An image-based screen for secreted proteins involved in breast cancer G0 cell cycle arrest

Abstract

Secreted proteins regulate the balance between cellular proliferation and G0 arrest and therefore play important roles in tumour dormancy. Tumour dormancy presents a significant clinical challenge for breast cancer patients, where non-proliferating, G0-arrested cancer cells remain at metastatic sites, below the level of clinical detection, some of which can re-enter proliferation and drive tumour relapse. Knowing which secreted proteins can regulate entry into and exit from G0 allows us to manipulate their signalling to prevent tumour relapse. To identify novel secreted proteins that can promote breast cancer G0 arrest, we performed a secretome-wide, image-based screen for proteins that increase the fraction of cells in G0 arrest. From a secretome library of 1282 purified proteins, we identified 29 candidates that promote G0 arrest in non-transformed and transformed breast epithelial cells. The assay we have developed can be adapted for use in other perturbation screens in other cell types. All datasets have been made available for re-analysis and our candidate proteins are presented for alternative bioinformatic refinement or further experimental follow up.

Fig. 1

Experimental workflow. (a) The screen methodology is depicted starting with the primary screen, including staining, imaging, analysis and hit ranking, followed by a secondary screen of hits and bioinformatic shortlisting. Representative images of each stain are depicted, with CKI (CDK inhibitors) representing p21 or p27. The image analysis pipeline is represented by screenshots from Harmony (PerkinElmer) software. (b) Flow chart depicting how the secretome library was refined at each stage to a final list of proteins which induce G0 arrest.

Fig. 2

Bioinformatic analyses of hits. (a) Correlations between the G0 arrest score per tumour and the expression of selected secretome genes (encoding proteins that were hits in the screen) within the same tumour, calculated pan-cancer (first column) or by cancer type. Abbreviations correspond to different cancer types as defined by TCGA. Positive values (blue) indicate a positive correlation between G0 arrest levels and the expression of the respective gene, negative values (red) indicate an inverse relationship. Significant correlations (Pearson p < 0.05) are marked with an asterisk. (b) Correlations between the G0 arrest score and the expression of selected secretome genes across ER positive and ER negative breast cancers. Colour gradient and asterisk annotations as in (a). (c) Secretome genes with prognostic value in TCGA. The results from Cox proportional hazards analyses for overall survival across all cancers are shown, after adjustment for cancer type and stage. Samples with gene expression in the upper quartile were compared with those whose expression is in the lower quartile for the same gene. Positive values indicate that cases with high expression of the respective gene present worse survival.

Fig. 3

ICAM3 and ICAM5 promote G0 arrest in hTERT-HMEC and MCF7 cells. Graphs show commercial protein supplies of ICAM3 (a) or ICAM5 (b) addition to hTERT-HMEC cells. Mean +/− stdev of three technical repeats is shown. Graphs show dose curves of commercial protein supplies of ICAM3 (c) or ICAM5 (d) addition to MCF7 cells. Mean + /− stdev of three technical repeats is shown (except PBS controls for ICAM5 were nine technical repeats are shown).

Fig. 4

Positive controls for the screen that increase G0 arrest. (a) Representative images of screening controls in MCF7 cells - neutral (PBS), and positive controls for G0 arrest - TGFb1 and Palbociclib. In merged images Hoechst is shown in blue, P-Rb in green, p21 in yellow and EdU staining in red. Scale bar = 50 μM. (b) Frequency distribution of single cell data showing the EdU ratio. Dashed red line represents the cut off between EdU negative (−) and EdU positive cells. (c) Quantification of EdU negative (G0 arrested) cells. Graphs showing the percentage of cells per well which are EdU-. Red line represents mean. Mock = PBS treatment and Vehicle = DMSO. One-way ANOVA was used for statistical significance. ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05, ns = not significant.

Fig. 5

Technical repeats for the secretome screen. Graphs to show the reproducibility of the raw data between the two replica plates for nuclear number (a) and EdU negative (EdU-) fraction (b) in MCF7 cells (p < 0.0001). Graphs to show the reproducibility of the raw data between the two replica plates for nuclear number (c) and EdU negative (EdU-) fraction (d) in hTERT-HMEC cells (p < 0.0001). Note that the data for hTERT-HMEC EdU- fraction is noisy for hTERT-HMEC cells but proteins are only called as a hit if they increased the EdU- (G0) fraction in both technical replicates.