Escherichia coli Common Pilus (ECP) Targets Arabinosyl Residues in Plant Cell Walls to Mediate Adhesion to Fresh Produce Plants*

Background: Bacterial fimbriae mediate binding to host tissue through specific interactions. Results: ECP interacts with arabinosyl residues in pectin and other plant cell wall components. Conclusion: ECP-arabinan interactions facilitate binding of E. coli to plant hosts. Significance: The prevalence of arabinan targets in produce plants together with ECP expression may explain the association of pathogenic bacteria in edible plants. Outbreaks of verotoxigenic Escherichia coli are often associated with fresh produce. However, the molecular basis to adherence is unknown beyond ionic lipid-flagellum interactions in plant cell membranes. We demonstrate that arabinans present in different constituents of plant cell walls are targeted for adherence by E. coli common pilus (ECP; or meningitis-associated and temperature-regulated (Mat) fimbriae) for E. coli serotypes O157:H7 and O18:K1:H7. l-Arabinose is a common constituent of plant cell wall that is rarely found in other organisms, whereas ECP is widespread in E. coli and other environmental enteric species. ECP bound to oligosaccharides of at least arabinotriose or longer in a glycan array, plant cell wall pectic polysaccharides, and plant glycoproteins. Recognition overlapped with the antibody LM13, which binds arabinanase-sensitive pectic epitopes, and showed a preferential affinity for (1→5)-α-linked l-arabinosyl residues and longer chains of arabinan as demonstrated with the use of arabinan-degrading enzymes. Functional adherence in planta was mediated by the adhesin EcpD in combination with the structural subunit, EcpA, and expression was demonstrated with an ecpR–GFP fusion and ECP antibodies. Spinach was found to be enriched for ECP/LM13 targets compared with lettuce. Specific recognition of arabinosyl residues may help explain the persistence of E. coli in the wider environment and association of verotoxigenic E. coli with some fresh produce plants by exploitation of a glycan found only in plant, not animal, cells.

E. coli O157:H7, one of the most important causative agents of fresh produce-associated outbreaks, is often linked with contaminated lettuce and spinach (11,12).
In mammalian hosts, bacterial adherence is often mediated by lectins present at the fimbrial tip that bind to complementary carbohydrates on the surface of the host tissues. Type 1 fimbriae and P fimbriae are the best characterized in the chaperone-usher family, encoding tip adhesins FimH and PapD, respectively. They recognize ␣-D-mannosylated proteins and ␣-D-galactopyranosyl-(134)-␤-D-galactopyranoside receptor epitope in the globoseries of glycolipids, respectively (13,14). Conversely, no specific adhesin targets in plant tissue have been elucidated either for phytopathogenic bacteria that encode chaperone-usher fimbriae or for human pathogenic bacteria that can colonize plants as secondary hosts.
The E. coli common pilus (ECP), 3 originally termed meningitis-associated and temperature-regulated (Mat) fimbria, was first identified in newborn meningitis and septicemia E. coli isolate IHE3034 (O18:K1:H7) when it was grown at 20°C (15). The ecp operon is ubiquitous across E. coli and even conserved for some other enteric species (15)(16)(17)(18)(19). ECP belongs to the chaperoneusher family encoded by the ecpRABCDE operon where EcpA encodes the pilin domain and EcpD encodes the polymerized tip adhesin. Unusually for classical tip adhesins, EcpD can be polymerized independently, which requires an N-terminal extension in EcpD, or with the major pilin domain (18). Several roles have been described for ECP, including binding to cultured human epithelial cells (16,17,20), colonization of infant mice (21), and biofilm development through interorganelle binding of EcpA (22). The regulator EcpR represses the flagellar master operon flhDC, supporting the role for EcpA in biofilms (23). Expression of ECP is strain-dependent in E. coli: some isolates belonging to the B2 phylogenetic group are able to produce these fimbriae in conditions mimicking those in the intestine or at 20°C, whereas others do not retain this capability (24).
Plant cell walls are complex configurations of structures composed mostly of carbohydrates present in polysaccharides and highly glycosylated proteins. Polysaccharides are represented by three different families: cellulose, hemicellulose, and pectin. Cellulose is present as long chains of ␤-1,4-glucose, and hemicelluloses are branched polysaccharides containing backbones of neutral sugars. Pectins are defined by the presence of uronic acids divided in four polysaccharide "domains": homogalacturonan (HG), rhamnogalacturonan I (RG-I), rhamnogalacturonan II (RG-II), and xylogalacturonan (25,26). The polysaccharides are linked together in varying abundances with HG as the most abundant followed by RG-I (27). Cell walls also contain abundant hydroxyproline-rich glycoproteins (HRGPs), a superfamily that encompasses extensins, proline-rich proteins, and arabinogalactan proteins (AGPs). HRGPs are mainly O-glycosylated with arabinose and galactose (28,29).
In the present study, we tested the hypothesis that bacteria interact with plant tissue by targeting specific glycans. We identified targets for two different ECP variants in plants by using high throughput plant glycan arrays and enzyme-linked immunosorbent assay (ELISA). We were able to capitalize on the extensive body of work for ECP/Mat from newborn meningitis and septicemia E. coli strain IHE3034 (O18:K1:H7) but expanded this to investigate the role of ECP from the human pathogen E. coli O157:H7 (strain Sakai). The interactions were characterized with plant polysaccharides and from functional in planta adhesion assays.

EXPERIMENTAL PROCEDURES
Bacterial Strains and Media-E. coli O157:H7 strain Sakai (Shiga-toxin negative) (30) and E. coli K-12 strain JT1 (31) were grown in either Luria-Bertani (LB) broth or rich defined MOPS supplemented with 0.2% glucose, thiamine, and essential and nonessential amino acids (32). Antibiotics were included where necessary to maintain transformed plasmids or for selection with adherence assays at the following final concentrations: 50 g ml Ϫ1 kanamycin, 12.5 g ml Ϫ1 chloramphenicol, 50 g ml Ϫ1 ampicillin, 10 g ml Ϫ1 tetracycline. Induction of ecp genes in recombinant E. coli strains was carried out with 5 M isopropyl ␤-D-thiogalactopyranoside. All media, antibiotics, and inducers were purchased from Sigma-Aldrich. E. coli strain JT1 was selected to overexpress ECP fimbriae. This strain lacks flagella and type 1 fimbriae, and it encodes a copy of the ecp cluster even though it belongs to serogroup K-12, which does not contain strong, active promoters for the expression of native ecp (24).
Cloning and Mutagenesis-ecpA (ECs0323) and ecpA-D (ECs0323-0320) deletions were constructed using allelic exchange as described previously (33). Primers used for crossover PCR are listed in Table 1. A PstI site in the ecpA upstream sequence required that the PCR products were blunt endcloned via T4 polynucleotide kinase into the pTOF24 vector. A Flip recombinase target-flanked tetracycline cassette was subcloned into the NotI site introduced into pTOF24, creating allelic exchange vectors pAH002 for ecpA and pAH003 for ecpA-D. The promoter region of ecpR (440 bp of the 5Ј-UTR) was PCR-amplified from E. coli O157:H7 Sakai genomic DNA with primers 0324.5.XbaI and 0324.3.XbaI (Table 1) and cloned into pKC026 using XbaI, creating the transcriptional fusion pAH001. Two overexpression constructs for E. coli, ecpA-D and ecpA-E, were generated by cloning the regions into pSE380 (Table 2), generating pYR006 and pYR007, using HindIII-HindIII and XbaI-HindIII, respectively. Enzymes were purchased from New England Biolabs (Ipswich, MA) or Roche Applied Science. Plasmid clones were confirmed by appropriate restriction digests and Sanger sequencing. Mutants were confirmed by phenotype and diagnostic PCR.
ECP Preparation and Antibodies-ECP from pYR006 (EcpA-E from E. coli O157:H7 Sakai) and pMat3 (EcpA-E from E. coli O18:K1:H7 IHE3034) transformed in E. coli JT1 were isolated as described with modifications (34). In short, bacteria were cultured in LB medium supplemented with isopropyl ␤-D-thiogalactopyranoside and ampicillin at 37°C for 16 h in static conditions, harvested by centrifugation, resuspended in cold Trisbuffered saline (TBS), detached from the bacterial cell by using a blender (three times for 30 s), and centrifuged twice at 4,000 ϫ g for 30 min. Ammonium sulfate was added slowly into the supernatant containing the fimbriae with vigorous stirring to achieve two-thirds saturation. After an overnight incubation at 4°C, the fimbriae were harvested by centrifugation at 15,000 ϫ g for 20 min at 4°C and then suspended in cold TBS ϩ 0.5% deoxycholate. The preparation was subjected to a sucrose gra- dient ultracentrifugation (60 -10%) in an Optima L-80 XP centrifuge for 16 h at 20,000 rpm at 4°C in an SW41 rotor and then fractioned in 1-ml lots. The fractions were visualized for density and absorbance at 280 nm. The fractions containing ECP were pooled, dialyzed against water, and freeze-dried. To confirm purity of the isolated ECP, the samples were subjected to SDS-PAGE after denaturation at low pH (35). The proteins were separated by 12.5% SDS-PAGE and electrotransferred to Hybond-P membrane (GE Healthcare). Protein concentrations were determined using the BCA protein assay (Thermo Fisher Scientific Inc., Waltham, MA). The predicted size of E. coli EcpA is 21 kDa. A mock fimbrial preparation was prepared in an identical manner using E. coli JT1 as a control. The anti-ECP fimbrial serum derived from E. coli strain JTI overexpressing pMat3 was obtained by immunizing rats according to standard procedures by Genosphere Biotechnologies (France). Antibodies used in this study were rabbit anti-E. coli (Abcam, Cambridge, UK; Ab137967), anti-rat IgG Alexa Fluor 568-conjugated antibodies, and anti-rabbit and anti-rat horseradish peroxidase (Sigma-Aldrich). For anti-plant polysaccharide/glycoprotein (Plantprobes, Leeds, UK) specificities and details, see Table 3. The presence of fimbriae was confirmed by electron microscopy. Fimbrial preparations were spotted (10 l) to carbon-coated copper grids and incubated for 1 min, and the grid was transferred to a 100-l drop of sterile distilled water and incubated for 2 s. Without drying, the grid was transferred to a 100-l drop of 2% phosphotungstic acid, pH 7.0 and incubated for 2 min. Excess liquid was then removed, the grid allowed to dry, and the images were obtained with a JEOL JEM-1400 electron microscope (Japan). Spinach Protein Enrichment-Extracts enriched in glycoproteins from spinach leaves or roots were obtained as described by Karlova et al. (37). Briefly, plant material was ground in liquid nitrogen and thawed in extraction buffer containing 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.5% Triton X-100, and 1ϫ plant protease inhibitor mixture (Sigma-Aldrich). The samples were incubated on ice for 30 min and then centrifuged at 400 ϫ g at 4°C for 3 min. Protein concentrations were determined using the BCA protein assay. The proteins were kept at Ϫ80°C for no longer than 1 week.

Plants and Growth Conditions-Spinach
Cell Wall Polymer Extraction-Lettuce and spinach leaves or roots were ground by mortar and pestle to a fine powder in liquid nitrogen. Six volumes of ethanol 70% were added to 1 volume of powder and incubated at room temperature for 10 min with rotation and then centrifuged at 3,000 ϫ g 10 min. This step was repeated five times. The pellet was rinsed twice with acetone and dried. This alcohol-insoluble residue preparation was then followed by sequential extraction using two solvents: 50 mM CDTA, pH 7.5 and 4 M NaOH. The two extraction solvents used are known to solubilize pectins and noncellulosic polysaccharides, respectively.
Glycan Array Printing and Screening-Glycan array printing was performed as described (38), and arrays were screened as described previously (4). For detailed information on all oligosaccharide samples, see Table 4. The results shown are based on six individual experiments. A competition assay with BSA-conjugated arabinoheptaose (Ara7) was performed by co-incuba-  tion with purified ECP (pYR006) at 20 g/ml and was based on two independent experiments. ELISA and mAb Screening for Extract Characterization and Bacterial Interaction-ELISAs were performed in 96-well microtiter plates (Nunc, Maxisorb) coated with spinach pro-teins or arabinan at 50 g/ml and at a 1:10 dilution of plant polysaccharide extract (in 0.1 M NaHCO 3 buffer, pH 9.6) in a volume of 100 l incubated at 4°C overnight, similar to the concentrations used by others (39). The plates were washed three times with TBS and blocked with protein-free blocking TABLE 4 List of plant derived BSA-conjugated oligosaccharides and layout of the glycan array X refers to xylose, L refers to galactose, and G refers to glucose. buffer (Thermo Fisher Scientific Inc.) for 1 h at room temperature. Plates were again washed three times with TBS and then incubated for 2 h with 100 l of bacteria at a concentration of ϳ2 ϫ 10 7 cfu/ml (A 600 adjusted to 0.02 in TBS) or primary antibodies (Table 3) (at 1:20 for all plant polysaccharides/glycoproteins). After 2 h, the wells were washed three times with TBS and incubated with a rabbit anti-E. coli antibody (Abcam) diluted at 1:500 in TBS or incubated directly with a horseradish peroxidase anti-rat IgG conjugate (diluted at 1:1000). For detection of bacteria, after 1 h, the wells were washed three times with TBS and incubated with 100 l of horseradish peroxidase anti-rabbit IgG conjugate (diluted at 1:1000). After 1 h, all secondary antibodies were washed three times with TBS. The color reaction was developed with 2,2Ј-azino-bis(3-ethylbenzthiazoline-6-sulfonic acid) solution: 22 mg of 2,2Ј-azinobis(3-ethylbenzthiazoline-6-sulfonic acid) (Sigma-Aldrich) diluted in 100 ml of citrate buffer (50 mM sodium citrate, 0.05% H 2 O 2 , pH 4.0). The absorbance at A 405 was measured on a Spec-traMax M5 microplate reader (Molecular Devices, Sunnyvale, CA). Extracts enriched in glycoproteins were pipetted onto nitrocellulose membrane as immunodot blots at 100 g each, allowed to dry, and probed with the monoclonal antibodies (Table 3) essentially as described for the ELISA with the following modifications: nonspecific interactions were blocked with 3% skimmed milk, primary antibodies were used at 1:100 dilution, the secondary anti-rat HRP conjugate was used at 1:5000 dilution, and the blots were washed with PBS containing 0.08% Tween 20. Gum arabic was used at 20 g as a positive control. ECL substrate was used to detect binding (Millipore, Billerica, MA) and was exposed to ECL Hyperfilm for 10 min.
Monosaccharide Analysis-Monosaccharide analysis of cell walls was carried out as described (40). Acid hydrolysis (2 M trifluoroacetic acid for 1 h at 120°C) was carried out, and the samples were loaded on a high performance anion exchange chromatography column and run on a Dionex ICS3000 system with a Dionex AS autosampler fitted with a 25-l loop. The monosaccharides fucose, galactose, glucose, xylose, mannose, galacturonic acid, and glucuronic acid were separated on 3 ϫ 150-mm PA-20 column with a 3 ϫ 30-mm guard column. Eluents were ultrapure water (18.2 megaohms), 200 mM NaOH, and 1 M sodium acetate as described. The monosaccharides arabinose and rhamnose were separated on 4 ϫ 250-mm PA-100 column with a 4 ϫ 50-mm guard column eluted isocratically at 1 ml min Ϫ1 with 200 mM NaOH. The flow rate was 0.4 ml min Ϫ1 (40). The data are presented as a percentage of the total (mg/mg dry weight) to allow direct comparison.
Bacterial Adherence Assays-Adherence assays were performed as described (4). Briefly, plant tissue was aseptically separated, washed, and incubated with a bacterial suspension at a concentration of ϳ2 ϫ 10 7 cfu/ml (A 600 adjusted to 0.02 in PBS) for 2 h at 18°C. The tissue was then washed three times and macerated, and the bacteria were enumerated on selective medium.
Confocal Microscopy-Plant tissues were analyzed as described (4). Briefly, plant tissues were fixed, deparaffinized, treated with enzymes, and incubated successively with proteinfree blocking buffer and antibodies. Primary antibodies were used at 1:50 and secondary (anti-rat conjugated to Alexa Fluor 568) at 1:5000 (Sigma-Aldrich). The sample was incubated for 10 min with Calcofluor white and then mounted with Dako fluorescent mounting medium (Dako, Carpentaria, CA). Fresh leaves were stained with polyclonal antiserum against ECP for 1 h (1:500) or propidium iodide. After three washes in TBS, the leaf samples were incubated for 45 min with an Alexa Fluor 568-conjugated anti-rat antibody (1:5000). Imaging was performed on a Leica TCS-SP2 AOBS microscope (Leica Microsystems, Germany). Photoshop CS software (Adobe Systems) was used for postacquisition image processing.
Sequence and Phylogenetic Analysis-The ecpD ORF derived from E. coli isolate Sakai was used to carry out a BLASTn analysis against a range of E. coli sequences (supplemental Table 1). The ORFs were identified, and amino acid sequences were derived using a bioinformatics program (CLC Bio, Aarhus, Denmark). These sequences were aligned using ClustalW (41). Sequences that did not encode a full-length ecpD ORF were excluded from subsequent alignment.
Statistical Analyses-The means Ϯ S.E. were calculated for each experimental group, and the statistical significance was evaluated with Student's t test or one-way analysis of variance. The results were considered as significant for a p value Յ0.05.

ECP Binds to Arabinose Oligosaccharides-Recombinant
ECP was purified from E. coli JT1 overexpressing ECP from high copy number plasmid pMat3 (ECP derived from newborn  DECEMBER 5, 2014 • VOLUME 289 • NUMBER 49 meningitis and septicemia E. coli strain IHE3034 (O18:K1:H7)) (15) to generate a polyclonal antibody. The antibody reacted with a protein of the predicted size of the main structural subunit, EcpA (Fig. 1, A-C). Induction of pMAT3 resulted in fimbrial elicitation with fine, hairlike structures (Fig. 1, D and E). The antibody reacted to a similar level to ECP derived from E. coli O157:H7 isolate Sakai, which encodes an identical copy of the mature peptide of the main structural subunit, EcpA, but not to the mock fimbrial preparation from E. coli JT1 (pSE380). Although the host strain encodes a native ecp cluster, it belongs to serogroup K-12, which expresses minimal or non-detectable levels of ECP (24). ECP derived from E. coli O157:H7 strain Sakai was cloned and purified. We used this strain because it was isolated from a massive foodborne outbreak in Japan in 1996 associated with white radish sprouts (42). Purified ECP fimbriae were used to screen a plant glycan array printed with oligosaccharides coupled to BSA including galacturonate, galactose, arabinose, mannose, xylose, xyloglucan, and glucose and probed with the ECP antibody (detailed in Table 4). ECP Sakai (pYR006) and ECP IHE3034 (pMat3) interacted only with BSA conjugates with arabinose oligosaccharides from three (␣-(135)-linked arabinotriose) to seven (␣-(135)-linked arabinoheptaose) residues ( Fig. 2A). Although detection required a relatively long exposure time (10 min), a specific signal was observed for the higher concentration of the conju-gated oligosaccharides (1 mg ml Ϫ1 ) ( Fig. 2A). No signal was detected with the negative control (no ECP). The signal intensity appeared to be proportional to the oligosaccharide complexities: the longer the oligosaccharides, the stronger the signal, and no signals were detected from complexes smaller than arabinotriose. To confirm the specificity, two arabinose-specific antibodies were used, LM6 and LM13. LM6 recognizes linear chains of (135)-␣-arabinan from pectin and some AGPs (39,43), whereas LM13 is specific to linear chains of (135)-␣arabinan from pectin but importantly only to oligosaccharides longer than arabinotriose (38,44,45). As expected, LM6 recognized arabinose residues on the glycan arrays of all sizes, whereas LM13 recognized residues from arabinotriose to arabinoheptaose present on the array (Fig. 2B). ECP was found to share the same specificity for longer arabinans as LM13. The ECP interaction was abrogated following co-incubation with free BSA-Ara7, confirming the interaction by competitive inhibition (Fig. 2B).

E. coli Common Pilus Binds to Arabinose
Arabinofuranosidase (AF) is a glycoside hydrolase that specifically targets ␣-L-arabinofuranosides containing (133) and/or (135) linkages found in RG-I as well as arabinoxylans and arabinogalactans. Pretreatment of the glycan array with AF prevented binding of purified ECP Sakai (pYR006) to any residues and severely affected LM13 binding to Ara7 so that a signal was only just visible after a long exposure time (Fig. 3A). In  Table 4 for detail and a pictorial map. B, glycan arrays probed with anti-arabinan antibodies LM6 and LM13 or co-incubated with purified ECP Sakai (pYR006) and Ara7. Numbers 1-7 refer to the number of arabinosyl residues in each oligosaccharide.
contrast, LM6 detection occurred for all BSA-arabinose conjugates. No obvious interaction was detected between ECP and any other purified components on the array.
Bacterial adhesion to BSA-conjugated arabinotriose and Ara7 mediated by ECP was confirmed by ELISA using E. coli JT1 overexpressing ECP IHE3034 (pMat3). Enzymatic treatment with AF to remove arabinans resulted in a significant decrease in the interaction between ECP and Ara7 (p Ͻ 0.05) but not for arabinotriose. The result was consistent with a decrease in the interaction of LM13 antibody with Ara7 and no interaction with arabinotriose (Fig. 3B). This assay was carried out with whole bacteria expressing ECP, which presumably accounts for the background level of binding to the substrate. As expected, LM6 antibody interacted with both arabinans, although enzymatic treatment was obviously not complete.
ECP Interaction with Spinach Arabinosyl Residues in Pectin-Because spinach (S. oleracea) accounts for a large proportion of E. coli O157:H7 fresh produce outbreaks (46 -49), we assessed ECP binding to spinach plant cell wall constituents. Pectin and hemicellulose components were extracted from total spinach polysaccharides following CDTA and NaOH treatment, respectively, and probed with LM6 for pectin (arabinose) and LM15 for hemicellulose (xyloglucans) (Fig. 4A). As expected, arabinose was only detected in the pectin fractions of spinach leaves and roots, and xyloglucans were present only in hemicellulose fractions. To demonstrate tropism of ECP for pectin, adhesion of E. coli JTI expressing either variant of ECP (pMat3 or pYR006 with pSE380 as the negative control) was assessed on pectin fractions using hemicellulose as a negative control (Fig.   4B). Commercial purified branched and debranched long arabinan polymers (Ͼ110 residues/molecule), which were recognized by LM6 and LM13 (data not shown), were included to better characterize the interaction. Both ECP variants (Sakai and IHE3034) bound to spinach pectin and not to hemicellulose or branched arabinan, demonstrating a tropism for pectin. Adherence of ECP to isolated pectin from roots was Ͼ2-fold higher than on leaves presumably because this tissue type is richer in arabinan as shown with LM6 (Fig. 4B). Interestingly, ECP Sakai expressed from E. coli JT1 (pYR006) was significantly more adherent to debranched arabinan ECP IHE3034 (pMat3) (p Ͻ 0.05) as well as pSE380. The ability of bacteria expressing ECP to inhibit antibody adhesion (LM1, LM6, and LM13) to the CDTA extract of spinach leaves was tested. The presence of bacteria expressing ECP Sakai or ECP IHE3034 reduced LM6 and LM13 binding by ϳ35% (Fig. 4C).
ECP Interaction with Spinach Arabinosyl Residues in Glycoproteins-An ELISA was performed with spinach leaf or root extracts enriched in glycoproteins and E. coli JT1 overexpressing ECP IHE3034 (from pMat3) or the empty plasmid (pSE380) as a control (Fig. 5A). Expression of ECP IHE3034 resulted in a ϳ2-fold increase in adherence to the extracts from both leaves and roots compared with the control. Enzymatic treatment with AF abolished the increase in binding of ECP-expressing bacteria to levels similar to those seen with the pSE380 control, whereas treatment for mannans (mannanase) or pectins (pectin lyase (PL)) did not affect the interaction. The presence of arabinans in the spinach extracts was detected with antibodies LM6 and LM13 and was reduced by AF enzymatic treatment (Fig. 5B). There was no recognition of extensin using the LM1 antibody. Characterization of the extracts with the cell wall antibodies confirmed the presence of AGPs (with LM2 and JIM8) and showed the presence of the major pectin polysaccharide domain, homogalacturonan, partially methyl-esterified, using JIM7 (Fig. 5C). This is not unexpected as it is not possible to generate completely pure glycoprotein preparations. LM13 was detected albeit in relatively low amounts. A positive control for AGPs, gum arabic, was strongly bound by LM2 (not recognized by LM13 or LM6). The binding data together with the presence of antibody epitopes and response to enzymatic treatment indicate an interaction of ECP with AGPs.
Functional Binding of ECP in Planta-ECP knock-out mutants were constructed to determine whether ECP plays a functional role for E. coli O157:H7 in planta on roots and leaves. The main structural subunit, EcpA, is involved in biofilm formation and cell-cell binding; indeed, imaging of purified ECP fimbriae indicated organelle interactions and bundling (Fig. 1, D and E). Therefore, to distinguish the biofilm effect from a specific interaction, adherence to plants for an ecpA mutant was compared with that for a whole cluster ecpA-D mutant where structural and adhesin genes were removed. Deletion of ecpA in E. coli O157:H7 Sakai resulted in an ϳ10fold reduction in the number of bacteria recovered from spinach leaves that was further reduced with the removal of the whole cluster ecpA-D (p Ͻ 0.05) (Fig. 6A). A significant reduction in adherence for the ecpA and ecpA-D mutants was also observed in roots, although the magnitude of the decrease was not as great. Complementation of the ecpA-D mutant (with  DECEMBER 5, 2014 • VOLUME 289 • NUMBER 49

JOURNAL OF BIOLOGICAL CHEMISTRY 34355
pYR007) restored the WT phenotype in leaves (p Ͻ 0.001) and roots (p Ͻ 0.005). Oddly, the mutant complemented with the control plasmid pSE380 became more adherent on spinach roots compared with Sakai WT, which we postulated was due to the effect of isopropyl ␤-D-thiogalactopyranoside (the inducing agent) elsewhere on the genome. Addition of isopropyl ␤-Dthiogalactopyranoside to E. coli O157:H7 Sakai WT and both ecp mutants resulted in an increase in bacterial adherence (data not shown), confirming this idea.
Pretreatment of spinach roots with AF, arabinanase, PL, or mannanase reduced the ability of E. coli O157:H7 Sakai to adhere, although the greatest difference was observed with PL treatment (p Ͻ 0.005), indicating the possibility of multiple interactions, including arabinose-dependent ECP (Fig. 6B). The PL buffer (pH 10) was included because the high pH has been demonstrated previously to degrade pectin backbones that break down under basic conditions (50). Moreover, arabinan can contain various esters (e.g. acetyl and feroyl) that are also base-sensitive (51,52).
In plant leaves, stomatal guard cell walls are rich in pectin and contain arabinan epitopes (53,54). Expression of functional ECP fimbriae significantly increased the number of bacteria associated with stomata for both E. coli O157:H7 isolate Sakai and E. coli JT1 overexpressing ECP Sakai (from pYR006) (Fig. 7,  A and B). To identify unspecific molecular binding mediated by ECP, natural wool was used as an inert, biological substrate to mimic plant root morphology in the absence of any plant glycans (Fig. 7C). There was no significant difference in the number of E. coli JT1 Ϯ ECP (pMat3 or pSE380) recovered from wool (p ϭ 0.721) in contrast to a significant increase exhibited by ECP IHE3034 (pMat3) on spinach roots (p Ͻ 0.05).
Identification of LM13 Epitopes in Spinach-The distributions of LM13 (long arabinans) and LM6 (total arabinans) epitopes in spinach roots were identified using antibody probes together with enzymatic treatment (Fig. 8). Cortical cells appeared to be enriched in pectin as indicated by LM6 and   (HRGPs) (B). Detection was carried out with or without (native) enzymatic treatments: AF, mannanase (Mase), and PL. Spinach root extracts were characterized from dot blots with the monoclonal antibodies (C). The labels refer to each antibody (Table 3), and "LM2 (gum)" refers to the interaction with gum arabic as a positive control. Significance levels are as follows: ***, p Յ 0.001; **, p Յ 0.005; *, p Յ 0.05; NS, non-significant. Error bars represent S.E. FIGURE 6. Functional adherence of ECP to plant tissue. A, bacteria recovered from spinach leaves and roots following a 2-h adherence assay for E. coli O157:H7 Sakai WT, mutants ⌬ecpA and ⌬ecpA-D, and the complemented ⌬ecpA-D mutant with plasmid pYR007 or the control pSE380. The average number of bacteria is expressed as log 10 cfu/g of fresh tissue. B, bacteria recovered from spinach roots following a 2-h adherence assay with AF, arabinanase (Arase), PL, and PL buffer treatments for E. coli O157:H7 Sakai WT. Significance levels are as follows: ***, p Յ 0.001; **, p Յ 0.005; *, p Յ 0.05; NS, non-significant. Error bars represent S.E. DECEMBER 5, 2014 • VOLUME 289 • NUMBER 49 LM13 detection and AF and PL treatment effects (Fig. 8, D, E, G,  and H). Although AF treatment reduced signals from all of the antibody probes, PL treatment had a greater effect on LM13 detection. JIM7 (homogalacturonan) (55) and PL buffer treatment alone were included as controls. Epitopes for LM13 and LM6 were also identified in spinach guard cells (Fig. 8, M and  N). It was not possible to assess enzymatic treatment on guard cells because of their destructive effects on stomata (53,54).

E. coli Common Pilus Binds to Arabinose
Comparative adherence assays for E. coli O157:H7 WT and ⌬ecpA-D showed a more pronounced role for ECP to spinach compared with lettuce (L. sativa) (Fig. 9) (the data are expressed as a ratio to compare plant species). To determine whether increased ECP-dependent adherence was a result of greater availability and/or abundance of the ECP target in spinach, cell wall polysaccharides from the roots and leaves of both plant species were analyzed using a range of antibodies to detect epitopes for pectin (LM5, LM6, and LM13), hemicellulose (LM21 and LM25), and extensin (LM1) ( Table 3). As expected, the pectin probes reacted most strongly with CDTA-derived extracts, whereas the hemicellulose probes reacted with NaOH-derived extracts (Fig. 9, C and D). LM13 and LM6 gave signals in both plant species, but higher levels were detected in spinach. The trend was the same for both leaves and roots albeit with differences in the complement of the specific residues. Monosaccharide quantification of plant extracts, including arabinose (Table 5), determined from HPLC analysis correlated with the epitope presence (Fig. 9E).
Functional Expression of ECP in Planta-E. coli O157:H7 has been shown to encode an active promoter site for the ecp operon (24). To observe ECP transcription and production in FIGURE 7. Interaction of E. coli with guard cells. A, confocal microscopy images of a stoma of a fresh spinach leaf inoculated with E. coli O157:H7 Sakai WT, E. coli O157:H7 Sakai ⌬ecpA-D, or E. coli JT1 expressing ECP Sakai (pYR006) or the control plasmid (pSE380). Bacteria are false colored blue, and chloroplast autofluorescence is purple. B, quantification of the different bacteria strains from five different stomata, expressed as the average number of bacteria per stoma. C, bacteria recovered from wool or spinach roots following a 2-h adherence assay for E. coli JT1 expressing ECP IHE3034 (pMat3) or the control plasmid (pSE380). The average number of bacteria is expressed as log 10 cfu/g of fresh tissue. Significance levels are as follows: ***, p Յ 0.001; *, p Յ 0.05; NS, non-significant. Error bars represent S.E. planta, we examined spinach leaves inoculated with E. coli O157:H7 Sakai transformed with a plasmid-encoded ecpR-GFP transcriptional fusion after anti-ECP detection. Expression of ecpR-GFP (green) was evident in most cells, and elicitation of ECP fimbriae (red) was also observed. Differences in the green fluorescence levels suggested that ecpR transcription may be heterogeneous in planta (Fig. 10). A different plant species was evaluated (a) to confirm that ECP expression is not restricted to spinach plants and (b) to observe any co-localization with particular plant tissues. N. benthamiana line mGFP5-ER expresses GFP fused to an endoplasmic reticulum protein (36). Inspection of a guard cell shows the green network of the endoplasmic reticulum together with bacteria expressing ecpR (green) and ECP fimbriae (red) adjacent to a chloroplast (purple) and on the leaf epidermis (Fig. 10B).
EcpD Sequence Variation-ECP fimbriae are almost ubiquitous in E. coli and are also encoded by related members of the Enterobacteriaceae, although the EcpA (main structural subunit) and EcpD (adhesin) coding sequences are not completely conserved. Alignment of 66 E. coli EcpD amino acid sequences  highlighted variable residues ( Table 6). The types of substitutions for the most frequent changes included conserved (e.g. L145M); non-polar aliphatic to polar, uncharged (e.g. R232L); and vice versa (e.g. A117T). The mature peptide of EcpD from E. coli from isolates Sakai and IHE3034 (used in this study) contains three amino acid substitutions.

DISCUSSION
Identification and characterization of the molecular targets underpinning the adhesion of bacteria (phytopathogenic or human pathogenic) to plants is limited compared with that for mammalian and animal hosts (57,58). So far, only an interaction between E. coli flagella and ionic lipids has been described where flagella can intercalate into plant plasma membranes (4). The interaction between ␣-(135)-L-arabinose in plant cell walls and ECP derived from a well known human pathogen associated with outbreaks from fresh produce is the first description of its type. L-Arabinose is synthesized by the plant kingdom where it is very common, whereas only some bacterial species synthesize D-arabinose polysaccharides (59). L-Arabinose is present in homo-as well as heteropolysaccharides and is O-linked on HRGPs (60). Interaction with ECP required oligomerization of at least three residues of arabinose, and specificity correlated with the antibody LM13. LM13 primarily recognizes linear epitopes of (135)-␣-L-arabinans present in RG-I polysaccharides of pectin as well as other arabinose residues, e.g. in AGPs where the structural context is correct (45). The overlap in recognition shows that ECP recognized pectic arabinans and potentially O-glycosylated hydroxyproline proteins (AGPs), which contain arabinose and galactose, ranging from a single residue to up to 75 arabinogalactan residues (28,61).
Stomatal guard cells, which are rich in pectic arabinan (53,54) and have been shown to be a target for bacteria (4,62), were targeted by bacteria expressing ECP. ECP fimbriae bound preferentially to pectin fractions in polysaccharides extracted from spinach compared with long oligomers of arabinan purified from beet roots that were either branched or unbranched (63). Sugar beet arabinan consists of an ␣-(135)-linked backbone of L-arabinosyl residues that are either single or double substituted (64). Treatment of unbranched arabinan to remove (132)-and (133)-␣-L-arabinofuranosyl branches revealed a specificity of ECP Sakai , indicating a potential affinity for ␣-(135)-linked L-arabinosyl residues and longer chains of arabinan. This may explain why ECP Sakai was slightly more adherent on pectin polysaccharides extracted from spinach roots than ECP IHE3034 . The significance of structural context may explain differences in the extent of ECP binding to plant extracts compared with that on the purified oligosaccharides.
Sequences of EcpD, the adhesin, are extremely well conserved in all E. coli genomes and especially between the serotype O157:H7 strains. However, differences in the EcpD mature protein may be sufficient to alter receptor specificity as has been observed for type 1 fimbriae (65). ECP tropism for pectic arabinan was confirmed with pectate lyase treatment, which degrades homogalacturonan chains, the main component of pectin polysaccharides (66), and releases the arabinose-substituted pectin polysaccharides. The high potency of pectate lyase activity means that it may have some arabinase activity even at low concentrations. Furthermore, because of the many intermolecular connections between cell wall components, enzymatic treatments that do not target arabinan specifically may nonetheless impact its availability. ECP interaction with spinach extracts enriched in glycoproteins was only affected by arabinofuranosidase treatment. The data indicate that despite the presence of HG backbone in the extracts pectic arabinans were not abundant and that ECP also interacts with AGPs. The apparent lack of pectate lysase activity on the HG component can be accounted for by the requirement for Ca 2ϩ for catalysis; Ca 2ϩ ions are absent in the methylated form of HG (67). In summary, it appears that arabinan from pectin polysaccharides and possibly plant glycoproteins play a role in E. coli ECP-dependent adhesion in planta. In turn, the distribution of arabinan could affect tropism for the bacteria: toward guard cells in leaves and cortical cells within root tissue just as is seen for PapD-dependent tropism to galactans in renal tissue.
Functional binding of ECP to plant tissue appears to be conferred by the adhesin, EcpD, and is supported by the presence of the main structural subunit, EcpA. Because ECP did not significantly contribute to binding to wool (non-plant substrate with a morphology broadly similar to plant root), we concluded that a biofilm-associated function of EcpA is negligible during short term incubation on plants. EcpD can be secreted and polymerized/oligomerized in the absence of EcpA (18) as an "afimbrial" adhesin, although the processes may not be as efficient as in the presence of EcpA. Moreover, as with type 1 fimbriae, EcpA itself could play a role in the specificity (68). Our data show that the EcpA fimbrial structure is required for presentation of EcpD to target cell wall arabinans during interactions with plant hosts. The fact that bacterial binding to plant tissue was not completely inhibited by removal of the ECP cluster or enzymatic treatment points to a role for other functional adherence factors, including flagella, to mediate plant interactions.
Functional adherence of ECP in planta occurred at submammalian host temperatures and was coupled with expression of the ecpR regulator. This supports the reports of low temperature expression at 20°C (15,24). Variation in the level of ecpR expression may in turn reflect the amount of ECP organelle expressed from this polycistronic operon. This finding is in line with the reported alternative transcriptional start sites for ecpR (24).
Both ECP variants from E. coli isolates Sakai and IHE3034 demonstrated enhanced adherence to spinach root polysaccharides compared with leaves, which supports previous findings of increased colonization in the roots (69). Furthermore, higher levels of ECP-mediated adherence of E. coli O157:H7 occurred in spinach compared with lettuce, which may be a result of the  reduced arabinan in lettuce. It is of note that verotoxigenic E. coli are isolated from spinach plants at higher rates compared with lettuce (70,71) and that consumption of raw spinach has risen by 180% in the United States from 1992 to 2005 (49). This information together with ECP specificity for arabinose may go some way to explain why verotoxigenic E. coli outbreaks are more frequently associated with spinach. Demonstration of adherence mediated by flagella together with specific recognition of a plant oligosaccharide by an adhesin (4) suggests a stepwise process of E. coli adherence to plant cells. First, a broad molecular recognition occurs between flagella and plasma membranes without any particular tropism. This is then followed by a specific, adhesin-mediated adherence, e.g. ECP to cell wall pectin polysaccharides. Exploitation of such an abundant plant-specific glycan may aid the spread of E. coli in the environment, and given the ubiquity of ECP, it could be a reasonably common mechanism of binding to plants (and for related enterobacteria). However, differences in ecp cluster expression and EcpD amino acid sequence are likely to affect both the occurrence of binding and the affinity.
The ability of mammalian pathogenic bacteria to target an intermediate host for survival and transmission is not without precedent: the spread of Vibrio cholerae, a well known human pathogen, through freshwater and marine environments (72,73) is dependent on its interaction with chitin polysaccharide in zooplankton (74,75). Furthermore, pathogenic strains of V. cholerae show increased fitness in bivalves compared with environmental isolates (76). Similarly, ECP-arabinan interactions may play an important role in allowing pathogenic E. coli to use plants as a secondary host and as a vehicle to spread to other primary hosts such as cattle or humans.