miR-1298 Inhibits Mutant KRAS-Driven Tumor Growth by Repressing FAK and LAMB3

Mutations in the KRAS oncogene are found in approximately 20% of human cancers and is the most commonly mutated gene in the RAS family of oncogenes (85% of RAS mutations are KRAS mutations). RAS is a GTPase that cycles between GTP-bound (active) state and GDP-bound (inactive) forms through the actions of guanine nucleotide exchange factors (GEF) and GTPase-activating proteins (GAP). Activating mutations found in cancers (most commonly at conserved codons 12 and 13) impair the GTPase activity of KRAS, resulting in the constitutive activation of KRAS and its downstream pathways, including the RAF–MEK–MAPK and PI3K–AKT pathways that regulate critical cell processes such as proliferation and survival (1, 2). Activating KRAS mutations are major driver mutations in non–small cell lung cancer (NSCLC) with mutations in KRAS found in about 30% of NSCLC cases, representing the largest fraction of known driver mutation associated NSCLC. KRAS mutations are also frequently identified in 40% to 50% of colorectal cancers, and KRAS mutational status is a key factor in deciding on treatment options for metastatic colorectal cancer patients due to the ineffectiveness of anti-EGFR antibody (cetuximab or panitumumab) therapy in metastatic colorectal cancer patients with mutant KRAS (2). In NSCLC, KRAS mutations are generally mutually exclusive to both EGFR mutations and ALK rearrangements, and KRAS-mutated cancers are generally resistant to targeted tyrosine kinase inhibitors targeting these driver mutations (erlotinib for EGFR mutations and crizotinib for ALK rearrangements).

Despite decades of effort, no targeted therapy is yet available clinically for KRAS-mutated cancers despite much effort. Because of the absence of well-defined binding pockets at key active sites (3), efforts to inhibit KRAS activity with small molecules have been largely fruitless (4–6). For this reason, synthetic lethal approaches, in which genetic perturbations or drugs that are normally not lethal to the cell become lethal in the presence of another genetic perturbation (in this case KRAS oncogenic mutation), have been used to identify potentially tractable targets for therapy. Several groups have conducted systematic RNA interference (7–11) or small-molecule (12) screens to identify synthetic lethal interactions with oncogenic KRAS. Such screens have identified novel targets such as PLK1, TBK1, STK33, and GATA2.

Prior functional genomic screens have largely focused on coding genes. Although more than 90% of the human genome is transcribed, only approximately 2% codes for proteins (13), highlighting the importance of noncoding RNAs (ncRNA) for biologic function. ncRNAs are now implicated as crucial regulators of diverse aspects of cellular processes. Among the different ncRNAs, small conserved sequences called miRNA have been the best characterized over the past decade and are known to be involved in crucial cellular processes and widely dysregulated in cancer. miRNAs are small (∼22 nucleotide) noncoding RNAs that regulate gene expression posttranscriptionally by binding to specific regions in the 3′-untranslated region (UTR) of mRNA targets to promote their degradation and/or inhibit their translation. Because a single miRNA can potentially bind to and regulate hundreds, even thousands, of targets, they can have profound effects on gene expression networks (14). Accordingly, numerous miRNAs have been shown to regulate vital cellular processes in both physiological and pathologic conditions (15).

miRNAs are generally downregulated in tumors (16) with many miRNAs acting as tumor suppressors. Impairment of miRNA maturation has been reported to enhance the transformed phenotype of cancer cells and accelerate tumor formation in KRAS-driven models of lung cancer (17). The well-studied let-7 family of miRNAs were the first miRNAs to be implicated with RAS (18) and have been shown to repress KRAS expression and inhibit mutant KRAS-dependent cancer cell growth in vitro and tumor growth in vivo (18–20). Other miRNAs have been implicated as KRAS tumor suppressors, miRNA-96 in KRAS pancreatic cancer (21), miRNA-181a in KRAS-mutant oral squamous cell carcinoma (22), and miRNA-30b in colorectal cancer (23).

In this study, we screened a library of human miRNA mimics to identify novel synthetic lethal interactions with oncogenic KRAS mutation. In screening for synthetic lethal interactions, two main approaches have been traditionally used: screening across large panels of different cell lines harboring different mutations or comparing paired isogenic cell lines that are derived from identical parental cell lines but differ in the mutation of interest. We used both approaches in this study, carrying out an initial screen in an isogenic colorectal cell line differing only in the presence or absence of an activating KRAS mutation, and then validating our hits in a panel of lung cancer cell lines with and without KRAS mutations. By combining cell lineages from both colorectal cancer and NSCLC with different KRAS-activating mutations, we sought to identify more conserved synthetically lethal KRAS interactions with potentially less lineage dependence. Using this approach, we identified miR-1298 and its targets, FAK and LAMB3, as suppressors of KRAS-dependent cell growth in both colorectal cancer and NSCLC contexts, and demonstrate that ncRNAs can be used for synthetic lethal screens to functionally annotate their function and uncover novel signaling regulatory pathways in cancer.

The paired isogenic HCT116 KRAS^WT/− and HCT116 KRAS^WT/G13D colon cancer cell lines were kind gifts from Dr. Bert Vogelstein (Johns Hopkins, Baltimore, MD). A549, NCI-358, NCI-H460, NCI-H522, NCI-H1568, NCI-H1792, NCI-H23, NCI-H1975, and NCI-H1437 were obtained from the ATCC. HCT116 cells were cultured in DMEM (Invitrogen) with 10% FBS (Invitrogen) and 1% penicillin/streptomycin (Invitrogen). All lung cancer cell lines were cultured in RPMI1640 (Gibco) with 10% FBS and 1% penicillin/streptomycin. HCT116 Cell lines were validated by PCR amplification and Sanger Sequencing to confirm the mutation at the genomic level. Strict bio-banking procedures were followed and cells were tested for contamination, including mycoplasma. All other cell lines were obtained from the ATCC and used within 2 to 4 months.

Cells were transfected with miRNA mimics or siRNAs by reverse transfection using RNAiMAX (Invitrogen), according to the manufacturer's recommendations. siRNAs were used as pools of 4 per gene (Dharmacon SMARTpools) unless specified. For the miRNA mimic library screening, the HCT116 isogenic pair (800 cells/well in 384-well plates) were reverse transfected with 10 nmol/L of a human miRNA mimic library containing approximately 858 mimics (Ambion) and 0.1 μL RNAiMax/well. Cell lines were used 5 days after transfection.

Cells were plated in triplicate at 3,000 cells per well in 384- or 96-well plates. Five days later, cell viability was determined using the ATPLite luminescent assay (PerkinElmer) according to the manufacturer's instructions. Luminescence signals were normalized and the average of triplicates was designated the cell viability score. A viability score of 1.0 indicates no difference between test cells and cells treated with a control miRNA mimic. For validation, proliferation was also measured with the CyQUANT NF Cell Proliferation Assay Kit (Invitrogen), which uses a fluorescent cell–permeable DNA-binding dye.

Lentiviruses were produced by transfection of 293T cells with pMIF-Zeo-cGFP-miR-1298 or pMIF-Zeo-cGFP-miR-Ctrl vectors (construction described below) together with the packaging plasmids FIV and VSV-G using Fugene HD (Roche). Culture supernatants containing lentiviruses were collected at 48 and 72 hours after transfection. Viral supernatants were pooled and stored at −80°C. NSCLC cells were infected with a 1:10 dilution of supernatant in polybrene-containing medium. Cells were selected with zeocin starting at 24 hours after infection. After 72 hours, 80% to 90% of the cells were GFP⁺.

NSCLC cells were infected with miR-Ctrl- or miR-1298–expressing lentiviruses for 72 hours washed, and resuspended. Groups of 6 Balb/c nude mice were injected subcutaneously into the left and right flanks with 10⁶ control or miR-1298–infected cells. Tumor dimensions were measured with calipers for up to 40 days, and volumes (mm³) were calculated as (0.5 × width) × (height²).

Cells were lysed in RIPA buffer. Protein samples were resolved by SDS-PAGE and then subjected to Western blot analysis with anti-LAMB3 (kindly provided by Dr. M. Peter Marinkovich, Stanford University, School of Medicine, Stanford, CA), anti-actin (Sigma), or anti-FAK (Cell Signaling Technology) antibodies.

NCI-H460 cells and NCI-H522 cells were transfected with miR-Ctrl or miR-1298 mimics, cultured for 5 days, and then fixed with 10% paraformaldehyde. TUNEL [terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate (dUTP) nick-end labeling] staining was performed using a TUNEL kit (Roche). Nuclei were counter-stained with 4′,6-diamidino-2-phenylindole (DAPI). The cells were imaged using a Leica DM13000 immunofluorescent microscope, and TUNEL⁺ cells were enumerated using the software.

Putative miR-1298–binding sites were identified with rna22 and TargetScan (http://www.targetscan.org) pattern-based algorithms. rna22 (https://cm.jefferson.edu/data-tools-downloads/rna22-v2-0/) generates full-length target sites; that is, each predicted heteroduplex extends beyond the corresponding miRNA seed region.

HeLa cells were seeded at 1 × 10⁵ cells per well in 12-well plates 24 hours before transfection. miR-1298 or miR-Ctrl mimics were cotransfected at a final concentration of 25 nmol/L with 200 ng of the psiCHECK-2 vector. This vector constitutively expresses firefly luciferase, allowing transfection efficiency to be controlled. Forty-eight hours after transfection, firefly and Renilla luciferase activities were measured consecutively with the Dual-Luciferase Reporter System (Promega). All luciferase assays were repeated a minimum of 3 times, each with biological duplicates.

For quantitative PCR analyses, total RNA was extracted from cells using TRIzol reagent (Invitrogen) according to the manufacturer's instructions. cDNA synthesis was performed with 500 ng of total RNA and Superscript II, according to the manufacturer's instructions (Bio-Rad). PCR was performed using SYBR Green PCR Master Mix with GAPDH as an internal control. Primer sequences were designed by the Roche Universal Probe Library and are available on request. All amplicons were analyzed using a Roche LightCycler 480 II.

The mammalian expression vectors for LAMB3 and FAK were purchased from Open Biosystems. KRAS^G12V in a pCMV-Myc vector was a gift from Dr. Huaxi Xu (Sanford Burnham Prebys Medical Discovery Institute, La Jolla, CA). cDNAs encoding LAMB3 and FAK were subcloned using pcDNA4.0.

pMIF-cGFP-Zeo vector (SBI), an FIV-based miRNA precursor construct, was used to express miR-1298. The construct consists of the miR-1298 stem-loop structure and 220 bp of upstream and downstream flanking genomic sequences. The mature miR-1298 sequence can be obtained from http://www.mirbase.org/.

psiCHECK-2 vector (Promega) was used for the luciferase assay. For LAMB3-MREs, the sequence corresponding to the rna22-predicted miRNA-MRE (or mutated sequence for LAMB3-MRE-MUT; Supplementary Fig. S4E) were synthesized as sense and antisense oligomers, then annealed and cloned into psiCHECK-2 directly downstream of Renilla luciferase. For FAK, the full 3′ UTR was cloned into psiCHECK-2. The mutant construct was generated by PCR-directed mutagenesis using a primer pair containing the mutant MRE sequence. The PCR product was subsequently used as a template for full-length PCR.

H460 cells or H1437 cells in a 10-cm dish were transfected with 30 nmol/L miR-1298 mimic or miR-Ctrl mimic. After 12 hours, the cells were reseeded into 6-well plates in DMEM plus 10% FBS and incubated for 2 weeks. The cells were then fixed with formalin, stained with 0.5% crystal violet, and the colonies were counted.

miRNA mimics were chemically modified at 2 distinct positions in the antisense/guide strand of miR-1298, as previously described (24). Briefly, psoralen (Pso) was covalently attached to the 2′ position of the sugar ring of a uridine within the miRNA-1298 seed sequence (position 5 from the 5′ end), and a biotin was conjugated to the 3′ end for affinity purification. To ensure a high yield of Pso-modified miRNAs, the coupling reaction was conducted using an excess molar ratio of activated Pso to amine-containing miRNA mimics (2′O protected, Dharmacon) in DMSO in the presence of N,N-diisopropylethylamine in the dark. After 24 hours, excess Pso was removed by precipitating the modified miRNAs with ethyl acetate, and the Pso-functionalized miRNA mimics were analyzed by UV spectroscopy and denaturing PAGE. Subsequent analysis of deprotected single-stranded RNAs by denaturing gel electrophoresis clearly showed single bands with slower electrophoretic mobility compared with the unmodified strand, confirming that the miRNAs were efficiently conjugated to Pso.

For crosslinking experiments, HCT116 cells were seeded in 15-cm dishes at a density of 2 × 10⁶ cells per dish. HCT116 were transfected with 25 nmol/L of the miR-1298-TRAP probe using Lipofectamine 2000 (Invitrogen), according to the manufacturer's protocol. Twenty-four hours later, the culture medium was removed and cells were rinsed three times with PBS, 10 mL of PBS was added, and the cells were irradiated for 5 minutes with UVA (360 nm) in a photochemical reactor (Rayonet, model RPR-100) equipped with 16 RPR-3500 light tubes. Cells were then lysed by addition of 3 mL lysis buffer (20 mmol/L Tris, pH 7.5, 200 mmol/L NaCl, 2.5 mmol/L MgCl₂, 0.5% NP-40, and 80 U RNaseOUT) and shaken on a rocker at 4°C for 5 minutes. Cells were collected by scraping the plates, and cell debris was removed by centrifugation at 13,000 rpm at 4°C for 15 minutes. The supernatant was removed and approximately 1 mL was mixed with streptavidin-conjugated Dynabeads M-280 (Invitrogen) and rotated at 4°C for 4 hours. The tubes were placed on a magnet for 2 minutes, the supernatant was removed, and the beads were washed with three times with 500 μL lysis buffer for 5 minutes each. TRIzol reagent (200 μL) was added to the bead sample and RNA was extracted according to the manufacturer's protocol. Final RNA samples were suspended in 7.5 μL RNase-free water and reverse transcribed using Superscript II (Invitrogen). qPCR was performed using a Roche LightCycler 480 II and SYBR Green master mix (Bio-Rad). qPCR primers were designed by Roche Universal Probe Library, and sequences are available upon request.

Associations between LAMB3, LAMA3, and LAMC2 expression and NSCLC patient survival were based on a published study (25) of gene expression data and survival outcomes from a large cohort of lung cancer patients. Statistical analysis of gene expression and prognosis of patients carrying KRAS^MUT versus KRAS^MUT tumors was as described previously (8). A KRAS gene signature was defined using a dataset of 84 lung adenocarcinomas (26), and subsequently validated with two independent cohorts of lung cancer (27, 28) for which the KRAS mutation status was confirmed. Because the Shedden datasets were generated by four different laboratories, gene expression values within each subset were normalized to SDs from the mean. The Bhattacharjee KRAS signature consisted of “up” and “down” genes. Within each Shedden subset, the average of “up” genes in the signature was compared with the average of “down” genes in the signature. Tumors with higher expression of the “up” or “down” genes (P < 0.01, t test) were classified as positive or negative for the KRAS signature. The signature identified 143 tumors as KRAS signature^+/pos and 116 tumors as wild-type or KRAS signature^–/neg from 442 lung cancer patients (8). The remaining tumors were excluded from subsequent analyses. Within the KRAS signature subgroups, tumors were defined as having high or low expression of LAMB3, LAMA3, or LAMC2 if mRNA levels were higher or lower, respectively, than the median for that subset. Survival was estimated using the Kaplan–Meier method and P values were calculated with the log-rank (Mantel–Cox) test. Microarray data and clinical information on the Shedden samples (25) are available at https://array.nci.nih.gov/caarray/project/jacob-00182.

To identify human miRNAs that are synthetic lethal in cancer cells harboring oncogenic KRAS, we took advantage of an isogenic pair of HCT116 colon carcinoma cell lines expressing either wild-type KRAS (KRAS^WT) or an oncogenic mutant KRAS (KRAS^MUT, in this case KRAS^G13D). These cell lines were kindly provided by Dr. Bert Vogelstein's laboratory. These cells were confirmed to show differential sensitivity to siRNA-mediated KRAS silencing (Supplementary Fig. S1A and S1B). We performed a high-throughput arrayed screen of a human miRNA mimic library (∼858 mimics) to identify miRNAs that selectively kill cells carrying mutant KRAS (Fig. 1A). miRNA mimics are designed to recapitulate the function of endogenous miRNAs but can also be modified to improve uptake and stability. Viability of HCT116 cells was measured with an ATP-based assay 5 days after library transfection, and three miRNA mimics were identified with selective lethality against HCT116 KRAS^G13D cells: hsa-miR-618, hsa-miR-1298, and hsa-miR-512-5p (Fig. 1B). The two selection criteria were that the miRNA mimic had little or no effect on the viability of KRAS^WT cells (a viability score of 0.5–1.0, in which 1.0 indicates no effect) and had a significantly greater effect on the viability score of KRAS^MUT cells than on KRAS^WT cells (≥3-fold). The three hits identified in the screen were validated in more detailed assays using an independent fluorescence-based cell proliferation assay (Fig. 1C). These experiments confirmed the specific and prolonged effects of miR-618, miR-1298, and miR-512-5p mimics on the viability of KRAS^MUT HCT116 cells.

We next asked whether the miRNA mimics identified in the HCT116 screen were also lethal in other KRAS-mutant cell lines, and if so, whether they showed specificity for particular KRAS mutants. For this, we examined a panel of 9 NSCLC lines; 4 carrying KRAS^WT (H1568, H1437, H522, H520) and five harboring KRAS mutations (G12C: H23, H358, H1792; G12S: A549, and Q61H: H460). All lines express wild-type EGFR (Supplementary Fig. S2). Among the three miRNA mimics tested, miR-1298 showed selective lethality on all of the KRAS^MUT cell lines, whereas the effects of miR-618 and miR-512-5p varied with the cell line (Fig. 2A). RT-qPCR analysis showed miR-1298 expression was lower in KRAS-mutant NSCLC cell lines relative to cell lines with wild-type KRAS (Supplementary Fig. S3A). On the basis of these results, we selected miR-1298 for further analysis.

The functional role of miR-1298 was assessed by examining colony formation in vitro and tumor growth in vivo. We found that the miR-1298 mimic inhibited foci formation by H460 (KRAS^MUT) cells but not by KRAS^WT H1437 cells (Fig. 2B), and significant H460 cell death by apoptosis was confirmed by TUNEL staining (Fig. 2C and Supplementary S4). To determine the effect on tumor growth in vivo, miR-1298 or a control miRNA was overexpressed in KRAS^Q61H and KRAS^G12S cell lines (H460 and A549, respectively) using a lentiviral vector and transduced cells were injected subcutaneously into Balb/c nude mice. Tumor growth was then followed for up to 6 weeks. As was observed for in vitro growth, tumors derived from expressing miR-1298 cells grew at significantly slower rates than tumors expressing a control miRNA (Fig. 2D). Thus, miR-1298 impairs the growth of oncogenic KRAS-driven cells both in vitro and in vivo.

miRNAs function by binding to their specific target mRNAs and promoting their degradation or inhibiting their translation. We took a two-pronged approach to identify putative downstream targets of miR-1298. First, we performed microarray analysis to identify differentially expressed genes in HCT116 cells transfected with miR-Ctrl or miR-1298. We then compared the most downregulated genes from the microarray analysis with the results of searches on rna22 (29) or TargetScan (http://www.targetscan.org); these programs use pattern-based predictive algorithms to identify potential miRNA target sites. From these two screens, the top 50 overlapping genes were chosen for further validation (Fig. 3A).

To demonstrate direct interactions between miR-1298 and its target mRNAs in vivo, we turned to a recently reported method miR-TRAP (miRNA target RNA affinity purification; ref. 24). For this, psoralen, a highly photoreactive probe, was conjugated to the 2′ position of the sugar ring of an amine-modified uridine residue 3′ of the miR-1298 seed sequence, and biotin was conjugated to the 3′ end for affinity purification (Fig. 3B). The miR-1298-TRAP probe was transfected into HCT116 cells, and the cells were exposed to UVA (360 nm) to crosslink miR-1298-TRAP to its complementary mRNA. The miRNA–mRNA complexes were then affinity purified from cell lysates, and the isolated RNA was analyzed by RT-qPCR, using primers specific for the 50 genes described above (Fig. 3B and Supplementary Fig. S5). Using this relatively high-throughput assay, we identified 4 mRNAs among the 50 examined that showed significant enrichment in miR-1298–TRAP complexes from UVA-crosslinked cells compared with nonirradiated, but otherwise identically treated, cells. These mRNAs were FAK (focal adhesion kinase), LAMB3 (laminin β3 subunit), PKP4 (plakophilin-4), and WWP1 (E3 ubiquitin ligase; Fig. 3C).

To determine whether one or more of the four miR-1298 targets contributed to the synthetic lethality of miR-1298 in KRAS^MUT cell lines, we knocked down FAK, LAMB3, PKP4, and WWP1 using SMARTpool siRNAs. Downregulation of both FAK and LAMB3 impaired the viability of KRAS^MUT cells, whereas knockdown of WWP1 and PKP4 was without effect (Fig. 4A). PLK1 (8) and GATA2 (9) were also examined because both have been identified to have synthetic lethal interactions with oncogenic KRAS in NSCLC, which we confirmed here (Fig. 4A). To rule out possible off-target effects of the LAMB3 and FAK siRNAs, we tested four siRNAs against each gene, and found that three of the four siRNAs effectively killed KRAS^MUT cells (Supplementary Fig. S6A)

Importantly, the cytotoxic effects of siFAK and siLAMB3 on KRAS^MUT lines correlated strongly with the effects of miR-1298 mimic (Fig. 4B and C), consistent with FAK and LAMB3 being effectors of miR-1298 in these cells. Moreover, the FAK dependence of KRAS^MUT cell lines was confirmed by treatment with a small-molecule inhibitor of FAK, VS-6063, which potently reduced KRAS^MUT cell viability (Fig. 4D).

To confirm that miR-1298 regulates the expression of FAK and LAMB3, we first examined their mRNA levels in miRNA-transfected cells. Indeed, both FAK and LAMB3 mRNA levels were significantly and selectively decreased in KRAS^MUT cell lines by miR-1298 (Supplementary Fig. S6C and S6D). Next, using predictive algorithms, we determined the putative miR-1298–binding sites on FAK and LAMB3 mRNA using TargetScan and rna22, respectively. On the basis of the sequences predicted, we cloned the full-length 3′UTR region of FAK and the miRNA response element (MRE; ref. 30) in the coding region of LAMB3 into the psiCHECK2 luciferase reporter plasmid, upstream of the Renilla luciferase gene (Supplementary Fig. S6E). HeLa cells cotransfected with the LAMB3 or FAK reporter plasmid together with the miR-1298 mimic showed specific reductions in the ratio of Renilla luciferase to the constitutively expressed firefly luciferase, indicating that miR-1298 specifically recognizes, binds, and directs the degradation of both targets (Supplementary Fig. S6F). miR-1298 recognition and cleavage was confirmed to be sequence specific by demonstrating that mutants of these targeted sequences rescued Renilla luciferase activity (Supplementary Fig. S6F).

Finally, we asked whether overexpression of FAK and/or LAMB3 could prevent death of KRAS^MUT lines induced by the miR-1298 mimic. Interestingly, the viability of miR-1298 mimic–transfected KRAS^MUT lines was not significantly affected by overexpression of either LAMB3 or FAK alone, but coexpression of both proteins fully rescued the lethal effects of miR-1298 in multiple KRAS^MUT lines (Fig. 5A). These results demonstrate that, although knockdown of either gene mimics the effect of miR-1298, both proteins are necessary to prevent miR-1298-induced death.

To determine how LAMB3 and FAK intersect with the oncogenic KRAS pathway, we examined their mRNA and protein expression after knocking down KRAS in KRAS^MUT lines and overexpressing KRAS^G12V in KRAS^WT cells. Interestingly, LAMB3 mRNA levels were markedly reduced by siKRAS treatment of KRAS^MUT lines (Fig. 5B) and dramatically increased in KRAS^WT cells overexpressing oncogenic KRAS^G12V (Fig. 5D). LAMB3 protein expression was similarly affected (Fig. 5C and E). Surprisingly, expression of FAK mRNA (Supplementary Fig. S7) and protein (Fig. 5C and E) was entirely unaffected by the same treatments. We also found that knockdown and overexpression of mutant KRAS similarly did not affect the expression of miR-1298 (Supplementary Fig. S3B and S3C). These findings suggest that LAMB3 is a downstream transcriptional target of KRAS, but that miR-1298 and FAK are not directly affected. On the other hand, FAK does not appear to have direct interactions within the KRAS pathway but mutant KRAS cells are highly sensitized to changes in FAK expression. This highlights the utility of unbiased functional synthetic lethal genomic screens in identifying interactions that could not be predicted on the basis of differential expression.

To determine the clinical significance of our observations, we analyzed gene expression data from a cohort of 259 lung cancer patients for whom the survival outcome and KRAS transcriptional signature status are known (25). The 143 KRAS signature^pos patients and 116 KRAS signature^neg patients were stratified according to LAMB3 and FAK mRNA expression, with high and low expression classified as greater than or less than the median expression levels, respectively. Survival outcome was then analyzed using a previously established method (8). Interestingly, high LAMB3 expression was significantly associated with poorer overall survival of patients harboring KRAS signature^pos tumors but not those with KRAS signature^neg tumors (Fig. 6A), whereas FAK expression was without prognostic value (Fig. 6B), which was also seen in another study of prognosis in early-stage NSCLC (31). Finally, because LAMB3 is one subunit of the laminin 332 heterotrimer, we also examined the possible involvement of the remaining subunits, LAMA3 and LAMC2. Interestingly, siRNA-mediated knockdown of either protein had no effect on the viability of KRAS^WT or KRAS^MUT cell lines (Supplementary Fig. S8A) and there was no detectable correlation between expression of LAMA3 or LAMC2 and the prognosis of patients with KRAS gene signature^pos NSCLC (Supplementary Fig. S8B). Collectively, these findings suggest that differences in LAMB3 expression are clinically relevant in KRAS^MUT lung cancers, and thus LAMB3 may be useful not only as a potential therapeutic target but also as a prognostic biomarker. In contrast with LAMB3, FAK expression is not affected by mutant KRAS. As such, the clinical outcome in KRAS signature^pos NSCLC is LAMB3 dependent but not FAK dependent.

KRAS-mutant cancers are among the most difficult to treat with poor survival and resistance to chemotherapy and newer targeted therapies and agents to target KRAS-mutant cancers remain an unmet clinical need in colorectal cancer and NSCLC. In this study, we performed a large-scale synthetic lethal screen in an isogenic colorectal cancer cell line pair to identify miRNAs that are selectively lethal for KRAS^MUT cells and confirmed these miRNAs in a panel of cell lines harboring different oncogenic KRAS mutations, and then used miR-TRAP approach to identify functional miRNA targets. Collectively, these experiments identified miR-1298 and its downstream targets LAMB3 and FAK as a potentially targetable regulatory axis in KRAS-mutated tumor cells (Fig. 6C). Furthermore, the clinical relevance of these findings was revealed by demonstrating a significant correlation between LAMB3 expression and survival specifically in lung cancer patients positive for a KRAS gene expression signature.

LAMA3, LAMB3, and LAMC2 encode the of α3, β2, and γ2 subunits, respectively, of the trimeric basement membrane protein laminin 332 (32). Laminin 332 has been shown to promote tumorigenesis in squamous cell carcinoma through interactions with cell surface receptors such as integrins, EGFR, and syndecan 1 (32, 33). In addition, human squamous cell carcinoma induced by HRAS^V12 and IκBα can be inhibited by blocking antibodies against laminin 332 or β4 integrins (34). In models of human epidermal carcinogenesis, LAMB3 activates PI3K through interactions with collagen VII (35). LAMB3 was also identified in a large siRNA synthetic lethal screen in similar isogenic KRAS colorectal cancer cell lines (11) further validating our synthetic lethal miRNA screen. We also demonstrated that LAMB3 expression is affected by KRAS oncogenic mutations and LAMB3 expression level correlated with survival. The association between KRAS mutations and LAMB3 expression raises the possibility of LAMB3 as a prognostic biomarker in KRAS-mutated NSCLC. Given our results, LAMB3 also may serve as a biomarker to identify patients in whom targeting the miR-1298/FAK/LAMB3 axis might have clinical benefit.

The second specific miR-1298 target mRNA we identified is FAK (PTK2), a tyrosine kinase component of focal adhesion complexes that has been implicated as a crucial mediator at the intersection of many different signaling pathways with diverse effects in cancer, including promoting tumor progression and metastasis, the maintenance of cancer stem cells, and modulation the tumor microenvironment and angiogenesis (36). Binding and phosphorylation of FAK tyrosine 925 by Src family kinases creates a Grb2 SH2-domain binding site and provides a link to activation of the RAS signal transduction pathway (37). Furthermore, previous studies have shown that RhoA–FAK signaling is required for maintenance of KRAS-driven adenocarcinomas (38), and defactinib (VS-6063) a FAK small-molecule inhibitor is currently in phase II clinical trial in KRAS-mutant NSCLC. Our data highlight the significance of FAK not just in RhoA signaling, but implicates FAK in oncogenic KRAS-driven cancers and suggests that FAK small-molecule inhibitors may be used to treat KRAS-driven cancers more generally than previously envisioned. Given the current complexity in terms of diverse mechanisms in FAK on human cancers, our finding of a novel regulatory axis through miR-1298 on the expression of FAK and LAMB3 as a potential vulnerability in KRAS-driven cancers suggests future mechanistic studies focusing on the interaction of integrins with FAK and RAS signaling may provide further insight toward targeting this signaling axis. Also, given the frequent use of anti-angiogenesis therapies in NSCLC and colorectal cancer, future studies on the association between FAK and angiogenesis may also provide further insight into potential novel therapies for KRAS-mutated cancers.

In summary, our study has revealed a novel role for miR-1298 in the survival of KRAS^MUT tumor cells and identified two target genes, LAMB3 and FAK, with potential therapeutic value for KRAS-driven cancers. Our work emphasizes the utility of miRNAs as tools with which to probe complex biological processes and to identify novel therapeutic targets.

No potential conflicts of interest were disclosed.

Conception and design: Y. Zhou, J. Moscat, T.M. Rana

Development of methodology: Y. Zhou, P. Aza-Blanc, T.M. Rana

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Zhou, J. Dang, P. Aza-Blanc

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Zhou, K.-Y. Chang, E. Yau, P. Aza-Blanc, T.M. Rana

Writing, review, and/or revision of the manuscript: Y. Zhou, J. Dang, K.-Y. Chang, E. Yau, J. Moscat, T.M. Rana

We are grateful to Dr. Bert Vogelstein for HCT116 and RasWT1 (HAF1) cell lines; Dr. Baigude for help in miR-TRAP experiments, Dr. Peter Marinkovich for laminin antibodies; Chad J. Creighton, Ji Luo, and Stephen J. Elledge for sharing the KRAS signature genes, and David G. Beer, Kevin K. Dobbin, and Guoan Chen for their generous assistance in mapping the sample IDs of the Shedden dataset.

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.