LPS Unmasking of Shigella flexneri Reveals Preferential Localisation of Tagged Outer Membrane Protease IcsP to Septa and New Poles

The Shigella flexneri outer membrane (OM) protease IcsP (SopA) is a member of the enterobacterial Omptin family of proteases which cleaves the polarly localised OM protein IcsA that is essential for Shigella virulence. Unlike IcsA however, the specific localisation of IcsP on the cell surface is unknown. To determine the distribution of IcsP, a haemagglutinin (HA) epitope was inserted into the non-essential IcsP OM loop 5 using Splicing by Overlap Extension (SOE) PCR, and IcsPHA was characterised. Quantum Dot (QD) immunofluorescence (IF) surface labelling of IcsPHA was then undertaken. Quantitative fluorescence analysis of S. flexneri 2a 2457T treated with and without tunicaymcin to deplete lipopolysaccharide (LPS) O antigen (Oag) showed that IcsPHA was asymmetrically distributed on the surface of septating and non-septating cells, and that this distribution was masked by LPS Oag in untreated cells. Double QD IF labelling of IcsPHA and IcsA showed that IcsPHA preferentially localised to the new pole of non-septating cells and to the septum of septating cells. The localisation of IcsPHA in a rough LPS S. flexneri 2457T strain (with no Oag) was also investigated and a similar distribution of IcsPHA was observed. Complementation of the rough LPS strain with rmlD resulted in restored LPS Oag chain expression and loss of IcsPHA detection, providing further support for LPS Oag masking of surface proteins. Our data presents for the first time the distribution for the Omptin OM protease IcsP, relative to IcsA, and the effect of LPS Oag masking on its detection.


Introduction
Shigella flexneri is an intracellular pathogen which causes bacillary dysentery, a disease characterised by the presence of severe mucoid bloody diarrhoea and by invasion of the gut epithelium [1,2]. IcsA (VirG) is a 120 kDa outer membrane (OM) protein localised at the cell pole [3]. It mediates intracellular cytoplasmic movement of S. flexneri in epithelial cells, and cell-to-cell spread, by the assembly of an F-actin comet-tail at one pole of the bacterium [4][5][6]. This type of movement is described as actin-based motility (ABM). IcsA is secreted primarily at the 'old pole' of Shigellae [7] which is opposite the 'new pole' (the pole derived from the site of septation of the parent cell [8]. The 36.9 kDa IcsP (SopA) OM protease of S. flexneri slowly cleaves IcsA at the Arg 758 -Arg 759 bond position [9] resulting in the release of a 95 kDa aminoterminal IcsA fragment that can be detected in culture supernatants [5,10]. Analysis of icsP/sopA mutants has shown that IcsA is detected across the entire surface of these bacteria with polar reinforcement [11,12]. Over-expression of IcsP results in the complete removal of IcsA from the cell surface [13].
IcsP belongs to the Omptin family of proteases which consists of 6 members; OmpT and OmpP of Escherichia coli, Pla of Yersinia pestis, PgtE of Salmonella enterica, Pla endopeptidase A of Erwinia pyrifoliae, and IcsP of Shigella flexneri. Immunogold labelling of overexpressed OmpP has shown that OmpP is symmetrically distributed over the cell surface [14]. However to date, no studies have attempted to describe the surface localisation of Omptins expressed at native levels. While it has been suggested that IcsP may also be located uniformly across the cell surface [13], its specific distribution is currently unknown. In contrast to many inner membrane proteins, such as FtsZ [15] and MreB [16] involved in cell division and cell shape, few OM proteins have had their subcellular distribution determined. An exception to this is the E. coli OM protein LamB which has been characterised to exist as two populations: one that diffuses in a helical pattern, and one that is relatively immobile [17,18]. The E. coli Iss and Bor proteins have been detected on the cell surface with no distinct pattern [19]. A number of non-specific E. coli OM proteins were suggested to be organised in stable helical swaths [20], and data by Shiomi et al. (2006) suggested that the general protein translocation Sec machinery itself may also be arranged in a helical array. Whether IcsP possesses a distribution similar to these OM proteins, or has an asymmetric distribution like IcsA, is the subject of this study.
In addition to the above, mutations affecting lipopolysaccharide (LPS) have also been shown to affect the observed distribution of OM proteins [21][22][23][24]. LPS is composed of three distinct regions: lipid A, core sugars and O antigen (Oag) polysaccharide chains.
Strains with LPS containing all 3 regions intact are known as smooth LPS strains. Shigella mutants lacking Oag are known as rough LPS strains. Such strains have been shown to have high levels of circumferentially distributed IcsA on the cell surface (at both cell poles and on lateral regions) [25,26], compared to the polar localisation of IcsA seen in smooth LPS strains. Treatment of Y serotype derivatives of smooth LPS S. flexneri with bacteriophage Sf6 tailspike protein (TSP) endorhamnosidase results in the hydrolysis of Oag chains and an increase detection of circumferential IcsA on the cell surface by indirect immunofluorescence (IF) staining [21]. This suggests that the presence of LPS Oag masks the observed distribution of IcsA on the cell surface and supports the idea that LPS Oag structure may block antibody accessibility to the detection of surface proteins [22,23]. The effect of LPS Oag structure on the detection and distribution of IcsP has not been investigated.
In this study, we investigated the distribution of IcsP by cell surface quantum dot (QD) IF labelling of functional, HA-tagged IcsP (IcsP HA ) in S. flexneri 2457T and establish that LPS Oag masks detection of IcsP HA on the cell surface by using tunicamycin to inhibit Oag synthesis. Additional IF labelling with anti-IcsA antibodies to mark the location of the old pole suggested that IcsP is preferentially localised to the new pole of non-septating cells and to the septa of septating cells. We also investigated the distribution of IcsP in a rough LPS 2457T strain to provide further support for the LPS Oag masking hypothesis. Overall, our data presents for the first time the cell surface distribution of the Omptin OM protease IcsP and the effect of LPS Oag masking on its detection. This distribution has implications for IcsA polarity determination, and a model is described to explain IcsP's contribution to IcsA polarity in S. flexneri.

Ethics Statement
The anti-IcsP and anti-IcsA antibodies were produced under the National Health and Medical Research Council (NHMRC) Australian Code of Practice for the Care and Use of Animals for Scientific Purposes and were approved by the University of Adelaide Animal Ethics Committee.

Bacterial Strains, Plasmids and Media
Bacterial strains and plasmids used in this study are listed in Table 1. Bacteria were routinely grown at 37uC in Luria-Bertani (LB) broth with aeration or on Congo red agar [26]. Antibiotics used were as follows: 50 mg ampicillin (Amp) ml 21 ; 25 mg chloramphenicol (Cml) ml 21 ; 50 mg kanamycin (Kan) ml 21 .

DNA Methods
E. coli K-12 DH5a was used for all cloning experiments. DNA manipulation, PCR, transformation and electroporation was performed as previously described [27,28]. Anti-HA monoclonal antibody (#H3663) was purchased from Sigma. Rabbit anti-IcsP and anti-IcsA antibodies were prepared as described previously [26,29]. The antibodies were produced under the National Health and Medical Research Council (NHMRC) Australian Code of Practice for the Care and Use of Animals for Scientific Purposes and were approved by the University of Adelaide Animal Ethics Committee.

Insertion of HA Epitope into IcsP
The sequence encoding the HA epitope (YPYDVPDYA) was inserted into the putatively non-essential IcsP OM loop 5 (based on sequence alignments with E. coli OmpT) using SOE PCR [30,31]. In the first part of this two-step PCR technique, upstream and downstream amplicons were amplified from S. flexneri 2457T genomic DNA using HA encoding primers (ET18/ET19) and icsPspecific primers (ET3/ET10) ( Table 2). The two amplicons from this primary PCR were then mixed and used as a DNA template for the second round of PCR with primers ET3/ET10. In this second reaction, the HA encoding regions of the primary PCR amplicons overlap and prime one another to give the final icsP PCR product with an inserted HA epitope. The icsP HA fragment was then cloned into pGEMT-Easy and primers ET22/ET25 (Table 2) were used to amplify an icsP HA product with KpnI and HindIII restriction enzyme sites from this construct. The resultant KpnI-HindIII fragment was digested and sub-cloned into likewise digested pBAD30 to give pBAD30::icsP HA , also referred to as pIcsP HA in text (Table 1). Primers ET22/ET25 were also used to amplify the icsP gene from 2457T genomic DNA, and cloned into pBAD30 to give pBAD30::icsP, referred to as pIcsP in text (Table 1). DNA sequencing was used to confirm that no mutation had been introduced by PCR into the sequence, and the presence of the in-frame HA epitope tag sequence.

Construction of S. flexneri icsP mutant
The S. flexneri 2457T icsPmutant strain was constructed using allelic exchange mutagenesis [27] to inactivate the icsP gene by insertion of a kanamycin resistance gene (kan R ). Initially, the icsP gene was PCR amplified with primers ET3/ET4 containing BamHI and SacI restriction sites from 2457T genomic DNA. The resultant PCR fragment was digested with BamHI and SacI and sub-cloned into likewise digested pSL1180 (Table 1). Further digestion with ClaI allowed insertion of the AccI-AccI digested kan R gene from pKTUWE (Table 1) to give pSL1180-icsP::kan R (Table 1). Following re-digestion with BamHI and SacI, the icsP::kan R fragment was cloned into pCACTUS (Table 1) and transformed into S. flexneri 2457T via electroporation. Allelic exchange mutagenesis was performed as previously described [27]. The icsP::kan R mutation in the virulence plasmid was confirmed by PCR with primers ET3/ET4 (Table 2) to give the 2457T icsPmutant ETRM22 ( Table 1).

Construction of S. flexneri icsP 2 /rmlD -Double Mutant
The S. flexneri 2457T icsP 2 /rmlDmutant strain was constructed using a modification of the l Red recombinase system to initially delete the rmlD gene [32]. Primers ET28/ET29 containing NheI restriction enzyme sites ( Table 2) were used to PCR amplify the kan R gene from pKD4 ( Table 1). The amplified product was ligated into pGEMT-Easy and pGEMT-Easy::kan R was digested with NheI. The NheI-kan R -NheI fragment was then subcloned into likewise digested pRMA718 [26] to give pRMA718-rmlD::kan R (Table 1). This plasmid was then digested with BamHI and the rmlD::Kan R fragment was cloned into the BamHI site of pCACTUS. The pCACTUS-rmlD::Kan R construct was then electroporated into S. flexneri 2457T and allelic exchange mutagenesis was induced to give the 2457T rmlD::kan R mutant ETRM230 (Table 1). ETRM230 was transformed with pCP20 at 30uC to flip out the FRT flanked kan R gene and give the 2457T rmlDmutant ETRM233 ( Table 1). The rmlD mutation was confirmed by LPS analysis. ETRM233 was further electroporated with pCACTUS-icsP::kan R (Table 1) and another round of allelic exchange mutagenesis was performed to give the final 2457T rmlD 2/ icsPdouble mutant ETRM240 ( Table 1).

Analysis of IcsP/IcsP HA Protein Production
For detection of native IcsP, strains were grown at 37uC in LB broth with aeration for 16 h, subcultured 1/20 into fresh broth and grown for another 3 h to an OD 600 of ,1. Strains harbouring pIcsP or pIcsP HA were grown in LB broth containing 0.2% (w/v) glucose for 16 h with aeration, subcultured 1/20 into fresh broth and grown for 1.5 h to an OD 600 reading of ,0.4. Cultures were then pelleted by centrifugation (22196g, 10 min, Sigma 3K15 centrifuge), washed 3 times in LB, and unless otherwise stated, induced with 0.03% (w/v) arabinose for 1 h to an OD 600 of ,1. Cells (5610 8 ) were then harvested by centrifugation and resuspended in 2X sample buffer [33]. Protein samples were solubilised at 100uC for 5 min, separated on SDS 15% polyacrylamide gels, and stained with Coomassie R-250, or subjected to Western immunoblotting on nitrocellulose membrane (Medos) with either polyclonal rabbit anti-IcsP antiserum (at 1/250 dilution) or monoclonal mouse anti-HA (at 1/500 dilution). Detection was performed with goat anti-rabbit (or anti-mouse) horseradish-peroxidase-conjugated antibodies (KPL) and chemi-luminescence reagent (Sigma). Benchmark prestained molecular weight markers (Invitrogen) were used as molecular size markers.

Sucrose Gradient Density Fractionation
Fractionation of the cell whole membrane (WM) into cytoplasmic membrane (CM) and OM fractions was performed by sucrose gradient centrifugation according to the method of Osborn and Munson [34]. In brief, 200 ml cultures were grown and induced with arabinose as described above, harvested by centrifugation (9,8006g, 15 min, 4uC, JA14 rotor, Beckman centrifuge J2-21M), washed in 50 mM Tris-HCl (pH 8.0) and resuspended in 5 ml 10 mM HEPES in 1 mM MgCl 2 . The bacterial suspension was then passed through a pre-cooled French Pressure cell (SLM Aminco) once and re-centrifuged to remove cell debri. WM pellets were collected by ultracentrifugation (115,0006g, 1 h, 4uC, 80 Ti rotor, Beckman Coulter Optima L-100 XP ultracentrifuge),   icsPmutant (icsP -) were grown in LB and whole cell lysate samples were taken at 0.5, 1, 1.5, 2, 2.5 and 3 h after subculture, followed by electrophoresis on a SDS 15 % polyacrylamide gel and Western immunoblotting with rabbit anti-IcsP antiserum; (B) strains 2457T, icsPand icsPharbouring pIcsP, pIcsP HA or pBAD30 (as indicated) were grown in LB for 1.5 h to an OD600 reading of ,0.4, washed 3 times, and induced with arabinose for 1 h. Pellet and supernatant protein samples were then prepared and electrophoresed on a SDS 15 % polyacrylamide gel, followed by Western immunoblotting with rabbit anti-IcsA antibodies. The size of the full length IcsA protein (120 kDa) and the cleaved form of IcsA (95 kDa) are indicated; (C) S. flexneri strains 2457T and icsPharbouring pIcsP HA or pBAD30 were grown in LB as described in (B), followed by induction with 0%, 0.003%, 0.006%, 0.0125%, 0.025%, 0.05%, 0.1% or 0.2% (w/v) arabinose for 1 h. Whole cell lysates were prepared and electrophoresed on a SDS 15 % polyacrylamide gel, followed by Western immunoblotting with rabbit anti-IcsP antiserum. The size of the full length IcsP protein (36 kDa) is indicated in (A) and (C). Each lane contains 5 x 10 7 bacterial cells of each strain. doi:10.1371/journal.pone.0070508.g001 Distribution of the Outer Membrane Protease IcsP PLOS ONE | www.plosone.org solubilised in 0.8 ml 25% (w/w) sucrose in 5 mM EDTA and applied to a 10 ml sucrose gradient of 30-50% (w/w) sucrose in 5 mM EDTA. Centrifugation to equilibrium was performed with a Beckman SW40Ti swing out rotor (217,0006g, 20 h, 4uC, Beckman Coulter Optima L-100 XP ultracentrifuge) and 0.5 ml fractions collected through the pierced bottom of the tube. 10 ml samples of each fraction were resuspended in 2X sample buffer [33] and IcsP protein detected as described above.

Detection of Cell Associated and Soluble IcsA
Whole cell and supernatant bacterial protein extracts were prepared as described previously [35]. IcsA protein was detected from 10 ml whole cell protein extracts and 20 ml supernatant protein extracts. Western immunoblotting was performed as described above, but with a polyclonal rabbit IcsA antibody and a goat anti-rabbit HRP conjugate.

LPS Analysis
LPS samples and gels were prepared as described previously [36,37].

LPS Depletion-regeneration Assay
Depletion and regeneration of LPS was performed as previously described [29] with the exception that 0.03% (w/v) arabinose induction was included in the final hour of tunicamycin/ polymyxin B nonapeptide (PMBN) treatment.

Formaldehyde Fixation of Bacteria for Immunofluorescence (IF) Microscopy
Bacteria were grown and induced as described above 1610 8 cells of induced bacteria were then harvested by centrifugation, washed once in PBS and resuspended in 100 ml 3.7% (w/v) formaldehyde (Sigma) in PBS for 20 min at room temperature (RT). Fixed bacteria were then pelleted, washed three times in PBS and resuspended in a final volume of 100 ml PBS.

Quantum Dot (QD) IF Staining and Epi-fluorescence Microscopy
Sterile glass coverslips were placed into wells of a 24-well tray and coated with 10% (v/v) poly-L-lysine solution (Sigma) in PBS for 1 h at RT. Coating solution was aspirated and 5 mL of formaldehyde fixed bacteria were spotted onto coverslips. The tray was then centrifuged to assist adherence of bacteria (Heraeus Labofuge 400R Centrifuge, 2,0006g, 5 min, 20uC). Bacteria were blocked for 1 h at RT with 10% (v/v) foetal calf serum (FCS) diluted in PBS. For labelling, bacteria were incubated for 2 h at RT with mouse anti-HA antibody (Sigma) and rabbit anti-IcsA antiserum diluted 1:50 and 1:100 respectively in PBS containing 10% (v/v) FCS. Bacteria were then washed 3 times with PBS, and then incubated for 1 h at RT with either QD 525 donkey antimouse antibody (Invitrogen) or QD 625 donkey anti-rabbit antibody (Invitrogen) diluted 1:50 and 1:100, respectively, in PBS containing 10% (v/v) FCS. After a final 3 more washes with PBS, coverslips were mounted on glass microscope slides with Mowiol 4-88 (Calbiochem). All microscopy images were captured using an Olympus IX-7-Microscope, with a phase contrast 100X oil immersion objective and a 1.5X enlarger, which was controlled by MetaMorph (Version 7.7.1.0, Molecular Devices). All IcsP HA (525 nm) and IcsA (625 nm) channel images were acquired with 1 sec and 0.1 sec exposures respectively using an X-Cite 120Q lamp set at high intensity as the excitation source. The excitation filter used was FF01-435/40-25 (Semrock) and the emission filters were FF01-525/15-25 and FF01-625/15-25 (Semrock). Semrock FF510-Di01-25636 dichroics were used.

Fluorescence Quantification of QD Labelled Surface IcsP HA
Data for intensity profile plots were extracted from images using MetaMorph's Line-scan function which averages intensities across the perpendicular axis of a point-to-point scan. Single scans were conducted from pole-to-pole with width (perpendicular axis) equal to the bacterium (approx. 20 pixels). Cumulative scans of the 525 nm wavelength (IcsP HA ) for non-septating cells were conducted for 50 bacteria from each independent sample. Each bacterium was scanned from the new pole to old pole where IcsA was used as a marker of the old pole. From the same samples, all septating cells from captured images were scanned starting from the septum to the old pole of one daughter cell chosen at random. An average of 26.8 septating bacteria were scanned for each sample. Intensity data was then exported to MS Excel using MetaMorph's Dynamic Data Exchange and subsequently analysed using GraphPad Prism. Statistical significances were tested by Student's two-tailed t-test.

Detection of IcsP Expression by Western Immunoblotting and Immunofluorescence
In initial experiments to detect cell surface IcsP, the optimal time point for IcsP expression in S. flexneri was determined. Western immunoblotting with a rabbit anti-IcsP was performed on whole cell lysates collected from wild-type S. flexneri 2457T and the 2457T icsPmutant (icsP -) grown for 0.5, 1, 1.5, 2, 2.5 and 3 h after subculture. A band consistent with the size of the IcsP protein (,36 kDa) was detected at time points after 1.5 h for 2457T (Fig. 1A, lanes 5, 7, 9 & 11) with high expression observed at 3 h (Fig.  1A, lane 11). We reasoned that IcsP levels at this time point were high enough for subsequent immunofluorescent detection of IcsP.
No expression of IcsP was observed for icsPas expected (Fig. 1A,  lanes 2, 4, 6, 8, 10 & 12). Subsequent attempts to detect IcsP on the surface of 2457T with rabbit anti-IcsP however were unsuccessful, even in a rough LPS strain (data not shown). We speculate that as the IcsP protein used to raise antisera was purified under denaturing conditions, this may have affected the resulting antibody's ability to detect native IcsP. However, difficulties with IF detection of cell surface IcsP with a polyclonal antibody have also been reported by others [13]. Hence, in an alternative approach to investigate the distribution of IcsP on the bacterial cell surface, a HA epitope was inserted into the IcsP protein.

Insertion of a HA Epitope into IcsP
IcsP is 60% identical in primary amino acid sequence to the E. coli protease OmpT and computer structure modelling predicts that both proteins possess similar b-barrel structures ( Fig. 2A). Based on sequence alignments with E. coli OmpT (Fig. 2B), a HA epitope tag (YPYDVPDYA) was hence inserted into the OM loop 5 of IcsP (Fig. 2B) using SOE PCR (as described in the Methods). An area within this loop region with some sequence diversity between IcsP and OmpT was selected as we reasoned that this sequence variability might allow the protein to accommodate the  The schematic (C) shows methodology of IcsP HA intensity scan directions from either new poles to old poles (as marked by IcsA) of non-septating bacteria, or septa to epitope insertion with little disturbance to the overall structure. The OM loop 5 region was also selected to increase the chance of surface detection by antibodies. OM loops 2 and 4 were not selected for HA tag insertion to avoid the proposed catalytic residues present in OmpT [38] which also exist in IcsP (Fig. 2C). The IcsP HA (as well as IcsP) coding regions were placed in front of the pBAD promoter in pBAD30 [39] to allow expression control with arabinose. Expression of IcsP/IcsP HA was confirmed by Western immunoblotting with anti-IcsP or anti-HA antibodies (Fig. 3A & B, lanes WM and Fig. 1A & C).

IcsP HA Activity on IcsA
To determine whether insertion of a HA epitope into the OM of IcsP affected IcsP's protease activity on IcsA, pellet and supernatant protein preparations of 2457T, icsPand arabinoseinduced icsPstrains expressing pIcsP HA , pIcsP and pBAD30 were subjected to Western immunoblotting with an anti-IcsA antibody. The full length 120 kDa IcsA protein was detected in cell pellet samples of all strains as expected (Fig. 1B, lanes 1, 3, 5, 7 and 9), while the 95 kDa cleaved form of IcsA was only detected in the supernatant sample of 2457T, icsP -[pIcsP HA ] and icsP -[pIcsP] (Fig. 1B, lanes 2, 6 & 8). These results suggest that the insertion of the HA tag into the OM loop 5 of IcsP does not affect its ability to cleave IcsA, and hence the IcsP HA protein is functional.

Localisation of IcsP/IcsP HA Protein to the OM
Since IcsP HA was modified compared to IcsP and expressed from the pBAD promoter, the presence of IcsP HA protein exclusively in the OM was confirmed by using sucrose density gradient centrifugation. The WM of icsP -[pIcsP HA ] and icsP -[pIcsP] were fractionated into CM and OM on sucrose gradients, and fractions subjected to SDS 15% polyacrylamide gel electrophoresis, prior to visualisation by Coomassie Blue staining and Western immunoblotting with anti-HA and anti-IcsP. Analysis of the sucrose gradient samples showed that fractions which were enriched with OM proteins (OmpF, OmpC, and OmpA) [40] contained the majority of the 36 kDa IcsP HA and IcsP proteins ( Fig. 3A & B, lanes 2 to 5). These results indicate that IcsP HA is localised to the OM similar to IcsP, and that the HA insertion did not result in any dramatic disruption of IcsP protein localisation.

Cell Surface Detection of IcsP HA Distribution in S. flexneri 2457T and the Effect of Tunicamycin Treatment
Having established an induction protocol that closely approximates the level of expression of IcsP HA to native IcsP, we next attempted to detect IcsP HA at these levels on the surface of 2457T icsPusing indirect QD IF microscopy. Strains icsP -[pIcsP HA ] and icsP -[pBAD30] were cultured with 0.03% (w/v) arabinose, fixed, and then probed for HA using a primary anti-HA antibody and a secondary QD 525 conjugated antibody. However, no IcsP HA was detected on the cell surface, and intensity scans of icsP -[pIcsP HA ] were equivalent to icsP -[pBAD30] (Fig. S1).
Since IcsP HA could not be detected on the cell surface of icsP -[pIcsP HA ], we reasoned that the presence of LPS Oag may mask the detection of surface IcsP as this has previously been shown for IcsA [21]. An LPS Oag depletion-regeneration assay [29] was hence carried out on icsP -[pIcsP HA ] and icsP -[pBAD30] induced with 0.03% (w/v) arabinose, followed by IF labelling. This assay involves the use of an inhibitor of the WecA enzyme necessary for Oag subunit biosynthesis using tunicamycin, and polymyxin B nonapeptide (PMBN) was used to improve OM penetration. Upon removal of these two chemicals from growing bacteria, LPS Oag is regenerated [29]. Analysis of the resulting LPS by SDS-PAGE and silver staining showed that icsP -[pIcsP HA ] and icsP -[pBAD30] samples treated with tunicamycin/PMBN (TP) had depleted LPS Oag (Fig. 4A, lanes 5 & 6), with Oag production restored (R) upon removal of TP (Fig. 4A, lanes 7 & 8). Untreated (U) and PMBN treated (P) samples showed no inhibition of Oag biosynthesis as expected (Fig. 4A, lanes 1-4).
Following successful depletion of LPS Oag, QD IF microscopy was performed on the above mentioned samples. IcsP HA could not be detected on untreated or PMBN treated icsP -[pIcsP HA ] cells, as expected (Fig. S2A). However, surface IcsP HA was detected on icsP -[pIcsP HA ] cells treated with TP, suggesting that LPS Oag is able to mask antibody accessibility to IcsP HA (Fig. S2A). Interestingly, IcsP HA appeared to be distributed asymmetrically over the cell surface of the majority of cells examined and fluorescence intensity line-scans revealed that IcsP HA localised preferentially to one pole of non-septating cells and the septa of septating cells (Fig. S2A). To determine if the distribution of IcsP HA on the cell surface segregated to either the new pole or the old pole, additional staining of the bacteria was conducted with anti-IcsA antibodies since IcsA is known to localise to the old cell pole [5]. Again, IcsP HA could not be detected on untreated or PMBN treated icsP -[pIcsP HA ] cells (Fig. 5A & B) but could be detected after TP treatment (Fig. 5C). Peak IcsP HA detection was consistently observed at the new pole (opposing IcsA at the old pole) and at the septum of septating bacteria as shown by linescans (Fig. 5C). As expected, no surface IcsP HA was detected for untreated, PMBN treated, or TP treated samples of icsP -[pBAD30] when stained for IcsP HA (Fig. S2B) or double stained for IcsP HA and IcsA (Fig. S3). Notably, IcsA on icsP -[pBAD30] was detected at higher amounts laterally and at the septa of bacteria (Fig. S3) as previously seen for DicsP strains [13].

Cell Surface Detection of IcsP HA Distribution in Rough LPS S. flexneri 2457T
To further investigate the effect of Oag on IcsP distribution, a 2457T icsP 2 /rmlDdouble mutant was constructed to indepen-old poles of septating bacteria. Student's two-tailed t-tests (D) were also conducted on the fold differences of mean IcsP HA intensities between discrete positions (new pole, old pole, and septum) of Oag-depleted icsPand icsP -/rmlDbacteria. SEM = standard error of the mean, GL = Grey level, N = non-septating cells, S = septating cells, P = p-value, ns = not significant. doi:10.1371/journal.pone.0070508.g008 dently assess the distribution of IcsP HA in a rough LPS background. Mutation of the rmlD gene in S. flexneri results in a strain which is unable to synthesise the precursor deoxythymidine diphosphate (dTDP)-rhamnose required for Oag repeat units and, hence, results in a rough LPS phenotype [29,41]. Analysis of the resulting LPS conferred by icsP 2 /rmlDstrains expressing pIcsP HA and pBAD30 showed that rough LPS was observed for both (Fig.  4B, lanes 1 & 2), with a band consistent with the size of the IcsP HA protein (36 kDa) detected only in the icsP 2 /rmlD -[pIcsP HA ] sample by Western immunoblotting, as expected (Fig. 4C, lane 1).
Fixed samples of icsP 2 /rmlD -[pIcsP HA ] and icsP 2 /rmlD -[pBAD30] cells were then probed for both IcsP HA and IcsA using the same QD IF staining protocol as previously conducted for smooth strains. Similarly to LPS Oag-depleted icsP -[pIcsP HA ], IcsP HA was detected on the bacterial surface most predominately at the new pole (Fig. 6A) and the septum (Fig. 6B). Again, the majority of septating cells had higher peak IcsP HA intensity at the septum than new poles of non-septating cells (line-scans Fig. 6A &  B). Single staining of these cells for IcsP HA was also conducted and yielded the same localisation results (Fig. S4). As expected, IcsP HA was not detected on the icsP 2 /rmlD -[pBAD30] strain in IF microscopy experiments when either single (Fig. S4) or double stained (Fig. 6B). Again, IcsA on icsP 2 /rmlD -[pBAD30] was detected at higher amounts laterally and at the septum (Fig. 6A &  B).
To again demonstrate the effect of LPS Oag masking of IcsP HA , smooth LPS structure was restored in the icsP 2 /rmlDstrain expressing pIcsP HA by transforming pRMA727 carrying a functional rmlD gene. Analysis of the resulting LPS conferred by icsP 2 /rmlD -[pIcsP HA ][pRMA727] showed restored smooth LPS phenotype (Fig. 4B, lane 3), and when probed for IcsP HA by QD IF microscopy, was barely detectable (Fig. 7A). The control icsP 2 / rmlD -[pIcsP HA ] strain carrying pACYC184 conferred a rough LPS phenotype (Fig. 4B, lane 4) and IcsP HA was detected by IF microscopy with the same distribution observed previously (Fig.  7B).

Multi-cell Line-scan Analysis of IcsP HA Surface Distribution
To determine if the observed sub-cellular preference of IcsP HA to the new poles and septa is a statistically significant phenomenon, cumulative QD IF line-scan analyses on both IcsP HA /IcsA double stained LPS Oag-depleted icsPstrains and rough icsP 2 / rmlDstrains was conducted for a larger population of cells. Scanning was initiated from the new pole to the old pole (as marked by IcsA) for non-septating cells, and from the septum to the old pole for septating cells (Fig 8C). From 9 independent samples of LPS Oag-depleted icsP -[pIcsP HA ] and icsP -[pBAD30], a total of 450 non-septating cells each were line-scanned. Additionally a total of 355 and 256 septating cells were scanned for icsP -[pIcsP HA ] and icsP -[pBAD30] respectively. The resultant mean fluorescence intensity profiles show that IcsP HA is preferentially localised at the new pole and tends to gradually decrease towards the old pole on non-septating cells expressing pIcsP HA (Fig. 8A). For septating cells, IcsP HA mean intensity is localised highest at the septum and declines more steeply towards the old pole (Fig. 8A). As expected, the intensity profiles of icsP -[pBAD30] cells were at a negligible level (Fig. 8A). Statistical analysis of mean IcsP HA intensity at discrete bacterial positions of LPS-depleted icsP -[pIcsP HA ] confirmed that the localisation of IcsP HA is: (i) 2fold higher at the new pole of non-septating cells than the old pole (P = 0.004), (ii) 2.9-fold higher at the septum of septating cells than the old pole (P,0.0001), and (iii) 1.5-fold higher at septa of septating cells compared to new poles of non-septating cells (P = 0.038) (Fig. 8D).
For the rough LPS strains, 9 independent samples of icsP 2 / rmlD -[pIcsP HA ] and icsP 2 /rmlD -[pBAD30] were investigated with a total of 450 non-septating cells each line-scanned. Additionally, a total of 172 and 173 septating cells were scanned for icsP 2 /rmlD -[pIcsP HA ] and icsP 2 /rmlD -[pBAD30] respectively. Again, the mean intensity profiles of non-septating cells expressing pIcsP HA shows IcsP HA is preferentially localised at the new pole and tends to gradually decrease towards the old pole (Fig. 8B). Likewise, IcsP HA mean intensity of septating cells is localised highest at the septum and declines very steeply towards the old pole (Fig. 8B). As expected, the mean intensity profiles of icsP 2 /rmlD -[pBAD30] were at a negligible level (Fig. 8B). Statistical analysis of mean IcsP HA intensity at discrete bacterial positions of icsP 2 /rmlD -[pIcsP HA ] again confirmed that the localisation of IcsP HA is: (i) 2.2fold higher at the new pole of non-septating cells than the old pole (P,0.0001), (ii) 3.6-fold higher at the septum of septating cells than the old pole (P,0.0001), and (iii) 1.7-fold higher at septa of septating cells compared to new poles of non-septating cells (P = 0.0002) (Fig. 8D). Although, IcsP HA fluorescence intensity was an average of 2.5 times higher on icsP 2 /rmlD -[pIcsP HA ] cells than LPS-depleted icsP -[pIcsP HA ] cells, the ratios of intensity between bacterial positions were comparable to the respective fold-changes observed for LPS Oag-depleted icsP -[pIcsP HA ], suggesting that IcsP HA is localised similarly in both types of Oag deficient cells. The lower level of IcsP HA detection on LPS-depleted icsP -[pIcsP HA ] compared to icsP 2 /rmlD -[pIcsP HA ] indicates that TP treatment is not 100% efficient in inhibiting Oag synthesis.

Discussion
The cell surface distribution of neither IcsP, or any other member of the Omptin family, had not been previously determined. In this study we investigated the distribution of IcsP on the cell surface of S. flexneri 2a 2457T using a HA-tagged IcsP protein (Fig. 2) under pBAD control. Characterisation of IcsP HA showed that it was functionally able to cleave IcsA, and was secreted into the OM comparably to IcsP (Fig. 1B & 3). However, when IcsP HA was expressed at native IcsP equivalent levels ( Fig.  1A & C), it was undetectable in the OM via QD IF microscopy in smooth LPS S. flexneri (Fig. S1) but detectable on both LPS Oagdepleted and rough LPS Shigella bacteria (Fig. 4, 5 & 6). Furthermore, this masking effect was restored in rough strains upon complementation of rmlD (Fig. 4B & 7). We suggest that the long LPS Oag chains of smooth strains sterically hinder the accessibility of antibodies to the OM surface. This type of protein masking by LPS Oag has also been shown for IcsA in 2457T [21], and several other OM proteins [22,23]. It is interesting to note that multi-cell line-scan analysis of IcsP HA detection on the cell surface showed that IcsP HA fluorescence intensity was higher for rough LPS cells than for LPS Oag-depleted cells (Fig. 8). Similar to its Omptin homolog, OmpT, IcsP possesses most of the putative LPSbinding sites found in OmpT (Fig. 2C) [38] and may also interact with LPS. Since TP is only partly effective at blocking LPS Oag synthesis, it is possible that the few LPS Oag molecules synthesised are still closely bound to IcsP. This may cause a small amount of masking and may explain the lower fluorescence intensity observed on LPS Oag-depleted cells (Fig. 8A) compared to the rough LPS cells (Fig. 8B).
IcsA distributions in DicsP strains previously provided indirect evidence to suggest that IcsP is distributed equally over the cell surface of S. flexneri [12,13,42]. However, once the masking effect of LPS Oag was circumvented, we were able to extensively study the distribution profile of IcsP HA in the OM. This work has shown that IcsP HA (and likely IcsP) has an asymmetrical distribution which may be cell cycle dependent. On dividing cells IcsP accumulates at high concentration at the septum and then declines steeply towards the old pole of the emerging daughter cell. However, once divided, detection of IcsP at the new pole (the pole derived from the septum) of daughter cells decreased significantly and then the decline towards the old pole is more moderate. To explain this, we propose that IcsP is directed to septa of dividing cells and that daughter cells 'inherit' a higher concentration of IcsP at new poles (model Fig. 9). As the daughter cell elongates, IcsP is laterally diluted, resulting in the gentle gradient on non-septating cells. In this model, lateral dilution of IcsP may occur by membrane insertion events during cell elongation similarly to that described for E. coli LamB [43]. This model predicts increased cleavage of IcsA at the septum and lateral regions of the cell in order to help set up a relatively higher concentration of IcsA at the old pole and to maintain the old pole preference of IcsA in daughter cells. Additionally, the model may explain, and also fits, the common observation that IcsA distributions and intensities vary between cells in a given population of the same strain.
Many proteins are known to accumulate at the inner membrane of the mid-cell in order to mediate septum formation and cell division -for instance, FtsZ polymerises at the cytoplasmic face to form the Z-ring and attract further divisome components [44]. However, few OM proteins have been shown to concentrate to the septum. A notable exception to this is the OM lipoprotein LpoB which localises as distinct foci at septa and complexes with periplasmic penicillin binding protein 1B tethering it to other divisome components [44,45]. Interestingly, septal localisation of LpoB is lost when peptidoglycan synthesis is inhibited [45]. Whether the localisation of integral OM protein IcsP to the septal OM requires similar interactions with divisome components, or is dependent on peptidoglycan synthesis, remains to be investigated.
In summary, this work has shown for the first time the surface localisation of IcsP, a member on the Omptin family of OM proteases. We have observed that: (i) the distribution of IcsP is masked by LPS Oag in S. flexneri 2457T, and (ii) IcsP is concentrated at new poles and at the septum of dividing cells. This distribution of IcsP explains the observed IcsA localisation defect in DicsP strains [11,12]. We have also proposed a model to explain the inheritance of OM proteins IcsP and IcsA through generations of cell division. Finally, unmasking of surface antigens via LPS Oag-depletion may be useful in the study of other minimally exposed OM proteins.  Figure S4 Single IcsP HA staining of 2457T icsP 2 /rmlD -. Rough LPS 2457T icsP 2 /rmlDstrains harbouring pIcsP HA (left columns) or pBAD30 (right columns) were subcultured LB broth for 1.5 h to an OD 600 reading of ,0.4. Cultures were then washed 3 times in LB, induced with 0.03% (w/v) arabinose for 1 h, fixed, and subjected to QD IF using antibodies for HA epitope. Nonseptating and septating life stages are shown in A and B respectively. Representative bacteria are shown. Scan = Single line-scans measuring the fluorescence intensity of IcsP HA detected along the surface of the bacterium, Bars = 1 mm, Arrows = direction of line-scan, Arrow heads = septa, GL = Grey level, Phase = phase contrast image, IcsP HA = image of fluorescence at 525 nm. (TIF)