PTPN13 induces cell junction stabilization and inhibits mammary tumor invasiveness

Clinical data suggest that the protein tyrosine phosphatase PTPN13 exerts an anti-oncogenic effect. Its exact role in tumorigenesis remains, however, unclear due to its negative impact on FAS receptor-induced apoptosis. Methods: We crossed transgenic mice deleted for PTPN13 phosphatase activity with mice that overexpress human HER2 to assess the exact role of PTPN13 in tumor development and aggressiveness. To determine the molecular mechanism underlying the PTPN13 tumor suppressor activity we developed isogenic clones of the aggressive human breast cancer cell line MDA-MB-231 overexpressing either wild type or a catalytically-inactive mutant PTPN13 and subjected these to phosphoproteomic and gene ontology analyses. We investigated the PTPN13 consequences on cell aggressiveness using wound healing and Boyden chamber assays, on intercellular adhesion using videomicroscopy, cell aggregation assay and immunofluorescence. Results: The development, growth and invasiveness of breast tumors were strongly increased by deletion of the PTPN13 phosphatase activity in transgenic mice. We observed that PTPN13 phosphatase activity is required to inhibit cell motility and invasion in the MDA-MB-231 cell line overexpressing PTPN13. In vivo, the negative PTPN13 effect on tumor invasiveness was associated with a mesenchymal-to-epithelial transition phenotype in athymic mice xenografted with PTPN13-overexpressing MDA-MB-231 cells, as well as in HER2-overexpressing mice with wild type PTPN13, compared to HER2-overexpressing mice that lack PTPN13 phosphatase activity. Phosphoproteomic and gene ontology analyses indicated a role of PTPN13 in the regulation of intercellular junction-related proteins. Finally, protein localization studies in MDA-MB-231 cells and HER2-overexpressing mice tumors confirmed that PTPN13 stabilizes intercellular adhesion and promotes desmosome formation. Conclusions: These data provide the first evidence for the negative role of PTPN13 in breast tumor invasiveness and highlight its involvement in cell junction stabilization.

and a heated capillary temperature of 250°C. Cycles of one full-scan mass spectrum (350-1500 m/z) at a resolution of 70,000, followed by ten data-dependent MS/MS spectra were repeated continuously throughout the nanoLC separation. All MS/MS spectra were recorded at a resolution of 17,500 with an isolation window of 2 m/z (AGC target 1e5, NCE 26). Data were acquired using the Xcalibur software (v 2.2, Thermo Fisher Scientific, Waltham, MA).
Raw data were analyzed using the MaxQuant software (V. 1.4.1.2). Retention timedependent mass recalibration was applied with the aid of a first search as implemented in the Andromeda software, and peak lists were searched against the UniProt human database (Complete proteome set with isoform; http://www.uniprot.org), 255 frequently observed contaminants as well as reversed sequences of all entries. The standard MaxQuant settings were used. Enzyme specificity was set to Trypsin/P. Up to two missed cleavages were allowed and only peptides with at least seven amino acids in length were considered. Methionine oxidation and serine, threonine and tyrosine phosphorylation were set as variable modifications. Peptide identifications were accepted based on their false discovery rate (FDR, 1%). Accepted peptide sequences were subsequently assembled by MaxQuant into proteins, to achieve a false discovery rate of 1% at the protein level. Relative protein quantification in samples to be compared was performed based on the median SILAC ratios, using the MaxQuant software with standard settings.

Statistical analysis
Proteins quantified in both biological replicates with at least a total number of three peptide evidences were selected. To analyze the normalized SILAC ratios of all peptide evidences of each protein, an empirical estimation of their FDR was performed based on the non-parametric Wilcoxon rank test. Proteins were accepted as differentially enriched based on their FDR (0.05%) and their SILAC ratio (>20% of variation in both biological replicates).

Improving the identification of enriched proteins in SILAC data
Although the SILAC approach is now frequently used, there are still no dedicated tools allowing an exhaustive analysis of the data and there is no consensus on straightforward statistical tests to identify proteins of which the phosphorylation is significantly changed, based on their SILAC ratios. The difficulty in interpreting the values of these ratios is related to the fact that: (i) there are few replicates (often only two), (ii) there is a wide diversity in the number of peptides identified per protein and (iii) there may be errors in assigning peptides to particular proteins.
Furthermore, in quantitative phosphoproteomics one has to combine peptide-level values with protein-level values and perform tests at each level. The correspondence can be achieved in several ways and classical statistical methods designed for microarrays (simple Student's t-test, LIMMA) do not cover this more complex situation or need a lot of replicates. We therefore designed our own bioinformatics pipeline to go through most of the analysis steps, including statistical tests. For the first step of the SILAC analysis, which is associating peptides to proteins and remove contaminants, we relied on the existing MaxQuant software. Then, our pipeline merges experiments reversing some ratios depending on the experimental protocol. Only proteins with at least 3 peptide evidences are selected. The next steps combine peptide-level values to obtain protein scores. We have used the non-parametric Wilcoxon rank test to analyse peptide counts. The p-values obtained for proteins with different number of peptides cannot be compared. To overcome this limitation, we performed false discovery rate (FDR) estimations by comparing the p-value obtained for each protein, with these obtained in thousands of random permutations. The resulting scores can be used to rank all proteins.