Caco-2 cells were obtained from ATCC (HTB-37, Manassas, VA, United States) and cultured as recommended by ATCC and grown as previously described (He et al., 2013; Shah et al., 2014) with passage numbers 22–30. In brief, cells were plated at a density of 10⁵/cm² in either a T25 or a 96-well plate after differentiation (Ouwehand and Salminen, 2003). Cells were maintained in DMEM/high modified (Thermo Scientific, Rockford, IL, United States) with 16.6% fetal bovine serum (FBS) (HyClone Laboratories, Logan, UT, United States), non-essential amino acids (Thermo Scientific, Rockford, IL, United States), 10 mM MOPS (Sigma-Aldrich Corp., St. Louis, MO, United States), 10 mM TES (Sigma-Aldrich Corp., St. Louis, MO, United States), 15 mM HEPES (Sigma-Aldrich Corp., St. Louis, MO, United States), and 2 mM NaH₂PO₄ (Sigma-Aldrich Corp., St. Louis, MO, United States). Cells were considered to be differentiated 14 days post confluence (Ouwehand and Salminen, 2003), and used for the adhesion and cross-linking assays. Lactobacillus acidophilus cultures were obtained from ATCC (700396, Manassas, VA, United States) and grown microaerophilically on MRS media at 37°C as described by Chou and Weimer (1999 #3704). Salmonella Typhimurium and Escherichia coli were obtained from ATCC (14028, 43895, Manassas, VA, United States). Salmonella and E. coli were grown aerobically on LB media at 37°C.

The bacterial surface was labeled with Sulfo-N-hydroxysuccinimidyl-2-(6-[biotinamido]-2-(p-azido benzamido)-hexanoamido) ethy-1,3′-dithioproprionate (Sulfo-SBED) (Thermo Fisher Scientific, Rockford, IL, United States). The total protein content on ∼10⁹ bacteria’s surface was measured using the BCA protein assay (Thermo Fisher Scientific, Rockford, IL, United States). Assuming an average molecular weight of 60,000 Da for proteins, a 10-fold molar excess Sulfo-SBED (dissolved in DMSO at 50 μg/μl) of the determined protein concentration, was added to 10⁹ bacteria/ml suspended in 1 ml of Tyrode’s buffer (140 mM NaCl, 5 mM KCl, 1 mM CaCl₂, 1 mM MgCl₂, 10 mM glucose, 10 mM sodium pyruvate, 10 mM HEPES, pH 7.4) The labeling reaction was conducted in dark on ice for 45 min with intermittent shaking. After incubation, the reaction was quenched by adding twofold molar excess glycine compared to Sulfo-SBED. The bacteria were washed twice with Tyrodes buffer by centrifugation at 6000 × g for 2 min and resuspended in 1 ml Tyrodes buffer.

Fibronectin (Sigma-Aldrich Corp., St. Louis, MO, United States), fibrinogen (Sigma-Aldrich Corp., St. Louis, MO, United States), and amyloid precursor protein (APP; Novus Biologicals, Littleton, CO, United States) were dissolved in Tyrodes buffer at 1 mg/ml and were labeled with 10-fold molar excess of sulfo-SBED as described above. After quenching the labeling reaction, the proteins were desalted with Tyrodes buffer using Microcon YM-30 (Millipore, Billerica, MA, United States) as described by the manufacturer’s instructions and resuspended at a concentration of ∼1 mg/ml.

The cell lysate (50 μl) was diluted with 50 μl of 2× Laemmli buffer (Shapiro et al., 1967) (52.5 mM Tri–HCL, pH 6.8, 25% glycerol, 2% SDS, 0.02% bromophenol blue) with or without the addition of 150 mM DTT and heated at 95°C for 10 min. Subsequently, the lysate was centrifuged at 12,000 × g in a micro-centrifuge for 5 min and the proteins (50 μl) in the supernatant were separated by SDS–PAGE using the Mini-PROTEAN electrophoresis system (Bio-Rad Laboratories, Hercules, CA, United States) as described by manufacturer’s insert at a constant current of 30 mAmp per gel, using 4–20% precast Tris–HCL Gels (Bio-Rad Laboratories, Hercules, CA, United States). The gels were stained overnight with Imperial protein stain (Thermo Fisher Scientific, Rockford, IL, United States) as per the manufacturer’s protocol. The gels resolved in reducing and non-reducing conditions were imaged using the Kodak Image Station 2000R (Carestream Health). The images for gels were visually compared for missing bands in the reduced gels that were targeted for excision and protein identification as using LC–MS/MS at the Center for Integrated BioSystems, Utah State University (Logan, UT, United States) (Liang et al., 2006; Supplementary Figure S1).

Deletion mutants in Salmonella were constructed as described by Datsenko and Wanner (2000). Briefly, mini-prep kit (Qiagen, Valencia, CA, United States) was used to isolate plasmid pKD46 containing ampicillin resistance and λ Red recombinase genes from E. coli BW25141 (CGSC 7634), plasmid pKD3 containing chloramphenicol resistance gene from E. coli BW25141 (CGSC 7631), and plasmid pKD4 containing kanamycin resistance gene from E. coli BW25141 (CGSC 7632). This plasmid pKD46 was electroporated in to S. Typhimurium and transformants were selected by growth on LB agar containing 100 μg/ml ampicillin (Sigma-Aldrich Corp., St. Louis, MO, United States). S. Typhimurium containing pKD46 was grown in LB broth in the presence of 100 μg/ml ampicillin and 100 mM L-arabinose to induce λ Red recombinase production. The chloramphenicol or kanamycin resistance genes were amplified using plasmid pKD3 or pKD4 templates, respectively, to ensure exact deletion sites. The primers used for amplification of the gene encoding the phage tail-like protein (STM2699) were STM2699 P1 and STM2699 P2. The primers used for amplification of the gene encoding for the integration host factor protein (ihfA, STM14_1626) were STM14_1626 P1 and STM14_1626 P2. The purified PCR products were electroporated into S. Typhimurium with induced λ Red recombinase. The transformants were selected on LB agar either with 10 μg/ml chloramphenicol or 40 μg/ml kanamycin. The gene deletion for ΔSTM2699 was confirmed by PCR using primers STM2699 J1, STM2699 J2, STM2699 F, and STM2699 R. The gene deletion for ΔSTM14_1626 was confirmed by PCR using primers STM14_1626 J1, STM14_1626 J2, STM14_1626 F, and STM14_1626 R. All primer sequences are available in Supplementary Table S1.

Microbial cells (∼10⁹ cfu/ml) were incubated with 1 ml of Sulfo-SBED labeled fibronectin, fibrinogen, or APP (1 mg/ml) for 30 min at 37°C. Subsequently, the suspension was placed under a 15-watt UV lamp (302 nm) at a distance of 5 cm for 10 min for cross-linking. The cells were washed twice with 1 ml Tyrodes buffer by centrifugation at 6000 × g for 2 min and resuspended in 500 μl lysis buffer (0.1% Triton, 150 mM DTT) and 250 μl of glass beads (0.1 mm) (BioSpec Products, Inc., Bartlesville, OK, United States). The samples were homogenized in a Mini-Beadbeater (BioSpec Products, Inc., Bartlesville, OK, United States) by giving three pulses at full speed for 30 s with intermittent 1 min incubation on ice (Chou and Weimer, 1999; Chou et al., 2001). The free biotin on the Sulfo-SBED reagent from the reduced non-cross-linked proteins was removed by passing the lysate through an YM3 Microcon ultrafiltration module (Millipore, Billerica, MA, United States) as described in manufacturer’s instructions. The retantate (volume brought up to 800 μl with Tyrodes buffer) (De Ridder et al., 1975) from the Microcon module was incubated with ∼30 avidin-coated glass beads (3 mm; Xenopore Corp., Hawthorne, NJ, United States) for 30 min on a shaking platform to capture the biotinylated proteins. The glass beads were subsequently washed three times with 5 ml of wash buffer (50 mM Tris, 1.5 M NaCl, pH adjusted to 7.2) to remove non-specifically bound proteins. The proteins on the washed beads were digested overnight using 300 ng of proteomics grade trypsin (Sigma-Aldrich Corp., St. Louis, MO, United States) in 1 ml of 100 mM ammonium bicarbonate buffer (Liang et al., 2006). The digested proteins were concentrated using a speedvac (Thermo Fisher Scientific, Rockford, IL, United States) to final volume of 50 μl and submitted to the Center for Integrated BioSystems, Utah State University (Logan, UT, United States) (Liang et al., 2006; Boudina et al., 2009; Wende et al., 2015) or the Mass Spectrometry and Proteomics Core Facility, University of Utah (Salt Lake City, UT, United States) for protein identification by LC/MS/MS. Analysis of Salmonella ligands to APP set the peptide mass tolerance at ±6 ppm and fragment mass tolerance at ±0.6 Da via LC/MS/MS. Peptides were identified using a MASCOT with the NCBInr database and confirmed against a custom Salmonella protein database constructed from the exact bacterial genome used in this experiment (NC_016856). The entire labeling protocol was repeated using unlabeled host binding components as negative controls in three biological replicates.

Prior to whole cell cross-linking the cells were maintained in osmotically balanced buffer to maintain cell membrane integrity and subsequently reduce interference from intracellular proteins during the whole cell experiment. Sulfo-SBED labeled bacteria (10⁹ bacteria/ml) were interacted with ∼10⁶ Caco-2 cells grown in a T25 in a final volume of 3.5 ml Tyrodes buffer for 60 min at 37°C. At the end of 60 min incubation, the bacterial suspension was aspirated from the flask and the host cell flask was placed under a 15-watt UV lamp (302 nm) at a distance of 5 cm for 10 min. The cross-linked Caco-2 cells and associated bacteria were resuspended in 500 μl of lysis buffer [8 M urea, 6.0% ampholytes pH range (3–10) (Bio-Rad Laboratories, Hercules, CA, United States), 0.4% CHAPS, 0.25% Triton 100, 0.15% n-dodecyl-B-β-D-maltoside, 0.002% bromophenol blue] and 250 μl of glass beads (0.1 mm, BioSpec Products, Inc., Bartlesville, OK, United States). The samples were homogenized in a Mini-Beadbeater (BioSpec Products, Inc., Bartlesville, OK, United States) by giving three pulses at full speed for 30 s with intermittent 1 min incubation on ice. The samples were stored at -70°C until further use. Cross-linked extracted samples were processed with 2D gels and LC/MS/MS for protein identification. InterProScan (Markowitz et al., 2006) was used to identify each protein in the partnership to verify the accuracy of the identified pairs.

The cross-linked samples were thawed and centrifuged at 12,000 X g in a microcentrifuge and the supernatant was used for 2D gel analyses (Liang et al., 2006; Pate et al., 2007, 2008). Isoelectric focusing (IEF) was done using 50 μl of sample in tube gels using the Model 175 tube cell (Bio-Rad Laboratories, Hercules, CA, United States) as per the manufacturer’s instructions. In brief, 4% polyacrylamide gels were cast in 1 mm diameter tubes to a height of 11 cm. Samples (50 μl) were loaded in to the tubes and electrophoresed at 200 V for 2 h, 400 V for 4 h, and finally at 800 V for 8 h. After the IEF run, gels were extruded from the tubes and equilibrated in transfer buffer (3% SDS, 0.07 M Tris–HCL) with or without 150 mM DTT for 15 min. The gels equilibrated in presence of DTT were subsequently alkylated in transfer buffer using 150 mM iodoacetamide for 15 min. The tube gels were then resolved in the second dimension using the Criterion electrophoresis system (Bio-Rad Laboratories, Hercules, CA, United States) exactly as per manufacturer’s recommendation at a constant current of 45 mAmp per gel using 4–20% precast Tris–HCL Gels (#345-0104, Bio-Rad Laboratories, Hercules, CA, United States). The gels were stained overnight with Imperial protein stain (Thermo Fisher Scientific, Rockford, IL, United States) exactly as per manufacturer’s recommendations.

The gels resolved under reducing and non-reducing conditions were imaged using the Kodak Image Station 2000R (Carestream Health, Rochester, NY, United States). The images were compared and spots missing in the reducing gels but present in the non-reduced gel were identified. Protein spots were selected if: (1) the spot disappeared in reduced conditions; (2) had host and microbe protein in the same spot; (3) the sum of molecular weight of the identified proteins matched the observed molecular weight on the gel. The gel spots were picked, in-gel digested, and the proteins identified using LC–MS/MS at the Center for Integrated BioSystems, Utah State University (Logan, UT, United States; Supplementary Figure S1).

To determine if the additional proteins found to bind fibronectin were relevant and indicative of a robust method, an investigation of the characteristics for LBA0222 was done using displayed Dendroscope (SSDB paralog search, SSDB domain analysis) (Aoki and Kanehisa, 2005; Jensen et al., 2009; Huson and Scornavacca, 2012) to discover the domain conservation.

The role of specific receptors and ligands in adhesion and invasion of the host by Salmonella was assayed by determining the changes in the amount of total host (Caco-2 cells) associated Salmonella after to receptor blocking by specific antibodies to specific host proteins identified by cross-linking. Caco-2 cells were cultured as described above, except it was done in a 96-well plate format. The bacteria were used after two transfers for the adhesion assays (Chen et al., 2017). Bacterial cells were collected from 2 ml of media after growth for 14 h, washed twice with an equal volume of PBS, and re-suspended at ∼10⁸ cfu/ml, in DMEM/high modified with 1× non-essential amino acids, 10 mM MOPS, 10 mM TES, 15 mM HEPES, and 2 mM NaH₂PO₄ but without the FBS. Caco-2 cells were incubated with dilutions of anti-SPTAN Ab (1:1000, 1:2000, 1:4000, and 1:8000) (Novus Biologicals, Littleton, CO, United States). Anti-APP Ab was used to block APP at a final dilution of 1:800, after an optimization that used additional concentrations (1:200, 1:400, 1:800, 1:1600), with an incubation of 60 min in a final volume of 50 μl at 37°C in 5% CO₂. At the end of 60 min, the Caco-2 cells were infected with Salmonella as previously described (Arabyan et al., 2016) (MOI 1:100) and incubated for 60 min at 37°C in 5% CO₂. The bacterial cell suspension was aspirated and the Caco-2 monolayer was washed thrice with 200 μl of Tyrodes to remove non-adhered bacterial cells from the monolayer. Intracellular bacteria were further quantified by a 2 h incubation in 40 μg/ml gentamicin solution (Sigma-Aldrich, St. Louis, MO, United States) and washed thrice with 200 μl of Tyrodes buffer to remove dead bacterial cells from the surface of the monolayer (Elsinghorst, 1994; He et al., 2013; Shah et al., 2014; Arabyan et al., 2016). DNA extraction buffer (AEX Chemunex, France; 50 μl) was used to lyse the monolayer and the bacteria associated with the host, and incubated at 37°C for 15 min followed by 95°C for 15 min. The resulting cell lysate was used to determine the number of bacteria associated with the Caco-2 cells. Quantitative analysis was done using qPCR with a CFX 96 Real Time System (Bio-Rad, Hercules, CA, United States). Reactions were performed in a final volume of 25 μl including 1 μl of cell lysate, 100 nM of PCR primers, and iQ SYBR Green Supermix (Bio-Rad, Hercules, CA, United States) as per manufacturer’s instructions. The primers used for the amplification are listed in Supplementary Table S1. The reaction parameters consisted of denaturation step at 95°C for 5 min, followed by 40 cycles of denaturation, annealing, and extension at 95°C for 15 s, 56°C for 30 s, 72°C for 30 s, respectively, and a final extension at 72°C for 1 min. The product was verified using a melt curve analysis from 50 to 95°C with a transition rate of 0.2°C/s. The number of bacterial cells and Caco-2 cells present in each well were determined by using a standard curve of C_T vs. Log₁₀ cfu and concentration of bacteria per Caco-2 was calculated. The data were normalized relative to control wells that were not given Ab treatment. The experiment was done in four biological replicates. Effects of the treatments were assessed using one-way ANOVA, and individual means were compared to that of control by Dunnett’s multiple comparison test. Means were considered significantly different at p < 0.05.

Early stationary phase (14 h) Salmonella cultures and the corresponding spent supernatant were interacted with purified APP protein (NBP1-99026, Novus Biologicals, Littleton, CO, United States) for 30 min at 37°C. Cells were suspended in 100 mM Tris–HCl buffer (pH 7.2) while spent Luria Broth (LB broth; BD Difco, Franklin Lakes, NJ, United States) was used as the spent supernatant. Following incubation, APP suspensions were examined using a 4–12% gradient SDS-page gel and stained with SYPRO Ruby Protein Gel Stain (Thermo Fisher Scientific, Petaluma, CA, United States).

Actively growing L. acidophilus NCFM was incubated with Sulfo-SBED-labeled fibronectin beads and cross-linked. LC–MS/MS analysis revealed two proteins from the same operon – cell separation protein (CdpA) and an uncharacterized protein encoded by LBA0222 – were bound to fibronectin. Additionally, four ribosomal proteins were cross-linked to fibronectin (Table 1). A negative control in which unlabeled fibronectin was interacted with the bacterial cells identified no proteins. This verified previous observations confirming CdpA (Johnson et al., 2015) and found additional bacterial proteins.

Functional analysis of CdpA revealed that this protein accounts for ∼80% of the adhesion capability of NCFM to gut epithelial cells (Altermann et al., 2004). Predicted homologs of this protein were found in Lactobacillus spp. containing orthologs of CdpA, S-layer proteins, and levansucrase (Supplementary Figure S2), confirming this approach found specific proteins that mediate adhesion of lactobacilli.

The hypothetical protein found during cross-linking is in an operon with cdpA that is unique to NCFM. Protein network analysis of CdpA found two predicted functional partners: two bacterial surface layer proteins – LBA0220 and LBA0221 that are in a single operon – as well as ribose-p-pyrophosphokinase (Prs) (Supplementary Figure S3), again confirming published results. Protein domain analysis (Cserzo et al., 1997; Aoki and Kanehisa, 2005; Jensen et al., 2009) found LBA0222 and CdpA to contain SLAP and FIVAR protein domains (Supplementary Figure S4-48pc). SLAP domains are found in numerous bacterial cell surface proteins, including S-layer proteins, amidases, and cell separation proteins, suggesting this domain is important in host receptor binding. As they are often found in combination with glycosyl–hydrolase domains, further suggesting that they are involved in hydrolyzing the host glycan as recently found to be important in Salmonella for invasion (Arabyan et al., 2016).

The protein encoded by the LBA0222 locus was also cross-linked to fibrinogen, and is in the same operon as cdpA (Supplementary Figure S4). Bioinformatic analysis of the LBA0222 protein determined that homologs were found only in other lactobacilli and one genome of Chlamydia (Supplementary Figure S2). The genetic neighborhood was unique to NCFM and is in the same operon as CdpA, as well as other organisms (Supplementary Figure S3). These observations found the exact results from previous reports (Johnson et al., 2013, 2015) in NCFM. This recapitulation of the results led to expansion of the approach to pathogenic bacteria with pure proteins suspected of binding pathogenic enteric bacteria where host interaction is more complex and less well described.

As with lactobacilli, we initiated verification of the methods using purified proteins. Each 2D gel spot contained fibrinogen as well as a bacterial protein (Table 2 and Supplementary Figure S1). Detection of S. Typhimurium and E. coli ligands to fibrinogen revealed one candidate spot on 2D gels for each organism. The cross-linked protein complex from S. Typhimurium LT2 contained fibrinogen and murein lipoprotein (lpp) [STM1376 (lppB), STM1377 (lpp)], while the cross-linked 2D gel spot from E. coli contained fibrinogen and a universal stress protein B (uspB; b3494). This confirmed previously known ligands associated with pathogenesis for these organisms and was shown that deletion of lpp resulted in lower infectivity (Sha et al., 2004; Daley et al., 2005).

Secondarily, S. Typhimurium was cross-linked to purified APP, which is a membrane-associated host protein whose proteolytic product recently emerged as a possible protective factor to enteric pathogens, including Salmonella (Kumar et al., 2016). Cross-linking resulted in identification of 39 bacterial proteins to be associated with APP. Of those identified, 15 had a sequence query match of sufficient strength (i.e., >50) to be considered viable for investigation (Table 3). Those ligands included bacterial membrane-associated proteins, entericidin B (EcnB; STM4336), FtsH protease regulator (HflK; STM4363), and integration host factor (IhfA; STM1339). Homologs of these proteins have been characterized in E. coli but not in relation to host interaction and do not have host association receptor proteins reported (Bishop et al., 1998 #2586; Kihara et al., 1998 #2588; Bandyopadhyay et al., 2010 #2589). Use of purified proteins with these pathogens verified bacterial proteins known to facilitate bacterial association using the approach using two different purified host proteins. Consequently, we proceeded to use whole bacterial and host cells in cross-linking experiments.

The ultimate utility of this approach is to find novel receptor/ligand partnerships during host/microbe association. The verification experiments with pure host proteins provided the basis to use whole Salmonella and host colonic cells in vitro to determine specific protein partnerships. After whole cell cross-linking was done, we observed SPTAN1 (host) to bind the protein produced from STM2699 (Fels-2 prophage tail) in Salmonella, while host proteins HSP90B1 and ACTN4 were bound to the protein from STM1956 (fliA) and STM4088 (yiiU) in Salmonella, respectively (Table 4). Of the four host proteins identified, SPTAN1 and HSP90B1 were previously reported to be involved in pathogenesis of Salmonella and Listeria (Cabanes et al., 2005; Ruetz et al., 2011; Martins et al., 2012), but no report of the microbial partner is known.

To examine the role of the putative whole cell binding receptor/ligand partners we conducted in vitro infection assays with Salmonella and colonic epithelial cells (Caco-2). We hypothesized that disrupting the receptor and the ligand partnerships would result in a reduction of Salmonella association in vitro. This hypothesis was tested using Ab-blocking of host receptors and genetic deletion of the bacterial ligand. Pre-treating Caco-2 cells with an α-SPTAN1 Ab significantly (p < 0.05) decreased the total host associated Salmonella by approximately twofold (Figure 1). Deletion of the Fels-2 prophage tail (STM2699) from the genome also resulted a significant twofold decrease (p < 0.05) in total Salmonella association. These observations confirm that SPTAN1 and STM2699 are cognate partners that mediate Salmonella association with colonic epithelial cells. STM2699 is annotated as Fels-2 prophage tail protein and has 100% identity to Fels-2 prophage tail proteins from other organisms (E-value = 5.8e-60) including nine other S. enterica ssp. enterica genomes, Enterobacter aerogenes strain 682_EAER, Pantoea ssp. At9b, and Erwinia billingiae. This protein is controversial in its role in Salmonella pathogenicity (Andrews-Polymenis et al., 2004; Pickard et al., 2008), which is possible since it can be move between bacteria via horizontal gene transfer and transduction, but may play a role in host adaptation in Salmonella (Andrews-Polymenis et al., 2004). Shah et al. (2014) reported SPTAN1 to be involved in Salmonella virulence and used during host invasion during abiotic stress. With this positive identification of binding partners that have reported biological impact on Salmonella association, we proceeded to examine a more complex binding partner – APP – to determine if the identified binding partners play a role in Salmonella association.

Unexpectedly, pre-treating the epithelial monolayer with anti-APP Ab prior to Salmonella association had no effect on host invasion (Figure 2), suggesting that this was not a primary binding partner for Salmonella. However, deletion of STM1339 (ihfA) resulted in a significant (p < 0.05) 10-fold increase in Salmonella host invasion. Since STM1339 (ihfA) regulates proteases in pathogenicity island 1 (SPI-1) in Salmonella (Mangan et al., 2006) we proceeded to examine possible digestion of APP by Salmonella during APP binding using the ihfA::Cm^R mutant. The ability of the mutant to digest APP did not differ from wild-type Salmonella; however, incubation of purified APP in the mutant spent supernatant resulted in a 25% decrease in intact APP (Figure 3). Deletion of STM1339 (ihfA) relieved the intricate regulation of the proteolytic cascade resulting in dysregulation of Salmonella protease system and export leading to a hypervirulent phenotype as observed previously in Salmonella (Heithoff et al., 2012). Taken together, these observations suggest that Salmonella produces a soluble protease(s) that digests APP during whole host cell association with Salmonella that results in a complex interaction that produces other biologically active end products.