Functional screening implicates miR-371-3p and peroxiredoxin 6 in reversible tolerance to cancer drugs

Abstract

Acquired resistance to cancer drug therapies almost always occurs in advanced-stage patients even following a significant response to treatment. In addition to mutational mechanisms, various non-mutational resistance mechanisms have now been recognized. We previously described a chromatin-mediated subpopulation of reversibly drug-tolerant persisters that is dynamically maintained within a wide variety of tumour cell populations. Here we explore a potential role for microRNAs in such transient drug tolerance. Functional screening of 879 human microRNAs reveals miR-371-3p as a potent suppressor of drug tolerance. We identify PRDX6 (peroxiredoxin 6) as a key target of miR-371-3p in establishing drug tolerance by regulating PLA2/PKCα activity and reactive oxygen species. PRDX6 expression is associated with poor prognosis in cancers of multiple tissue origins. These findings implicate miR-371-3p as a suppressor of PRDX6 and suggest that co-targeting of peroxiredoxin 6 or modulating miR-371-3p expression together with targeted cancer therapies may delay or prevent acquired drug resistance.

ARTICLE

Received 2 Sep 2015 | Accepted 23 Jun 2016 | Published 3 Aug 2016

DOI: 10.1038/ncomms12351 OPEN

Functional screening implicates miR-371-3p and peroxiredoxin 6 in reversible tolerance to cancer drugs

Nisebita Sahu1, Jean-Philippe Stephan2,w, Darlene Dela Cruz3, Mark Merchant3, Benjamin Haley4, Richard Bourgon5, Marie Classon6 & Jeff Settleman1,w

Acquired resistance to cancer drug therapies almost always occurs in advanced-stage patients even following a signicant response to treatment. In addition to mutational mechanisms, various non-mutational resistance mechanisms have now been recognized. We previously described a chromatin-mediated subpopulation of reversibly drug-tolerant persisters that is dynamically maintained within a wide variety of tumour cell populations. Here we explore a potential role for microRNAs in such transient drug tolerance. Functional screening of 879 human microRNAs reveals miR-371-3p as a potent suppressor of drug tolerance. We identify PRDX6 (peroxiredoxin 6) as a key target of miR-371-3p in establishing drug tolerance by regulating PLA2/PKCa activity and reactive oxygen species. PRDX6 expression is associated with poor prognosis in cancers of multiple tissue origins. These ndings implicate miR-371-3p as a suppressor of PRDX6 and suggest that co-targeting of peroxiredoxin 6 or modulating miR-371-3p expression together with targeted cancer therapies may delay or prevent acquired drug resistance.

1 Department of Discovery Oncology, Genentech, 1 DNA Way, South San Francisco, California 94080, USA. 2 Department of Protein Chemistry, Genentech, 1 DNA Way, South San Francisco, California 94080, USA. 3 Department of Translational Oncology, Genentech, 1 DNA Way, South San Francisco, California 94080, USA. 4 Department of Molecular Biology, Genentech, 1 DNA Way, South San Francisco, California 94080, USA. 5 Department of Bioinformatics and Computational Biology, Genentech, 1 DNA Way, South San Francisco, California 94080, USA. 6 Department of Cancer Targets, Genentech, 1 DNA Way, South San Francisco, California 94080, USA. w Present address: Models and HTS Department, Institut de Recherche SERVIER, 125 Chemin de Ronde, 78290 Croissy-Sur-Seine, France (J.-P.S.); Calico Life Sciences, 1170 Veterans Boulevard, South San Francisco, California 94080, USA (J.S.). Correspondence and requests for materials should be addressed to J.S. (email: mailto:[email protected]

Web End [email protected] ).

NATURE COMMUNICATIONS | 7:12351 | DOI: 10.1038/ncomms12351 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 1

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12351

The relatively rapid emergence of resistance to cancer drug therapies is a major cause of treatment failure for patients18. Numerous mechanisms have now been

identied that contribute to drug resistance, which include both mutational (genetic) and non-mutational (presumably epigenetic) mechanisms that ultimately render tumours unresponsive to a treatment that was initially effective9,10. Drug-tolerant persisters (DTPs) constitute a subpopulation of tumour cells that emerge at relatively high frequency upon treatment of largely drug-sensitive cancer cell populations with various anti-cancer agents11. DTPs are widely observed in various cancer contexts and exhibit a reversible ability to survive otherwise lethal drug exposures, implicating epigenetic regulation1,2,9,10.

Accumulating evidence suggests a role for microRNAs (miRNAs) in epigenetically regulating various phenotypic states in cancer cells1,2,1218. miRNAs can impact genetic programs through post-transcriptional silencing of target genes either by promoting degradation of target messenger RNAs (mRNAs) or by inhibiting their translation19,20. miRNAs have been implicated in regulation of various aspects of cancer biology, including drug resistance, cancer cell stemness, epithelial-to-mesenchymal transition and metastasis19,20. We therefore hypothesize that miRNAs may also regulate genes required to engage the DTP state, and test this possibility by conducting a genome-wide functional screen of miRNAs to identify those with the ability to affect the establishment of the DTP state.

ResultsFunctional screening reveals miR-371-3p as a regulator of DTPs. To investigate whether miRNAs are required to engage or maintain the drug-tolerant state, we functionally screened 879 miRNA precursors and 885 miRNA inhibitors (anti-miRs) for cooperation with anti-cancer drugs to inhibit the emergence of DTPs. In a primary screen, EGFR mutant non-small cell lung cancer cells (PC9 cell line) were transfected with the miRNA libraries and cultured in the presence or absence of the EGFR kinase inhibitor erlotinib for 3 days, followed by a 3-day drug-free recovery phase to identify miRNAs that specically affect DTPs (Fig. 1a). Sixty-seven per cent of miRNA precursors caused general cytotoxicity (o60% viability), even in the absence of drug, suggesting the requirement for a large number of miRNAs in critical biological control networks. Thirty-nine miRNAs (6%) signicantly reduced the number of DTPs formed on erlotinib treatment without having any detectable effect on cell viability in the absence of the drug (Fig. 1b,d; Supplementary Fig. 1i; Supplementary Data 1). In contrast, 55% of the anti-miRs increased DTPs (Supplementary Fig. 1j). Only two of the anti-miRs signicantly decreased DTPs (Fig. 1c,e; Supplementary Fig. 1j; Supplementary Data 2). However, overexpression of these miRNAs in the miR-mimic screen did not increase cell viability. Hence, further validation of ve of the top hits from the miR mimic screen that caused 495% cell death upon overexpression and a 1.52-fold increase in cell viability upon inhibition was undertaken using stable PC9 cells overexpressing each miRNA precursor (pre-miRs).

Erlotinib-treated PC9 cells stably expressing pre-miRs revealed 2 miRNAs (pre-mir371 and pre-miR-548n) that substantially reduced DTPs without affecting the parental population (Fig. 1f; Supplementary Fig. 1a,c). The most robust effect was seen with pre-miR-371, which yields mature miR-371-3p and miR-371-5p isoforms (Fig. 1gi; Supplementary Fig. 1bd). Expression of miR-371-3p was decreased in DTPs in comparison with parental cells and in drug-tolerant expanded persisters (DTEPs), which eventually emerge following continuous exposure of cancer cells to appropriate drugs11, suggesting that miR-371-3p specically plays an important role in the survival of DTPs (Supplementary

Fig. 1e). Overexpressed pre-miR-371 caused similar effects in other cancer cell line models of different tissue origins with distinct drug sensitivities, implicating miR-371-3p in a widely used drug tolerance mechanism (Fig. 1j; Supplementary Data 3). Notably, some cell lines experienced cytotoxicity upon miR-371-3p overexpression in the absence of drug, possibly reecting its importance in cell viability in the non-DTP fraction of some cancers.

Identication of relevant miR-371-3p targets. Next, candidate gene targets of miR-371-3p were identied using the TargetScan prediction algorithm. The KDM5A and IGF1R genes, which encode proteins previously found to regulate DTPs, did not display any 30-untranslated region (UTR) regions that could be potentially targeted by miR-371. Direct regulation by miR-371-3p of the top 70 predicted targets was assessed using a luciferase reporter assay following erlotinib treatment in the presence of miR-371-3p. In addition, KDM-3B, which displayed partial sequence homology in the 30UTR with miR-371, was explored using the 30UTR luciferase reporter assay. However, miR-371 was unable to target the 30UTR of KDM-3B, suggesting that miR-371 does not regulate DTPs via KDM-3B (Supplementary Fig. 1h). Treatment with the histone deacetylase inhibitor trichostatin A also downregulated the expression of miR-371-3p, suggesting that miR-371-3p-mediated regulation of DTP formation is distinct from KDM5A/IGF-1R-mediated regulation (Supplementary Fig. 1c). Signicantly, 21 genes displayed reduced luciferase activity upon erlotinib treatment, implicating these genes as bona de targets of miR-371-3p (Fig. 2a). Microarray analysis of PC9 cells conrmed downregulation of only 12 of the predicted targets upon pre-miR-371 overexpression (Fig. 2b).

We then determined whether short interfering RNAs (siRNAs) directed against the 21 genes could mimic the effects of pre-miR-371 on DTPs (Fig. 2c; Supplementary Data 4). Transfection of siPRDX6, siPLCb4 or siSTX12 into PC9 cells decreased DTPs following erlotinib treatment without having an effect on cell viability in the absence of erlotinib treatment, reproducing the effects of pre-miR-371 (Fig. 2d; Supplementary Figs 2 and 3). These genes were also selectively downregulated by pre-miR-371 overexpression and upregulated by anti-miR-371 expression (Fig. 2e; Supplementary Fig. 1f). Mutation of the single putative miR-371-3p recognition site within PRDX6, PLCb4 and STX12 30UTR sequences abolished the ability of miR-371-3p to inhibit luciferase reporter expression, conrming its direct regulation of these genes and suggesting that a single recognition element is sufcient for their regulation by miR-371-3p (Fig. 2f).

Gene expression analysis revealed increased expression of only PRDX6 and STX12 in DTPs (Fig. 2g,h; Supplementary Fig. 10); however, stable knockouts or knockdowns of PRDX6, STX12 and PLCb4 genes in PC9 cells conrmed their requirement for DTP formation without affecting viability of the overall population in the absence of treatment (Fig. 2i,j; Supplementary Figs 4 and 5). Double knockdown of PRDX6 and PLCb4, and triple knockdown of PRDX6, PLCb4 and STX12 produced the greatest effects on DTPs, implicating PRDX6 and PLCb4 as the key mediators of miR-371-3ps function in drug tolerance (Fig. 2j; Supplementary Fig. 4). Indeed, knockdown of PRDX6, PLCb4 or STX12 in ve additional cancer cell lines similarly affected DTP formation (Supplementary Figs 2a and 3).

Downregulation of PLA2/PKCa activity disrupts drug tolerance. Functional classication of the miR-371-3p targets revealed that four of the genes, including PLCb4 (phospholipase C, beta 4) and

PRDX6 (peroxiredoxin 6), encode proteins associated with phospholipase function21,22 (Supplementary Data 5). In contrast,

2 NATURE COMMUNICATIONS | 7:12351 | DOI: 10.1038/ncomms12351 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12351 ARTICLE

a

Cyquant assay (InCell 6000)

Day 0

Day 2

Day 5

Day 8

2

Media

or

drug

(72 h)

Recovery

(72 h)

Final read out

12

miRNA

mimics/anti-miRs

plates

Transfection

library and controls

Image analysis Nuclei count

b

d

f

200

6 miR-mimics

879 miR mimic

library screen

(in duplicate) 60100% viability

in media

427 miRNAs

>95% cell death

in erlotinib

86 miRNAs

90100% viability in media

>95% cell death in erlotinib

39 miRNAs

>95% cell death

in erlotinib

4

miR-371-3p NTC siTOX

Z score (erlotinib)

% change in DTPs

150

*

*

2

4

6

0

miR-196a miR-575 miR-223 miR-371-3p miR-569 miR-548k miR-548i miR-548n miR-936 miR-409-3p

2

100

0

50

hsa-miR-371-3p

GFP vector

miR-371 prec

0 miR-548n prec

miR-196a prec

miR-575 prec

miR-223 prec

8 50 40 30 20 10 0 10

g

DMSO

Erlotinib

120

P =0.053

Z score (DMSO)

100

*

80

60

c

e

% cell viability

40

>90% cell viability

in media miR-196a miR-575 miR-223 miR-371-3p miR-569 miR-548k miR-548i miR-548n miR-936 miR-409-3p

6

20

885 anti-miR

library screen

(in duplicate)

Anti-miRs

NTC siTOX

Z score (erlotinib)

Anti-miR-371-3p

4

60100% viability

in media

880 miRNAs

>75% cell death

in erlotinib

48 miRNAs

90100% viability in media

>95% cell death in erlotinib

2 miRNAs

2

h

DMSO Erlotinib

2

4

635 30 25 20 15 10 5 0 5 10

siNTC miR-371-3p

0 Anti-miR-371-3p

GFP vector

Pre-miR-371

Z score (DMSO)

i

j

COLO-205 (colon) MKN-45 (stomach)

DMSO

200 DMSO

AZ628 (RAFi)

NCI-H596 (lung) DMSO

200

140

200

Paclitaxel (chemo)

*

120

150

Crizotinib (METi) P =0.0178

P =1.23e5

150

P =1.78e4

miR-371-3p (pmol)

*

*

100

100

% cell viability

% cell viability

% cell viability

150

100

50

50

80

100

60

10

4

3

50

40

5

2

20

1

0

GFP vector

miR371-3p prec

0 Anti-miR371-3p

siNTC

miR-371-3p

0 siNTC

miR-371-3p

0 siNTC

miR-371-3p

Figure 1 | Functional screening identies miR-371-3p as a regulator of DTPs. (a) miRNA mimic or inhibitor screening strategy. (b) miR-mimic library and (c) anti-miR library screen summary. (d) miR mimic screen data and (e) anti-miR screen data plotted as z scores of miRNAs in erlotinib (y axis) versus DMSO (x axis). (f) DTP count in miRNA mimic-expressing PC9 cells. (g) Effects of miR-371-3p, anti-miR371-3p and controls on DTPs.(h) Representative uorescence images of GFP control or pre-miR-371-expressing parental cells and DTPs. Scale bar, 20 mm (i) Expression of miR-371-3p in a pre-miR-371-or anti-miR-371-expressing PC9 line validated by absolute quantitative PCR. (j) miR-371-3p overexpression reduces DTPs in COLO-205,

MKN-45 and NCI-H596 cells upon drug treatment. All experiments were performed in triplicate and data are representative of at least two independent experiments. Data are represented as means.e.m. For f, g and i, *Po0.05, Students t-test. For j, P represents interaction P value.

STX12 (syntaxin 12) is implicated in vesicle-mediated trafcking of EGFR and Src23. We therefore tested inhibitors of phospholipase activity as well as PKCa, a key component of phospholipase signalling. Phospholipase A2 (PLA2) inhibitors (cinnamycin and AACOCF3) and PKCa inhibitors

(Bisindolylmaleimide I, IX, XI and PKC-20-28) reduced DTP formation (Fig. 3a,b; Supplementary Fig. 6a,b). PLA2 activity was also signicantly reduced upon pre-miR-371 overexpression and upregulated by anti-miR-371 expression (Supplementary Fig. 1i). Knockdown of miR-371-3p targets, PRDX6, PLCb4 and STX12,

NATURE COMMUNICATIONS | 7:12351 | DOI: 10.1038/ncomms12351 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 3

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12351

a

b

c

2

1.0

Log2 ratio (targeting / non-

targeting miRNA)

8 6 4 2 0 2 4 6 8

Genes

PRDX6

PLC4

STX12

NTC

siTOX

Z score (erlotinib)

0

* * * * * * * * * * * * * * * * * * * * * *

0.5

Log2 fold change

0.0

* * * * * * * * * * * *

2

0.5

4

1.0

6

1.5

miR-371-3p_3UTR

PIGK

PRDX6

PLCB4

ATP6V1E1

GDE1

PLD6

HMOX1

STX12

TOMM20

SLC37A2

ZNF213

GLT25D1

ZNF576

FAM73B

Empty_3UTR

GAPDH_3UTR

R04_3UTR

WNT7B

REM2

GABBR2

SRXN1

FGFR1

CXorf23

WDR33

PIGK

PRDX6

PLCB4

ATP6V1E1

GDE1

PLD6

HMOX1

STX12

TOMM20

SLC37A2

ZNF213

GLT25D1

ZNF576

FAM73B

WNT7B

REM2

GABBR2

SRXN1

FGFR1

CXorf23

WDR33

25 20 15 10 5 0 5

Z score (DMSO)

d

e

200

% cell viability

DMSO Erlotinib

PLC4

STX12

150

1.5

PRDX6

P =1.73E05

1.5 P =0.003

1.5

P =0.012

100

*

50

1.0

1.0

1.0

30

RQ

RQ

RQ

20

*

0.5

0.5

0.5

**

**

*

10

**

**

**

**

**

**

**

**

**

**

0.0

0.0

0.0

0 NTC siPRDX6 #1

siPRDX6 #2

siPRDX6 #3

siPRDX6 #4

siPLCB4 #1

siPLCB4 #2

siPLCB4 #3

siPLCB4 #4

siSTX12 #1

siSTX12 #2

siSTX12 #3

siSTX12 #4

GFP vector

Pre-miR-371-3p

GFP vector

Pre-miR-371-3p

GFP vector

Pre-miR-371-3p

f

0.5 kb 1 Kb

0.5 kb 1 Kb 1.5 Kb 2 Kb

0.5 kb 1 Kb 1.5 Kb 2 Kb

miR-371 predicted site

TTTTGGCACTAACAT TTTTGTAACTAACAT

3UTR

miR-371 predicted site

3UTR

miR-371 predicted site

3UTR

PRDX6 3UTR WT

PRDX6 3UTR mutant

PLCB4 3UTR WT

PLCB4 3UTR Mutant TTTTGTAACTAACAT

TTTGGGCACTATATT STX12 3UTR WT STX12 3UTR mutant

TTTGGTAACTATATT

TTTTGGCACTAACAT

150

Media+siNTC Erlotinib+siNTC Media+miR371-3p Erlotinib+miR371-3p

150

Media+siNTC Erlotinib+siNTC Media+miR371-3p Erlotinib+miR371-3p

Media+siNTC Erlotinib+siNTC Media+miR371-3p Erlotinib+miR371-3p

150

* *

NS

* * * * *

NS

NS NS NSNS

RLU (%)

100

* * * * *

100

100

* *

RLU (%)

RLU (%)

50

50

50

0

PRDX6 WT

0 PRDX6 Mut

EMPTY_3UTR

miR371_syn

PLCB4 WT

PLCB Mut

EMPTY_3UTR

miR371_syn

STX12 WT

0 STX12 Mut

EMPTY_3UTR

miR371_syn

g h

13.6 *

i j

PC9

PC9

13.4

6

Logintenisty

Parental

DTP

5

5 WT shPRDX6 shPLC4 shSTX12

sh(PRDX6+PLC4) sh(PRDX6+STX12) sh(PLC4+STX12)

sh(PRDX6+PLC4+STX12)

13.2

4

# of DTPs 10

4

10.0

*

(kDa)

3

3

9.6

25 25

37

9.2

PRDX6

STX12

# of DTPs 10

0.6

2

9

PC9

0.4

*

8

*

* *

*

0.2

1

* *

-actin

WT PRDX6 KO

PLC4 KO

0.0

*

*

7

0

*

Parental

DTP

Parental

DTP

Parental

DTP

STX12 KO

PRDX6 PLC4 STX12

Figure 2 | Identication of relevant miR-371-3p targets. (a) Downregulation of luciferase by miR-371-3p upon erlotinib treatment from 30UTR reporters corresponding to candidate miR-371-3p target genes *Po0.05. (b) Microarray analysis of mRNA expression of miR-371-3p targets in pre-miR-371-expressing PC9 cells *Po0.05. (c) siRNA screen data corresponding to miR-371-3p targets plotted as z scores of siRNAs in erlotinib (y axis)

versus DMSO (x axis)-treated cells. (d) Validation of siRNA knockdown effects of miR-371-3p targets, PRDX6, PLCb4 and STX12 on DTPs from PC9 cells. * and ** indicate signicant differences from DMSO- and erlotinib-treated NTC controls, respectively. (e) Quantitative PCR of gene expression of PRDX6,

PLCb4 and STX12 in pre-miR-371-expressing PC9 cells. (f) Mutations of seed sequences of the putative miR-371-3p recognition element in target 30UTRs prevented inhibition by miR-371-3p. (g) PRDX6, PLCb4 and STX12 mRNA, and (h) protein expression in PC9 parental cells and DTPs. (i) DTP count of PC9 cells with PRDX6, PLCb4 and STX12 CRISPR/Cas9 knockouts, and (j) PC9 cells expressing shPRDX6, shPLCb4 and shSTX12 single, double or triple knockdowns. * indicates signicant differences from wild-type (WT) controls. All experiments were performed in triplicate and data are representative of at least two independent experiments. Data are represented as means.e.m. * and **Po0.05, Students t-test. NS, not signicant.

as well as treatment with cinnamycin or bisindolylmaleimide XI, also decreased PLA2 activity (Fig. 3c,d). PKCa activity was reduced in the DTP subpopulation upon pre-miR-371 overexpression, depletion of PRDX6, PLCb4 and STX12, and following co-treatment with erlotinib and bisindolylmaleimide XI (Fig. 3e,f; Supplementary Fig. 10). Knockdown of PRKCA (PKCa)

also decreased DTPs and downregulated PLA2 activity (Fig. 3a,g). Furthermore, DTP formation was increased upon treatment with the PKC agonist, phorbol 12-myristate 13-acetate, implicating PLA2/PKCa signalling as a key determinant of DTP formation

(Fig. 3h). The decrease in PLA2 and PKCa activity on STX12 knockdown was unexpected since its product, Syntaxin 12, has not been previously implicated in these pathways, and its specic function in this context remains unclear.

To determine whether PLA2 and PKCa inhibition targets the same DTP subpopulation within which expression of PRDX6 is upregulated, PC9-derived DTPs expressing GFP (DTP-GFP) were co-cultured with parental cells (PC9-RFP) in a 2.5:1 ratio and treated with cinnamycin or erlotinib and cinnamycin for 6 days. Cinnamycin reduced DTPs when cultured in the absence of

4 NATURE COMMUNICATIONS | 7:12351 | DOI: 10.1038/ncomms12351 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12351 ARTICLE

a

b

PC9

COLO-205 HCC-1954

# of DTPs 10

HCC-1954 MKN-45

30 P = 3.23e8 P = 0.0164 P = 8.95e6

COLO-205

40

80

40

P = 5.84e7 P = 0.00383 P = 0.0186

40

# of DTPs 10

4

# of DTPs 10

# of DTPs 10

30

60

30

30

3

20

# of DTPs 10

20

# of DTPs 10

40

20

20

2

1.0

2

2

1.5

1.0

0.6

1.0

0.4

0.5

1

1

0.5

0.5

0.2

0.0

0

0.0

DMSO

Erlotinib

Bis XI

Erl+ Bis XI

Bis XI

GDC-0980+Bis XI

DMSO

0 AZ628

Bis XI

AZ628+Bis XI

DMSO

GDC-0980

0.0 AZ628 + Cinn

DMSO

AZ628

0.0 GDC-0980 + Cinn

Cinn

DMSO

GDC-0980

Cinn

DMSO

Crizotinib

Cinn

Criz+ Cinn

c

d

e

150

150

150

150

150

150

RFU

(PLA2 activity)

RFU

(PLA2 activity)

RU

(PKC kinase activity)

RU

(PKC kinase activity)

100

100

100

100

100

100

*

* * *

RU

(PKC kinase activity)

RU

(PKC kinase activity)

*

* *

*

50

50

* *

*

50

50

50

* * *

50

WT siPKC

siPLC4

siSTX12

0 DMSO

Cinn

Bis XI

PC9 parental

0 PC9 DTP

GFP vector

0 pre-miR371

0 WT

PRDX6 KO

PLC4 KO

STX12 KO

DMSO

0 Erlotinib

Bis XI

Erl+Bis XI

f

g h

PRDX6 KO

PRKCA (PKC)

100 * *

*

PC9

150

% cell viability

DMSO Erlotinib

30

# of DTPs 106

P = 3.78e8

WT

PLC4 KO

*

20

(kDa) (kDa) (kDa)

WT

STX12 KO

DTP

PRDX6

WT

STX12

50

10

(kDa) 75 75 37

PLC4

25

25

75 75 37

30

4

p-PKC

PKC -Actin

Parental

PC9

PC9

p-PKC

PKC -Actin

PC9

p-PKC

PKC -Actin

75 75

37

PC9

100 p-PKC PKC -Actin

20

NS

75 75

37

2

10

**

0 NTC siRNA #1

siRNA #2

** siRNA #3

** siRNA #4

DMSO

0 Erlotinib

PMA

Erl+PMA

i

Erlotinib Cinnamycin Erl+Cinn

DTP (GFP)+

j

T : Top panel M : Middle panel B : Bottom panel

DTP (GFP) DTP (GFP)

parental (RFP) DTP (GFP) Parental (RFP)

T

M

B

DTP GFP (in PC9 RFP mix)

PC9 RFP (in DTP GFP mix)

DTP GFP PC9 RFP

300 *

*

* *

*

300

200

*

% cell viability

% cell viability

% cell viability

200

200

150

100

*

100

100

10

20

15

8

5

NS

10

5

4

Erlotinib

0 Cinn

Erl +Cinn

Erlotinib

0 Cinn

Erl +Cinn

Erlotinib

0 Cinn

Erl +Cinn

Figure 3 | Downregulation of PRDX6/PLCb4-associated PLA2/PKCa activity inhibits drug tolerance. (a) Co-treatment with bisindolylmaleimide XI reduces DTPs in PC9, COLO-205 and HCC-1954 cells. (b) Co-treatment with cinnamycin reduces DTPs in COLO-205, HCC-1954 and MKN-45 cells.

Decreased PLA2 activity upon (c) siRNA knockdown of PRDX6, PLCb4 or STX12 and (d) treatment with PLA2 inhibitors cinnamycin and the PKCa inhibitor bisindolylmaleimide XI. * indicates signicant differences from wild-type (WT) or DMSO controls. (e) PKC activity in PC9 cells and DTPs, stable cell lines expressing pre-miR-371 or PRDX6, PLCb4 or STX12 knockouts and upon treatment with bisindolylmaleimide XIerlotinib. * indicates signicant differences from DMSO or WTcontrols. (f) Immunoblot of PKCa activation in PC9 DTPs and PC9 cells with PRDX6, PLCb4 or STX12 knockouts. (g) siPRKCA knockdown in PC9 cells reduces DTPs. * and ** indicate signicant differences from DMSO- and erlotinib-treated NTC controls respectively. (h) Co-treatment with the

PKC activator phorbol-12-myristate13-acetate (PMA) increases DTPs in PC9 cells i, Fluorescence microscopy images of surviving PC9 DTPs co-cultured with parental cells (T: top panel), PC9 DTPs (M: middle panel) or parental cells (B: bottom panel) after 6-day treatment with cinnamycinerlotinib, Scale bar, 50 mm and (j) viable cell population counts of imaged cells in f on day 6 of treatment. All experiments were performed in triplicate and data are representative of at least two independent experiments. Data are represented as means.e.m. For ce and g, * or ** Po0.05, Students t-test,

NS, not signicant. For a, b and h, P represents interaction P value.

parental cells, whereas viability of parental cells was unaffected (Fig. 3i,j). Co-treatment with erlotinib and cinnamycin reduced both populations under single as well as co-culture conditions. Sequential treatment with PLA2 or PKCa inhibitors (before or after erlotinib) demonstrated that co-treatment most effectively reduced DTPs, suggesting that dual inhibition of the EGFR and PLA2/PKCa signalling pathways is required for efcient targeting

of both the parental and DTP population (Supplementary Fig. 6c,d). These inhibitors similarly affected DTPs in other tested cancer cell models (Fig. 3a,b). PLA2/PKCa inhibition also prevented the establishment of actively proliferating DTEPs (Fig. 4a,b). DTEP formation was also reduced upon stable miR-371 expression or following knockout or knockdown of PRDX6, PLCb4 and STX12 (Fig. 4a,b).

NATURE COMMUNICATIONS | 7:12351 | DOI: 10.1038/ncomms12351 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 5

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12351

a b

PC9 Pre-miR371 PRDX6 KO PLC4 KO STX12 KO

shPRDX6

600

Erlotinib

500

# of DTEP colonies

shPLC4 shSTX12 shPRDX6

+shPLC4

shPRDX6 +shSTX12

shPLC4 +shSTX12

shPRDX6+ sh(PLC4+STX12)

400

300

150

Erlotinib

100

50

Cinn Erl+Cinn Bis XI

Erl+Bis IX Erl+Bis XI

Bis IX

* *

* * *

* * * * *

*

* *

*

PC9

0 WT

miR-371

PRDX6 KO

PLCB4 KO

STX12 KO

shPRDX6

shPLCB4

shSTX12

sh(PRDX6+PLCB4)

sh(PRDX6+STX12)

sh(PLCB4+STX12)

sh(PRDX6+PLCB4+STX12)

Cinnamycin

Erl+Cinn

Bis IX

Erl+Bis IX

Bis XI Erl+Bis XI

PMA

Erl+PMA

PMA Erl+PMA

PC9

c d

Erlotinib (day 13)

e

PC9+GFP vector (vehicle) PC9+GFP vector (erlotinib) PC9+pre-miR371 (vehicle) PC9+pre-miR371 (erlotinib)

0 0 4 8 12 16 20 24 28

600

250 GFP vector Pre-miR-371

6

GFP vector (vehicle) GFP vector (erlotinib)

*

*

200

Tumour volume (mm)

Tumour volume (mm)

500

150

4

RQ

400

100

2

50

300

0

0

PRDX6 PLC4 STX12

No treatment (day 25)

Tumour volume (mm)

200

600 GFP vector Pre-miR-371

2.0

Pre-miR-371 (vehicle) Pre-miR-371 (erlotinib)

100 *** *

500

1.5

**

400

300

RQ

1.0

200

0.5

100

Erlotinib (day 013)

No treatment

(day 1325)

0

0.0

PRDX6 PLC4 STX12

f g

1.0

(Lung)

HR = 1.75 (1.372.23) logrank P = 6e-06

1.0

0.0

HR = 1.28 (1.041.57)

(Gastric) logrank P = 0.021

0 50 100 150 Time (months)

Probability (overall survival)

Probability (overall survival)

0.8

0.8

miR-371-3p

PRDX6

0.6

0.6

0.4

0.4

0.2

Expression

0.2

Expression

Low PRDX6 (n =360) High PRDX6 (n =360)

0.0

Low PRDX6 (n =191) High PRDX6 (n =402)

0 50 100 150 200

PLA2

ROS

Time (months)

1.0

0.0

(Colorectal) HR = 2.62 (1.146.07) P value = 0.0255

Expression Low PRDX6 (n =63) High PRDX6 (n =114)

(Breast) HR = 1.75 (1.013.04) P value = 0.0086

Probability (overall survival)

Probability (disease-free survival)

1.0

0.0

0.8

0.8

PKC

0.6

0.6

0.4

0.4

DTP survival

0.2

0.2

Expression Low PRDX6 (n =222) High PRDX6 (n =27)

0 20 40 60 80 100 120Time (months)

140

0 2 4 6 8 10 12

Time (years)

Figure 4 | PRDX6-mediated PLA2/PKCa activity is required for drug tolerance. (a) DTEP formation after erlotinib treatment is reduced in PC9 cells on pre-miR-371 overexpression, single, double or triple knockdowns of PRDX6, PLCb4 or STX12, knockouts of PRDX6, PLCb4 or STX12 and treatment with

PLA2/PKCa inhibitors or phorbol-12-myristate13-acetate (PMA; n 3). (b) DTEP colony counts obtained in a. Data are representative of at least two

independent experiments. Data are represented as means.e.m. * (Po0.05, Students t-test) indicate signicant differences from wild-type (WT) controls. (c) Xenograft studies with PC9 GFP and PC9 pre-miR-371 cell lines using erlotinib (n 15). Data are represented as means.e.m. Interaction

P values *P 0.0288, **P 5.17e 05, ***P 0.00408. (d) Change in tumour volumes during drug treatment and after treatment cessation.

(e) Quantitative PCR for expression of PRDX6, PLCb4 and STX12 mRNAs in xenograft tumours (n 5). *Po0.05, Students t-test. (f) KaplanMeier plots of

overall survival of lung adenocarcinoma, colorectal cancer, gastric cancer and breast cancer patients; patient groups were separated based on PRDX6 mRNA expression. (g) Model depicting proposed mechanism for regulation of DTP survival via miR-371-3p and PRDX6.

6 NATURE COMMUNICATIONS | 7:12351 | DOI: 10.1038/ncomms12351 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12351 ARTICLE

In addition to its PLA2 activity, PRDX6 has a second functionregulation of oxidative stress via a peroxidase activity, and could potentially contribute to drug tolerance through this additional mechanism21,2426. Notably, our previous studies demonstrated that increasing the levels of reactive oxygen species (ROS) selectively promotes cell death within the DTP subpopulation27. In PC9 cell lines expressing the three miR-371-3p target knockouts, as well as in DTP populations, ROS levels were decreased upon erlotinib treatment (Supplementary Fig. 7a,b). However, PC9 cells with stable knockout of PRDX6 and STX12 showed increased baseline ROS levels relative to parental and PLCb4 knockout cells, which were reduced in all of the lines upon erlotinib treatment (Supplementary Fig. 7b). Furthermore, low ROS levels were observed upon overexpression of miR-371-3p in the absence of drug treatment, but led to a signicant increase in ROS levels on erlotinib treatment in comparison with the parental cell line (Supplementary Fig. 7c). These data further support our observations with knockout cell lines for miR-371-3p target genes, PRDX6 and STX12, and suggest that miR-371-3p overexpression targets PRDX6 and regulates ROS levels only upon cytotoxic drug treatment. Thus, increased baseline ROS levels due to the reduced PRDX6 and STX12 activity could contribute to increased cell death during drug treatment, further implicating ROS modulation in DTP survival.

miR-371 overexpression enhances drug-induced tumour regression. We next assessed the role of pre-miR-371 overexpression in drug tolerance in vivo using PC9 and MKN-45 xenograft mouse models. Mice harbouring tumours derived from PC9 GFP

vector and PC9 pre-miR-371 cell lines were treated with

erlotinib. The growth rate of PC9 tumours was moderately reduced upon pre-miR-371 expression in the vehicle-treated group (Fig. 4c). Regression was observed in both PC9 GFP

vector and PC9 pre-miR-371-derived tumours upon erlotinib

treatment, but was signicantly enhanced in the pre-miR-371-expressing tumours. After stopping erlotinib treatment for 12 days, tumour relapse was observed in 80% of mice harbouring PC9 GFP-derived tumours (Fig. 4c,d). In contrast, only 38% of

mice with pre-miR-371-expressing tumours relapsed. Furthermore, expression of PRDX6, PLCb4 and STX12 was upregulated in erlotinib-treated PC9 GFP tumours and was downregulated

in erlotinib-treated PC9 pre-miR-371 tumours, suggesting that

downregulation of these genes contributes to tumour regression and prevention of tumour regrowth during drug treatment (Fig. 4e). Similar effects were observed in mice harbouring tumours derived from MKN-45 pre-miR-371 cell lines, in

contrast to MKN-45 GFP vector-bearing mice, upon treatment

with the MET kinase inhibitor crizotinib and upon stoppage of treatment (Supplementary Fig. 2b,c). Notably, The Cancer Genome Atlas (TCGA) analysis of patient samples of several cancer types also revealed that relatively high expression of the miR-371-3p target PRDX6, but not PLCb4 or STX12, signicantly correlated with poor prognosis in lung adenocarcinoma, colorectal, gastric and breast cancer patients (Fig. 4f; Supplementary Figs 8 and 9).

DiscussionTogether, these ndings reveal the miR-371-3p target gene PRDX6 as a key regulator of the reversible drug tolerance that frequently emerges within heterogeneous cancer cell populations. miR-371-3ps role in drug tolerance reects its regulation of genes that encode components of both PLC/PKCa signalling and oxidative stress management. Recent studies have also implicated PKCa as the central regulatory node in breast cancer cells that are

drug resistant and contribute to the emergence of more aggressive and treatment-refractory tumours28. The identication of PKCa as a regulator of drug tolerance implicates PKCa as a key driver of the emergence of drug-tolerant subpopulations of cancer cells. A PKC-dependent mechanism has also been recently reported to confer resistance to ALK inhibition in lung cancer29. In conclusion, our ndings reveal a major regulatory role for miRNAs in the emergence of reversible drug tolerance, and suggest that combining current targeted therapies with inhibitors of PLA2/PKCa and ROS accumulation could provide a treatment strategy to delay some types of drug resistance in human cancer.

Methods

Cell lines and reagents. PC9 cells were from Kyushu University, Japan. EVSA-T and MKN-45 cells were from DSMZ. All other cell lines were from American Type Culture Collection. All cell lines were banked at the Genentech cell line core facility that routinely performs single nucleotide polymorphism (SNP) and short tandem repeats (STR) analysis to conrm cell line identity and detect possible mycoplasma contamination. PC9 cells were cultured in RPMI medium (Gibco) containing4.5 g l 1 of D-glucose and supplemented with 10% fetal bovine serum, 2 mM L-glutamine, and 100 U ml 1 of penicillin and streptomycin. All other cell lines were maintained in RPMI medium (Gibco) supplemented with 10% fetal bovine serum, 2 mM L-glutamine, and 100 U ml 1 of penicillin and streptomycin. Lapatinib was from LC Laboratories, cinnamycin was from Sigma, AACOCF3 was from Santa

Cruz Biotechnology, paclitaxel was from US Biologicals, carboplatin was from Selleckchem, AZ628 was from Tocris Bioscience, Bisindolylmaleimide I, XI and XI were from Cayman Chemicals, PKC-20-28 was from EMD Millipore, phorbol-12-myristate13-acetate was from VWR Scientic, and erlotinib and GDC-0980 were synthesized at Genentech.

miRNA screen. PC9 cells were transfected with the Dharmacon miRIDIAN human miRNA mimic library (CS-001010; 879 miRNA mimics) or the Dharmacon miRIDIAN human miRNA inhibitor library (IH-001010; 885 anti-miRs) in 96-well plates. Reverse transfections were performed using Dharmafect 1 in duplicate with12.5 nM of each miRNA or anti-miR. A total of 1,000 PC9 cells were plated per well. Forty-eight hours post transfection, media was replaced with media containing dimethylsulphoxide (DMSO) or 1 mM erlotinib and cells were incubated for 72 h. Drug-containing media was then removed and cells were cultured for another 72 h in drug-free medium, since DTPs were detectable in this short-term treatment regimen, which also enabled the robust quantication of viable drug-tolerant cells with good signal-to-noise ratio. Drug-tolerant cells were stained using the CyQUANT Direct Cell Proliferation Assay kit (Life Technologies) and cell counts were performed using an automated uorescent microscope (In Cell Analyzer 6000; GE Healthcare). Cell viability for miR/anti-miRs-expressing cells was determined by calculating the ratio: (cell count in miR- or anti-miR-expressing cells)/(cell count in cells expressing non-targeting control (NTC) in DMSO). The z score (the number of s.d.s from the mean) for each miRNA was calculated using Screensifter. Twenty-six of the top hits obtained from the screens were tested in eight additional cancer cell linesCOLO-205, MKN-45, NCI-H596, EVSA-T, M14, HCC-1954, NCI-H441 and SKBr3. COLO-205, EVSAT and HCC-1954 cells were transfected with RNAiMax, NCI-H596 and SKBr3 cells were transfected with Dharmafect 1, M14 cells were transfected with Dharmafect 3, and MKN-45 cells were transfected with Dharmafect 4. Drug treatments used for these cell lines were: COLO-205 (AZ628, 1 mM), MKN-45 (crizotinib, 1 mM), NCI-H596 (paclitaxel,6 mM), EVSA-T (GDC-0980, 2.5 mM), M14 (AZ628, 1 mM), HCC-1954 (GDC-0980, 2.5 mM), NCI-H441 (carboplatin, 2.5 mM) and SKBr3 (lapatinib, 1 mM).

30UTR luciferase reporter screen. Human 30UTRs of potential targets for miR371-3p were selected from the LightSwitch genome-wide collection of 30UTR luciferase reporters (Active Motif). A total of 89 30UTR luciferase fusion reporters were selected from the genome-wide collection based on the presence of one or more predicted hsa-miR-371-3p target sites. An additional ve reporters were used as negative controls, determined by a lack of predicted hsa-miR-371-3p target sites. The negative set included the empty vector, two 30UTR constructs from housekeeping genes and two controls containing a random, non-genic, non-conserved sequence. A synthetic miRNA target sequence consisting of sequence repeats that are fully complementary to miR-371-3p and cloned downstream of the RenSP reporter gene in the pLightSwitch_3UTR vector was used as a positive control. Ninety-six-well plates were seeded with 15,000 PC9 cells 24 h before transfection to achieve 80% conuency at the time of transfection. Each transfection included0.3 ml of DharmaFECT DUO transfection reagent (Dharmacon), 100 ng of 30UTR reporter and mimic or non-targeting control miRNA (Dharmacon) to yield a nal concentration of 50 nM in a total volume of 100 ml per well. Each construct was transfected in triplicate separately with either the miR-371-3p mimic or the non-targeting control. After 24 h, cells were treated with 1 mM erlotinib for an additional 24 h. A measure of 100 ml of LightSwitch Luciferase Assay Reagent (Active Motif) was then added to each well, plates were incubated at room temperature for 30 min

NATURE COMMUNICATIONS | 7:12351 | DOI: 10.1038/ncomms12351 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 7

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12351

and then read on a SpectraMax L luminometer (Molecular Devices). To identify genes that were signicantly repressed, a P value (two-tail t-test) and log2-ratio was calculated for each reporter from the average luminescence values of the mimic and non-targeting control transfections.

siRNA screen. A siRNA screen was performed following the same protocol as the miR screen using siRNAs corresponding to 21 hits short-listed from the 30UTR luciferase reporter screen. Three or four single siRNAs per gene were obtained from Dharmacon for all the genes. Cell counts were performed after cells were stained with the CyQUANT kit using the In Cell Analyzer 6000. The z score (the number of standard deviations from the mean) for each siRNA was calculated using Screensifter. Cell viability was also measured at the end of the screen using the Cell Titer-Glo Luminescent Cell Viability Assay kit (Promega). Genes down-regulated with two or more siRNAs and caused o40% cell death in the absence of drug treatment and 490% cell death on erlotinib treatment were considered for further analysis. The target sequences of siPRKCA (PKCa) used are as follows: siPRKCA #1: GAAGGGTTCTCGTATGTCAsiPRKCA #2: GGACTGGGATCGAACAACAsiPRKCA #3: GGATTGTTCTTTCTTCATAsiPRKCA #4: TAAGGAACCACAAGCAGTA

Site-directed mutagenesis. Mutations of seed sequences for miR-371-3p were generated using a modication of the QuikChange (Stratagene) protocol. Five nucleotides were mutated within a single target site in each 30UTR reporter construct. After sequence conrmation, mutant reporters were tested alongwith wild-type controls in triplicate using the same protocol used for the 30UTR luciferase reporter assay screen.

Generation of DTPs and DTEPs. DTPs were generated by plating 106 cells in 10 cm plates and after reaching 80% conuency, treating with 1 mM of relevant drugs for 9 days, replacing with fresh media containing drug every 72 h. The number of DTPs was determined using a Nexcelom Cellometer. DTEPs were generated by continuing the treatment regimen used for generating DTPs for30 days. After 30 days, DTEP colonies were washed with PBS, and xed and stained with 0.5% crystal violet containing 20% methanol. Individual DTEP colonies were counted and in cases where colonies were too numerous to count, they were given an arbitrary number of 500 colonies in the graphs plotted. All data represent means.e.m. of three independent experiments.

Immunouorescence microscopy. PC9 cells stably expressing GFP were plated in six-well plates and treated with 1 mM erlotinib for 9 days. On day 9, the number of DTPs formed was counted by uorescent microscopy using an In Cell Analyzer 6000. PC9 cells stably expressing RFP were then plated at a 2.5:1 ratio with DTPs. After 24 h, cells were treated with 1 mM erlotinib and/or cinnamycin for 6 days.

After 6 days, surviving DTP and parental cells were counted by uorescent microscopy using an In Cell Analyzer 6000.

PKC activity assay. Cultured cells were treated for 24 h with relevant drugsand lysates were collected after 24 h. Cell lysates were prepared using kinase buffer (20113, Millipore) containing protease inhibitor cocktail (ab65621, Abcam)and sonicated for 8 20 s intervals. Protein concentration was then quantied

by BCA assay and 5 mg of protein lysates was used for each assay. ThePKC activity assay was performed as per the manufacturers instructions (ADI-EKS-420A, Enzo) and measured at an absorbance of 450 nm. Data were normalized to values from control, DMSO-treated cells.

PLA2 activity assay. The PLA2 assay was conducted with cell lysates obtained using the same protocol described for the PKC activity assay. An amount of 5 mg of protein lysates was used and the assay was performed as per the manufacturers instructions (E10217, Invitrogen). Fluorescence was measured at 450 nm using a SpectraMax M5 (Molecular Devices) and normalized to control cells.

Cellular ROS detection assay. Cells were treated with DMSO or 1 mM erlotinib for 5 h and ROS levels were detected using the ROS detection agent, DCFDA,as per the manufacturers instructions (ab113851, Abcam). Fluorescence was measured with excitation wavelength at 485 nm and emission wavelength at535 nm using a SpectraMax M5 (Molecular Devices) and normalized to untreated PC9 parental cells. ROS levels in PC9 parental and DTP populations were measured by ow cytometry after performing the assay as per the manufacturers instructions (C10491, Molecular Probes).

Quantitative RTPCR. For miRNA isolation, an RNA isolation kit (Exiqon) was used and complementary DNA was generated using a universal cDNA synthesis kit (Exiqon). PCR with reverse transcription (RTPCR) was conducted using an ExiLENT SYBR Green Master Mix kit and LNA primer sets for miR-371-3p(cat #204299) as per the manufacturers instructions. Absolute quantication

of miR-371-3p was performed via standard curve method using synthetic oligonucleotides for miR-371-3p from Exiqon. RTPCR was performed using an ABI Prism 7500 Real-Time PCR System as per the manufacturers instructions.

For mRNA isolation, an RNAeasy kit (Qiagen) was used and RTPCR was performed using Taqman RNA-to-CT kit (Applied Biosystems) and Taqman probes from Applied Biosystems for PRDX6 (Hs00705355_m1), PLCb4 (Hs00168656_m1), STX12 (Hs00295291_m1) and GAPDH (Hs02758991_g1). RTPCR was performed using an ABI Prism 7500 Real-Time PCR System. mRNA results were normalized to GAPDH and RQ was analysed using the ABI software.

CRISPR/Cas9-mediated knockouts. Human codon optimized S. pyogenes Cas9 was cloned into a pRK vector and expressed via the human cytomegalovirus (CMV) immediate-early promoter. Individual sgRNAs were cloned downstream of and expressed from the human U6 promoter of the pLKO.5 vector (Sigma, product SHC-201). A total of 6 105 cells were plated in six-well plates. After 24 h, cells

were transfected with Cas9 and sgRNA constructs using Lipofectamine 2000. After 5 days of transfection, cells were selected with 2 mg ml 1 of puromycin for 24 h.

Genomic DNA was isolated using a DNA Isolation kit (Qiagen) and deletion of the segment between each sgRNA pair was validated by performing PCR with locus-specic primers. Single-cell clones were obtained by sorting cells using a BD FACSAria Cell Sorter.

The following sgRNAs were used for generating knockouts in PC9 cells.

Gene Target sequence 1 Target sequence 2PRDX6 TTTCCTCCCAGTCTTGCCATAGG TTTCTAGCACGCAAGTACACTGG PLCb4 GCTCCCTCATCAACAGTATTCGG TTTGAGTATAATGGATCGTGCGG

STX12 CCAGTTTCTCCGTCAGCCTGCGG AGGCGGGAGGATCAGTATATCGG

The following primers were used to validate CRIPSR/Cas9-mediated knockout in genomic DNA using PCR.

Gene Forward sequence Reverse sequencePRDX6 GTGCACACCTCCACAATCTT GCACAGGTTTAAAAGAATGAAGC PLCb4 CCCTGGGACTTAATTCGATG TCCCTCAAACCCAGCTTAAA

STX12 GTTCCGGGTCTGAGTTTCAA CTGTGGAAAAGGGAAGTGGA

Lentiviral infection. Lentiviral particles were generated as described previously30. For miRNA overexpression, human pre-miRNA Expression Constructs from System Biosciences that were used are: (Lenti-miR-371: PMIRH371PA-1), (Lenti-miR-196a: PMIRH196a1PA-1), (Lenti-miR-575: PMIRH575PA-1), (Lenti-miR-223: PMIRH223PA-1), (Lenti-miR-548n: PMIRH548nPA-1) and (Scramble control hairpin in pCDH-CMV-MCS-EF1-copGFP (CD511B-1): PMIRH000PA-1). For shRNA knockdowns, shPLCb4 and shSTX12 constructs in a lentiviral GFP vector were obtained from Origene, and shPRDX6 constructs were obtained from Genecopoeia.

Protein expression analysis and antibodies. Cell lysates were prepared using RIPA lysis buffer (Thermo Scientic) containing a protease inhibitor cocktail (Thermo Scientic) and protein concentration was measured using the BCA assay (Thermo Scientic). An amount of 5 mg total protein was resolved by

SDSpolyacrylamide gel electrophoresis and transferred to the nitrocellulose membrane (Bio-Rad). Immunoblotting was performed using 1:1,000 dilutionsof the following antibodies, p-PKCa (sc-12356, Santa Cruz Biotechnology), PKCa (sc-208, Santa Cruz Biotechnology), PRDX6 (13585-1-AP, ProteinTech), PLCb4 (sc-404, Santa Cruz Biotechnology), STX12 (Syntaxin 12) (sc-368438, Santa Cruz Biotechnology) and b-actin (3700S, Cell Signaling).

Proliferation assays. Kinetic proliferation assays were performed by live-cell imaging using the IncuCYTE system (Essen Biosciences). Cells were plated in 96-well plates in triplicates and a proliferation time course was measured in the IncuCYTE for 72hrs. Data points were taken every 2 h.

Bioinformatics. Predicted miR-371-3p target-binding sites were determined using TargetScan Human v.6.2. Gene functional classication for potential miR371-3p targets was performed using DAVID Bioinformatics Resources 6.7 (ref. 31; http://david.abcc.ncifcrf.gov/home.jsp

Web End =http://david.abcc.ncifcrf.gov/home.jsp ).

Statistical analysis. All data are represented as means.e.m, unless otherwise specied. All replicates (n) were biological and represent two or three independent experiments. Interaction testing was performed for assessing interactions between miRNA and drug (Fig. 1i and xed time points in Fig. 4c) or between two drugs (Fig. 3hj). Standard linear models to regress log-scale response were used against the two main effects plus their interaction. The Wald statistic associated with the interaction term was used for an interaction P value. For all other gures, Students t-test (two-tailed) was used to compare two groups and to calculate P values. Po0.05 was considered signicant. All data were normally distributed and the variances were similar between the groups being statistically compared. The sample

8 NATURE COMMUNICATIONS | 7:12351 | DOI: 10.1038/ncomms12351 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12351 ARTICLE

size for in vitro studies was based on the previous experience with experimental variability and no statistical method was used to predetermine the sample size. For in vivo xenograft studies, 1015 tumour-bearing mice per experimental or control cohort were used based on power analyses that called for an n of 7 or greater to achieve a condence level of 90% as per the variation in tumour volume between different groups and s.d. of tumour volumes within each group. No samples were excluded from the analysis. The investigators were not blinded to the group allocation during the experiment or for outcome assessment.

Xenograft studies. All studies involving animals complied with protocols approved by the Institutional Animal Care and Use Committee at Genentech and all studies were carried out in an Association for the Assessment and Accreditation of Laboratory Animal Care-accredited facility. For PC9 GFP vector and PC9

pre-miR371 xenografts, 5 106 cells were implanted subcutaneously into the right

anks of 68-week-old female Harlan Athymic Nude mice. For MKN-45 GFP

vector and MKN-45 pre-miR371 xenografts, 10 106 cells were implanted

subcutaneously into the right anks of 68-week-old female NCR Athymic Nude mice. Mice were randomized into two groups after tumours reached B200 mm3 and treated for 13 days with vehicle (MCT) or erlotinib (50 mg kg 1) for PC9 xenografts, and with vehicle (30% PEG400/0.5% Tween 80/5% propylene glycol) or crizotinib (50 mg kg 1) for MKN-45 xenografts. Drug treatment was then stopped for the next 12 days (days 1325). Tumour sizes were measured throughoutthe study and graphed in the gures without any exclusion of samples. The investigators were not blinded to the group allocation during the experiment or for outcome assessment. At the end of the study, RNA was extracted from tumours using the RNeasy Plus kit from Qiagen as per the manufacturers instructions.

KaplanMeier survival curves. KaplanMeier survival analysis of patient samples was generated either using KaplanMeier plotter tool32 (http://kmplot.com

Web End =http://kmplot.com) or PrognoScan33. The patients were divided into high versus low expressors based on the median value for each gene.

Microarray analysis. For gene expression microarray analysis, PC9 GFP vector

and PC9 pre-miR371-expressing stable cell lines were treated with DMSO or

2 mM erlotinib for 24 h and RNA was extracted using the RNeasy kit (Qiagen). For microarray analysis of DTPs, RNA was extracted from PC9 GFP vector treated

with DMSO or 2 mM erlotinib for 9 days using the RNA isolation kit (Exiqon). Gene expression proling was performed using the Affymetrix Hgu133plus2 platform. Gene expression data were normalized by quantile normalization34 and differentially expressed genes were identied using the LIMMA package, with false discovery rate and P value limited to o0.05 and fold changes 41.5.

Data availability. Data generated during this study are available within the article, its Supplementary Information les. All microarray data generated in this study have been deposited in GEO with accession no. GSE83122.

References

1. Holohan, C., Van Schaeybroeck, S., Longley, D. B. & Johnston, P. G. Cancer drug resistance: an evolving paradigm. Nat. Rev. Cancer 13, 714726 (2013).

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

3. Cara, S. & Tannock, I. F. Retreatment of patients with the same chemotherapy: implications for clinical mechanisms of drug resistance. Ann. Oncol. 12, 2327 (2001).

4. Kurata, T. et al. Effect of re-treatment with getinib (Iressa, ZD1839) after acquisition of resistance. Ann. Oncol. 15, 173174 (2004).

5. Yano, S. et al. Retreatment of lung adenocarcinoma patients with getinib who had experienced favorable results from their initial treatment with this selective epidermal growth factor receptor inhibitor: a report of three cases. Oncol. Res. 15, 107111 (2005).

6. Lee, H. J. et al. Drug resistance via feedback activation of Stat3 in oncogene-addicted cancer cells. Cancer Cell 26, 207221 (2014).

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

8. Wheeler, D. L., Dunn, E. F. & Harari, P. M. Understanding resistance to EGFR inhibitors-impact on future treatment strategies. Nat. Rev. Clin. Oncol. 7, 493507 (2010).

9. Glasspool, R. M., Teodoridis, J. M. & Brown, R. Epigenetics as a mechanism driving polygenic clinical drug resistance. Br. J. Cancer 94, 10871092 (2006).

10. Trumpp, A. & Wiestler, O. D. Mechanisms of Disease: cancer stem cells--targeting the evil twin. Nat. Clin. Pract. Oncol. 5, 337347 (2008).

11. Sharma, S. V. et al. A chromatin-mediated reversible drug-tolerant state in cancer cell subpopulations. Cell 141, 6980 (2010).

12. Lujambio, A. & Lowe, S. W. The microcosmos of cancer. Nature 482, 347355 (2012).

13. Migliore, C. & Giordano, S. Resistance to targeted therapies: a role for microRNAs? Trends Mol. Med. 19, 633642 (2013).

14. Song, S. J. et al. MicroRNA-antagonism regulates breast cancer stemness and metastasis via TET-family-dependent chromatin remodeling. Cell 154, 311324 (2013).

15. Ma, L., Teruya-Feldstein, J. & Weinberg, R. A. Tumour invasion and metastasis initiated by microRNA-10b in breast cancer. Nature 449, 682688 (2007).

16. Calin, G. A. et al. Frequent deletions and down-regulation of micro- RNA genes miR15 and miR16 at 13q14 in chronic lymphocytic leukemia. Proc. Natl Acad. Sci. USA 99, 1552415529 (2002).

17. Lu, J. et al. MicroRNA expression proles classify human cancers. Nature 435, 834838 (2005).

18. Kasinski, A. L. & Slack, F. J. Epigenetics and genetics. MicroRNAs en route to the clinic: progress in validating and targeting microRNAs for cancer therapy. Nat. Rev. Cancer 11, 849864 (2011).

19. Bartel, D. P. MicroRNAs: genomics, biogenesis, mechanism, and function. Cell 116, 281297 (2004).

20. Bartel, D. P. MicroRNAs: target recognition and regulatory functions. Cell 136, 215233 (2009).

21. Rolfs, F. et al. Dual role of the antioxidant enzyme peroxiredoxin 6 in skin carcinogenesis. Cancer Res. 73, 34603469 (2013).

22. Lyon, A. M. & Tesmer, J. J. Structural insights into phospholipase C-beta function. Mol. Pharmacol. 84, 488500 (2013).

23. Williams, K. C. & Coppolino, M. G. SNARE-dependent interaction of Src, EGFR and beta1 integrin regulates invadopodia formation and tumor cell invasion. J. Cell. Sci. 127, 17121725 (2014).

24. Wang, Y., Feinstein, S. I. & Fisher, A. B. Peroxiredoxin 6 as an antioxidant enzyme: protection of lung alveolar epithelial type II cells from H2O2-induced oxidative stress. J. Cell. Biochem. 104, 12741285 (2008).

25. Sundar, I. K. et al. Peroxiredoxin 6 differentially regulates acute and chronic cigarette smoke-mediated lung inammatory response and injury. Exp. Lung. Res. 36, 451462 (2010).

26. Zhang, S. et al. Protein and miRNA proling of radiation-induced skin injury in rats: the protective role of peroxiredoxin-6 against ionizing radiation. Free Radic. Biol. Med. 69, 96107 (2014).

27. Raha, D. et al. The cancer stem cell marker aldehyde dehydrogenase is required to maintain a drug-tolerant tumor cell subpopulation. Cancer Res. 74, 35793590 (2014).

28. Tam, W. L. et al. Protein kinase C alpha is a central signaling node and therapeutic target for breast cancer stem cells. Cancer Cell 24, 347364 (2013).

29. Wilson, F. H. et al. A functional landscape of resistance to ALK inhibition in lung cancer. Cancer Cell 27, 397408 (2015).

30. Moffat, J. et al. A lentiviral RNAi library for human and mouse genes applied to an arrayed viral high-content screen. Cell 124, 12831298 (2006).

31. Huang, da, W., Sherman, B. T. & Lempicki, R. A. Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nat. Protoc. 4, 4457 (2009).

32. Gyorffy, B. et al. An online survival analysis tool to rapidly assess the effect of 22,277 genes on breast cancer prognosis using microarray data of 1,809 patients. Breast Cancer Res. Treat. 123, 725731 (2010).

33. Mizuno, H., Kitada, K., Nakai, K. & Sarai, A. PrognoScan: a new database for meta-analysis of the prognostic value of genes. BMC Med. Genomics 2, 18 (2009).

34. Smyth, G. K. Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat. Appl. Genet. Mol. Biol. 3, Article3 (2004).

Acknowledgements

We thank Zora Modrusan, Robert Soriano, Peng Yue, Pan Du, Robert Yauch, the Genentech IVCC group and the IVP dosing group for their technical support and the Genentech gCell facility for providing cancer cell lines and conducting cell line authentication.

Author contributions

N.S., J.-P.S., M.C. and J.S. conceived and designed the experiments, N.S., J.-P.S.,B.H. and D.D.C. performed the experiments; N.S., J.-P.S. and M.M. analysed the data and interpreted the results; N.S. and R.B. performed the computational analysis; N.S. and J.S. directed the studies and wrote the manuscript.

Additional information

Accession codes: Gene Expression Omnibus, GSE83122.

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

NATURE COMMUNICATIONS | 7:12351 | DOI: 10.1038/ncomms12351 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 9

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12351

Competing nancial interests: All authors were employees of Genentech at the time that the work was conducted and declare 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: Sahu, N. et al. Functional screening implicates miR-371-3p and peroxiredoxin 6 in reversible tolerance to cancer drugs. Nat. Commun. 7:12351doi: 10.1038/ncomms12351 (2016).

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) 2016

10 NATURE COMMUNICATIONS | 7:12351 | DOI: 10.1038/ncomms12351 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications