Sialylation of Outer Membrane Porin Protein D: A Mechanistic Basis of Antibiotic Uptake in Pseudomonas aeruginosa

Pseudomonas aeruginosa (PA) is an environmentally ubiquitous, extracellular, opportunistic pathogen, associ-ated with severe infections of immune-compromised host. We demonstrated earlier the presence of both (cid:2) 2,3-and (cid:2) 2,6-linked sialic acids (Sias) on PA (PA (cid:3) Sias ) and normal human serum is their source of Sias. PA (cid:3) Sias showed decreased complement deposition and exhibited enhanced association with immune-cells through sialic acid binding immunoglobulin like lectins (Siglecs). Such Sias-siglec-9 interaction between PA (cid:3) Sias and neutrophils helped to subvert host immunity. Additionally, PA (cid:3) Sias showed more resistant to (cid:4) -lactam antibiotics as re-flected in their minimum inhibitory concentration required to inhibit the growth of 50% than PA (cid:5) Sias . Accordingly, we have affinity purified sialoglycoproteins of PA (cid:3) Sias . They were electrophoresed and identified by matrix-assisted laser desorption-ionization time-of-flight/time-of-flight mass spectrometry analysis. Sequence study indicated the presence of a few (cid:2) 2,6-linked, (cid:2) 2,3-linked, and both (cid:2) 2,3- and (cid:2) 2,6-linked sialylated proteins in PA. The outer membrane porin protein D (OprD), a specialized channel-forming protein, responsible for uptake of (cid:4) -lactam antibiotics, is one such identified sialoglycoprotein. Accordingly, sialylated (OprD (cid:3) Sias ) and separately. sensorgrams both similar of

Sialic acids (Sias) 1 are nine carbon atom containing acidic residues characteristically found in the terminal position of glycoproteins and glycolipids (1)(2)(3)(4). Structural diversity of sialic acids is because of the modification of one or more hydroxyl groups in various positions of the core structure by different groups like acetyl-, methyl-, sulfate-, lactyl-, or phosphate (1,(5)(6)(7). More than fifty derivatives of Sias has been reported both in vertebrate and invertebrate systems. It functions as ligand for various cellular communications and also act as masking element for glycoconjugates (8 -12).
Sialic acid binding immunoglobulins (Ig)-like lectins (siglecs) selectively expressed on the hematopoetic cells and interact with an array of linkage-specific Sias on a glycan structure express on the same cells or other cells (13). Siglecs can also recognize terminal sialylated glycoconjugates on several pathogens (14 -16). After recognizing, they carry out various functions like internalization, attenuation of inflammation, restraining cellular activation along with inhibition of natural killer cell activation (17).
Pseudomonas aeruginosa (PA) is a Gram-negative, rodshaped bacterium. This human pathogen has remarkable capacity to cause diseases in immune compromised hosts. This colonizing microbial pathogen is responsible for infection in chronic cystic fibrosis, nosocomial infections; severe burn, transplantation, cancer, and AIDS and other immuno-supressed patients (18).
We have reported earlier the presence of linkage-specific Sias on PA. Normal human serum (NHS) is possibly one of the sources of these Sias (19). PA utilizes these Sias to interact through siglecs present on the surface of different immune cells. PA ϩSias showed enhanced association with neutrophils through ␣2,3-linked Sias-siglec-9 interaction which facilitated their survival by subverting innate immune function of host (20).
The treatment of PA-infected patient depends upon the extent of the disease and the concerned organs. Conventional ␤-lactam, cephalosporins, and aminoglycosides group of antibiotics are most common for such treatment (21). ␤-lactam antibiotics inhibit cell wall synthesis by disrupting the synthesis of the peptidoglycan layer of bacterial cell walls (22). When PA showed resistant to ␤-lactam antibiotics, new generation of ␤-lactam with increased doses or other broad spectrum antibiotics like tetracyclines or fluoroquinolones are prescribed (23). PA isolates from intensive care unit (ICU) patients in general showed higher rates of ␤-lactam resistance among other hospitalized patients (24). The increasing frequency of resistance to ceftazidime, piperacillin, imipenem, fluoroquinolone, and aminoglycoside were 36.6%, 22.3%, 22.8%, 23.8%, and 17.8% respectively in PA (25).
The outer membrane of Gram-negative bacteria is, in general, semipermeable through which hydrophilic molecules including antibiotics of below exclusion limit size (0.6 kDa) can pass through the channel-forming proteins generally called porins e.g. OprD, OprF, OprG etc. (26,27). PA shows lower outer membrane permeability with respect to many other Gramnegative bacteria like Acinetobacter baumannii, Stenotrophomonas maltophilia, Burkholderia cepacia, hence the diffusion rate of ␤-lactam antibiotics is decreased (27).
However, PA being such a notorious organism, it might have many other different mechanisms to fight against antibiotics for their survival. Therefore, it is worthwhile to explore newer mechanism to understand how antibiotics penetrate inside this bacterium. Here we addressed the following questions. Does sialylation of glycoproteins demonstrated on PA play any role in the entry of antibiotics that might facilitate their survival within host?
Accordingly, we have affinity purified a few sialoglycoproteins from PA. Sequence analysis identified twenty six ␣2,3and ␣2,6-linked sialoglycoproteins. One such identified sialoglycoprotein is OprD porin protein. The presence of Sias on OprD was conclusively confirmed. We have demonstrated that Sias on OprD protein isolated four different clinical isolates hampered its interaction with ␤-lactam antibiotics. This might be one of the new mechanisms for ␤-lactam antibiotic resistance of PA and thereby facilitates their survival in host.

EXPERIMENTAL PROCEDURES
Bacteria-Pesudomonas aeruginosa (PA 14) is a wild type, virulent burn-wound isolate, gifted by Prof. Richard D. Cummings, Emory University School of Medicine (Atlanta, GA, USA). Three more stains of PA were isolated from urine (PA Urine ), pus (PA Pus ), and sputum (PA Sputum ) of the patients hospitalized at All Indian Institute of Medical Science, New Delhi, India. The Institutional Human Ethical Committee had approved the study and samples were taken with the consent of the patients. They were grown Trypticase soy broth (TSB, DIBCO) in a microaerobic atmosphere and harvested after overnight growth. Additionally, PA was also grown either in sialic acid free medium [Heme-L-histidine (4.0 ml), RPMI 1640 (191 ml), Minimum essential medium (2 ml, 100 mM), ␤-nicotinamide adenine dinucleotide (2 ml, 1.0 mg/ml H 2 O), uracil (10 ml, 2.0 mg/0.1 N ml NaOH), and inosine (20 ml, 20 mg/ml H 2 O), pH ϭ 7.5] (PA ϪSias ) or in presence of 10% heat inactivated normal human serum (HI-NHS) used as a source of sialic acids (PA ϩSias ) (20). Bacterial suspensions were counted by using a spectrophotometric method and confirmed by pour plate colony counts. Bacterial suspensions were extensively washed with phosphate buffered saline (0.02 M sodium phosphate, 0.15 M saline, pH 7.2; PBS).
Survival of PA-The anti-microbial activities of viable PA ϩSias / PA ϪSias of four clinical isolates including PA14 were measured by using two known ␤-lactam antibiotics such as piperacillin and ceftazidime (Sigma). Different doses of ceftazidime (0, 2, 4, 8, 10, 15, and 20 g/ml) or piperacillin (0, 4,8,15,30, and 50 g/ml) were added to culture PA ϩSias /PA ϪSias [optical density at 600 nm (OD 600 nm ) ϭ 0.3] separately and allowed to grow at 37°C with shaking. Bacterial suspensions were counted by measuring OD 600 nm . The survival of PA was further confirmed by pour plate colony counts and the percent survival was calculated by considering bacteria grown in absence of antibiotics as 100%.
Surface Plasmon Resonance (SPR)-Two sialic acid binding plant lectins Sambucus nigra agglutinin (SNA) and Maackia amurensis agglutinin (MAA), that recognize ␣2,6and ␣2,3sialogalactosyl residues, (Vector labs, Burlingame, CA) respectively were used for this study. The pattern of binding of SNA and MAA with the membrane fraction of PA was determined using SPR (Biacore 2000; Biacore, Uppsala, Sweden). Carboxymethyl-dextran sensor chips (CM5 sensor chips) were equilibrated with running buffer (10 mM HEPES, pH 7.4, 0.15 M KCl and 0.001% Tween 20) for a period of 10 min at 25°C at a flow rate of 5 l/min. Four flow cells of a CM5 sensor chip were activated using of a mixture (70 l, 1:1) of 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (200 mM) and N-hydroxy-sulfosuccinimide (50 mM) at a flow rate of 5 l/min in a 14 min pulse. SNA and MAA (200 g/ml in 10 mM sodium acetate buffer, pH 4.3) were immobilized separately by injecting over the flow cells 2, 3, and 4 respectively to reach 6500 resonance units (RU). Un-reacted groups were blocked with ethanolamine (1 M, 50 l, pH 8.5) for 7 min. Flow cell 1 was considered as the reference. Dialyzed membrane fractions (10 -100 g/30 l) against coupling buffer (5 mM Na 2 HPO 4 , pH 7.4 and 150 mM NaCl) at 4°C were injected for 300 s at a constant flow rate of 5 l/min. The sample was first injected over the reference surface (flow cell 1) and subsequently on flow cell 2, 3, and 4 with immobilized lectins. Bound ligands were removed with acetate buffer (10 mM, pH 4.3, 30 l) during a regeneration step.
Similarly, interaction of sialylated and non-sialylated OprD proteins (OprD ϩSias or OprD ϪSias ) with ␤-lactam antibiotics was examined by SPR. OprD ϩSias and OprD ϪSias (200 g/ml in acetate buffer) were immobilized separately over the flow cells. Non-covalently bound OprD proteins were removed by two fluxes of HCl (20 mM) for 2 min. ␤-lactam antibiotics (10 -100 M) such as piperacillin and ceftazidime (Sigma) were injected separately onto OprD ϩSias or OprD ϪSias coated CM5-sensor chip at a flow rate of 10 l/min (42,43). Bio-Evaluation 3.0 software (Biacore) was used to analyze sensorgrams. The dissociation constant (K D ) for antibiotic-OprD protein association was calculated as the ratio of the backward (k d ) and forward rates (k a ).
MAA lectin (2.0 mg) was dissolved in coupling buffer (2.0 ml) containing sodium bicarbonate (0.1 M), sodium chloride (0.5 M), pH 8.5. CNBr-activated Sepharose 4B (0.6 g) was socked with HCl (1 mM) for 15 min at 25°C and washed with coupling buffer. MAA solution was mixed with the activated Sepharose 4B beads and kept for overnight on gentle shaker at 4°C. MAA concentration was measured in the supernatant by Bradford method. The active site of Sepharose 4B was blocked with glycine (0.2 M, pH 7.2, 50 l/5 ml bead) for 2 h at 25°C. MAA coupled beads were washed with 40 ml of coupling buffer and 40 ml of acetate buffer (0.1 M sodium acetate, 0.5 M NaCl, pH 4) alternatively, followed by 40 ml of TBS. MAA-Sepharose 4B was stored in TBS at 4°C with sodium azide (0.01%). Similarly SNA was also coupled with CNBr-activated Sepharose 4B. The percentages of coupled MAA and SNA with Sepharose 4B were 83.33% and 87.87% respectively.
The PA membrane protein (2.0 mg) was passed through the SNA-Sepharose and MAA-Sepharose affinity column separately for several times and incubated for overnight at 4°C. The columns were washed with TBS to remove unbound proteins. Bound sialoglycoproteins were eluted with ethylenediamine (20 mM) at 4°C, neutralized immediately with dilute HCl (5:3 ratio) and exchanged with TBS by using Viva spine 6 (GE healthcare, 5 kDa cut off). Purified sialoglycoproteins was stored in presence of protease inhibitors mixture at Ϫ70°C for further use (20,44).
Isoelectric Focusing (IEF)-Purified OprD ϩSias and OprD ϪSias (2 g/100 ml) were incubated for overnight on IPG gel strip (7 cm) with a gradient range of pH 4 -7. IEF was carried out for 20 min at 250 V at linear mode; 2 h at linear mode from 250 to 4000 V; at 4000 V up to 10000 V-h in rapid mode. Separated proteins according to their iso-electric point (pI) were fixed with methanol-acetic acid-water (30: 15:55) and stained by Coomassie. The proteins were desialylated and processed similarly. The pI was determined from the pI of known proteins used as standards (12,45).
Gel pieces were washed with acetonitrile, dried and further rehydrated with trypsin containing digestion buffer. Peptides were extracted from gel, dried, and dissolved in acetonitrile (50%) in trifluoroacetic acid (0.1%). Subsequently, they were spotted on a target MALDI plate using ␣-cyano hydroxy cinnamic acid (CHCA) as a matrix (45,46) and analyzed using MALDI-TOF mass spectrometer (Applied Biosystem, Foster City, CA). Spectra were calibrated using the matrix and tryptic auto-digestion ion peaks of Calmix, a standard mixture of six peptides.
Spectral data were analyzed from PMF in combination with MS/MS spectra by searching against the database using the MASCOT (Matrix Science Ltd., London, UK) version 2.2 and basic local alignment search tool (BLAST) of ABI GPS Explorer software, version 3.6 (Applied Biosystems). For database searching the following parameters were used. Peak list-generating software: 4000 series explorer software version 3.5; taxonomy: all entries; database: MSDB version 2.1.0 dated 27.02.2005; No of entries: Database-MSDB20050227 (1942918 sequences; 629040812 residues); cleavage enzyme: trypsin; variable modifications: oxidation on methionine; fixed modification: carbamidomethylation; missed cleavages permitted: one missed cleavages; minimum signal to noise ratio (S/N): 10; peptide charge: ϩ1; precursor mass tolerance: Ϯ 100 ppm; mass tolerance for the MS/MS search: Ϯ 0.2 Da. Significance of data was selected according to their p value (p Ͻ 0.05) where p is the probability that the observed match in a random event. Therefore Mascot search engine is setting the threshold ions score [-10*Log(p)] on its own based on the type of analysis, number of spectra to be analyzed etc.
To examine the quality and accuracy of data, false discovery rate (FDR) values were determined using Mascot software (Matrix Science). Briefly, raw combined MS and MS/MS data were converted into Mascot generic format (MGF) file using following parameters ( Purification of OprD Proteins by Anion Exchange Chromatography-For outer membrane OprD porin proteins purification (47), PA14, PA Urine , PA Pus , and PA Sputum membrane pellets were separately mixed with Tris-HCl buffer (10 mM, pH 8, 5 ml) containing ␤-octyl glucoside (68 mM), EDTA (5 mM) and protease inhibitor mixture and sonicated (5 pulses, 10 s each). The supernatant obtained after ultracentrifugation (100,000 ϫ g) for 30 min at 4°C contained the membrane proteins. Diethylaminoethyl (DEAE)-anion exchange column (5.0 ml) was equilibrated with buffer containing Tris-HCl (10 mM, pH 8), octa-ethleneglycol dodecyl ether (5 mM) and 1 mM EDTA (Buffer A, 47). Membrane protein (2 mg) was loaded onto the column and incubated overnight at cold. Unbound or loosely adhered proteins were removed by washing with Buffer A. The bound protein was eluted at 4°C using a linear gradient of NaCl (0 -0.50 M) in Buffer A. Eluted fractions (1 ml each/microfuge) were checked by measuring OD at 280 nm. Each twenty fractions were checked on 10% SDS-PAGE through Coomassie brilliant blue staining. OprD protein-enriched fraction was found in #6 -10 microfuges (5.0 ml).
OprD protein-enriched fraction was further re-chromatographed using DEAE-anion exchange column. Accordingly, OprD protein-enriched fraction was exchanged with Tris-HCl (10 mM, pH 8), ␤-octyl glucoside (34 mM), and 1 mM EDTA (Buffer B). Unbound protein was removed by washing with buffer B, bound protein was eluted (1 ml/microfuge) using a linear gradient of NaCl (0 -0.3 M) in buffer B. Eluted OprD protein (#6 -9 tubes) was combined, concentrated, and concentration was estimated by Bradford protein assay (48). The purity of OprD was checked on SDS-PAGE by Coomassie brilliant blue staining using different amounts of protein.
For further confirmation, gel band was excised, digested with trypsin and identified by MALDI-TOF/TOF-MS. PMF and MS/MS combined spectra were compared with MSDB database sequences.
MALDI-TOF MS Analysis-HPLC fraction of lyophilized DMB-sialic acids purified from OprD protein was dissolved in water (2 l) and processed by dried-droplet procedure. An equal volume of 5-dihydroxybenzoic acid (DHB, 10 g/l) in ethanol (60%) as matrix was mixed with DMB-sialic acids. The mixture (0.50 l) was placed on the target and processed as described above (50).
Quantitation of Sias by Fluorimetric Acetyl Acetone Method-The total Sias content of OprD ϩSias or OprD ϪSias (50 g each) of PA14, PA Urine , PA Pus , and PA Sputum was measured by mild oxidation with sodium metaperiodate separately using pure Neu5Ac as standard (51). The fluorogen formed was detected at 510 nm upon excitation at 410 nm with an F-4010 spectrofluorimeter (Hitachi, Tokyo, Japan).
Enzymatic Release of N-Linked Glycans-Sialylated and non-sialylated OprD proteins were purified from PA ϩSias and PA ϪSias respectively. Proteins were de-salted using spin filter (10 kDa cut off, Milli-Pore, Billerica, MA) by Glycocore. N-glycans were isolated from 200 g of protein using peptide N-glycosidase F (New England Biolabs, Ipswich, MA). Briefly, the samples were denatured using denaturing buffer (New England Biolabs) at 100°C for 10 min, followed by blocking of excess SDS using Nonidet P-40 buffer (New England Biolabs) and finally treating the samples with 5 mU of PNGaseF (New England Biolabs) in enzyme reaction buffer at 37°C for 20 h. N-Glycans were purified using sequential passing of the reaction mixture over SepPak C18 (100 mg, 1 ml, Waters, Milford, CA) and Poly-Graphitized-Charcoal cartridge (50 mg, 1 ml, Thermo Scientific, Waltham, MA). N-glycans trapped in the PGC was eluted with 15-30% acetonitrile containing 0.1% TFA and dried down on speed vac. N-Glycans isolated from known amount of proteins were used for their chemical characterization. For monosaccharide analysis by high performance anion-exchange chromatography with pulsed amperometric detection (HPAEC-PAD) 20 g of protein was hydrolyzed using 2NTFA at 100°C for 4 h. Acid was removed by dry nitrogen flush and samples dissolved in known volume of water and injected on Car-boPac PA-1 column attached to PAD detector (ICS3000,Dionexnow Thermo Scientific, Sunnyvale, CA). Sialic acid was done using an ultra performance liquid chromatography with fluorescence (UPLC-FL, Aquity UPLC System, Waters, Milford, MA) on 10 g of protein.
Briefly, samples were hydrolyzed using 2 M AcOH at 80°C for 3 h, followed by removal of the acid in speed vac and 10 kDa spin filtration. The filtrate was dried and reacted with DMB (Sigma Aldrich, Milwaukee, WI) and injected on a BEH-C18 column (2.1 ϫ 50 mm) (Waters). For fluorescent tagging purified N-glycans equivalent to 20 g of protein was used. Fluorophore 2-AB (Sigma Aldrich) was used to reductive couple on the pure N-glycans followed by removal of excess reagents by Glycoclean S-cartridge (Prozyme, Hayward, CA). Fluorescent labeled N-glycans were also identified by HPAEC coupled with an online fluorescent detector (HPAEC-FL). 2-AB-labeled glycans were dissolved inwater and separated by using a PA-1 column (4 ϫ 250 mm; Dionex, Sunnyvale, CA). For high mannose glycans the elution profile was matched with the retention times of 2-AB-labeled N-glycans from RNaseB (Sigma Aldrich) and for sialylated glycans the spectra was compared with N-glycans from bovine fetuin (Sigma Aldrich).
Purified N-glycans isolated from 150 g of protein were characterized by MALDI-TOF mass spectrometry (4800 Plus MALDI-TOF/TOF, AB-Sciex). Prior to doing MALDI, N-glycans were permethylated following by modified Ciucannu and Kereks method (52). Briefly, dried N-glycans were dissolved in anhydrous DMSO (Sigma Aldrich) followed by addition of NaOH slurry in DMSO and Methyl-iodide (Sigma Aldrich). After vigorous stirring for 45 min the reaction was stopped adding ice-cold water and permethylated glycans were extracted using chloroform-water partitioning. The chloroform later containing permethylated glycans were dried using dry nitrogen flush and dissolved in 200 l of methanol l.50 l of sample was dried in speed vac and re-dissolved in 5 l of methanol followed by addition of1 l of water containing 0.1% TFA. 1.2 l of this sample solution was mixed with super-DHB matrix (Sigma Aldrich) in 1:1 ratio and spotted on stainless steel MALDI plate. All the spectra were acquired in positive mode using 337 nm solid state lasers (53).
Chemical Release of O-Linked Glycans-Reductive beta elimination reaction was performed for O-glycan release on 150 g of protein sample. Briefly samples were treated with 50 mM NaOH in presence of 1 M NaBH 4 at 45°C for 16 h on a stirring cum heating block (ReactiTherm, Pierce, Rockford, IL). The samples were cooled over ice-water bath, neutralized with ice-cold 30% acetic acid and desalted using cation-exchange resine (Dowex-50W, BioRad) and lyophilized. Dried sample was co-evaporated three times each with acidified methanol (MeOH: AcOH 9:1 v/v mixture) and MeOH and to remove the borate passed over SepPak C18 cartridge (Waters) and permethylated prior to doing mass spectral analysis using MALDI. Plausible N-and O-glycan structures were searched and annotated by selecting the consortium for functional glycomics (CFG) data base of GlycoWorkbench software version 1.2.4105 (53)(54)(55).
Liposome-swelling Assay-Egg phosphatidylcholine (6.2 mol) and dicetylphosphate (0.2 mol) were dissolved in acetone and dried at the round bottomed flask by vacuum centrifugation. The phospholipids film was resuspended in double distilled water (0.2 ml) and mixed with purified OprD ϩSias /OprD ϪSias or sialidase-treated-OprD ϩSias /OprD ϪSias proteins of PA Urine , PA Pus , and PA Sputum along with PA14 or BSA (10 g) separately. The mixtures were sonicated (5 s pulse, five times) in an ice bath until suspension turn from opaque to translucent. Suspensions were then dried in the same tubes by warming them in a water bath (45°C) and further using vacuum pump via a copper sulfate (CuSO 4 ) tube. Purified channel-forming porin protein OprD was incorporated into the phospholipid bilayer. The liposomes were stored overnight in an evacuated desiccator containing CuSO 4 . Finally, the film was suspended in a solution (0.40 ml) typically containing dextran T-20 (Sigma, 17%, W/V), sodium nicotinamide adenine dinucleotide (NAD, 4 mM, pH 6) and imidazole-NAD (1 mM, pH 6.0) buffer. The suspension in the tubes was kept for 2 h at 25°C without agitation and then gently resuspended by hand shaking. The suspended liposome solutions were then passed through Millipore membrane filter (8 m) to remove larger particles. The filtered suspension (20 l for each tube) was diluted (0.6 ml) with antibiotic (piperacillin or ceftazidime, 20 mM) solution separately. Because of presence dextran T-20 (17%) of OprD ϩSias protein-containing liposome solution increased the turbidity of the liposome suspension after a rapid dilution into the solutions of antibiotics. Because of the osmotic pressure applied by the dextran in the intravascular space, penetration of piperacillin or ceftazidime through liposomes caused the liposomes to swell and the swelling was detected as a decrease in turbidity. The optical density was measured at 400 nm with a spectrophotometer (Shimadzu, Japan) for 90 s (56).
Statistical Analysis-Values show the means Ϯ standard deviations of triplicate results from single representative experiments. For all data shown, three independent experiments were routinely carried out, unless indicated otherwise. Student's t test was used to determine the level of statistical significance.

PA ϩSias Showed Increased Survival Against Antibiotics-
We have reported earlier the presence of linkage-specific Sias on PA. As increased frequency of multi-drug-resistance in PA is very common, here we wanted to address whether the Sias on PA play any role toward ␤-lactam antibiotic resistance. Accordingly, PA were treated with two different types of ␤-lactam antibiotics and their survival was quantitated by colony-forming unit (CFU). Near minimum inhibitory concentration required to inhibit the growth of 50% (MIC 50 , 8 g/ml) of ceftazidime antibiotics, PA ϩSias (5.92 Ϯ 0.25 ϫ 10 5 ) showed higher CFU compared with 1.43 Ϯ 0.18 ϫ 10 5 shown by PA ϪSias (Fig. 1A). In parallel, PA ϩSias with piperacillin (6.79 Ϯ 0.34 ϫ 10 5 ) also showed enhanced CFU as compared with PA ϪSias with piperacillin (3.12 Ϯ 0.21 ϫ 10 5 ) in the vicinity of MIC 50 (15 g/ml) indicating Sias on PA ϩSias may play some crucial role for their resistivity toward antibiotics (Fig. 1B). However, both PA ϩSias and PA ϪSias showed low but similar CFU value (0.26 Ϯ 0.06 ϫ 10 5 ) in presence of higher doses of either ceftazidime (20 g/ml) or piperacillin (50 g/ml). CFU counts indicated that all three clinical isolates namely PA Urine (Fig. 1C, 1D), PA Pus (Fig. 1E, 1F), and PA Sputum (Fig. 1G, 1H) showed similar degree of resistance against those ␤-lactam antibiotics like PA14.
PA ϩSias Membrane Showed Increased Association with Sias-binding Lectins-Encourage by this observation, we aimed to understand the various sialoglycoproteins present on PA ϩSias at molecular level. Accordingly, at the initial phase, we determined the binding capacity of membrane protein isolated from PA ϩSias with two Sias-binding lectins (SNA and MAA). For analysis of such interaction, SNA and MAA were separately immobilized on the activated CM5 sensor chip to reach 6500 RU. The best level of activation of CM5 chips was observed following activation for 14 min with freshly prepared carbodiimide and hydroxysuccinimide buffers, used immediately after mixing.
The sensorgram was obtained by chip activation and immobilization of lectins followed by deactivation of the remaining sites with ethanolamine. A representative sensorgram was shown for immobilization of SNA (supplemental Fig S1). Bac-terial membrane fractions (10 -100 g/30 l) were injected for 6 min over the covalently bound SNA and/or MAA surface separately. Representative sensorgrams showed that both SNA and MAA exhibited a similar pattern and magnitude of binding ( Fig. 2A, 2B). Dose dependent increased association of PA membrane protein (10, 20, 30, 50, and 100 g/30 l) with SNA (648, 872, 1107, 1431, and 1675 RU) as well as MAA (755, 847, 966, 1135, and 1342 RU) respectively was observed. The bacterial membrane-lectin interactions with both SNA and MAA reduced significantly after de-sialylation as represented by a lower RU value indicating specificity of the binding (Inset, Fig. 2A, 2B).
PA ϩSias Exhibited More Number of ␣2,6-linked Compared with ␣2,3-linked Sialylated Glycoproteins-Cell surface sialoglycoproteins of pathogens play essential role in disease biology. To check which glycoproteins of PA become sialylated, all sialoglycoproteins having ␣2,6and ␣2,3-linkages were purified from total membrane protein of PA ϩSias by using SNA and MAA as their respective affinity ligands. Approximately 8.9 Ϯ 1.23% and 7.45 Ϯ 1.38% of total membrane proteins are ␣2,6-linked and ␣2,3-linked of purified sialoglycoproteins respectively.
Equal amount (10 g) of total membrane proteins from PA ϩSias , purified ␣2,6and ␣2,3-linked sialoglycoproteins as well as unbound proteins from affinity column were separated by SDS-PAGE (Fig. 3A). As expected a few bands in purified fractions were coincided with total membrane proteins whereas a few of them were absent in unbound fraction. Purified SNA/MAA bound sialoglycoproteins (120 g) were separately analyzed by 2D-SDS-PAGE and images were taken. There were 22 spots corresponding to ␣2,6-linked sialoglycoproteins (Fig. 3B) whereas ␣2,3-linked sialoglycoproteins showed only 14 spots (Fig. 3C) indicating PA membrane fractions contained more number of ␣2,6-linked compared with ␣2,3-linked sialylated glycoproteins.
Linkage-specific Sialoglycoproteins Identified on PA ϩSias -Molecular identification of these linkage-specific sialoglyco-proteins is necessary to understand their role in PA ϩSias . Accordingly, Coomassie-stained 2D SDS-PAGE gel spots were excised, destained and treated with trypsin. Tryptic fragments of all the spots were analyzed by MALDI-TOF/TOF-MS. Using the MASCOT software, the resulting combined PMF (supplemental Figs S6 -S31) and MS/MS (supplemental Figs S58 -S83) spectrum was compared with the MSDB sequence database. A few ␣2,6-linked, ␣2,3-linked (Table I) and both ␣2,3and ␣2,6-linked sialylated proteins (Table II) were matched with the MSDB database sequences of PA using search criteria "All entries" that will consider mammalian, bacteria, fungi etc. together. Additionally, to examine the quality and accuracy, false discovery rate (FDR) of all data was calculated. FDR values emphasized that 22 out of 26 proteins were identified with zero false positive identifications (Table I and II). Information about the identified peptides for each protein and their individual scores have documented in supplemental Figs S32-S57. Although we believed that these identified proteins possibly are sialylated based on their purification using lectin-affinity chromatography, however, details investigation of each protein is needed to further validate this observation.
Purification and Characterization of OprD ϩSias Protein-Antibiotic resistance of PA is very common in hospitalized patients especially in underdeveloped country. After careful monitoring of these linkage-specific sialoglycoproteins identified on PA ϩSias , we have selected OprD porin proteins, having both ␣2,3and ␣2,6-linked sialic acids, to understand the detailed functional role of such acquired sialic acids with respect to the entry of ␤-lactam antibiotics inside the cell.
Accordingly, in search for the possible reasons behind antibiotic resistance of PA in protein level, we have purified OprD proteins form PA ϩSias (OprD ϩSias ) and PA ϪSias (OprD ϪSias ) separately by using classical method like DEAE-cellulose as a matrix of anion exchange chromatography. The enriched fraction of OprD protein was further purified by changing buffer composition and salt gradient using another column. The yield of purified OprD ϩSias and OprD ϪSias was 0.045 Ϯ 0.08 mg and 0.056 Ϯ 0.10 mg from 6 ϫ 10 14 cells, corresponding to 2.25 Ϯ 0.07% and 2.80 Ϯ 0.08% respectively. Coomassie brilliant blue stained SDS-PAGE (10%) of purified OprD ϩSias and OprD ϪSias appeared as a single band (supplemental Fig S2A). Different amounts of purified protein also showed single band on SDS-PAGE (data not shown). These proteins exhibited single band in IEF ( supplemental Fig S2B).
For further confirmation, gel band (lane 4, Fig. 4A) was excised, digested with trypsin and identified by MALDI-TOF/ TOF-MS. Combined PMF (Fig. 4B) and MS/MS (supplemental Fig S3) spectra were compared with MSDB database se-quences using search parameter "All entries" which matched with OprD of P aeruginosa. Database search identified fifteen proteins from these spectra according to the protein score (supplemental Fig S4). Among these 15 proteins, protein score of OprD precursor is 669 as compare with other protein score being around 50 -60. Sequence coverage is 24% (Fig.  4C, shown in red color) and protein identity is S23771. Matched peptide sequences are also shown in Fig. 4D.
OprD ϩSias is Highly Sialylated-Before and after neuraminidase treatment, OprD ϩSias showed variation in mobility when analyzed on 10% SDS-PAGE (supplemental Fig S2A). Difference in molecular mass (1.25 kDa) were more prominent in 7.5% gel (Fig. 5A). On contrary, OprD ϪSias showed negligible difference in molecular mass upon neuraminidase treatment indicating presence of less Sias. Such dissimilarity in mobility between OprD ϩSias and OprD ϪSias further suggested that they may differ also in charge and/or structure. Charge heterogeneity of OprD ϩSias and OprD ϪSias was subsequently investigated by IEF. In general, both the proteins showed a single band confirming their purity and the presence of a single molecular entity (supplemental Fig S2B). Interestingly, neuraminidase-treated OprD ϩSias showed increase in pI from 4.95 (lane 2) to 6.05 (lane 3) demonstrating the presence of more Sias. On the other hand, pI of OprD ϪSias is only 6.10, which did not change appreciably after neuraminidase treatment suggesting minimal presence of Sias (supplemental Fig. S2B, lane 5).
Presence of Enhanced Sias on OprD ϩSias was Confirmed by Different Analytical Methods-To check the status of glycosidically bound Sias, OprD ϩSias were subjected to acid hydrolysis and purified through Dowex cation and anion exchange columns. The eluted free sialic acids were DMB derivatized and separated by fluorimetric-HPLC (Fig. 5B). The chromatogram of DMB-Sias of OprD ϩSias exhibited well-re-solved intense peak of N-acetyl neuraminic acid (Neu5Ac), co-migrating with Neu5Ac, N-glycolyl neuraminic acid (Neu5Gc), 5-N-acetyl-9-O-acetyl neuraminic acid (Neu5,9Ac 2 ) derived from BSM and used as internal standards. In parallel OprD ϪSias was similarly processed. As expected, DMB-Sias from OprD ϪSias exhibited a small peak of Neu5Ac corroborating presence of negligible Sias.
Additionally, fluorimetric-HPLC fraction corresponding to Neu5Ac was analyzed by MALDI-TOF MS (Fig. 5C) and yielded the expected signal for the sodium cationized molecular ion having m/z at 448.7 convincingly demonstrated their occurrence on OprD ϩSias .
In parallel, Sias on OprD ϩSias were shown using orcinolstained TLC plates, further demonstrating the presence of Neu5Ac (supplemental Fig. S2C). The R F values corresponded to standard Neu5Ac and free Sias purified from BSM. No such spot was observed with OprD ϪSias . Presence of Sias on OprD ϩSias or OprD ϪSias was quantitated by fluorimetric acetyl acetone method (Fig. 5D). OprD ϩSias showed significantly (p Ͻ 0.0001) higher amount of Sias (1.98 Ϯ 0.15 g) compared with OprD ϪSias (0.15 Ϯ 0.02 g). Additonally, we have also purified sialylated and nonsialylated OprD proteins from three more clinical isolates of Occurrence of Neu5Ac showed 23-fold higher in sialylated OprD as compared with non-sialylated analog. However, total amount of monosaccharide (excluding sialic acid) as determined by HPAEC-PAD present was comparable within the range of 30.19 -36.21 ng/g of OprD ϩSias and OprD ϪSias .
Predicted N-glycan structures from their respective masses were shown in the profiles which revealed the presence of high mannose, complex bi-antennary and different extent of sialylation on tri-antennary type of glycans on OprD ϩSias (Fig.  6A). On the other hand N-glycans isolated from OprD ϪSias is devoid of sialylation (except a minor contribution from a mono-sialylated glycan) (Fig. 6B). The presence of high mannose and sialylated branched N-glycans were also confirmed by HPAEC-PAD of 2-AB tagged N-glycans isolated from Three sialylated (m/z 2792.39, 3602.74, and 3963.90) N-glycan structures of OprD ϩSias were observed (Fig. 6A). However, OprD ϪSias showed only one sialylated N-glycans signal corresponding m/z 2880.40 (Fig. 6B).
O-Glycan data also confirmed the presence of mono (m/z 895.58) and di-sialylated Core-1 type structures (m/z 1256.79) in OprD ϩSias . However such sialylated O-linked glycans were absent in OprD ϪSias (Fig. 6C-6D). Therefore, we may suggest that one of the sialylglycoprotein, OprD, may have mammalian N-and O-glycan structures.
OprD ϩSias Showed Reduced Association with ␤-lactam Antibiotics-To investigate whether these enhanced sialylation of OprD porin play any role in the binding with ␤-lactam antibiotics, we initially checked the association of two representative drugs e.g. piperacillin (Fig. 7A-7B) or ceftazidime ( Fig. 7C-7D) with OprD ϩSias proteins by SPR analysis. Proteins were separately immobilized on activated CM5 sensor chips to reach 6500 RU. To demonstrate the patterns and magnitude of binding, different amounts (10 -100 M) of piperacillin or ceftazidime were injected over OprD ϩSias immobilized sensor chips separately. OprD ϪSias was always used for comparison.

Permeability of Piperacillin or Ceftazidime Decreased
Through OprD ϩSias -Searching for the crucial role of sialic acids found on this porin protein (OprD ϩSias ) from PA14 in the penetration of antibiotics inside the bacteria, we checked permeabilization of piperacillin (Fig. 8A) or ceftazidime (Fig.  8B) through artificially formed liposome membrane with OprD ϩSias by liposome swelling assay. Decrease in OD was observed upon the dilution of OprD ϪSias or sialidase treated-OprD ϪSias or sialidase treated-OprD ϩSias protein-containing liposomes in either piperacillin or ceftazidime. In contrast, little decrease of absorbance was noticed upon the dilution of OprD ϩSias or BSA containing liposomes upon addition of antibiotics. Moreover, membrane permeability of antibiotics through liposome using sialylated and non-sialylated OprDs from other three clinical isolates of PA PA Urine , PA Pus , and PA Sputum (Fig. 8C, 8E, 8G) demonstrated similar results with piperacillin. These three strains also exhibited comparable results with ceftazidime (Fig. 8D, 8F, 8H). This result indicated that non-sialylated porin is highly capable to permeabilize antibiotics through them. Because of presence of sialic acid on the porin protein, it hindered to bind antibiotics leading to inefficient uptake of antibiotics which possibly responsible for bacterial resistivity toward ␤-lactam antibiotics.

DISCUSSION
Both ␣2,3and ␣2,6-linked sialic acids found on the PA's surface help them to resist from the serum complement deposition (19). Correlation of sialylation of glycoproteins in PA and their survival in host is relatively untouched field. Therefore, we considered it may be essential to identify these sialylated glycoproteins that might play an important molecular determinant on PA for their survival.
The main achievements of the current investigations include purification of a few sialoglycoproteins of PA, their molecular identification, characterization and more impor- FIG. 7. OprD ؉Sias showed decreased interaction with ␤-lactam antibiotics. Association of OprD ϩSias with piperacillin (A) and OprD ϩSias with ceftazidime (C) or OprD ϪSias with piperacillin (B) or OprD ϪSias with ceftazidime (D) was examined by SPR as described in materials and methods. OprD ϩSias or OprD ϪSias (200 g/ml) were immobilized in CM5 sensor chip. Several doses (0, 10, 20, 50, and 100 g/ml labeled as 1, 2, 3, 4, and 5 respectively) of antibiotics were allowed to interact with OprD ϩSias -and OprD ϪSias -immobilized chip at a flow rate of 10 l/min. The sensorgram was evaluated by the BiaEvaluation 3.0 software. tantly established the role of one such sialoglycoprotein (OprD) on PA. Treatments of PA ϩSias as well as purified OprD ϩSias with two commonly used ␤-lactam antibiotics (ceftazidime and piperacillin), demonstrated that Sias might play an essential responsibility for their uptake during infection and thereby implement a new strategy for the successful survival of PA within host.
The PMF spectrum of the purified linkage-specific sialoglycoproteins is compared with the MSDB sequence database and the sequence homology matched with PA origin. A few such identified sialoglycoproteins are outer membrane protein OprD, OprF, OprG, flagellin type B etc. are known to execute different survival strategies of PA. In general, ␤-barrel shaped, channel forming porins (OprD and OprF) play the central role in outer membrane permeability of drug (57). OprF plays the crucial role to maintain structural integrity of PA (58), whereas OprG protein is tightly regulated by anaerobiosis and contributes to the cytotoxicity of this bacterium during the early infection (59). OprD is specific protein through which PA uptakes basic amino acids, peptides and ␤-lactam antibiotics (60). PA can colonize in the various surfaces to form biofilm with the help of flagellin type B protein and becomes impervious to therapeutic concentrations of many antibiotics (61). We have demonstrated sialylation in all these vital proteins. Therefore, it is expected that Sias must play some key role to carry out these fundamental functions of PA.
Other major important outer membrane proteins are OprB, OprE, OprI, OprL, OprP, and iron repressible outer membrane proteins already reported in PA (62). Interestingly, under the experimental conditions, they did not show any sialylation. Glycosylation of these proteins demands future investigation.
Among these different mechanisms of antibiotics-resistance, we investigated the involvement of Sias of OprD porin protein to find out the possible role of Sias toward the uptake of drug in this bacterium. Sequence study revealed that OprD porin is a highly sialylated glycoprotein having both ␣2,6and ␣2,3-linked Sias.
N-and O-linked glycosylation are the two most common forms of glycosylation in proteins. Glycoproteomic analysis of OprD ϩSias /OprD ϪSias reveals that they may have mammalian like N-and O-glycan structures. Though comparable amount of total carbohydrates were found in OprD ϩSias and OprD ϪSias , a 23-fold higher Sias was observed only in OprD ϩSias . Such a huge amount of Sias may be responsible for vast modulation of its structure. The enhanced presence of high-mannose contained N-glycans in OprD ϩSias possibly playing an additional role for their proper folding and dynamic stability through hydrogen bonds. Literature search reveals that mammalian like N-glycans are also present on E. coli (63). Such high-mannose containing N-glycans are also been reported on influenza virus (64).
Increased presence of Sias in OprD ϩSias has further been convincingly demonstrated by TLC, fluorimetric-HPLC and MALDI-TOF-MS. Isoelectric point of OprD ϩSias has been revealed in acidic region and a huge shift of pI after removal of Sias established enhanced sialylation compared with OprD ϪSias . Increased sialylation in OprD ϩSias possibly causes structural modification of this protein.
Structural characterization revealed higher glycans in OprD ϩSias compared with OprD ϪSias . It may be envisaged that all the OprD proteins are not glycosylated in similar extent, however it needs further detailed investigation.
Next we addressed whether sialylation of OprD play any role in resistance of two commonly prescribed ␤-lactam antibiotics (ceftazidime and piperacillin). SPR analysis showed that ceftazidime-OprD ϩSias and piperacillin-OprD ϩSias interactions are relatively weak compared with OprD ϪSias suggesting Sias are possibly creating some problem for antibiotics to interact with porin protein for their entry. Both the antibiotics also showed higher membrane permeability through liposome containing only OprD ϪSias . In contrast, lower capability of OprD ϩSias revealed important role of Sias in drug permeabilization in PA.
This was further corroborated in live cell experiment. Near the MIC 50 values of both the drugs, antibiotics-treated PA ϩSias showed higher CFU count with respect to PA ϪSias . The main deference between PA ϩSias and PA ϪSias is sialic acids, therefore Sias definitely play some crucial role for hindering antibiotics uptake in PA that leads to resistance against drugs.
We have used four clinical isolates of PA to show the relationship between sialylation of protein and the antibiotic resistance. All four clinical isolates showed comparable rate of antibiotic resistance. Additionally, sialylation of purified OprDs from these four isolates exhibited similar degree of inhibition of drug permeabilization. These observations indicated a possible link between sialylation of OprD and the antibiotic confrontation in PA. This may be a general trend as all four clinical isolates showed comparable degree of resistance against antibiotics.
In conclusion, extensive glycoproteomic analysis gave us some information on sialylated N-and O-glycan compositions of purified OprD porin protein which may be responsible for their structural modulation; hence functional impairment. Both sialylated PA and purified OprD showed more resistance toward ␤-lactam antibiotics. This is because of lower penetration capability of drugs through this highly sialylated porin protein of PA ϩSias . This may be an alternative drug resistance mechanism of PA.
To the best of our knowledge, this is the first report where sialic acids acquired by OprD protein have been assigned a vital function for antibiotics uptake in four different clinical isolates of PA. The findings may help to design newer drug which can enter cells freely even in presence of Sias or block sialylation or cleave sialic acids. However consequence of such study needs further in depth investigation.