Selective targeting of point-mutated KRAS throughartificial microRNAsMario Acunzoa,b,1,2, Giulia Romanoa,b,1, Giovanni Nigitaa, Dario Venezianoa, Luigi Fattorec, Alessandro Laganàd,Nicola Zanesia, Paolo Faddaa, Matteo Fassane, Lara Rizzottof, Raleigh Kladneya,g, Vincenzo Coppolaa,and Carlo M. Crocea,2

aDepartment of Cancer Biology and Genetics, College of Medicine, The Ohio State University, Columbus, OH 43210; bDivision of Pulmonary Disease andCritical Care Medicine, School of Medicine, Virginia Commonwealth University, Richmond, VA 23298; cIstituto Nazionale per lo Studio e la Cura dei Tumori“Fondazione G. Pascale”, Naples 80131, Italy; dDepartment of Genetics, Genomic Sciences Icahn School of Medicine at Mount Sinai, New York, NY 10029;eDepartment of General Anatomic Pathology, University of Padua, 35100 Padova, Italy; fDepartment of Internal Medicine, Comprehensive CancerCenter, The Ohio State University, Columbus, OH 43210; and gDepartment of Molecular Genetics, College of Biological Sciences, The Ohio StateUniversity, Columbus, OH 43210

Contributed by Carlo M. Croce, March 22, 2017 (sent for review December 16, 2016; reviewed by Karen E. Knudsen and Frank McCormick)

Mutated protein-coding genes drive the molecular pathogenesisof many diseases, including cancer. Specifically, mutated KRAS is adocumented driver for malignant transformation, occurring earlyduring the pathogenesis of cancers such as lung and pancreaticadenocarcinomas. Therapeutically, the indiscriminate targeting ofwild-type and point-mutated transcripts represents an importantlimitation. Here, we leveraged on the design of miRNA-like artificialmolecules (amiRNAs) to specifically target point-mutated genes,such as KRAS, without affecting their wild-type counterparts.Compared with an siRNA-like approach, the requirement ofperfect complementarity of the microRNA seed region to a giventarget sequence in the microRNA/target model has proven to bea more efficient strategy, accomplishing the selective targetingof point-mutated KRAS in vitro and in vivo.

artificial microRNA | KRAS | RNAi

Single point mutations, categorized according to their abilityto affect the structure and function of the translated protein,

are involved in numerous diseases, including cancer (1). Specif-ically, Kras, a GTPase protein, is frequently mutated in codon12 of its coding sequence (CDS) in large fractions of specifichuman cancer types (2, 3). For example, the G12S mutation, inwhich a single nucleotide modification leads to the substitutionof a glycine (Gly) for a serine (Ser), confers constitutive activa-tion of KRAS, inducing cell proliferation (4–6). Additionally, theG12S mutation induces resistance to all drugs targeting upstreamof any pathway in which KRAS is involved (i.e., gefitinib) (7).Several attempts to target KRAS-mutant cancers have usedsmall noncoding RNA (sncRNA), in particular, small-interferingRNA (siRNA) targeting KRAS pathways (8). Given the intrinsiclimitations of siRNA targeting, there are not effective siRNA-based strategies to specifically target KRAS mutations withoutaffecting KRAS WT (9), and such indiscriminate targeting islikely to be toxic (10–12). Alternatively, use of microRNAs(miRNAs), a class of endogenous 22- to 24-bp-long sncRNAsinvolved in multiple cellular processes, pathways, and diseases, in-cluding cancer (13), could provide a more tailored targeted therapydesign. Generally, miRNAs target the 3′ untranslated region (UTR)of mRNA transcripts, despite having also been proven to bind theCDS, as well as the 5′ UTR (14) like siRNAs, causing translationinhibition and/or inducing mRNA degradation (15). Neverthe-less, unlike siRNAs, miRNAs exhibit only partial complemen-tarity to the binding sequence on their targets. The efficacy of theinteraction in this partial binding is based on a 6- to 8-nucleotide(nt)-long region on the 5′ end of the miRNA sequence termed the“seed” sequence. A single mismatch in the seed region (specificallyin positions 2–7) can dramatically compromise the interaction be-tween an miRNA and its target (16, 17). Furthermore, the commonpresence of a central bulge in the binding interaction supports theimportance of a perfect seed match to assure the stability of the

duplex (18). Leveraging these important traits, we designed a set ofartificial miRNAs (amiRs) (19) with seed regions that perfectlymatch an 8-nt sequence encompassing the single point-mutation inthe CDS of the KRAS (G12S and G12D), as well as BRAF(V600D) mRNAs, while imperfectly matching KRAS WT andBRAF WT transcripts. We have developed an miRNA-basedmethodology for the selective targeting of single point-mutatedgene transcripts. Here, we demonstrate the impact of an amiRon pathways affected by KRASG12S in non-small-cell lung cancer(NSCLC), while also investigating the possibility of off-targetbiological effects.

ResultsDesign of amiRs for Selective mRNA Targeting. To generate a set ofartificial RNA-based molecules for selective targeting of specificpoint-mutated KRAS mRNAs, while sparing WT counterparts,we elected to use an amiR-based design, rather than an siRNA-based one, leveraging the innate miRNA seed-mediated targetrecognition (16, 17). We had reported the effectiveness of amiRsfor gene silencing (19). The rationale for an amiR-based approachstems from the ability of this type of molecule to better discriminatebetween two targets that differ only by a single base in its seed targetregion (SI Appendix, Fig. S1A), as opposed to the use of siRNAs,whose function may not be greatly affected by a single mismatch (SIAppendix, Fig. S1B). We thus decided to test our method by de-signing a set of 22-bp-long amiRs to target mutated KRAS, one ofthe best-investigated oncogenes in human cancer. We equipped ouramiRs with seed sequences perfectly matching the region of KRAS

Significance

Recently, small interfering RNAs have been used to specifi-cally target point-mutated KRAS, yet without sufficientlydiscriminating its wild-type counterpart. Here, we describe aninnovative approach based on the development of artificialmicroRNAs able to efficiently target mutated KRAS, leavingtheir normal counterpart unaffected and preventing majorside effects.

Author contributions: M.A., G.R., and C.M.C. designed research; M.A., G.R., L.F., P.F., M.F.,L.R., R.K., and V.C. performed research; G.N., D.V., A.L., and N.Z. contributed new re-agents/analytic tools; G.N., D.V., and N.Z. analyzed data; and M.A., G.R., and D.V. wrotethe paper.

Reviewers: K.E.K., Kimmel Cancer Center, Thomas Jefferson University; and F.M., Univer-sity of California, San Francisco.

The authors declare no conflict of interest.1M.A. and G.R. contributed equally to this work.2To whom correspondence may be addressed. Email: carlo.croce@osumc.edu or mario.acunzo@vcuhealth.org.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1620562114/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1620562114 PNAS | Published online May 8, 2017 | E4203–E4212

GEN

ETICS

PNASPL

US

mRNA, including the G-to-A point mutation (14) at codon 12(G12S) and, as a consequence, mismatching for the same region inWT KRAS. Specifically, the amiRs were designed such that theseed mismatch on KRAS WT would fall on different seed bases,between and including nucleotide positions 2 and 7. The remainingportions of the amiRs were designed to favor a binding site with theKRAS target characterized by both a central bulge preceded by theseed region as well as a single 3′ supplementary site (20) thatconfers a greater specificity (21) (Fig. 1A). The WT and mutated(G12S) KRAS portions of the coding region were, respectively,cloned into a PGL3 control reporter vector and tested for luciferaseactivity, after cotransfection with the amiR-KS set. Each of thedesigned amiRs was superior in suppressing KRAS G12S expres-sion vs. KRAS WT (Fig. 1B), thus showing that seed-driven designcan discriminate between mutant and WT, with the exact positionof the mismatch within the seed region as almost irrelevant fortargeting efficiency.Among the amiRs from the KRAS G12S (KS) set, we selected

amiR-KS3 for its strong ability to most significantly repress

KRAS G12S, while not affecting KRAS WT expression (Fig. 1Band SI Appendix, Fig. S2A). To prove the hypothesis that only amolecule with miRNA characteristics (seed region followed bycentral bulge) can discriminate the two forms of KRAS (point-mutated and WT), we created an siRNA with a sequence iden-tical to amiR-KS3, except for the bulge, thus displaying perfecttarget complementarity, in compliance with siRNA structure andfunction (22) (Fig. 1C). siRNA- and amiR-KS3 were, respectively,tested on KRAS WT and KRAS G12S by luciferase assay to verifytheir effects on the two distinct KRAS forms. The siRNA-KS3 couldnot discriminate between KRAS WT and KRAS G12S, causingstrong repression of both. On the contrary, amiR-KS3 was con-firmed to selectively target KRAS G12S (Fig. 1D).We then tested the effect of amiR-KS3 in both KRAS WT and

KRAS mutated cancer cell lines. For this purpose, we usedH1299 and H292 NSCLC cell lines that express KRAS WT, aswell as A549, a NSCLC cell line that homozygously expressesendogenous KRAS G12S (Fig. 1 E–G, Left). amiR- and siRNA-KS3 were, respectively, transfected in all three cell lines, and

Fig. 1. amiRs specifically targeting KRAS G12S. (A) Representation of the amiR KS set designed for KRAS G12S targeting (six sequences) and correspondingmatch with the KRAS binding site in the CDS including the G12S mutation (Left) and the WT version (Right), respectively. (B) Graph of luciferase activitypercentage for amiR KS set sequences cotransfected with the KRAS WT (green) and KRAS G12S (orange) CDS portions, respectively, cloned downstream of theluciferase reporter gene and including the corresponding amiR KS binding site. (C) Representation of amiR- and siRNA-KS3 matching the same binding site onthe KRAS WT and G12S CDS, respectively. (D) Graph of luciferase activity percentage for amiR- and siRNA-KS3 sequences cotransfected with the KRAS WT(green) and KRAS G12S (orange) CDS portions, respectively. (E–G) Chromatogram of sequenced KRAS endogenously expressed in H1299 and H292 cell lines(blue dot) and A549 cell line (orange dot), respectively, all compared with the WT KRAS reference sequence (green dot) (Left); Western blots show theexpression of endogenous KRAS in all cell lines after the transfection of amiR- and siRNA-KS3, respectively (Right). *P < 0.05; **P < 0.01; ***P < 0.001. Scr,scramble; seq., sequence.

E4204 | www.pnas.org/cgi/doi/10.1073/pnas.1620562114 Acunzo et al.

KRAS levels were assessed by Western blot. Our results showthat amiR-KS3 targeted the mutant form of KRAS, but not itsWT, as opposed to the siRNA-KS3, which did not distinguishbetween the two KRAS versions (Fig. 1 E–G, Right). The effectof the amiR-KS set on endogenous KRAS G12S and WT mRNAexpression on H1299 and A549 cell lines was further validated byRT-PCR, and amiR-KS set expression was confirmed by usingcustom probes (SI Appendix, Fig. S2 B–E). The vast majority ofthe KS-set amiRs suppressed KRAS G12S in A549, while notdown-regulating KRAS WT in H1299.To further test the efficacy of our approach, we designed two

additional sets of amiRs, an amiR-KD set for the KRAS G-to-Apoint mutation in codon 12 (G12D) and an amiR-VD set forthe BRAF GU-to-UA double mutation in codon 600 (V600D),comprising 6 and 5 amiR sequences, respectively, targetingmutated versions of KRAS or BRAF mRNAs, and evaluatedtheir efficacy by luciferase assays, as described for KRAS G12S.The sequences of these amiRs and their binding sites on thetarget sequences are shown in SI Appendix, Fig. S3A for KRASG12D and in SI Appendix, Fig. S4A for BRAF V600D. Weperformed a luciferase assay using reporter plasmids (KRASWTand G12D and BRAF WT and BRAF V600D coding regions,respectively) cotransfected with each amiR from each set. Asshown in SI Appendix, Fig. S3B and S4B, the amiRs designed totarget KRAS G12D and BRAF V600D, respectively, repressedthe luciferase activity of the mutated versions of KRAS andBRAF better than they repressed the respective WT versions.We also validated the KD set of amiRs by transfecting eachamiR into pancreatic carcinoma cell lines BX-PC3 (expressingKRASWT) and PC1 (expressing only KRAS G12D-mutant) andtested for endogenous Kras expression levels by Western blot (SIAppendix, Fig. S3C). Similarly, we transfected the amiR-VD setinto melanoma cell lines SK-Mel2 (expressing BRAF WT) andWM266 (expressing only BRAF V600D mutant) and verified theendogenous Braf expression levels (SI Appendix, Fig. S4C). Inboth cases, a number of the designed amiRs were able to effectivelydown-regulate the point-mutated versions of the Kras and Brafendogenous proteins, while not down-regulating their WTcounterparts.

amiR-KS3 Inhibits KRAS G12S Activity. The mutant forms of KRASdrive the induction of the cell cycle by activating the ERK andPI3K pathways (11, 23, 24). We first evaluated the effects ofamiR-KS3 in the A549/KRAS G12S cell line by comparing theexpression of 729 genes of the nCounter PanCancer PathwaysPanel, in biological triplicates, with effects of a commercialKRAS siRNA (SI Appendix, Fig. S5 A–C). KRAS was signifi-cantly down-regulated in both cases. We performed a functionalenrichment analysis of the significantly deregulated genes byusing the Ingenuity Pathway Analysis (IPA) software and fo-cused our attention on the ERK/MAPK signaling and cell-cycleG1/S checkpoint regulation pathways, strictly linked to the KRASpathway activity in lung cancer (24, 25). Our results show how bothmolecules (siRNA-KRAS and amiR-KS3) were significantly in-volved in these two pathways (SI Appendix, Fig. S5D).We elected to validate our enrichment predictions by trans-

fecting amiR- and siRNA-KS3, respectively, evaluating the phos-phorylation levels of Erk1/2 and Akt in the A549, H292, andH1299 NSCLC cell lines. In Fig. 2A, we show that only amiR-KS3 resulted in down-regulation of the KRAS G12S effect on itsdownstream effectors, whereas the siRNA-KS3 acted indiscrim-inately on KRAS WT (H1299 and H292) and KRAS G12S(A549). Additionally, siRNA KS3 indiscriminately reduced cellviability in KRAS WT and KRAS mutant-expressing cell lines (SIAppendix, Fig. S6A).The H1299, H292, and A549 (KRAS G12S) cell lines were

also transfected with amiR-KS3 and evaluated for cell growthdensity (Fig. 2B), cell viability (Fig. 2C), and cell proliferation

(Fig. 2D) time course assays. We showed that amiR-KS3 causedreduction of all three functions only in KRAS-mutated A549cells. Finally, we performed a cell-cycle study demonstrating thatamiR-KS3 maintained KRAS-mutant cells in the G1/G0 phase(Fig. 2E and SI Appendix, Fig. S6B). The results thus show theability of amiR-KS3 to affect the cell cycle only in A549 cells,whereas no significant effect was recorded in the H1299 andH292 KRAS WT cell lines.Additionally, several studies have shown that mutated KRAS

induces tumor cell migration through the activation of theMAPKs and PI3K/AKT pathways (26, 27). We assessed the ef-fects of amiR-KS3 on cell migration in both KRAS WT andmutated NSCLC cell lines. As shown in Fig. 3 A and B by woundhealing assay, amiR-KS3 led to reduced migration of A549(KRAS G12S), whereas no effect was noted in H1299 and H292(KRAS WT). We also conducted a transwell migration assay inall three cell lines to reinforce our results (Fig. 3C).Finally, we tested amiR-KS3 effects in vivo. Through intra-

tumoral injection of A549 (KRAS G12S) cells, we inducedoverexpression of amiR-KS3, which resulted in marked in-hibition of tumor growth and cell proliferation, as assessed by theKi-67 proliferative index, starting at 12 d of treatment (Fig. 3 Dand E). A significant enhanced tumor necrosis was also noted inthe amiR-KS3–treated group (Fig. 3F). We confirmed the levelsof amiR-KS3 in the xenograft tumors by quantitative RT-PCR(RT-qPCR) at day 17 (SI Appendix, Fig. S7A). Together, ourresults clearly show that our amiR design strategy allows develop-ment of molecules to effectively target KRAS G12S, perturbing itsbiological effects in vitro and in vivo, without significantly affectingKRAS WT.

amiR-KS3 Induces Gefitinib Sensitivity. Gefitinib acts on the EGFreceptor by inhibiting its pathway activation, thereby reducingproliferation and increasing apoptosis (28, 29). A549 cells aregefitinib-resistant because of the KRAS G12S mutation (30–32)(Fig. 4A). We thus tested the ability of amiR-KS3 to sensitizeA549 cells to gefitinib treatment. We overexpressed amiR-KS3in A549 cells and assessed the percentage of apoptotic cellsthrough the Annexin V-5 assay. As shown in Fig. 4B, A549 cellstreated with amiR-KS3 and gefitinib (5 μM) showed an increasedapoptotic fraction vs. scrambled control, amiR-KS3/DMSO, orscrambled control/gefitinib (Fig. 4C). We also performed a timecourse treatment with gefitinib, evaluating apoptosis throughcaspase 3/7 activity (Fig. 4D) and Trypan Blue assays (Fig. 4E),confirming the Annexin V-5 results. In conclusion, the resultsconfirm the ability of amiR-KS3 to induce gefitinib sensitivity inA549 NSCLC cell lines by targeting KRAS G12S.

amiR-KS3 Does Not Present Off-Targets. Evidence shows that en-dogenous miRNAs possess multiple targets (33). It has also beenproven that the presence of one or more perfect matches in 3′UTR sequences of unintended targets can lead to considerableoff-target effects, as with siRNAs when acting as miRNAs (34,35). Because our artificial RNA molecules were designed to actas endogenous miRNAs, possessing a central bulge, they alsodisplay extensive complementarity to the corresponding targetsequence, like siRNAs, thus potentially reducing off-targeting.Considering the “hybrid” nature of our molecules, we thus de-cided to evaluate the targeting specificity of amiR-KS3, esti-mating its potential off-targeting effects. To accomplish this, weagain transfected amiR-KS3 into H1299 and H292 (KRAS WT)cell lines, respectively, to evaluate the expression of 729 genes ofthe nCounter PanCancer Pathways Panel, in biological triplicates(SI Appendix, Fig. S7B). Because of the general effects of en-dogenous miRNAs in gene regulation, we only considered thedown-regulated transcripts in both experiments, as well as in theprevious transfection in A549 cells, obtaining 65 genes in A549,25 in H1299, and 68 in H292 as potential targets of amiR-KS3.

Acunzo et al. PNAS | Published online May 8, 2017 | E4205

GEN

ETICS

PNASPL

US

Remarkably, KRAS results down-regulated only in A549 (KRASG12S) (Fig. 5A). All down-regulated transcripts from all threeexperiments were then intersected together as shown in Fig. 5Band SI Appendix, Fig. S7C. We then performed a targeting pre-diction consensus using the miRNA target prediction tools mi-Randa (36), PITA (37), and miRiam (38) for amiR-KS3 on all 5′UTR, CDS, and 3′ UTR human transcript sequences, obtainingthree distinct sets of putative direct targets (Dataset S1). Each ofthese three sets was then intersected with each of the three setsof significantly down-regulated genes. As shown in Fig. 5B and SIAppendix, Fig. S7C, only very few transcripts were shared amongat least two of the three sets of significantly down-regulatedgenes. Moreover, only three of these transcripts resulted as puta-tive direct targets (Dataset S2), indicating the absence of a diffusedoff-targeting phenomena.To more formally evaluate the off-targeting effects in an un-

biased manner, we bioinformatically evaluated the statistical en-richment of binding sites for amiR-KS3 across the transcriptome ofinvestigated cell types. We thus used the Sylamer software (39) on

the list of transcripts of both panels (Dataset S3), as well as on thesame panel previously run for amiR-KS3 transfection in the A549KRAS G12S cell line. Specifically, Sylamer was used to determinethe hypergeometric statistical enrichment of three 6-bp 5′ UTR,CDS, and 3′ UTR sequences complementary to the 6-mer amiR-KS3 canonical, as well as two offset seeds (33), across the threetranscript lists, each ranked from most down-regulated to most up-regulated compared with a control scrambled transfection (Fig. 5Cand SI Appendix, Fig. S8). Results show that in all three cell lines (in5′ UTR, CDS, and 3′ UTR), there was no occurrence bias forany of the three 6-mer amiR-KS3 seeds, thus showing no off-target signals from the expression data.

DiscussionThe development of molecules to specifically target oncogeneshas in recent years shifted to the use of noncoding RNAs, such assiRNA and miRNA (40). Our laboratory has previously de-veloped an approach for the design of amiRs, thus leveraging on theirnatural targeting for the down-regulation of multiple transcripts of

0

10

20

30

0h 48h 72h 96hH12

99 (K

RA

S W

T) C

ells

104

amiR KS3

Scr

0

25

50

75

100

48h 72h

H12

99 (K

RA

S W

T)

ytilibaivllecfo% H29

2 (K

RA

S W

T)%

of c

ell v

iabi

lity

0

25

50

75

100

48h 72h

SG2/M G1/G0

0

20

40

60

80

100

H1299(KRAS WT)

H292(KRAS WT)

A549(KRAS G12S)

A

C

D

E

0

25

50

75

100

48h 72h

A54

9 (K

RA

S G

12S)

%

of c

ell v

iabi

lity

*** **

B

aeraecnerefe

R

aeradennac

S

H1299 (KRAS WT)

48h

Scr amiR-KS3

72h

0

25

50

75

100

48h 72h

H12

99 (K

RA

S WT)

ytisned

slleC

48h

Scr amiR-KS3

72h

H292 (KRAS WT)

H29

2 (K

RA

S WT )

C

ells

den

sity

0

20

40

60

80

100

48h 72h

amiR

-KS

3S

cr

A549 (KRAS G12S)

48h

Scr amiR-KS3

72h

0

25

50

75

100

48h 72h

A54

9(K

RA

S G

12S)

%

of C

ells

den

sity

*** ***

0

10

20

30

0h 48h 72h 96hA54

9 (K

RA

SG12

S ) C

ells

104

amiR KS3Scr

***

**

A549 (KRAS G12S)

pErk1/2

Erk 1/2

Akt

pAkt

H292 (KRAS WT)

Erk 1/2

pErk1/2

Akt

pAkt

H1299 (KRAS WT)

pErk1/2

Erk 1/2

Akt

pAkt

0

10

20

30

0h 48h 72h 96hH29

2 (K

RA

S W

T) C

ells

104

amiR KS3Scr

amiR-KS3Scr

% V

alue

s

Fig. 2. amiR-KS3 specifically reduces the proliferation of the A549 G12S mutated cell line. (A) H1299 and H292 expressing KRAS WT and A549 expressingKRAS G12S transfected with amiR- and siRNA-KS3, respectively; phospho-AKT and -Erk1/2 levels detected by Western blot in all three cell lines, with total AKTand total ERK1/2 measured by Western blot as expression control. (B) Bar graphs of cell-density growth assays (Lower) measured after 48 and 72 h fromtransfection of amiR-KS3 and scramble control in H1299 and H292 KRAS WT, as well as in A549 KRAS G12S cell lines. (B, Upper) Representative image of thecell density assays for each cell line. (C) Cell viability assay after 48 and 72 h from transfection of amiR-KS3 and scramble control in H1299 and H292 KRAS WT,as well as in A549 KRAS G12S cell lines. (D) Graph of cellular growth assay after 48, 72, and 96 h from transfection of amiR-KS3 and scramble control inH1299 and H292 KRAS WT, as well as in A549 KRAS G12S cell lines. (E) Flow cytometry analysis using propidium iodide (PI) after 48 h from transfection ofamiR-KS3 and scramble control in H1299 and H292 KRAS WT, as well as in A549 KRAS G12S cell lines. *P < 0.05; **P < 0.01; ***P < 0.001. Scr, scramble.

E4206 | www.pnas.org/cgi/doi/10.1073/pnas.1620562114 Acunzo et al.

choice in multiple biological pathways (24). In the presentwork, we provide a design methodology for a class of amiRsable to target a point-mutated transcript of choice, discriminatingit from its WT version specifically in the location of the seed in-teraction within the binding site.KRAS targeting represents a particularly important opportu-

nity for impacting cancer therapy, given that KRAS mutationsare involved in the development of several types of cancers (3).Attempts at targeting point-mutated KRAS have consisted ofsiRNAs directly targeting this oncogene and associated pathways.However, there are currently no methodologies for the efficienttargeting of single point-mutated transcripts, discriminating themfrom their WT form. To accomplish this goal, we demonstrated thatan miRNA-like design is necessary, as opposed to an siRNA-likeone, to specifically discriminate the point-mutated form from theWT, allowing such methodology to be applied in the case of otherimportant point-mutated transcripts. Despite our miRNA-like de-sign results being diversely effective compared with an siRNA one,

our approach could replace such siRNA-oriented strategiesfor specific point-mutated gene repression in light of the greaterspecificity it provides. For this reason, we have designed such mol-ecules in a very specific manner, assuring an exact matching in theseed segment, as well as at the 3′ end of the binding region (as withsiRNA), yet with a central bulge. Such an approach further provesthe necessity for perfect complementarity of the seed region withthe target sequence (as in endogenous miRNAs).Having tested our amiRs on a mutation of KRAS, G12S, we

have identified a best-performing candidate to verify its efficacyin inhibiting the cancerous pathways associated with the G12Smutation. We, thus, demonstrate, in vitro and in vivo, that ourdesign creates molecules that can neutralize specific pathwayscharacterized by point-mutated oncogenes, inhibiting proliferationand migration and inducing necrosis, while also providing proof ofeffectiveness in the ability to sensitize gefitinib-resistant cell linesto gefitinib.

Fig. 3. amiR-KS3 effects on A549 KRAS G12S mutated cell line migration, as well as in vivo proliferation and necrosis. (A) “Wound healing” migration assaysmeasured after 8 and 24 h from transfection of amiR-KS3 and scramble control in H1299, H292, and A549 cell lines. (B) Wound repair percentage graphsrelative to A. (C) Bar graphs (Lower) of transwell migration assays (Upper) measured after 24 h from transfection of amiR-KS3 and scramble control in inH1299, H292, and A549 cell lines. (D, Left) Mouse xenograft A549 injection protocol. (D, Center) Tumor sizes for two representative mice, respectively pickedfrom the scramble and amiR-KS3 groups at the end of a 17-d treatment. (D, Right) Tumor volume increase graph after 17 d of treatment for two groups of10 mice each (5 male and 5 female), scramble and amiR-KS3 intratumoral-injected, respectively. (E and F) Percentage of Ki-67 proliferation marker-positivetumor cells (E) and graph representing the mean percentage of necrotic areas on the whole tumor area (F), for xenograft mice tumors after 17 d of scrambleand amiR-KS3 treatment, respectively (Right). (E and F, Left) Representative image of Ki-67 expression (immunohistochemistry) (E) and of H&E stain (F) inscramble and amiR-KS3–treated xenograft tumors. The necrosis border is represented by a black dotted line. *P < 0.05; **P < 0.01; ***P < 0.001. Migr.,migrated; Scr, scramble.

Acunzo et al. PNAS | Published online May 8, 2017 | E4207

GEN

ETICS

PNASPL

US

Fig. 4. amiR-KS3 induces gefitinib sensitivity in an A549 KRAS G12S mutated cell. (A) Effects of gefitinib on proliferation when KRAS is WT (green) and G12Smutated (orange). (B) Representative plots of Annexin V5 assay on A549 KRAS G12S transfected with amiR-KS3 and scramble and treated with DMSO (Upper)and gefitinib 5 μM (Lower) for 48 h. (C) Bar graph representing the results shown in B. (D) Caspase 3/7 activity assay after amiR-KS3 and scramble transfection,respectively, and treated with DMSO and gefitinib 5 μM for 48 and 72 h. Each bar is normalized to its own corresponding DMSO-treated. (E) Apoptotic cellpercentage of A549 treated with DMSO and gefitinib 5 μM for 48, 72, and 96 h after Trypan Blue cell viability test. Statistical significance for amiR-KS3 withgefitinib (green) was calculated in relation to amiR-KS3 with DMSO (yellow). *P < 0.05; **P < 0.01; ***P < 0.001. Gef., gefitinib; Scr, scramble.

E4208 | www.pnas.org/cgi/doi/10.1073/pnas.1620562114 Acunzo et al.

Additionally, we have analyzed gene expression in NSCLC celllines transfected with our amiR-KS3 and carrying KRAS WTand G12S, respectively, and, through computational analyses, wehave detected no relevant off-target effects.The main goal of our methodology thus aims for its applica-

tion to several other point-mutated oncogenes in addition toKRAS, potentially also providing a major advantage to RNA-based therapy for cancer in regard to delivery. In fact, the ad-vantages of our strategy lie in our artificial molecules requiringan efficient, yet nonspecific, delivery, having little to no biologicaleffect on cells expressing only the WT form of the desired target.Considering recent technology developed for the delivery of smallRNAs, such as nanoparticles and liposomes (41), our methodwould provide a strategy that could allow one to compensate forthe lack of specificity of the delivery system, through its ability toexclusively affect specific targets expressed only in cancer cells.Furthermore, coupling our strategy with more specific deliverysystems (i.e., associating antibodies or aptamers) could even ad-ditionally reduce any potential “leakage” of such systems.We have clearly shown that our methodology for the design of

artificial noncoding RNAs can be used to effectively and exclu-sively target point-mutated KRAS. The application of such adesign technique can be extended to other point-mutated on-cogenes and to other diseases caused by point-mutation gain offunction, providing a major advantage compared with otherRNA-based therapies.

MethodsCell Culture, Transfection, and Chemicals. All cell lines (H292, A549, H1299, BX-PC3, PC1, SK-Mel2, WM266, and Hek-293) were grown in RPMI with 10% FBS,L-glutamine, and antibiotics (Invitrogen). All of the transfections were per-formed by using Lipofectamine 3000 (Invitrogen) as suggested by themanufacturer. All cell lines used were transfected at 60% (72 and 96 h ex-periment) or 80% confluence in 60-mm or 96-well plates with a serum-freemedium without antibiotics and then transfected with 100 nmol of artificial-premiR oligonucleotides (Qiagen), custom siRNA KS3 (Qiagen), Kras-siRNA (SantaCruz), or negative control for 48, 72, or 96 h. pGL3-Control (Promega) Kras-CDS12WT, pGL3-Control Kras-CDS12S (mutated in G12S), pGL3-Control Kras-CDS12D, pGL3-Control B-RAFCDS600WT, and pGL3-Control B-RAFCDS600D(mutated in V600D) were transfected as described in Luciferase Assay.

RNA Extraction. Total RNA was extracted with TRIzol solution (Invitrogen),according to the manufacturer’s instructions, followed by an RNA extractionkit (NORGEN).

Western Blot Analysis. H292, A549, H1299, BX-PC3, PC1, SK-Mel2, andWM266 cells were seeded and grown in appropriate mediumwith 10% FBS inp60 plates for 24 h before the transfection. After transfection, cells werewashed with cold PBS and collected for RNA and protein extraction. Cells forprotein extraction were subjected to lysis in a lysis buffer (50 mM Tris·HCl,1 mM EDTA, 20 g/L SDS, 5 mM DTT, and 10 mM phenylmethylsulfonyl fluo-ride), and then equal amounts of protein lysates (50 mg each) and rainbowmolecular weight marker (Bio-Rad Laboratories) were separated by 4–20%SDS/PAGE and then electrotransferred to nitrocellulose membranes. Themembranes were blocked with a buffer containing 5% nonfat dry milk orBSA (depending on the antibody) in Tris-buffered saline with 0.1% Tween-20 for at least 1 h and incubated overnight with antibodies at 4 °C. After asecond wash with Tris-buffered saline with 0.1% Tween 20, the membraneswere incubated with peroxidase-conjugated secondary antibodies (GEHealthcare, Amersham) and developed with an enhanced chemiluminescencedetection kit (Pierce).

Antibody Used for Western Blot Analysis. Vinculin (Abcam) was used as aloading control. K-RAS was from Abcam; pAkt pERK1/2, PARP antibodieswere from Cell Signaling Technologies (Danvers); BRAF was from Santa Cruz.

Real-Time qRT-PCR. Real-time PCRwas performed by using a standard TaqManPCR Kit protocol on an Applied Biosystems 7900HT Sequence DetectionSystem (Applied Biosystems). The 10-μL PCR included 0.67 μL of RT product,1 μL of TaqMan Universal PCR Master Mix (Applied Biosystems), 0.2 mMTaqMan probe, 1.5 mM forward primer, and 0.7 mM reverse primer. Thereactions were incubated in a 96-well plate at 95 °C for 10 min, followed by

40 cycles of 95 °C for 15 s and 60 °C for 1 min. All reactions were run intriplicate. The threshold cycle (Ct) is defined as the fractional cycle number atwhich the fluorescence passes the fixed threshold. The comparative Ctmethod for relative quantization of gene expression (Applied Biosystems)was used to determine miRNA expression levels. The y axis represents the2(″DCt), or the relative expression of different miRs. amiR expression wascalculated relative to U44 and Kras relative to GAPDH expression. Exper-iments were carried out in triplicate for each data point, and data analysiswas performed by using software (Bio-Rad). All custom probes were fromApplied Biosystems.

Luciferase Assay. To generate Kras and B-RAF luciferase reporter constructs, apart of CDSs were amplified by PCR and cloned downstream of the lucif-erase CDS in the PGl3Ctr vector at the XbaI restriction site (Promega).Mutations were introduced into the miRNA-binding sites by using theQuikChange Mutagenesis Kit (Stratagene). HEK-293 cells were cotransfectedwith renilla plasmid, pGL3-Control Kras-CDS12WT, pGL3-Control Kras-CDS12S(mutated in G12S), pGL3-Control Kras-CDS12D, pGL3-Control B-RAFCDS600WT,and pGL3-Control B-RAFCDS600D (mutated in V600D) plasmids and amiRs byusing Lipofectamine 3000 (Invitrogen). After 24 h, cells were lysed and assayedwith Dual Luciferase Assay (Promega) according to the manufacturer’sinstructions.

The primers used for the cloning were as follows:

K-RAS CDS forward (FW): 5′-TCTAGATCAGCGGCTCCCAGGTGC-3′

K-RAS CDS reverse (RV): 5′-TCTAGAGCCCTCCCCAGTCCTCATGTAC-3′

BRAF CDS FW: 5′-GGCTCGAGTGAAGTAGGAGTACTCAGG-3′

BRAF CDS RV: 5′-TTGCGGCCGCGCCAGCAGCTCAAT-3′

K-RAS G12S FW: 5′-GTGGTAGTTGGAGCTAGTGGCGTAGGCAAGA-3′

K-RAS G12S RV: 5′-TCTTGCCTACGCCACTAGCTCCAACTACCAC-3′

K-RAS G12D FW: 5′-CTCTTGCCTACGCCATCAGCTCCAACTACCA-3′

K-RAS G12D RV: 5′-TGGTAGTTGGAGCTGATGGCGTAGGCAAGAG-3′

BRAFCDSV600D FW: 5′-GACCCACTCCATCGAGATTTCACTGTAGCTAGACC-AAAATCA-3′

BRAFCDSV600D RV: 5′-TGATTTTGGTCTAGCTACAGTGAAATCTCGATGGA-GTGGGTC-3′

Transwell Migration Assay. To measure cell migration, 8-nm pore size cultureinserts (Transwell; Costar) were placed into the wells of 24-well culture plates,separating the upper and lower chambers. In the lower chamber, 400 μL ofRPMI (10% FBS) was added. A549, H1299, and H292 cells were grown in RPMIcontaining 10% FBS to ∼60% confluence and transfected with 100 nmol ofamiR-KS3 precursor molecule or negative control. After 48 h, the cells wereharvested by trypsinization, and 1 × 105 cells were added to the upper chamberin no-FBS medium. After 16 h of incubation at 37 °C with 5% CO2, the cells thathad migrated through the pores were fixed and colored by Brilliant Blue, andthe number was quantified by counting 12 independent visual fields under themicroscope (Zeiss) using 20× magnification.

In Vitro Migration Scratch. Confluent A549, H1299, and H292 cells weretransfected by using Lipofectamine 3000 (Invitrogen) with amiR-KS3 orcontrol oligonucleotides. At 48 h after transfection, a cell scratch spatula wasused to make a scratch in the cell monolayer, after which the cell monolayerswere rinsed and further incubated. Pictures of the scratches were taken byusing a digital camera system coupled to a microscope. The cells were in-cubated for 8 and 24 h, and pictures of the scratches were taken by using adigital camera system coupled to a microscope. To determine the migrationdistance (micrometers), we used a software image analysis program, ImageJ(a version of NIH Image, https://imagej.nih.gov/ij/), as the reduction of thewide of the open area.

Cell Death and Cell Proliferation Assays.Counting. A549, H1299, and H292 cells, at an appropriate time after trans-fection, were harvested by trypsinization, and the number of them wascounted on a hemocytometer.AlamarBlue assay. Cells were plated in 96-well plates in triplicate and trans-fected as described. Cell viability was examined after 48 and/or 72 h withalamarBlue (Bio-Rad), according to the manufacturer’s protocol. Metaboli-cally active cells were detected by adding 10 μL to each well. After 1 h ofincubation, the plates were analyzed in a Multilabel Counter (Bio-Rad).

Acunzo et al. PNAS | Published online May 8, 2017 | E4209

GEN

ETICS

PNASPL

US

Growth density assay. The cells were harvested 8 h after transfection bytrypsinization, and 5 × 105 of them were seeded into 10-mm plates in trip-licates. The cells were then stained and fixed, and density was assessed byusing the QuantityOne program (Bio-Rad).Cell-cycle analysis. The cells, 48 h after transfection, were trypsinized, washedin PBS, and fixed with ice-cold 70% ethanol while vortexing. Cells wererehydrated in PBS and stained 30 min at room temperature with propidiumiodide (PI; 50 mg/mL PI and 0.5 mg/mL RNase in PBS). Samples were analyzedon a FACSCalibur flow cytometer (BD Bioscience). Percentages of cells in theG0/G1, S, and G2/M phases of the cell cycle were calculated by using theMODFIT-LT software (Verity Software House).Apoptosis. Apoptosis was assessed by using Annexin V-FITC apoptosis de-tection kits followed by flow-cytometric analysis and caspase 3/7 activity.A549 cells were transfected and, after 24 h, treated with 5 μMgefinitib. Afterincubation, cells were washed with cold PBS and removed from the plates bytrypsinization. The resuspended cells were washed with cold PBS and stainedwith FITC-conjugated annexin V antibody according to the manufacturer’sinstructions (Roche Applied Science). Samples were allowed to stand for15 min in the dark and immediately analyzed. Samples were analyzed on aFACSCalibur flow cytometer (BD Bioscience). The percentage of apoptoticcells was calculated by using FlowJo software.

For detection of caspase 3/7 activity, cells were cultured and transfected in96-well plates; after 24 h, they were treated with 5 μMgefinitib and analyzed

by using the Caspase-Glo 3/7 Assay kit (Promega) according to the manu-facturer’s instructions. The percentage of caspase activity was corrected forbackground levels found in the corresponding untreated controls.

In Vivo Studies. The animal studies were approved by The Ohio State Uni-versity Institutional Animal Care and Use Committee (OSU-IACUC). ViableA549 cells were injected s.c. (total cells 7 × 106) into the left flanks of 6-wk-old nude mice. After 1 wk, we started to intratumorally transfect 11 μg (finalconcentration) of scrambled oligonucleotides or amiR-KS3, by using Turbo-Fect in vivo Transfection Reagent (Invitrogen), according to the protocol ofthe manufacturer. We treated three times a week for 17 d. Tumor sizes weremeasured before each treatment. Three days after the last treatment, micewere killed, and tumors were excised; a number of the tumors collectedwere lysed for RNA extraction, and the rest was put in formalin along withparaffin. Tumor volumes were determined by using the equation V (in mm3) =A × B 2/2, where A is the largest diameter and B is the perpendicular diameter.

Tissue Analysis. The tissue was processed in a Leica ASP 300S Tissue Processorfor ∼5 h. Samples were then embedded in Leica Surgipath Paraplast Plusembedding medium. The tissue was cut at 5 μm and placed on white frosted,positive-charged slides. Slides were placed on a slide warmer at 66 °C for∼15 min. The slides were then placed in the Leica Autostainer XL by using

Fig. 5. Assessment of off-target effects for amiR-KS3. (A) Heat maps of down-regulated transcripts detected through the nCounter PanCancer PathwaysPanel in H1299, H292 (KRAS WT), and A549 KRAS G12S NSCLC cell lines. (B) Venn diagrams of down-regulated genes after amiR-KS3 transfection in A549,H1299, and H292 cell lines, respectively. (C) Sylamer hypergeometric statistical enrichment landscape plots for three words corresponding to the 6-bp 3′ UTRsequences complementary to the 6-mer amiR-KS3 canonical and two offset seeds (listed according to enrichment, above; or depletion, below; from left toright in each graph), in transfection experiments in A549, H1299, and H292 cell lines, respectively. In each case, the x axis represents the sorted gene list frommost down-regulated (C, Left) to most up-regulated (C, Right). The y axis shows the hypergeometric significance for each word at each leading bin. Positivevalues indicate enrichment [−log10(P value)], and negative values indicate depletion [log10(P value)]. The horizontal red dotted lines in each graph represent Evalue thresholds (Bonferroni corrected) of 0.01 for both enriched and depleted significances. Scr, scramble.

E4210 | www.pnas.org/cgi/doi/10.1073/pnas.1620562114 Acunzo et al.

the standard H&E staining program. Slides were coverslipped by usingTissue-Tek Coverslipping Film.

For immunohistochemistry, all sections were stained by using a Bond Rxautostainer (Leica). Briefly, slides were baked at 65 °C for 15 min, and au-tomated software performed dewaxing, rehydration, antigen retrieval,blocking, primary antibody incubation, postprimary antibody incubation,detection (DAB), and counterstaining by using Bond reagents (Leica).Samples were then removed from the machine, dehydrated throughethanols and xylenes, mounted, and coverslipped. An antibody for thefollowing marker was diluted in Antibody diluent (Leica): rabbit antibodyK-i67 (1:200; Abcam).

Differential Expression Analysis in amiR-KS3 and siRNA-KS3 Transfections. Weperformed differential expression analysis of 729 genes of the nCounterPanCancer Pathways Panel, in biological triplicates. Sixty nanograms of totalRNA was hybridized overnight with nCounter Reporter (20 μL) probes inhybridization buffer and in excess of nCounter Capture probes (5 μL) at 65 °Cfor 16−20 h. After overnight hybridization, excess probes were removed byusing two-step magnetic beads-based purification on an automated fluidichandling system (nCounter Prep Station). Biotinylated capture probe-boundsamples were immobilized and recovered on a streptavidin-coated cartridge.The abundance of specific target molecules was then quantified by using thenCounter digital analyzer. Individual fluorescent barcodes and target mol-ecules present in each sample were recorded with a CCD camera by per-forming a high-density scan (600 fields of view). Normalization wasperformed by using the nSolver Analysis Software (Version 3.0; NanoStringTechnologies), applying the geometric mean of the negative controls fornegative background subtraction; the positive controls for the technicalnormalization; and of the Housekeeping genes for the biological normali-zation in all samples, as recommended by NanoString. P values were calcu-lated by using the LIMMA package from the Bioconductor R project. P valueswere used to rank all genes, retaining those under a significant threshold of0.05 and with average expression of >50 counts, after negative backgroundsubtraction, within triplicates of at least one of the compared conditions.These data were used to generate heat maps through the employment ofthe R package heatmap.2.

We created the Venn diagrams that represent the intersection betweenthe sets of down-regulated genes by using the web-based tool InteractiVenn(www.interactivenn.net/index2.html).

Consensus Targeting Prediction for amiR-KS3. We respectively predictedbinding sites on 5′ UTR, CDS, and 3′ UTR sequences (UCSC.hg19) of down-regulated genes through a consensus of three miRNA target predictiontools: miRanda36 (Version 3.3a), PITA37 (Version 6.0), and miRiam38, our in-house tool, enhanced with the scoring function described in Laganà et al. (19).

Standard parameters were used for the tools miRanda and PITA. miRiam’sparameters were set to detect canonical binding sites only (6-mer, 7-mer-A1,7-mer-m8, and 8-mer), allowing no mismatches in the seed (e.g., wobblepairs). miRiam’s scoring function is based on the combination of the tree-based multiple linear regression learning system M5P with CTree and takesinto account six different features of miRNA/target interactions: type ofseed match, miRNA nucleotide composition, pairing of the three regionsof the miRNA, adenylate–uridylate (AU) content of the binding site and itsflanking regions, structural accessibility of the binding site, and presenceof AU-rich element and cytoplasmic polyadenylation element motifs up-stream of the binding sites.

Pathway Enrichment Analysis After amiR-KS3 and siRNA-KS3 Transfection inA549 Cell Lines. We performed functional enrichment analysis on theretained set of differentially expressed genes (mentioned above) by using thesoftware IPA by Ingenuity. Settings used included experimentally observeddata for human species, specifically, pathways exclusively associated to theA549 NSCLC cell line.

Off-Target Analysis for amiR-KS3. The 729 genes of the nCounter PanCancerPathways Panel were ranked from most down-regulated to most up-regulated to generate a ranked transcript list. Sylamer software39 wasthen used to respectively assess 5′ UTR, CDS, and 3′ UTR amiR-KS3 seedmatch enrichment P values across the ranked transcript list. Three FASTAfiles, respectively, containing 5′ UTR, CDS, and 3′ UTR sequences of refseqtranscripts (UCSC.hg19), were provided to Sylamer, along with the rankedtranscript list, after they were purged of duplicate sequences (bioinfo5.ugr.es/srnatoolbox) and sequences referring to noncoding RNAs, re-dundant segments were eliminated through the web tool purge-sequence ofthe RSAT package (rsat-tagc.univ-mrs.fr/rsat/), and, finally, low-complexity re-gions were masked with the tool RepeatMasker by using default settings(www.repeatmasker.org/). Sylamer settings were 6-bp motifs, fixed number ofbins 40, Markov correction 4, scan restricted to sequences in the ranked genelist, and searching for words (option –words) corresponding to the threepossible 6-nt motifs representing the reverse complementary sequences of thethree 6-nt-long seeds of amiR-KS3, from positions 1–6, 2–7, and 3–8 from the5′ end.

ACKNOWLEDGMENTS. We thank Prof. F. Kay Hubner for her useful supportduring the drafting process and Prof. Patrick Nana-Sinkam for his precioussuggestions during the experimental design and the scientific discussion.This work was supported by National Institutes of Health Grant NIH U01CA166905-04. D.V. was supported by Italian Foundation for Cancer ResearchGrant 16572.

1. Vogelstein B, et al. (2013) Cancer genome landscapes. Science 339:1546–1558.2. Malumbres M, Barbacid M (2003) RAS oncogenes: The first 30 years. Nat Rev Cancer 3:

459–465.3. Cox AD, Fesik SW, Kimmelman AC, Luo J, Der CJ (2014) Drugging the undruggable

RAS: Mission possible? Nat Rev Drug Discov 13:828–851.4. Quinlan MP, Settleman J (2009) Isoform-specific ras functions in development and

cancer. Future Oncol 5:105–116.5. Prior IA, Lewis PD, Mattos C (2012) A comprehensive survey of Ras mutations in

cancer. Cancer Res 72:2457–2467.6. Johnson L, et al. (2001) Somatic activation of the K-ras oncogene causes early onset

lung cancer in mice. Nature 410:1111–1116.7. Misale S, et al. (2012) Emergence of KRAS mutations and acquired resistance to anti-

EGFR therapy in colorectal cancer. Nature 486:532–536.8. Yuan TL, et al. (2014) Development of siRNA payloads to target KRAS-mutant cancer.

Cancer Discov 4:1182–1197.9. McCormick F (2015) KRAS as a therapeutic target. Clin Cancer Res 21:1797–1801.10. Shimizu T, et al. (2012) The clinical effect of the dual-targeting strategy involving

PI3K/AKT/mTOR and RAS/MEK/ERK pathways in patients with advanced cancer. Clin

Cancer Res 18:2316–2325.11. Samatar AA, Poulikakos PI (2014) Targeting RAS-ERK signalling in cancer: Promises

and challenges. Nat Rev Drug Discov 13:928–942.12. Gysin S, Salt M, Young A, McCormick F (2011) Therapeutic strategies for targeting ras

proteins. Genes Cancer 2:359–372.13. Croce CM (2009) Causes and consequences of microRNA dysregulation in cancer. Nat

Rev Genet 10:704–714.14. Helwak A, Kudla G, Dudnakova T, Tollervey D (2013) Mapping the human miRNA

interactome by CLASH reveals frequent noncanonical binding. Cell 153:654–665.15. Acunzo M, Romano G, Wernicke D, Croce CM (2015) MicroRNA and cancer—a brief

overview. Adv Biol Regul 57:1–9.16. Brennecke J, Stark A, Russell RB, Cohen SM (2005) Principles of microRNA-target

recognition. PLoS Biol 3:e85.

17. Doench JG, Sharp PA (2004) Specificity of microRNA target selection in translationalrepression. Genes Dev 18:504–511.

18. Ye W, et al. (2008) The effect of central loops in miRNA:MRE duplexes on the effi-ciency of miRNA-mediated gene regulation. PLoS One 3:e1719.

19. Laganà A, et al. (2014) miR-Synth: A computational resource for the design of multi-site multi-target synthetic miRNAs. Nucleic Acids Res 42:5416–5425.

20. Bartel DP (2009) MicroRNAs: Target recognition and regulatory functions. Cell 136:215–233.

21. Grimson A, et al. (2007) MicroRNA targeting specificity in mammals: Determinantsbeyond seed pairing. Mol Cell 27:91–105.

22. Elbashir SM, et al. (2001) Duplexes of 21-nucleotide RNAs mediate RNA interferencein cultured mammalian cells. Nature 411:494–498.

23. Tape CJ, et al. (2016) Oncogenic KRAS regulates tumor cell signaling via stromal re-ciprocation. Cell 165:910–920.

24. Downward J (2003) Targeting RAS signalling pathways in cancer therapy. Nat RevCancer 3:11–22.

25. Westcott PM, To MD (2013) The genetics and biology of KRAS in lung cancer. Chin JCancer 32:63–70.

26. Campbell PM, Der CJ (2004) Oncogenic Ras and its role in tumor cell invasion andmetastasis. Semin Cancer Biol 14:105–114.

27. Okudela K, et al. (2004) K-ras gene mutation enhances motility of immortalized air-way cells and lung adenocarcinoma cells via Akt activation: Possible contribution tonon-invasive expansion of lung adenocarcinoma. Am J Pathol 164:91–100.

28. Cragg MS, Kuroda J, Puthalakath H, Huang DC, Strasser A (2007) Gefitinib-inducedkilling of NSCLC cell lines expressing mutant EGFR requires BIM and can be enhancedby BH3 mimetics. PLoS Med 4:1681–1689; discussion 1690.

29. Sordella R, Bell DW, Haber DA, Settleman J (2004) Gefitinib-sensitizing EGFR muta-tions in lung cancer activate anti-apoptotic pathways. Science 305:1163–1167.

30. Janmaat ML, Rodriguez JA, Gallegos-Ruiz M, Kruyt FA, Giaccone G (2006) Enhancedcytotoxicity induced by gefitinib and specific inhibitors of the Ras or phosphatidylinositol-3 kinase pathways in non-small cell lung cancer cells. Int J Cancer 118:209–214.

Acunzo et al. PNAS | Published online May 8, 2017 | E4211

GEN

ETICS

PNASPL

US

31. Pao W, et al. (2005) KRAS mutations and primary resistance of lung adenocarcinomasto gefitinib or erlotinib. PLoS Med 2:e17.

32. Garofalo M, et al. (2011) EGFR and MET receptor tyrosine kinase-altered microRNAexpression induces tumorigenesis and gefitinib resistance in lung cancers. Nat Med18:74–82.

33. Bartel DP (2004) MicroRNAs: Genomics, biogenesis, mechanism, and function. Cell116:281–297.

34. Jackson AL, et al. (2006) Widespread siRNA “off-target” transcript silencing mediatedby seed region sequence complementarity. RNA 12:1179–1187.

35. Birmingham A, et al. (2006) 3′ UTR seed matches, but not overall identity, are asso-ciated with RNAi off-targets. Nat Methods 3:199–204.

36. John B, et al. (2004) Human microRNA targets. PLoS Biol 2:e363.37. Kertesz M, Iovino N, Unnerstall U, Gaul U, Segal E (2007) The role of site accessibility

in microRNA target recognition. Nat Genet 39:1278–1284.38. Laganà A, et al. (2010) Prediction of human targets for viral-encoded microRNAs by

thermodynamics and empirical constraints. J RNAi Gene Silencing 6:379–385.39. van Dongen S, Abreu-Goodger C, Enright AJ (2008) Detecting microRNA binding and

siRNA off-target effects from expression data. Nat Methods 5:1023–1025.40. Lam JK, Chow MY, Zhang Y, Leung SW (2015) siRNA versus miRNA as therapeutics for

gene silencing. Mol Ther Nucleic Acids 4:e252.41. Miele E, et al. (2012) Nanoparticle-based delivery of small interfering RNA: Challenges

for cancer therapy. Int J Nanomedicine 7:3637–3657.

E4212 | www.pnas.org/cgi/doi/10.1073/pnas.1620562114 Acunzo et al.