The sorting protein PACS-2 promotes ErbB signalling by regulating recycling of the metalloproteinase ADAM17

The metalloproteinase ADAM17 activates ErbB signalling by releasing ligands from the cell surface, a key step underlying epithelial development, growth, and tumour progression. However, mechanisms acutely controlling ADAM17 cell-surface availability to modulate the extent of ErbB ligand release are poorly understood. Here, through a functional genome-wide siRNA screen, we identify the sorting protein PACS-2 as a regulator of ADAM17 trafficking and ErbB signalling. PACS-2 loss reduces ADAM17 cell-surface levels and ADAM17-dependent ErbB ligand shedding, without apparent effects on related proteases. PACS-2 co-localizes with ADAM17 on early endosomes and PACS-2 knockdown decreases the recycling and stability of internalized ADAM17. Hence, PACS-2 sustains ADAM17 cell-surface activity by diverting ADAM17 away from degradative pathways. Interestingly, Pacs2-deficient mice display significantly reduced levels of phosphorylated EGFR and intestinal proliferation. We suggest that this mechanism controlling ADAM17 cell-surface availability and EGFR signalling may play a role in intestinal homeostasis, with potential implications for cancer biology.

The amount of co-immunoprecipitated PACS-2 was normalized to the amount of immunoprecipitated mature ADAM17. The unstimulated negative control for each experiment was then set to 1, the other raw data were normalized to this value, and finally the average of all individual experiments was calculated. Data were analysed by unpaired two-tailed Student's t-test.  Fig. 6. Levels of total EGFR were assessed in small intestinal (SI) lysates from 6 control and 6 Pacs2-/-mice. The fold change was obtained by normalizing EGFR levels to actin, calculating the average value in controls and setting this to 1, then normalizing all the other raw data to this value, and finally calculating the average for each genotype. EGFR pY1068 levels were assessed in intestinal lysates from 5 control and 7 Pacs2-/-mice. pEGFR levels were quantified and shown as a ratio relative to total EGFR levels. The fold change was obtained by first calculating the average value in controls and setting this to 1, then normalizing all the other raw data to this value, and finally calculating the average for each genotype. (b) Total cellular levels of EGFR were assessed in control and Pacs2-/-MEFs serum-starved overnight. Samples were normalized to input actin, and the fold change calculated by setting the control for each experiment to 1, normalizing the Pacs2-/-data to this value, and calculating the average of all individual experiments.
As in (b), pEGFR levels are shown as a ratio relative to total EGFR levels. Data were compiled from 5 individual experiments for total EGFR and from 3 individual experiments for pEGFR. On blots, # denotes a non-specific band. Graphs show mean values ± standard error of the mean. Data were analysed by unpaired two-tailed Student's t-test. *p<0.05, **p<0.01.

Criteria used to define positive hits
All data were processed using the cellHTS2 software in R Bioconductor (http://www.bioconductor.org/packages/devel/bioc/html/cellHTS2.html). Raw fluorescence intensity values were normalized to the plate median value and Z-scores were calculated by subtracting the population mean from the individual raw values and dividing the difference by the population standard deviation. Z-scores from plate replicates were averaged and genes preventing loss of AP-HB-EGF cell-surface levels after PMA treatment with no effect on cell number were then selected according to the criteria: mean Z-score ≥ 3.5 after PMA treatment and no effect on cell number (mean Z-score ≥-1 and ≤1) (Supplementary Fig. 1a+b). In addition, a few genes with mean Z-score ≥ 10 after PMA treatment were allowed a minor loss in cell number (mean Z-score ≥-1.3) (Supplementary Fig. 1a). Finally, genes affecting unstimulated AP-HB-EGF levels were identified using the following criteria: mean Z-score ≥ 3.0 or ≤-3.0 without PMA treatment and no effect on cell number (Supplementary Fig. 1a).
Based on this selection, the primary screen resulted in the identification of 645 candidate genes affecting AP-HB-EGF levels after PMA treatment and 187 genes affecting unstimulated AP-HB-EGF levels.

Deconvolution screen
To validate the selected screen hits, the deconvoluted siGENOME library from Dharmacon was used. Several genes selected in the primary screen were not targeted by the newer deconvoluted library. In brief, the 4 individual siRNAs constituting the SMARTpool were each tested in triplicate using the same assay conditions as for the primary screen, except that depending on whether selected genes affected unstimulated or PMA-stimulated AP-HB-EGF cell-surface levels, they were only tested under those conditions. Moreover, rather than normalizing raw values to the plate median, normalization to mean RISC-free negative controls on each plate was performed. In order to qualify as a hit, at least 3 out of the 4 individual siRNAs from the pool had to give rise to more than 500% increased mean fluorescence intensity as compared to the negative control ( Supplementary Fig. 1c).
A total of 81 genes (15% of the 548 deconvoluted hits) were confirmed with at least 3 independent siRNA oligonucleotides resulting in more than 500% increased mean fluorescence intensity after PMA treatment, as compared to non-targeting siRNA controls (Supplementary Fig. 1c+d and Supplementary Table 1). Both ADAM17 and protein kinase C (PKC)  were among these, ratifying the screen design. Moreover, 7 genes affected the surface level of AP-HB-EGF in unstimulated cells (3 independent siRNAs causing more than 200% increased mean fluorescence intensity or exhibiting less than 50% mean fluorescence intensity, as compared to non-targeting siRNA controls) ( Supplementary Fig. 1d). Yet, these 7 genes showed opposite effects in the deconvolution screen versus the primary screeni.e. increased levels in the primary screen and reduced levels in the secondary screen, and were therefore excluded. Hits were further evaluated using an independent colorimetric AP shedding assay (see experimental procedure below), which revealed that knockdown of 24 genes phenocopied ADAM17 knockdown (Supplementary Table 1).
We functionally categorized the genes using the DAVID bioinformatics database (http://david.abcc.ncifcrf.gov/). Multiple categories of genes were enriched based on Panther Molecular Function Gene Ontology terms 1 (Supplementary Fig. 1e and Supplementary Table 1). No major signalling pathways appeared using Ingenuity pathway analysis (www.ingenuity.com). All screen data are available at http://hts.cancerresearchuk.org/db/public/index.php.