Outer Membrane Protein of Gut Commensal Microorganism Induces Autoantibody Production and Extra-Intestinal Gland Inflammation in Mice

Gut commensal microorganisms have been linked with chronic inflammation at the extra-intestinal niche of the body. The object of the study was to investigate on the chronic effects of a gut commensal Escherichia coli on extra-intestinal glands. The presence of autoimmune response was diagnosed by autoantibody levels and histological methods. Repeated injection of E. coli induced mononuclear cell inflammation in the Harderian and submandibular salivary glands of female C57BL/6 mice. Inflammation was reproduced by adoptive transfer of splenocytes to immune-deficient Rag2 knockout mice and CD4+ T cells to mature T cell-deficient TCRβ-TCRδ knockout mice. MALDI TOF mass spectrometry of the protein to which sera of E. coli-treated mice reacted was determined as the outer membrane protein A (OmpA) of E. coli. Multiple genera of the Enterobacteriaceae possessed OmpA with high amino-acid sequence similarities. Repeated injection of recombinant OmpA reproduced mononuclear cell inflammation of the Harderian and salivary glands in mice and elevation of autoantibodies against Sjögren’s-syndrome-related antigens SSA/Ro and SSB/La. The results indicated the possibility of chronic stimuli from commensal bacteria-originated components as a pathogenic factor to elicit extra-intestinal autoimmunity.


Introduction
The structure of the human gut microbiota is established early in life, and thereafter maintains a symbiotic relationship with their host in an anaerobic environment for decades [1,2]. Escherichia coli, a facultative anaerobe from the Proteobacteria phylum and a key member of the Enterobacteriaceae family [3], colonize the intestine of human infants within hours of birth [4], by their ability to respire oxygen in the intestine of newborns [5,6]. Although early colonization of E. coli is indispensable for the development of oral tolerance [7], Enterobacteriaceae constitute only a small fraction of less than 1% of the gut microbiota in healthy adults [8] due to restricted carbohydrate sources that they need for growth, for which they compete with the obligate anaerobe counterparts [9]. Nitrate can be used for anaerobic respiration in Enterobacteriaceae, thus conferring growth advantage to E. coli by a high protein diet lifestyle [10][11][12][13]. Dysbiosis of increased facultative anaerobes exceeds commensal tolerance and stimulates innate immune activation beyond the intestinal niche [14], such as atherosclerosis [14,15]. Several of the bacterial components have been implicated in the development of autoimmunity, including the microbial-von Willebrand factor type A domain protein (vWFA) [16] and the flagellin FliC [17] in sialoadenitis of Sjögren's syndrome (SS), and the outer membrane protein OmpC in the synovial membrane of rheumatoid arthritis [18].

Effect on Extra-Intestinal Gland Inflammation by Systemic Treatment of Mice with Bacterial Components
To examine whether chronic exposure of bacteria or bacterial cell wall components could induce autoimmunity in mice by systemic immunization, six weeks old mice were treated with E. coli, peptidoglycan (PGN), muramyl dipeptide (MDP), lipoteichoic acid (LTA), lipopolysaccharide (LPS) or phosphate buffered saline (PBS), by intraperitoneal injection once a week for a total of eight weeks. Harderian glands and salivary glands were histologically examined fifteen weeks after the final treatment. Harderian glands of E. coli-treated mice showed infiltration of inflammatory cells ( Figure 1B,M), in comparison to no inflammation in those of mice treated with PBS alone ( Figure 1A). The incidence of at least one foci of inflammation in the Harderian gland occurred in all seven mice of the E. coli-treated group ( Figure 1O). Inflammation in the Harderian gland was not detected in mice receiving PGN, MDP, LTA or LPS ( Figure 1C-F). Inflammatory cells were present in the salivary glands of E. coli-treated mice ( Figure 1H,N). Incidence of at least one foci in the salivary gland occurred in three of a total of seven mice in the E. coli-treated group ( Figure 1O). Inflammation in the salivary gland was not detected in mice treated with PGN, MDP, LTA, LPS, or PBS ( Figure 1G,I-L). Quantification of the inflammatory foci showed that foci were scattered within one sample organ ( Figure 1O).

Cell-Mediated Immunity in Bacteria-Treated Mice
E. coli induced accumulation of inflammatory cells in the Harderian glands (Figure 2A,I,L). No inflammatory cells were observed in mice treated with PBS alone ( Figure 2E,L). Infiltrating cells in the Harderian glands of the E. coli-treated mice were in part CD3 positive cells detected by immunohistochemistry ( Figure 2D), which were not detected in the PBS-treated mice. IgG1 positive cells were scattered within the stromal tissue of the Harderian glands of mice by E. coli treatment ( Figure 2H). IgG1 cells were not observed in the Harderian glands of mice treated with PBS.
Adoptive transfer experiments were undertaken to elucidate the role of cell-mediated immune responses in mice induced by repeated E. coli injection. Adoptive transfer of bulk splenocytes to Rag2 KO mice from E. coli-treated donor mice developed accumulation of inflammatory cells in the Harderian glands ( Figure 2B,J,L), compared to those from donor mice receiving PBS alone ( Figure 2F,L). Similarly, Harderian glands of TCRβ-TCRδ KO mice subsequent to adoptive transfer of spleen-derived  N) fifteen weeks after the final inoculation of E. coli were visualized by hematoxylin and eosin (HE) staining and were compared to those of mice treated with PBS alone (A,G, respectively). A representative inflammatory focus in one mouse of seven mice per treatment group is presented (A-L). A representative area of 2 mm × 2 mm is indicated as a white-lined square in a histologic section of the Harderian gland (M) and the salivary gland (N), respectively, visualized at low magnification (4×). Scale bar, 100 µm. Quantification of inflammatory foci in Harderian and salivary glands. The number of inflammatory foci, each containing at least 50 mononuclear cells, was defined as # foci (O). Data presented are individual values (n = 7) plotted with a bar indicating the group mean. Differences between the groups of measurements were analyzed by Kruskal-Wallis with post-test comparison. Focus scores of the Harderian glands in E. coli-treated group were significantly higher (p < 0.05) compared to all other measurement groups (O). PBS: phosphate buffered saline; PGN: peptidoglycan; MDP: muramyl dipeptide; LTA: lipoteichoic acid; LPS: lipopolysaccharide.
( Figure 2H). IgG1 cells were not observed in the Harderian glands of mice treated with PBS.
Adoptive transfer experiments were undertaken to elucidate the role of cell-mediated immune responses in mice induced by repeated E. coli injection. Adoptive transfer of bulk splenocytes to Rag2 KO mice from E. coli-treated donor mice developed accumulation of inflammatory cells in the Harderian glands ( Figure 2B,J,L), compared to those from donor mice receiving PBS alone ( Figure  2F,L). Similarly, Harderian glands of TCRβ-TCRδ KO mice subsequent to adoptive transfer of spleen-derived CD4 + T cells from E. coli-treated donor mice showed infiltration of inflammatory cells ( Figure 2C,K,L), but not those from donor mice receiving PBS alone ( Figure 2G). Figure 2. Histology of mice injected with E. coli. Harderian glands of E. coli-treated (A,D,H,I) and PBS-treated (E) C57BL/6 wild type mice were histologically examined fifteen weeks after the final injection. Harderian glands of Rag2 KO mice (B,F,J) and TCRβ-TCRδ KO mice (C,G,K) were histologically examined one week after adoptive transfer of splenocytes (B,F,J) or CD4 + T cells (C,G,K) from E. coli-inoculated (B,C,J,K) or PBS-inoculated (F,G) C57BL/6 donor mice. Inflammatory infiltrates were examined by HE-staining (A-C,E-G,I-K) and immune-staining for CD3 (D) and IgG1 positive cells (H). A representative inflammatory focus in one mouse of six to seven mice per treatment group is presented. Inflammatory foci are indicated inside dotted lines in a representative 4 mm 2 histology section of the Harderian gland of one mouse from each treatment group (I-K). Scale bar, 100 µm. Quantification of inflammatory foci in the Harderian and salivary glands of wild type (n = 7), Rag2 KO (n = 6), and TCRβ-TCRδ KO (n = 6) mice (L). Data presented are individual values plotted with a bar indicating the group mean. Differences between the groups of measurements were analyzed by Kruskal-Wallis with post-test comparison. Focus scores of the Harderian glands in E. coli-treated group were significantly higher (p < 0.05) compared to all other measurement groups (L).

Identification of OmpA from E. coli as a Representative Immunogenic Protein in Mice
We sought to identify the cellular component of E. coli that contributes to the generation of autoimmunity. Cell surface extracts of E. coli separated on 2-D gels showed numerous numbers of proteins at molecular weights between 25-100 kDa and pI 6-11 ( Figure 3A). Western blotting of 2-D PAGE gel-transferred nitrocellulose membrane probed with sera of the experimental dacryoadenitis mice showed an immune-reactive protein migrating at 41 kDa/pI 6.24 ( Figure 3B), which was not detected with sera of mice inoculated with PBS alone ( Figure 3C). Analysis of the 41 kDa/pI 6.24 spot excised from the 2-D PAGE-gel of E. coli surface fraction by MALDI-TOF mass spectrometry was identified as OmpA (theoretical Mw/pI of 35,081.25/5.42), with MS-fit sequence coverage of 38% ( Figure 3D). To elucidate the immunogenicity of OmpA in vivo, sera of mice treated once a week during eight weeks with E. coli was validated by the production of antibodies against OmpA. Serum anti-OmpA antibody titer levels were elevated by systemic E. coli injection and were increased over time up to 10 months after completion of the final injection of E. coli ( Figure 3E).

Analysis of the Amino Acid Sequences of OmpA from Phylum Proteobacteria
Phylogenetic analysis of OmpA from phylum Proteobacteria showed that the bacteria from ten genera of the Enterobacteriaceae family (Escherichia, Salmonella, Citrobacter, Cronobacter, Enterobacter, Klebsiella, Pantoea, Serratia, Erwinia, and Yersinia) possessed OmpA with highly similar amino acid residues ( Figure 4A,B). Residues corresponding to OmpA G80 to P305 of E. coli ATCC 25922 were aligned for analysis by using ClustalW, gap open penalty 1, gap extention penalty 0.05. The similarity scores of OmpA amino acid sequences of Enterobacteriaceae compared to E. coli were 69-92%, indicating that the amino acid residues were highly conserved among Enterobacteriaceae. Excluding the Enterobacteriaceae family, other orders of the class Gammaproteobacteria (genera

Analysis of the Amino Acid Sequences of OmpA from Phylum Proteobacteria
Phylogenetic analysis of OmpA from phylum Proteobacteria showed that the bacteria from ten genera of the Enterobacteriaceae family (Escherichia, Salmonella, Citrobacter, Cronobacter, Enterobacter, Klebsiella, Pantoea, Serratia, Erwinia, and Yersinia) possessed OmpA with highly similar amino acid residues ( Figure 4A,B). Residues corresponding to OmpA G80 to P305 of E. coli ATCC 25922 were aligned for analysis by using ClustalW, gap open penalty 1, gap extention penalty 0.05. The similarity scores of OmpA amino acid sequences of Enterobacteriaceae compared to E. coli were 69-92%, indicating that the amino acid residues were highly conserved among Enterobacteriaceae.
Excluding the Enterobacteriaceae family, other orders of the class Gammaproteobacteria (genera Pasteurella, Gallibacteruim, Glaesserella, Bibersteina, Acinetobacter, Actinobacillus, Haemophillus, Vibrio, and Pseudomonas) were less similar to E. coli in the OmpA sequence ( Figure 4A). Of the Proteobacteria phyla, classes Betaproteobacteria (genera Burkholderia and Bordetella, Figure 4A) and Alphaproteobacteria (genus Rickettsia) were also less similar to E. coli in the OmpA sequence.  Figure 4A). Of the Proteobacteria phyla, classes Betaproteobacteria (genera Burkholderia and Bordetella, Figure 4A) and Alphaproteobacteria (genus Rickettsia) were also less similar to E. coli in the OmpA sequence.

Induction of Extra-Intestinal Gland Inflammation by OmpA
OmpA of E. coli was cloned and expressed as an N-terminal 6×His-tagged recombinant protein by IPTG induction ( Figure 3F, lanes 4 and 5). Purification of 6×His-OmpA ( Figure 3F, lane 2) using Ni 2+ -affinity column resulted in elution of a single band, which reacted to OmpA-specific monoclonal antibody [28] ( Figure 3F, lane 3). To investigate on the effect of the immunogenic OmpA protein, recombinant OmpA was injected to mice intraperitoneally once a week for eight weeks. Inoculation of the full length OmpA resulted in inflammatory cell infiltration of the Harderian glands at fifteen weeks ( Figure 5A,E) and at twenty-two weeks ( Figure 5F,H), after the final injection. Inflammation of the Harderian glands was observed in all eight mice per group, whereas no inflammatory cells were detected in those after inoculation of PBS ( Figure 5I). Infiltrating cells in the Harderian glands of the OmpA-treated mice were in part CD3-( Figure 5B,G), CD4-( Figure 5C), and CD8-( Figure 5D) positive cells detected by immunohistochemistry, which were not detected in those of the PBS-treated mice. OmpA induced inflammation in the salivary glands in three of the eight mice per group, but not in the pancreas ( Figure 5I).

Induction of Extra-Intestinal Gland Inflammation by OmpA
OmpA of E. coli was cloned and expressed as an N-terminal 6×His-tagged recombinant protein by IPTG induction (Figure 3F, lanes 4 and 5). Purification of 6×His-OmpA ( Figure 3F, lane 2) using Ni 2+ -affinity column resulted in elution of a single band, which reacted to OmpA-specific monoclonal antibody [28] ( Figure 3F, lane 3). To investigate on the effect of the immunogenic OmpA protein, recombinant OmpA was injected to mice intraperitoneally once a week for eight weeks. Inoculation of the full length OmpA resulted in inflammatory cell infiltration of the Harderian glands at fifteen weeks ( Figure 5A,E) and at twenty-two weeks ( Figure 5F,H), after the final injection. Inflammation of the Harderian glands was observed in all eight mice per group, whereas no inflammatory cells were detected in those after inoculation of PBS ( Figure 5I). Infiltrating cells in the Harderian glands of the OmpA-treated mice were in part CD3-( Figure 5B,G), CD4-( Figure 5C), and CD8-( Figure 5D) positive cells detected by immunohistochemistry, which were not detected in those of the PBS-treated mice. OmpA induced inflammation in the salivary glands in three of the eight mice per group, but not in the pancreas ( Figure 5I).

Cytokine Production in Sera of Mice with Extra-Intestinal Inflammation Induced by OmpA
Levels of cytokines in mice sera were measured, since systemic inflammatory responses may be able to trigger extra-intestinal gland inflammation. IFN-γ and IL-17A inflammatory cytokine levels and IL-1β innate inflammatory cytokine level were up-regulated in the sera of OmpA-and E. coli-treated mice, compared to those of the PBS-inoculated control mice. However, E. coli or OmpA administration did not affect IL-10, IL-5, and TNF-α production in sera ( Figure 6).

Cytokine Production in Sera of Mice with Extra-Intestinal Inflammation Induced by OmpA
Levels of cytokines in mice sera were measured, since systemic inflammatory responses may be able to trigger extra-intestinal gland inflammation. IFN-γ and IL-17A inflammatory cytokine levels and IL-1β innate inflammatory cytokine level were up-regulated in the sera of OmpA-and E. coli-treated mice, compared to those of the PBS-inoculated control mice. However, E. coli or OmpA administration did not affect IL-10, IL-5, and TNF-α production in sera ( Figure 6).

Effect of Autoantibody Production by OmpA
Anti-SSA/Ro and anti-SSB/La antibodies are autoantibodies elevated in sera of patients with systemic autoimmune diseases such as rheumatoid arthritis, lupus, and SS [30,31]. Thus, we examined whether OmpA-treated mice produced autoantibodies similar to those in autoimmune patients. Serum levels of anti-SSA/Ro antibodies and anti-SSB/La antibodies were elevated in OmpA-and E. coli-inoculated mice, compared to those inoculated with PBS ( Figure 7). The data

Effect of Autoantibody Production by OmpA
Anti-SSA/Ro and anti-SSB/La antibodies are autoantibodies elevated in sera of patients with systemic autoimmune diseases such as rheumatoid arthritis, lupus, and SS [30,31]. Thus, we examined whether OmpA-treated mice produced autoantibodies similar to those in autoimmune patients. Serum levels of anti-SSA/Ro antibodies and anti-SSB/La antibodies were elevated in OmpA-and E. coli-inoculated mice, compared to those inoculated with PBS ( Figure 7). The data together with the histology of the inflammation in the extra-intestinal glands suggested that our model mouse had developed systemic autoimmune responses. Figure 6. Cytokine production in mice injected with OmpA. Levels of IL-1β, IL-5, IL-10, IL-17A, IFN-γ, and TNF-α in sera of PBS (n = 8)-, E. coli (n = 8)-, and OmpA (n = 8)-treated mice were assessed using ELISA. Data presented as mean ± SEM analyzed by Kruskal-Wallis with post-test comparison. * p < 0.05.

Effect of Autoantibody Production by OmpA
Anti-SSA/Ro and anti-SSB/La antibodies are autoantibodies elevated in sera of patients with systemic autoimmune diseases such as rheumatoid arthritis, lupus, and SS [30,31]. Thus, we examined whether OmpA-treated mice produced autoantibodies similar to those in autoimmune patients. Serum levels of anti-SSA/Ro antibodies and anti-SSB/La antibodies were elevated in OmpA-and E. coli-inoculated mice, compared to those inoculated with PBS ( Figure 7). The data together with the histology of the inflammation in the extra-intestinal glands suggested that our model mouse had developed systemic autoimmune responses.

Discussion
The human gut harbors a highly complex microbial community that allows the digestion of diet and has profound influence on the immune health, hence dysbiosis has been implicated in the immune tolerance modulation or the development of autoimmunity. We previously reported on the significant role of a commensal bacterial strain of E. coli in the systemic immune activation and the generation of exocrinopathy in mice [32]. In the mice of the present study, the Harderian and salivary glands showed infiltrates of inflammatory cells by E. coli-treatment. Dysbiosis of the gut commensal bacteria have been reported to be associated with autoimmune disease severity [16,[33][34][35]. In the present study, we found that repeated inoculation of OmpA from commensal E. coli could induce autoantibody production accompanied by inflammation of the Harderian and salivary glands in mice. Anti-SSA/Ro and anti-SSB/La antibodies have been reported to be elevated in sera of patients with autoimmune diseases upon exposure to intracellular autoantigens, which are discharged into the microenvironment through secretion of autoantigen-containing exosomes or by cell death [31]. The development of the pathologic responses in mice by OmpA may be due to chronic systemic immune activation and autoimmunity.
Microbe-derived structures have been reported to possess the ability to induce autoimmune diseases in experimental animals [36]. Chronic stimuli of the immune system may be provided by extra-intestinal translocation or overgrown microorganisms of a dysbiotic gut. To address the role of microbe-derived molecules in the patho-etiology of autoimmunity, normal mice were stimulated repeatedly with a broad range of pathogen-associated molecular patterns (PAMPs) (PG, TLR-2 ligand; LTA, TLR-2 ligand; MDP, NOD-2 ligand; LPS, TLR-4 ligand). Stimulation with PGN, MDP, LTA, or LPS did not induce infiltration of inflammatory cells in the Harderian or salivary glands. It may be possible that the molecules did not possess the combined functionality of immunogenic antigens and bystander adjuvants [37], which are indispensable for autoimmunity [38]. In addition to the microbial molecules investigated in the present study, we previously studied the flagellar filament structural protein FliC, a TLR5-ligand, and an NLR apoptosis inhibitory protein (NAIP) activator, derived from commensal E. coli, which was repeatedly inoculated following the same regimen of injection as the present study [17,39]. In the reports, we had found that sialadenitis and pancreatitis could be induced by FliC in mice within the fifteen weeks of observation after the final inoculation [17,39]. The results of the present study indicated that OmpA was highly antigenic to induce humoral adaptive immune response to elicit antibodies and to encounter histological inflammation of the Harderian and the salivary glands, but not of the pancreas, indicating that different microbe-derived proteins had initiated inflammation in exocrine glands with tropism.
In our previous study, pancreas of mice after E. coli-derived FliC treatment showed infiltration of inflammatory cells, and the majority of the infiltrated cells were CD3 + T cells [39]. In the present study, inflammation of the Harderian glands was reproduced in immune-deficient Rag2 KO mice and in mature T cell-deficient TCRβ-TCRδ KO mice by adoptive transfer of splenocytes or CD4 + T cells, respectively, from E. coli-treated mice. The results suggested that the exogenously administered E. coli-derived peptides possessed the ability to encounter cell-mediated immune response. Enterobacteriaceae, such as E. coli and Klebsiella pneumoniae, have the ability to use its cell surface OmpA to bind and to incorporate into dendritic cells (DCs) [40,41]. Salmonella OmpA induces expression of MHC class II and co-stimulatory molecules on DCs in mice [42] and differentiate naïve CD4 + T cell biased towards Th1 and Th17 [42][43][44][45]. In our mouse model, IL-1β, IFN-γ, and IL-17A were inflammatory cytokines that were predominantly produced in sera, similar to the cytokine production by FliC [17]. In addition to cytokine production, OmpA of Klebsiella pneumoniae after endocytosis by DCs has been shown to gain access to the cytosolic MHC class I presentation pathway to elicit cytotoxic T cells (CTLs), in the absence of CD4 + T cell help or adjuvant. This property of OmpA has been explained by its ability to bind scavenger receptors, such as LOX-1 [21,46], and favor antigen targeting to and cross-priming by DCs [40], which may explain the positivity of CD8 + T cells in the Harderian glands in OmpA-treated mice in the present study.
Although exact mechanisms remain unknown, shared epitopes between proteins of the host and the microbiome have been proposed in the initiation of autoimmune diseases. In autoimmune pathology, sequence similarity of skin, oral, and gut commensal bacteria with autoantigens of SS and lupus [16], and cross-reaction of envelope proteins in spondyloarthritis patients have been reported [47], which may in part explain the link between dysbiosis and the pathology of autoimmunity [14]. The Enterobacteriaceae family possesses OmpA with high post-transcriptional sequence similarities, indicating that the conserved sequence could be shed from a broad spectrum of microbiota. It may thus be speculated that not only E. coli, as in the present study, but multiple microorganisms belonging to the Enterobactericacea family may potentially trigger initiation of autoimmunity. Validation of sequence differences of OmpA in patient isolates will allow further elucidation of the role of OmpA in the pathogenesis of autoimmunity.
In conclusion, we found a novel antigen of the commensal bacteria with a possible association to the pathogenesis in autoimmunity in mice. The highly immunogenic OmpA may be responsible for E. coli to induce innate and Th1/Th17-associated pro-inflammatory cytokines and CTLs, which may contribute to the acinar cell destruction of exocrine glands. Since we were not able to elucidate the interaction of OmpA with the exocrine glands, future analysis is needed to determine responsible OmpA residues for autoantibody production. Peptide sequence differences may explain the inflammatory involvement in the pancreas by FliC, but not by OmpA, despite similar cytokine production by the two Enterobactericeae derivatives.

Treatment of Mice with Bacteria and Bacterial Components
Mice were treated with or without bacteria or bacterial cell wall components. Heat-killed E. coli (2 × 10 7 colony forming units), 10 µg of PGN, MDP, LTA, LPS or recombinant OmpA in 200 µL PBS, respectively, were injected intraperitoneally once a week for a total of 8 weeks. Each treatment group consisted of seven to eight mice. Mice were sacrificed to obtain samples subsequent to euthanasia with carbon dioxide at the indicated time after the final injection.

Isolation of Extra-Intestinal Gland Tissues
Harderian glands and submandibular salivary glands were removed following euthanasia. Tissue was fixed in 10% buffered formalin.

Sera Collection
Mice were bled by cardiac puncture after euthanasia. Samples were placed at room temperature for 2 h, centrifuged at 900× g for 20 min, and sera were collected and stored at −20 • C until use.

Adoptive Transfer Experiments
Single-cell suspensions from spleen of C57BL/6 wild type mice injected once a week during eight weeks with E. coli, OmpA, or PBS, were obtained at one week after the final inoculation. Bulk spleen cells, or splenic T cells separated using anti-mouse CD4 antibody by magnetic beads on AutoMACS (Miltenyi Boitec, Bergisch Gladbach, Germany) were transferred intravenously to Rag2 KO and TCRβ-TCRδ KO mice on C57BL/6 background, respectively. Each treatment group consisted of six mice. Mice were sacrificed one week after cell transfer to obtain tissue samples.

Focus Scoring
Inflammation in the exocrine gland was quantified using focus scores, as previously described in Reference [48]. Inflammation of at least 50 mononuclear cells was considered as a focus, and the numbers of inflammatory foci per 4 mm 2 were determined as focus scores. Digital measurement of the surface area of the sections and the cell counts were performed on BZ-II Analyzer version 2.00 (Keyence). Inflamed specimen was determined by the presence of more than one focus. Total focus score was calculated for each individual mouse.

Two Dimensional Polyacrylamide Gel Electrophoresis (2-D PAGE) of E. coli Cell Surface Proteins
Water solubilization of E. coli ATCC 25922 surface proteins was performed as previously described in Reference [39]. No DNA was detectable in the spectrophotometric analysis at 260 nm. Then, 2-D PAGE was performed with 200 mg of the surface protein extract on Ettan IPGphor3 IEF using Immobiline DryStrip pH 6-11 (GE Healthcare, Little Chalfont, Buckinghamshire, UK), and 10% acrylamide gels with molecular weight markers between 25-100 kDa (Nippon Gene, Tokyo, Japan). Spot detection and matching were performed using Multi Gauge Software on FLA5100 (Fuji Film, Tokyo, Japan).

Western Blotting of Immunoreactive E. coli Surface Proteins
Proteins resolved in 2-D PAGE were transferred to nitrocellulose membrane (Bio-Rad Laboratories, Hercules, CA, USA), which was blocked in TBS-TM buffer (10 mM Tris-HCl pH7.4, 0.9% NaCl, 0.05% Tween 20, 10% skimmed milk) and incubated with serum of a mouse two months after the eighth inoculation of E. coli, or serum of a mouse treated with PBS alone, diluted 1:2000 in TBS-TM. Immunoreactive spots were detected using an Immunostar enhanced chemiluminescence kit (Wako Pure Chemical Industries, Osaka, Japan).

Identification of Proteins Using MALDI-TOF-MS
MALDI-TOF mass spectrometry (AutoflexII; Bruker Daltonics) was used for protein identification, as previously described in Reference [39]. Briefly, spots from SYPRO-Ruby (Life Technologies) stained 2-D PAGE gel were excised on EXQuest (Bio-Rad), and digested with trypsin (Trypsin Gold, Mass Spectrometry Grade; Promega Corporation, Fitchburg, WI, USA). Mass spectra were calibrated using a peptide calibration standard mono (Bruker Daltonics) and ran in the positive ion reflector mode and in a mass-to-charge ratio (m/z) range of 600-4000 Da. Peptide mass fingerprint (Matrix Science) was conducted using the non-redundant NCBI database (US National Library of Medicine) with MASCOT search engine (Matrix Science), through BioTools 3.0 interface (Bruker Daltonics).

Recombinant OmpA Protein Purification and Anti-OmpA Antibody Measurement
DNA fragments of the ompA gene encoding for OmpA1-325, was amplified by polymerase chain reaction (PCR) using InFusion DNA polymerase (TaKaRa Bio) and genomic DNA of E. coli ATCC 25922 as template. PCR products were inserted into TAGZyme pQE2 (Qiagen, Hilden, Germany) at SphI and HindIII sites (New England Biolabs) in E. coli DH5α, which were selected with 50 µg/mL ampicillin in LB medium. After IPTG induction of E. coli BL21 transformed with constructed plasmids, cells were lysed in the presence of EDTA-free protease inhibitor cocktail (Sigma Chemicals). Then, 6×His-tagged recombinant OmpA was purified by Ni 2+ -affinity chromatography (GE Healthcare). Protein purity was assessed by SDS-PAGE stained with Coomassie brilliant blue and Western blotting probed with anti-OmpA monoclonal antibody 49.415 [30]. After treatment with Detoxin Gel Endotoxin Removing Columns (TaKaRa Bio), endotoxin levels in the recombinant protein solution used for experiments were <0.005 EU/ml measured by using the Toxin Sensor Chromogenic Limulus Amebocyte Lysate Endotoxin Assay Kit (GenScript, Piscataway, NJ, USA). Protein concentration was quantified with the bicinchoninic acid (BCA) assay (BioRad). Anti-OmpA antibody levels in sera were measured by detection with HRP-conjugated goat anti-mouse antibodies (BioSource, Camarillo, CA, USA) against purified recombinant OmpA on nitrocellulose membrane strips, followed by determination of the intensities using the ImageJ program/gel analyzer option (US National Institute of Health, Bethesda, MD, USA), as previously described in Reference [32].

Statistical Analysis
Multiple comparisons of the data were performed using Kruskal-Wallis with a post-test comparing each group to all other groups on GraphPad Prism version 5 for Windows (GraphPad Software, San Diego, CA, USA). p-values < 0.05 were considered as statistically significant.