Scaffolding protein connector enhancer of kinase suppressor of Ras 1 (CNKSR1) regulates MAPK inhibition responsiveness in pancreas cancer via crosstalk with AKT signaling

Abstract

Combinatorial molecular therapy in pancreatic ductal adenocarcinoma (PDAC) has yielded largely disappointing results in clinical testing to-date as a multitude of adaptive resistance mechanisms is making selection of patients via molecular markers which capture essential, intersecting signaling routes challenging. Here, we report the scaffolding protein connector enhancer of kinase suppressor of Ras 1 (CNKSR1) as mediator of resistance to mitogen-activated protein kinase (MEK) inhibition. MEK inhibition in CNKSR1^high cancer cells induces translocation of CNKSR1 to the plasma membrane where the scaffolding protein interacts with and stabilizes the phosphorylated form of AKT. CNKSR1-mediated AKT activation following MEK inhibition was associated with increased cellular p-PRAS40 levels and reduced nuclear translocation and cellular levels of FoxO1, a negative regulator of AKT signaling. In clinical PDAC specimens, high cytoplasmatic CNKSR1 levels correlated with increased cellular phospho-AKT and mTOR levels. Pharmacological co-blockade of AKT and MEK ranked top in induced synergies with MEK inhibition in CNKSR1^high pancreas cancer cells among other inhibitor combinations targeting known CNKSR1 signaling. In vivo, CNKSR1^high pancreatic tumors treated with AKT and MEK inhibitors showed improved outcome in the combination arm compared to single agent treatment, an effect not observed in CNKSR1^low models.

Our results identify CNKSR1 as regulator of adaptive resistance to MEK inhibition by promoting crosstalk to AKT signaling via a scaffolding function for the phosphorylated form of AKT. CNSKR1 expression might be a possible molecular marker to enrich patients for future AKT-MEK inhibitor precision medicine studies.

Introduction

Molecular therapies targeting oncogenic signaling have yet to yield impactful improvements of overall outcome in pancreatic ductal adenocarcinoma (PDAC). PDAC, a lethal disease with dismal 5-year survival rates below 10 percent, is notoriously resistant to most forms of systemic therapy (1, 2). The anti-EGFR inhibitor erlotinib (Tarceva^™), the first molecular therapy agent approved by FDA, received regulatory approval based on minimal survival gains over single agent gemcitabine. (3). More recent regulatory approvals of molecular therapy agents apply to small subgroups of pancreatic cancer patients only. These include the PARP inhibitor olaparib approved for maintenance treatment of metastatic PDAC in the context of germline BRCA mutations which are detected in 5–7% of PDAC patients (4). The immune checkpoint inhibitor pembrolizumab has been approved for microsatellite instability-high (MSI-H) and the neurotrophic receptor tyrosine kinase (NTRK) inhibitors larotrectinib and entrectinib for NTRK gene fusion-positive tumors (5, 6).

The oncogene KRAS is known to be affected by activating mutations in more than 90 percent of PDAC cases. Thus, targeting oncogenic signal transduction of KRAS effectors may impact the overall PDAC outlook substantially. Unfortunately, early targeted therapies blocking mitogen-activated protein kinase (MAPK) pathway signaling via mitogen-activated protein kinase (MEK) inhibition, either alone, or in combination with systemic chemotherapy, did not show any improvement in clinical outcomes compared to standard of care (7–9). After single arm combination molecular therapy trials such as combined MEK and EGFR inhibition failed to show efficacy, the large Intergroup trial S1115 (NCT01658943) tested the combination of orally administered MEK and AKT inhibitors by randomizing patients with advanced PDAC who had received 1^st line gemcitabine-based chemotherapy to the MEK inhibitor selumetinib (AZD6244) and AKT inhibitor MK-2206 combination versus modified oxaliplatin plus fluorouracil (mFOLFOX) chemotherapy (10, 11). While survival outcomes in the chemotherapy arm were superior, there were objective response and stable disease observations in patients treated with dual MEK-AKT inhibition which may indicate antitumor activity by the combination molecular therapy regimen in some patients (11). Enrichment of patients more likely to benefit from the combination approach via e.g. molecular markers which indicate cross talk and adaptive activation of AKT and MAPK pathway signaling might have improved outcome in the experimental therapy arm.

Adaptive resistance to molecular therapy targeting single pathways is notoriously common in PDAC and other solid organ cancers and typically occurs rapidly via induction of genes and activation of redundant signaling pathways circumventing the original blockade (12, 13). Considering the interconnectivity of the MAPK and PI3K/AKT pathways, molecular regulators mediating crosstalk between the two signal transduction axes appear to be attractive candidates for biomarkers for combinatorial MEK-AKT inhibition. Indeed, in preclinical models mammalian target of rapamycin (mTOR) complex 1 (mTORC1) activation, insulin receptor substrate 1 (IRS-1)-mediated positive feedback, or unique KRAS effector signaling via 90 kDa ribosomal S6 kinase (RSK) mediate MAPK-AKT pathway crosstalk, are linked with cooperative responses to dual small molecule inhibition, and have been suggested as predictors of response to dual MEK-AKT inhibition (12, 14, 15). In this regard, it is surprising that scaffolding proteins have not yet been investigated as potential markers for MAPK-AKT pathway crosstalk and possible guides for dual molecular therapy. Protein scaffolds integrate output of signal transduction cascades by regulating spatiotemporal assembly of pathway components, both within individual signal transduction axes as well across different signaling pathways. By either increasing signaling output, or imparting assembling and reducing effector signaling protein scaffold have been shown to modulate drug responses including response to MEK inhibition (16–18). For example, kinase suppressor of Ras (KSR) and connector enhancer of kinase suppressor of Ras 1 (CNKSR1), some of the more well studied scaffold proteins, interact with and regulate a diverse range of binding proteins and intracellular regulators which are not limited to kinases, small GTPases, or the cell cycle regulator Ras association domain family 1 isoform A (RASSF1A) but also include membrane receptors or phosphoinositide (PtdIns) second messenger lipids (19–24).

Here, we conducted a loss-of-function whole kinome RNAi screen in MEK-resistant pancreas cancer cells and identified the scaffold protein CNKSR1 as a novel mediator of resistance to MEK inhibition. CNKSR1 mediated MEK resistance via crosstalk to AKT. MEK inhibition induced plasma membrane localization of CNKSR1, interaction, and stabilization of the phosphorylated form of AKT. Co-treatment improved MEK inhibitor responses in CNKSR1^high but not CNKSR1^low cells. In line with increased AKT activation levels detected in pancreatic cancer tissues with high CNKSR1 expression levels, these findings suggest that pancreatic cancers with CNKSR1^high expression status might be selectively susceptible to dual MAPK-AKT pathway blockade.

Materials and Methods

Cell lines and materials

The origins of the human pancreatic cancer cell lines derived from primary and secondary sites used in this study are listed in Suppl. Table 1. Cells were maintained according to instructions from supplier or in RPMI 1640 medium with 10% (v/v) FBS and incubated at 37°C in a 5.0% CO₂ atmosphere. In accordance with AACR practices identities of used cell lines were confirmed by SNP genotyping using Illumina MiSeq sequencing. Cells were tested for the presence of mycoplasma using the MycoAlert^™ Mycoplasma Detection Kit from Lonza, (Basel, Switzerland). Antibodies including used dilutions used for immunoblotting and immunofluorescence studies are listed in Suppl. Table 1. Selumetinib (AZD6244; Cat.#S1008), Rapamycin (AY-22989, Cat.#S1039), MK-2206 2HCL (Cat.#S1078) were purchased from Selleck Chem, Inc. (Houston, TX).

RNAi screening

YAPC cells were obtained from the Leibniz Institute DSMZ German Collection of Microorganisms and Cell Cultures GmbH (Braunschweig, Germany) and maintained in RPMI containing 10% fetal bovine serum (FBS). Transfections were performed in 384-well plates (Cat# 3570, Corning Inc, NY). Cell viability was measured using CellTiter Glo (Promega). For transfections, 20μL of serum free media containing Lipofectamine RNAiMax (0.08μL) was added to wells containing siRNA (0.8μpmol). Lipid and siRNA were allowed to complex for 45 min at ambient temperature before addition of 1,000 cells in RPMI, 20% FBS to yield final transfection mixtures containing 20 nM siRNA in RPMI, 10% FBS. The kinome screen was conducted using Ambion’s Silencer Select Human Kinase Library V4. This library targets 704 human genes with 3 independent siRNAs per gene. Each siRNA is arrayed in an individual well. Selumetinib (10 μM AZD6244) or vehicle (0.1% DMSO) was added to the entire plate 48 h post-transfection and viability using the CellTiter-Glo^® Luminescent Cell Viability Assay (CellTiter Glo, Promega) was determined 72 h later on a PerkinElmer Envision plate reader. Ambion Silencer Select Negative Control #2 was incorporated on all screening plates for normalization (16 wells per plate). Qiagen’s AllStars Cell Death control was incorporated as a positive transfection control (16 wells per plate).

To select candidate genes that modulate selumetinib activity, the ratio of vehicle only (VO)-treated cell viability (%siNeg) versus selumetinib-treated cell viability (%siNeg) was calculated for each siRNA and these values were converted into robust z-scores. Genes were prioritized if more than one corresponding siRNA exhibited a robust z-score > 2. Genes selected from either replicate screen were considered as top candidates (i.e. the union of candidates from both screens). All screening data can be found in Supplemental Table 1.

Ingenuity Pathway Analysis (Ingenuity^® Systems, www.ingenuity.com) was performed to identify enriched protein-protein interaction networks among the identified candidates. The IPA network Score was based on the hypergeometric distribution and calculated as the -log of a right-tailed Fisher’s Exact Test using only direct relationships of the identified candidates versus the 704 genes represented in the screen as background. Networks were ranked by IPA network scores.

NanoPro capillary isoelectric focusing immunoassays

NanoPro 1000 is an automated capillary based isoelectric-focusing immunoassay system (ProteinSimple, Santa Clara, CA; http://www.proteinsimple.com). Protein isolation, detection and quantification were done as per manufacturer’s instructions. Immunoprobing was done using primary antibodies obtained from Cell Signaling Technology (anti-phospho-ERK) and Millipore (anti-ERK1/2), and then probed with horseradish peroxidase (HRP)-conjugated, goat anti-rabbit secondary antibodies (Jackson ImmunoResearch). All antibody incubation and wash steps were programmed and performed automatically in the NanoPro system; incubation time for primary antibodies was 2 hrs and for secondary antibodies 1 hr. A 1:1 The digital image was analyzed and quantified with Compass software (ProteinSimple) according to the manufacturer’s instructions. Experiments were conducted in triplicates.

Cell proliferation and apoptosis assay

The inhibitory concentration 50 (IC₅₀) dose level of selumetinib (AZD6244) was calculated for each cell line using the CellTiter-Glo^® assay, after exposure to a serial dilution of drug in a 96 well plate format for 72 h. Live cells were plated at 3,000 to 5,000 cells/well in 100uL standard media. Following incubation overnight, serial dilations of a 10mM selumetinib stock solution in a volume of 1μL were added. After 72 hrs relative cell growth was assessed using the CellTiter-Glo^® assay with levels of untreated cells normalized to 100%. Apoptosis was evaluated at the same time using the Caspase-Glo^®3/7 assay (Promega, Madison, WI).

Selumetinib dose response and cell proliferation following siRNA transfection

Cells were plated at a concentration of 3000–5000 cells/well in 100uL media containing antibiotic free media with the addition of 0.3–0.4 μL RNAi Max transfection agent (Life Technologies, Carlsbad, CA) and 50–100nM siRNA (GeneSolution, Qiagen, Valencia, CA) reconstituted in RNase free water. Following transfection for 48 h, cells were then exposed to selumetinib for 72 h and final cell viability and apoptosis were measured using the CellTiter-Glo^® assay and Caspase-Glo^®3/7 assay, respectively. To determine impact of siRNAs onto cell viability in the absence of selumetinib, dose response testing analyzed as a percentage of the number of cells present at treatment start. Data analysis was performed using GraphPad Prism version_7 software. Drug response curves were created using a four-parameter equation fitting technique. Cell proliferation of siRNA transfected cells was determined at 24, 48, 72 and 96 hours using the CellTiterGlo assay.

Gene expression analysis of microarrays

Total RNA was extracted from cancer cell lines grown at 60 to 80% confluency using RNeasy Mini Kit (Qiagen, Valencia, CA). RNA concentration and integrity were analyzed with Agilent 2100 Bioanalyzer system (Agilent Technologies, Columbia, MD). Biotinylated cRNA was generated with Illumina TotalPrep^™-96 RNA Amplification Kit (Life Technologies, Grand Island, NY). Biotinylated cRNA was then hybridized to HumanHT-12 v4 Expression BeadChip (Illumina, San Diego, CA). Array signal was normalized to expression levels of housekeeping genes and log2-transformed. Unsupervised hierarchical clustering analysis was conducted using GeneSpring v12.6 (Agilent Technologies, Columbia, MD). Per-probe normalization was applied by subtracting the log2 signal intensity of the median value for a specific probe from the log2 signal intensity of each cell line. Hierarchical clustering was then performed using GeneSpring default settings, the default workflow options (Threshold: 0; p value: 1; Signals: both) were used for analysis. Pathway analysis was performed using Ingenuity Pathway Analysis (IPA; https://digitalinsights.qiagen.com/) software, top-ranking networks of imputed gene probes are shown.

Phospho-immunoblot analysis and immunoprecipitation

Pancreas cancer cells were lysed with M-PER^™ Mammalian Protein Extraction Reagent (Cat.# 78501, Thermo Scientific^™, Waltham, USA). Protein concentration was determined via BCA analysis kit (ThermoScientific, Waltham, USA). For phospho-immunoblotting using p-ERK Thr202/Tyr204 (Cat.#4376), p-AKT S473 (Cat.#4060), approximately 200–300 μg of protein was loaded, for total anti-ERK (Cat#9101), AKT (Cat#C67E7), actin (Cat#4967, all Cell Signaling, Danvers, USA) and CNKSR1 (Cat# sc-514607, Santa Cruz Biotechnology, Dallas, TX) 5–10 μg, onto 4–20% SDS/Polyacrylamide gels. Proteins were transferred to nitrocellulose blotting paper via the iBlot 2 Dry Blotting System (ThermoScientific, Waltham, USA). Bands were visualized via the Odyssey luminescence scanner (Li-Cor, Lincoln, USA). For co-immunoprecipitation studies, anti-CNK1 Antibody (G-7) conjugated to agarose (Cat#sc-514607, Santa Cruz Biotechnology, Dallas, USA), anti-p-AKT (1:50, Cat#12694, Cell Signaling Tech, Danvers, USA) or anti-t-AKT (1:50; Cat#2920, Cell Signaling Tech, Danvers, USA) was incubated at 4°C overnight with 2000 μg cell lysate. After 6-hour incubation with the mixture of antibody-lysate at 4°C, Protein G Magnetic Beads (Cat#88848, ThermoScientific, Waltham, USA) were washed and eluted via addition of 40 μL elution buffer and 20 μL NuPAGE^™ LDS sample buffer (Cat# NP0007, ThermoScientific, Waltham, USA) at 70°C, 10 mins. Agarose beads were heated at 100°C with 60ul NuPAGE^™ LDS sample buffer. Elute was analyzed by immunoblotting with the Quick Western Kit (Li-Cor, Lincoln, USA).

Tissue microarray (TMA) and immunohistochemistry

De-identified pancreatic ductal adenocarcinoma tissues included in NCI Surveillance, Epidemiology, and End Results (SEER) Residual Tumor Registries pancreatic cancer tissue microarray (TMA) were previously described (25). Appropriate ethical and transfer of material approvals were obtained from originating sites, as well as the NCI. Immunohistochemical staining for CNKSR1 (mouse monoclonal antibody CNKSR1 (clone 46), Santa Cruz Biotechnology, TX, USA, #sc-135,870; dilution 1:200) was performed on a Leica BOND-MAX autostainer (Leica Microsystems, IL, USA). Antigen retrieval was for 25 min with Bond Epitope Retrieval Solution 2 (Leica Biosystems Newcastle, UK, #AR9640). Primary antibody was incubated for 30 min at room temperature. For detection the BondMax avidin biotin free polymer-based detection system (Bond Polymer Refine Detection #DS9800) was used with diaminobenzidine as chromogen. CNKSR1 cytoplasmatic staining was quantified as previously described (25). CNKSR1 expression was evaluated based on intensity semi-quantitatively on a four-tier scale (0 = negative, 1 = weak/background, 2 = moderate/positive, 3 = strongly positive). CNKSR1 showed minimal expression in lymphoid tissues according to RNA-Seq data and immunohistochemical staining from the Human Protein Atlas (Human Protein Atlas available from www.proteinatlas.org) and samples of lymph nodes and intratumoral lymphocytes were therefore used as negative controls. Immunostaining for p-AKT (S473) and p-mTOR (S2448) was developed on adjacent or consecutive sections using a rabbit monoclonal antibody at 1:200 dilution. Phosphorylated AKT and mTOR staining was predominantly cytoplasmatic with very few cases demonstrating nuclear staining as well. Scoring of p-AKT and p-mTOR was done blinded to the CNKSR1 results using standard intensity scores above.

Immunofluorescence and proximal ligation assay

50,000 cells were seeded onto 8-well chamber slides (Cat.#80827, Ibidi, WI, USA) and treated with 10μM selumetinib (AZD6244) or vehicle for 24 hrs at 37°C followed by fixation with 4% paraformaldehyde for 15 min, permeabilization with 0.3% Triton X-100 for 5 min, and blocking with 3% BSA in PBS for 1 h. After blocking, cells were incubated with primary antibody (Suppl. Table 1) for 1 h at RT. Alexa Fluor^® 488 goat anti-rabbit IgG (H + L) (Cat. #A27034; ThermoFisher Scientific, Inc., Waltham, MA) or Alexa Fluor^® 555 goat anti-mouse IgG (H + L) (Cat. #A-21422) secondary antibody was then applied for 1 hr at room temperature. Slides were mounted with Vectashield/DAPI (Vector Laboratories, Burlingame, CA). Images were captured using a Zeiss LSM 510 UV or Zeiss LSM 780 confocal microscope (Zeiss, Thornwood, NY). Three separate images for each treatment group containing about 50 cells were analyzed using ImagePro software (Media Cybernetics, Inc, Rockville, MD). PLAs via the conducted using the kit Duolink^® In Situ Red Starter Kit Mouse/Rabbit (Millipore Sigma, Burlington, MA) following the manufacturer’s instructions.

YAPC and PSN-1 orthotopic pancreatic cancer model

All animal procedures were approved by the National Cancer Institute Animal Care and Use Committee (ACUC) of NIH. 100,000 YAPC or PSN-1 cells were transplanted via left subcostal incision, dissection of the distal pancreas, and injection into the tail of NOD-scid IL2Rgamma^null (NSG) immuno-deficient mouse pancreata (mouse strain #005557, The Jackson Laboratory, FL). Mice were administered either vehicle, 25mg/kg of selumetinib (AZD6244) administered in 4% DMSO+30% PEG300+5% Tween 80+ddH2O, 120mg/kg MK-2206 administered in 20% SBE-β-CD in saline, or the combination of both via intraperitoneal (IP) injection daily starting 21 days post-transplantation and continuing until mice became moribund or were required to be euthanized as directed by the veterinary team which was blinded to the treatment allocations. Survival was analyzed using the Kaplan-Meier method and groups were compared with the log-rank test.

Flow cytometry analysis

Pancreatic tumors were harvested, washed with PBS, minced with a scalpel, and digested using Tumor Dissociation Kit, mouse with a Gentle Macs Agitator (Miltenyi Biotec, San Diego, CA). Tumor lysates were passed through a 70μm filter, washed in PBS, and stained with the Live/Dead Fixable Blue Dead Cell Stain Kit (ThermoFisher Scientific) and antibodies listed in Supplementary Table 1. Stained cells were washed with FACS buffer prior to sample acquisition with the BD LSRFortessa SORP I flow cytometer (BD Bioscience). Flow cytometry data was analyzed using FlowJo software (TreeStar, Ashland, OR).

Statistical analysis

Continuous data were compared by paired Student t test using GraphPad Prism (Version 7.01). Flow data analysis was performed by one-way ANOVA with the post-hoc Tukey test. Error bars indicate standard error of the means (SEM) unless otherwise indicated. Calculated p values were given by number and asterisk(s). *, p < 0.05; **, p < 0.01; ***, p < 0.001.

Results

RNAi loss-of-function screening identifies CNKSR1 as modulator of MEK inhibition in pancreas cancer cells

To identify target genes that sensitize pancreas cancer cells to inhibition of the MAPK pathway using the allosteric MEK1/2 inhibitor selumetinib, we conducted a kinome-wide small interfering RNA (siRNA) loss-of-function screen in YAPC pancreatic cancer cells treated with selumetinib. YAPC pancreatic cancer cells harbor KRAS^G12V and TP53^H179R-loss of function mutations and are resistant to MAPK pathway inhibition (Suppl. Fig. 1A). In line with previous findings that MAPK inhibition in pancreas cancer induces a cytostatic growth arrest effect rather than induction of cell death, neither MEK nor ERK inhibition induced cleaved caspase (up to 7 days) in YAPC cells despite phosphorylation levels of ERK isoforms remaining suppressed (Suppl. Fig. 1B to D) (26). The RNAi screen comprised the Ambion Silencer Select Human Kinase Library V4 targeting 704 individual genes with 3 independent siRNAs per gene. For the screen, YAPC cells were transfected with siRNA and treatment with selumetinib (10μM) or vehicle (0.1% DMSO) was started 48 h post-transfection (Fig. 1A). Viability was assayed after an additional incubation of 72 h. Screening was performed in duplicate. Replicate screens exhibited good correlation both in terms of vehicle and selumetinib treatment conditions (Suppl. Fig. 1E). To select candidate genes that modulate the activity of selumetinib, median absolute deviation (MAD) z-scores of vehicle- versus selumetinib-treated cell viability were calculated for each siRNA after normalization to Ambion Silencer Select Negative (siNeg) control (Fig. 1B). The ratio of vehicle-treated cell viability (% siNeg) versus selumetinib-treated cell viability (% siNeg) was calculated for each siRNA and genes were prioritized if more than one corresponding siRNA exhibited a robust z-score > 2 (Fig. 1C). Genes which were per se lethal after silencing in the vehicle-treated group (viability ≤50% compared to Ambion Silencer Select Negative (siNeg) control) were not considered. This analysis identified 17 candidates with multiple siRNAs scoring ≥2 z-scores. Ingenuity Pathway Analysis (IPA) of these top candidates revealed an enrichment for MAPK pathway regulators and protein-protein interactions (Fig. 1D).

Figure 1.

Kinome-wide, loss-of-function screening identifies network of genes sensitizing MEK-resistant pancreas cancer cells. A. Scheme of loss-of-function screen in selumetinib-treated YAPC pancreas cancer cells. Cell numbers were determined after siRNA knockdown and treatment with vehicle or selumetinib. B. Activity of individual siRNAs in vehicle-only (x-axis) and AZD6244-treated (y-axis) YAPC cells. Dark line indicates siRNAs with >2 Median absolute deviation (MAD) Z scores in AZD6244- vs vehicle-treated cells indicating increased loss of viability by selumetinib administration. C. Distribution of siRNA activity by AZD6244- versus vehicle only-treated Z score ratios. siRNA with >2 MADs lower Z scores in AZD6244-treated screen indicated are highlighted. Right, 17 candidate genes with ≥2 active siRNAs (robust Z score > 2). D. Top candidates are enriched in known protein-protein interactions of MAPK signaling as determined by IPA. The top scoring IPA network (score = 29) is shown, top gene candidates highlighted in red.

To select hits for secondary validation, genes without known involvement in tumor cell growth were omitted. Mitogen-activated protein kinase 3 (MAPK3; alias ERK1) and p38 alpha (MAPK14) as already known drug targets in pancreas cancer were also not considered (27, 28). Ten hits from the initial RNAi screen were selected for secondary validation studies. Hits evaluated in ten-concentration dose response testing included previously known regulators of resistance to MAPK pathway inhibition like mitogen-activated protein kinase kinase kinase 8 (MAP3K8; c-COT kinase), glycogen synthase kinase-3 beta (GSK3B), ribosomal protein S6 kinase alpha-5 (RPS6KA5), or phosphatidylinositol 4,5-bisphosphate 3-kinase subunit alpha (PIK3CA) (Fig. 2A,B; Suppl. Fig. 2A). Comparing dose responses to selumetinib in YAPC cells transfected with scramble siRNA versus target gene siRNA, silencing of genes encoding ribosomal protein S6 kinase A5 (RPS6KA5), serine/threonine-protein kinase WNK2 (WNK Lysine Deficient Protein Kinase 2), scaffolding protein connector enhancer of kinase suppressor of Ras 1 (CNKSR1), membrane palmitoylated protein 6 (MPP6), and integrin-linked protein kinase (ILK) induced the most pronounced sensitizing effect to MEK inhibition (Fig. 2B). Next, to exclude genes which in the absence of MEK inhibition have a lethal effect on cancer cells per se and might have not been excluded after the RNAi screen, dose response testing was repeated and analyzed as a percentage of the number of cells present at the treatment start (lethal dose response format; reduction of cell growth in the absence of selumetinib to RNAi control indicative of essential lethality of target gene). Gene targets which sensitized cells to MEK inhibition without inducing cytotoxicity in untreated cells included, in addition to CNKSR1 and RPS6KA5, MAPK pathway modulators ILK, MPP6, or serine/threonine-protein kinase WNK2 (Fig. 2C; Suppl. Fig. 2B). Next, we validated the loss-of-function results of response to MEK inhibition of the top 10 hits in the second pancreas cancer cell line PK-59. Loss of WNK2, CNKSR1, RPS6KA6 and RPS6KA5 in KRAS^G12D-mutant PK-59 pancreas cancer cells showed the greatest improvement on response to selumetinib (Fig. 2D; Suppl. Fig. 3). As ribosomal protein S6 kinase signaling was recently identified as an essential Ras effector pathway in pancreas cancer and serine/threonine-protein kinase WNK2 had been reported as a tumor suppressor and negative regulator of MAPK signaling in pancreas cancer, these hits were not selected for further follow-up study (12, 28, 29).

Figure 2.

Secondary validation of candidate genes in dose response testing of YAPC and PK-59 cells. A. Percent sensitization of YAPC cells treated with 1 and 10μM selumetinib for 72 hours after knockdown of indicated candidate genes. Bars indicate percent growth of cells transfected with indicated siRNAs compared to YAPC cells transfected with scramble control siRNA which was set at 100 percent. Comparisons of decrease in cell viabilities to scramble siRNA control indicated by * using student’s t-test. B. Dose response of YAPC cells transfected with scramble siRNA (blue curves) and CNKSR1 and RPS6KA5 siRNA (red curves) incubated for 72 hrs with selumetinib. Y-axis depicts percent growth compared to untreated vehicle control in a four-parameter fit of depicted concentrations, in triplicate. Right, table lists IC₅₀ values of YAPC cells treated with selumetinib after transfection and knockdown with siRNAs against indicated targets (in μM). C. YAPC cells incubated with increasing concentrations of selumetinib for a period of 72 h after transfection with scramble (blue lines) and CNSKR1 and RPS6KA5 siRNAs (red lines). Effect on viability was recorded and plotted as a percentage of the number of cells present at the treatment start. D. Percent growth of PK-59 cells treated with 1 or 10μM selumetinib after transfection with indicated candidate gene siRNA, scramble control was set at 100 percent.

CNKSR1 expression is associated with activation of AKT signaling

Our group has previously shown that CNKSR1 expression status is associated with clinical outcome and prognosis in pancreas cancer (25). Thus, we decided to focus on the role of CNKSR1 as a mediator of MEK resistance and examined expression levels of CNKSR1 in a panel of twenty-two pancreas cancer cells lines next. In line with CNKSR1 heterogenous expression levels in clinical pancreas cancer specimens, CNKSR1 expression varied significantly across human pancreatic cancer cell lines (Fig. 3A). CNKSR1 has previously been reported in different cell-based contexts to regulate other intracellular signal transduction processes including NF-κB, c-JUN, RhoGTPase signaling, or AKT signaling (24, 30–33). To identify leads for intracellular signal transduction cues associated with CNKSR1 function in pancreas cancer, we compared the transcriptome of pancreas cancer cell lines with high CNKSR1 expression levels to cell lines with no or minimal CNKSR1 expression next (Fig. 3B). The top differentially expressed genes (DEGs) between the two groups included the erb receptor 3 (ERBB3; HER3), PI3K regulating tyrosine-protein phosphatase non-receptor type 13 (PTPN13), or subunits of the phosphoinositide 3-kinase (PI3K) family (phosphatidylinositol-4-phosphate 3-kinase C2 domain-containing beta polypeptide, PIK3C2B) with some gene transcripts measuring ≥10-fold higher in cell lines with high CNKSR1 expression levels. The top-ranking network of the top 28 DEGs after IPA identified HER2-HER3 signaling as the most enriched network in CNKSR1^high cells (Fig. 3C). To examine whether the observed gene expression changes in cells with high CNKSR1 expression like Erb receptor tyrosine kinase and PI3K signaling translates into increased activation of AKT we used the median of CNKSR1 expression levels normalized to the housekeeping protein actin to classify cells into CNKSR1^high and CNKSR1^low and determined p-AKT levels by immunoblotting which showed increased phosphorylation levels in the CNKSR1^high compared to CNKSR1^low group (p<0.05; Fig. 3D). To assess whether the association of elevated CNKSR1 expression and AKT activation is also present in clinical samples of pancreatic cancer tissues we performed immunohistochemical staining of the Surveillance, Epidemiology, and End Results (SEER) Program (SEER) pancreas cancer tissue microarray (TMA) (25). Using the previously described semiquantitative scoring algorithm to classify tissues as CNKSR1^high (moderate and strong staining; scored as 2+ and 3+) and CNKSR1^low (no and weak staining; 0 and 1+) (23), both cytoplasmatic levels of phospho-AKT, and to a lesser degree phospho-mTOR, correlated with CNKSR1 expression status affirming observations in the cell line panel (Fig. 3E; Suppl. Fig. 3B). In summary, in line with CNKSR1’s scaffolding function, these findings suggest CNKSR1 as a possible regulator of AKT signaling in pancreas cancer.

Figure 3.

CNKSR1 expression correlates with AKT activation in pancreas cancer. A. CNKSR1 expression is heterogenous across pancreas cancer cell lines. Immunoblot of total lysates of indicated pancreas cancer cell lines probed with anti-CNKSR1 (top) and actin (bottom) antibodies. B. CNKSR1^high cell lines harbor select transcriptomic alterations compared to CNKSR1^low cells. Unsupervised hierarchical cluster analysis and heat map of baseline transcriptomic profiles of CNKSR1^low (MIAPaCa-2, SK-PC-3, MDA-Panc-28) and CNKSR1^high (YAPC, PK-59, KLM-1) cells. Rows represent individual probes, columns represent individual cell lines, top 30 differentially expressed genes (DEGs) are shown. Color of each probe reflects log2 ratio of normalized expression values for each cell line compared to the median from all cell lines. C. Top-ranking IPA network of DEGs. D. Immunoblot of cell lysates of CNKSR1^high vs CNKSR1^low cells determined by median of CNKSR1 signal normalized to actin control probed with indicated antibodies. Right, quantification of p-AKT and total AKT normalized to actin activity and p-AKT to total AKT ratios of immunoblots by ImageJ. N=2 independent experiments were performed. E. CNKSR1^low pancreas cancers show less AKT phosphorylation at S473 and less downstream mTOR activation. Representative images of SEER Pancreas Cancer TMA sections stained with H&E, anti-CNKSR1, anti-p-AKT^S473 and anti-p-mTOR. Representative images of CNKSR1^low tumor after histochemical staining scored as +1 (weak staining) shown in top row, representative images of CNKSR1^high tumor scored as +3 (strong staining) shown in bottom row. Quantification of graded p-AKT immunohistochemical staining (0-1, 2, 3, and +4) of CNKSR1^low (tumors with no or weak staining; 0 and +1) and CNKSR1^high cases (moderate-to-strong staining; +2 and +3) on right.

CNKSR1 regulates AKT activity in response to MAPK pathway blockade

To investigate how CNKSR1 might regulate the response of pancreas cancer cells to MEK inhibition we first examined the impact of loss of CNKSR1 expression on cell growth. Silencing of CNKSR1 did not affect cell proliferation after three days or increased cleaved caspase 3 levels (Fig. 4A, B; Suppl. Fig. 4A). When cells were incubated with the MEK inhibitor selumetinib, which exerts a cytostatic effect in pancreas cancer cells (Suppl. Fig. 1B), loss of CNKSR1 induced apoptosis at nanomolar concentrations of selumetinib suggesting that CNKSR1 protects MEK-inhibited cells from apoptosis (Fig. 4B). MEK inhibition did not alter CNKSR1 expression levels and enforced expression of CNKSR1 in MEK-sensitive CNKSR1^low PSN-1 cells shifted dose response curves to the right increasing resistance to selumetinib (Suppl. Fig. 4B, C). In view of the association of CNKSR1 expression and AKT activation, we examined whether AKT activation regulated by CNKSR1 is a possible adaptive mechanism of resistance to MEK inhibition in pancreas cancer cells next. Treatment with selumetinib increased phosphorylation of AKT in CNKSR1^high but not CNKSR1^low pancreas cancer cell lines (Fig. 4C; Suppl. Fig. 4D). Co-immunoprecipitation experiments in CNSKR1^high YAPC, PK-59, and PANC-1 cell lines treated for 24 hrs with selumetinib or vehicle using anti-CNKSR1, anti-p-AKT, and anti-total AKT antibodies, showed that CNKSR1 predominantly interacts with the phosphorylated form of AKT as immunoprecipitation with anti-phospho AKT antibodies showed co-immunoprecipitation of CNKSR1 whereas the use of anti-total AKT antibodies did not (Fig. 4D). Of note, silencing of CNKSR1 prior to treatment with selumetinib prevented the increase of AKT phosphorylation, reduced endogenous p-AKT, but not total AKT levels (Fig. 4E; Suppl. Fig. 4E, F). In line with increased AKT phosphorylation induced by MEK inhibition, Forkhead box protein O1 (FoxO1), a downstream substrate of AKT promoting cell death and inducing cell cycle arrest upon nuclear translocation, showed reduced nuclear and cytoplasmatic levels in selumetinib-treated cells (Fig. 4F). Increased phosphorylation of FoxO1 retains the factor in the cytoplasm and phosphorylated FoxO1 is ubiquitinated in the cytoplasm followed by proteasomal degradation and reduced overall FoxO1 levels. Enforced expression of CNKSR1 in CNKSR1^low cell lines increased AKT activation (Fig. 4G). These findings suggest that CNKSR1, in line with its known function as a scaffolding protein, promotes stabilization and activation of the phosphorylated isoform of AKT after MEK inhibition and is intertwined in the crosstalk between the MAPK and AKT pathways.

Figure 4.

CNKSR1 governs AKT activation upon MEK inhibition in pancreas cancer cells. A. Impact of CNKSR1 knockdown on proliferation of pancreas cancer cell. Relative cell growth of indicated cell lines transfected with scramble siRNA and CNKSR1 siRNA. Cell proliferation was measured with CellTiter-Glo^® Assay, cell number post-transfection was normalized to 1, relative changes of CellTiter-Glo^® measurements are shown. Cell numbers at 24, 48, 72 and 96 hours timepoints in triplicates between scramble and CNKSR1 siRNA were compared by student’s t-test. N=3 independent experiments were performed. B. CNKSR1 protects selumetinib-treated pancreas cancer cells from apoptosis. Relative Caspase-Glo^® 3/7 Assay System measurement of YAPC and PK-59 cells transfected with scramble siRNA and CNKSR1 siRNA and treated with vehicle and selumetinib at 0nM, 10nM, 30nM, 100nM, 300nM and 1μM. Caspase-Glo^® 3/7 Assay System measurements of DMSO-treated cells were normalized to 1, values indicate relative change. Representative graph of two independent experiments, in triplicates. C. MEK inhibition increases phospho-AKT levels in CNKSR1^high cell lines. Immunoblots of YAPC (CNKSR1^high) and MIACaPa-2 (CNKSR1^low) cell lysates treated with 10μM selumetinib and probed with p-AKT, p-PRAS40 and p-ERK antibodies. D. CNKSR1 interacts with the phosphorylated form of AKT. Immunoprecipitation of CNKSR1, p-AKT, and t-AKT from YAPC and PANC-1 cell lysates treated with vehicle and selumetinib, immunoblots with indicated antibodies after immunoprecipitation are shown. Input control (lysate probed with β-actin pre-immunoprecipitation) shown on bottom, representative image of N=3 independent experiments. E. Immunoblots of CNKSR1^high YAPC cells transfected with scramble and CNKSR1 siRNA and treated with 10μM selumetinib, quantification of p-AKT and t-AKT levels, after normalizing to β-actin control and setting the value in lysate control to 1, of N=3 independent plots by ImageJ shown on right. F. Representative immunofluorescence images by confocal microscopy after staining with anti-FoxO1 antibodies and DAPI in indicated CNKSR1^high pancreas cancer cell lines treated with vehicle (DMSO; left) and selumetinib (right). G. Overexpression of CNKSR1 increases p-AKT levels in CNKSR1^low cells. Confocal microscopy images after immunofluorescent staining of CNKSR1^low cell lines transfected with empty vector and CNKSR1 (CNKSR1 O.E.) with anti-CNKSR1 (green) and anti-p-AKT473 (red). Quantification of CNKSR1 and p-AKT fluorescence after normalizing of total fluorescence of individual markers in vehicle-treated cells to 1 is shown on right. Top, immunoblots of lysates of CNKSR1^low MIAPaca-2 and PSN-1 cells transfected with empty vector and CNKSR1 (CNKSR1 O.E.) and probed with indicated antibodies.

MEK inhibition increases membranous localization of CNKSR1

The scaffolding function of CNKSR1 is intimately linked to CNKSR1 localizing to the inner surface of the plasma membrane (22, 23). CNKSR1 senses via its pleckstrin homology domain and unique post-translational modifications phosphatidylinositol lipids of the inner plasma membrane and, upon engagement with its respective effectors, stimulates Ras-Raf signaling or, in the case of insulin-like growth factor 1 receptor tyrosine kinase signaling, ADP-ribosylation factor (Arf) GTPase and PI3K/AKT signaling (23, 34). To examine whether blockade of MEK drives intracellular redistribution and enhances CNKSR1 p-AKT co-localization at the plasma membrane, we conducted a series of immunocytochemistry studies next. Treatment with selumetinib increased recruitment and colocalization of CNKSR1 and p-AKT at the plasma membrane in CNKSR1^high cell lines (Fig. 5A; Suppl. 5A). We confirmed these findings in a proximal ligation assay (PLA) which detects proximity of target proteins using antibody probes from different species in situ. Selumetinib treatment increased co-localization of CNKSR1 and p-AKT in CNKSR1^high but not CNKSR1^low cells (Fig. 5B; Suppl. Fig. 5B). The PLA signal was dependent on the presence of both primary antibodies (Suppl. Fig. 5C). Silencing of CNKSR1 in the CNKSR1^high cell lines abrogated the PLA signal (Suppl. Fig. 5D). As both CNKSR1 and AKT are anchored to the plasma membrane via their pleckstrin homology domains which recognize 3-phosphorylated phosphoinositides, we examined phosphatidylinositol (3,4,5)-trisphosphate (PIP3) levels, which are generated at the inner surface of the plasma membrane, as a possible mechanism for the translocation of CNKSR1 to the membrane after MEK treatment next. Using anti-PtdIns(3,4,5)P3 antibodies for the detection and quantification of PIP3, selumetinib increased PIP3 levels in CNKSR1^high YAPC cells and increased the colocalization signal with CNKSR1 compared to vehicle control (Suppl. Fig. 5E). Next, we explored whether increased CNSKR1 p-AKT colocalization is associated with a decrease of CNKSR1 MAPK pathway scaffolding function. CNKSR1 is known to regulate both AKT and MAPK output and cooperates with KSR in the activation of RAF and MAPK pathway output (33–35). Recent crystallographic studies have resolved structures of KSR1-MEK complexes incubated with different MEK inhibitors showing unique alterations in KSR conformation states and disruption of the scaffolding complex (16). Using above PLA in the three CNKSR1^high pancreas cancer cell lines YAPC, PK-59, and PANC-1, MEK inhibition with selumetinib disrupted MEK and KSR, and MEK and CNKSR1 proximity signals, and to a lesser degree CNKSR1 and KSR (Fig. 6A). MEK inhibition did not alter proximity signals of the scaffolding proteins CNKSR1, KSR, and MEK in CNKSR1^low MIAPaca-2 and PSN-1 cells (Suppl. Fig. 6A). PLA signals generated through proximity of scaffolding proteins and MEK depended on the presence of both primary antibodies (Suppl. Fig. 6B). These findings suggest that MEK inhibition-induced disruption of scaffolding complexes might redirect CNKSR1 to the plasma membrane, increase CNKSR1’s scaffolding function for phospho-AKT, and increase AKT signaling in CNKSR1^high pancreas cancer cell lines (Fig. 6B).

Figure 5.

MEK inhibition induces membranous recruitment and co-localization of CNKSR1 and phospho-AKT. A. Increased membranous CNKSR1 p-AKT co-staining in CNKSR1^high cell lines after treatment with selumetinib. Representative confocal microscopy images after immunofluorescent staining of indicated cell lines co-stained with fluorogenic membrane dye staining plasma membrane (yellow), CNKSR1 (green), and p-AKT (red) after 24 hrs treatment with vehicle or 10μM selumetinib. Linear distance analysis by confocal microscopy shown on right, emission peaks across indicated distances (in nanometer; white line) shows redistribution of green (CNKSR1) and red (p-AKT) signals towards peaks of membrane dye (yellow). Blue, DAPI. Y-axes, intensities of green and blue channels on left, red and yellow channels on right. B. Proximity ligation assay for protein interaction between CNKSR1 and p-AKT in indicated CNKSR1^high pancreas cancer cells treated with vehicle vs selumetinib. Representative images shown on the left, a red spot represents a single interaction. Quantification after normalizing the number of detected red spots to the total number of quantified cells per group. Experiments were repeated two times.

Figure 6.

MEK inhibition disrupts MEK and KSR, and MEK and CNKSR1 interactions in CNKSR1^high cell lines. A. Proximity ligation assay for indicated protein interactions in CNKSR1^high YAPC, PK-59 and PANC-1 cells treated with vehicle vs selumetinib. Representative images shown on left, and quantification of PLA signal shown on right using ImageJ. Y-axis indicates number of red fluorescent objects. B. Schematic summarizing CNKSR1-mediated MAPK pathway AKT crosstalk following MEK inhibition with selumetinib.

AKT and MEK inhibition cooperate in CNKSR1^high pancreas cancer lines

To examine whether CNKSR1 expression status can guide combination therapy with AKT pathway blockade in MEK-inhibited pancreatic tumors, we first evaluated whether loss of CNKSR1 sensitizes additional pancreas cancer lines to MEK inhibition. Treatment with selumetinib after silencing of CNKSR1 in an additional three CNKSR1^high and CNKSR1^low cell lines improved response to MEK inhibition in CNKSR1^high but not CNKSR1^low cell lines suggesting that the protective effect of CNKSR1 to MEK inhibition is not restricted to the previously tested YAPC and PK-59 lines (Suppl. Fig. 6C, D). Next, we determined synergistic drug responses with selumetinib in the CNKSR1^high YAPC and PANC-1 cell lines within a matrix screen next. Small molecule inhibitors in preclinical or early clinical development targeting intracellular signal transduction pathways previously shown to be regulated by CNKSR1 were tested across multiple concentrations in combination with selumetinib (AZD6244) and synergy was determined via ΔΔbliss coefficients as previously described by our group (Suppl. Fig. 7A) (36). In line with CNKSR1-mediated activation of AKT signaling induced by MEK blockade, the top three synergistic combinations included the AKT inhibitors MK-2206 and perifosin as well as the dual mTORC1/2 inhibitor AZD8055 (Fig. 7A, B; Suppl. Table 1). Ten concentration dose response testing confirmed cooperativity of selumetinib in combination with AKT or mTOR inhibition in CNKSR1^high cells administered at dose levels not lethal to respective cell lines (Fig. 7C; Suppl. Fig. 7B). There was no cooperativity of AKT and MEK inhibition in CNKSR1^low cells (Fig. 7C). In line with these findings, pancreatic tumors generated after orthotopic implantation of CNKSR1^high high cells showed improved overall survival and reduction of tumor weights in the combination group compared to animals treated with single agent MEK and AKT inhibition compared to tumors generated from CNKSR1^low cells which did not show an additive effect (Fig. 7D). Quantitative flow cytometry analysis using anti-phospho antibodies of tumor digests obtained from animals randomized to the four treatment arms showed increased cell fractions positive for phospho-AKT in MEK-inhibitor treated animals and elevated cleaved caspase 3 levels in the combination arm compared to single agent AKT and MEK inhibition in the CNKSR1^high model. (Fig. 7E). Number of cancer cells as a fraction of total alive cells was decreased in the MEK – AKT inhibitor combination group in CNKSR1^high but not CNKSR1^low tumors compared to vehicle control. In contrast, there was no change in cleaved caspase 3 levels in CNKSR1^low tumors. Cleaved caspase 3 flow cytometry findings were validated by immunohistochemical staining of tumoral sections which showed increased levels of nuclear cleaved caspase 3 activity in the CNKSR1^high model treated with the AKT and MEK inhibitor combination (Fig. 7E). In summary, CNKSR1-mediated AKT activation following MEK inhibition appears to be an adaptive, pro-survival mechanism of resistance to MEK inhibition and CNKSR1^high pancreatic tumors might selectively respond to AKT MEK combination targeted therapy.

Figure 7.

CNKSR1 expression status governs drug response of MEK combination therapy. A. Synergistic drug-drug interactions in pancreas tumor cell lines YAPC and PANC-1 for combinations of selumetinib against a set of 26 compounds. Each drug pair was tested in a 10 × 10 matrix block, reflecting nine doses and DMSO control for each individual matrix block. Correlation plot of synergy as measured by Sum Negative ΔBliss metric, calculated as the sum of negative (synergistic) delta bliss values for a 10 × 10 matrix block, for each combination of selumetinib with all the 26 compounds in YAPC cells (x-axis) and PANC-1 cells (y-axis). B. %Response (top) and ΔBliss (bottom) heat maps for selumetinib combined with MK-2206, an AKT inhibitor, and selumetinib with AZD8055, an mTORC1/2 inhibitor, for YAPC and PANC-1. C. Dose response curves of CNKSR1^high YAPC, PK-59 and PANC-1 and CNKSR1^low MiaPaCa-2 and PSN-1 cells treated with selumetinib in combination with vehicle (black), 50nM mTOR inhibitor rapamycin (orange), or 2μM AKT inhibitor MK-2206 (red). Representative curve of two independent experiment is shown, in triplicate. D Kaplan-Meier estimates of overall survival of orthotopic YAPC (left) and PSN-1 (right) tumors treated with vehicle (black), MK-2206 (blue), selumetinib (orange) and the combination (red). Tumoral weights (in grams) of YAPC (right) and PSN-1 (left) xenografts, representative images of YAPC tumors for indicated treatment groups shown on left. E. Flow cytometry analysis of tumor digests of animals randomized to indicated treatments. Y-axis indicates fraction of cells positive for indicated markers of individual cell populations. Measured cell fractions in individual animals randomized to the four groups are plotted, datapoints indicate individual animals. Statistical comparison by one way ANOVA with post-hoc Tukey test. Immunohistochemical stains with anti-cleaved caspase 3 (CC3) antibodies depict YAPC tumors treated with vehicle vs MK-2206 and selumetinib.

Discussion

Targeted therapy has yet to make a larger impact on the outcome of PDAC patients. With the exception of therapies targeting rare molecular alterations single agent molecular therapy has seen only incremental improvements in outcome as the rapid activation of alternate signaling pathways dwarfs therapeutic efficacy (37). To overcome these limitations, the field has started to move towards combination targeted therapy strategies which, to-date, have met considerable challenges (10, 11, 37). For example, unexpected clinical toxicities, narrow therapeutic windows, or uncertain pharmacokinetic-pharmacodynamic relationships with uncertain engagement of the intended target(s) have all contributed to negative clinical findings (11, 38). Additionally, an incomplete understanding of the exact mechanism of resistance has also led to the inclusion of patients unlikely to optimally respond to such combination regimens (39, 40). Thus, enriching eligible patients by molecularly defined subgroups or via rationale-derived markers for targeted combination therapies might be a promising avenue to improve responses.

MEK inhibition has shown limited efficacy as single agent therapy in PDAC and other Kras-mutated tumors (7, 41). The rapid activation of escape pro-survival signaling networks like receptor tyrosine kinase (RTK) signaling via release of negative feedback inhibition, the suppression of mTORC1 signaling which relieves negative feedback on IGF-IR/IRS-1 which leads to PI3K/AKT signaling activation, or via redundant mitogen-activated protein kinase kinase kinase 8 (MAP3K8; or cancer osaka thyroid (COT) kinase) and p90 ribosomal S6 kinase (RSK) signaling are well investigated mechanisms of resistance to MEK inhibition (15, 18, 28, 42, 43). In line with these findings, COT and RSK kinases have been observed as top hits as mediators of resistance to MEK inhibition in our primary kinome RNAi screen. However, more recent probes into MEK resistance outside the immediate MAPK network and MAPK signal transduction revealed an increasingly diverse class of regulators like fibroblast growth factor receptor 1 (FGFR1), SH2 containing protein tyrosine phosphatase-2 (SHP2), or SETD5 complexes regulating chromatin reprogramming as novel, previously not considered as mechanisms of adaptive resistance to MEK inhibition in Kras-mutant tumors including PDAC (44–46).

In this study, we employed kinome-wide RNAi screening in MEK-resistant pancreas cancer cells and identified the scaffold protein CNKSR1 as a major mediator of resistance to MEK inhibition. MEK inhibition induced membranous localization of CNKSR1, interaction and activation of AKT signaling in CNKSR1^high cell lines which is, in some aspects, in line with prior knowledge on this scaffolding protein (22, 34). CNKSR1 is a multi-domain scaffolding protein containing a sterile alpha motif (SAM) domain which is essential for AKT-dependent phosphorylation of CNKSR1 triggering CNK1 oligomerization, a conserved region in CNK (CRIC) domain, and a PSD-95/DLG-1/ZO-1 (PDZ) domain, as well as a pleckstrin homology (PH) domain which anchors the protein via binding to inositol metabolites to cell membranes (22, 47). CNKSR1 has no known catalytic function and is not a catalytically deficient pseudokinase. With regard to CNKSR1 scaffolding mechanism of action in mammalian cells, Lim and colleagues showed that CNKSR1 interacts at the plasma membrane with cytohesins functioning as GTP exchange factors to promote insulin receptor tyrosine kinase (RTK) signaling and PI3K/AKT signaling activation. While Lim and coworkers did not report a direct interaction of CNKSR1 and AKT, Fritz and colleagues showed in 293T and breast cancer cells via a series of elegant co-immunoprecipitation studies that CNKSR1 interacts with the phosphorylated form of AKT, a finding also made in pancreas cancer cells in our study, and that CNKSR1 regulates AKT phosphorylation and downstream activity of the effector FoxO1 (32). Enforced expression of CNKSR1 increased cell proliferation in the study by Fitz and colleagues, an effect which was dependent on functional PI3K-AKT signaling (32). In addition, using an optogenetic approach, Fisher and coworkers showed that CNKSR1 can either stimulate AKT or MAPK signaling and that epidermal growth factor (EGF) concentrations can guide AKT activation and reciprocally reduce RAF signaling (33). These findings align with our observations showing that MEK inhibition-induced disruption of KSR/CNKSR1-MEK scaffolding complexes enhances membranous recruitment of CNKSR1, interaction with the phosphorylated form of AKT, and increases AKT activity which does not occur in CNKSR1^low cell lines. To link these findings to the mechanism of the MEK inhibitor selumetinib, we speculated that MEK inhibition disassembles scaffolding complexes of MEK which involve CNKSR1. Non-engaged CNKSR1 might then become available for scaffolding of other effectors at the plasma membrane like stabilization of the phospho form of AKT explaining the induced membranous co-localization of CNKSR1 and p-AKT, CNKSR1 and p-AKT interaction, and AKT activation upon MEK inhibition. Indeed, recent crystallographic studies of MEK bound to the scaffold KSR incubated with different MEK inhibitors showed that different MEK inhibitors induce different conformational changes with different impacts on the MEK-KSR1 interface (17). While MEK inhibitors including selumetinib other than trametinib were not found to engage KSR1 and 2 directly, selumetinib effectively competed with KSR1 for MEK inducing a large shift in the activation loop conformation of the MEK enzyme. While these co-crystallographic in vitro studies, unlike our proximal ligation assay findings, did not show a disruption of the MEK-KSR1 interface by selumetinib, it is conceivable that in a cell-based context with additional regulators and post-translational modifications selumetinib might induce conformational changes which ultimately lead to dissociation of MEK from its scaffolds KSR and CNKSR1. The concomitant loss of the proximity signal of KSR and CNKSR1 observed upon selumetinib treatment seems to be in line with the disruption of both MEK-KSR and MEK-CNKSR1 scaffolds by MEK inhibition and, possibly, findings in Drosophila where both scaffolds showed a unique pattern of cooperativity in the regulation of MAPK pathway output (16, 17, 48).

To explore whether the finding that CNKSR1 connects AKT and MAPK pathway signaling can be exploited as a potential biomarker for combination anti-AKT-MEK therapy in CNKSR1^high pancreas cancer cells, we tested the combination of selumetinib and small molecule inhibitors targeting intracellular networks previously shown the be regulated by CNKSR1 including NF-κB, IKK, JUN, p38, COT kinase, or ERK inhibitors within a matrix drug screen. The combination with AKT and mTOR inhibitors showed the strongest synergy with MEK blockade in CNKSR1^high cell lines supporting AKT activation as one of the predominant adaptive resistance mechanisms to MEK inhibition. These findings align with the recent characterization of inhibiting CNKSR1 directly. Recognizing CNKSR1 as a drug target, Indarte and colleagues recently presented a first-in-class small molecule lead, PHT-7.3, occupying CNKSR1’s PH domain which prevented plasma membrane recruitment of CNKSR1 and selectively inhibited growth of Kras-mutant cancer cell lines (23). In vivo, similar to combined AKT-MEK inhibition in the CNKSR1^high model, PHT-7.3 cooperated with MEK inhibition in a Kras-mutant lung cancer xenotransplantation model previously shown to be sensitive to combined MEK-AKT inhibition. While PHT-7.3 is an exciting, novel approach for improved targeted therapy, the high daily dose of 200mg/kg and possible off-target effects on other PH domain-containing regulators involved in physiologically important processes might currently still favor dual anti-MEK anti-AKT targeting using clinically tested small molecule inhibitors compared to the PHT-7.3 lead. CNKSR1 expression status might be used to enrich for patients likely to be sensitive to dual AKT-MEK inhibition until PHT-7.3, or future clinical candidates derived from analogues, are further matured and have been moved into clinical testing.

Our study is not without limitations. For example, we did not investigate the mechanism of increased PIP3 levels following MEK inhibition. CNKSR1 senses PIP2 with greater affinity than PIP3, the PI3K product recognized by PDK1 and AKT (23). Thus, while the immunofluorescence studies suggest that CNKSR1 recognizes PIP3 lipid second messengers and co-localizes with AKT at the plasma membrane to aid its activation, we only can speculate about the mechanism for the observed increased PIP3 levels. Interestingly, prior studies have identified receptor tyrosine kinase (RTK)-mediated PIP2 generation via phosphatidylinositol-4-phosphate 5-kinases (PIP5K) activation. Since RTK activation via release of negative feedback inhibition is a well-studied mechanism of MEK inhibition resistance, CNKSR1 might sense elevated PIP2 and PIP3 levels and redistribution of the scaffold might thus be directly linked to MEK inhibition-induced RTK activation (22). We can also not rule out that CNKSR1 through its scaffolding function on RAF and MEK, at least, in part hinders the action of the allosteric inhibitor selumetinib. For allosteric inhibitors to exert their full potential in Ras-mutant cancer cells, disruption of RAF-MEK and in particular disruption of RAF dimers has been shown to sensitize cells to allosteric MEK inhibitors (49, 50). Thus, it is conceivable that CNKSR1’s scaffolding function engaging both RAF and MEK promotes a MEK inhibitor-insensitive state which is either independent of, or additive to, the pro-survival, resistance signaling of activated AKT. We also did not stratify whether cases with more prominent membranous CNKSR1 staining had a stronger correlation with AKT phosphorylation. Fritz and coworkers previously showed in a large series of normal breast, ductal carcinoma in situ (DCIS), and mammary ductal carcinoma specimens that the fraction of membranous staining increased from normal breast tissues to DCIS and invasive carcinoma but did not comment on a possible correlation of membranous localization of CNKSR1 and AKT activation.

In summary, the scaffold protein CNKSR1 mediates adaptive resistance to MEK inhibition in pancreas cancer cells via its scaffolding function of phosphorylated AKT. Co-blockade of AKT improves drug response to anti-MEK therapy in CNKSR1^high cells possibly affording the opportunity to identify MAPK-AKT pathway crosstalk and escape from MEK inhibition via CNKSR1 expression status in future dual AKT-MEK therapy trials. Prospective combinatorial AKT-MEK inhibition in patients selected by high tumoral CNKSR1 expression might yield improved outcomes compared to the administration of the inhibitor combination to unselected patients.

Implications:

The CNKSR1 scaffold, identified within a RNAi screen as a novel mediator of resistance to MEK inhibition in pancreas cancer, connects MAPK pathway and AKT signaling and may be adopted as a biomarker to select patients for combined MEK AKT blockade.