Identification of E-cadherin signature motifs functioning as cleavage sites for Helicobacter pylori HtrA

The cell adhesion protein and tumour suppressor E-cadherin exhibits important functions in the prevention of gastric cancer. As a class-I carcinogen, Helicobacter pylori (H. pylori) has developed a unique strategy to interfere with E-cadherin functions. In previous studies, we have demonstrated that H. pylori secretes the protease high temperature requirement A (HtrA) which cleaves off the E-cadherin ectodomain (NTF) on epithelial cells. This opens cell-to-cell junctions, allowing bacterial transmigration across the polarised epithelium. Here, we investigated the molecular mechanism of the HtrA-E-cadherin interaction and identified E-cadherin cleavage sites for HtrA. Mass-spectrometry-based proteomics and Edman degradation revealed three signature motifs containing the [VITA]-[VITA]-x-x-D-[DN] sequence pattern, which were preferentially cleaved by HtrA. Based on these sites, we developed a substrate-derived peptide inhibitor that selectively bound and inhibited HtrA, thereby blocking transmigration of H. pylori. The discovery of HtrA-targeted signature sites might further explain why we detected a stable 90 kDa NTF fragment during H. pylori infection, but also additional E-cadherin fragments ranging from 105 kDa to 48 kDa in in vitro cleavage experiments. In conclusion, HtrA targets E-cadherin signature sites that are accessible in in vitro reactions, but might be partially masked on epithelial cells through functional homophilic E-cadherin interactions.


H. pylori HtrA mediates different cleavage patterns of E-cadherin in vitro and on gastric epithelial cells.
Several pathogenic bacteria secrete the HtrA protease, which cleaves off the extracellular NTF domain of E-cadherin on epithelial host cells to open intercellular junctions 20,21,25,26 . However, it is completely unknown how HtrA interacts with E-cadherin at the molecular level. The aim of this study was the identification of HtrA cleavage sites as the basis for the future generation of potential pharmacological tools. Mature human E-cadherin (hCdh1) is expressed as a 125 kDa glycoprotein, which consists of the five extracellular domains EC1-EC5, a linker region, a single transmembrane domain (TMD) and an intracellular (IC) domain (Fig. 1A, hCdh1 D 155 -D 882 ). Several antibodies are available that selectively recognise individual domains of the E-cadherin protein. SHE78-7 detects the EC1 domain, the HecD1 antibody recognises the EC2 domain and H108 binds to the EC5 domain 19,30 . The specificity and selectivity of these antibodies was confirmed by using deletion mutants of recombinant E-cadherin lacking the EC1, EC1-2, EC1-3, EC4-5 and EC5 domains (see Supplementary Fig. S1A, Fig. S1B, Table S1). To investigate the fragmentation of E-cadherin, NCI-N87 cells were infected with H. pylori for the indicated time periods and aliquots of the supernatants were analysed for E-cadherin fragments containing the EC1, EC2 or EC5 domain. After 16 h of infection, the anti-EC1 antibody detected a strong increase in the 90 kDa NTF fragment of E-cadherin (hCdh1 NTF ). We observed a similar picture for the anti-EC2 antibody. When the supernatant was analysed with the anti-EC5 antibody, the hCdh1 NTF fragment enriched during H. pylori exposure was already visible after 4 h of H. pylori exposure and the levels increased during the 16 h of infection (Fig. 1B, upper panels). Further fragments were not detectable, even after longer exposures of the membranes (not shown). These results indicate that H. pylori induces production of a stable 90 kDa hCdh1 NTF fragment which contains the EC1 to EC5 domains. Although H. pylori adhered slightly more strongly to AGS than to MKN-28 or NCI-N87 cells (see Supplementary Fig. S1C), the formation of a stable 90 kDa fragment of E-cadherin in the supernatants was reproducibly detected for all cell lines (Fig. 1B, see Supplementary Fig. S1D). Corresponding to the increase in the 90 kDa hCdh1 NTF , we observed a significant drop in signals for the 125 kDa full-length E-cadherin protein (hCdh1 FL ) in lysates of H. pylori-infected NCI-N87 cells (Fig. 1B, middle panels). A slight decrease in hCdh1 FL could be detected using the anti-EC1, anti-EC2 and anti-EC5 antibodies during early time periods of H. pylori infections and an obvious loss of hCdh1 FL was visible after 16 h. An anti-IC antibody confirmed these results, with β -actin as loading control (Fig. 1B, lower panels).
E-cadherin forms functional cis-and trans-interactions between the extracellular domains of two neighbouring cells 3 which could overlay additional cleavage sites. Therefore, we examined fragment formation using a recombinant E-cadherin chimera (rCdh1 D 155 -I 707 ). rCdh1 contains the five EC domains and the linker region (A 698 -I 707 ), but lacks the TMD. Instead of the IC domain, rCdh1 is fused to the Fc region and a His 6 tag (Fig. 1A, rCdh1 D 155 -I 707 ). Incubation of rCdh1 with recombinant H. pylori HtrA (rHtrA) for the indicated time periods showed a different cleavage pattern compared to the H. pylori infection experiments. Using the anti-EC1 antibody, we only detected the loss of the full-length E-cadherin protein (rCdh1 FL ) ( Fig. 2A). Even after long exposure, the Western blots did not reveal further fragments, suggesting an immediate loss of the EC1 domain (see Supplementary Fig. S2). The anti-EC2 antibody detected an additional high molecular weight (~110 kDa) fragment ( Fig. 2A, see Supplementary Fig. S2). A ladder of putative E-cadherin fragments of approximately 105 kDa, 90 kDa, 60 kDa and 48 kDa was observed using the anti-EC5 antibody ( Fig. 2A). Similar fragments appeared when an anti-His tag antibody was applied ( Fig. 2A, see Supplementary Fig. S2), indicating that fragmentation of E-cadherin begins at the N-terminus with the cleaving of individual EC domains (Fig. 2B). According to the fragment sizes observed during H. pylori infections and in in vitro cleavage experiments, we suggest that E-cadherin is N-terminally and C-terminally processed by HtrA. Upon H. pylori infection, a stable 90 kDa NTF originating . E-cadherin consists of five extracellular domains (EC1-EC5) and a linker region. hCdh1 contains a transmembrane domain (TMD) and an intracellular domain (IC). rCdh1 is expressed as a Fc-tag/His 6 fusion protein. The monoclonal antibodies recognise the EC1 (SHE78-7) or the EC2 domain (HecD1). The polyclonal antibody H108 is directed against the EC5 domain. (B) NCI-N87 cells were infected with H. pylori at a MOI of 100 (+) for the indicated time periods or left untreated (−). E-cadherin fragments in the supernatant (upper panels) and in the lysates (lower panels) of infected cells were detected using antibodies recognising the EC1, EC2, EC5, or the IC domains. β -actin was detected as a control.
Scientific RepoRts | 6:23264 | DOI: 10.1038/srep23264 from hCdh1 is formed, while in in vitro cleavage experiments, rCdh1 is mainly N-terminally processed resulting in a series of different fragments (Fig. 2B).

Identification of HtrA cleavage sites in E-cadherin.
To identify the HtrA cleavage sites in E-cadherin, we performed a label-free mass spectrometry-based proteomic analysis of HtrA/trypsin digests of rCdh1. Overall, we detected 128 peptides of E-cadherin origin, among which we identified 46 semi-tryptic peptides that were processed solely by HtrA at the non-tryptic cleavage site. These peptides were commonly cleaved after the residues I (37%), V (28%), T (17%), A (9%) and S (9%). Sequence analysis revealed that HtrA cleavage takes place favourably between hydrophobic amino acids (Fig. 3A,B). We also performed N-terminal Edman sequencing of rCdh1 cleavage products to validate the cleavage pattern obtained from the same recombinant Cdh1 protein.
These results agree with the mass spectrometric analysis. Four of five detected N-terminal cleavage sites (T ↓ T, , which are located within the four calcium-binding sites and which we have named "signature sites" S1, S2, S3 and S4 (Fig. 3C). The signature site S3 between EC3 and EC4 was less frequently targeted by HtrA. This consensus cleavage pattern for HtrA-catalysed proteolysis is related to that of its E. coli orthologue DegP, which also favourably cleaves between hydrophobic residues 31 . We further observed that HtrA cleaved most frequently in the fourth signature sequence of E-cadherin (Fig. 3B). Its I ↓ L cleavage site was detected with the highest intensity of all peptides.
Identification of substrate-derived inhibitory peptides. Since deletion or mutation of the htrA gene in the H. pylori genome is not possible, we aimed at the generation of a peptide-based E-cadherin cleavage inhibitor. We synthesised five peptides (P1− P5) around the cleavage signature sequence between E-cadherin domains EC4 and EC5 (positions L 583 -P 593 ) (Fig. 3C, see Supplementary Fig. S4A). The 21-residue peptide P1 (see Supplementary Fig. S3A, Fig. S4A) effectively inhibited in vitro HtrA-mediated E-cadherin cleavage (Fig. 4A). Shorter peptides (see Supplementary Fig. S4A) did not exhibit any suppressive effect (Fig. 4A,B). The direct binding affinities of the peptides to immobilised HtrA were measured by surface plasmon resonance (SPR). Peptide P1 showed strong binding, while only one other cleavage-sequence peptide (P5) evoked a positive binding response above the significance threshold (Fig. 4C). N-terminal truncation of the P1 peptide led to a loss of enzyme inhibition, indicating that the N-terminal part of P1 is important for substrate recognition by HtrA. In a competition experiment, we analysed the short peptide P1_NT, which contains the hydrophobic amino acid stretch TGTLLLI. As controls, P1_NT (TG) and P1_NT (GT) were included as scrambled 7-residue peptides (see Supplementary  Fig. S4A). None of these peptides blocked HtrA-induced E-cadherin cleavage or interfered with P1-mediated HtrA inhibition (see Supplementary Fig. S4B). We then substituted key amino acids in the C-terminal (P1_CT mut ) or N-terminal (P1_NT mut ) regions with glycine (see Supplementary Fig. S4A). Both P1_CT mut and P1_NT mut only partially inhibited HtrA-mediated E-cadherin cleavage (see Supplementary Fig. S4C). Therefore, we postulate that the 21-residue P1 peptide is required for efficient HtrA inhibition. We also synthesised P1 with a C-terminal alpha-amide and subjected both P1 variants (P1-OH, P1-NH 2 ) to SPR analysis. Although both peptides exhibited concentration-dependent binding (see Supplementary Fig. S3B, Fig. S3C), the P1-peptide with the natural C-terminal carboxylic acid evoked stronger SPR signals, reflecting its binding capacity to HtrA (Fig. 4D). This result is in agreement with a study by Krojer et al., who compared peptide substrates of the E. coli orthologue DegP 32 .
Peptide P1 strongly inhibited HtrA-mediated rCdh1 cleavage in a concentration-dependent manner (Fig. 5A). To investigate the bioactivity of P1, we quantitated the inhibition of HtrA in H. pylori-infected AGS cells that were transfected with hCdh1 FL . We observed a significant dose-dependent decrease in the formation of the hCdh1 NTF (Fig. 5B) in comparison with the corresponding loading controls (see Supplementary Fig. S4D). This effect was also observed in H. pylori infections of several other cell lines and at different infection doses (data not shown). We then analysed the epithelial transmigration of H. pylori in a transwell filter assay. Partially polarised MKN-28 cell monolayers were infected and the transmigration of E-cadherin-cleaving H. pylori was determined over time. Polarisation of MKN-28 cells requires functional tight junctions, which can be monitored by the measurement of the transepithelial electrical resistance (TEER). Corresponding to previous studies 21,25,33,34 , MKN-28 cells showed partial polarisation, as reflected by a resistance of ∼ 150 Ω/cm 2 after 14 days of cultivation (see Supplementary  Fig. S5A). Infections with H. pylori (see Supplementary Fig. S5B) did not significantly alter the polarisation of host cells. However, we observed a strong concentration-dependent reduction of H. pylori transmigration by P1 (Fig. 5C), indicating that P1 efficiently blocks HtrA-mediated E-cadherin cleavage. These results confirm the bioactivity of P1 as the first inhibitory peptide blocking bacterial HtrAs.

Discussion
E-cadherin is an important cell adhesion molecule and tumour suppressor. Ectodomain shedding and an increase in soluble E-cadherin is associated with several cancer types, such as breast cancer, pancreatic cancer and GC 15 . In recent years, several E-cadherin-cleaving proteases have been described. MMPs, such as MMP-7 (matrilysin) 35 and MMP-3 (stromelysin-1) 36 , or proteases such as ADAM- 10 37 are examples that cleave off the extracellular E-cadherin domain. In H. pylori-infected cells, activated ADAM-10 (a disintegrin and metalloprotease) has been suggested as a protease responsible for the cleavage of E-cadherin on host cells 24 . Furthermore, H. pylori induces the expression of several MMPs, including MMP-3 and MMP-7, which can also contribute to E-cadherin shedding 21,24 . Recently, HtrA was added to this collection as the first E-cadherin-cleaving protease expressed by H. pylori 21 . Generally, knowledge of H. pylori proteases is still rare and no additional H. pylori protease has yet been identified as an E-cadherin-cleaving protease. H. pylori expresses an active collagenase (Hp0169), which is crucially important for in vivo colonisation 38 , but this has not been tested for E-cadherin cleavage. Previous studies have predicted several putative H. pylori proteases with an extracellular localisation 39,40 . Among these, Hp0506 is required for bacterial cell shape and pole formation 41 , but, like Hp0657 and Hp1012, it does not target E-cadherin as a substrate 21 . Since E-cadherin acts as an important suppressor of carcinogenesis, increased understanding of the direct interaction between HtrA and E-cadherin is required. Therefore, we report here on a comprehensive cleavage site map for E-cadherin and the development of an effective peptide inhibitor based on an HtrA-targeted substrate sequence.
For MMPs, the amino acid sequence L 585↓ S ↓ D 587 in the EC4 domain and P 700↓ V 701 have been suggested as putative cleavage sites 17,19 . In our study, we identified the [LIV]-x-[LIV]-x-D-x-N-D-[NH]-x-P sequence pattern in the E-cadherin molecule as a target for H. pylori HtrA, which is in line with the L 585↓ S ↓ D 587 cleavage site for MMPs 17 . This sequence pattern is located in three out of four calcium-binding regions of the E-cadherin molecule, which we have designated as signature sites S1, S2, S3 and S4. The accessibility of E-cadherin signature sites as cleavage sites for H. pylori HtrA and possibly for H. pylori-regulated MMPs might explain why we observed different fragmentation patterns in H. pylori infections and in vitro (Fig. 6). We included several different gastric epithelial cell lines in our study, which might be differentially colonised by H. pylori. It had been previously shown that H. pylori infection of MKN-28 cells induces the formation of the stable 90 kDa NTF 20,21 . Although AGS cells showed a slightly stronger binding of H. pylori in comparison to MKN-28 and NCI-N87 cells, an unaltered fragmentation pattern was detected, as reflected by the stable 90 kDa NTF. These data indicate that possible differences in H. pylori adherence to host cells do not influence E-cadherin shedding by secreted HtrA, while cis and/or trans interactions of E-cadherin might be important. The interaction of the ectodomains at the molecular level is still not fully understood, but has been described as a highly dynamic process 3 . It has been suggested that the EC1 domain forms an initial weak intercellular adhesion, while strong adhesion requires at least EC1-3 42 (Fig. 6). In in vitro cleavage experiments, rCdh1 likely exists as a monomer that presents all signature sites as possible cleavage sites (Fig. 6). In contrast, epithelial cells form functional E-cadherin-mediated adhesions through cis and trans interactions of the extracellular E-cadherin domains. This tight binding might overlay the signature sites S1 to S3 and probably S4 42,43 , or induce changes in their structural accessibilities. In this model, H. pylori could still target the S4 signature site or an undiscovered cleavage site in the linker region leading to the formation of the 90 kDa fragment (Fig. 6). Due to the glycosylation of E-cadherin, the size of the E-cadherin sequence from D 155 to the S4 site is not definitively assessable. Therefore, we cannot exclude the possibility that HtrA also targets a sequence in the linker region near the transmembrane domain which was not detectable after tryptic digests.
In conclusion, we have identified signature sites in the E-cadherin protein that are cleaved by H. pylori HtrA leading to the fragmentation of E-cadherin, depending on the cis and trans interactions between the extracellular domains.
Since an htrA knock-out H. pylori mutant is still lacking, we focused on pharmacological inhibition of H. pylori HtrA. Several small molecule compounds targeting H. pylori HtrA have been successfully developed which can function as lead structures 21,[44][45][46] . Here, we present the first substrate-derived peptide that efficiently binds and inhibits H. pylori HtrA. The P1 peptide TGTLLLILSDVNDNAPIPEPR corresponds to the E-cadherin sequence T 578 -R 598 surrounding the signature site S4. The P1 peptide binds with high specificity and efficiently blocked E-cadherin cleavage in vitro. The inhibitory effect of P1 in H. pylori infections was weaker, but still significant at higher concentrations. The bioactivity of P1 was further confirmed by the inhibition of H. pylori transmigration. Although not investigated in this study, the P1 peptide might be captured by other E-cadherin binders or might be degraded. Although our data provide direct evidence for HtrA inhibition by the P1 peptide, which corroborates the idea that HtrA enzymatic activity is critical for E-cadherin cleavage, an optimisation of substrate-based inhibitors is still necessary.
E-cadherin could play a central role in H. pylori-dependent invasiveness of tumour cells, which is indicated by the finding that the H. pylori virulence factor cytotoxin-associated gene A (CagA) can interact with the IC of E-cadherin, interfering with its tumour-suppressing function 47,48 . It would be compelling to investigate if truncated E-cadherin still binds the intracellular catenin complex or CagA, and if catenin signalling is affected. Importantly, the soluble 90 kDa NTF fragment of E-cadherin does not only serve as a biomarker for GC 49 , but can also impair cell adhesions in a paracrine manner 35 . These data imply that gastric E-cadherin ectodomain shedding is also important in H. pylori-dependent carcinogenesis. In conclusion, our results provide an inspiring novel entry point for innovative anticancer drug discovery.

Cell culture and infection experiments.
The human gastric adenocarcinoma cell line AGS is E-cadherinnegative and was derived from an adenocarcinoma of the stomach of a 54 year-old Caucasian female (ECACC, 89090402). AGS cells expressing hCdh1 were generated by transfection using a plasmid containing E-cadherin wild-type cDNA 50 which has been kindly provided by Rohan D. Teasdale (University of Queensland, Australia). E-cadherin-positive MKN-28 cells (JCRB, 0253, obtained from the Max-Planck Institute for Infection Biology) were originally described as being isolated from a moderately differentiated tubular adenocarcinoma and have been frequently used to study E-cadherin-dependent processes 20,21,23,25 . The gastric epithelial NCI-N87 cells (ATCC, CRL-5822) were included as an additional E-cadherin-expressing cell line 24 and were derived from metastasis in the liver. All cells have been described previously 24,51 and were grown in RPMI 1640 medium containing 4 mM glutamine (Invitrogen) and 10% FCS (Sigma) in a humidified atmosphere at 37 °C. H. pylori wild-type strains P12 52 and Hp26695 53 were cultured on agar plates containing 10% horse serum under microaerophilic conditions for 48 h at 37 °C before infection experiments. Cells were infected with H. pylori at a MOI of 100. Adherence of H. pylori was quantitated in a cfu (colony-forming unit) assay. Briefly, bacteria were suspended in culture medium, added to the cells at a MOI of 100 and co-incubated with host cells for 16 h. The number of cfu was determined by growth on agar plates for 72 h. To investigate the influence of the P1 inhibitor peptide on bacterial transmigration, transwell infections were performed in the presence or absence of 100, 200 or 500 μM peptide. The human gastric adenocarcinoma cell line MKN-28 was grown to confluence on transwell filters and polarised for 14 days. Establishment of functional tight epithelial cell monolayers was further confirmed by measuring the transepithelial electrical resistance (TEER). Infection of these cells was carried out with bacteria at a MOI of 50 for the indicated periods of time per experiment. The number of transmigrated bacteria was determined as cfu by taking aliquots from the basal chambers 21,54 . All infection assays were done in triplicate.
Western blotting. Proteins were separated by SDS PAGE and analysed by Western blot using specific antibodies against HtrA 21 , the extracellular domains EC5 (H108, Santa Cruz Biotechnology and ab40772, Abcam), EC1 (SHE78-7, Invitrogen) and EC2 (HecD1, Calbiochem) of E-cadherin, or the intracellular domain (24E10, Cell Signaling). His 6 -tagged proteins were analysed using an anti-His 5 -antibody (Qiagen). GAPDH (Cell Signaling) and β -actin (Sigma) were detected as loading controls. Where indicated, signals of protein bands were quantified using the ImageLab software (BioRad). Graphical presentation and statistical evaluations (Student's t-test) were performed with GraphPad Prism 5 and p-values of ***p ≤ 0.001, **p ≤ 0.01 and *p ≤ 0.05 were considered statistically significant. 3). All samples were incubated for 12 h at 37 °C in a horizontal shaker. Samples 1 and 2 were applied in sextuplicate, and sample 3 was done in triplicate. Finally, 2 μl of sequencing-grade modified porcine trypsin was added (stock solution of 50 ng/μl in 50 mM HEPES, 1 mM CaCl 2 , pH 7.3, Promega). Protease digestion was carried out overnight at 37 °C in a horizontal shaker. After tryptic digestion, Samples 2 and 1 were mixed at a ratio of 1:10 to obtain Sample 4. Peptides were desalted, purified and concentrated using C18 microcolumns OMIX tips (Agilent Technologies). After lyophilisation, peptides were stored at -20 °C.

In vitro
Nanocapillary HPLC with automated on-line fraction spotting onto MALDI target plates. Peptides were separated by reverse-phase high-performance liquid chromatography using an EASY-nLC system (Proxeon, now Thermo Fisher Scientific) with mobile phase A: 0.1% trifluoroacetic acid (TFA) in water; mobile phase B: 0.1% TFA in acetonitrile; flow rate of 300 nl/min. Lyophilised peptides were dissolved in 23 μl of buffer A, of which 18 μl were loaded onto the column (inner diameter = 75 μm; length = 15 cm; filled with ReproSil-Pur C18 AQ, 3 μm, 120 Å beads; Dr. Maisch GmbH). The peptides were eluted with a gradient of 5-33% B for 62 min, 33-48% B for 15 min, 48-100% B for 2 min and 100% B for 10 min. The column was then equilibrated with 5% B for 20 min before the next sample was analysed. Eluting fractions were mixed with a solution of 3 mg/ml α -cyano-4-hydroxycinnamic acid, 187.5 pmol/ml of each of the four internal standard peptides ([des-Arg 9 ]-bradykinin, neurotensin, angiotensin I and adrenocorticotropic hormone fragment 1-17; all from Sigma), 0.1% TFA and 70% acetonitrile in water and deposed on a blank MALDI target plate (416 spots Scientific RepoRts | 6:23264 | DOI: 10.1038/srep23264 per sample) using an online SunCollect system (SunChrom). The final concentration of each internal peptide standard was 50 fmol per spot.

Mass spectrometric analysis. MALDI-TOF/TOF analysis was carried out with a 4800 MALDI TOF/TOF
Analyzer (AB SCIEX). All spectra were acquired with a solid-state laser (355 nm) at a laser repetition rate of 200 Hz. After measuring all samples in the MS mode, a maximum of 12 precursors per spot were automatically selected for subsequent fragmentation by CID using the mass spectrometer control software (4000 Series Explorer V3.7, AB SCIEX). The resulting spectra were processed and analysed using the Global Protein Server Workstation version 3.6 (GPS Explorer, AB SCIEX), which uses internal MASCOT version 2.1 (Matrix Sciences) software for matching MS and MS/MS data against databases of in silico-digested proteins. Both fully tryptic and semi-tryptic peptides were taken into account. MS/MS data was searched against a database consisting of H. pylori HtrA (UniProt ID: G2J5T2) 55 , glutathione S-transferase class-mu 26 kDa isozyme from Schistosoma japonicum (UniProt ID: P08515), human recombinant E-cadherin Fc His 6 and typical contaminants from recombinant protein expression (yielding a total of 27455 proteins). Possible contaminants included amino acid sequences of human keratins, Escherichia coli and proteins from foetal calf serum downloaded from UniProt (www.uniprot.org). The following analysis settings were used for the identification of peptides and proteins: (i) precursor tolerance: < 15 ppm, (ii) MS/MS fragment tolerance: 0.5 Da, (iii) maximal missed cleavages: 1, (iv) one variable modification (oxidation of methionine). Peptides were considered correct calls if the confidence interval exceeded 95%. After MS acquisition, the data related to the individual peaks (fractions, intensities, mass-to-charge ratios) were loaded into the DeepQuanTR software 56 , which performed a normalisation of individual signal intensities to the internal standard peptides and an annotation (peptide identification and association with a parent protein). Normalised intensities for the individual peptides from all samples of each group (Samples 1-4) were used for the computation of DeepQuanTR peptide and protein scores, indicating the relative abundance of individual peptides and proteins in the sample groups.
N-terminal sequencing. Edman sequencing was performed on an acid-etched glass fibre disk or on a PVDF membrane of an ABI Procise 494 sequencer. Prominent E-cadherin cleavage product bands of the HtrA cleavage assay were eluted and N-terminal sequencing was performed by Alphalyse A/S (Odense).
Comparative model of human E-cadherin. We generated a preliminary protein "homology" model using a mouse E-cadherin crystal structure as template (PDB ID: 3Q2V; resolution = 3.4 Å; sequence identity to human E-cadherin = 82%; coverage = 98%). We used the software Modeller 9.9 after aligning the template with the human E-cadherin sequence (D 155 -I 707 ; UniProt ID: P12830) with ClustalW (www.ebi.ac.uk/Tools/msa/clus-talw2/) 57,58 . The final model was selected by analysing the Ramachandran plots of the computed structures, which we computed with EBI PDBsum Generate (www.ebi.ac.uk/thornton-srv/databases/pdbsum/Generate.html). The Scientific RepoRts | 6:23264 | DOI: 10.1038/srep23264 best model had 434 (90%) residues in the most favoured regions, with only one problematic residue (V 310 ). We did not perform subsequent molecular dynamics relaxation because we only used the model for visual inspection of H. pylori HtrA cleavage sites.