Dual inhibition of protein kinase C and p53-MDM2 or PKC and mTORC1 are novel efficient therapeutic approaches for uveal melanoma

Uveal melanoma (UM) is the most common cancer of the eye in adults. Many UM patients develop metastases for which no curative treatment has been identified. Novel therapeutic approaches are therefore urgently needed. UM is characterized by mutations in the genes GNAQ and GNA11 which activate the PKC pathway, leading to the use of PKC inhibitors as a rational strategy to treat UM tumors. Encouraging clinical activity has been noted in UM patients treated with PKC inhibitors. However, it is likely that curative treatment regimens will require a combination of targeted therapeutic agents. Employing a large panel of UM patient-derived xenograft models (PDXs), several PKC inhibitor-based combinations were tested in vivo using the PKC inhibitor AEB071. The most promising approaches were further investigated in vitro using our unique panel of UM cell lines. When combined with AEB071, the two agents CGM097 (p53-MDM2 inhibitor) and RAD001 (mTORC1 inhibitor) demonstrated greater activity than single agents, with tumor regression observed in several UM PDXs. Follow-up studies in UM cell lines on these two drug associations confirmed their combination activity and ability to induce cell death. While no effective treatment currently exists for metastatic uveal melanoma, we have discovered using our unique panel of preclinical models that combinations between PKC/mTOR inhibitors and PKC/p53-MDM2 inhibitors are two novel and very effective therapeutic approaches for this disease. Together, our study reveals that combining PKC and p53-MDM2 or mTORC1 inhibitors may provide significant clinical benefit for UM patients.


Supplementary Data
Supplementary Materials:

Dosage schedule for in vivo experiments
Gene expression analyses

RPPA analyses
List of antibodies used in the study for western blot analyses. Table S1: Biological characteristics of the five UM PDXs used in the study.

Evaluation of tumor growth of in vivo experiments
Tumor growth was evaluated by measuring with a caliper two perpendicular tumor diameters twice a week. Individual tumor volume, relative tumor volume (RTV) and tumor growth inhibition (TGI) were calculated according to a standard method (35). Tumor stability or shrinkage was defined as a RTV < 1 at the end of experiments. To evaluate the response to each treatment according to individual mouse variability, we have considered each mouse as one tumor-bearing entity. We have defined a relative tumor volume variation (RTVV) of each treated mouse: RTVV=Vt/Vc, where Vt is the volume of the treated mouse and Vc the median tumor volume of the corresponding control group, at the end of treatment. For each mouse, we calculated an overall response rate (ORR) using the formula: ORR=[(RTVV)-1]. A tumor was considered as responding to therapy when ORR was lower than -0.5. Since data were normalized to each control group, results of independent in vivo experiments could be merged. The main interest of the ORR consists of the possibility to combine various models in a same representation due to the normalization performed for each treated PDX and therefore, to allow statistical comparisons in all models included in the experimental program.

Dosage schedule for in vivo experiments
AEB071 was administered per os twice a day, 5 days/week at a daily dose of 120 or 240 mg/kg according to the in vivo experiment design. MEK162 was administered per os twice a day, 5 days/week at a daily dose of 3.5mg/kg. RAD001 was administered per os, 5 days/week at a daily dose of 5mg/kg. CGM097 was administered per os, 5 days/week at a daily dose of 100mg/kg. LEE011 was administered per os, 5 days/week at a daily dose of 75mg/kg. All these doses have been defined based on previous preclinical data. In particular, as RAD001 has been reported at daily doses ranging from 1 (1, 2) to 10 mg/kg/day (3), and because of the absence of toxicity when RAD001 was administrated alone and in combination with various other compounds using a dosage of 5 mg/kg per day, we have decided to maintain this schedule in order to perform the efficacy studies in our UM PDXs.
Similarly, the dose of CGM097 was defined according to previous studies performed by Novartis, showing that a dose of 100 mg/kg appears to be well tolerated and efficacious in multiple xenografts studies, and leads to an exposure that is currently observed in patients (Novartis unpublished data).
There were two sets of in vivo independent experiments: a first one testing dose-dependent efficacy of AEB071 administered alone, and a second one testing AEB071 (120 mg/kg) alone and in combinations (AEB071+RAD001/CGM097/MEK162/LEE011). Hence, in this second set of experiments, control groups are similar in each model for all tested combinations, and the doses of each compound were identical in all tested combinations and in each model.

Gene expression analyses
Transcriptome profiling of patient tumors and corresponding PDXs was defined using Affymetrix Human Exon 1.0 ST array. RNA isolation, preparation, hybridization and raw data were obtained as previously described (Laurent C et al 36). Briefly, gene expression levels were estimated using custom Brainarray CDF (version 13) based on EntrezGene database.
Data were summarized and normalized using RMA method. Raw and normalized data are publicly available on GEO repository under the accession number GSE78033 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE78033).
Transcriptome profiling of each studied PDX model at passage 1 were studied. Among these 5 samples, 500 most variant genes based on IQR measure (Inter Quartile Range) were selected. Clusters of samples based on these most informative genes were identified by hierarchical clustering using Pearson's correlation as metric to measure the distance between observations and Ward's method as linkage criterion to measure the distance between sets of observations.

RPPA analyses
An untreated tumor sample was collected for each model and from similar in vivo passages between them. These samples were analyzed for the normalized expression level of 93 proteins or phospho-proteins by RPPA as described in Laurent et al., 2013. The list of antibodies used in this study is shown in Supplementary Table S7. The five models were classified as low or high responders for each treatment. We performed pairwise comparisons by t-tests (two sided, 95% confidence intervals) between the both response groups for each treatment when it was possible (presence of a differential response and at least two models per group) and for each protein. All significant results are summarized in Supplementary   Table S8.      ° All combinations induced complete remissions, 1/7 after AEB071 + MEK162 (14%), 4/7 after AEB071 + RAD001 (57%), 2/7 after AEB071 + CGM097 (29%), and 3/7 after AEB071 + LEE011 (43%).

Table S6: p-values of combinations versus monotherapy treatments in vivo
See pdf file p-values were calculated for each two by two treatment comparisons. Each comparison was done between the treatments indicated in the columns B and C. Significant differences are highlighted in red.

See pdf file
In total, we selected 93 proteins or phospho-proteins (center column) belonging to 18 signaling pathways or functions (left column) frequently altered in cancer. The supplier and reference of each antibody used in this RPPA study are specified in the right column.             and S13. Five 3-fold serial dilutions were tested. AEB071 was tested from 0 to 2 µM in all lines. RAD001 and CGM097 were used from 0 to 0.1 µM and from 0 to 2 µM respectively.
The averages between triplicate points were made and percentages of growth inhibition relative to DMSO-control treatment are represented ± SEM.