs, opgGIH, two of which are

similar to the Escherichia coli operon, mdoGH, governing

OPG backbone synthesis Inactivation of opgG, opgI or

opgH abolished OPG production and complementation analysis indicated that the three genes are necessary for backbone synthesis In contrast, inactivation of a gene similar to ndvB, encoding the OPG-glucosyl transferase in Sinorhizobium meliloti, had no consequence on OPG syn-thesis in Rhodobacter sphaeroides Cassette insertions in opgHhad a polar effect on glucan substitution, indicating that opgC is in the same transcription unit Expression of opgIHCin E coli mdoB/mdoC and mdoH mutants allowed the production of slightly anionic and abnormally long linear glucans

Keywords: periplasm; osmoregulation; cyclic glucans; glucosyl transferase; operon

Osmoregulated periplasmic glucans (OPGs) are found in

the periplasmic space of proteobacteria [1] These

oligosac-charides exhibit quite different structures among various

species but they share four common characteristics: (a) a

small size, with a degree of polymerization (DP) in the range

of 5–24; (b) D-glucose being the only sugar unit; (c)

b-glucosidic bonds being the main type of linkages; (d) the

periplasmic concentration increasing in response to a

decrease of environmental osmolarity Four families of

OPGs are described on the basis of structural features of the

polyglucose backbone [1]: family I, heterogeneously sized

linear and branched b-1,2;b-1,6 glucans; family II,

hetero-geneously sized cyclic b-1,2 glucans; family III,

homogen-eously sized cyclic and branched b-1,3;b-1,6 glucans;

family IV, homogeneously sized cyclic b-1,2;a-1,6 glucans

In several bacterial species, OPGs are substituted by one or

several of a series of different residues, originating from

either the membrane phospholipids (phosphoglycerol,

phosphoethanolamine, and phosphocholine) or from inter-mediate metabolism (acetyl, succinyl, and methylmalonyl) Thus, depending on the bacterial strain and growth conditions, OPGs can be found unsubstituted, neutral or anionic

The function of OPGs in the bacterial envelope remains obscure However, mutants defective in OPG synthesis have

a highly pleiotropic phenotype, indicative of an overall alteration of their envelope properties When some bacteria interact with a eucaryotic host, as pathogens or symbionts, mutants defective in backbone synthesis are partially or completely impaired in this interaction [1] This is the case for mutants of Agrobacterium tumefaciens, Pseudomonas syringae, Bradyrhizobium japonicum, P aeruginosa, Erwinia chrysanthemiand Brucella abortus [2–8] One highly attenu-ated Salmonella (enterica) typhimurium mutant resulted from a transposon insertion in the mdoB gene known to govern OPG substitution by phosphoglycerol in Escherichia coli[9] In contrast, a S meliloti mutant impaired in OPG substitution by phosphoglycerol effectively nodulated alfalfa [10], but the anionic character of these OPGs was more or less retained by an increase of succinyl substitution Obviously, further genetic analyses of different model organisms are needed to understand the OPG function(s) Rhodobacter sphaeroides is a free-living photohetero-trophic bacterium of the alpha subdivision of the proteo-bacteria, whose genome is composed of two distinct circular chromosomes [11] Genetic analysis is highly developed in this organism, which is a model for the study of bacterial photosynthesis [12] R sphaeroides produces OPGs belong-ing to family IV They mainly consist of a cyclic glucan homogenous in size (DP

Correspondence to J.-P Bohin, CNRS UMR8576, Baˆt.C9, U.S.T.L.,

59655 Villeneuve d’Ascq Cedex, France.

Fax: + 33 3 20 43 65 55, Tel.: + 33 3 20 43 65 92,

E-mail: Jean-Pierre.Bohin@univ-lille1.fr

Abbreviations: DP, degree of polymerization; SIS, Sistrom’s succinic

acid minimal medium; LOS, low-osmolarity medium; Amp,

ampicillin; Kan, kanamycin; Rif, rifampicin; Spc, spectinomycin;

Str, streptomycin; Tet, tetracycline; Tmp, trimethoprim; Cml,

chloramphenicol; MP, maximum-parsimony; TMS, transmembrane

segment; ACP, acyl carrier protein.

(Received 19 October 2001, revised 11 March 2002,

accepted 22 March 2002)

are linked by b-1,2 linkage and one glucose unit is linked

by a-1,6 linkage This backbone is substituted to various

degrees by two kinds of residues: O-acetyl residues (0–2 per

mol) and O-succinyl residues (1–7 per mol) that confer a

highly anionic character to these OPGs [13] Because

R sphaeroidesshows a close relationship to those organisms

that share an interactive lifestyle with a eucaryotic host, but

itself does not, it represents a model organism in which to

study OPG synthesis

Our initial purpose was to obtain OPG defective mutants

by screening a transposon insertion library with a thin layer

chromatographic assay [14] A single mutant clone was

detected that produced neutral OPGs Actually, this

partic-ular phenotype allowed the demonstration of acetyl

substi-tution of the OPGs [13] The transposon insertion was

located just downstream of a series of genes similar to the

E coli mdoGHoperon that govern the synthesis of OPGs

belonging to family I In this paper, we describe the

molecular and functional characterization of these genes,

opgGIH,necessary for OPG backbone synthesis and of a

new gene, opgC, necessary for OPG succinylation

M A T E R I A L S A N D M E T H O D S

Bacterial strains and media

The bacterial strains and plasmids used are detailed in

Table 1 R sphaeroides strains were grown at 30C in

Sistrom’s succinic acid minimal medium (SIS; [22]);

anaero-bically, illuminated at 100 WÆm)2 To determine OPG

production, aerobic chemoheterotrophic cultures were

grown, with shaking, in Luria–Bertani broth [23] E coli

strains were grown at 37C in Luria–Bertani When low

osmolarity medium was required, Luria–Bertani without

NaCl or low-osmolarity medium (LOS; [15]) was used

Solid media were obtained by adding agar (15 gÆL)1)

Antibiotics were added to the medium at the following

concentrations: ampicillin (Amp), 100 lgÆL)1; kanamycin

(Kan), 25 lgÆL)1; rifampicin (Rif) 100 lgÆL)1;

spectino-mycin (Spc), 50 lgÆL)1; streptomycin (Str), 50 lgÆL)1;

tetra-cycline (Tet), 1.0 lgÆL)1; trimethoprim (Tmp), 50 lgÆL)1for

R sphaeroidesand ampicillin, 50 lgÆL)1; chloramphenicol

(Cml), 25 lgÆL)1; kanamycin, 50 lgÆL)1; tetracycline,

10 lgÆL)1; trimethoprim, 50 lgÆL)1for E coli

Transformation and mating

E colicells were made competent and transformed using

the rubidium chloride technique [24] The broad host range

plasmids (originating from pLA2917, pRK415 or pSUP202)

were mobilized from S17-1 into R sphaeroides strains

Matings were performed on nitrocellulose filters laid on

Luria–Bertani plates and the exconjugants selected on Tet,

Tmp or Kan Luria–Bertani plates, or SIS Str+Spc or SIS

Kan plates

Transposon mutagenesis

The mobilizable suicide plasmid pSUPTn5TpMCS [25,26]

was introduced into R sphaeroides WS8 by mating at 30C

with E coli S17-1 and spread onto Luria–Bertani Tmp Str

plates After a 3-day incubation, clones of R sphaeroides

containing transposon insertions (conferring trimethoprim

resistance) were picked and arranged to construct libraries

of putative mutants

Thin-layer chromatographic screening method Each Tn5TpMCS mutant generated from WS8 was screened for production of OPGs as described previously:

4 mL of an overnight culture in Luria–Bertani without NaCl were treated to give a 15-lL extract in water [14] Samples of 5 lL were analyzed by chromatography on aluminium silica gel 60 plates (Merk) in ethanol/butanol/ water (5 : 5 : 4) solvent Glucans were revealed by spray-ing dried plates with 0.2% orcinol in 20% sulfuric acid followed by heating at 110C The same procedure was used for rapid determination of the OPG synthesis by the various mutants obtained from strain 2.4.1 This proce-dure was poorly quantitative and used only to checkthe presence or absence of OPGs and their anionic or neutral character

DNA purification, restriction and modification enzymes and ligase

Standard procedures [27] were used for large scale plasmid isolation and rapid analysis of recombinant plasmids Genomic DNA extraction was done as described by Davis

et al [28] Restriction endonucleases (Eurogentec or Gibco BRL), the large (Klenow) fragment of DNA polymerase I and T4 DNA ligase (Gibco BRL) were used according to manufacturer’s recommendations

In vitro construction of plasmids For the sequencing of a fragment of the opgC gene, a genomic DNA fragment of the opgC1::Tn5TpMCS strain NFB4000 was cloned in the EcoRI site of plasmid pUC19 This restriction enzyme cut once in the Tn5TpMCS DNA outside the gene conferring trimethoprim resistance to the transposon Trimethoprim-resistant clones harboring a plasmid containing a 6-kb DNA insert were isolated on plates containing trimethoprim and ampicillin In this plasmid, called pNFR2, a translational fusion occurred fortuitously between the eighteenth codon of the a-lacZ fragment present in the vector and the third codon of opgI, while opgH was intact and opgC inactivated

For complementation tests in R sphaeroides, a 4.5-k b SalI fragment from cosmid pUI8166 [21], containing an intact copy of opgC (opgH being truncated), was inserted into the SalI site of pUC19 to give pNFR12 The 4.5 kb fragment was then liberated by digesting pNFR12 with HindIII and KpnI and inserted into the broad host rang mobilizable vector pRK415 [18] digested with HindIII and KpnI, to give pNFR13 (Fig 3)

A 2.2-k b EcoRV–BglII fragment from pUI8166, con-taining an intact copy of opgH, was inserted into Litmus

28 (Ampr; New England Biolabs), to give pGAK115 Then, the 2.2-kb fragment was liberated by digesting pGAK115 with HindIII and BglII, and inserted into pRK415 digested with HindIII and BamHI, to give pGAK136 (Fig 3)

A 3.5 SalI fragment from cosmid pUI8166, containing an intact copy of opgGI, was inserted into the SalI site of pBS II SK(+) (Ampr; Statagene), to give pUI2509 Then, the

3.5 kb was liberated by digesting pUI2509 with HindIII and

KpnI, and inserted into pRK415 digested with HindIII and

KpnI, to give pGAK247 (Fig 3)

A 1.2-k b BglII–KpnI fragment from pUI2509, containing

an intact copy of opgI, was inserted into the BamHI site of

pRK415, to give pGAK135 (Fig 3)

A 1.8-k b BamHI fragment from pUI2509, containing an

intact copy of opgI, was inserted into the BamHI site of

pRK415, to give pGAK245 (opgI in the tet promoter

orientation) and pGAK246 (opgI opposite to the tet

promoter orientation; Fig 3)

A 5-k b ApaI–BalI fragment from cosmid pUI8166,

containing an intact copy of opgGIH (opgC being truncated),

was blunt ended and inserted into the SmaI site of pUC19, to

give pNFR14 A 5-kb fragment was liberated by digesting

pNFR14 with HindIII and KpnI, and inserted into pRK415

digested with HindIII and KpnI, to give pNFR21 (Fig 3)

A 2.0-k b HindIII–EcoRV fragment from pNFR14 was

inserted between the HindIII and SmaI sites of pUC19 to

give pNFR15 A 2.0-kb fragment was liberated by digesting pNFR15 with HindIII and KpnI, and inserted into pRK415 digested with HindIII and KpnI, to give pNFR20 (Fig 3)

A 1.8-k b StuI–SmaI fragment from pNFR12, containing opgC, was inserted into the SmaI site of pUC19 to give pNFR18 A 1.8-kb fragment was liberated by digesting pNFR18 with KpnI and SmaI, and inserted into the expression vector pYZ4 to give pNFR25 (Fig 3)

A 3-k b EcoRI fragment from pNF14, containing opgIH, was inserted into the EcoRI site of pUC19 in the same orientation as a-lacZ, to give pNFR35 (Fig 3) A 2.2-kb HindIII–BglII from pNFR35 and a 1.2-kb BglII–EcoRI from pNFR25 were ligated and inserted into the site created

by digestion with HindIII and EcoRI of pUC19, to give pNFR37 (Fig 3) This plasmid is similar to pNFR2 except that opgC is now intact

A 4-k b EcoRI–HindIII fragment from pNFR2 was blunt ended and inserted into the SmaI site of pYZ4, to give pNFR30

Table 1 Bacterial strains and plasmids used in this study.

Escherichia coli

S17-1 thi pro hsdR–hsdM+recA RP4 plasmid integrated Tc::Mu-Km::Tn7 [15]

DH5a F–recA1 endA1 gyrA96 thi-1 hsdR17 glnV44 relA1 DlacU169 k (/80 dlacZDM15) Lab stock

NFB216 D(lac-pro) ara mdoH200::Tn10 pyrC46 rpsL thi (/80 dlacZDM15) [16]

NFB702 D(lac-pro) ara mdoG202::neo pyrC46 rpsL thi (/80 dlacZDM15) [16]

NFB1933 his pgi::Mu D(zwf-edd)1 eda-1 rpsL mdoB214::Tn10 mdoC1::Tn5 [14]

Rhodobacter sphaeroides

Plasmids

pUI8166 pLA2917-derived cosmid clone from R sphaeroides 2.4.1Tlibrary [21]

pGAK245 Tet r , pRK415 carrying opgI + (in the tet promoter orientation) This work pGAK246 Tetr, pRK415 carrying opgI+(opposite of the tet promoter orientation) This work

Construction of mutants

A 2.6-k b EcoRI fragment from pUI2509 was inserted into a

pBS II SK(+) derivative in which the BamHI was filled in,

to give pUI2511 The W interposon was liberated by

digesting pHP45W with BamHI and inserted in the unique

BamHI of pUI2511 The construct, containing opgG

disrupted 244 bp from its predicted start codon, was

subcloned as a 4.6-kb EcoRI fragment into the EcoRI site

of the broad host rang mobilizable vector pSUP202 [15] to

give pUI2513 pUI2513 was mobilized from S17-1 into

R sphaeroides 2.4.1 Exconjugants were selected on SIS

Str+Spc plates and the structure of the Tet-sensitive clones

was confirmed by Southern hybridization [29]

A 1.3-k b Eco47III fragment was deleted from pUI2509

to give pGAK112 Then, a 2.1-kb SalI fragment from

pGAK112 was inserted into the SalI site of pUC19 to give

pGAK114 A kan cassette was liberated as a 1.2-kb SalI

fragment from pUC4K (Pharmacia), and inserted (in both

orientations) into the unique XhoI site of pGAK114 The

two resulting constructs, containing opgI disrupted 19 bp

from its predicted start codon, were subcloned as 3.8 kb

AatII-Eco47III fragments between the ScaI and AatII sites

of the broad host range mobilizable vector pSUP202D

(M Gomelsky, Department of Microbiology and

Molecu-lar Genetics, University of Texas Health Center, Houston,

Texas, USA)

2 to give pGAK118 (kan and opgI in the same

orientation) and pGAK119 (kan and opgI in opposite

orientations) The two plasmids were mobilized into

R sphaeroides 2.4.1 Exconjugants were selected on SIS

Kan plates and the structure of the Tet-sensitive clones was

confirmed by Southern hybridization

The 1.2 kb SalI fragment, containing the kan cassette

described above, was inserted (in both orientations) into the

unique SalI site of pGAK115 The two resulting constructs,

containing opgH disrupted 609 bp from its predicted start

codon, were subcloned as 3.5 EcoRV–SnaBI fragments into

the ScaI site of pSUP202D to give pGAK120 (kan and

opgH in the same orientation) and pGAK121 (kan and

opgH in opposite orientations) The two plasmids were

mobilized into R sphaeroides 2.4.1 Exconjugants were

selected on SIS Kan plates and the structure of the Tet

sensitive clones was confirmed by Southern hybridization

A 9.5-k b SacI–KpnI cosmid DNA fragment mapping

to contig 12 of the R sphaeroides 2.4.1 genome (http://

mmg.uth.tmc.edu/sphaeroides/) was inserted into Litmus 28

to give pGAK227 This DNA sequence contains a portion

of the ndvB gene beginning 569 nucleotides downstream of

the purported ndvB start codon The W interposon was

liberated by digesting pHP45W with SmaI and combined

with the 8.7 kb MscI fragment of pGAK227 to give

pGAK232 A 7.9-kb EcoRI fragment from pGAK232 was

subcloned into pSUP202 (pGAK238) and mobilized into

R sphaeroides 2.4.1 Exconjugants were selected on SIS

Str+Spc plates and the structure of the Tet-sensitive clones

was confirmed by Southern hybridization

DNA sequencing

Small scale DNA sequencing was carried out with the

Sequenase version 2.0 kit (USB corporation) except

for the Tn5 insertion point where the oligonucleotide

5¢-CATGGAAGTCAGATCCTGG-3¢ (Eurogentec),

cor-responding to the end of both IS50 delineating Tn5 was used as a primer

The opgGIHC DNA sequence was determined by primer walking and performed at the DNA Core Facility of the Department of Microbiology and Molecular Genetics (University of Texas Health Science Center, Houston, Texas, USA), on an ABI 377 automatic DNA sequencer using the Big Dye terminator sequencing kit (Perkin-Elmer, Applied Biosystem Division) Gibco BRL and Integrated DNA Technology synthesized custom primers Assembly and analysis of the DNA sequences were performed using the DNA Strider (Institut de Recherche Fondamentale, Commisariat a` l’Energie Atomique, Paris, France),PHRED AND PHRAP(CodonCode Corporation) andGCG(Genetics Computer Group, Wisconsin Package, Madison, Wiscon-sin, USA) softwares The opgGIHC nucleotide sequence has been deposited in GenBankunder accession no AF016298 The DNA sequences and deduced amino-acid seq-uences were analyzed by using computer programs and sequence data made freely available from Infobiogen (http://www.infobiogen.fr/) and from ERGO (http://wit integratedgenomics.com/IGwit/CGI/)

A preliminary alignment of the full-length sequences of MdoH homologues was generated by CLUSTAL W, using default gap penalties TheCLUSTAL W alignment was then refined by manually deleting N- and C-terminal noncon-served sequences However, the P1 domain [30], which is almost absent in a series of MdoH homologues, was considered as phylogenetically relevant and included (in MdoH, 637 out of the 847 amino acids were thus considered) Phylogenetic trees were constructed by using maximum-parsimony (MP) and neighbor-joining methods The MP analyses used the programPROTPARSimplemented

in PHYLIP(Phylogeny Inference Package, Joe Felsenstein, Department of Genetics at the University of Washington) ThePHYLIPprograms SEQBOOT, PROTPARS, and CONSENSE were used sequentially to generate an MP tree that was replicated in 100 bootstraps; on this basis bootstrap confidence levels were determined

Analysis of OPGs fromR sphaeroides Cultures (100–500 mL) of R sphaeroides were grown overnight in Luria–Bertani without NaCl After 20 min centrifugation at 10 000 g, OPGs were extracted by 70% ethanol from the cell pellets The extracts were concentrated

by rotary evaporation, and lipids and proteins were then removed by the addition of a mixture of chloroform and methanol (2 : 1) The aqueous phase, containing OPGs, was chromatographied on a Biogel P4 column (Bio-Rad) The column (1.5 cm in diameter, 68 cm in height) was equili-brated with acetic acid 0.5% and eluted at a rate of

15 mLÆh)1in the same buffer Fractions (1.5 mL) contain-ing OPGs were pooled, concentrated by rotary evaporation, desalted on a Biogel P2 column (Bio-Rad), and fractions containing OPGs were pooled and lyophilized Sugar content was determined colorimetrically by using the anthrone-sulfuric acid reagent procedure [27]

Analysis of OPGs fromE coli Strain NFB1933 and its derivatives were grown in LOS medium (5 mL) supplemented with 0.24 m -[U-14

C]glu-cose (125 MBqÆmmol)1) OPGs were extracted by the

char-coal adsorption procedure [15] Pyridine extract obtained by

this procedure was chromatographied on a Biogel P4

column (Bio-Rad) and then on DEAE-Sephacel column

(Pharmacia)

Determination of neutral and anionic characteristics

of OPGs

OPGs were desalted on a PD10 column (Pharmacia) and

equilibrated with a Tris/HCl 10 mMpH 7.4 buffer

OPG-containing fractions were pooled and chromatographied on

a DEAE-Sephacel (Pharmacia) column (1.5 cm in diameter,

38 cm in height) equilibrated with Tris/HCl 10 mMpH 7.4

and eluted with the same buffer containing increasing

concentrations of NaCl ranging from 0 to 0.2Mby steps of

0.05M A volume of 60 mL was used for each NaCl

concentration and the volume of each collected fraction was

4 mL

Matrix-assisted laser desorption-ionization (MALDI)

mass spectrometry (MS)

To remove all substituents, 250 lg glucose equivalent of

lyophilized OPGs from E coli were dissolved in 100 lL

of fluorohydric acid (HF) and left for 60 h at 4C

OPGs were then neutralized by the addition of seven

volumes of saturated lithium hydroxide (LiOH) solution

The LiF precipitate was separated by centrifugation and

washed several times The different supernatants were

pooled and neutralized with AG 50 W-X8 (H+ form,

Bio-Rad), and then desalted on a Biogel P2 column

(Bio-Rad) Fractions containing OPGs were pooled and

lyophilized

For removal of the succinyl and acetyl substituents,

OPGs from R sphaeroides were de-esterified in 0.1MKOH

at 37C for 1 h After neutralization with AG 50 W-X8

(H + form, Rad), the samples were desalted on a

Bio-Gel P-2 column

The matrix used for carbohydrate analysis was

3-amino-quinolin (10 gÆL)1 in water; [13]) Lyophilized

oligosac-charides samples were redissolved in doubly distilled water

and then diluted with an appropriate volume of the matrix

solution (1 : 5, v/v) One microliter of the resulting solution

was deposited onto a stainless steel target, and the solvent

was evaporated under gentle stream of warm air

The experiments were carried out on a VISION 2000

(Finnigan MAT) time-of-flight mass spectrometer, as

pre-viously described [13]

R E S U L T S

Isolation of a mutant with a OPG altered phenotype

A random Tn5TpMCS mutagenesis was performed in

R sphaeroides WS8 and glucans extracted from 436

random Tn5TpMCS insertion mutants were analyzed by

thin layer chromatography to find a clone (Fig 1, lane 2)

whose OPGs showed a slower migration when compared to

the wild type (Fig 1, lane 1) An isolated clone was called

NFB4000

Mild alkali treatment of OPGs removes substituents

attached to the glucan backbone by O-ester linkages As

expected, treated wild-type OPGs (Fig 1, lane 3) exhibited

a reduced migration similar but not identical to that of mutant OPGs (Fig 1, lane 2) To observe identical migration between the two types of OPGs, mild alkali treatment of the mutant OPGs was necessary (Fig 1, lane 4) The growth rates and growth yields of the two strains grown in Luria–Bertani without NaCl were identical Accurate measurements showed that their OPG levels were also identical (22 ± 2 lg of glucose equivalent per

mg of cell protein), indicating that the observed change was not the consequence of a reduction in glucose backbone synthesis

OPGs isolated from the mutant are neutral OPGs produced by the mutant and wild-type strains were further analyzed by DEAE-Sephacel chromatography, which allows for the separation of subfractions of glucan

by their anionic character OPGs extracted from WS8 were separated into five main subfractions eluted at increasing NaCl concentration higher than 100 mM, showing their highly anionic character (Fig 2) OPGs produced by wild-type R sphaeroides are essentially homogeneous in size with

a degree of polymerization of 18 glucose residues [13] These are substituted to various degrees with succinyl residues that

Fig 1 Thin layer chromatographic analysis of OPGs extracted from WS8 (wild type, lanes 1 and 3) or NFB4000 (opgC1::Tn5TpMCS, lanes

2 and 4) Extracts (see Materials and methods) were applied directly to thin-layer chromatography plates (lanes 1 and 2) or first subjected to mild alkali treatment to remove substituents attached by O-ester linkages (lanes 3 and 4) Arrows on the left side indicate the position of three different levels of OPG substitution.

are negatively charged at pH 7.4, and acetyl residues that

are neutral Thus, one could expect that each subfraction

separated by DEAE-Sephacel corresponded to an

increas-ing number of succinyl residues The second, third and

fourth subfractions were collected separately, desalted and

analyzed by MALDI-mass spectrometry This analysis

revealed that these subfractions were still heterogeneous

with different degrees of substitution by succinyl residues

(data not shown) Thus, we must consider that each

subfraction corresponded to a different charge-to-mass

ratio due to various levels of substitution by succinyl and

acetyl residues

OPGs extracted from NFB4000 were not adsorbed on

the DEAE-Sephacel column, showing their neutral

charac-ter (Fig 2) An accurate structural analysis revealed that the

OPGs from NFB4000 lacked succinyl residues but remained

substituted by acetyl residues [13] The gene interrupted by

the Tn5TpMCS insertion was called opgC, by analogy with

the mdoC gene that governs OPG succinylation in E coli

[14]

opgC lies downstream from a locus similar

tomdoGH of E coli

The opgC1::Tn5TpMCS mutation was cloned into the

pUC19 vector from genomic DNA using the trimethoprim

resistance conferred by the transposon as a selection

Several clones were obtained that contained plasmids with the same 6 kb insert in one or the other orientation The DNA sequence of this 6 kb insert was determined using a primer from within the IS50 DNA (see Materials and methods) This sequence was compared with the avail-able sequence data from R sphaeroides 2.4.1 (http:// mmg.uth.tmc.edu/sphaeroides/) and found 99% identical with an ORF present in the vicinity of cerRI, a locus present

on chromosome I and governing the synthesis of an acylhomoserine lactone signal [21] Previous sequencing of cosmid pUI8166 [21] revealed the presence upstream from opgCof three ORFs, two of which (named opgG and opgH) are very similar to the genes mdoG and mdoH of E coli (Fig 3) In this organism, mdoG and mdoH form an operon under osmotic control that governs the synthesis of linear OPGs [16] When pUI8166 was introduced into the mutant strain NFB4000, restoration of the anionic character of the OPGs was observed by thin layer chromatography (Table 1) The opgC gene was further subcloned as a 4.5-kb SalI fragment in plasmid pNFR13 that still complements the NFB4000 defect

Analysis of nucleotide sequence of opgGIHC

ofR sphaeroides The first ORF, opgG, encodes a 540-amino-acid polypep-tide OpgG starts with an AUG and no alternative initiation codon (GUG or UUG) is found in its vicinity Analysis of the first 70 amino acids with theSIGNALPprogram (http:// www.cbs.dtu.dk/services/SignalP/) allowed the prediction

of a 38-amino-acid signal peptide and of a 502-amino-acid mature protein This protein is 40% identical and 58% similar to the mature MdoG protein

The second ORF was named opgI because it appeared to

be necessary to OPG backbone synthesis (see below) It overlaps the opgG stop codon (TGATG) and is predicted to encode a 66-amino-acid polypeptide No similarity was detected between OpgI and sequences available in the databases, except with an ORF conserved in the opgGIHC locus of R capsulatus This locus is 64% identical to its

R sphaeroidescounterpart over 5070 nucleotides Thus, this observation strengthens that hypothesis that the existence of opgIis not the result of a sequencing error

The third ORF, opgH, starts five nucleotides after the opgIstop codon and encodes a 595-amino-acid polypeptide Thus, OpgH is shorter than MdoH (847 amino acids) MdoH consists of three large cytoplasmic domains separ-ated by eight transmembrane segments (TMS); the topology

is N-terminal, two TMS, central, six TMS, C-terminal [30] For both proteins, the TOPPRED2 program (http:// bioweb.pasteur.fr/seqanal/interfaces/toppred.html) predic-ted seven TMS while the eighth segment was demonstrapredic-ted experimentally for MdoH [30] Thus, OpgH exhibits the same organization as MdoH with the major difference that the N-terminal domain is almost absent and the C-terminal domain is much shorter Finally, within the conserved regions, OpgH and MdoH are 40% identical and 70% similar

Tn5TpMCS was found to be inserted 747 bp down-stream from the putative start codon of the fourth ORF, opgC, encoding a putative 399-amino-acid polypeptide OpgG and OpgH present a high degree of sequence similarities to MdoG and MdoH of E coli In contrast,

Fig 2 DEAE-Sephacel anion exchange column chromatography

pro-files of OPGs from strains WS8 (Top) and NFB4000 (Bottom) 1200

(WS8) and 400 (NFB400) lg of glucose equivalent were loaded on the

column Ionic strength was increased by steps of 0.05 M NaCl at the

fractions indicated by the arrows Fractions (4 mL) were collected and

sugar content was determined colorimetrically (see Materials and

methods) The concentration of each fraction is indicated as percent of

the total fractions.

while OpgC and MdoC have similar sizes (399 and 385

amino acids, respectively), they do not show any significant

sequence similarities However, OpgC and MdoC exhibit

stretches of hydrophobic amino acids over their entire

length For MdoC [14], 10 TMS have been predicted by the

TOPPRED2 program For OpgC, the same program allowed

the prediction of 11 TMS

Two inverted repeats of 12 bp (CGAAGGCACCCTC

ACGGGTGCCCTTGG) followed by TCGTTT are found

144 nucleotides downstream from the stop codon of opgC

This could be an intrinsic transcription terminator as no

sizeable ORF starts before this point

OpgG, OpgI, and OpgH are necessary to OPG backbone

synthesis, but not NdvB

A locus similar to ndvB, the gene governing OPG synthesis

in S meliloti, has been partially described previously on

chromosome II [31] Thus, the question was to determine

whether ndvB, or opgGIH, or both are necessary for OPG

synthesis in R sphaeroides Therefore, each of the four

genes were inactivated separately in R sphaeroides 2.4.1 and

the resulting strains subjected to OPG analysis (see

Mate-rials and methods)

None of the mutants exhibited any particular phenotype

on plate when compared to the wild-type strain 2.4.1

Growths of the various strains were compared in four

different liquid media (SIS, Luria–Bertani, Luria–Bertani

without NaCl, and LOS; see Materials and methods)

No differences were observed in the growth rates and the

growth yields of these cultures

When OPGs were extracted from the ndvB mutant and

analyzed by thin layer chromatography, spots

correspond-ing to anionic OPGs were observed These OPGs were

further purified, deesterified, and analyzed by MALDI-MS

The resulting spectra (Fig 4) revealed the presence of

one quasimolecular ion with the calculated mass for an

[M + Na]+ ion based on an unsubstituted 18-member cyclic glucan The glucan produced seems to be mostly homogeneous in size, and only minor species corresponding

to cyclic glucans composed of 16, 17, 19, 21, 22, 23, and 24 glucose residues are also present (Fig 4) These data allowed two major conclusions: (a) NdvB is not necessary for OPG synthesis in R sphaeroides; (b) OPG structures are identical amongst various strains of R sphaeroides as identical spectra were obtained for OPGs extracted from strains 2.4.1 and WS8 [13] Similar results were previously observed for different strains of X campestris [13] When extracts from the opgG, opgI, or opgH mutants were analyzed by thin layer chromatography, no spots corresponding to OPGs could be detected These results

Fig 3 Restriction map of a 8-kb fragment

present in cosmid pUI8166 and its derivatives.

Arrows indicated ORFs of opgGIHC.

Horizontal bars indicate the structure of the

various inserts of the relevant plasmids.

Fig 4 Positive-ion MALDI mass spectra of OPGs extracted from the

R sphaeroides 2.4.1 derivative EG238 (ndvB) Mass assignments are based on an external calibration The number on the top of each peak refers to the degree of polymerization of glucose residues.

were confirmed by Biogel P4 chromatography of extracts

obtained from 100-mL cultures Complementation analysis

were performed by introducing various plasmids into each

of the mutants tested (Table 2)

The opgG mutation was complemented with a plasmid

containing only the putative promoter and opgG The W

insertion had no polar effect on downstream gene

expres-sion One can suggest the presence of a secondary promoter

inside the opgG ORF as previously observed in E coli and

E chrysanthemi[7]

The opgI mutations behaved differently according to the

orientation of the kan cassette When the cassette was in the

same orientation as the opg genes, the opgI mutation could

be complemented by a plasmid containing only opgI and the

putative secondary promoter When the plasmid contained

a shorter sequence upstream of opgI, complementation was

not possible This allowed a more accurate localization of

this promoter between the BamHI and BglII sites of opgG

When the cassette was in the opposite orientation,

comple-mentation was not observed, indicating a polar effect on

opgHexpression

opgC is cotranscribed with opgH

The putative start codon of opgC is located five nucleotides

before the stop codon of opgH, which strongly suggests

cotranscription Actually, when opgH was disrupted by a

cassette in the same orientation (EG18) the mutation could

be complemented with only opgH and the OPGs produced

had an anionic character When the cassette was in the

opposite orientation (EG7), the OPGs produced were

neutral When cosmid pUI8166 was introduced into the

EG7 and EG18 strains, both synthesis and anionic

substi-tution of the OPGs were restored whatever the cassette

orientation Therefore, we concluded that opgC and opgH

are cotranscribed

Phylogenetic analysis of MdoH homologues The E coli mdoH gene product shows structural features

of a glycosyltransferase belonging to family 2 [1] With the growing number of complete genomes sequenced, phylogenetic analyses of MdoH homologues is now pos-sible Figure 5 shows a phylogenetic tree obtained by the maximum-parsimony method (the neighbor-joining method gave similar results) Two cellulase synthase from cyano-bacteria can be considered as an out-group and indicate the possible location of the root No similarities were found between the MdoH and the NdvB homologues A very similar tree was observed for the MdoG homologues, but with no out-group (data not shown) One may wonder if a correlation exists between the phylogenetic position of a particular OPG-glucosyltransferase and the structure of the OPGs produced At the present time, we have only partial information However, the OPGs produced by X campestris and R sphaeroides belong to the same family and the corresponding OPG-glucosyltransferases appear related (Fig 5)

In our current working model [1], the OPG-glucosyl-transferase H is assisted by the MdoG-like protein, possibly

a transglycosidase Depending on the proteins considered, the glucan produced would be a linear b,1-2 with b,1-6 branches as in E coli, or cyclic b,1-2 with an a,1-6 closure like in R sphaeroides If this hypothesis is true, it may be possible to obtain b,1-2 polymerization of glucose residues

in E coli when expressing the opgH of R sphaeroides

opgIH ofR sphaeroides can complement a mdoH mutation inE coli

Several difficulties emerged when trying to express

R sphaeroides genes in E coli We have known that

R sphaeroidesgenes are not readily expressed in E coli,

Table 2 OPG production in various opg mutants of R sphaeroides 2.4.1.

Strain

Chromosomal

OPG synthesis

OPG character

even in an in vitro system [32] Moreover, the first gene of

this locus encodes a periplasmic protein translated with an

uncommonly long signal-peptide that may not be

recog-nized by the E coli secretory machinery Actually, the

introduction of cosmid pUI8166 in either mdoG or mdoH

mutants of E coli failed to restore any OPG synthesis

(data not shown) Placing the opgGIH genes downstream of

the lac promoter in pUC19 (pNFR14) was also

unsuc-cessful However, when pNFR2 was introduced into a

mdoH200::Tn10 strain, material corresponding to the OPGs

was detected by thin layer chromatography Sequencing of

the plasmid revealed that a translational fusion should have

fortuitously occurred between the eighteenth codon of the

a-lacZ fragment present in the vector and the third codon of

opgI,thus allowing the expression of the downstream opgH

gene This was confirmed by the construction of pNFR30

where an opgIH-containing fragment with cohesive ends

was blunt ended using the Klenow fragment of DNA

polymerase I (see Materials and methods) Among plasmids

presenting the correct orientation of insert with respect to

the lac promoter, some were able to complement a mdoH

mutation, others not

Plasmids pNFR30 and pNF309 where opgIH (Fig 3)

and mdoH [4], respectively, are governed by the lac

promoter in the same vector, were introduced in the same

mdoHmutant strain OPGs were extracted and analyzed by

gel filtration chromatography The amount of OPGs was

threefold lower with pNFR30 than with pNF309 As OPG

synthesis is increased by a factor of 1.5 when mdoH+is

present on a multicopy plasmid like pNF309 [30], the level

observed with the R sphaeroides genes was considered to

be the result of an efficient complementation The OPGs

produced by the two strains were treated to remove all

substituents and then subjected to a MALDI-MS analysis

With pNF309, the spectra were characteristic of

linear-branched glucans found in E coli, with a DP of five to 14

glucose residues, the three principal species containing six,

seven, and eight glucose residues (Fig 6A) It should be

noted that the presence of pNF309 induced a slight

increased of the maximal DP normally observed [33] With

pNFR30, the spectra were very similar but the maximal DP was at least of 18 glucose units and the principal species contained five, six, seven, eight, and nine glucose residues (Fig 6B)

opgC ofR sphaeroides transfers succinyl residues

to OPGs inE coli

As shown above, very similar proteins are implicated in the synthesis of quite different glucosidic backbones, but

Fig 6 Positive-ion MALDI mass spectra of OPGs extracted from

E coli strains NFB1100 (pmdoH + /mdoH, panel A) or from NFB4234 (popgI + H + /mdoH, panel B) Mass assignments are based on an external calibration.

Fig 5 Unrooted phylogenetic tree for MdoH homologues prepared using the maximum parsimony method as described in Materials and methods Numbers on forks are bootstrap confidence levels Cell synth., cellulose synthase Full species names are as follows: Caulobacter crescentus; Nitrosomonas europeae; Pseudomonas fluorescens; Ralstonia eutropha; Rhodopseudomonas palustris; Shewanella putrefaciens; Vibrio cholerae; Xanthomonas citri; Xylella almond; Xylella oleander; Xylella fastidiosa; Yersinia pseudotuberculosis All sequence data are from ERGO with the exception of X citri (E G M Lemos, School of Agronomy and Vetenary Sciences, Campus of Jaboticabal, Sao Paulo State University, Brazil, personal communication)

membrane proteins, with no sequence similarities, are

probably involved in the transfer of succinyl residues

through the cell membrane to OPGs present in the

periplasmic space [14] The open question was whether the

R sphaeroidesgenes can be expressed in E coli, and what

properties they confer in this context

The highly anionic character of OPGs synthesized by

E coli is due to the presence of phosphoglycerol and

succinyl residues Two genes mdoB and mdoC govern these

two kind of substitution and a mdoB mdoC double mutant

produce neutral OPGs [14] To test the transfer ability of

OpgC in a heterologous context, pNFR37 (Fig 3) was

introduced in strain NFB1933 We had previously observed

that pNFR12, which contains only opgC, was ineffective

Plasmid pNFR37 is expected to express opgI as a

transla-tional fusion with a-lacZ and opgH, and opgC which

overlaps opgH The specifically labeled OPGs produced in

the presence of pNFR37 were purified and analyzed by

DEAE-Sephacel chromatography Under these conditions,

OPGs synthesized by the recipient strain are totally neutral

and are not retained by the column [14] After introduction

of the plasmid, 20% of the radioactivity, corresponding to

anionic glucans, were retained by the column and eluted

into two subfractions by increasing the ionic strength

(Fig 7)

D I S C U S S I O N

This paper describes the isolation of an R sphaeroides

mutant defective in OPG succinylation No OPG-defective

mutants (our initial purpose) were obtained, as was the case

while screening potential mutants (several thousands in

total) in S meliloti [10], E coli [14], and E chrysanthemi

(V Cogez, & J.-P Bohin, unpublished data) The reason for

this fact remains unknown The gene inactivated in the

mutant, opgC, was then isolated and characterized OpgC is

predicted to be a highly hydrophobic protein, most

probably inserted into the cell membrane As we have

previously postulated for MdoC, OpgC should transfer

succinyl residues, probably provided by the succinyl-CoA

pool, through the membrane, to the OPGs accumulating in

the periplasmic space [14] The opgC gene is located downstream of opgGIH, three genes necessary for OPG backbone synthesis opgC overlaps opgH and a polar mutation in opgH prevent opgC expression Thus, opgC forms an operon with opgGIH

When expressed together, OpgI and OpgH can comple-ment a mdoH mutation in E coli That means that proteins that normally catalyze the synthesis in R sphaeroides of b-1,2; a-1,6 cyclic glucans comprised of 18 glucose residues can catalyze the synthesis in E coli of b-1,2; b-1,6 linear glucans comprised of varying numbers (five to 18) of glucose residues The small ORF encoded by opgI was found necessary for OPG synthesis in R sphaeroides In

E coli, opgI is most probably necessary too, but we cannot exclude that only translational coupling is necessary for opgHexpression OpgH is shorter than MdoH, and OpgI appears to correspond to the N-terminal domain of MdoH [30], a domain poorly conserved among the different homologues found in the proteobacteria One should notice that another small protein, the acyl carrier protein (ACP), participates in OPG synthesis in E coli [33] ACP from

E coli(or closely related species) functions in an unknown way that does not require the presence of the phospho-pantetheine prosthetic group ACP from R sphaeroides was shown to be inactive in the in vitro glucosyltransferase reaction both as an activator or an inhibitor [34] Thus, OpgI could play in R sphaeroides, a role similar to that of ACP in E coli

MdoH and OpgH have typical motifs found in glycosyl transferases and one can imagine that both proteins catalyze the polymerization of long chains of b-1,2 glucose residues

We have previously postulated that this kind of protein, embedded in the membrane by a number of TMSs, could be directly involved in the translocation of the nascent glucan chains to the periplasmic face of the membrane [30] Therefore, other proteins such as MdoG or OpgG may rearrange this backbone to add branches (MdoG) or make

a cyclic molecule (OpgG) As mutants of this second protein

do not accumulate any glucan molecules (this work; [4]), the periplasmic and the membrane-bound proteins must inter-act in a very coordinate manner during the process Thus, the abnormal control of the degree of polymerization of the OPG synthesized when the R sphaeroides opgIH genes were expressed, would be the result of a partially defective interaction between MdoG and OpgH Until now, we had not obtained the expression of OpgG in E coli This experiment will be important in order to determine to what extent this protein determines the structural differences between the OPGs produced by the two different bacterial species

When expressed in E coli, together with OpgI and OpgH, OpgC can transfer charged residues to the OPGs Together with the loss of succinyl transfer in the opgC mutant, this is a confirmation that OpgC is an OPG-succinyltransferase However, this activity remained at a low level in E coli, especially if we consider the highly anionic character of the R sphaeroides OPGs Two hypo-theses, not mutually exclusive, can be formulated to explain this observation: one regarding the protein organization in the membrane and one regarding the substrate specificity

As already mentioned above, OpgC and MdoC are two functional homologues that do not share significant amino-acid sequence similarity In E coli, succinylation occurs

Fig 7 DEAE-Sephacel anion exchange column chromatography

pro-files of [U-14C] glucose labeled OPGs from strain NFB4245

(popgI+H+C+/mdoC mdoB) Ionic strength was increased by steps of

0.05 M NaCl at fractions indicated by the arrows Fractions (4 mL)

were collected and radioactivity was determined on aliquots.