Kindlin‐1 modulates the EGFR pathway and predicts sensitivity to EGFR inhibitors across cancer types

Dear Editor, Despite the development of epidermal growth factor receptor (EGFR)-targeted therapies that have revolutionised the treatment of a broad range of cancers, the efficacy of EGFR inhibitors is limited by the lack of primary response or acquired resistance.1-3 In this study,we demonstrated that Kindlin-1 interacts with and regulates the EGFR pathway.4,5 We explored the influence of Kindlin1 expression in cancer cell lines, patient-derived xenograft (PDX) models and cohorts of human tumours to reveal its key role in predicting sensitivity to EGFR inhibitors. Previous works demonstrated that Kindlin-1 and EGFR interact in keratinocytes.6 Thus, we analysed both proteins in a panel of breast cancer cell lines and found that their expression levels were highly correlated (Figure 1A; Figure S1A,B). We next assessed the interaction of Kindlin-1 and EGFR at the endogenous level in BT20 cells. As shown in Figure 1B, EGFR specifically co-immunoprecipitated with Kindlin-1 and reciprocally Kindlin-1 co-immunopreciptated with EGFR (Figure 1B). Immunofluorescence experiments revealed the subcellular colocalisation of both proteins. In the absence of epidermal growth factor (EGF), EGFR mainly localised at the plasma membrane, while Kindlin-1 exhibited dot-like staining at the perinuclear region of BT20 cells. However, after EGF stimulation, promoting EGFR activation, the receptor was internalised and colocalised with Kindlin-1 (Figure S1C). Strikingly, the depletion of Kindlin-1 in these cells resulted in an impaired response to EGF; EGFR remained at the plasma membrane and was no longer internalised (Figure 1C). In addition, Kindlin-1-depleted cells showed decreased levels of extracellular signalregulated kinase (ERK) and EGFR phosphorylation (1.8and 4.5-fold change, respectively; Figure 1D; Figure S1D). Conversely, enforced expression of Kindlin-1 increased ERK phosphorylation (Figure 1E). Thus, in agreement with previous works, our findings suggest that Kindlin-1 interacts with EGFR, affects its internalisation6 and controls EGFR activity in breast cancer cells.

Notably, the activation of EGFR signalling increased Kindlin-1 protein levels, while inhibition of EGFR decreased Kindlin-1 expression, suggesting that Kindlin-1 may act as a downstream effector of the EGFR pathway ( Figure 1F; Figure S1E).
We next analysed Kindlin-1 in 58 breast cancer cell lines from CCLE ( Figure S2A). Kindlin-1 and EGFR transcripts were positively correlated ( Figure S2B), and the correlation was even stronger at the protein level ( Figure 1G-H). Moreover, a gene set enrichment analysis (GSEA) showed enrichment of several oncogenic pathways associated with EGFR/RAS/MAPK signalling in cells highly expressing Kindlin-1 (Figures 1I; Figure S2C and Table S1).
To evaluate the clinical relevance of these findings, we examined Kindlin-1 and EGFR transcripts in breast tumours from The Cancer Genome Atlas (TCGA). Kindlin-1 transcripts were more abundant in EGFR-overexpressing tumours (Figure 2A), and these results were corroborated in an independent series of breast cancer patients treated at Institut Curie (Table S2 and Figure S2D). As shown for breast cancer cell lines, the EGFR pathway was significantly enriched in breast cancer patients with high Kindlin-1 expression ( Figure 2B). Furthermore, an immunohistochemical analysis of both proteins in the same patients (n = 62) also showed high Kindlin-1 expression in tumours exhibiting higher levels of EGFR ( Figure 2C-D). Notably, concomitant Kindlin-1 and EGFR expression was associated with poor patient outcomes; 33% of TCGA triple-negative breast cancer patients with higher expression of Kindlin-1 and EGFR showed a lower metastasis-free survival rate than patients with single overexpression or low expression of both genes ( Figure 2E). We next examined Kindlin-1 and EGFR in other EGFRdriven cancers (TCGA and OncoSG cohorts). 7 Figure S4B). Immunohistochemical analyses also showed increased Kindlin-1 expression in those tumours presenting higher levels of EGFR in HNSCC tumours or mutations of EGFR in lung cancers ( Figure 3C,G; Figure S5). Furthermore, concomitant Kindlin-1 and EGFR expression was associated with poor overall survival in HNSCC and lung and bladder cancers (Figure 3D,H; Figures S3C,D and S4C).
Hence, we evaluated whether Kindlin-1 expression was associated with sensitivity to EGFR inhibitors by interrogating the Genomics of Drug Sensitivity in Cancer database. Kindlin-1 expression was associated with IC50 values for a broad range of EGFR inhibitors in breast, lung, HNSCC and bladder cancer cell lines, suggesting Kindlin-1 as a predictive biomarker of EGFR-targeted therapies regardless of epithelial tissue lineage ( Figure 4A; Figures  S6A and S7). Interestingly, in breast cancer cells, the correlation with the response to cetuximab, a monoclonal antibody widely used in the clinic, was less significant with EGFR (chi square test, P = 0.001 vs. 0.02, respectively; Figure 4B). We then used a cohort of 15 PDXs from triplenegative breast cancers (TNBCs) to test the response to the lapatinib tyrosine kinase inhibitor ( Figure 4C). A significant correlation was observed between Kindlin-1 levels and the tumour growth inhibition induced by lapatinib. The treatment affected PDXs highly expressing Kindlin-1 but not PDX models with low Kindlin-1 levels ( Figure 4D; Figure S6B).
Finally, we analysed a data set of 40 HNSCC patients treated with a first-line cetuximab-based combination. 8 Patients presenting high Kindlin-1 expression experienced a longer progression-free survival (PFS) on cetuximabbased therapy than patients with low expression (median of 468 vs. 254; Figure 4E). These results were validated in an independent series of 18 HNSCC patients treated with cetuximab monotherapy at the Institut Curie. Kindlin-1 was found to be significantly increased in patients whose tumours were sensitive to cetuximab compared to nonre-sponder patients ( Figure S8A; Figure 4F, median of 170 vs. 46). In both HNSCC cohorts, receiver operating characteristic (ROC) analyses showed significant accuracy in predicting cetuximab outcome, with area under the curve (AUC) values indicating that Kindlin-1 was strongly associated with disease control (Figure 4E,F). Similar results were obtained in a series of 110 colorectal cancer patients enrolled in a cetuximab monotherapy trial and a series of 20 non-small cell lung cancer (NSCLC) patients treated with erlotinib. 9,10 Patients with Kindlin-1-overexpressing tumours presented a delayed progression of the disease under EGFR inhibitor treatment compared to patients with low Kindlin-1 tumours ( Figure 4G,H; Figure S8B-D).
In conclusion, our data support Kindlin-1 as a promising predictive biomarker of the response to EGFRtargeted inhibitors. Moreover, our findings provide a rationale for the development of Kindlin-1-based therapeutic approaches that may offer new treatment opportunities for a broad range of cancer patients.