Identification of essential loops and residues of glucosyltransferase V (GtrV) of Shigella flexneri

Abstract

Lipopolysaccharide (LPS), particularly the O-antigen component, is one of many virulence determinants necessary for Shigella flexneri pathogenesis. O-antigen modification is mediated by glucosyltransferase (gtr) genes encoded by temperate serotype-converting bacteriophages. The gtrV and gtrX genes encode the GtrV and GtrX glucosyltransferases, respectively. These are integral membrane proteins, which catalyze the transfer of a glucosyl residue via an α1,3 linkage to rhamnose II and rhamnose I of the O-antigen unit. This mediates conversion of S. flexneri serotype Y to serotype 5a and X, respectively. Essential regions in the topology of GtrV protein were identified by in vivo recombination and a PCR-mediated approach. A series of GtrX-GtrV and GtrV-GtrX chimeric proteins were constructed based on the fact that GtrV and GtrX share sequence similarity. Analysis of their respective serotype conversion abilities led to the identification of two important periplasmic loops: loops No 2 and No 10 located in the N- and C-termini, respectively. Within these two loops, three conserved motifs were identified; two in loop No 2 and one in loop No 10. These conserved motifs contain acidic residues which were shown to be critical for GtrV function.

• Bacteriophage
• glucosyltransferase
• chimeras
• dual reporter
• Shigella flexneri

Introduction

Shigella flexneri is the major cause of shigellosis or bacillary dysentery. S. flexneri possesses many virulence factors; one of these factors is the lipopolysacharide (LPS) extending from the outer membrane (Allison & Verma [2000]). LPS plays an important role in pathogenesis and in protection against host defences. It consists of three main segments: lipid A, core polysaccharide, and the O-antigen (Brahmbhatt et al. [1992], Korres & Verma [2004], Makela & Stocker [1984], Okada et al. [1991], Sandlin et al. [1995], Sansonetti [1998], Sansonetti [2001]). The O-antigen consists of repeating tetrasaccharide units comprised of 2)-β-L-Rha-(1–2)-α- L-Rha-(1–3)-α-L-Rha-(1–3)-α-D-GlcpNAc-(1 (Allison & Verma [2000]). Specific serotypes arise by the addition of glucosyl residues (glucosylation) to specific sugars of the repeat unit via specific linkages. It has recently been shown that glucosylation of LPS facilitates bacterial invasion of target cells by enhancing the type III secretion system, without compromising the protective properties of LPS (West et al. [2005]). The genes associated with the glucosylation process are carried by temperate phages and have been identified in a three-gene cluster containing gtrA, gtrB and gtr_[type]. For complete serotype conversion all three genes are required, although gtrA and gtrB are highly homologous and interchangeable between serotypes. The gtr_[type] is unique and highly specific in adding the glucosyl residue to the correct sugar of the O-antigen via a specific linkage. The membrane topology of all three proteins encoded by this cluster has recently been determined and a model of how these three S. flexneri membrane proteins modify the O-antigen unit proposed (Korres et al. [2005], Korres & Verma [2004], Markine-Goriaynoff et al. [2004]). GtrV adds a glucosyl residue via an α1,3 linkage to rhamnose II of the O-antigen chain, while GtrX adds a glucosyl residue to rhamnose I via the same linkage (Allison & Verma [2000], Markine-Goriaynoff et al. [2004]). This gives rise to serotypes 5a and X which have the type V and group 7,8 antigens, respectively. The unmodified serotype Y possesses the 3,4 group antigen (Allison & Verma [2000]).

Identification of domains and essential regions for catalysis of membrane proteins can be achieved by genetic approaches. These approaches have successfully been used to identify functional domains of various membrane proteins where chimeras were constructed between homologous genes (Buck & Amara [1994], Cosgriff et al. [2000], Giros et al. [1994], Nishizawa et al. [1995]). Such studies provided important information on domain existence and the formation of catalytic sites, and ultimately on the function of the proteins. In this paper, we report the presence of two essential loops in the topology of GtrV and the identification of critical residues in both loops. These essential segments (loops) were determined by the construction of chimeras between GtrV and its closest homologue, GtrX. Segment swapping between the two proteins revealed that GtrV and GtrX consist of two functional segments which include periplasmic loop No 2 located in the N-terminus and periplamsic loop No 10 located in the C-terminus. Identification of potentially important residues in both segments was achieved by multiple alignments between the S. flexneri Gtrs. Conserved residues identified in both loops include negatively charged amino acids such as aspartic acid and glutamic acid. These were mutated in GtrV by site-directed mutagenesis and were shown to abolish serotype conversion from serotype Y to serotype 5a.

Materials and methods

Bacterial strains, plasmids and DNA methodology

All bacterial strains used in this study are shown in Supplementary Table 1 with their genotypes (online version only). The bacterial strains were grown aerobically at 37°C in Luria Bertani broth (LB). Antibiotics were added to liquid and solid media as described previously (Sambrook & Russell [2001]). Alkaline phosphatase (AP) and β-galactosidase (BG) activities were detected as described previously in Korres and Verma ([2004]).

Plasmids used in this study are listed in Supplementary Table I (online version only). Primers were synthesized by Invitrogen or Proligo. Double-stranded plasmid sequencing was performed at the Biomolecular Resources Facility, John Curtin School of Medical Research, Australian National University. Sequencing primer PHOSEQ specific for the 5′ end of the phoA-lacZ fusion was used to determine the exact point of fusion. Also primers gtrV1780FSeq (GCACGCCAATGTTACTAATG) and gtrX1760FSeq (CTCACTGGTATTTATCATTG) were used to sequence within the gtrV and gtrX genes, respectively. PCR was performed using Pfu polymerase as specified by the manufacturer (Promega). DNA techniques were used as described previously by Sambrook and Russell ([2001]).

Construction of chimeras by homologous recombination

Construction of chimeras by this method requires that gtrV and gtrX be in tandem with two unique restriction sites between them to successfully linearize the plasmid (Cosgriff et al. [2000]). In brief, gtrV was amplified from pNV323 using primers gtrVFSacI (GATGAGCTCTGAGAAACAAAAATGAAAAGCC) and gtrVRXbaI (ACGTCTAGAACCATTCAACATTAAGGC). These primers have the SacI and XbaI sites incorporated in them, thus it was possible to ligate this fragment into pBC SK using the same sites, giving rise to pNV1077. gtrX was then amplified from pNV574 using the primers gtrXFHindIII (GGTAAGCTTCGGGATTGGTGTTCTCGG) and gtrXRKpnI (GGAGGTACCCAGGCTACTATAAACACG). These primers have the HindIII and KpnI sites incorporated in them, therefore it was possible to ligate this fragment just downstream of gtrV in pNV1077. This gave rise to pNV1082, in which both genes are in tandem with BamHI and EcoRI sites between them. pNV1082 was used to construct GtrV→GtrX chimeras. Functionality of both genes was confirmed as described below. In order to create chimeras with the GtrX→GtrV orientation, the gtrX was amplified using different primers gtrXFSacI (GGTGAGCTCCGGGATTGGTGTTCTCGG) and gtrXRXbaI (GGATCTAGACAGGCTACTATAAACACG), then cloned in the respective sites of pBC SK to create pNV1057. gtrV was amplified with the primers, gtrVFHindIII (GATAAGCTTGGATTGGTGTTCTCGGTG) and gtrVRKpnI (ACGGGTACCACCATTCAACATTAAGGC), and ligated into the respective sites in pNV1057, giving rise to pNV1063. The functionality of both genes was confirmed as described below. pNV1063 has the same restriction sites between gtrV and gtrX as pNV1082, and differs only in the order of gtrV and gtrX.

GtrX→GtrV chimeras were created using pNV1063. In brief, pNV1063 was linearized using BamHI and EcoRI. After inactivation of the restriction enzymes, the linearized DNA (100–500 ng) was electroporated into JM109. Preliminary screening of transformants was performed by colony cracking as described previously (Sambrook & Russell [2001]) and later confirmed by sequencing. This technique gave rise to 2 chimeras: pNV1066 which consists of gtrX from 1–417bp and gtrV from 418–1221bp and pNV1067, which consists of gtrX from 1–378bp and gtrV from 379–1224bp. The functionality of chimeras was checked as described below. In order to increase the frequency and range of chimeras generated in this orientation, the gtrV gene was recloned again in pNV1057 in the HindIII and KpnI sites using the primers gtrVFHindIII (GATAAGCTTTGAGAAACAAAAATGAAAAGCC) and gtrVRKpnI (ACGGGTACCACCATTCAACATTAAGGC). This time gtrV was amplified exactly from its start as opposed to the previous construct which included 100bp upstream of the start codon. This gave rise to pNV1079. As with the previous construct the frequency and diversity of chimeras was limited and no further chimeras were obtained. The same problem was encountered when pNV1082 was used to construct GtrV→GtrX chimeras. No chimeras were obtained in this orientation, so a PCR mediated approach was employed.

Construction of GtrV→ GtrX chimeras by PCR

Construction of GtrV→GtrX chimeras was performed by PCR. Primers were designed to amplify in-frame portions of both genes. pNV1082 was used in conjunction with the primers gtrXFchimL3 (GCCAATATAACAAATGCTCACTGG) and gtrVRchimL3 (GTGCACTTCTGATACTTCCGGCATTG). The amplified product, which included portions of both genes (start of gtrV and end of gtrX) without the intervening region and the rest of the vector, was gel purified and self-ligated. The ligated product was then transformed into JM109. Transformants carrying in-frame chimeric fusions were confirmed by sequencing. This gave rise to pNV1261 which consists of gtrV from 1–408bp and gtrX from 409–1275bp. The same procedure was followed for pNV1188 which consists of gtrV 1–687bp and gtrX 688–1254bp, using primers chimgtrVR (GGTCTCAGAGGATGTCATG) and chimgtrXF (AGATCACATGCTCCATTAG). In addition, pNV1189 was constructed using primers gtrXchimL7 (GGATAATCACCTTTTATCG) and gtrVchimL7 (CATAAAGGTGGCAAAAAGTCACC), and consists of gtrV 1–795bp and gtrX 796–1254bp. Primers gtrXFchimL10 (CCTGACATGAACTGGTCTC) and gtrVRchimL10 (GAGTGGGGTAAGATTGAA) gave rise to pNV1197 which consists of gtrV 1–1134bp and gtrX from 1135–1245bp (Table I).

Table I.  Chimeras GtrV→GtrX and GtrX→GtrV constructed by a PCR mediated approach and homologous recombination. P, positive. This table is reproduced in colour in Molecular Membrane Biology online.

		Agglutination with antisera
	Construct model	V	X	3,4	Dual indicator coloration/Protein synthesis – assembly
pNV1066		+	−	+	Red/P
FL1460
-
pNV1067		+	−	+	Red/P
SFL1630
-
pNV1261		−	−	+	Blue/P
-
pNV1188		−	−	+	Blue/P
pNV1189		−	−	+	Blue/P
SFL1554
pNV1197		−	−	+	Blue/P
pNV1232		−	−	+	Blue/P
pNV1337		−	−	+	Blue/P
pNV1077		+	−	+	Blue/P
Agglutination example with serotype V antisera

Also a double mutant was constructed consisting of gtrV-gtrX(reentrant loop)-gtrV-gtrX portions (Table I). pNV1189 was used since it already contains the gtrX end from 796–1254bp. In this construct two unique SmaI/EcoRV sites were introduced by site-directed mutagenesis. The primers gtrVSmaIreloopF (GTTTATTATAACAATGCCCGGGGTATCAGAAGTGC) and gtrVSmaIreloopR (GCACTTCTGATACCCCGGGCATTGTTATAATAAAC) were used for SmaI introduction, giving rise to pNV1221. pNV1221 was then used in conjuction with the primers gtrVEcoRVreloopF (CTATTATCATGATATCCTCTGAGACCAGAG) and gtrVEcoRVreloopR (CTCTGGTCTCAGAGGATATCATGATAATAG) to introduce the EcoRV site, giving rise to pNV1224. The reentrant loop of gtrX was amplified from pNV574 using the gtrXEcoRVreloopR primer (GGAGATATCAGAATTATTGATGTTGCCTGTATC) and the gtrXFHindIII primer (GGTAAGCTTCGGGATTGGTGTTCTCGG). Since there is an endogenous SmaI site in gtrX it was possible to cut the insert with SmaI and EcoRV and insert it in-frame into pNV1224. In-frame insertion was checked by sequencing. This gave rise to pNV1232 which consists of gtrV1–390bp-gtrX391–696bp-gtrV697–816bp-gtrX817–1275bp (Table I).

The construction of the following chimera consisting of gtrV-gtrX(reentrant loop)-gtrV was performed by inserting the SmaI/EcoRV sites in pNV1077 as done previously (double mutant). Prior to this the SmaI and EcoRV sites located within the vector's multiple cloning site had to be eliminated. pNV1077 was cut with both EcoRV and SmaI and self-ligated, giving rise to pNV1319. By using pNV1319 and the primers outlined in the construction of the double mutant, we created pNV1323, which contains the SmaI site. Then pNV1323 was used for the insertion of the EcoRV site as described previously giving rise to pNV1324. The reentrant loop of gtrX was extracted from pNV1232 and blunt-end ligated to the EcoRV/SmaI sites of pNV1324. This gave rise to pNV1337 which consists of gtrV1–390bp-gtrX 391–696bp-gtrV 697–1272bp (Table I).

The functionality of all the chimeras was checked by introducing them into either SFL1444 or SFL1616 and subsequent serotype conversion was checked by agglutination using either type V or group 7,8 agglutination sera (SEIKEN).

To determine whether the chimeric proteins were synthesized, the dual reporter phoA/lacZα was ligated to unique sites such as HpaI and EcoRV found in the chimeric genes. In brief, the dual reporter was extracted with EcoRV/SmaI and inserted in-frame using the HpaI site of pNV1188, 1189, 1197, 1232 and 1261, giving rise to pNV1285, 1286, 1287, 1288 and 1318, respectively. The dual reporter was also inserted in pNV1066, pNV1067 and pNV1337 using the EcoRV site for in-frame insertion, giving rise to pNV1289, pNV1322 and pNV1338, respectively. Sequencing was used to confirm in-frame insertion. Constructs were then plated onto dual indicator plates as described previously (Korres & Verma [2004]) and coloration was observed. Constructs pNV1066 and 1067 were further investigated by western immunoblotting using mouse anti-E.coli PhoA antisera (Chemicon) as described below, in order to investigate protein synthesis and assembly.

Site-directed mutagenesis

Site-directed mutagenesis was performed as described by the manufacturer (Stratagene). Plasmids pNV1077 and pNV1105 containing gtrV and gtrV/phoA/lacZα, respectively, were used as the templates. Primers F42E (CAGTTTTGGGCACAAGATGCTCATGTATGG) and R42E (CCATACATGAGCATCTTGTGCCCAAAACTG) were used to mutate E42 to Q using pNV1077, creating pNV1092. Primers F43D (CAGTTTTGGGCAGAAAATGCTCATGTATGG) and R43D (CCATACATGAGCATTTTCTGCCCAAAACTG) were used to mutate D43 to N using pNV1077, creating pNV1094. Primers F4243ED (CAGTTTTGGGCACAAAATGCTCATGTATGG) and R4243ED (CCATACATGAGCATTTTGTGCCCAAAACTG) were used to mutate E to Q and D to N at positions 42 and 43 in succession using pNV1105, creating pNV1173. Primers F168D (CAAAAAGCACACAATTATATAGCGA) and R168D (TCGCTATATAATTGTGTGCTTTTTG) were used to mutate D168 to N using pNV1077, and primers F168A (CAAAAAGCACACGCTTATATAGCGA) and R168A (TCGCTATATAAGCGTGTGCTTTTTG) were used to mutate D168 to A using pNV1105, creating pNV1085 and pNV1187, respectively. Primers F175A (GCGATAATCATAGCTGGACTAAGTGGCCC) and R175A (GGGCCACTTAGTCCAGCTATGATTATCGC) were used to mutate C175 to A using pNV1105, creating pNV1193. Primers F380D (CCCACTCCCAAATTACGGGTG) and R380D (CACCCGTAATTTGGGAGTGGG) were used to mutate D380 to N using pNV1077, creating pNV1087. Primers F380A (CCCACTCCCAGCTTACGGGTGG) and R380A (CCACCCGTAAGCTGGGAGTGGG) were used to mutate D380 to A using pNV1105, creating pNV1168. Primers F397A (GTTTAGCACCAGGAGCATCGTATTCTTTC) and R397A (GAAAGAATACGATGCTCCTGGTGCTAAAC) were used to mutate E397 to A using pNV1105, creating pNV1180.

In addition, constructs generated during the topology analysis of GtrV (Korres & Verma [2004]) using pNV1090, which includes the dual reporter phoA/lacZα, were functionally assayed in this study. For example pNV1102, which consists of GtrV in-frame with the dual reporter at amino acid 380, missing most of the C-terminal end was analysed for function by introducing it into SFL1444 and performing agglutination tests with type V antisera. The same was done for plasmids pNV1104 (in frame to dual reporter at 382 amino acids), pNV1159 (in frame to dual reporter at 384 amino acids), pNV1110 (in frame to dual reporter at 385 amino acids), pNV1101 (in frame to dual reporter at 392 amino acids) and pNV1105 (in frame to dual reporter at 412 amino acids). In this way we were able to monitor the function (and thus protein production and assembly) in both mutants created by site-directed mutagensis and C-terminal truncations.

Immunogold labelling and electron microscopy

To determine the degree of serotype conversion occurring on chimeras pNV1067 and 1066, the bacteria were adhered to carbon-coated copper 200-mesh grids, negatively stained, and analysed under a Hitachi H-7100FA transmission electron microscope. Immunogold labelling of the bacteria was performed according to the previously published protocol (Giron et al. [1991], Lehane et al. [2005]).

Fractionation of E. coli cells, protein SDS-PAGE and western blotting

The method used to isolate proteins in the insoluble fraction (inner membrane [IM] and outer membrane [OM]) has been described previously (Achtman et al. [1979], Morona et al. [1983], Korres et al. [2005]). Samples were solubilized at 100°C in sample loading buffer before SDS electrophoresis as described previously (Sambrook & Russell [2001]) using iGels (12%) from Gradipore. Equal amounts of protein were loaded on the gel (functional chimeras) as assayed by BioRad protein assay. Detection of GtrV/AP/BG fusions and chimeric sandwich fusions was achieved by transfer of SDS-PAGE separated samples to nitrocellulose membranes (Morona et al. [1991]). The primary antibody used was mouse anti-E. coli PhoA antisera obtained from Chemicon. Goat anti-rabbit peroxidase conjugate was used as the secondary antibody. Staining was performed as described by the manufacturer Sigma (Sigma Fast, 3,3-diaminobenzidine tetrahydrochlorite with metal enhancer).

Whole cell colony agglutination

The functionality of the resultant constructs was checked by introducing them into SFL1444 and/or SFL1616. In brief, the mutant plasmids were introduced by electroporation into SFL1444 and/or SFL1616, which are derivatives of SFL124 except that they carry pNV1060 (gtrA and gtrB, Amp^R) and pNV1241 (Km^R), respectively. Since the whole three-gene cluster has to be present for complete serotype conversion, this allows functional examination of gtrV with gtrA and gtrB which are carried on another plasmid. Serotype conversion was tested by slide agglutination as described by Korres and Verma ([2004]).

Results

Characterization of essential regions by homologous recombination and PCR

Regions essential for the function in GtrV were identified by constructing chimeras consisting of GtrV as the N-terminal segment and GtrX as the C-terminal segment and vice versa. Due to the close homology between GtrV and GtrX it was possible to create chimeras by homologous recombination which occurs between homologous stretches of DNA ranging between 6–15bp (Cosgriff et al. [2000], Tommassen et al. [1985], Wang [2000]). Also, the fact that a good level of DNA homology exists only between the first 450bp (total size 1.25kb), as seen from best fit analysis further suggests that two such essential segments exist (data not shown). The only chimeras that were successfully obtained by this method were pNV1066 and 1067 of the GtrX→GtrV orientation. The limitation in the range of chimeras can be attributed to the fact that DNA homology gets weaker past 450bp. Also, the longer stretches of homology in the N-terminal sequence will be more frequently favoured in recombination than shorter stretches found along the genes. Furthermore, no chimeras were obtained in the GtrV→GtrX orientation after several attempts, suggesting that these constructs would be unstable. Chimeras pNV1066 and 1067 were transformed into SFL1444 and showed positive agglutination with type V antisera (serotype V), and no agglutination with group 7,8 antisera (serotype X) (Table I).

The agglutination observed with type V antisera was not as strong and fast as observed for the wild-type gtrV gene, suggesting that the chimeras were not fully functional. This can be seen in Figure 1 where both chimeras showed gold particle binding compared to the negative control pNV1060, which does not carry a Gtr and was therefore unmodified. Binding to the positive control (pNV1077) was stronger than that observed for the chimeras, further supporting the hypothesis that function was somehow obscured.

Figure 1.  Functional analysis of mutants. Electron micrographs (X50 000 magnification) of SFL1444 containing pNV1066 and 1067 tested with type V antibody. Secondary antibody used is conjugated to 10nm gold particles. Presence of gold particles is indicative of O-antigen modification. Description of each strain is shown below the image.

In order to investigate whether the effect on function was related to specific segments or amino acid interactions rather than altered protein synthesis levels and assembly, the dual reporter phoA/lacZα was introduced in the above chimeras giving rise to pNV1289 and pNV1322. Coloration was observed over time on dual indicator plates when these constructs were transformed in JM109 (Korres & Verma [2004]). Red coloration intensity was observed equally for all chimeras when compared among themselves and to the positive control (data not shown), indicating equal level of expression of Gtr protein produced. Western immunoblot with PhoA antisera (Figure 2) further confirmed that a similar amount of protein was produced by the functional chimeras and the positive control pNV1105. Thus the weaker binding observed by the chimeras was not due to decreased protein production levels but perhaps due to altered function.

Figure 2.  Western immunoblot of functional chimeras. Western immunoblot of chimeras pNV1066 and 1067 using mouse anti-E.coli PhoA antisera. pNV1105, positive control; pNV1060, negative control.

As mentioned earlier, the construction of chimeras in the GtrV→GtrX orientation was unsuccesful using homologous recombination, so a PCR-mediated approach was adopted. This method gave rise to pNV1261, pNV1188, pNV1189 and pNV1197 (see Supplementary Table I online for details). The GtrX→GtrV functional chimeras suggested that a chimera of the opposite orientation should be able to convert serotype Y to serotype X. This hypothesis was tested by chimera pNV1261, when transformed into SFL1616. This chimera failed to agglutinate with either type V or group 7,8 antisera, clearly indicating that this chimera is non-functional. The chimeras described next were constructed at the same time as pNV1261. In order to investigate whether conversion to serotype X can be established, a range of chimeras were constructed (Supplementary Table I online). In brief, the chimeras constructed were pNV1188, pNV1189 and pNV1197. All of these have progressively larger N-terminal segments and subsequently smaller C-terminal segments (Table I). For example, pNV1197 consists of the whole GtrV protein except for the C-terminal end after the TPLPD motif (see site-directed mutagenesis section below) which is substituted with the C-terminal end of GtrX. All of the above chimeras were transformed into SFL1444 and tested for their ability to mediate serotype conversion. All chimeras failed to convert serotype Y to either serotype 5a or serotype X, suggesting that the protein was either non-functional or that protein synthesis and assembly did not occur. In order to investigate whether protein synthesis and assembly occurred, the dual reporter was fused in frame to unique restriction sites in all of the above mentioned chimeras (see Materials and methods). All of the chimeras were produced and assembled in the inner membrane as indicated by the blue color produced on dual indicator plates (Table I) (Korres & Verma [2004]). The inability of the chimera derived from pNV1197 to induce serotype conversion suggests that the periplasmic C-terminal loop No 10 of GtrV must be required for O-antigen modification (Figure 4).

To investigate the role of the reentrant loop in the GtrV topology (Korres & Verma [2004]), a double mutant (pNV1232) was created. As observed from pNV1189, the reentrant loop of GtrX is not part of this segment, thus it was hypothesized that the C-terminal segment requires its own reentrant loop (same origin) for proper function as observed in the GtrX→GtrV functional chimeras. When this chimeric double mutant pNV1232 was transformed into SFL1616 to determine its serotype conversion ability, the transformed bacteria failed to agglutinate with either type V or group 7,8 antisera (Table I). Also the blue coloration observed in transformants carrying the dual reporter (see Materials and methods), indicated that the protein constructs were produced and assembled (Table I). Since this construct was non-functional, no insights were gained into the importance of the reentrant loop.

In order to further investigate the importance of the reentrant loop, a second mutant was created which involved the substitution of only the reentrant loop. Thus both the N- and C-terminal segments were of GtrV origin and the reentrant loop was of GtrX origin. The resultant mutant pNV1337 (see Materials and methods) when tested in SFL1616 for serotype conversion with type V and group 7,8 antisera, failed to produce agglutination with either (Table I). The dual reporter was also introduced in pNV1337 to check for protein production. As observed from dual indicator plates blue color production confirmed protein synthesis and assembly. This clearly suggests that the reentrant loop is of functional significance.

The above results indicate that two distinct functional regions do exist in the GtrV protein with the N-terminal loop No 2 possibly interacting in a conserved manner with UndP-Glucose while the C-terminal loop No 10 is involved in the specific attachment of the glucosyl residue to the O-antigen unit. These two essential segments must contain critical residues which form the active site. Also the double and single mutants constructed further support the notion that the reentrant loop is important for GtrV function and has to be part of the C-terminal segment.

Identification of critical residues by site-directed mutagenesis

Truncations generated in an earlier study of GtrV topology were assayed for function (Korres & Verma [2004]) (see Materials and methods). These constructs mainly have the C-terminal periplasmic end truncated at different positions. When assayed in SFL1444 it was found that all truncations were able to abolish function, except pNV1105 which retained its function despite the presence of the dual reporter after residue 412. In addition, the presence of the dual reporter confirmed that the fusion proteins were present in the inner membrane. This, together with the results gathered from the chimeras, further support that the C-terminal end of GtrV is functionally important and contains critical residues.

Residues for site-directed mutagenesis were chosen on the basis of conservation between the Gtrs. Bioinformatics was of limited use since the Gtrs share little sequence homology. GtrV and GtrX are the only proteins that show significant homology (35.7% identity and 61.3% similarity; Figure 3A).

Figure 3.  Best fit analysis of Gtrs. (A) Best fit analysis of GtrV and GtrX. Consensus is shown below sequence and in boxes. Note consensus regions boxed (red), indicative of motifs containing negatively charged residues. In vertical boxes shaded grey are the residues that were targeted by site-directed mutagenesis. (B) Best fit analysis of first periplasmic loop No 2 of N-terminus present in all Gtrs. The conserved aspartic acid (D) is shown in italics (grey box). Also in bold are either aspartic or glutamic acids part of the shown motif. This Figure is reproduced in colour in Molecular Membrane Biology online.

From this a number of conserved motifs were identified: WAEDA and KxxD (where x is any amino acid) in the N-terminal segment which includes loop No 2 and TPLPD in loop No 10 (C-terminus). All of these motifs are characterized by the presence of E and/or D residues. Furthermore, a closer inspection of residues located in the first periplasmic loop (which is topologically conserved in Gtrs; data not shown) revealed a conserved aspartic acid in all Gtrs. Also this residue is close to another acidic residue (Figure 3B), forming a kind of a DxD, ED or DE motif. This is included in the motif WAEDA found in GtrV and GtrX. Since these conserved negatively charged residues are located in the periplasm it was postulated that they might interact with UndP-Glucose (loop No 2) or involved in attaching the glucosyl group to the specific rhamnose of the O-antigen (Loop No 10). Also, a cysteine residue in the reentrant loop was mutated to investigate whether this residue forms important disulfide bonds with another cysteine of either segment, since the reentrant loop is important for function.

The mutations E42Q, D43N, E42QD43N, D168N, D168A, C175A, D380N, D380A and E397Q were introduced by site-directed mutagenesis (see Materials and methods). The positions of these residues on the topology of GtrV are shown in Figure 4A.

Figure 4.  Position of critical GtrV residues and western immunoblot. Targeted residues in respect to the recently solved topology of GtrV. (A) Highlighted residues (bold-underlined) were mutated to residues shown next or above them. Note that important residues ED in the first periplasmic loop No 2 (N-terminus) were mutated to QN respectively, while residue D in loop No 10 (C-terminus) was mutated to N as well as A. (B) Western immunoblot of site-directed mutants pNV1173 and 1168 using mouse anti-E. coli PhoA antisera. A 98kDa band which includes both GtrV and dual reporter is observed in both constructs. This Figure is reproduced in colour in Molecular Membrane Biology online.

The template used was pNV1077 for E42Q, D43N, D168N and D380N and pNV1105 for the rest. pNV1105 has the full GtrV protein with the dual reporter fused in frame at the end of the C-terminus after residue 412. Despite the presence of the dual reporter this construct is still functional and can readily convert serotype Y to 5a. In addition, the dual reporter gives rise to blue coloration on dual indicator plates due to its periplasmic position (Korres & Verma [2004]). The use of this construct enables the rapid testing of whether non-functional constructs are assembled in the inner membrane in JM109.

Constructs encoding the mutated protein were transformed into SFL1444 (see Materials and methods). The E42QD43N, E42Q, D43N and D380A mutations were found to abolish serotype conversion to 5a (Table II).

Table II.  Site directed mutagenesis applied on selected GtrV residues. Mutations in bold are critical in function. P, Positive.

		Agglutination with Antisera		Dual Indicator Plate Coloration/Protein Synthesis	Agglutination slides
Strain/Wild type Residues	Mutated Residues	V	3.4		Mutant	Positive Control pNV1077
pNV1092/E42, pNV1094/ D43, pNV1173/E42D43	Q42, N43 Q42N43	−	+	Blue/P
pNV1085, pNV1187/D168	N168, A168	+	+	Blue/P	Positive	Positive
pNV1193/ C175	A175	+	+	Blue/P	Positive	Positive
pNV1087, pNV1168/ D380	N380, A380	+, −	+	Blue/P
pNV1180/E397	A397	+	+	Blue/P	Positive	Positive

The rest of the mutants were functional (Table II). Interestingly, when residues E42 and D43 were mutated to their basic forms (Q and N), function was destroyed. However when D380 was mutated to its basic form (N) there was no effect, but when mutated to alanine (a neutral amino acid), function was abolished. This might indicate a specific role for the amino acids at positions 42 and 43.

To determine whether the mutated non-functional constructs were assembled in the inner membrane of JM109, JM109 bearing constructs pNV1173 and pNV1168 were plated onto dual indicator plates. Blue coloration indicated that both of the encoded mutants were assembled in the inner membrane (Korres & Verma [2004]). In support, Figure 4B shows a western immunoblot of the non-functional constructs pNV1173 and pNV1168. When compared to the negative control pNV1060, a distinctive band is apparent at 98kDa which includes GtrV and the dual reporter. This further supports the notion that the effect of the mutations is based on protein function and not on protein synthesis and assembly.

Discussion

The topology of GtrV enabled us to further characterize its mode of action by identifying important regions and residues, thus giving us greater insight into how these glucosyltransferases are able to modify the O-antigen and confer serotype specificity (Korres & Verma [2004]). These segments were identified by creating chimeras between GtrV and its closest homologue, GtrX. Although this method has been used previously to identify conserved segments between related proteins (Cosgriff et al. [2000], Wang [2000]), this is the first report where integral membrane protein chimeras constructed using homologous recombination were able to maintain function despite exhibiting only 35.7% identity.

The two chimeras constructed in the GtrX→V orientation showed serotype conversion to 5a and not to X. Conversion to serotype 5a indicated that despite substituting the N-terminus with GtrX, serotype conversion to 5a was maintained. Thus the N-terminal segment (loop No 2) is potentially performing a conserved function in interacting with UndP-Glucose while the C-terminal segment (loop No 10) may be responsible for specifically attaching the glucosyl residue to the specific rhamnose. The N-terminal segment identified in this above orientation consists of three N-terminal transmembrane segments (1–417bp; 139aa) while the C-terminal segment consists of six C-terminal transmembrane segments (418–1221bp; 268aa). These chimeric constructs show reduced function compared to the wild-type protein. This could be due to limited or restricted interactions between essential segments and specific residues. This has also been observed in previous studies in which chimeric activity was decreased compared to the wild type (Cosgriff et al. [2000], Diallinas et al. [1998], Moore & Blakely [1994], Nishizawa et al. [1995], Will et al. [1998]). However, construction in the opposite orientation (GtrV→GtrX) showed absence of function. This indicated that specific amino acid interactions between segments or donor and acceptor substrate interactions are maintained in the first orientation and not in the second, possibly due to improper protein folding. Although function is lost, protein synthesis and assembly in the membrane are maintained, as shown by the blue color produced by the dual reporter phoA/lacZα. This phenomenon has also been observed in previous studies (Cosgriff et al. [2000], Diallinas et al. [1998], Moore & Blakely [1994], Nishizawa et al. [1995], Will et al. [1998]).

Other glycosyltransferases that use similar donor substrates such as DolP-linked glycosyltransferases found in eukaryotes, also possess a topologically similar periplasmic N-terminal loop after transmembrane segment I (Breton & Imberty [1999]). Domains have been identified in other glycosyltransferases of different function such as ScPmt1p (an O-mannosyltransferse) and MurG:UDP-GlcNAc responsible for forming the glycosidic linkage between N-acetyl muramyl pentapeptide and N-acetylglucosamine in the biosynthesis of the bacterial cell wall. Domain identification and active site formation in the latter was investigated by 3D crystallography (Girrbach et al. [2000], Hu et al. [2003]).

Supplementary Table I. Bacterial strains and plasmids used in this study.

Strain or plasmid	Relevant characteristics^?	Source reference
E. coli K-12
DH5?	supE44 ΔlacU169 (φ80 lacZ ΔM15) hsdR17 recA1 endA1 gyrA96 thi-1 relA1	(Hanahan [1983])
JM109	recA1 supE44 endA1 hsdR17 gyrA96 relA1 thi Δ(lac-proAB) F’ [ traD36 proAB+ lacI^q lacZ ΔM15]	(Yanisch-Perron et al. [1985])
XL1-Blue MRF’	supE44 hsdR17 recA1 endA1 gyrA46 thirelA1 lac- Δ(mcrA)183 Δ(mcr CB-hsd SMR-mrr)173	(Jerpseth et al. [1992])
S. flexneri
SFL124	Attenuated Shigella vaccine serotype Y (ΔaroD)	(Lindberg et al. [1988])
SFL1444	SFL124 carrying pNV1060	This study
SFL1616	SFL124 carrying pNV1241	A. Lehane
Plasmids
pBC SK	cloning vector	Stratagene
pMA632	pBluescript II SK (+) carrying ‘phoA-lacZ?’ cassette	(Alexeyev & Winkler [1999])
pNV323	SfV 3 gene cluster (gtrA, gtrB and gtrV in pUC19)	(Huan et al. [1997])
pNV574	SfX 3 gene cluster (gtrA, gtrB and gtrX in pBC KS II, Amp^r)	(Huan et al. [1997])
pNV1060	Amp^r, 1.64kb HindIII/BglII fragment containing gtrA and gtrB in low copy number pACYC177
pNV1066	chimeric gene consisting of gtrX 1-417bp and gtrV 418-1221, produced using pNV1063 by homologous recombination	’’
pNV1067	chimeric gene consisting of gtrX 1-378 and and gtrV 379-1224, produced using pNV1063 by homologous recombination	’’
pNV1077	Cm^r pBCSK containing SacI/XbaI gtrV (1.25kb) fragment	This study
pNV1085	Using pNV1077, gtrV mutated at amino acid 168 from D to N	’’
pNV1087	Using pNV1077, gtrV mutated at amino acid 380 from D to N	’’
pNV1092	Using pNV1077, gtrV mutated at amino acid 42 from E to Q	’’
pNV1094	Using pNV1077, gtrV mutated at amino acid 43 from D to N	’’
pNV1105	Cm^r derivative of pNV1090 carrying fusion of truncated gtrV and dual reporter phoA/lacZ at amino acid position 412 of GtrV	This study
pNV1168	pNV1105 carrying GtrV and dual reporter mutated at 380 amino acid GtrV from D to A	’’
pNV1173	pNV1105 GtrV mutated at amino acids 42 and 43 from E to Q and D to N respectively	’’
pNV1180	pNV1105 GtrV mutated at amino acids 397 from E to A	’’
pNV1187	pNV1105 GtrV mutated at amino acid 168 from D to A	’’
pNV1188	Chimeric gene consisting of gtrV 1-687bp and gtrX 688-1254, produced by PCR mediated method using pNV1082	’’
pNV1189	Chimeric gene consisting of gtrV 1-795bp and gtrX 796-1254, produced by PCR mediated method using pNV1082	’’
pNV1197	Chimeric gene constisting of gtrV 1-1134bp and gtrX1135-1245bp, produced by PCR mediated method using pNV1082	’’
pNV1193	pNV1105 GtrV mutated at amino acid 175 from C to A	’’
pNV1232	Double mutant consisting of gtrV 1-390, gtrX 391-696 reentrant loop, gtrV 697-816 gtrX 817-1275 produced using pNV1189	’’
pNV1241	Km^r, 1.64kb BamHI/PstI fragment containing gtrA and gtrB in low copy number pACYC177	A. Lehane
pNV1261	Chimeric gene consisting of gtrV 1-408bp and gtrX 409-1275, produced by PCR mediated method using pNV1082	This Study
pNV1337	Double mutant consisting of gtrV 1-390, gtrX 391-696 reentrant loop, gtrV 697-1272 produced using pNV1324	’’

^??The genetic nomenclature is that described by Bachman ([1990]). Allele numbers are indicated where known. Cm^r, chloramphenicol resistance; Amp^r, ampicillin resistance.

The significance of the reentrant loop was also investigated. Constructs pNV1232 and pNV1337, which did not mediate serotype conversion to either serotype 5a or X, revealed that the reentrant loop must be protein and segment specific. As mentioned in Korres and Verma ([2004]), the reentrant loop can allow the protein to be flexible at a particular point so that the C-terminus (potentially loop No 10) can position itself to the correct rhamnose after binding to the donor substrate (Korres & Verma [2004]). By switching reentrant loops, this flexibility and/or specific movement is abolished thus specific interactions cannot be maintained. GtrV attaches the glucosyl residue to the second rhamnose, while GtrX attaches it to the first rhamnose, thus as seen in pNV1337 which contains just the GtrX reentrant loop, the glucosyl residue cannot be positioned to the correct rhamnose, as the C-terminus is not of GtrX origin. Therefore specific interactions and positioning cannot be maintained by the reentrant loop of GtrX and C-terminus of GtrV. As seen in pNV1067 and 1066 in which the reentrant loop and the C-terminus (loop No 10) are of the same origin, serotype conversion to 5a is maintained.

To further elucidate the mechanism of O-antigen glucosylation, critical residues were identified by best fit analysis in both the essential regions. Three critical residues were identified: E42, D43 and D380. The mutated constructs were fused to the dual reporter phoA/lacZα to ensure that non-functional constructs are still synthesized and assembled in the inner membrane.

Mutating residues E42 and D43 to their basic uncharged forms abolished function. The same was observed for mutations in the DxD motifs of the GT-A superfamily of glycosyltransferases (Breton & Imberty [1999]). These motifs are thought to form the active site and require the association of metal ions for their function. Glutamic or aspartic acids of these glycosyltransferases are believed to assist in deprotonating the nucleophilic hydroxyl group of the acceptor sugar (Breton & Imberty [1999]). Although the acidic residues are not separated by another residue in GtrV and GtrX, these residues might perform a similar function or interaction. Also an equivalent residue in GtrII (E40) has been mutated and shown to abolish GtrII function (Lehane et al. [2005]). GtrII adds a glucosyl residue via an α1,4 linkage to rhamnose III of the O-antigen chain. Potentially equivalent residues have been identified in DolP-linked glycosyltransferases, especially in the 2nd loop. Although these residues have not been mutated, they are thought to perform a similar action. In the case of GtrV, these residues are located in loop No 2 and may be involved in the initial interaction with the donor substrate, in this case UndP-Glucose, before C-terminal (loop No 10) involvement and specific substrate addition. The existence of two acidic residues in loop No 2 of all the Gtrs (Figure 3B) supports the involvement and existence of a conserved loop as shown in this study and points to a possible conserved mechanism of action.

Residue D380 located in loop No 10 was also found to be critical for function. In contrast to E42D43, D380 had to be mutated to a neutral amino acid such as alanine for function loss. This might indicate that D380 is involved in binding to the acceptor substrate (the O-antigen) and/or in indirectly interacting with loop No 2 for proper substrate binding to the O-antigen. Furthermore, the truncated mutants provide further evidence that the C-terminal end (also includes the amino acid D380) is required for proper function. GtrV and GtrX attach a glucosyl residue to different rhamnoses via the same linkage. This supports the notion that the critical residues identified in this study (E42, D43 and D380) may be involved in the formation of the active site and be responsible for the specific linkage α1,3, while the reentrant loop with specific rotational movements facilitated by specific amino acid interactions is able to facilitate the addition of the glucosyl residue to the correct rhamnose of the O-antigen. This movement has also been observed in the MurG:UDP-GlcNAc protein which undergoes a 10°C-terminal body rotation relative to its position when UDP-GlcNaAc is bound, facilitating proper function (Hu et al. [2003]). Although conserved motifs were identified between GtrV and GtrX, similar motifs are absent in other Gtrs except in the 1st periplasmic loop of the N-terminal segment. This indicates that over time the Gtrs from different serotypes have evolved in such a way to recognize a different acceptor and attach a glucosyl residue with different linkages. The same has been observed in DolP-linked glycosyltransferases which, despite recognizing the same donor and acceptor substrates, share few regions of sequence homology (Breton & Imberty [1999]).

In this study for the first time we have identified two GtrV essential regions: the N-terminal segment which is potentially responsible for the conserved interaction with UndP-Glucose and the C-terminal segment which is potentially involved in the specific addition of the glucosyl residue to the O-antigen. Also, three critical residues were identified – two in the N-terminus (loop No 2; glutamic and aspartic acid) and one in the C-terminus (loop No 10; glutamic acid). These residues are thought to be involved in the interaction with the donor and acceptor substrates. In addition, the reentrant loop identified in the previous study (Korres & Verma [2004]) was proven to be of functional importance. This provides the basis for the localization of the active site and understanding the mode of action of the remaining S. flexneri Gtrs.

We would like to thank Sally Stowe, Lily Shen and Cheng Huang at the Electron Microscopy Unit for their technical assistance and Adele Lehane and Rob Stagg for critical reading of the manuscript. This work was supported by the National Health and Medical Research Council of Australia.