The Transcription Factor SLUG Uncouples Pancreatic Cancer Progression from the RAF–MEK1/2–ERK1/2 Pathway

Activating mutations of KRAS drive the vast majority of pancreatic adenocarcinomas and ∼25% of all cancers. Intracellular signaling downstream of Ras branches into several pathways. One of them, the RAF-MEK1/2-ERK1/2 pathway, culminates in the onset of a transcriptional program that includes the upregulation of cyclin D, and subsequent progression through the cell cycle (1). RAF-MEK1/2-ERK1/2 signaling is considered particularly relevant for cancer progression. Consistently, BRAF is mutated in a substantial percentage (8%) of all tumors, and BRAF and RAS mutations tend to be mutually exclusive (2).

Despite intense past and ongoing efforts, activated RAS has remained an undruggable target in the clinic (3), shifting focus to the inhibition of the RAF, MEK1/2, and ERK1/2 kinases. BRAF inhibitors have shown clinical efficacy against tumors with activating BRAF mutations, particularly cutaneous melanoma (4–6). Mek1/2 inhibition is also efficacious, albeit to a lesser extent, against the same tumors (7), but shows little or no effect on tumors driven by mutant KRAS, such as pancreatic cancer (8). This failure prompted the search for mechanisms of resistance to MEK1/2 inhibition of cancer cells bearing KRAS mutations (9, 10).

Virtually all the mechanisms described so far resulted in the reactivation of ERK1/2 activity in the presence of MEK1/2 inhibitors (11–17). These included genetic lesions activating components of the pathway, such as activating mutations of MEK1 and MEK2 or amplifications of BRAF, as well as the relief of negative feed-back loops. These negative feed-back loops prevent excessive activation of the pathways in homeostasis: activated ERK1/2 limits its own activity by phosphorylating and thereby attenuating its upstream activators, receptor tyrosine kinases (RTK), BRAF and MEK1/2. Pharmacologic inhibition of MEK1/2, and subsequent inhibition of ERK1/2, releases these negative feed-back loops, restoring ERK1/2 activity to levels that allow progression through the cell cycle. Because mutations within the pathway do not seem to explain resistance of mutant KRAS-driven tumors to MEK1/2 inhibition in the clinic in most cases, it is presently not clear whether the lack of efficacy is due to the relief of the negative feed-back loops or to the lack of addiction to activated MEK1/2–ERK1/2 signaling.

To shed light on these issues, we generated and characterized a cell-based model of resistance. We showed that chronic treatment with MEK1/2 inhibitors leads to the uncoupling of cell proliferation from the RAF-MEK/12-ERK1/2 pathway. The mechanism used by resistant cells offers new therapeutic opportunities to reinstate sensitivity to MEK1/2 inhibition and, therefore, drug combination strategies with durable antitumor activity.

The pancreatic cancer cell lines MIA PaCa-2, HPAF-II, and BxPC-3 were purchased from ATCC. MIA PaCa-2 and HPAF-II were grown at 37°C in presence of 5% CO₂ and in DMEM/F-12 (Invitrogen) supplemented with 10% FBS, 2.5% horse serum, 2 mmol/L L-glutamine and 1× antibiotic–antimycotic (Gibco). BxPC-3 cells were grown at 37°C in presence of 5% CO₂ and in RPMI1640 (Invitrogen) supplemented with 10% FBS, 2 mmol/L L-glutamine and 1× antibiotic–antimycotic. Cells were routinely tested every month for the absence of Mycoplasma contamination using the MycoAlert Mycoplasma Detection Kit (Lonza). Cell lines were not authenticated in-house and were not passaged more than 30 times.

The list of inhibitors used in this study is in the Supplementary Materials and Methods section. For in vivo administration, dTAG-13 was resuspended in a 5% DMSO and 20% solutol saline solution and administered by intraperitoneal injection. For assessing in vitro degradation, cells were incubated 24 hours whereas for assessing degradation in vivo, tumors were collected 6 hours after the last treatment.

In the first approach (R1), cells were treated with 1.2 μmol/L MEK162, which corresponded to ∼10 × IC₅₀ of MEK162 on MIA PaCa-2 cells. After 2 days, cell culture was replenished with drug-free medium until cells were confluent. In the second strategy (R2), MIA PaCa-2 cells were treated with increasing concentrations of MEK162, starting from 0.25 μmol/L (∼2 × IC₅₀) and increasing over time, until reaching 1.2 μmol/L. In this strategy, MEK162 was maintained in the medium until confluence. Resistant cells from P-PDX #57 were obtained following the second strategy, with increasing concentrations of MEK162 (started with 45 nmol/L and finished with 5800 nmol/L). As control, we cultured MIA PaCa-2 and P-PDX #57 parental cells in the presence of vehicle alone during the same period and under the same conditions.

To generate resistance, MIA PaCa-2 cells were treated after attachment with increasing concentrations of AMG-510, starting from 5.5 nmol/L and increasing over time, until reaching 60 nmol/L. AMG-510 was maintained in the medium until confluence.

Cell proliferation was determined by crystal violet staining.

Protein extracts obtained from whole cell lysates were fractionated in SDS-PAGE gels, transferred onto Immobilion-P (Millipore) membranes, and subjected to immunoblot analysis (antibody list in Supplementary Materials and Methods).

IHC was performed as described previously (18). Primary antibodies list in Supplementary Materials and Methods.

The relative levels of phosphorylation of 43 kinase phosphorylation sites and two related total proteins were assessed by using a commercial human phosphokinase antibody array (R&D Systems), according to the manufacturer's protocol.

Exome sequencing was performed and analyzed as described previously (18). Details in Supplementary Materials and Methods.

RNA was extracted using the RNeasy Mini Kit (Qiagen) according to the manufacturer's recommendations. TruSeq Stranded Total RNA Kit protocol (Illumina Inc.) was used to prepare the RNA-seq libraries. The size and quality of the libraries were assessed with a High Sensitivity DNA Bioanalyzer assay (Agilent Technologies). Libraries were sequenced in a HiSeq2000 instrument, with a read length of 2 × 100 bp. The RNA-seq data in this study have been deposited in sequence read archive (SRA) database and are accessible through the SRA Bioproject (PRJNA719991).

Pathway enrichment was assessed through the preranked version of gene set enrichment analysis (GSEA), and we used gene sets derived from the HALLMARK database (19). Only statistical gene sets are represented in Fig. 2C (NOM P < 0.05).

Real-time quantification of transcript abundance was determined by qRT-PCR using Taqman Gene Expression probes (list in Supplementary Materials and Methods) and TaqMan Universal Master Mix II, with UNG (Thermo Fisher Scientific) following the manufacturer's protocol.

Cells were labeled with 5 μmol/L of CellTracker Green reagent (Invitrogen), following the manufacturer's protocol, and kept overnight in serum-free medium. Twenty-four hours later, 5 × 10⁴ cells were seeded in triplicates in 24-well plates previously coated with 300 μg/mL Corning Matrigel matrix (Corning) in 1× DMEM F-12 for 2 hours. For migration or invasion, cells were seeded in triplicates in 10 μg/mL fibronectin (Sigma-Aldrich) (migration) or 300 μg/mL Matrigel-coated (invasion) Fluoroblok BD Biocoat Cell Culture Inserts in serum-free medium, whereas the wells were loaded with complete growth medium. After incubation, cells were washed with 1× PBS and fixed with 4% formalin. Cells adhered or migrated to the basolateral side was visualized and counted using the Fiji software.

To assess the ability to form three-dimensional (3D) spheroids, cells were mixed in growth-reduced factor Matrigel (BD Bioscience, #354234) in a concentration of 1,000 cells per 25 μL of Matrigel and cultured in 3D culture media. Details in Supplementary Materials and Methods.

Animal work was performed according to protocols approved by the Ethical Committee for the Use of Experimental Animals at the Vall d'Hebron Institute of Oncology. MIA PaCa-2, R1–16, and R2–39 luciferase cells were suspended in Matrigel to 80,000 cells/μL. Ten microliters of cell suspension was injected into the tail of the pancreas of 5-week-old female BALB/c nude mice. Mice were killed after 40 to 60 days, and all major organs were resected in order to determine the existence of distant metastasis by means of luciferase signal with the IVIS-200 imaging system from Xenogen (PerkinElmer). Metastasis was considered positive if luciferase signal was greater than the background.

All PDXs have been established at VHIO following institutional guidelines. The Institutional Review Boards at Vall d'Hebron Hospital provided approval for this study in accordance with the Declaration of Helsinki. Written informed consent was obtained from all patients who provided tissue samples. To obtain primary cell cultures from pancreatic cancer PDXs, tumor pieces of approximately 750 to 1,000 mm³ were minced with scalpels and digested in digestion solution: 300 U/mL collagenase IA (Sigma-Aldrich) in 10 mL pancreatic medium for 1 hour at 37°C. The digested tissue was centrifuged at 300 × g for 5 minutes, and cultured in pancreatic medium. All PC-PDXs were previously sequenced by amplicon sequencing.

For virus production, HEK293T cells were transfected as described previously (18). Infected cells were selected with 1 μg/mL puromycin, starting 2 days after infection, and subsequently maintained with 0.5 μg/mL puromycin in the growth media. Details of plasmids in Supplementary Materials and Methods.

A total of 0.8×10⁶ MIA PaCa-2 and MEK162-resistant cells were resuspended in 50% Matrigel:PBS and injected subcutaneously into BALB/c nude mice. When tumors reached 100 to 200 mm³, mice were treated with MEK162 daily. In the case of metastasis assessment from subcutaneous injection, luciferase-expressing cells were injected, and tumors were resected when they were above 1,000 mm³. Lung metastases were measured after 94 days.

For in vitro and in vivo experiments, comparisons between two groups were made by two-tailed Student t test.

To generate cellular models of resistance to MEK1/2 inhibition, we treated cultures of a KRAS mutant pancreatic cancer cell line (MIA PaCa-2) with the MEK1/2 inhibitor MEK162 (also known as Binimetinib, ARRY-162, or ARRY-438162) during 6 months. Treatment consisted of constant or increasing concentrations and, as a result, we generated several independent clones that we named R1 and R2 according to the dosage scheme (Fig. 1A). Compared with parental cells, both types of clones had high proliferating capabilities in micromolar concentrations of the MEK1/2 inhibitor. We observed some variability between experiments; for parental MIA PaCa-2 cells, IC₅₀ varied between 0.08 and 0.25 μmol/L, for resistant clones it varied between ∼5 and > 7.3 μmol/L (Fig. 1B and successive figures). Resistant cells were also insensitive to trametinib (also known as Mekinist or GSK1120212; Supplementary Fig. S1A), a more potent inhibitor that prevents feedback reactivation because it also inhibits the phosphorylation of MEK1/2 by Raf kinases (15). Culturing resistant clones during three months without MEK162 did not modify their resistance (Fig. 1C; Supplementary Fig. S1B), showing that the mechanism of resistance does not require continuous selective pressure to be maintained.

To determine the relevance of the differences in IC₅₀ observed in vitro, we analyzed the sensitivity of resistant cells in vivo. Although the growth of subcutaneously engrafted MIA PaCa-2 cells was prevented by treating mice with MEK162, the inhibitor had little or no effect on the in vivo growth of a randomly chosen resistant clone (Fig. 1D).

The majority of mechanisms of acquired resistance described to date reinstate the activation or ERK1/2 in the presence of the MEK1/2 inhibitor (10). However, we found that pERK1/2 levels in our resistant cells were very similar to those of parental cells and that MEK162 reduced the phosphorylation of ERK1/2 to the same extent in all cases (Fig. 1E). These results suggest that resistant cells are no longer dependent on the ERK1/2 signaling. Confirming this hypothesis, resistant cells showed a significantly reduced sensitivity to the ERK1/2 inhibitor SCH772984 (Fig. 1F, left; Supplementary Fig. S1C).

Cells can acquire resistance by activating different RTKs that activate compensatory pathways (14, 17). Because the PI3K-AKT pathway has been previously shown to induce resistance to MEK1/2 inhibition in different contexts (14, 20, 21), we analyzed the levels of phospho-AKT and its downstream target, phospho-EBP1, and found that they were similar in parental and resistant cells (Supplementary Fig. S1D). In addition, analysis of lysates of one randomly chosen resistant clone using a phosphokinase antibody array, which included AKT phosphorylated at different sites (S473 and T308) showed no evidence of activation of any compensatory pathway (Supplementary Fig. S1E).

Because resistant cells seem to be independent from two major signaling pathways branching from RAS, we tested the sensitivity of resistant cells to AMG-510, a compound that inhibits KRAS with the G12C mutation, precisely the driving mutation in MIA PaCa-2 cells. The IC₅₀s of resistant cells were significantly higher than that of parental cells (Fig. 1F, right; Supplementary Fig. S1F), indicating the independency of the cells selected for resistance to MEK1/2 inhibition of RAS-driven pathways.

To search for previously described genetic lesions shown to confer resistance to MEK1/2 inhibition, we compared the exome sequences of the different resistant clones with that of MIA PaCa-2 cells. Although we did not find common mutations in all resistant clones, we found 41 and 9 common mutated genes in R1 and R2 clones, respectively (Supplementary Fig. S2A). The algorithm PolyPhen-2 (22) predicted relevant functional effects of 17 and 1 of the common mutations found in R1 and R2 clones, respectively (Supplementary Fig. S2B). However, the corresponding genes did not have any obvious functional relationship to the KRAS-RAF-MEK1/2-ERK1/2 or alternative cell proliferation pathways. Therefore, we concluded that the resistance to MEK1/2 inhibition is not likely caused by gene mutation.

Collectively, these results indicated that the cells obtained constitute a model for a previously unidentified mechanism of resistance.

Transcriptomic analysis by RNA-seq showed that 239 and 218 genes were acutely up- or downregulated in R1 and R2 resistant cells compared wit parental MIA PaCa-2 cells (≥4-fold; q < 0.0001), respectively. The expression of 68 of these genes was altered both in R1 and R2 (Fig. 2A and B), indicating that a common mechanism could explain their resistances. GSEA identified numerous biological processes that differed between parental and resistant cells (Fig. 2C).

The enrichment of the epithelial-to-mesenchymal transition (EMT) signature (Fig. 2C and D) was consistent with the upregulation of SNAI2, the gene encoding SLUG (Fig. 2B), one of the transcription factors that mediates the EMT, a process that increases cell motility and is associated with resistance to chemotherapy (23). Analysis of the expression of SLUG (mRNA and protein) confirmed its upregulation in all resistant cells (Fig. 2E; Supplementary Fig. S3A). MIA PaCa-2 have several mesenchymal characteristics (24). Consistently, we were not able to detect E-CADHERIN (a classical epithelial marker) in our cultures, and we detected several mesenchymal markers such as LAMININ B, αSMA, COLLAGEN 1, VIMENTIN, or FIBRONECTIN. Although we did not find differences in LAMININ B, αSMA, or COLLAGEN 1, consistent with a more mesenchymal phenotype, resistant cells expressed higher levels of VIMENTIN and FIBRONECTIN (Supplementary Fig. S3A). Thus, prolonged inhibition of the MEK1/2 pathway results in resistance to MEK1/2 inhibitors and, concomitantly, to the apparent acquisition of a more mesenchymal phenotype. This result was not a particularity of MIA PaCa-2 cells. Treatment of cultures from a pancreatic cancer patient derived xenograft (P-PDX #57) with MEK162 during 3 months gave rise to resistant cells that also overexpressed Slug (Supplementary Fig. S3B).

Different transcription factors induce similar EMT phenotypes. However, although resistant cells expressed slightly different levels of five well-known EMT transcription factors, none was consistently overexpressed in resistant cells (Supplementary Figs. S3C and S3D). Thus, the enrichment in the EMT signature seems to be due solely to the expression of SLUG.

Short-term treatment of MIA PaCa-2 cells with MEK162 resulted in a rapid increase of SNAI2 transcript levels (Fig. 2F, top), albeit to levels much lower than those expressed by resistant cells (∼10 fold in parental cells treated during 48 hours, compared with ∼100 fold in resistant cells). Importantly, MEK1/2 inhibition did not affect the expression of other EMT transcription factors (Supplementary Fig. S3E). The upregulation of SNAI2 expression in parental cells was reversible; upon removal of the inhibitor, the levels of SNAI2 transcript decreased. However, repeated treatments with the MEK162 inhibitor led to a progressive increase in the expression of SNAI2 mRNA, even after removing the inhibitor (Fig. 2F, bottom). In view of this result, we hypothesized that long-term inhibition of MEK1/2 results in progressive upregulation of SLUG to the levels displayed by resistant cells.

Different MAP kinase pathways cross-talk with the MEK1/2–ERK1/2 pathway (10). However, the phosphoprotein array shown in Supplementary Fig. S1C showed no evidence of activation of the p38 alpha or JNK1/2/3 pathways. Two independent lines of evidence pointed to a role of MEK5/ERK5 (25), which is not included in the phosphoprotein array, in resistance. Inhibition of MEK1/2–ERK1/2 results in the activation of ERK5 (also known as MAPK7 or BMK1; ref. 26). On the other hand, the MEK5–ERK5 pathway can upregulate the expression of SLUG (27). Confirming the potential role of MEK5/ERK5, compared with parental cells, all the resistant clones expressed constitutively high levels of pERK5 (Fig. 2G). Analysis of parental MIA PaCa-2 cells confirmed that treatment with the MEK1/2 inhibitor induced the activation of the MEK5-ERK5 pathway (Fig. 2H). Inhibitors of MEK5 or ERK5 prevented the upregulation of SLUG expression induced by MEK162 (Fig. 2I). Furthermore, treatment of resistant cells with inhibitors of MEK5 or ERK5 reduced the levels of SLUG (Fig. 2J). We concluded that the activation of ERK5 induced by the inhibition of MEK1/2 causes a stable increase in SLUG expression.

By expressing SLUG, cells are expected to acquire metastatic potential (23). To determine if this was the case of resistant cells, we compared their behavior to that of parental cells in several assays. Cell adhesion was assessed with Matrigel-coated plates and migration and invasion with Transwells coated with fibronectin and Matrigel, respectively. Resistant cells had increased ability to adhere, migrate, and to invade (Fig. 3A–C). MIA Paca-2 cells form spheres when cultured in 3D (see also Fig. 3D; ref. 28). Although resistant cells also grew as spheres, in line with the previous assays, they tended to be more migratory (Fig. 3D).

To confirm these results in vivo, we orthotopically implanted parental and resistant cells expressing luciferase into pancreases of immunodeficient mice. After 8 weeks, we detected metastatic growth in liver, diaphragm, spleen, lungs, and lymph nodes (Fig. 3E). Quantitative analyses showed that resistant cells metastasized more efficiently than parental cells (Fig. 3F).

Thus, concomitantly with the acquisition of resistance to MEK1/2 inhibition, cells increased expression of the transcription factor SLUG, acquired mesenchymal traits, and became more metastatic.

Overexpression of SLUG tagged at the C-terminus with the Flag epitope resulted in an IC₅₀ for the MEK1/2 inhibitor comparable with that of resistant cells (Fig. 4A). Some transcription factors have overlapping functions with SLUG (23). To analyze whether these transcription factors can also cause resistance to MEK1/2 inhibition, we overexpressed the most similar to SLUG, that is, SNAIL. We found that SNAIL had no effect on the sensitivity to MEK162 (Supplementary Fig. S3F). Thus, resistance is specifically induced by SLUG.

To silence the expression of SLUG we tested five independent shRNAs. Two of them effectively reduced the levels of SLUG in resistant cells (Supplementary Fig. S3G). The most efficient shRNA (shSlug 239) restored the sensitivity to MEK162 (Fig. 4B and C). The less efficient shRNA also increased sensitivity to MEK162, albeit to a lesser extent (Supplementary Fig. S3H). Thus, gain and loss of function analyses showed that SLUG reversibly promoted the transition to a condition in which cell proliferation is independent of the MEK1/2 pathway.

Cellular proliferation requires accumulation of D-type cyclins, which allow progression through the cell cycle (29). The MEK1/2 pathway is a major regulator of the expression of D-cyclins (30). Consistently, treatment of parental cells with the mek1/2 inhibitor led to a marked decrease in the levels of CYCLIN D1 and arrest in the G₁ phase of the cell cycle (Fig. 4D and E). In contrast, in resistant cells, MEK1/2 inhibition had little or no effect on the levels of CYCLIN D1 and the distribution of cells in the different phases of the cell cycle (Fig. 4D and E). Thus, acquisition of resistance led to the uncoupling of CYCLIN D1 regulation from the MEK1/2 pathway. Gain- and loss-of-function analyses showed that SLUG is sufficient and required to promote such uncoupling (Fig. 4F).

Finally, we showed that downregulation of SLUG from resistant cells reinstated sensitivity to MEK1/2 inhibition in vivo (Fig. 4G). Thus, gain- and loss-of-function analyses show that SLUG promotes resistance to MEK1/2 inhibition.

Because R1 and R2 resistant cells are also resistant to inhibition of KRAS (Fig. 1F, right; Supplementary Fig. S1F), we reasoned that Slug can also mediate the resistance to the KRAS^G12C inhibitor AMG-510. To test this hypothesis in an unbiassed manner, we cultivated MIA PaCa-2 cells in increasing concentrations of AMG-510 (Supplementary Fig. S4A). The resulting cells (AMG-R) showed a marked resistance to AMG-510 (Supplementary Fig. S4B), and overexpressed SLUG to levels similar to those found in cells resistant to MEK1/2 inhibition (Supplementary Fig. S4C). SLUG knockdown confirmed that SLUG is responsible for the resistance to KRAS^G12C inhibition (Supplementary Figs. S4D and S4E).

Thus, characterization of resistant cells showed that overexpression of Slug results in cells that do not depend on the KRAS–MEK1/2–ERK1/2 pathway to proliferate.

Compared with parental cells, MIA PaCa-2 cells overexpressing SLUG displayed higher adhesive, migratory, and invasive abilities in vitro (Fig. 5A). Conversely, the knockdown of SLUG from resistant cells resulted in a decrease of the same abilities (Fig. 5B and C).

In vivo analyses were consistent with the in vitro assays. Overexpression of SLUG in MIA PaCa-2 cells led to increased metastases from pancreas to diaphragm and spleen (Fig. 5D). Analysis of the fate of subcutaneously engrafted cells confirmed that SLUG overexpression causes increased metastasis growth (Supplementary Figs. S5A and S5B). Conversely, we observed a decrease in the metastatic growth of resistant cells engineered to underexpress SLUG (Fig. 5E and F). These results showed that the acquisition of resistance to MEK1/2 inhibition through SLUG overexpression resulted in a more aggressive tumor growth, which was also dependent on SLUG expression.

To further determine the influence of SLUG in the aggressiveness of pancreatic cancer, we queried the TGCA https://www.cancer.gov/tcga) and the International Cancer Genome Consortium (ICGC; ref. 31) databases. Consistent with the aggressiveness conferred by SLUG in our models, high SNAI2 transcript levels correlate with worse pancreatic cancer patient outcome in the two data sets (Fig. 5G). In contrast, tumors with high levels of expression of similar transcription factors did not have different outcomes, compared with tumors with low levels (Fig. 5G; Supplementary Fig. S5C). Further, high SNAI2 expression conferred similar bad prognosis, whether patients with tumors with high SNAI1 expression were included or not. Conversely, SNAI1 was not a prognostic factor, whether patients with tumors expressing high levels of SNAI2 were included or not (Supplementary Fig. S5D). Thus, SNAI2, but not SNAI1 or other EMT transcription factors, is critical for determining the prognosis of pancreatic cancers.

We have previously shown that sensitivity to MEK1/2 inhibition varies widely among pancreatic cancer cell lines (18). To determine if SLUG could be related to these differences, we measured its levels in a panel of cell lines, as well as in cultures from P-PDXs previously established in our laboratory (18). With the exception of one cell line (HPAF-II), which express moderate levels of SLUG and has the highest IC₅₀ value, there was an inverse correlation between the levels of SLUG and the sensitivity to MEK1/2 inhibition (Fig. 6A). It also should be noted that BxPC-3 cells have been previously shown resistant to trametinib (32). The different sensitivities to both MEK1/2 inhibitors, MEK162 and trametinib, should be further investigated. We also found that P-PDXs previously shown to be primarily resistant to MEK1/2 inhibitors in vivo (see also Supplementary Fig. S6A; ref. 18), expressed higher levels of SLUG, compared with sensitive P-PDXs (Supplementary Fig. S6B). Only one of the sensitive PDXs, expressed moderate levels of SLUG (P-PDX-10; Supplementary Figs. S6A and S6B).

To functionally characterize the role of SLUG in some of these cells, we overexpressed SLUG in the low expressing P-PDX #57. As a result, we observed an increase in the IC₅₀ for MEK162 of ∼10-fold (Fig. 6B). When we attempted to downregulate SLUG in cells with high endogenous levels, we observed a marked loss of viability (see, e.g., Supplementary Fig. S6C), showing that SLUG is required for the survival of these cells and precluding the analysis of sensitivity to MEK1/2 inhibition in knock-down cells. We did achieve viable cells upon downregulation of SLUG from P-PDX #61 and showed that it resulted in acquisition of sensitivity to MEK162 (∼100-fold decrease in IC₅₀; Fig. 6C). Collectively, correlations and gain and loss of function experiments support that SLUG regulates the sensitivity of a variety of pancreatic cancer cells to MEK1/2 inhibition.

All the cell lines expressing detectable levels of SLUG, except BxPC-3, displayed detectable levels of pERK5 (Fig. 6D), indicating that the MEK5–ERK5 pathway upregulates SLUG in different pancreatic cancer cell lines. In BxPC-3 Slug overexpression is likely due to a mechanism independent of the MEK5–ERK5 pathway. Inhibition of ERK5 in cells displaying higher levels of pERK5 (HPAF-II and P-PDX #27) resulted in the marked downregulation of SLUG (Fig. 6E), showing that in these cell lines the overexpression of SLUG is mediated by MEK5–ERK5 signaling. Thus, SLUG confers primary resistance to MEK1/2 inhibition in different pancreatic cancer cells and, in some of them, the expression of SLUG is mediated by the MEK5–ERK5 pathway.

The results presented show that the knockdown of Slug restores the dependency of pancreatic cancer cells on the MEK1/2 pathway (Fig. 4B, C, G, and 6C). Although SLUG remains an undruggable protein, as shown in ref. 27 and Fig. 2J, its expression can be downmodulated through inhibition of the MEK5–ERK5 pathway. On the other hand, Proteolysis-targeting chimeras (PROTAC) have recently emerged as a means of targeting previously undruggable factors. Thus, we hypothesized that MEK5–ERK5 inhibition or targeted degradation of SLUG, would restore the sensitivity to inhibitors of the MEK1/2–ERK1/2 pathway.

Inhibitors of MEK5 or ERK5 had little or no effect on the sensitivity of MIA PaCa-2 cells to MEK162 (Supplementary Fig. S7A). In contrast, and consistent with our hypothesis, MEK5 or ERK5 inhibitors resensitized resistant cells to MEK1/2 inhibition (Fig. 7A).

To determine if the resensitization to MEK162 induced by ERK5 inhibition is due to the downregulation of SLUG, we infected resistant cells with a retroviral vector encoding SLUG. Note that the slower migration of exogenous SLUG is due to a Flag epitope located at the C-terminus (Fig. 7B). Treatment with the ERK5 inhibitor led to a reduction of the expression of endogenous SLUG to levels comparable to those of parental MIA PaCa-2 cells, but did not affect exogenous SLUG, which is under the control of a constitutively active promoter (Fig. 7B). Importantly, ectopic expression of SLUG prevented the reduction of the IC₅₀ of MEK162 induced by ERK5 inhibition (Fig. 7C), confirming that SLUG is responsible for resistance.

Next, we determined the effect of ERK5 inhibition on additional pancreatic cancer cell lines. We chose P-PDX #27 and HPAF-II cells because they are resistant to MEK1/2 inhibition and express high levels of SLUG in an ERK5-dependent manner (Fig. 6A and D). As a control, we used BxPC-3 cells because they are also resistant to MEK1/2 inhibition and express high levels of SLUG, but in an ERK5-independent manner (Fig. 6D). Although ERK5 inhibition had little effect on the sensitivity of BxPC-3 cells to MEK162, the same treatment sensitized HPAF-II cells and cells from P-PDX #27 to MEK1/2 inhibition, resulting in an IC₅₀ similar to that of parental MIA PaCa-2 cells (Supplementary Fig. S7B).

These results reinforce the central role of SLUG in determining the sensitivity of pancreatic cancer cells to inhibition of the MEK1/2–ERK1/2 pathway and show that ERK5 inhibition, through the regulation of SLUG, sensitizes cells to MEK1/2 inhibitors, unveiling a potentially efficacious drug combination.

PROTACs are hetero-bifunctional compounds that simultaneously bind to a target protein and to an E3 ubiquitin ligase, thereby promoting the degradation of the former by the proteasome (reviewed in ref. 33). Although PROTACs that recruit different E3 ligases, such as cereblon, are available, the development of a chemical moiety to bind a given protein, such as Slug, is a considerable endeavor. To overcome this hurdle, different tag-based methods have been developed. One of these methods, the dTAG system, consists in fusing a FKBP12 mutant protein (FKBP12^F36V) to the protein of interest, and recruiting the fusion protein to cereblon using a compound, dTAG-13, which binds to FKBP12^F36V and cereblon (34), see also (Fig. 7D).

Transduction of SLUG tagged at the C-terminus with FKBP12^F36V resulted in resistance to MEK162 (IC₅₀ = 7.2 μmol/L; Fig. 7E and F, left) comparable with that of R1 and R2 resistant cells (IC₅₀s = 6.5–7.3 μmol/L, Fig. 1B) or cells transfected with SLUG tagged at the C-terminus with Flag (IC₅₀ = 5.28, Fig. 4A). In contrast, transduction SLUG tagged at the N-terminus with FKBP12^F36V resulted in a smaller increase in IC₅₀ (1.86 μmol/L; Supplementary Figs. S7C and S7D), arguing that modifying the N-terminus of SLUG impairs protein functionality and thus impacts its ability to confer resistance to MEK1/2 inhibition. Further experiments were therefore performed using SLUG tagged at the C-terminus.

Consistent with our previous observations (Fig. 1F, right; Supplementary Fig. S4), cells expressing SLUG-FKBP12^F36V resulted in resistance to the KRAS^G12C inhibitor AMG-510 (Fig. 7F, right).

Treatment with dTAG-13 induced the degradation of the Slug-FKBP12^F36V fusion protein in a dose-dependent manner, with near maximal degradation evident at 0.5 μmol/L in vitro (Fig. 7G). Treatment of MIA PaCa-2 cells expressing SLUG-FKBP12^F36V with 0.5 μmol/L dTAG-13 restored sensitivity to MEK1/2 inhibition (Fig. 7F, left), showing that degradation of SLUG resensitized cells to Mek1/2 inhibition. Treatment of parental MIA PaCa-2 cells with a high concentration of dTAG-13 (2.5 μmol/L) had no effect on the sensitivity to MEK162 (Supplementary Fig. S7E), arguing against the existence of off-target effects. Treatment with dTAG-13 also resensitized cells to AMG-510 (Fig. 7F, right), confirming that degradation of SLUG restores dependency of KRAS.

We next aimed to confirm degradation of SLUG-FKBP12^F36V in vivo. A short-term treatment of MIA PaCa-2 cells expressing SLUG-FKBP12^F36V xenograft tumors led to marked loss of fusion protein expression (Fig. 7H), confirming effective degradation in vivo with dTAG-13. We next evaluated tumor burden in response to MEK162, dTAG-13, or MEK162 and dTAG-13. As expected, treatment with MEK162 did not prevent the growth of tumors generated by MIA PaCa-2 cells expressing SLUG-FKBP12^F36V (Supplementary Fig. S7F) and treatment with dTAG-13 alone did not alter tumor growth either (Fig. 7I). However, dTAG-13 fully restored sensitivity to MEK162 (Fig. 7I). Confirming our in vitro observations, MEK162 was equally efficient in decreasing the levels of phosphorylated ERK1/2 independently of the presence or absence of SLUG, and effectively impaired cell proliferation in its absence (Supplementary Fig. S7G). Furthermore, the proliferation marker Ki67 was decreased only in tumors from mice treated with the combination (Fig. 7J) and the number of cells expressing active CASPASE-3 was increased by the same treatment (Supplementary Fig. S7G). These results suggest that therapies targeting SLUG will be efficacious to treat pancreatic cancers that become resistant to inhibitors or the KRAS–RAF–MEK1/2–ERK1/2 pathway.

The intense effort to pharmacologically target the RAF–MEK1/2–ERK1/2 pathway has resulted in dozens of inhibitors, but very limited successes, restricted mainly to cutaneous melanomas. Even in these successful cases, tumor cells invariably develop resistance and resume their progression (10).

The resistance to MEK1/2 inhibition of KRAS-driven tumors, singularly pancreatic cancers, has been explained by a variety of mechanisms that either reactivate the ERK1/2 signaling or activate compensatory pathways, such as the PI3K-AKT pathway (35–38). To avoid reactivation of the ERK1/2 pathway, a variety of inhibitors that prevent the effect of released feed-back loops, as well as combinations of BRAF, MEK1/2, and ERK1/2 inhibitors have been used. To counteract the activation of compensatory pathways, additional combinations have been designed (such as MEK1/2 inhibitors plus PI3K inhibitors). The downside of these combinatorial approaches is that the wide variety of reactivation and compensatory mechanisms points to an ever-increasing complexity of the combinations, leading to serious toxicities. By showing that cells resistant to MEK1/2 inhibitor may rely on SLUG to proliferate, our results unveil an additional vulnerability that may help to design simpler and more effective combinations.

Because SLUG cannot be directly inhibited yet, we aimed to unpair its expression by targeting the MEK5–ERK5 pathway, which had been previously shown to be activated by MEK1/2 inhibition (26) and upregulates SLUG expression (27). We found that, indeed, MEK5–ERK5 sustains the overexpression of SLUG in cells with acquired resistance to MEK1/2 inhibition. Further, we showed that in some pancreatic cancer cells, constitutive high levels of SLUG are maintained by MEK5-ERK5 signaling. Importantly, in these cells the combined inhibition of MEK1/2 and ERK5 impaired cell proliferation because the inhibition of ERK5 downmodulated SLUG and, as a consequence, cells regained sensitivity to MEK1/2 inhibition.

Remarkably, the same combination, MEK1/2 plus ERK5 inhibition, has been recently shown effective against pancreatic cancer cells because it promotes the degradation of MYC (26). Two signatures of MYC targets are downregulated in resistant cells (Fig. 2C), indicating that MYC downregulation may affect the proliferation of resistant cells. Because MYC is a regulator of apoptosis (39), future work should determine if the low levels of MYC in resistant cells result in reduced levels of apoptosis in resistant cells.

Importantly, the combined inhibition of MEK1/2 and ERK5 is only effective in cells in which the overexpression of SLUG is due to ERK5 activation. In cells where SLUG is overexpressed by other means, ERK5 inhibition does not affect sensitivity to MEK1/2 inhibitors. Thus, our results show that the effect of ERK5 inhibition is mediated by SLUG. Several factors have been shown to regulate the expression of SLUG, including KLF4 (40), SIM2 (41), FOXA2 (42), TCF21 (43), ESR1 (44), and miRNA-630 (45). Predictably, inhibition of some of these factors will downregulate SLUG in certain cellular backgrounds and it will resensitize them to MEK1/2 inhibition.

The development of degradation-based approaches has opened additional therapeutic strategies, by allowing the specific degradation of any intracellular protein. Because the expression of SLUG is negligible in several cell lines sensitive to MEK1/2 inhibition, and expression of a SLUG-FKBP12^F36V fusion protein confers resistance, we could test the feasibility of targeting SLUG with a PROTAC. As a result, we showed that degradation of SLUG resensitized cells to MEK1/2 inhibition. Thus, the identification of SLUG as a crucial regulator of a subset of pancreatic warrants the development of direct-acting PROTACs or molecular glue compounds targeting SLUG.

Analysis of public databases shows that ∼60% of pancreatic cancers express high levels of SLUG and that these patients have worse prognosis. It has been recently shown that the inhibition of KRAS using CRISPR/Cas-mediated genome editing in pancreatic cancer cells led to the appearance of tumor cells with enriched EMT signatures that predict poor prognosis in patients (46). This result is in agreement with ours and shows that more than half of pancreatic carcinomas are more aggressive and not dependent on the RAF–MEK1/2–ERK1/2 pathway. The identification of SLUG as a crucial regulator of these subsets of pancreatic cancer open new avenues to more effective treatments.

No disclosures were reported.

F. Bilal: Conceptualization, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. E.J. Arenas: Conceptualization, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. K. Pedersen: Investigation. A. Martínez-Sabadell: Formal analysis, investigation, methodology, writing–review and editing. B. Nabet: Resources, methodology. E. Guruceaga: Formal analysis, investigation, methodology. S. Vicent: Formal analysis, investigation, methodology. J. Tabernero: Resources. T. Maraculla: Resources. J. Arribas: Conceptualization, resources, formal analysis, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.

This work was supported by Instituto de Salud Carlos III Project Reference number AC15/00062 and the EC under the framework of the ERA-NET TRANSCAN-2 initiative co-financed by FEDER, Instituto de Salud Carlos III (CB16/12/00449 and PI19/01181). E.J. Arenas was funded by the Spanish Government (Juan de la Cierva Formación FJCI-2017-34900). A. Martínez-Sabadell was funded by the Spanish Government (PFIS FI20/00188). B. Nabet was supported by an American Cancer Society Postdoctoral Fellowship (PF-17-010-01-CDD) and the Katherine L. and Steven C. Pinard Research Fund.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.