Codon bias imposes a targetable limitation on KRAS-driven therapeutic resistance

Abstract

KRAS mutations drive resistance to targeted therapies, including EGFR inhibitors in colorectal cancer (CRC). Through genetic screens, we unexpectedly find that mutant HRAS, which is rarely found in CRC, is a stronger driver of resistance than mutant KRAS. This difference is ascribed to common codon bias in HRAS, which leads to much higher protein expression, and implies that the inherent poor expression of KRAS due to rare codons must be surmounted during drug resistance. In agreement, we demonstrate that primary resistance to cetuximab is dependent upon both KRAS mutational status and protein expression level, and acquired resistance is often associated with KRAS^Q61 mutations that function even when protein expression is low. Finally, cancer cells upregulate translation to facilitate KRAS^G12 -driven acquired resistance, resulting in hypersensitivity to translational inhibitors. These findings demonstrate that codon bias plays a critical role in KRAS-driven resistance and provide a rationale for targeting translation to overcome resistance.

ARTICLE

Received 9 Sep 2016 | Accepted 12 Apr 2017 | Published 8 Jun 2017

Moiez Ali1,*, Erin Kaltenbrun1,*, Grace R. Anderson1, Sarah Jo Stephens1, Sabrina Arena2,3, Alberto Bardelli2,3, Christopher M. Counter1,** & Kris C. Wood1,**

KRAS mutations drive resistance to targeted therapies, including EGFR inhibitors in colorectal cancer (CRC). Through genetic screens, we unexpectedly nd that mutant HRAS, which is rarely found in CRC, is a stronger driver of resistance than mutant KRAS. This difference is ascribed to common codon bias in HRAS, which leads to much higher protein expression, and implies that the inherent poor expression of KRAS due to rare codons must be surmounted during drug resistance. In agreement, we demonstrate that primary resistance to cetuximab is dependent upon both KRAS mutational status and protein expression level, and acquired resistance is often associated with KRASQ61 mutations that function even when protein expression is low. Finally, cancer cells upregulate translation to facilitate KRASG12-driven acquired resistance, resulting in hypersensitivity to translational inhibitors. These ndings demonstrate that codon bias plays a critical role in KRAS-driven resistance and provide a rationale for targeting translation to overcome resistance.

DOI: 10.1038/ncomms15617 OPEN

Codon bias imposes a targetable limitation on KRAS-driven therapeutic resistance

1 Department of Pharmacology and Cancer Biology, Duke University, Durham, North California 27710, USA. 2 Department of Oncology, University of Torino, 10060 Candiolo, Torino, Italy. 3 Candiolo Cancer Institute-FPO, IRCCS, 10060 Candiolo, Torino, Italy. * These authors contributed equally to this work.** These authors jointly supervised this work. Correspondence and requests for materials should be addressed to C.M.C. (email: mailto:[email protected]

Web End [email protected] ) or to K.C.W. (email: mailto:[email protected]

Web End [email protected] ).

NATURE COMMUNICATIONS | 8:15617 | DOI: 10.1038/ncomms15617 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 1

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15617

Primary and acquired resistance to targeted therapies place major limitations on the clinical efcacy of these drugs, despite their promising preclinical potential1. One proposed

method to overcome resistance has been to identify genetic2, epigenetic and signalling alterations that underlie sensitivity and resistance, and then use these as biomarkers to better select patients that will benet from specic mono- or combination therapies35. In this regard, numerous studies have specically implicated activating mutations in KRAS as a major driver of both primary and acquired resistance to diverse targeted therapies.

In the area of primary resistance, activating mutations in KRAS track with poor responses to EGFR inhibition with the monoclonal antibodies (mAbs) cetuximab and panitumumab, or small molecule EGFR inhibitors like getinib and erlotinib, in colorectal cancer (CRC) and non-small cell lung cancer (NSCLC) patients610. Activating mutations in KRAS have also been proposed as a mechanism of primary resistance to the tyrosine kinase inhibitor (TKI) imatinib in KIT- or PDGFRA-mutant gastrointestinal stromal tumours11 and to JAK inhibitors in JAK2-mutant myeloproliferative neoplasms12.

In addition to cases of primary resistance, the emergence of de novo KRAS mutations has also been linked to acquired resistance in multiple cancer types and contexts: to both anti-EGFR therapy and MEK1/2 inhibitors in CRC13,14, to imatinib in chronic myelogenous leukaemia15, and to BRAF/ MEK inhibitors in melanoma16. In CRC, despite achieving initial responses, patients who originally present with no detectable mutations in KRAS (wild-type (WT) KRAS, or KRASWT, disease) develop acquired resistance to anti-EGFR therapy that is commonly driven by KRAS mutations, limiting the clinical benet of this therapy13,1719. Curiously, the mutations detected in this setting of acquired resistance are a balance of G12/G13 and Q61 mutations, the latter of which are rarely found in treatment naive CRC20,21. An improved understanding of the biology and signalling that support KRAS-mediated resistance may therefore give rise to new therapeutic strategies for these refractory tumours.

KRAS belongs to a family of three genes, the other two being HRAS and NRAS. Interestingly, it is KRAS that is the most commonly mutated of the three in a wide spectrum of cancers and in the setting of resistance22. Despite this apparent contrast in epidemiological data, the encoded proteins are very similar, and in fact share B85% sequence identity22. However, we discovered that the coding nucleotide sequence varies extensively between these three genes. Specically, HRAS is enriched in common codons that yield robust translation and hence high protein expression. KRAS is characterized by rare codons, yielding poor translation and low expression, while NRAS has a mixture of common and rare codons and intermediate expression22.

Here, we show that this rare-codon bias, entrenched in the nucleotide sequence of KRAS, plays a critical role in both primary and acquired resistance and may underlie observations of elevated KRAS expression in models of resistance to the mAb cetuximab, the emergence of more rare KRAS amino acid site mutations in patients with acquired cetuximab resistance, and provide a novel therapeutic avenue to combat resistance.

ResultsMutant HRAS confers greater drug resistance than mutant KRAS. Previously, we described a systematic approach to identify novel drug resistance pathways by screening cancer cell lines with a pooled library of 36 mutant complementary DNAs (cDNAs) that constitutively activate or inhibit 17 key oncogenic signalling pathways23. Using this approach, we observed that across nine

genotype- and lineage-dened groupings of 14 cell lines, treated with a total of 14 targeted therapies, there was a marked differential resistance conferring potential between oncogenic HRASG12V and KRASG12V (Fig. 1a). Specically, oncogenic HRASG12V conferred resistance in 27 of 29 screens with a broad spectrum of therapeutics using a previously established scoring threshold (Fig. 1b)23, and it scored as the top overall hit in 22 of 29 of the screens (Fig. 1c). By contrast, KRASG12V scored only rarely (6 of 29 screens), and never as the top overall hit. Moreover, even in screens where both HRASG12V and KRASG12V reached scoring criteria, HRASG12V consistently achieved higher enrichment scores (the relative abundance of each construct in the presence of drug normalized to the same value in the absence of drug), implying stronger resistance (Fig. 1d). To determine whether the ability to score in our screens correlated with the expression of each construct, we performed western blot analysis using extracts from cell lines in which both constructs (A375) or only HRASG12V (SKBR3, PC9, NCIH508) scored. In all cases examined, HRASG12V was detected at higher levels compared to KRASG12V (Fig. 1e). Thus, the level of RAS expression correlated with resistance, with HRAS consistently expressed at higher levels.

Codon bias underlies increased resistance conferred by HRAS. One feature of RAS genes that could explain the enhanced resistance conferred by HRAS relative to KRAS is codon bias. To address this hypothesis, we created native-codon and codon-modied cDNAs encoding FLAG-tagged, oncogenic HRAS and KRAS. Specically, we created a rare codon-enriched, oncogenic HRASG12D cDNA by converting key common codons to rare codons (termed HRASR,G12D). Reciprocally, we also created a common codon-enriched, oncogenic KRASG12D cDNA by exchanging rare codons for common codons (termed KRASC,G12D). These four constructs as well as an empty vector control were individually stably expressed in three very different human cancer cell lines, namely the BRAFV600E mutation-positive melanoma cell line UACC-62 (Fig. 2ac), the EGFR mutation-positive NSCLC cell line PC9 (Fig. 2df), and the PDGFR-amplied NSCLC cell line NCIH1703 (Fig. 2gi). In agreement with the known effect of codon bias on RAS protein expression22, oncogenic HRASG12D was readily detected by immunoblot analysis, and its expression was greatly reduced following the exchange of common codons for rare codons in all three cell lines (Fig. 2a,d,g). Conversely, oncogenic KRASG12D was very poorly expressed in all three cell lines, an effect that was reversed by changing rare codons to common (Fig. 2a,d,g). When these cell lines were treated with targeted inhibitors against the driver oncogene in each line, namely vemurafenib to inhibit oncogenic BRAF in UACC-62 cells, getinib to inhibit oncogenic EGFR in PC9 cells and sunitinib to inhibit PDGFR in NCIH1703 cells, we found that in each case the degree of resistance matched RAS protein expression. Namely, the RAS isoforms with the most common codons (HRASG12D and KRASC,G12D) imparted greater resistance than the versions with more rare codons (HRASR,G12D and KRASG12D) (Fig. 2b,e,h, quantied in Fig. 2c,f,i). The same was also true in the completely independent assay of anchorage-independent growth (Supplementary Fig. 1). Thus, the ability of oncogenic HRAS to impart resistance to a broad spectrum of cell lines and drugs appears to be due, at least in part, to its high-level protein expression due to its inherent bias towards common codons.

Codon bias underlies KRAS-mediated resistance to cetuximab. The aforementioned nding appears, at face value, at odds with the clinical observation that oncogenic KRAS has been extensively

2 NATURE COMMUNICATIONS | 8:15617 | DOI: 10.1038/ncomms15617 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15617 ARTICLE

a b c

HRAS (G12V)

KRAS (G12V)

HcRed

Luciferase

100

27/29

100

Percentage of screens

scoring as top overall hit

Drug

(Target) Cell line

90

Percentage of screens

scoring

Genotype

80

80

90 22/29

A375 Colo679 UACC62

70

70

60

60

50

50

40

40

30

AZD6244

(MEK)

A375 Colo679 UACC62

VX-11E

(ERK)

A375 Colo679 UACC62

6/29

30

BRAF (V600)

melanomas

PLX4720

(RAF)

20

20

10

10

0/29 0/29 0

0/29

0/29 0/29

0 HRASG12V KRASG12V

HcRed Luciferase

HRASG12V KRASG12V

HcRed Luciferase

Construct

Construct

T47D MD-453

BT-20

d

115

MK-2206

(AKT)

T47D MD-453

BT-20

20

PIK3CA (H1047R)

breast cancers

BKM-120

(PI3K)

Torin1

(mTOR)

T47D MD-453

BT-20

Enrichment score

15

PLX4720 (Colo679) AZD6244 (A375) AZD6244 (UACC62) VX11E (Colo679) BKM120 (T47D) BKM120 (BT20) MK2206 (MB-453) Torin1 (T47D)

Torin1 (BT20) Rapamycin (MB-453) Lapatinib (SkBr3) Sunitinib (NCI-H1703) Gefitinib (PC9) Cetuximab (NCI-H508)

Rapamycin

(mTORC1)

T47D MD-453

BT-20

10

HER2 amplified

breast cancer

Lapatinib

(HER2) SkBr3

FGFR amplified

gastric carcinoma

PD173074

(FGFR) Kato III

PDGFR amplified

NSCLC

5

Sunitinib

(PDGFR)

NCI

H1703

EGFR mutant

NSCLC

Gefitinib

(EGFR) PC9

ALK mutant

NSCLC

0

HRASG12V

HcRed Luciferase Construct

KRASG12V

TAE684

(ALK) H2228

JAK2-mutant MPN Ruxolitinib

(JAK) UKE-1

EGFR driven CRC Cetuximab NCI-H508

Oxco-2

Enrichment score = 1.5 3.0

e

A375 SKBR3 PC9 NCIH508

Enrichment score > 3.0

+ HRASG12V

+ KRASG12V

+ HcRed

+ HRASG12V

+ KRASG12V

+ HcRed

+ HRASG12V

+ KRASG12V

+ HcRed

+ HRASG12V

+ KRASG12V

+ HcRed

HRAS

KRAS

-actin

29

29

45

Figure 1 | Pathway activating screens reveal differential resistance conferring potential between ectopic HRASG12V and KRASG12V. (a) Enrichment

scores for four lentivirally delivered cDNA constructs (encoding HRASG12V, KRASG12V, HcRed and Luciferase) across 29 resistance screens involving diverse kinase inhibitors and cell lines. Enrichment scores reect the relative abundance of each construct in the presence of drug normalized to the same value in the absence of drug, with scores exceeding 1.5 considered to be indicative of resistance. (b) Percentage of screens in a in which indicated constructs scored (enrichment score41.5). (c) Percentage of screens in a in which indicated constructs scored as the top overall hit from a library of 40 pathway activating constructs and controls. (d) Enrichment score for each construct in each screen. (e) Immunoblot analysis with antibodies against HRAS, KRAS or b-actin using extracts isolated from the indicated cell lines stably transduced with a vector encoding the indicated transgene. Images are cropped for clarity.

implicated as a driver of primary and acquired resistance. We posited that perhaps poor expression of KRAS due to rare codons is a barrier to resistance, such that in addition to an oncogenic mutation, resistance may require an increase in KRAS expression. To this end we focused on CRC, a cancer in which oncogenic KRAS is well established to be a central driver of primary and acquired resistance to anti-EGFR therapies13. The CRC cell line NCIH508, which is KRASWT, sensitive to the drug cetuximab, and can be rendered resistant to cetuximab through expression of oncogenic RAS, was stably transduced with the aforementioned vectors encoding no transgene or HRASG12D, HRASR,G12D, KRASG12D or KRASC,G12D. As before, expression of the RAS isoforms tracked with codon bias (Fig. 3a). Further, we observed that common codon-enriched RAS constructs conferred

resistance to cetuximab in colony growth assays relative to rare codon-enriched RAS constructs (Fig. 3b,c). Identical ndings were observed upon treatment of cells with an eight-log dilution series of cetuximab, where half-maximal growth inhibition values for both HRASG12D and KRASC,G12D were signicantly greater than for their rare codon-enriched counterparts (Fig. 3d,e). These results were reproducible in another KRASWT CRC cell line, LIM1215 (Supplementary Fig. 2A,B). Immunoblots performed in the presence and absence of cetuximab in cells expressing controls or common and rare RAS constructs demonstrated on-target inhibition of the mitogen-activated protein kinase (MAPK) pathway, as evidenced by modest to substantial decreases in phospho-ERK (p-ERK) levels in both NCIH508 and LIM1215 cell lines (Fig. 3f, Supplementary Fig. 2C). This

NATURE COMMUNICATIONS | 8:15617 | DOI: 10.1038/ncomms15617 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15617

a

b c

0

0.2 M vemurafenib

2.0

Relative cell growth

*

* * *

Empty

HRAS

HRAS

KRAS

UACC-62

KRAS

1.5

Empty

Empty

HRAS

HRAS

KRAS

KRAS

1.0

FLAG

-actin

FLAG

-actin

FLAG

-actin

0.5

0.0

Empty

HRASG12D

HRASR,G12D

KRASG12D

KRASC,G12D

d

e f

0

0.02 M gefitinib

1.0

Relative cell growth

**

** ** **

0.8

PC-9

NCIH1703

Empty

HRAS

HRAS

KRAS

KRAS

Empty

Empty

HRAS

HRAS

KRAS

KRAS

0.6

0.4

0.2

0.0

Empty

HRASG12D

HRASR,G12D

KRASG12D

KRASC,G12D

g

h i

0

12.9 nM sunitinib

NS

1.5

Relative cell growth

* * *

Empty

HRAS

HRAS

KRAS

KRAS

Empty

Empty

HRAS

HRAS

KRAS

KRAS

1.0

0.5

0.0

Empty

HRASG12D

HRASR,G12D

KRASG12D

KRASC,G12D

Figure 2 | Codon bias underlies the differential resistance conferred by oncogenic HRAS versus KRAS. (a,d,g) Immunoblot analysis with anti-FLAG and anti-b-actin in extracts from indicated cell lines stably transduced with a vector encoding the indicated transgene (or no transgene, empty). All HRAS and

KRAS constructs are FLAG tagged. (b,e,h) Representative plate of three technical replicates of colony growth assay of various HRAS and KRAS expressing derivatives in the presence of indicated drug. (c,f,i) Averages.e.m. number of colonies in b,e,h normalized to HRASG12D expressing cells (n 3 technical

replicates per condition). Images are cropped for clarity. *Pr0.05; **Pr0.01. P values were calculated with unpaired, two-tailed Students t-tests.

effect on MAPK signalling correlated with the degree of cell death, specically quantication of the extra long-BIM splice variant (EL-BIM) as well as total BIM levels (T-BIM), where decreased induction of this pro-apoptotic protein was observed in common codon-expressing cells when treated with cetuximab, an effect that was lost in the same cells expressing the rare codon counterparts (Fig. 3g, Supplementary Fig. 2D). Furthermore, Annexin V staining revealed that treatment with cetuximab led to signicant increases in Annexin V-positive cells in control and rare-codon RAS-expressing cells, and that this induction was reduced in cells expressing HRASG12D and KRASC,G12D (Fig. 3h, Supplementary Fig. 2E). Collectively, these results demonstrate that codon bias underlies the isoform-specic effects of RAS-driven resistance to cetuximab in CRC cell lines, and that common codon-expressing RAS isoforms drive resistance, at least partially through the suppression of drug-induced apoptosis. This nding suggests that overcoming poor translation of oncogenic KRAS may promote greater resistance to EGFR inhibitors.

Clinical model reveals selection of potent KRAS mutations. The demonstration that codon bias limits KRAS-driven therapeutic resistance in CRC suggests several testable, clinically relevant hypotheses. For example, low levels of KRAS expression may

select for more potent activating mutations that drive acquired resistance without fully overcoming suppressed protein synthesis. Recent work has established that although mutations in KRAS codons 12, 13 and 61 are all oncogenic, codon 61 mutations directly disrupt GTPase activity and more potently activate KRAS than allosteric codon 12 and 13 activating mutations24,25. Consistent with this idea, although it has been well established that activating mutations in codons 12, 13 and 61 of KRAS are all considered drivers of acquired resistance to anti-EGFR mAbs in metastatic CRC patients17,20,26, there appears to be an unexplained preponderance of codon 61 mutations in patients with anti-EGFR refractory disease, mutations which are seldom observed in patients prior to the initiation of therapy21,27. We therefore hypothesized that this detection of Q61 site mutations in tumours with acquired resistance may reect a selection for more potent mutations to overcome the inherent poor translation of KRAS. Indeed, by analysing three separate cohorts of metastatic CRC patients with acquired resistance to anti-EGFR therapy2730, we found that in all three groups, and in one combined subset of the groups, there was a signicant enrichment in the number of KRASQ61 site mutations in comparison to a treatment naive background population (Table 1). We next modelled two different point mutations in

4 NATURE COMMUNICATIONS | 8:15617 | DOI: 10.1038/ncomms15617 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15617 ARTICLE

a

b c

0

34.3 108M Cetuximab

10

Relative cell growth

*

* * **

8

Empty

HRAS

HRAS

KRAS

KRAS

Empty

Empty

HRASG12D

HRASR,G12D

KRASG12D

KRASC,G12D

6

NCIH508

29 FLAG

140 Vinculin

4

2

0 Empty

HRASG12D

HRASR,G12D

KRASG12D

KRASC,G12D

d

e

1.5

NS

*** ****

0.0006

Half-maximal cetuximab (M)

NS

0.0004

**** ***

****

Cell viability

Empty

KRASG12D

KRASC,G12D

HRASG12D

HRASR,G12D

1.0

0.0002

3106

0.5

2106

1106

0.0

0

13 12 11 10 9 8 7 6 5 4

Empty

log [Cetuximab(M)]

KRASG12D

KRASC,G12D

HRASG12D

HRASR,G12D

f

g

HRASR,G12D

KRASG12D

KRASC,G12D

2.0

Fold

EL-BIM protein (a.u.)

Empty

HRASG12D

1.5

+ + + + +

1.0

Cetuximab:

FLAG

p-ERK1/2

T-ERK1/2

T-Bim

Vinculin

29

140

0.5

0.0

40

Empty

HRASG12D

HRASR,G12D

KRASG12D

KRASC,G12D

40

h

50 * *

* NS

* NS

**

20

40

Annexin V+ / 7-AAD

30

20

10

0

Empty(+V)

Empty(+C)

KRAS (+V)

KRAS (+C)

KRASC (+V)

KRASC (+C)

HRAS (+V)

HRAS (+C)

HRASR (+V)

HRASR (+C)

Figure 3 | Rare codons limit KRAS-mediated resistance to cetuximab in colorectal cancer. (a) Immunoblot analysis with anti-FLAG or anti-vinculin antibody in NCIH508 cells stably transduced with a vector encoding the indicated transgene (or no transgene, empty). All HRAS and KRAS constructs are FLAG tagged. (b) Representative plate of three technical replicates of colony growth assay of various HRAS and KRAS expressing derivatives in the presence of cetuximab or vehicle control. (c) Number of colonies in b normalized to cells expressing no transgene, determined when the indicated cell line transduced with vectors encoding the indicated transgenes were treated with cetuximab. (d) Nonlinear regression growth curves depicting cell viability as detected by Cell Titer Glo of cells transduced with indicated constructs and treated with an eight-log dilution series of cetuximab. (e) Derived half-maximal growth inhibition values of curves in d. (f) Immunoblotting of indicated constructs for signalling of phospho- and total-ERK1/2 and T-BIM. All HRAS and KRAS constructs are FLAG tagged as in a, and vinculin is a loading control. (g) Fold change in extra long-BIM (EL-BIM) protein levels from the immunoblots in f comparing cetuximab and vehicle treated cells. (h) Apoptosis quantication (Annexin V /7-AAD ) of indicated constructs in the presence or absence of cetuximab. In all cases, data are averages.e.m. (n 3 technical replicates per condition). Immunoblotting images are cropped for clarity.

*Pr0.05; **Pr0.01; ***Pr0.001, ****Pr0.0001. P values were calculated with unpaired, two-tailed Students t-tests.

codons 12 or 61 of KRAS in the aforementioned KRASWT CRC cell lines (NCIH508 and LIM1215), then calculated the sensitivities of these lines to cetuximab. KRASQ61 site mutations were stronger drivers of resistance to anti-EGFR therapy than oncogenic KRASG12 site mutations (Fig. 4a,b, Supplementary Fig. 3A,B). Similar to our ndings with rare and common

codon-enriched RAS constructs, we observed on-target activity of cetuximab in these cells, as evidenced by reduced p-ERK levels (Fig. 4c, Supplementary Fig. 3C) and induction of EL-BIM and T-BIM that was suppressed by the resistance-conferring KRASQ61 mutations (Fig. 4d, Supplementary Fig. 3D). Importantly, these effects were independent of KRAS protein expression levels,

NATURE COMMUNICATIONS | 8:15617 | DOI: 10.1038/ncomms15617 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15617

Table 1 | KRAS mutational status prevalence among treatment naive and treatment resistant cohorts.

KRAS mutation

No. (patients)

Total P value

Treatment naive28 G12/13 76 224

Q61 4 224Treatment resistant29 G12/13 20 24 0.0108

Q61 10 24 o0.0001 Treatment resistant30 G12/13 3 16 0.42

Q61 8 16 o0.0001 Treatment resistant27 G12/13 NA NA o0.0001

Q61 9 27

Treatment resistant29,30

G12/13 23 40 0.0864

Q61 18 40 o0.0001

P values as assessed by two-tailed Fishers test. NA, not applicable. Superscripts signify reference number.

which were similarly low across the various 12 and 61 point mutations tested (Fig. 4e, Supplementary Fig. 3E). Together, these data suggest that selection for stronger KRAS mutations can overcome poor KRAS expression imposed by codon bias, a result that may explain the enrichment of KRASQ61 site mutations in patients with acquired cetuximab resistance.

KRAS expression determines intrinsic resistance to cetuximab. Aside from the possibility of selection for a more potent oncogenic mutation, another potential mechanism to overcome the inability of oncogenic KRAS to potently impart drug resistance is by producing more of the protein. As such, primary KRAS-driven resistance to cetuximab may be dependent not only on the presence of activating mutations, as is commonly understood17, but also on the degree to which cells surmount poor KRAS expression prior to therapy. To test this hypothesis, we examined a panel of 21 KRASG12/G13 mutant CRC cell lines that were previously characterized for their intrinsic sensitivity to cetuximab31 to determine whether the degree of resistance to this drug correlated with the expression level of mutant KRAS protein. Indeed, after stratifying cell lines on the basis of responsiveness to cell growth inhibition in the presence of cetuximab (with responsive lines being dened as exhibiting 410% growth inhibition by cetuximab and non-responsive lines showing no growth inhibition), we found a signicant elevation in KRAS expression in non-responsive cell lines relative to responsive lines (Fig. 5a,b). We then compared global translation rates in cetuximab responsive and non-responsive cell lines by labelling with the translation elongation inhibitor puromycin, which leads to direct puromycin incorporation into nascent polypeptides that can be detected with anti-puromycin antibodies32. Following 15 min of puromycin treatment, immunoblotting revealed the rate of global translation to be higher in cetuximab non-responsive cell lines than cetuximab responsive cell lines, and quantication of labelling showed a signicant correlation with KRAS expression in these cell lines (Supplementary Fig. 4A,B). This observed correlation between intrinsic resistance, KRAS protein expression and global protein synthesis suggests that KRAS mutational status alone may not fully predict primary resistance to EGFR inhibition in CRC. Instead, the expression level of mutant KRAS is likely to be a second key contributor to resistance, a notion that directly implicates codon bias and the ability of cancer cells to adopt mechanisms that overcome the translational barrier it imposes.

To begin discerning what cellular programs might be involved in upregulating mutant KRAS protein expression, we analysed a

broader panel of KRAS-mutant lung, pancreatic and CRC cell lines, and observed that lines that displayed high endogenous KRAS protein expression (KRAS-high lines) also tended to exhibit elevated expression of an ectopic KRAS transgene when compared to cell lines with low endogenous KRAS protein expression (KRAS-low lines) (Fig. 5c). This was unexpected, as KRAS cDNAs are typically expressed poorly in the absence of codon optimization (refer to Fig. 1), suggesting the intriguing possibility of a mechanism(s) in these cells to overcome poor KRAS expression that is transferable to a plasmid-borne KRAS cDNA. To investigate whether this feature reects a more general strategy by cancer cells dependent on mutant KRAS expression to overcome poor oncogene translation, we next stably expressed an AKT3 cDNA, which is similarly enriched in rare codons and poorly expressed22, in this panel of KRAS mutant cell lines. Interestingly, many KRAS-high lines also displayed elevated levels of AKT3 when compared to their KRAS-low counterparts (Fig. 5d, Supplementary Fig. 4C). By contrast, we did not observe a correlation between expression levels of ectopic KRAS and ectopic NRAS, the latter of which contains approximately equal percentages of rare and common codons (Supplementary Fig. 4D,E). To further test this concept, we stably expressed the gene pair ORMDL3/1, which is analogous to HRAS/KRAS in that it exhibits divergent codon bias and protein expression22, in KRAS-high/low cells. Divergent codon-dependent expression of ORMDL1/3 was completely normalized in the KRAS-high cell line SW900, even in the absence of codon optimization of the rare codon-enriched ORMDL1, but was not observed in the KRAS-low cell line AsPC1 (Fig. 5e). Finally, to compare global translation in these KRAS-high/low lines, a time course of puromycin labelling was used, revealing the rate of global translation to be higher in KRAS-high cells (SW900) than KRAS-low cell cells (AsPC1) (Fig. 5f). Collectively, these data suggest that cancer cells can overcome poor KRAS expression by globally upregulating protein synthesis through mechanisms that also enable cancer cells to overcome the effects of codon bias.

Selective targeting of anti-EGFR resistant colorectal cancer. The above results suggest the intriguing possibility that tumours with acquired KRASG12 or KRASG13 site mutations must overcome poor translation to achieve resistance. To test this hypothesis, we measured the levels of KRAS protein in the parental LIM1215 cell line and in two independently derived, KRASG12/G13 mutant clones selected for resistance to cetuximab13. In both clones, KRAS protein levels were higher than the parental cells (Fig. 6ac). The same result was observed when the experiment was repeated using the cell line OXCO2 (Supplementary Fig. 5A,B). Consistent with our previous observations, upregulated KRAS expression in resistant derivatives was associated with increased global translation rate (Supplementary Fig. 5C). Upregulation of global translation is sufcient to increase KRAS protein expression levels, as ectopic expression of eIF4E, a key factor in protein synthesis, drove increases in KRAS expression in parental LIM1215 cells to levels comparable to those observed in a resistant derivative (Supplementary Fig. 5D). (We note that, as expected, eIF4E-driven upregulation of KRASWT expression was insufcient to drive resistance to cetuximab, as resistance also requires mutational activation of KRAS (Supplementary Fig. 5E).) Given the evidence that cells overcome KRAS codon bias by upregulating translation (Fig. 5), we hypothesized that inhibitors of translation may be selectively potent in CRC cells with acquired KRASG12 or KRASG13 site mutations, thereby revealing a synthetic lethality of the resistant state and a potential target in anti-EGFR resistant disease. We selected a

6 NATURE COMMUNICATIONS | 8:15617 | DOI: 10.1038/ncomms15617 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15617 ARTICLE

a b

NCIH508

****

**

1.5

0.0006

****

****

Cell viability

Empty

KRASQ61R

0.0002

KRASQ61L

Half-maximal cetuximab (M)

0.0004

1.0

KRASG12D

KRASG12V

5105

0.5

4105

3105

2105

1105

0.0

0

13 12 11 10 9 8 7 6 5

Empty

log [Cetuximab(M)]

KRASG12D

KRASG12V

KRASQ61L

KRASQ61R

c d

Empty

KRASG12D

KRASG12V

KRASQ61L

KRASQ61R

2.5

Cetuximab:

p-ERK1/2

Vinculin

Fold

EL-BIM protein (a.u.)

+ + + + +

2.0

1.5

40

40

140

1.0

T-ERK1/2

T-Bim 20

0.5

0.0

Empty

KRASG12D

KRASG12V

KRASQ61L

KRASQ61R

e

Empty

KRASG12D

KRASG12V

KRASQ61L

KRASQ61R

FLAG

Vinculin

29

140

Figure 4 | Model of clinical KRAS-mediated resistance to cetuximab reveals selection of more potent KRAS mutations. (a) Nonlinear regression growth curves depicting cell viability as detected by Cell Titer Glo of NCIH508 cells transduced with a vector encoding no transgene (empty) or the indicated RAS transgene in the presence of cetuximab. (b) Derived half-maximal growth inhibition values of curves in a. (c) Immunoblotting of indicated constructsfor phospho- and total-ERK1/2 and T-BIM in NCIH508 cells. (d) Fold change in extra long BIM (EL-BIM) protein levels from the immunoblots in (c) comparing cetuximab and vehicle treated cells. (e) Immunoblotting of indicated constructs for FLAG-tagged KRAS. In all cases, data are averages.e.m. (n 3 technical

replicates per condition). Images are cropped for clarity. **Pr0.01; ****Pr0.0001. P values were calculated with unpaired, two-tailed Students t-tests.

panel of translational inhibitors, including Rapamycin (an allosteric mTORC1 inhibitor), AZD2014 and MLN0128 (ATP- competitive mTORC1/2 inhibitors), BEZ235 (a dual ATP-competitive mTOR/PI3K inhibitor), and 4EGI-1 (a direct eIF4E inhibitor), and found that in all cases, acquired-resistant derivatives were signicantly more sensitive to these inhibitors. For the case of allosteric and ATP-competitive mTOR inhibitors in particular, many of which are FDA approved or in advanced clinical development, we observed growth inhibition of resistant clones at doses 10- to 100-fold lower than their matched parental counterparts (Fig. 6d). Further, levels of KRAS protein were diminished in resistant cells when compared to parental cells within 4 h of treatment with each translational inhibitor (Fig. 6e). KRAS is differentially required for the growth of resistant cells relative to their parental counterparts as evidenced by short hairpin RNA (shRNA) knockdown experiments (Supplementary Fig. 5F,G). As such, the enhanced growth inhibitory effects of translational inhibitors in resistant cells parallels the effects of KRAS knockdown. Together, these data suggest that in CRC cells with KRASG12/G13 mutation-driven acquired resistance, a translationally primed state overcomes codon bias and facilitates expression of mutant KRAS at levels sufcient to potentiate resistance. Thus, the targeting of this translation-dependent state may provide a novel, selective

therapeutic target for the treatment of anti-EGFR refractory disease.

DiscussionDespite providing a clinically signicant survival benet to patients with metastatic KRAS WT CRC, acquired resistance to the anti-EGFR mAbs cetuximab and panitumumab can arise via three predominant mechanisms: (1) genetic alterations involving downstream EGFR effectors, including KRAS, (2) activation of parallel receptor tyrosine kinase pathways, including HER2 and c-MET and(3) mutations in the extracellular domain of EGFR33. KRAS mutations appear to be the most common driver of acquired resistance to anti-EGFR mAbs. In one study, serial sampling of patient sera revealed that 9 out of 24 (38%) patients with initially KRAS WT tumours developed detectable mutations in KRAS between 5 and 6 months after the initiation of panitumumab monotherapy34. In another study, analysis of metastases from patients who developed resistance to anti-EGFR mAbs showed the emergence of KRAS amplication in one sample and acquisition of secondary KRAS mutations in 60% (6 out of 10) of the cases examined13. Although KRAS has been shown to play a central role in primary and acquired resistance to anti-EGFR therapy in CRC as well as in other therapeutic contexts, no effective pharmacological inhibitors of KRAS, or other RAS oncoproteins, have reached the

NATURE COMMUNICATIONS | 8:15617 | DOI: 10.1038/ncomms15617 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 7

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15617

a

b

Responsive Non-responsive

3.0

KRAS expression (a.u.)

Cetuximab profile of KRASmut CRC cell lines

*

2.5

SW620

SKCO1

FET

2.0

29

KRAS

-actin

LIM2099

RW7213

COGA2

1.5

1.0

45

0.5

0.0

Responsive Non-responsive

c d

AsPC-1

SU.86.86

CFPAC1

H727

H358

SW620

SW480

SW900

HPAFII

Capan-2

40

LoVo

ectKRAS levels

(relative to ACTIN)

30

29

29

FLAG (KRASG12V)

KRAS Ectopic

Endogenous

-actin

20

2=0.6516

P=0.0048

10

0

45

0 500 1,000 1,500

ectAKT3 Levels (relative to ACTIN)

e f SW900 (KRAS-high)

AsPC-1 (KRAS-low)

0 5 10 15 0 5 10 15

SW900 (KRAS-high)

AsPC-1 (KRAS-low)

Time (min)

60

30

45

ORMDL3

ORMDL1

ORMDL3

ORMDL1

ORMDL1op

ORMDL1op

75

MYC

23 KRAS

45 -actin

Puromycin

29 KRAS

-actin

Figure 5 | Expression levels of mutant KRAS correlate with intrinsic resistance to cetuximab and are mediated by translational upregulation.(a) Endogenous KRAS protein expression (normalized to b-actin) in 21 KRAS mutant CRC cell lines classied as responsive (410% growth inhibition following treatment with cetuximab, ten cell lines) or non-responsive (no growth inhibition following treatment with cetuximab, 11 cell lines).

(b) Immunoblot analysis for endogenous KRAS in representative responsive or non-responsive cell lines. (c) Immunblot analysis for endogenous WT KRAS and exogenous FLAG-tagged KRASG12V in a panel of KRAS mutant cell lines. (d) Linear correlation plot of ectopic MYC-AKT3 protein levels and ectopic FLAG-KRASG12V protein levels in cell lines. Immunoblots corresponding to these data are presented in Supplementary Fig. 4. (e) Immunoblotting of indicated MYC-tagged constructs in representative KRAS-high/low lines, where op signies codon-optimized construct. (f) Immunoblotting time course of puromycin labelling in representative KRAS-high/low lines. Images are cropped for clarity. Error bars show datas.e.m. (n 3 technical replicates per

condition). *Pr0.05. P values were calculated with unpaired, two-tailed Students t-tests.

clinic to date35. This stands in contrast with the other predominant mechanisms of anti-EGFR resistance. For example, alternative receptor tyrosine kinase pathways such as HER2 and c-MET can be targeted with selective TKIs36. Similarly, extracellular domain mutations in EGFR, which occur in about 20% of CRC patients treated with anti-EGFR mAbs33, can be treated with oligoclonal antibodies like MM-151 and Sym004, which have been shown to overcome this resistance mechanism in model systems and are now being explored clinically as secondary therapies for patients who have relapsed on anti-EGFR mAbs33,37,38. Thus, there is a particularly compelling need for therapeutic strategies that block KRAS-driven resistance to EGFR inhibitors in CRC.

Our ndings suggest that the resistance-conferring ability of mutant KRAS is limited by codon bias, a mechanism that controls

the expression and subsequent downstream activities of RAS proteins. Indeed, cells with acquired cetuximab resistance driven by canonical KRASG12/G13 site mutations exhibit higher levels of this protein. Alternatively, the selection for more potent KRAS mutations is an alternative mechanism to overcome codon bias, a concept that may explain the paradoxical enrichment of KRASQ61 mutations observed in patients with anti-EGFR refractory disease. Combined, these studies suggest that KRAS-mediated therapeutic resistance may require higher levels of KRAS activity than KRAS-mediated tumorigenesis, an idea that is consistent with other signal amplication-based resistance mechanisms39. To examine these concepts further, future studies to determine the direct relationship between KRAS expression levels in patient tumours and subsequent clinical response to anti-EGFR mAbs are

8 NATURE COMMUNICATIONS | 8:15617 | DOI: 10.1038/ncomms15617 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15617 ARTICLE

a b

1.5

c

LIM1215

P R1 R2

P = LIM1215R1 = CETUX clone (KRASG12R) R2 = CETUX clone (KRASG13D)

*

KRAS

Vinculin

29

140

KRAS expression (a.u.)

1.4

1.0

1.3

Fold

1.2

Cell viability

0.5

1.1

1.0

0.9

0.0

P R1 R2

12 11 10 9 8 7 6 5 4

Cetuximab

d

18

Half-maximal Rapamycin (M)

Half-maximal MLN0128 (M)

5

1.0

16

4

AZD2014 (M)

0.8

Half-maximal

14

3

0.6

0.4

2

0.4

** ***

0.2

1

0.2

** **

*** ***

0.0

0

0.0

P R1 R2

P R1 R2

P R1 R2

Half-maximal BEZ235 (M)

Half-maximal 4EGI-1 (M)

80

60

60

40

40

0.5

* **

0.4

0.3

20

0.2

* *

0.1

0.0

0

P R1 R2

P R1 R2

e

Rapamycin AZD2014 MLN0128 BEZ235 4EGI-1

P R1 R2 P R1 R2 P R1 R2 P R1 R2 P R1 R2

KRAS

Vinculin

29

140

Figure 6 | Translation-dependent state of KRASG12/G13-mutant cells enables selective targeting of anti-EGFR resistant colorectal cancer.

(a) Immunoblot analysis for endogenous KRAS levels in parental (P) and derived KRAS mutant, cetuximab-resistant (R1 and R2) clones. (b) Fold change in KRAS protein levels, normalized to loading control, between resistant derivatives (R1 and R2) and parental cells (P). (c) Nonlinear regression growth curves depicting cell viability as assessed by Cell Titer Glo (CTG) of parental (P) and matched resistant derivatives (R1 and R2) treated with cetuximab.(d) Derived half-maximal growth inhibition values of parental (P) and matched resistant derivatives (R1 and R2) treated with Rapamycin, AZD2014, MLN0128, BEZ235 and 4EGI-1. (e) Immunoblot analysis of KRAS protein levels in parental (P) and resistant derivatives (R1 and R2) following 4 h treatment with the corresponding inhibitors in d. Images are cropped for clarity from the same exposure of the same membrane. Error bars show datas.e.m. (n 3

technical replicates per condition). *Pr0.05; **Pr0.01; ***Pr0.001. P values were calculated with unpaired, two-tailed Students t-tests.

warranted. Similarly, direct comparison of KRAS expression levels before and following the onset of KRASG12-mediated acquired resistance will shed further light on the universality of translational upregulation in this context.

Previously, it has been demonstrated that oncogenic activation of translation initiation and/or elongation can support tumorigenesis by driving selective upregulation of specic mRNA transcripts40. Indeed, seminal studies revealed that overexpression of eIF4E, involved in the translation initiation complex of protein synthesis, was sufcient to drive tumorigenesis in cell lines and spontaneous tumorigenesis in mice4143. In addition, deregulation of translation has more recently been identied as both a primary downstream consequence of oncogenic signalling, as well as a central mediator of resistance to clinical therapies, including drugs targeting the MAPK and PI(3)K-AKT-mTOR signal transduction

pathways44,45. These ndings, along with data showing that cancer cells co-opt translational machinery to support tumour growth, have driven increased interest in targeting translational control as a method to selectively kill cancer cells46.

By treating cells with direct and/or indirect inhibitors of protein synthesis, it is possible to selectively target the translationally primed state that facilitates KRASG12/G13-mediated acquired resistance. Although our studies indicate that these resistant cells exhibit hypersensitivity to diverse inhibitors of translation, future studies are required to determine the reason for this sensitivity and the specic mechanisms imparting greater translation of KRAS, which may lead to more precise and rened approaches to target these tumours. Importantly, however, indirect inhibitors of translation such as rapamycin and MLN0128, and their analogues, are already in clinical development. As such, studies to examine the sensitivity of

NATURE COMMUNICATIONS | 8:15617 | DOI: 10.1038/ncomms15617 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 9

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15617

Table 2 | TRC shRNA sequences.

Construct TRC ID Sequence1. shKRAS-1 TRCN0000033262 CCTATGGTCCTAGTAGGAAAT2. shKRAS-2 TRCN0000040151 CCTACAGGAAGCAAGTAGTAA

patients with KRASG12/G13-mediated resistance to these agents, or the ability of upfront treatment with these agents in combination with anti-EGFR mAbs to shape resistance evolution, are warranted.

Methods

Cell lines and reagents. All cell lines were grown at 37 C in 5% CO2. A375, PC9, NCIH508, UACC-62, H1073, LIM1215, SW620, SW480, SW900, HPAFII, Capan-2, AsPC-1, SU.86.86, CFPAC1, H727, H358, OXCO2 and LoVo were cultured in RPMI supplemented with 10% FBS and 1% penicillin/streptomycin. SKBR3 was cultured in McCoys 5a supplemented with 10% FBS and 1% penicillin/streptomycin. All other cell lines were purchased from American Type Culture Collection (ATCC) or Duke University Cell Culture Facility (CCF). LIM1215, OXCO2 and matched anti-EGFR resistant derivatives, as well as KRAS mutant CRC lines and cetuximab, were used as previously described18. All other drugs were purchased from Selleck Chemicals, ChemieTek, MedChemExpress, Ontario Chemicals, Sigma-Aldrich, or Apex Bio.

Colony formation and soft agar growth assays. For colony formation assays, UACC-62, PC-9, H1073 and NCI-H508 cells selected for stable expression of the indicated RAS constructs were seeded in duplicate into six-well plates at 250 cells per well. Twenty-four hours later, DMSO or the indicated drug (in DMSO)were added to cells. DMSO or drug-containing media was replaced every 2 days and the assays were cultured for 1421 days. Plates were then rinsed with phosphate-buffered saline (PBS), xed in 10% formalin for 5 min and stained with0.1% crystal violet stain for 30 min. Plates were rinsed in distilled water and scanned. ImageJ software was used to quantify colony growth area as a percentage of the well covered. Anchorage-independent growth was assayed in six-well plates with 1 ml of 0.6% bactoagar media solution (nal concentration 1 DMEM, 10%

FBS, 1 penicillin/streptomycin) as a bottom support layer. A total of 4 104

UACC-62 cells selected for stable expression of indicated RAS constructs were resuspended in RPMI (10% FBS, 1 penicillin/streptomycin) and mixed 1:1 with

0.6% bactoagar media solution to give a nal bactoagar concentration of 0.3%,0.2 mM vemurafenib (or equivalent amount DMSO). Cells were then plated in triplicate to give nal density of 1 104 cells per well. Each well was fed with fresh

media and drug on days 7 and 14. Colonies were counted on day 21.

Short-term growth-inhibition assay (GI50). Cells were seeded into 96-well plates at 1,000 cells per well for inhibition assays to cetuximab or at 5,000 cells per well for all other drugs tested. To generate GI50 curves, cells were treated with vehicle (PBS or DMSO) or an eight-log serial dilution of drug. Each treatment condition was represented by at least three replicates. Seven days after cetuximab addition or 3 days after all other drug additions, cell viability was measured using Cell Titer Glo (Promega). Relative viability was then calculated by normalizing luminescence values for each treatment condition to control treated wells. To generate GI50 curves for exogenously expressed KRAS experiments, slight modications were made. Cells were seeded at 300,000 cells per well in six-well plates. The following day cells were infected with the desired retroviral constructs. Following 2 days of puromycin selection (2 mg ml 1), the cells were seeded into 96-well plates at 5,000 cells per well. Doseresponse curves were t using

GraphPad/Prism 6 software.

Western blotting and antibodies. Immunoblotting was performed as previously described47, and membranes were probed with the following primary antibodies at the corresponding dilutions: vinculin, b-actin, BIM, p-ERK1/2, ERK, p-AKT (S473,T308), AKT, eIF4E at 1:1,000 (Cell Signaling); HRAS, NRAS and KRAS at 1:100 (Santa Cruz Biotechnology); FLAG at 1:1,000 (M2; Sigma Aldrich); Myc at 1:1,000 (Invitrogen) and Puromycin at 1:10,000 (Millipore). To determine correlation between ectKRAS and ectAKT3 protein levels, immunoblots were scanned and quantied using ImageJ software. For each cell line analysed, the band density of ectKRAS and ectAKT3 (relative to actin) was determined and plotted graphically. For quantication of EL-BIM, T-BIM and KRAS protein levels, immunoblots were scanned, quantied using ImageJ software, and the band density of each protein (relative to vinculin) was determined and plotted graphically.

Pathway activating screens. Data are excerpted from screens previously reported12,18,23.

Puromycin incorporation. Cell lines were cultured to B70% conuence were treated with DMSO or 5 mM puromycin for the indicated time periods. After incubation, cells were washed with PBS to halt puromycin incorporation. Cells were lysed and subjected to immunoblotting with anti-puromycin antibody (Millipore, 1:10,000).

Lentiviral and retroviral production and infection of cells. Experiments performed using HRAS, KRAS and NRAS point mutants and codon-modied constructs were conducted as previously described22,48. Lentiviral experiments were performed as previously described23,49. In brief, HEK293T cells were transfected with a mixture of VSVG, PsPAX2, the construct of interest, and Opti-MEM and FuGENE 6 Transfection Reagent. The following day, transfection media was replaced with 20 ml of harvest media (RPMI 30% FBS 1% PS)

collected at 48 h post media change. Collected virus was passed through a 0.45 mM lter and frozen at 80 C prior to use in infection of various cell lines. We note

here that the ability to successfully ectopically express oncogenic RAS was variable from cell line to cell line.

shRNA constructs. TRC shRNA clones (Table 2) were obtained from the Duke RNAi Facility as glycerol stocks. Constructs were prepared in lentiviral form and used to infect target cells, as previously described50.

Quantication of apoptosis by Annexin-V. Cells were seeded in 12-well plates and treated the next day with no agent, cetuximab (34.3 10 8 M) or vehicle

(PBS). Cells were incubated for 5 days and then washed twice with ice-cold PBS, and resuspended in 1 Annexin V binding buffer (10 mM HEPES, 140 mM NaCl,

2.5 mM CaCl2; BD Biosciences). Surface exposure of phosphatidylserine was measured using APC-conjugated Annexin V (BD Biosciences). 7-AAD (BD Biosciences) was used as a viability probe. Experiments were analysed at 20,000 counts per sample using BD FACSVantage SE. Gatings were dened using untreated/unstained cells as appropriate.

Statistics. Results are expressed as the meanss.e.m. Unless otherwise specied, for comparisons between two groups, P values were calculated with unpaired, two-tailed Students t-tests.

Data availability. All data generated or analysed during this study are included in this published article (and its Supplementary Information les).

References

1. Ellis, L. M. & Hicklin, D. J. Resistance to targeted therapies: rening anticancer therapy in the era of molecular oncology. Clin. Cancer Res. 15, 74717478 (2009).

2. Tortora, G. et al. Overcoming resistance to molecularly targeted anticancer therapies: rational drug combinations based on EGFR and MAPK inhibition for solid tumours and haematologic malignancies. Drug Resist. Updat. 10, 81100 (2007).

3. Sequist, L. V. et al. Genotypic and histological evolution of lung cancers acquiring resistance to EGFR inhibitors. Sci. Transl. Med. 3, 75ra26 (2011).

4. Engelman, J. A. & Settleman, J. Acquired resistance to tyrosine kinase inhibitors during cancer therapy. Curr. Opin. Genet. Dev. 18, 7379 (2008).

5. Engelman, J. A. Targeting PI3K signalling in cancer: opportunities, challenges and limitations. Nat. Rev. Cancer 9, 550562 (2009).

6. Ciardiello, F. & Tortora, G. EGFR antagonists in cancer treatment. N. Engl. J. Med. 358, 11601174 (2008).

7. Pao, W. et al. KRAS mutations and primary resistance of lung adenocarcinomas to getinib or erlotinib. PLoS Med. 2, e17 (2005).

8. Pao, W. et al. EGF receptor gene mutations are common in lung cancers from never smokers and are associated with sensitivity of tumors to getinib and erlotinib. Proc. Natl Acad. Sci. USA 101, 1330613311 (2004).

9. Paez, J. G. et al. EGFR mutations in lung cancer: correlation with clinical response to getinib therapy. Science 304, 14971500 (2004).

10 NATURE COMMUNICATIONS | 8:15617 | DOI: 10.1038/ncomms15617 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms15617 ARTICLE

10. Lynch, T. J. et al. Activating mutations in the epidermal growth factor receptor underlying responsiveness of non-small-cell lung cancer to getinib. N. Engl. J. Med. 350, 21292139 (2004).

11. Miranda, C. et al. KRAS and BRAF mutations predict primary resistance to imatinib in gastrointestinal stromal tumors. Clin. Cancer Res. 18, 17691776 (2012).

12. Winter, P. S. et al. RAS signaling promotes resistance to JAK inhibitors by suppressing BAD-mediated apoptosis. Sci. Signal 7, ra122 (2014).

13. Misale, S. et al. Emergence of KRAS mutations and acquired resistance to anti-EGFR therapy in colorectal cancer. Nature 486, 532536 (2012).14. Little, A. S. et al. Amplication of the driving oncogene, KRAS or BRAF, underpins acquired resistance to MEK1/2 inhibitors in colorectal cancer cells. Sci. Signal 4, ra17 (2011).

15. Agarwal, A. et al. An activating KRAS mutation in imatinib-resistant chronic myeloid leukemia. Leukemia 22, 22692272 (2008).

16. Hugo, W. et al. Non-genomic and immune evolution of melanoma acquiring MAPKi resistance. Cell 162, 12711285 (2015).

17. Karapetis, C. S. et al. K-ras mutations and benet from cetuximab in advanced colorectal cancer. N. Engl. J. Med. 359, 17571765 (2008).

18. Misale, S. et al. Vertical suppression of the EGFR pathway prevents onset of resistance in colorectal cancers. Nat. Commun. 6, 8305 (2015).

19. De Roock, W. et al. Effects of KRAS, BRAF, NRAS, and PIK3CA mutations on the efcacy of cetuximab plus chemotherapy in chemotherapy-refractory metastatic colorectal cancer: a retrospective consortium analysis. Lancet Oncol. 11, 753762 (2010).

20. Loupakis, F. et al. KRAS codon 61, 146 and BRAF mutations predict resistance to cetuximab plus irinotecan in KRAS codon 12 and 13 wild-type metastatic colorectal cancer. Br. J. Cancer 101, 715721 (2009).

21. Arena, S. et al. Emergence of multiple EGFR extracellular mutations during cetuximab treatment in colorectal cancer. Clin. Cancer Res. 21, 21572166 (2015).

22. Lampson, B. L. et al. Rare codons regulate KRas oncogenesis. Curr. Biol. 23, 7075 (2013).

23. Martz, C. A. et al. Systematic identication of signaling pathways with potential to confer anticancer drug resistance. Sci. Signal 7, ra121 (2014).

24. Buhrman, G., Wink, G. & Mattos, C. Transformation efciency of RasQ61 mutants linked to structural features of the switch regions in the presence of Raf. Structure 15, 16181629 (2007).

25. Buhrman, G., Holzapfel, G., Fetics, S. & Mattos, C. Allosteric modulation of Ras positions Q61 for a direct role in catalysis. Proc. Natl Acad. Sci. USA 107, 49314936 (2010).

26. Amado, R. G. et al. Wild-type KRAS is required for panitumumab efcacy in patients with metastatic colorectal cancer. J. Clin. Oncol. 26, 16261634 (2008).

27. Morelli, M. P. et al. Characterizing the patterns of clonal selection in circulating tumor DNA from patients with colorectal cancer refractory to anti-EGFR treatment. Ann. Oncol. 26, 731736 (2015).

28. Cancer Genome Atlas, N. Comprehensive molecular characterization of human colon and rectal cancer. Nature 487, 330337 (2012).

29. Bettegowda, C. et al. Detection of circulating tumor DNA in early- and late-stage human malignancies. Sci. Transl. Med. 6, 224ra24 (2014).

30. Siravegna, G. et al. Clonal evolution and resistance to EGFR blockade in the blood of colorectal cancer patients. Nat. Med. 21, 795801 (2015).

31. Medico, E. et al. The molecular landscape of colorectal cancer cell lines unveils clinically actionable kinase targets. Nat. Commun. 6, 7002 (2015).

32. Schmidt, E. K., Clavarino, G., Ceppi, M. & Pierre, P. SUnSET, a nonradioactive method to monitor protein synthesis. Nat. Methods 6, 275277 (2009).

33. Arena, S. et al. MM-151 overcomes acquired resistance to cetuximab and panitumumab in colorectal cancers harboring EGFR extracellular domain mutations. Sci. Transl. Med. 8, 324ra14 (2016).

34. Diaz, Jr L. A. et al. The molecular evolution of acquired resistance to targeted EGFR blockade in colorectal cancers. Nature 486, 537540 (2012).

35. Cox, A. D., Fesik, S. W., Kimmelman, A. C., Luo, J. & Der, C. J. Drugging the undruggable RAS: mission possible? Nat. Rev. Drug Discov. 13, 828851 (2014).

36. Misale, S., Di Nicolantonio, F., Sartore-Bianchi, A., Siena, S. & Bardelli, A. Resistance to anti-EGFR therapy in colorectal cancer: from heterogeneity to convergent evolution. Cancer Discov. 4, 12691280 (2014).

37. Sanchez-Martin, F. J. et al. The rst-in-class anti-EGFR antibody mixture Sym004 overcomes cetuximab resistance mediated by EGFR extracellular domain mutations in colorectal cancer. Clin. Cancer Res. 22, 32603267 (2016).

38. Leslie, M. Potential therapy for refractory colon cancer. Cancer Discov. 6, 336337 (2016).

39. Wood, K. C. Mapping the pathways of resistance to targeted therapies. Cancer Res. 75, 42474251 (2015).

40. Truitt, M. L. & Ruggero, D. New frontiers in translational control of the cancer genome. Nat. Rev. Cancer 16, 288304 (2016).

41. Lazaris-Karatzas, A., Montine, K. S. & Sonenberg, N. Malignant transformation by a eukaryotic initiation factor subunit that binds to mRNA 50 cap. Nature 345, 544547 (1990).

42. Lazaris-Karatzas, A. & Sonenberg, N. The mRNA 50 cap-binding protein, eIF-4E, cooperates with v-myc or E1A in the transformation of primary rodent broblasts. Mol. Cell Biol. 12, 12341238 (1992).

43. Ruggero, D. et al. The translation factor eIF-4E promotes tumor formation and cooperates with c-Myc in lymphomagenesis. Nat. Med. 10, 484486 (2004).

44. Boussemart, L. et al. eIF4F is a nexus of resistance to anti-BRAF and anti-MEK cancer therapies. Nature 513, 105109 (2014).

45. Ilic, N., Utermark, T., Widlund, H. R. & Roberts, T. M. PI3K-targeted therapy can be evaded by gene amplication along the MYC-eukaryotic translation initiation factor 4E (eIF4E) axis. Proc. Natl Acad. Sci. USA 108, E699E708 (2011).

46. Bhat, M. et al. Targeting the translation machinery in cancer. Nat. Rev. Drug Discov. 14, 261278 (2015).

47. Wood, K. C. et al. MicroSCALE screening reveals genetic modiers of therapeutic response in melanoma. Sci. Signal 5, rs4 (2012).

48. Pershing, N. L. et al. Rare codons capacitate Kras-driven de novo tumorigenesis.J. Clin. Invest. 125, 222233 (2015).49. Yang, X. et al. A public genome-scale lentiviral expression library of human ORFs. Nat. Methods 8, 659661 (2011).

50. Root, D. E., Hacohen, N., Hahn, W. C., Lander, E. S. & Sabatini, D. M. Genome-scale loss-of-function screening with a lentiviral RNAi library. Nat. Methods 3, 715719 (2006).

Acknowledgements

We thank Channing Der, Adrienne Cox, Carlotta Cancelliere, and members of the Bardelli, Counter, and Wood labs for helpful discussions and technical assistance.

This work was supported by Duke University School of Medicine start-up funds (K.C.W.), a scholar award from the NIH Building Interdisciplinary Research Careers in Womens Health Program (K12HD043446, K.C.W.), a Golfers Against Cancer Research Award (K.C.W.), a Stewart Trust Fellowship (K.C.W.), and research grants from the Pancreatic Cancer Action Network-American Association for Cancer Research (K.C.W.), NIH (U01CA199235 and R01CA207083 to K.C.W., R01CA12031, R01CA154630 and P01CA203657 to C.M.C., and F32CA192715 to E.K.), NSF (DGE-1106401, G.R.A.), the Duke Clinical and Translational Science Award (UL1TR001117, K.C.W.), AIRC 2010 Special Program Molecular Clinical Oncology 5 per mille, Project no. 9970 (A.B.), Fondazione Piemontese per la Ricerca sul Cancro-ONLUS 5 per mille 2011 Ministero della Salute (A.B.), and AIRC IG n. 16788 (A.B.).

Author contributions

Conceptualization: M.A., E.K., C.M.C. and K.C.W.; experimentation: M.A., E.K., G.R.A. and S.J.S.; formal analysis: all authors; writing of original manuscript: M.A., C.M.C. and K.C.W.; writing, review and editing: all authors; resources: S.A. and A.B.; supervision: A.B., C.M.C. and K.C.W. M.A. and E.K. contributed equally as co-rst authors. G.R.A. and S.J.S. contributed equally as co-second authors.

Additional information

Supplementary Information accompanies this paper at http://www.nature.com/naturecommunications

Web End =http://www.nature.com/ http://www.nature.com/naturecommunications

Web End =naturecommunications

Competing interests: The authors declare no competing nancial interests.

Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

Web End =http://npg.nature.com/ http://npg.nature.com/reprintsandpermissions/

Web End =reprintsandpermissions/

How to cite this article: Ali, M. et al. Codon bias imposes a targetable limitation on KRAS-driven therapeutic resistance. Nat. Commun. 8, 15617 doi: 10.1038/ncomms15617 (2017).

Publishers note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional afliations.

This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the articles Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Web End =http://creativecommons.org/licenses/by/4.0/

r The Author(s) 2017

NATURE COMMUNICATIONS | 8:15617 | DOI: 10.1038/ncomms15617 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 11