Proximity-labeling proteomics reveals remodeled interactomes and altered localization of pathogenic SHP2 variants

This data respository contains data associated with a study to characterize how missense mutations in PTPN11, the gene encoding the protein tyrosine phosphatase SHP2, alter the protein-protein interactions and localization. SHP2 has a catalytic domain that dephosphorylates proteins and two phosphotyrosine-binding SH2 domains. One part of this respository contains data that were used to map the sequence recognition profiles of the two wild-type SH2 domains of SHP2. These sequence recognition profiles were used to predict binding sites for each SH2 domain across the human proteome, and they are juxtaposed with the proteomics data presented in the parent study. Another part of the repository contains microscopy data examining the mitochondrial localization of wild-type SHP2. A third part includes a structural model of SHP2 bound to an interactor, PPIF, identified from proximity-labeling proteomics experiments.

Methods
The deep sequencing data were generated using a high-throughput method for profiling the sequence specificities of SH2 domains. This method has been documented extensively in two published manuscripts:

https://doi.org/10.7554/elife.82345

https://doi.org/10.1073/pnas.2407159121

The overall approach entails the use of degenerate, genetically-encoded peptide libraries, with a structure X5-Y-X5, where X is any of the 20 canonical amino acids and Y is tyrosine. The libraries (approximately 1 million random sequences) are displayed on the surface of E. coli cells, then enzymatically phosphorylated using a mixture of tyrosine kinases. Then, SH2 domains immobilized on magnetic beads are used to enriched those phosphorylated cells with optimal peptide sequences for the particular SH2 domain. These cells are isolated, and DNA encoding the peptides is amplified by PCR and subject to deep sequencing. An input library that has not been selected by any SH2 domain is also sequenced.

Next, the sequenced libraries are translated from DNA to peptide sequences. The frequency of each amino acid at each position in the selected library is counted and normalized to that in the input library. These normalized matrices are used to calculate a binding score for all documented tyrosine phosphorylation sites seen in the PhosphoSitePlus database (https://www.phosphosite.org/homeAction.action). The scoring method is described in detail here: https://doi.org/10.7554/elife.82345

The raw fastq files, the translated sequence files, count matrices, and calculated scores, along with the scripts to generate these scores, are provided in this data repository. The analysis pipeline is also extensively documented here: https://doi.org/10.7554/elife.82345

The confocal microscopy data were acquired under oil immersion 60x magnification (Nikon, MRD71670) using a confocal spinning disk microscope (Andor Dragonfly) coupled to a Nikon Ti-2 inverted epifluorescence microscope with automated stage control, Nikon Perfect Focus System and a Zyla PLUS 4.2-megapixel USB3 camera. Illumination was done with 100 mW 405 nm, 50 mW 488 nm, 50 mW 561 nm and 140 mW 640 nm solid-state lasers. All hardware was controlled using Andor Fusion software. Lasers, laser powers, exposure times, objectives and experiment-specific acquisition parameters are 100% power 100ms exposure for all the images. Images were acquired with 11 z-slices at 2.0-μm intervals (Total scan size 20 μm). The images are in .ims format and can be opened in software such as ImageJ/Fiji. A metadata file for the microscopy data is also provided.

The AlphaFold 3 model is based on the canonical isoforms of full-length human SHP2 (Uniprot Q06124) and full-length (Uniprot P30405), modeled as a 1:1 complex using the AlphaFold Server (https://alphafoldserver.com/). The default parameters were used, with no post-translational modifications on either chain.