Inverse Correlation of STAT3 and MEK Signaling Mediates Resistance to RAS Pathway Inhibition in Pancreatic Cancer

Pancreatic ductal adenocarcinoma (PDAC) is currently the third leading cause of cancer mortality in the United States and estimated to become the second leading cause of cancer-related death by 2020 (1). Despite advances in understanding the biology of PDAC, the clinical course is still hampered by therapeutic resistance and lack of progress in improving survival outcomes. Activating Kras mutations, present in over 90% of PDAC tumors (2), are difficult to target due to the interdependence of redundant signaling pathways and feedback loops. Another hallmark of PDAC is the presence of a dense desmoplastic reaction composed of stromal cells, including stellate cells, immune cells, and cancer “stem” cells (CSC) that support the tumor microenvironment (TME), acts as barrier to drug delivery and also contributes to therapeutic resistance of PDAC (3, 4). There remains an incomplete understanding of interactions among the different TME components that support cancer growth and promote resistance to therapy.

In PDAC, Kras-activating mutations promote proliferation and inhibit apoptosis through the RAF/MEK/ERK and PIK3/AKT pathways (5–7). Because targeting RAS has remained elusive, efforts have focused on targeting downstream effectors of RAS through MEK inhibition (5, 8). The clinical success of MEK inhibitors in Kras-mutant melanoma and lung cancer implicate their therapeutic potential in PDAC (9, 10). Unfortunately, clinical trials of MEK inhibition (MEKi), have been unsuccessful in PDAC (7, 9, 11), likely due to the activation of resistance pathways.

Shedding of EGFR ligands from PDAC cells results in autocrine feedback and subsequently facilitates Kras activation to promote the onset of pancreatic neoplasia (12, 13). EGFR ligands amphiregulin (AREG) and TGFα reside as transmembrane glycoproteins in their precursor form and are enzymatically shed via metalloproteinase TNFα-converting enzyme (TACE, ADAM17) to regulate EGFR signaling (12, 14–16). We have shown that TACE and AREG, but not TGFα, are overexpressed in colorectal cancer and PDAC (14). AREG appears to be the primary ligand for promoting TACE-mediated activation of EGFR, MAPK, and STAT3 signaling.

The poor response of patients with PDAC to targeted and systemic therapies is also partially due to the dense desmoplastic reaction, which impedes drug delivery into tumors (17–20). We have shown that targeted inhibition of STAT3 (STAT3i) in combination with gemcitabine enhances drug delivery into tumors by remodeling collagen fibers and increasing microvessel density (19, 20). Therapeutic resistance in PDAC is also conferred by increased percent of CSCs in the TME following chemotherapy (21).

In this study, we have provided proof that STAT3 activation is a novel molecular mechanism underlying the therapeutic resistance of PDAC to RAS pathway inhibition. We show a direct inverse correlation between ERK and STAT3 signaling in PDAC and provide evidence that a major resistance mechanism impairing MEKi occurs through TACE-induced, AREG-mediated EGFR pathway activation. We further show combined MEKi/STAT3i results in sustained inhibition of both ERK and STAT3 signaling, overcoming therapeutic resistance associated with the AREG-EGFR–mediated parallel feedback. In addition, we show MEKi/STAT3i inhibits tumor fibrosis and downregulates CSCs to enhance therapeutic response in PDAC. Furthermore, elevated serum AREG levels are decreased with MEKi/STAT3i in PKT mice, suggesting a role for serum AREG as a prognostic biomarker of therapeutic response as well as a biomarker of therapeutic resistance to EGFR, MEK, and STAT3 inhibition.

Mouse pancreatic intraepithelial neoplasia (PanIN), primary PDAC (PDA), and liver metastatic (LMP) cell lines were derived from the LSL-Kras^G12D/+; Pdx1^Cre/+ (KC) and LSL-Kras^G12D/+; Trp53^R172H/+; Pdx1^Cre/+ (KPC) genetically engineered mouse models (GEMM) of PDAC (kindly provided by Dr. Andrew Lowy, University of California San Diego, San Diego, CA; refs. 22, 23) and maintained as described previously (20). Authenticated human PDAC cell lines MiaPaCa2, PANC1, CFPAC, Capan2, Capan1, AsPC1, SW1990, Panc04, Panc02, Panc10, and BxPC3 were obtained and maintained according to ATCC guidelines. The PC-13 and PC-17 cell lines were derived from patient-derived xenografts (PDX; kindly provided by Dr. Anthony Capobianco, University of Miami, Coral Gables, FL) with known mutational status (PC-13: Kras^Q61H/+/PIK3CA^M/+/p53^R72P/+; PC-17: APC^M/M/Kras^G12D/+/p53^R72P/R72P).

Cell authentication was performed by using short tandem repeat DNA profiling (latest date: June 16, 2016) and cell lines tested negative for Mycoplasma via Genetica cell line testing using eMYCO plus Kit (iNtRON Biotechnology). Cells with relative low passage numbers (<20) were used in the study.

Primary antibodies used for Western blot analysis (Supplementary Table S1), IHC (Supplementary Table S2), and flow cytometry analysis (Supplementary Table S3) were summarized in the Supplementary Tables S1–S3. The chemical agents used are presented at Supplementary Table S4.

Previously constructed tissue microarray (TMA; ref. 20) slides were concurrently evaluated by 2 of the authors (C. Shi and N.B. Merchant). Nuclear and cytoplasmic staining was scored as follows: staining index was considered as the sum of the intensity score (0, no staining; 1+, weak; 2+, moderate; 3+, strong) and the distribution score (0, no staining; 1+, staining of <33% of cells; 2+, between 33% and 66% of cells; and 3+, staining of >66% of cells). Staining indices were classified as follows: 3+ or higher, strong staining; 1+ to 2+, weak staining; and 0, negative staining. pERK and pSTAT3 were scored as positive if any detectable nuclear or cytoplasmic staining was present.

Female athymic nude mice [Foxn1 nu/nu (4–5 weeks old)] were purchased from Harlan Sprague Dawley, Inc. Ptf1a^cre/+;Tgfbr2^flox/flox and LSL-Kras^G12D/+;Tgfbr2^flox/flox mice were provided by Dr. Hal Moses (Vanderbilt University Medical Center, Nashville, TN). These two lines were intercrossed to generate Ptf1a^cre/+;LSL-Kras^G12D/+;Tgfbr2^flox/flox mice (PKT) on a C57Bl/6 background. Genotyping of alleles was performed using oligonucleotide primers as described previously (20, 24). LSL-Kras^G12D/+, Pdx1^Cre/+and p53^R273H/+mice were intercrossed to generate indicated LSL-Kras^G12D/+; Trp53^R172H/+; Pdx1^Cre/+ (KPC) animals (22, 23).

Subcutaneous tumors were established by injecting 2 × 10⁶ PANC1, MiaPaCa2, or BxPC3 cells into the flank of 6-week-old Fox1-nu/nu mouse [n = 4 (PANC1) or n = 3 (MiaPaCa2, BxPC3) in each group]. Treatment was initiated when the subcutaneous tumors reached 75–100 (MiaPaCa2 or PANC1) or 200–250 (BxPC3) mm³ size. Drug treatment was initiated at the same time point. AZD1480 (30 mg/kg/day), AZD6244 (25 mg/kg/day), both drugs together or vehicle (hydroxypropyl methyl cellulose/Tween 80) was administered by oral gavage for 41 (MiaPaCa2), 49 (PANC1), and 27 (BxPC3) days. The data were obtained as described in the Supplementary Materials and Methods.

PKT mice were treated with vehicle, AZD1480 (JAK/STAT3 inhibitor), AZD6244 (MEK inhibitor), or a combination of AZD6244 and AZD1480. Mice in the AZD1480 (30 mg/kg/day) and AZD6244 (25 mg/kg/day) arm received by oral gavage 5 days/week, starting at 4 weeks of age. Mice were euthanized and dissected after 2 weeks unless they were part of the survival arm. Because of the irregularity of the tumor dimensions, size was determined by weighing the entire tumor. Tumor tissue was processed for further IHC examination. Overall survival was determined by log-rank analysis using statistical software package R (version 3.3.2).

Descriptive statistics were calculated using Microsoft Excel and Prism software (Graphpad Software Inc.). Results are shown as values of mean ± SD unless otherwise indicated. Statistical analyses of IHC data were performed using the ANOVA followed by Tukey multiple comparisons test to determine P values. The two-sided Student t test was used for statistical analysis, with P < 0.05 taken as significant, except when indicated otherwise. Kaplan–Meier survival analysis was performed, and survival differences between groups were assessed with the log-rank test. Pearson correlation coefficient or Pearson r analysis was conducted using Prism software.

All animal experiments were carried out using protocols approved by the Institutional Animal Care and Use Committees at the Vanderbilt University Medical Center (Nashville, TN) and the University of Miami (Miami, FL; #15-099).

For additional experimental procedures please refer to Supplementary Materials and Methods.

A direct inverse correlation in expression levels of phosphorylated ERK (pERK) and STAT3 (pSTAT3) was observed both in vitro and in vivo (Fig. 1). Inverse baseline expression of pERK and pSTAT3 was confirmed in 11 human PDAC cell lines (Fig. 1A, left). Cell lines with low or minimal pERK expression showed high levels of pSTAT3 expression, whereas cell lines with high pERK expression showed little or no pSTAT3 expression. This pattern was also seen in a PanIN-derived cell line from the LSL-Kras^G12D/+; Pdx1^Cre/+ (KC) GEMM, and from PDA and LMP cell lines derived from the LSL-Kras^G12D/+; Trp53^R172H/+; Pdx1^Cre/+ (KPC) GEMM (Fig. 1A, right; ref. 22). Results from a human tumor TMA confirmed that the majority of patient tumors with strong pERK expression had low pSTAT3 expression, whereas tumors with high pSTAT3 expression had low pERK expression (Fig. 1B; Supplementary Fig. S1A). Knockdown of STAT3 in PANC1 cells (sh-STAT3; ref. 20) showed increased expression of pERK when compared with sh-scrambled cell lysate (Fig. 1C). In addition, pancreatic tumors from Ptf1a^cre/+;LSL-Kras^G12D/+; Tgfbr2^flox/flox (PKT) mice (Fig. 1D; Supplementary Fig. S1B) treated with the JAK/STAT3 inhibitor (STAT3i) AZD1480 showed increased pERK levels, whereas mice treated with MEKi AZD6244 showed increased pSTAT3 expression. Both STAT3i and MEKi showed increased pEGFR expression (Fig. 1D).

Treatment with the MEK inhibitors, AZD6244 or MEK162, effectively suppressed pERK while inducing pSTAT3 activation in PDAC cells in a time-dependent manner (Fig. 1E). Conversely, STAT3 inhibitors, Stattic and AZD1480, resulted in pERK activation in a time- (Fig. 1F; Supplementary Fig. S2) and dose-dependent manner (Supplementary Fig. S3A–S3C). Even SW1990 cells, which have very low basal levels of pSTAT3 (Fig. 1A, left), increased pSTAT3 and pEGFR levels with MEKi in a time- (Fig. 1G) and dose-dependent manner (Supplementary Fig. S3D).

To determine the effects of cytotoxic chemotherapy on STAT3 or ERK signaling, we treated MiaPaCa2 cells with gemcitabine with or without AZD1480 (Supplementary Fig. S4A and S4B) or erlotinib (Supplementary Fig. S4C). Gemcitabine treatment alone resulted in a dose-dependent activation of both pSTAT3 and pERK. Cells treated with both gemcitabine and AZD1480 exhibited similar effects on STAT3 and ERK signaling compared with AZD1480 (STAT3i) alone, demonstrating that gemcitabine did not alter the reciprocal activation of STAT3 or ERK signaling. EGFR inhibition with erlotinib attenuated EGFR phosphorylation.

To confirm this reciprocal activation of MEK and STAT3 as a mechanism of resistance, we targeted both STAT3 and MEK in PDAC cell lines. MiaPaCa2 (Fig. 2A) cells treated in combination with U0126, a highly selective inhibitor of MEK1/2 and Stattic, a small molecule that specifically inhibits the SH2 domain of STAT3 or PANC1 and MiaPaCa2 (Fig. 2B) with combined AZD6244/AZD1480 showed sustained inhibition of both ERK and STAT3 activation. Combined MEKi/STAT3i in MiaPaCa2 cells resulted in enhanced apoptosis (Fig. 2C), reduced colony formation (Fig. 2D, left), decreased cell invasion (Fig. 2D, right), and reduced spheroid growth (Fig. 2E). Furthermore, spheroid growth of PDX cells, PC-13 and PC-17 (Fig. 2F), were significantly reduced with combined STAT3i/MEKi when compared with control cells. These results show that combined MEKi/STAT3i suppresses pancreatic tumorigenicity.

RAS signaling pathway is a known mediator of tumor progression and survival that is activated downstream of the EGFR (25). Having shown, both in vitro and in vivo, reciprocal reactivation of pERK or pSTAT3 upon STAT3i or MEKi, respectively (Fig. 1C–G), we hypothesized that STAT3i would also result in EGFR activation, upstream of RAS signaling. As seen in Fig. 3A, STAT3i with AZD1480 resulted in pSTAT3 inhibition, with subsequent activation of pERK and pEGFR in both time- and dose-dependent manner (Fig. 3A; Supplementary Fig. S3C). Interestingly, over time, pSTAT3 becomes reactivated following activation of EGFR even with STAT3i. Again, as pSTAT3 is reactivated, pERK levels decrease, further confirming the inverse correlation of pSTAT3 and pERK signaling. These data suggest STAT3i-mediated ERK activation may be dependent on EGFR signaling. To further explore the mechanism by which STAT3i promotes EGFR tyrosine phosphorylation, we examined the upstream activators of EGFR signaling.

Previously, we and others have shown that TACE is regulated by G protein–coupled receptors (GPCR) that act as mediators of EGFR transactivation (14, 26, 27). TACE, when activated, induces cleavage and shedding of soluble forms of the EGFR ligands, AREG, and TGFα. Importantly, we have shown that TACE and AREG, but not TGFα, are overexpressed in pancreatic tumors when compared with normal human pancreatic tissues and are involved in EGFR/ERK/STAT3 signaling activation (14). First, we confirmed whether EGFR inhibition with erlotinib activates STAT3 phosphorylation (Supplementary Fig. S4C). To determine the mechanism by which STAT3i leads to EGFR activation, we analyzed the effects of STAT3i or MEKi on TACE phosphorylation (Fig. 3B). Interestingly, STAT3i increased TACE phosphorylation in a time-dependent manner. Similarly, MEKi also showed increase in TACE phosphorylation. Combined MEKi/STAT3i attenuated TACE activation (Fig. 3C), suggesting that TACE phosphorylation mediates the reciprocal activation of STAT3 and ERK through EGFR. We then assessed MEKi/STAT3i effects on AREG shedding in conditioned media (CM) from BxPC3, PANC1, and MiaPaCa2 cells (Fig. 3D). STAT3i with AZD1480 and MEKi with AZD6244 resulted in significantly increased AREG levels in the CM from all three cell lines. This observation further confirms AREG's involvement in MEKi and STAT3i-mediated activation of EGFR signaling.

We then sought to determine whether AREG is the ligand that activates EGFR signaling in this process. First, we confirmed whether AREG activates ERK phosphorylation (Supplementary Fig. S5). To further delineate the role of STAT3 in AREG-mediated EGFR–ERK activation, MiaPaCa2 cells were treated with AZD1480 and cetuximab (C225), a mAb that inhibits EGFR ligand binding (Fig. 3E, left) or erlotinib, a tyrosine kinase inhibitor of EGFR (Fig. 3E, right). Combined treatment with AZD1480 and C225 or erlotinib completely inhibited ERK and STAT3 activation along with EGFR signaling. Taken together, these data suggest that EGFR tyrosine kinase activity mediates STAT3-dependent ERK activation.

Because TACE activation correlates with AREG-dependent ERK activation (14), we sought to address the role of AREG release during in vivo PDAC progression. Serum was collected from patients with PDAC and normal individuals (Fig. 3F, left). Serum from patients with PDAC had significantly higher AREG levels when compared with normal patients, suggesting increased TACE activity in patients with PDAC as we have previously shown.(14) Furthermore, we compared AREG serum levels between KPC and wild-type (WT) mice (Fig. 3F, right). Serum AREG levels were significantly higher in KPC mice compared with WT mice, further corroborating the roles of AREG and activated TACE in PDAC progression. Collectively, these data show TACE-AREG–mediated EGFR pathway activation is a major resistance mechanism that impairs the efficacy of individual MEKi or STAT3i. These results suggest that combined EGFR and STAT3 or MEK and STAT3 inhibition may be therapeutically effective in PDAC. Furthermore, AREG may serve as a potential circulating biomarker of resistance to MEK or STAT3 inhibition.

To determine the effects of MEKi/STAT3i on tumor growth in vivo, PANC1 (Fig. 4A) and MiaPaCa2 (Fig. 4B) xenografts were subjected to vehicle, AZD1480, AZD6244, or combined treatment by daily oral gavage. Combined MEKi/STAT3i significantly suppressed tumor growth compared with vehicle or individual agent treatment. We did not observe any significant changes in animal body weight, suggesting there is no added toxicity from the combined treatment (Supplementary Fig. S6A and S6B). IHC staining of xenograft tumors showed decreased Ki67 (Fig. 4C) and increased cleaved caspase-3 (Fig. 4D) staining in the combined treated tumors compared with treatment with either agent alone.

To determine therapeutic efficacy in a GEMM of PDAC, we treated PKT mice with the same treatment regimen (vehicle, AZD1480, AZD6244, or combination) beginning at 4 weeks of age when pancreatic tumors are firmly established in this GEMM (Fig. 4E–G). There was a significant reduction in pancreatic tumor weight with combined AZD1480/AZD6244 treatment compared with vehicle or treatment with either agent alone (Fig. 4E). Furthermore, while pancreata of vehicle or monotherapy-treated mice were completely replaced with tumor burden, mice treated with combined MEKi/STAT3i had little or no tumor burden and maintained their pancreatic integrity (Fig. 4F and G). Again, there were no significant changes in animal body weight, suggesting no added toxicity from the combined treatment (Supplementary Fig. S7A). To further analyze pancreatic integrity, PKT tumors were assessed for cytokeratin 19 (CK-19), alcian blue, collagen 1, and Ki67 staining by IHC (Fig. 4G). Vehicle-treated PKT mice had abundant staining with the proliferation marker Ki67, extracellular matrix protein collagen type 1 and alcian blue, as well as CK-19 staining for adenocarcinoma. Pancreata of PKT mice treated with AZD1480 or AZD6244 alone had a slight reduction in staining of these markers compared with vehicle-treated mice. However, PKT mice treated with combined AZD1480/AZD6244 had significantly more normal pancreatic architecture and histology, demonstrating a staining profile similar to their WT littermates (Fig. 4G). These results show that combined MEKi/STAT3i significantly reduced tumor burden, reduced tumor cell proliferation, and retained pancreatic integrity in this aggressive PDAC GEMM.

In a separate study, PKT mice were analyzed for survival under the same treatment arms. Again, tumor growth was significantly reduced in PKT mice treated with combined MEKi/STAT3i (Fig. 5A). Furthermore, survival of PKT mice treated with MEKi/STAT3i was significantly improved when compared with vehicle-treated and STAT3i-treated mice (53 vs. 85 or 63 vs. 85 days respectively; P = 0.0002, log-rank test). Although there was an improvement in survival of mice treated with combined MEKi/STAT3i compared with MEKi-treated mice (85 vs. 76 days, Fig. 5B), these results were not statistically significant. Western blot analysis of whole tumor lysates showed high pERK and pSTAT3, but low pEGFR expression in vehicle-treated mice (Fig. 5C). Although STAT3i completely inhibited pSTAT3 expression, it increased expression of both pERK and pEGFR. MEKi inhibited pERK expression, but had limited effect on pSTAT3 expression. In contrast, combined MEKi/STAT3i significantly reduced the expression of both pSTAT3 and pERK without activating pEGFR. IHC staining showed a significant decrease in Ki67 staining (Fig. 5D) and enhanced cleaved caspase-3 (Fig. 5E) expression in tumors treated with combined MEKi/STAT3i, compared with tumors treated with vehicle or single agents alone. These results confirm the enhanced in vivo effects of combined MEKi/STAT3i in overcoming therapeutic resistance in an aggressive PDAC GEMM.

Combined MEKi/STAT3i attenuated TACE activation (Fig. 3C; Supplementary Fig. S7B), suggesting that TACE phosphorylation mediates the reciprocal activation of STAT3 and ERK through EGFR. Furthermore, to assess the effect of MEKi/STAT3i on systemic AREG levels, we measured AREG in the serum of treated PKT mice. Serum AREG levels were negligible and significantly lower in PKT mice treated with combined MEKi/STAT3i compared with treatment with single agents alone (Fig. 5F). These results further corroborate the effect of combined MEKi/STAT3i in overcoming AREG-EGFR–mediated resistance to RAS and STAT3 signaling inhibition in PDAC.

CSCs are self-renewing cells in the TME that are resistant to cytotoxic and targeted therapies and are often the cause of tumor recurrence despite initial therapeutic response. STAT3 signaling is active in tumor-associated CSCs (3, 28, 29). To determine whether MEKi/STAT3i affects CSCs in PDAC, PKT mouse tumors treated with MEKi and/or STAT3i were evaluated for CSC cell markers by flow cytometry. Analysis showed significantly decreased EpCAM⁺CD44⁺ (Fig. 6A) and EpCAM⁺CD133⁺ (Fig. 6B) CSCs in tumors treated with combined MEKi/STAT3i when compared with vehicle treatment. These results indicate that the enhanced in vivo effects of combined MEKi/STAT3i further overcome therapeutic resistance by targeting CSCs in PDAC. To further assess the impact on CSCs, BxPC3 xenografts were treated with vehicle or combined MEKi/STAT3i for 27 days and then treatment was stopped and tumor growth was assessed. There was regrowth of tumors from day 6 to day 17 after stopping MEKi/STAT3i treatment (Supplementary Fig. S7C), suggesting that MEKi/STAT3i suppress CSCs, but does not completely eliminate the CSC population.

A major therapeutic challenge in PDAC is its oncogene-driven innate and acquired chemoresistance. KRAS is the key oncogenic driver of tumorigenesis and malignant progression. Due to the molecular characteristics of RAS protein, direct inhibition of KRAS has yet to be achieved. We and others have developed strategies to inhibit RAS indirectly by targeting its downstream effectors in the RAF/MEK/ERK signaling cascade. MEK is a key node in this axis and serves as the primary mediator for KRAS signal transduction. However, MEK inhibition (MEKi) has not achieved clinical efficacy in PDAC due to redundant signaling pathways and molecular crosstalk that induce compensatory survival signaling. Furthermore, we have also previously identified activated STAT3 signaling as a key biomarker of resistance in PDAC (20, 30).

This study helps to define the heterogeneous response and therapeutic resistance of PDAC to RAS and STAT3 pathway inhibition. We have identified a novel molecular mechanism whereby STAT3i leads to reciprocal activation of MEK–ERK signaling, whereas MEKi results in activation of STAT3 signaling. Furthermore, we show that STAT3i results in TACE-AREG-EGFR–dependent activation of ERK. These results provide a mechanistic rationale of acquired resistance to downstream RAS and STAT3 pathway inhibition. This mechanism of ERK reactivation is relevant to other studies that have shown that ERK signaling reactivation plays a critical role in primary and acquired resistance to targeted therapies (8, 9, 31). Preclinical studies have identified distinct mechanisms by which cells acquire resistance to MEKi, including amplification of mutant BRAF (31, 32), PI3K upregulation (8, 33), or EGFR activation (8, 34). Dual inhibition of these pathways has provided benefit in some patients (10, 25, 35).

STAT3 activation has been observed in a variety of human tumors (20, 28) through engagement of cytokine or growth factor receptors. We have previously shown that STAT3i with gemcitabine remodels the tumor stroma and enhances in vivo drug delivery to the tumor, thereby improving survival in PKT mice (20). However, STAT3i results in reciprocal activation of ERK signaling, thereby preventing a sustained therapeutic response. The results of this study support targeting two components of the EGFR-RAS-MEK-JAK/STAT3 pathway to overcome primary and acquired resistance of targeted monotherapies.

Combined MEKi/STAT3i results in sustained inhibition of both ERK and STAT3 signaling. Importantly, the combination of MEKi/STAT3i induces robust apoptosis in vitro and tumor regression in vivo in PDAC tumors. In PDAC cells, EGFR is activated upon STAT3i, perhaps as part of an initial effort by the cell to escape apoptosis or reengage the cell cycle. Furthermore, EGFR activation is suppressed only when STAT3 is inhibited in combination with ERK. With combined MEKi/STAT3i, the cellular escape mechanisms are shut down and growth of PDAC cells is effectively suppressed.

TACE and AREG play a primary role in activating EGFR signaling and are overexpressed during PDAC progression (14). The pro-EGFR ligand, AREG, has been implicated in resistance to other therapies (36, 37) and is required to induce EGFR-mediated ERK signaling (14, 27, 36). We have previously shown that knockdown of AREG prevents STAT3 and ERK activation in response to deoxycholic acid in PDAC cells (14). Previous studies have also suggested inhibiting AREG activity may be necessary to overcome resistance to EGFR-targeted therapies in NSCLC and hepatocellular carcinoma (38, 39), which is consistent with recent reports showing that inhibition of AREG is associated with a better response to cetuximab therapy in colorectal cancer (40, 41).

Our results show AREG levels are increased following STAT3i treatment of PDAC cells, leading to EGFR phosphorylation and subsequent reactivation of ERK and STAT3 signaling. We found that STAT3i results in increased EGFR activity through TACE activation (Fig. 3A–C). Sera from KPC mice and human patients with PDAC have significantly elevated levels of AREG (Fig. 3F), suggesting that serum AREG levels may serve as a noninvasive prognostic biomarker. Furthermore, treatment with MEKi/STAT3i results in a significant decrease in serum AREG levels, suggesting its role as a biomarker of therapeutic response to EGFR-, MEK-, or STAT3-targeted therapies. We now demonstrate that STAT3i in combination with specific blockade of the AREG-mediated EGFR receptor using cetuximab or erlotinib completely inhibits ERK reactivation. Therefore, our study supports the potential of targeting TACE-AREG-EGFR axis as a potent approach to overcome resistance to MEKi or STAT3i.

Combined MEKi/STAT3i results in enhanced therapeutic efficacy not only through sustained inhibition of the RAS pathway–associated redundant feedback loop reactivation, but also by suppressing the TME through reducing CD44⁺ and CD133⁺ CSCs. The interplay between tumor cells and surrounding supportive cells, such as fibroblasts, immune cells, and CSCs, clearly indicates that targeting a single component will not result in sustained inhibition of tumor growth. Targeting multiple signaling pathways and different components of the TME is the means of overcoming either primary or acquired resistance to targeted monotherapy. This approach increases the likelihood of a sustained response by concurrently affecting multiple diverse mechanisms of action associated with cancer development and therapeutic resistance.

We have identified a novel mechanism associated with resistance to RAS–MEK–ERK pathway inhibition through activation of STAT3 signaling and resistance to STAT3 pathway inhibition resulting in TACE-AREG–mediated EGFR and ERK reactivation. We provide strong evidence supporting the role of targeting two components of the AREG–EGFR–MEK–STAT3 pathway to overcome therapeutic resistance associated with RAS pathway reactivation in PDAC. In addition, combined MEKi and STAT3i inhibits CSCs and maintains pancreatic integrity, providing a secondary mechanisms by which this combination therapy may enhance antitumor response and prevent recurrence of disease. In addition, our results suggest that serum AREG levels may serve as a key circulating prognostic biomarker of PDAC and a potential biomarker of therapeutic resistance and response to EGFR, MEK, and STAT3 inhibition.

No potential conflicts of interest were disclosed.

Conception and design: N.S. Nagathihalli, J. Castellanos, M.N. VanSaun, N.B. Merchant

Development of methodology: N.S. Nagathihalli, J. Castellanos, M.N. VanSaun, N.B. Merchant

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N.S. Nagathihalli, J. Castellanos, P. Lamichhane, F. Messaggio, M.N. VanSaun, N.B. Merchant

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N.S. Nagathihalli, J. Castellanos, F. Messaggio, X. Dai, X. Chen, N.B. Merchant

Writing, review, and/or revision of the manuscript: N.S. Nagathihalli, J. Castellanos, P. Lamichhane, C. Shi, X. Dai, P. Rai, M.N. VanSaun, N.B. Merchant

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): N.S. Nagathihalli, P. Lamichhane, X. Dai, M.N. VanSaun, N.B. Merchant

Study supervision: N.S. Nagathihalli, N.B. Merchant

Other (pathology and IHC): C. Shi

The authors thank Mr. Frank Revetta, Ms. Jennifer Barretta, Dr. Alexander Gaidarski III, and Ms. Yanhua Xiong for their technical assistance and Dr. Nilesh Kashikar (pathologist) for evaluating stained tissues. This work was supported by the NIH R01 CA161976, Pancreatic Cancer Action Network (PanCan)-AACR Translational Research Grant (15-65-25-MERC), NIH T32 CA211034, James & Ether King Biomedical Research Program, Florida Department of Health 8JK08, and Sylvester Comprehensive Cancer Center (to N.B. Merchant); NIH NCI R21 CA209536, American Cancer Society IRG 98-277-13, and Stanley Glaser Foundation Research Award (UM SJG 2017-24; to N.S. Nagathihalli); and American Cancer Society IRG 98-277-13 (to M.N. VanSaun). Flow Cytometry Core service was performed through the Sylvester Comprehensive Cancer Center (SCCC) support grant (to N.B. Merchant and N.S. Nagathihalli).

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.