Inhibition of miR-21 Regulates Mutant KRAS Effector Pathways and Intercepts Pancreatic Ductal Adenocarcinoma Development

Pancreatic ductal adenocarcinoma (PDA) is increasing in incidence, and remains among the deadliest cancers with an overall median survival of 4–6 months (1). Along with factors such as late diagnosis poor prognosis results from its complex tumor microenvironment (TME) composed of transformed epithelial cells and tumor-promoting stromal cells that renders PDA refractory to radiation, chemotherapies, and immunotherapies (2, 3). The stromal compartment consists of PDA-recruited cancer-associated fibroblasts (CAF), pancreatic stellate cells, immune cells, and extracellular matrix proteins that drive disease progression from the earliest premalignant pancreatic intraepithelial neoplasias (PanIN) to PDA (4).

PDA tumorigenesis is initiated, propagated, and maintained by the expression of the driver mutation KRAS (mKRAS; refs. 2, 5). mKRAS is one of the earliest catalytic events and most highly expressed mutation in over 90% of human PanIN1 lesions and PDA (6). mKRAS persistently activates downstream signaling pathways resulting in increased cell proliferation, survival, and metabolic reprogramming (7). Concurrent to the initiation of PanIN development, mKRAS also triggers the formation of the cancer-supportive microenvironment via recruitment of fibro-inflammatory stroma (2). These cancer-associated cell populations express additional signals that support PDA development, progression, and metastasis (8, 9).

Attempts to develop therapies that directly target mKRAS have been met with challenges (10). Most strategies have focused on inhibiting mKRAS directly or downstream effectors of mKRAS signaling (10). However, the clinical benefit of these downstream therapeutics has been limited by compensatory mechanisms, narrow therapeutic windows with associated drug toxicities, the accumulation of additional genetic alterations in tumor cells, and the complexity of signaling by cancer-associated stromal cells (10, 11). Although it is difficult to target mKRAS therapeutically, it may be possible to develop strategies that intercept mKRAS in early premalignancies, before additional genetic and regulatory mechanisms occur (12).

miRNAs are cellular regulators that have been associated with the etiology of many cancers at various stages of development and progression (13, 14). miRNAs are an emerging class of novel therapeutic targets being tested in clinical trials for cancer and other diseases. miRNAs are short ncRNAs, 19–25 nucleotides in length, that posttranscriptionally regulate gene expression by binding to complementary 3′ untranslated regions of target mRNAs resulting in their degradation or translational repression (13, 15). Currently, three miRNA therapeutics have advanced to clinical trials for the treatment of solid malignancies and leukemia (13, 16). Recently, interfering RNAs as a therapeutic approach to PDA have been explored using tumor-penetrating nanoparticles as a delivery modality that can penetrate the dense tumor stroma (17, 18). This technique has demonstrated efficacy in xenograft murine models with dysregulated miRNAs, specifically when both miR-21 and KRAS were simultaneously targeted (18, 19). Although miRNA expression in PDA has been reported, progressive profiling of dysregulated miRNAs with PanIN progression has been minimally studied.

We identified miR-21 and miR-224 as cell-specific and compartment-specific regulators in both PanINs and PDA. To study miRNA regulation of early PanIN development and progression, we used transgenic Kras^G12D/+;Trp53^R172H/+;Pdx-1-Cre (KPC) mice that exhibit stepwise progression through multiple stages of PanINs (PanIN1–3) to invasive PDA (20). KPC mice resemble human PDA by genetic instability and histology, including desmoplastic TME (20). Endogenous inhibition of miR-21 in primary KPC tumor cells decreased protumorigenic functionalities relative to baseline levels, while overexpression of miR-224 in normal pancreatic fibroblasts conferred activated phenotypes. Inhibition of miR-21 in tumor cells simultaneously downregulated multiple tumorigenic pathways downstream of activated mKRAS. In vivo studies demonstrated that miR-21 inhibition increased survival in mice with established PDA, while early administration of a miR-21 inhibitor intercepted tumorigenesis and delayed premalignant progression. Collectively, our studies identify miR-21 as a novel target for interception of mKRAS-driven PDA development.t

All animal studies were approved by the Institutional Animal Care and Use Committee of Johns Hopkins University (Baltimore, MD). Lox-STOP-Lox-Trp53^R172H/þ; Lox-STOP-Lox-Kras^G12D/þ; and Pdx-1-Cre strains on a mixed C57BL/6 background were gifted from Dr. David Tuveson (Cold Spring Harbor Laboratory, Cold Spring, NY). These mice were backcrossed to the C57BL/6 genetic background for 12 generations and interbred to obtain KPC mice. All wild-type (WT) C57BL/6 mice were purchased from The Jackson Laboratories. All animals were kept in pathogen-free conditions and treated in accordance with the Institutional Animal Care and Use Committee of the Johns Hopkins University and American Association of Laboratory Animal Care–approved guidelines. Animals were fed Teklad global 18% protein extruded rodent diet. The Allentown Individually Ventilated Caging System was used to house animals with Tek-Fresh paper bedding that is also an acceptable enrichment modality. Interventions were performed during light cycle.

Pancreata were isolated from aged KPC (4–20 weeks) or WT C57BL/6 mice and fresh-frozen in Tissue-Tek OCT (Sakura Finetek USA) or formalin-fixed, paraffin-embedded (FFPE) by the Johns Hopkins University Oncology Tissue Services. Guide slides were sectioned and stained with hematoxylin and eosin by the Johns Hopkins University Oncology Tissue Services. Guide slides containing various grades of PanINs were identified by a pathologist. Ten, 10 μm frozen sections adjacent to guide sections were mounted on membrane slides (Zeiss) and stained with Cresyl Violet (Ambion). FFPE sections were stained with the Arcturus Paradise Plus Reagent System Staining Kit (Thermo Fisher Scientific). Normal, low-grade PanIN1, high-grade PanIN2/3, and PDA tissue were microdissected using laser capture microdissection (LCM; Leica LMD7000).

RNA was extracted from fresh-frozen LCM samples using the RNAqueous-Micro Total RNA Isolation Kit (Thermo Fisher Scientific), quantified with a NanoDrop 1000 Spectrophotometer (Thermo Fisher Scientific), and converted to cDNA using the TaqMan MicroRNA Reverse Transcription Kit (Thermo Fisher Scientific) with Megaplex RT Primers Rodent Pool A and B (Thermo Fisher Scientific). cDNA was preamplified 18 cycles using the TaqMan PreAmp Master Mix (Thermo Fisher Scientific) and Megaplex PreAmp Primers Rodent Pool A and B (Thermo Fisher Scientific). miRNA microarray was performed by the Johns Hopkins School of Medicine Genetic Resources Core Facility using the TaqMan OpenArray Rodent MicroRNA Panel (Thermo Fisher Scientific) and TaqMan OpenArray Real-Time PCR Master Mix (Thermo Fisher Scientific). All cycle threshold (C_t) values were normalized to endogenous levels of U6 snRNA. Fold changes were quantified utilizing the ΔΔC_t method in which all PanIN groups were normalized to the normal group (normal ducts isolated from WT C57BL/6 mice). P values for differential expression statistics were generated and adjusted for multiple hypotheses testing using ExpressionSuite (Thermo Fisher Scientific).

Statistical analysis was performed using GraphPad Prism v7.0b (GraphPad Software). The data are presented as the means ± SEM. Comparisons between groups were made using two-tailed unpaired Student t tests and one-way or two-way ANOVA. Survival data were plotted using a Kaplan–Meier curve and differences in treatment groups were tested using the log-rank (Mantel–Cox) test. For all analyses, statistical significance is defined and shown as *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Statistical analyses for RNA-sequencing (RNA-seq) and The Cancer Genome Atlas (TCGA) are performed using R as described in the Supplementary Materials and Methods.

miRNA microarray data are available in the Gene Expression Omnibus under the accession number GSE143648. RNA-seq data are available in the Gene Expression Omnibus under the accession number GSE143326.

We conducted a comprehensive miRNA profiling study of multiple PanIN stages and PDA using KPC mice. We performed microdissections on increasing grades of PanINs from sections of KPC pancreata followed by a miRNA microarray analysis utilizing the TaqMan OpenArray platform to quantify the levels of 750 known murine miRNAs.

miRNA candidates were selected using specific criteria. First, candidate miRNAs had C_t values <30 and within the detection limits of the assay. Second, candidate miRNAs had large fold changes in expression that were consistently increasing or decreasing with PanIN progression to PDA. Third, we chose miRNAs with significant fold changes (P_adj < 0.05) in each PanIN stage relative to normal ducts of WT C56BL/6 mice. Confirmatory qPCR was completed to validate all miRNAs identified by OpenArray. Finally, the miRNAs had to have either putative or proven roles in cancer-promoting pathways.

Our selection criteria resulted in 14 miRNA candidates. Supplementary Tables S1 and S2 summarize the fold changes observed with increasing stages of PanIN development to PDA of our top 14 candidates and the primers used for subsequent validation. We found significant dysregulation of some miRNAs (miR-21, miR-16, and miR-28) as early as low-grade PanIN1 (P1), while dysregulation of others (miR-224, miR-28, and miR-216b) began at higher PanIN grades (P2/3). All 14 candidate miRNAs have significantly dysregulated expression in PDA compared with expression levels in normal ducts of WT mice.

We chose to further evaluate miR-21 and miR-224 because these two miRNAs have been reported to regulate targets in cancer-promoting inflammatory pathways (21) and epithelial–mesenchymal transition (EMT; ref. 22), respectively. We performed miRNA FISH (miR-FISH; Fig. 1A and B) to examine the endogenous expression of miR-21 and miR-224 in specific cell populations in early and late PanINs and in PDA in late PanINs and in PDA (Fig. 1C and D). miR-FISH allows the fluorescent visualization of miRNAs in sections of whole tissue. The intensity of the fluorescent signal directly correlates with levels of endogenous miRNA expression. We used modified bicyclic locked-nucleic acid (LNA) probes that are complimentary in sequence and therefore will stably and accurately hybridize to miRNAs of interest at specific hybridization temperatures in situ.

miR-FISH staining of WT ducts, increasing grades of KPC PanINs, and PDA tissue demonstrates spatial–temporal expression of miR-21 is specifically concentrated in the ductal epithelial cells of PanIN lesions and increases in expression with progression to PDA (Fig. 1C). The quantified raw fluorescent intensities of miR-21 in PDA tissue is significantly increased compared with normal WT ducts, while the normalized fluorescent signal indicates that high-grade PanIN2/3 lesions have a significant 3.5-fold increase and PDAs have a significant 6-fold increase of miR-21 expression compared with normal ducts (Fig. 1C), which correlates closely to the 7-fold increase of miR-21 expression in KPC tumor cells relative to WT pancreatic epithelial cells quantified by qPCR (Supplementary Fig. S1). In contrast, miR-224 expression localized in the invading stromal compartment surrounding high-grade PanIN2/3 lesions and persists with progression to PDA (Fig. 1D). These data are consistent with our findings in KPC CAFs that miR-224 expression is upregulated relative to normal pancreatic-associated fibroblasts (PNAF) early in the premalignant stromal cell compartment of developing PDAs. Moreover, the normalized fluorescent signal of miR-224 produced by miR-FISH (Fig. 1D) demonstrates that the significant 10-fold upregulation of miR-224 in PDA stroma equals the 10-fold increase of miR-224 expression in KPC CAFs as compared with PNAFs (Supplementary Fig. S1). The miR-FISH results qualitatively and quantitatively confirm qPCR results and further inform the compartmental-specific expression of miR-21 and miR-224 throughout early PDA development.

To investigate the functional roles of miR-21 and miR-224 in regulating early PDA development, we stably transduced the primary cell lines with miRNA inhibitors and mimics. Because KPC tumor cells express higher levels of miR-21 than WT epithelial cells and KPC CAFs express higher levels of miR-224 than PNAFs, we used lentiviral vectors to stably inhibit endogenous levels of miR-21 in KPC tumor cells and miR-224 in KPC CAFs. The inhibition occurs through expression of miRNA inhibitors that bind to target miRNAs via sequence complementation to prevent the loading of target miRNAs onto the RNA-induced silencing complex. Conversely, we transduced WT pancreatic epithelial cells and PNAFs with lentivirus overexpressing miR-21 and miR-224, respectively. Successfully transduced cells were selected using antibiotics specific to each construct, and further enriched using FACS (Supplementary Figs. S2 and S3A–S3F). Supplementary Fig. S3G summarizes the lentiviral vectors, target cell lines, fluorescent markers, and antibiotic selection used to generate stable cell lines expressing their associated miRNA mimic or inhibitor.

The degree of miRNA inhibition was measured by qPCR. miRNA inhibitors significantly reduced the expression of miR-21 in KPC tumor cells and miR-224 in KPC CAFs about 2-fold compared with nontransduced cells. However, inhibition was incomplete due to residual miR-21 and miR-224 expression being higher than in WT epithelial cells (Fig. 2A) and PNAFs, respectively.

Next, we investigated functional changes in KPC tumor cells following reduced expression of miR-21. miR-21 inhibition significantly reduced proliferation by 2-fold relative to nontransduced KPC tumor cells (Fig. 2B). More importantly, downregulation of miR-21 significantly reduced both migration and invasion to levels comparable with that of WT pancreatic epithelial cells (Fig. 2C and D), demonstrating that reduction of miR-21 levels in KPC tumor cells can revert the cell's migratory and invasive capacities to that of normal untransformed epithelial cells. Interestingly, downregulation of miR-224 in KPC CAFs did not affect cell proliferation (Fig. 3A and B). However, the migratory capacity of KPC CAFs was significantly increased by 1.7-fold in miR-224–inhibited CAFs compared with nontransduced CAFs (Fig. 3C). Invasion was not affected by downregulation of miR-224 in KPC CAFs (Fig. 3D).

PNAFs overexpressing miR-224 have a significant 2-fold increase in cell proliferation compared with nontransduced PNAFs (Fig. 3E). Scratch assay demonstrated upregulation of miR-224 expression significantly increased PNAF migratory capacity by 2-fold compared with all other control groups to achieve 100% wound closure (Fig. 3F). This migratory capacity is comparable with KPC CAFs, indicating that miR-224 overexpression can confer activated phenotypes like migration to PNAFs. In addition, PNAFs overexpressing miR-224 have significantly increased cell invasion compared with nontransduced PNAFs (Fig. 3G). Collectively, this demonstrates increased expression of miR-224 in normal fibroblasts produces CAF-like activated cellular functions.

RNA-seq was employed to investigate global miRNA-mediated modulation of key genes and pathways regulated by miR-21 in tumor cells and miR-224 in CAFs. Log fold changes for all gene transcripts were compared between (i) miR-21–inhibited and scramble-inhibited KPC tumor cells and (ii) miR-224–inhibited and scramble-inhibited KPC CAFs. Generally Applicable Gene-set Enrichment (GAGE) (23) analyses on the differentially expressed gene sets produced by miR-21 and miR-224 inhibition revealed miR-21 inhibition significantly downregulated the MAPK, mTOR, and actin cytoskeleton Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways in tumor cells (Supplementary Fig. S4A), while miR-224 inhibition significantly downregulated the DNA replication, cell cycle, and p53 signaling KEGG pathways in CAFs (Supplementary Fig. S4B).

The heatmaps for the three KEGG pathways significantly downregulated by miR-21 inhibition in KPC tumor cells (highlighted top four rows) are shown in Supplementary Fig. S4A. Although not statistically significant (P = 0.06), the RAS pathway was included for analysis due to the biological importance and relevance of the pathway to both early PanIN progression and PDA development. All four heatmaps display downregulation in all four pathways (in blue) for the two miR-21–inhibited tumor samples compared with the two scramble-inhibited tumor samples, which have higher gene expression. In contrast, there is less consistent downregulation in the CAF samples (bottom four rows), indicating that the regulation of these four KEGG pathways are specific to miR-21's role in PDA cells. The significant downregulation of MAPK, mTOR, and actin cytoskeleton pathways by miR-21 inhibition in tumor cells is in accordance with the reduced proliferation, migration, and invasion observed with stable miR-21 inhibition. We selected five differentially expressed genes from each pathway for qPCR validation of fold changes (Supplementary Fig. S5A–S5E; Supplementary Table S3).

The heatmaps for the three KEGG pathways that were significantly downregulated by miR-224 inhibition in KPC CAFs (highlighted bottom four rows) are shown in Supplementary Fig. S4B. Despite being statistically significant, the heatmaps of all three pathways lack consistent gene expression changes between samples within each treatment group. In addition, there is minimal clustering within the two samples of scramble-inhibited KPC CAFs. The transcriptional variability between biological replicates may indicate that there is greater heterogeneity of expressed genetic pathways following miR-224 inhibition in CAFs than can be adequately measured by this approach. We chose five genes (Shisa2, Cxcl12, Rasgef1b, Vav3, and Dusp9), regardless of pathway classification, that were significantly regulated by miR-224 inhibition for qPCR validation of fold changes (Supplementary Fig. S5F). Of the genes screened, Shisa2 was most significantly downregulated by miR-224 inhibition in CAFs, with a 255-fold reduction by qPCR (greater than 86-fold reduction by RNA-seq) when compared with the expression in CAFs transduced with scramble inhibitor. Shisa2, a regulatory endoplasmic reticulum protein, inhibits FGF signaling by preventing the maturation and cell surface expression of FGFRs (24). FGF signaling through FGFRs have been shown to promote cell migration in many cancer types and embryogenesis (25, 26). Therefore, the significant reduction in Shisa2 levels by miR-224 inhibition and potential increase in FGFR may explain the increased migration of miR-224–inhibited CAFs in vitro (Fig. 3C).

We next investigated the effects of inhibiting miR-21 in vivo. Between 60% and 90% of primary PDA tumor volume is fibrotic stroma (8, 27). Therefore, to more accurately recapitulate the major components of the PDA TME, we established subcutaneous tumors in C57BL/6 mice with a 2 CAF:1 KPC tumor cell ratio of different combinations of miRNA-inhibited or normal KPC tumor cells or CAFs. Mice with tumors composed of miR-21–inhibited KPC tumor cells and normal KPC CAFs (T 21i/CAF) had the lowest tumor burden and longest survival compared with all other groups (Fig. 4A and B). IHC staining of resected tumors with the proliferation marker Ki67, demonstrated that T 21i/CAF tumors had the lowest percentage of actively proliferating Ki67⁺ cells (Fig. 4C and D), suggesting that reduced tumor size is because of decreased proliferative capacity of miR-21–inhibited tumor cells. Tumors composed of normal KPC tumor cells and miR-224–inhibited CAFs (T/CAF 224i) were comparable in size with scramble-inhibited tumors (T Scri/CAF Scri) and larger than normal nontransduced tumors (T/CAF). Collectively, these data show that only miR-21 inhibition of tumor cells reduces tumor burden and confers survival benefit in established PDA.

Systemic administration of miR-21 and miR-224 inhibitors delayed progression of PanIN lesions to PDA in KPC mice. To deliver miR-21 and miR-224 inhibitors, oligonucleotides were chemically modified with a LNA ribose ring structure and phosphorothioate backbone for enhanced pharmacokinetic stability in vivo (28). Four- to 5-week old KPC mice, which are expected to already have low-grade PanIN1 lesions, were dosed with negative control LNA-scramble inhibitor, LNA-miR-21 inhibitor, or LNA-miR-224 inhibitor once every 5 days for 6 weeks. At the end of the dosing regimen, the pancreas, liver, lung, and kidney from treated and untreated mice were collected and analyzed for drug efficacy and toxicity. Each pancreas was pathologically evaluated for the highest grade lesion present. KPC mice receiving early systemic administration of LNA-miR-21 inhibitor developed the fewest advanced PanINs compared with other treatment groups (Fig. 5A). The average grade of premalignant lesions in the pancreas of mice dosed with LNA-miR-21 inhibitor was significantly lower than that of untreated mice. Furthermore, mice that received LNA-miR-21 inhibitor did not develop any lesions beyond the low-grade PanIN1 stage, demonstrating that early systemic miR-21 inhibition successfully delays premalignant progression. In addition, our studies show that as a monotherapy, LNA-miR-21 inhibitor effectively intercepts PDA tumorigenesis when administered early at the time of mKRAS activation. No adverse effects of systemic miR-21 inhibition were observed in post-mortem examination (Supplementary Fig. S6).

qPCR quantification of miRNA levels in the whole pancreas of dosed KPC mice showed that LNA-miR-21 inhibitor–treated mice had a significant 112-fold decrease and 50-fold decrease of miR-21 levels in their pancreas compared with untreated mice and LNA-scramble inhibitor–treated mice, respectively (Fig. 5B). Mice that received LNA-miR-224 inhibitor had a significant 19-fold decrease and 14-fold decrease of miR-224 levels in their pancreas as compared with untreated mice and LNA-scramble inhibitor–treated mice, respectively (Fig. 5C). In addition, the data demonstrate that systemic miR-224 inhibition does not affect localized expression of miR-21 and vice versa, thereby demonstrating the specificity of each miRNA inhibitor. Collectively, these data show that systemically administered miRNA inhibitors successfully localized to the pancreas and inhibited their targets with a high degree of specificity.

To examine translation of our findings to human PDA, we investigated miR-21 and miR-224 expression in the PDA cohort of TCGA. Specifically, we looked for a correlation between miR-21 expression and epithelial cell content as well as miR-224 expression and fibroblast content. This corresponds to our murine findings suggesting miR-21 and miR-224 cellular and compartmental specificity. Using the publicly available xCell quantification for TCGA (29), we compared epithelial and fibroblast content with miR-21 and miR-224 expression in 177 patients in the PDA cohort. Correlation analysis revealed a significant direct relationship between miR-21 expression and tumor epithelial cell content with P values of 8.7e-7 and 2.1e-13 for miR-21-3p and miR-21-5p, respectively (Fig. 6A). This relationship was not seen with comparison between miR-21 expression and tumor fibroblast content and is consistent with our findings in KPC model.

Similar comparison for miR-224 expression revealed an inverse relationship between miR-224 expression and tumor fibroblast content and a direct relationship between miR-224 expression and tumor epithelial cell content (Fig. 6B). Although the miR-224 findings were not consistent with our in vitro and in vivo murine findings, the broad classification of fibroblasts in the dataset may not be limited to CAFs, which would impact our analysis.

MiRNAs are regulatory ncRNAs controlling signaling pathways in both normal and tumor cells. We identified miR-21 and miR-224 as regulators of specific cell types within developing mKRAS-driven PDAs. We demonstrate that miR-21 is a potent regulator of EMT, modulating transformed epithelial cell invasion and migration. We show for the first time that miR-21 can serve as a target for intercepting PDA development following mKRAS activation in early PanIN lesions. We also identified miR-224 as a key regulator of TME fibroblasts, activating normal fibroblasts to exhibit CAF activity. Finally, we extrapolated our findings to the PDA cohort of TCGA and demonstrated a correlation between miR-21 expression and tumor epithelial cell content.

Because mKRAS targeting remains a challenge in the clinical setting, inhibiting downstream effector pathways is a leading strategy for developing mKRAS therapeutics (10, 11). Our study demonstrates miR-21 inhibition simultaneously downregulates protumorigenic effector pathways including MAPK, mTOR, and actin cytoskeleton pathways, downstream of mKRAS in KPC tumor cells. Previous studies demonstrated miR-21 directly inhibits individual tumor suppressor gene targets such as PTEN (30, 31), programmed cell death protein 4 (PDCD4; ref. 32), Sprouty2 (Spry2; ref. 33), tropomyosin (TPM1; ref. 34), and RAS p21 protein activator 1 (RASA1; refs. 35, 36) to modulate cell proliferation, migration, invasion, and apoptosis in various cancer types. However, here we show miR-21 inhibition is capable of simultaneously downregulating multiple mKRAS-activated biologic pathways responsible for tumor cell growth, metabolism, and EMT, indicating that miR-21 is a more potent regulator of mKRAS signaling than previously realized. In addition, miR-21 has been identified as a downstream transcriptional target of KRAS activation via transcription factor ELK1, which directly binds to miR-21's proximal promoter region, leading to increased miR-21 expression (35). Thus, miR-21 is a ncRNA effector of KRAS activation, providing additional rationale for its inhibition in mKRAS-driven PDA. Studies have shown that targeting individual downstream KRAS effectors such as MEK1/2, Erk1/2, or Akt has limited clinical efficacy due to compensatory mechanisms swiftly leading to acquired resistance, coupled with substantial toxicities (10, 11). We have also demonstrated that miR-21 expression is directly related to tumor epithelial cell content for the PDA cohort of TCGA. miR-21 is a promising therapeutic target because its inhibition can block oncogenic KRAS signaling through the regulation of multiple signaling proteins in multiple key protumorigenic pathways, thereby potentially reducing mechanisms of acquired resistance and toxicities associated with combination drug treatments.

Our study shows for the first time that miR-224 is specifically upregulated in activated KPC CAFs compared with normal fibroblasts, and its increased spatial–temporal expression in the developing TME is specifically localized in the stromal compartment surrounding high-grade PanIN2/3 lesions and PDA. The in situ localization of miR-224 contrasts that of miR-21's, whose expression is concentrated in transformed ductal epithelial cells. Moreover, endogenous overexpression of miR-224 in normal pancreatic fibroblasts significantly increased cell proliferation, migration, and invasion, indicating that overexpression of miR-224 may generate CAFs by conferring activated phenotypes to normal fibroblasts.

We postulate that the early upregulation of miR-224 may be because of the increased expression of hypoxia-inducible factor-1 α (HIF1α) and NF-κB in the developing TME (8, 37). Both HIF1α in gastric cancer and NF-κB in hepatocellular carcinoma transcriptionally regulate miR-224 by binding directly to its promoter region (38, 39). Although only demonstrated in tumor cells, this same mechanism of action may lead to increased miR-224 expression and potential miR-224–mediated activation of normal resident fibroblasts in the premalignant and PDA microenvironment. Conversely, endogenous inhibition of miR-224 in KPC CAFs did not affect cell proliferation or invasion, but rather increased cell migration, an effect possibly mediated by significant downregulation of Shisa2 in miR-224–inhibited CAFs quantified by RNA-seq. Shisa2, a negative regulator of FGF signaling (24), may promote cell migration upon downregulation in KPC CAFs. Although miR-224 inhibition has been shown to reduce protumorigenic functionalities of tumor cells in many cancer types (38, 40), the effects of miR-224 inhibition in CAFs have not been reported. Future studies are warranted to determine how best to regulate miR-224 activity in CAFs to revert or reduce their activated phenotype, and to determine whether such reversion contributes to reduced tumorigenicity of PDA, either through reduced tumor cell growth and metastasis or through improved anticancer therapy access to the PDA TME, or both.

Our in vivo studies demonstrate that miR-21 inhibition produces therapeutic effects at multiple stages of PDA development. We report for the first time that early systemic administration of the LNA-miR-21 inhibitor significantly decreased the occurrence of high-grade PanINs and PDA in KPC mice. Our study shows that inhibiting miR-21 may be a promising strategy to prevent or slow the development of other KRAS-driven cancers. Our data are consistent with previous findings that prove miR-21 to be an oncogenic miRNA and that its inhibition impedes tumorigenesis in vivo. Utilizing a doxycycline-inducible mir-21^LSL-Tetoff mouse model, induction of miR-21 produced pre-B-cell lymphoma, while miR-21 inactivation led to complete tumor regression (41). In addition, in a spontaneous transgenic mouse model of KRAS-driven NSCLC, overexpression of miR-21 enhanced tumorigenesis, while knockout of miR-21 reduced tumor formation in KRAS^LA2 mice (42). In this study, we report that early systemic miR-21 inhibition can specifically impede early mKRAS-driven PDA tumorigenesis.

Our data along with those of previous groups show miR-21 as a potential diagnostic marker for the presence of PanINs and PDA due to miR-21's increased expression throughout premalignant to PDA development (43, 44). Stable exosome encapsulated miR-21 is also elevated in the serum of patients with premalignancies and PDA, demonstrating that miR-21 is a potential circulating diagnostic marker for noninvasive early and established PDA screening reflective of increased in-tissue expression (45, 46). Therefore, elevated levels of miR-21 may indicate when miR-21 inhibition treatment should be initiated during early disease development. However, more accurate PanIN and PDA detection is necessary for the success of preventative therapies.

Our studies demonstrate that miR-21 is an important regulator of KPC tumor cell function both in vitro and in vivo. However, increasing evidence supports that miR-21 is also a regulator of fibroblast activation and CAF functionality. miR-21 expression is increased in CAFs compared with normal fibroblasts and miR-21 overexpression in normal fibroblasts enhances levels of CAF-specific markers, cell motility, and proliferation, indicating that miR-21, like miR-224, may induce CAF formation (47, 48). Moreover, inhibition of miR-21 reduces CAF migration, invasion, and proliferation (47). miR-21 achieves this regulation in fibroblasts by directly targeting and downregulating SMAD7, an inhibitor of the TGFβ pathway, thereby sensitizing fibroblasts to TGFβ-mediated CAF formation and fibrosis (49). In addition to cancer models, fibroblast activation in models of diabetic nephropathy and pulmonary fibrosis has been shown to occur via the same TGFβ/miR-21/SMAD7 regulatory axis (49, 50). In fact, in vivo inhibition of miR-21 attenuated fibrosis in a mouse model of fibrotic lung disease (50), a finding that is highly applicable to PDA intervention because the fibrotic stroma is an integral component of the TME that propagates PDA. Furthermore, the reported findings that miR-21 inhibition reduces fibrosis provide another mechanism of action for the PanIN-intercepting efficacy of our early systemic miR-21 inhibitor treatment reported in this study. Systemic dosing of LNA-miR-21 inhibitor exposed the entire premalignant microenvironment to miR-21 inhibition, which inhibited transformed epithelial cells and possibly the activation of fibroblasts.

In addition to CAFs in the stromal compartment, miR-21 has also been shown to promote protumorigenic immune populations. miR-21 has been shown to induce the polarization of tumor-supportive anti-inflammatory M2 macrophages (51), the differentiation of naïve CD4⁺ T cells to regulatory T cells (52), and the expansion of myeloid-derived suppressor cells that exhibit augmented suppressive capabilities upon miR-21 overexpression (53). Inhibition of miR-21 in the premalignant microenvironment may act to simultaneously reduce fibrosis and suppress the generation of immunosuppressive cell populations, which may cumulatively delay premalignant progression and reduce tumor growth. Therefore, miR-21 exerts its oncogenic effects on multiple cell populations, providing abundant evidence for inhibiting this miRNA in early disease progression. miR-21 inhibition shows efficacy as a monotherapy; however, future studies should test miR-21 inhibition in combination with other agents to modulate several components of the TME simultaneously to more effectively impede disease development.

In summary, we have shown that miR-21 and miR-224 are dysregulated throughout early premalignant progression to established PDA, and demonstrate cell-specific overexpression in the developing microenvironment. These data provide new opportunities for intercepting KRAS-driven premalignancy through impeding epithelial cell transformation and by reducing the PDA stromal barrier to therapeutic access.

N.J. Chu reports grants from NIH/NCI during the conduct of the study. E.J. Fertig reports personal fees from Champions Oncology outside the submitted work. A.C. Hopkins reports grants from NIH during the conduct of the study. T.D. Armstrong reports grants from NCI during the conduct of the study. E.M. Jaffee reports grants from NIH and grants from Lustgarten Foundation during the conduct of the study, personal fees from CSTONE, grants from AduroBiotech, grants from Bristol Myer Squibb, personal fees from Achilles, personal fees from DragonFly, personal fees from Genocea, personal fees from Adaptive Biotechnologies, nonfinancial support from Parker Institute, nonfinancial support from National Cancer Advisory Board, and nonfinancial support from AACR Board of Directors outside the submitted work. J.W. Zimmerman reports grants from NIH during the conduct of the study. No potential conflicts of interest were disclosed by the other authors.

Conception and design: N.J. Chu, R.A. Anders, B.P. Keenan, T.D. Armstrong, E.M. Jaffee

Development of methodology: N.J. Chu, R.A. Anders, M. Cao, E.M. Jaffee, J.W. Zimmerman

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N.J. Chu, M. Cao, A. Popovic, J.W. Zimmerman

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): N.J. Chu, R.A. Anders, E.J. Fertig, M. Cao, A.C. Hopkins, E.M. Jaffee, J.W. Zimmerman

Writing, review, and/or revision of the manuscript: N.J. Chu, R.A. Anders, E.J. Fertig, A.C. Hopkins, T.D. Armstrong, E.M. Jaffee, J.W. Zimmerman

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): N.J. Chu, J.W. Zimmerman

Study supervision: N.J. Chu, R.A. Anders, T.D. Armstrong, E.M. Jaffee

The authors acknowledge financial support from the NIH/NCI grant R01CA184926 (to E.M. Jaffee), Bloomberg Philanthropies, The Bloomberg Kimmel Institute for Cancer Immunology, The Skip Viragh Center for Pancreatic Cancer, The McGlothlin Foundation, and Lefkovsky Family Foundation grant 90058175 (to E.M. Jaffee). R.A. Anders and E.M. Jaffee have support from NCI SPORE P50 CA062924. J.W. Zimmerman was supported by 5T32 CA009071. The authors thank the Sidney Kimmel Comprehensive Cancer Center (SKCCC), the Skip Viragh Center for Pancreas Cancer, and the Bloomberg∼Kimmel Institute for Cancer Immunotherapy at Johns Hopkins. The authors also thank the Johns Hopkins University Oncology Tissue Services for all histology and slide preparation, the SKCCC Cell Imaging Core for fluorescent and infrared imaging, the SKCCC Flow Cytometry Core for cell sorting, the SKCCC Experimental and Computational Genomics Core for RNA-seq, and the Johns Hopkins School of Medicine Genetic Resources Core Facility for miRNA microarray profiling. The authors thank L. Zheng (Johns Hopkins University School of Medicine, Baltimore, MD) for primary KPC and PNAF cell lines.

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.