A Transposon Screen Identifies Loss of Primary Cilia As a Mechanism of Resistance to SMO Inhibitors

Published OnlineFirst September 18, 2017; DOI: 10.1158/2159-8290.CD-17-0281

Research Article

A Transposon Screen Identifies Loss of Primary Cilia as a Mechanism of Resistance to SMO Inhibitors

Xuesong Zhao1,2, Ekaterina Pak1,2, Kimberly J. Ornell1,2, Maria F. Pazyra-Murphy1,2, Ethan L. MacKenzie1,2, Emily J. Chadwick1,2, Tatyana Ponomaryov1,2,3, Joseph F. Kelleher4, and Rosalind A. Segal1,2

15-CD-17-0281_p1436-1449.indd 1436 11/17/17 2:34 PM Published OnlineFirst September 18, 2017; DOI: 10.1158/2159-8290.CD-17-0281

aBstRact Drug resistance poses a great challenge to targeted cancer therapies. In Hedgehog pathway–dependent cancers, the scope of mechanisms enabling resistance to SMO inhibitors is not known. Here, we performed a transposon mutagenesis screen in medulloblastoma and identifi ed multiple modes of resistance. Surprisingly, mutations in ciliogenesis genes represent a frequent cause of resistance, and patient datasets indicate that cilia loss constitutes a clinically relevant category of resistance. Conventionally, primary cilia are thought to enable oncogenic Hedgehog signaling. Paradoxically, we fi nd that cilia loss protects tumor cells from susceptibility to SMO inhibitors and maintains a “persister” state that depends on continuous low output of the Hedgehog program. Per- sister cells can serve as a reservoir for further tumor evolution, as additional alterations synergize with cilia loss to generate aggressive recurrent tumors. Together, our fi ndings reveal patterns of resistance and provide mechanistic insights for the role of cilia in tumor evolution and drug resistance.

SIGnIFICAnCE: Using a transposon screen and clinical datasets, we identifi ed mutations in ciliogenesis genes as a new class of resistance to SMO inhibitors. Mechanistically, cilia-mutant tumors can either grow slowly in a “persister” state or evolve and progress rapidly in an “aggressive” state. Cancer Discov; 7(12); 1436–49. ©2017 AACR.

See related commentary by Goranci-Buzhala et al., p. 1374.

iNtRODUctiON 2012 and 2015 respectively, vismodegib and sonidegib (NVP- LDE225) became the fi rst FDA-approved SMO inhibitors for Aberrant Hedgehog (HH) signaling is implicated in many anticancer treatment in advanced BCC. Several clinical trials in cancers ( 1 ) and is particularly critical in medulloblastoma, the medulloblastoma are in progress. Despite the initial success of most common malignant brain tumor in children, and in basal SMO inhibitors in mouse models and subsequently in patients, cell carcinoma (BCC), the most common human cancer overall long-term effi cacy is limited by the emergence of drug resistance ( 2–4 ). Hyperactivity of the HH pathway is frequently caused ( 9–13 ). Thus far, preclinical and clinical studies have uncovered by inactivating mutations in Patched (Ptch) , which encodes a number of point mutations in SMO that confer resistance. the receptor for HH ligands ( 5–7 ). In the absence of PTCH Although most mutations located in the drug-binding pocket function, Smoothened (SMO), a G-protein-coupled receptor– of SMO directly impair drug binding, an additional group like molecule, traffi cs into the primary cilia, a distinctive, of mutations distally located to drug-binding sites may have microtubule-based signaling organelle. SMO functions within allosteric effects for drug binding or promote a constitutively the primary cilia to inhibit the negative regulators Suppres- activated state of SMO that is less sensitive to drug inhibition sor of fused (SUFU) and protein kinase A (PKA), triggering a ( 9, 10 ). Alterations that activate HH signaling downstream of signaling cascade that culminates in the nuclear import of GLI SMO, such as loss of SUFU or amplifi cation of GLI2 , as well as transcription factors, which activates an HH transcriptional mutations that activate intersecting oncogenic pathways such program that drives proliferation and tumor growth. as RAS/MAPK, PI3K, or aPKC, have also been reported ( 14–16 ). The importance of the HH pathway in human cancers has However, the scope of resistance mechanisms to SMO inhibi- stimulated great interest in developing targeted therapy that tors and the process by which HH pathway–dependent tumors antagonizes HH signaling ( 1, 8 ). These efforts have resulted evolve and recur following therapy are not yet known. in the approval of SMO inhibitors as anticancer agents. In To identify resistance mechanisms, we used a genome-wide transposon mutagenesis screen in HH pathway–dependent medulloblastoma cells and identifi ed recurrent mutations in 1Departments of Cancer Biology and Pediatric Oncology, Dana-Farber Cancer Institute, Boston, Massachusetts. 2 Department of Neurobiology, the ciliogenesis gene Oral-facial-digital syndrome 1 (Ofd1 ) , in Harvard Medical School, Boston, Massachusetts. 3 Institute of Cardio- cells resistant to SMO inhibitors. We demonstrate that resist- vascular Sciences, College of Medical and Dental Sciences, University of ant tumors that lack primary cilia enter a “persister state” of 4 Birmingham, Birmingham, United Kingdom. Novartis Institutes for Bio- slow, GLI2-dependent growth and show that this mechanism medical Research, Cambridge, Massachusetts. of therapeutic resistance occurs in patients. note: Supplementary data for this article are available at Cancer Discovery Online (http://cancerdiscovery.aacrjournals.org/). X. Zhao and E. Pak are co-fi rst authors of this article. ResUlts Corresponding Authors: Rosalind A. Segal, Dana-Farber Cancer Institute, Transposon-Mediated Drug Resistance Screen 450 Brookline Avenue, Boston, MA 02215. Phone: 617-632-4737; Fax: 617-632-2085; E-mail: [email protected] ; and Xuesong Zhao, Identifi es Recurrent Mutations in Sufu and Ofd1 [email protected] To identify genetic changes capable of circumventing SMO doi: 10.1158/2159-8290.CD-17-0281 inhibition, we launched a genome-wide transposon mutagen- ©2017 American Association for Cancer Research. esis screen in SMB21 cells, a murine cell line derived from a

December 2017 CANCER DISCOVERY | 1437

15-CD-17-0281_p1436-1449.indd 1437 11/17/17 2:34 PM Published OnlineFirst September 18, 2017; DOI: 10.1158/2159-8290.CD-17-0281

RESEARCH ARTICLE Zhao et al.

Ptch−/− medulloblastoma (16). SMB21 retains key character- Loss of OFD1 Causes Resistance to SMO Inhibitors istics of SHH-subgroup medulloblastoma, is dependent on Identification of recurrent insertions inSufu provides a HH signaling for proliferation, and is exquisitely sensitive to good validation of the screen, as SUFU is a well-known nega- SMO inhibition (16). The engineered piggyBac transposon sys- tive regulator of the HH pathway that functions by binding tem utilized in this screen relies on a transposon vector that and sequestering GLI transcription factors, and mutations can either enhance or disrupt gene expression in the vicinity in Sufu have been identified in clinical tumors resistant to of the insertion site (Fig. 1A). To deploy the transposon-medi- SMO inhibitors (9–11, 17). To establish a causal relationship ated screen, SMB21 parental cells were transfected with both between mutations in Ofd1 and the resistant phenotype, we piggyBac transposon and transposase vectors for mutagenesis, first conducted an Flp/FRT-based rescue experiment (Fig. and were then selected for clones of cells able to grow in the 2A). As the transposon is engineered with two FRT sites presence of 1 μmol/L sonidegib (Fig. 1B). From 30 million flanking the DNA cargo, lentivirus expressing Flp was used cells, 29 resistant clones were isolated and 27 clones were to remove the DNA cargo. In clones where the transposon successfully propagated. All resistant clones showed robust is inserted in introns, such as R25, this approach success- resistance to sonidegib in subsequent multidose survival fully re-expresses OFD1 and restores sensitivity to sonidegib, assays (Supplementary Fig. S1A). A barcoded splinkerette- as assessed by survival assays and by GLI1 expression (Fig. PCR approach coupled with parallel sequencing was utilized 2B–D). However, this strategy cannot revert disruptions in to identify transposon insertion loci in each clone (Sup- genes if transposons are inserted in exons, such as in clones plementary Fig. S1B). A total of 182 transposon insertions R14 and R16, because the remaining transposon terminal were identified (Supplementary Fig. S1C and Supplementary repeat sequence outside of the FRT sites still causes truncated Table S1). The two genes with the greatest number of recur- mRNA (Supplementary Fig. S3A and S3B). In a second res- rent transposon insertion sites were Sufu (13 clones) and Ofd1 cue experiment, we asked whether expression of a full-length (3 clones; Fig. 1C and D). The robust resistance effects in OFD1 rescues the resistant phenotype in Ofd1 mutants. Sufu- and Ofd1-mutant cells were evident by the dramatic shift Indeed, expression of a full-length OFD1 restored sensitiv- of the growth-inhibitory concentration of sonidegib (Fig. 1E) ity to sonidegib in HH signaling and in cell viability assays and were further validated using additional SMO inhibitors, (Fig. 2E and F). To independently demonstrate that loss of vismodegib and cyclopamine (Supplementary Fig. S1D). OFD1 mediates resistance to SMO inhibition, we specifically We next evaluated the consequence of transposon inser- targeted the Ofd1 locus in parental cells using lentiviral- tions in Sufu and Ofd1. Although most of the transposons CRISPR/Cas9. Depletion of OFD1 confers resistance to SMO landed in introns and inserted in single alleles (Fig. 1C inhibition, demonstrating a causal relationship between loss and D), they effectively disrupted gene transcription and of OFD1 and drug resistance (Fig. 2G and H). eliminated SUFU or OFD1 protein levels in mutant cells (Fig. 1F and G; Supplementary Fig. S1E and S1F). As Ofd1 is on the X chromosome and the SMB21 parental line was Mutations in Ciliary Genes Confer derived from a tumor of a Ptch male, disrupting the single Resistance to SMO Inhibition allele of Ofd1 is sufficient to eliminate all gene products Ofd1 is an X-linked ciliopathy gene essential for primary cil- (Fig. 1G). Intriguingly, complete loss of the SUFU full- ium formation (18), and mutations in Ofd1 cause oral-facial- length protein was observed in nearly all Sufu-mutant clones. digital syndrome type 1 (19). Immunostaining and electron Genomic copy-number analysis of the Sufu locus revealed microscopy demonstrated that primary cilia are absent in loss of the second copy of the Sufu gene in 10 of 13 resist- Ofd1-mutant medulloblastoma cells (Fig. 3A–C). To determine ant clones with Sufu mutations (Supplementary Fig. S2). whether disruption of primary cilia per se confers resistance to This finding indicates that a second genetic alteration event SMO inhibition, we tested multiple genes essential for cili- occurred in the Sufu locus in a subset of mutant cells, thus ogenesis including Ift88, encoding an intraflagellar transport causing a complete loss of SUFU protein in these cells. Both protein, and Kif3a, encoding a subunit of the Kinesin-2 motor Sufu- and Ofd1-mutant cells displayed persistent activation complex. Either shRNA-mediated depletion of IFT88 or of HH signaling in the presence of SMO inhibitors, as dem- CRISPR/Cas9-mediated depletion of KIF3A in SMB21 paren- onstrated by expression of GLI1, a typical readout of HH tal cells confers resistance to SMO inhibitors (Fig. 3D and E; signaling (Fig. 1H). These results suggest that mutations in Supplementary Fig. S3C and S3D). Similar to Ofd1-mutant Sufu and Ofd1 confer resistance to SMO inhibition through cells, cells depleted of IFT88 or KIF3A exhibit persistent reactivation of HH signaling output. low-level activation of HH signaling (Fig. 3D; Supplementary

Figure 1. Transposon-mediated drug resistance screen identifies recurrent mutations inSufu and Ofd1. A, Schematic diagram of piggyBac transposon vectors in the screen. The CMVp-PBase vector contains a mouse codon-optimized transposase under the CMV promoter. The PB transposon vector has two cassettes: (i) UbC promoter–driven GFP and puro-TK fusion genes and (ii) CMV promoter–driven splice donor (SD) to promote overexpression of genes adjacent to the transposon-insertion sites. The DNA cargo is flanked by FRT sites (red arrows) and placed between a pair of PB terminal repeats (PB 5′ and PB 3′, black arrows). B, Outline of screen strategy to identify SMO inhibitor resistance genes. C and D, Location and orientation of transposon insertions in Sufu and Ofd1. Transcription start site of gene is on the left. E, Sufu-mutant (R1, R13) and Ofd1-mutant (R14, R16, R25) cells confer resistance to sonidegib. Survival analysis of cells treated with indicated sonidegib concentrations for 72 hours (relative to DMSO-treated controls; mean ± SEM; n = 3–4 independent experiments). F, Loss of SUFU proteins in Sufu-mutant clones. G, Loss of OFD1 protein in Ofd1-mutant clones. H, Immunoblot for GLI1 in parental and mutant cells treated with sonidegib (0, 10, 100, and 1,000 nmol/L) for 24 hours.

1438 | CANCER DISCOVERY December 2017 www.aacrjournals.org

15-CD-17-0281_p1436-1449.indd 1438 11/17/17 2:34 PM Published OnlineFirst September 18, 2017; DOI: 10.1158/2159-8290.CD-17-0281

Cilia Loss as New Paradigm of Resistance to Smo Inhibitors RESEARCH ARTICLE

A CMVp-PBase CMV promoter piggyBac transposase

FRTFRT

PB transposon PB 5′ UbCp GFP PTK CMVp SD PB 3′

B Transposon Resistance selection: mutant Isolate sonidegib- mutagenesis pool exposed to sonidegib resistant colonies

Identify the genes and Map transposon Secondary screen: resistance mechanisms insertion sites survival assay

C F R8 R18 R1 R9 R13 R12 R19 R4 R15 R5 R3 R10 R7 ParentalR1 R9 R10 R4 R18 SUFU

Actin Sufu 10 kb

D G R14 R16 R25 ParentalR14 R16 R25

OFD1

Actin Ofd1 10 kb

EHParental Ofd1 (R14) GLI1 Actin Sufu (R1) Ofd1 (R16) 150 Sufu (R13) Ofd1 (R25) Sonidegib 125 Parental 100

75 OFD1 (R25)

% Survival 50 OFD1 (R14) 25 SUFU (R13) 0 −2 −10 123 Sonidegib [log (nmol/L)] SUFU (R1)

December 2017 CANCER DISCOVERY | 1439

15-CD-17-0281_p1436-1449.indd 1439 11/17/17 2:34 PM Published OnlineFirst September 18, 2017; DOI: 10.1158/2159-8290.CD-17-0281

RESEARCH ARTICLE Zhao et al.

A BCDMSO D DMSO Sonidegib Sonidegib 1.5 ls 2.0

FRT FRT ve ** *** *** ** ** 100 PB5 PB3 1.5 * 1.0 ** 1.0 Flippase 50 Gli1 mRNA levels

0.5 Ofd1 mRNA le % Survival n.s. 0.5 PB5 PB3

FRT 0 0.0 0.0 Relative Relative Sonidegib Relative Parental Parental R25 R25 R25 R25 R25 R25 LV-Flp LV-Flp LV-Tomato LV-Flp LV-Tomato LV-Flp Parental Parental LV-Tomato LV-Tomato EF GH n.s. n.s. Parental R16 R16 *** *** *** *** LV-Tomato LV-Tomato LV-Ofd1 100 100 CRISPR sgEGFP sgOfd1 ival Sonidegib −+−+−+ ival Sonidegib −+−+ GLI1 50 50 % Su rv % Su rv GLI1 OFD1 OFD1 0 0 Actin Sonidegib Actin Sonidegib Parental Parental R16 R16 CRISPR: sgEGFP sgOfd1 LV-Tomato LV-Ofd1 LV-Tomato LV-Ofd1

Figure 2. Loss of OFD1 causes resistance to SMO inhibitors. A, Schematic for removal of transposon DNA cargo by Flp/FRT-mediated DNA recom- bination. B–D, Transposon removal restores the inhibitory effect of sonidegib on cell survival and HH signaling in R25 cells. B, Survival analysis of cells infected with lentivirus expressing Flp or tdTomato (control) and treated with sonidegib (0, 10, 100, and 1,000 nmol/L) for 72 hours (mean ± SEM, n = 3 independent experiments). C and D, qRT-PCR analysis of Gli1 and Ofd1 mRNA levels in parental and Ofd1-mutant cells treated with vehicle (DMSO) or 1 μmol/L sonidegib for 24 hours (mean ± SEM, n = 3 independent experiments, two-way ANOVA with Tukey correction). E and F, Expression of wild- type Ofd1 rescues resistant phenotype of Ofd1-mutant cells. E, Immunoblot for indicated antibodies in parental and Ofd1-mutant cells infected with lentivirus expressing OFD1 or tdTomato (control) and treated with DMSO or 1 μmol/L sonidegib for 24 hours. F, Survival analysis of parental and rescued Ofd1-mutant cells treated with sonidegib (0, 10, 100, and 1,000 nmol/L) for 72 hours (mean ± SEM, n = 3–4 independent experiments). G and H, CRISPR/ Cas9-mediated depletion of Ofd1 confers resistance to sonidegib in parental cells. G, Immunoblot for indicated antibodies after CRISPR/Cas9-mediated depletion of Ofd1. Cells were treated with DMSO or 1 μmol/L sonidegib for 24 hours. H, Survival analysis of cells treated with sonidegib (0, 10, 100, and 1,000 nmol/L) for 72 hours (mean ± SEM, n = 3 independent experiments) *, P < 0.05; **, P < 0.005; ***, P < 0.0005; n.s., not significant; one-way ANOVA with Dunnett correction unless otherwise noted.

Fig. S3C). Depletion of OFD1 or KIF3A also confers resist- we examined expression and posttranslational processing ance in two additional medulloblastoma lines, SMB55 and of GLI transcription factors. In both parental and Ofd1 SMB56 (Supplementary Fig. S3E and S3F). Given that hun- mutant cells, Gli2 mRNA is highly expressed, whereas no dreds of genes are critical for primary cilia formation (20), Gli3 mRNA can be detected by either qRT-PCR or RNA- mutations in ciliary genes are likely to represent a broad new seq (Supplementary Fig. S4). Thus GLI2, not Gli3, is the class of therapeutic resistance. To determine whether loss of primary effector mediating HH signaling transcription in primary cilia causes drug resistance in a nonspecific man- medulloblastoma cells. We then evaluated how posttransla- ner, we tested targeted and cytotoxic chemotherapy agents, tional processing of GLI2 is regulated in both parental and including BKM120 (PI3K inhibitor), BEZ235 (a dual inhibi- cilia mutant cells. In parental cells, GLI2 is predominantly tor for PI3K and mTOR), cisplatin, and vincristine. These in the full-length form (GLI2-F) with a small fraction in agents exhibit similar potency profiles in both parental and the truncated repressor form (GLI2-R), whereas inhibition cilia mutant cells (Supplementary Fig. S3G), suggesting that of Smo drastically reduces GLI2-F and increases GLI2-R loss of primary cilia does not cause a general resistance profile (Fig. 4C). These data indicate that shutdown of HH signal- but specifically results in resistance to SMO inhibitors. ing by Smo inhibitors in parental cells involves conversion of GLI2-F to GLI2-R. In contrast, in cilia mutant cells, only Loss of Primary Cilia Enables Tumor Cells to Evade GLI2-F is detected and levels are not affected in the presence Drug Inhibition and Maintain a “Persister” State of Smo inhibitors (Fig. 4C and D), suggesting that impaired The primary cilium is a signal transduction center for Gli processing due to cilia loss could explain persistent acti- HH signaling and is essential for SMO activation and vation of downstream HH signaling. Cellular fractionation GLI processing (21, 22). In the absence of primary cilia, experiments showed that GLI2-F is associated with nuclear SMO cannot transduce positive signal to GLI. Nonetheless, chromatin in cilia mutant cells, indicative of transcriptional Ofd1-mutant cells express HH pathway signature genes as activity (Fig. 4E and F). Importantly, knockdown of GLI2 assessed by Gene Set Enrichment Analysis (GSEA) of RNA- effectively reduced both transcriptional output of HH sig sequencing (RNA-seq) data, maintaining an active, albeit naling and cell viability in both parental and resistant cells limited, HH signaling output (Fig. 4A and B). To ascertain (Fig. 4G and H), indicating that proliferation and survival of how active HH signaling is maintained in cells lacking cilia, these cells depends on GLI2-mediated HH signaling output.

1440 | CANCER DISCOVERY December 2017 www.aacrjournals.org

15-CD-17-0281_p1436-1449.indd 1440 11/17/17 2:34 PM Published OnlineFirst September 18, 2017; DOI: 10.1158/2159-8290.CD-17-0281

Cilia Loss as New Paradigm of Resistance to Smo Inhibitors RESEARCH ARTICLE

A BC Parental R25 Parental Parental Parental Ofd1 mutant 100 80 60 R14 R16 40 R14 R25 20 % Cells with cilia 0

DAPI γ -tubulin Acetylated tubulin R14R16 R25 Parental

100

CRISPR SgEGFP Sg Kif3a #1 Sg Kif3a #2 Sg Kif3a #3 Sonidegib −+−+−+−+ 50

GLI1 % Survival KIF3A 0 Actin Sonidegib CRISPR: sgEGFP sgKif3a#1 sgKif3a#2 sgKif3a#3

Figure 3. Mutations in ciliary genes confer resistance to SMO inhibition. A–C, Ofd1-mutant cells are defective in primary cilia formation. A, Immu- nostaining of primary cilia in parental and mutant cells. Acetylated tubulin (green) marks the ciliary axoneme, and γ-tubulin (red) marks basal bodies at the cilium base. Nuclei are marked by DAPI. Scale bar, 10 μm. B, Quantification of cilia immunostaining based on percentage of DAPI-positive cells with cilia (mean ± SEM, n = 4 experiments). C, Transmission electron microscopy images of primary cilia (yellow arrows) and centrioles (blue arrows) in parental and Ofd1-mutant cells. Scale bars, 500 nm. D and E, CRISPR/Cas9-mediated depletion of KIF3A confers resistance to SMO inhibition in parental SMB21 cells. Assayed by relative survival at sonidegib concentrations (0, 10, 100, and 1,000 nmol/L; 72 hours, relative to DMSO-treated controls; mean ± SEM, n = 3 independent experiments).

Together, these results suggest that loss of cilia abolishes of Ptch status. Together, these data suggest that loss of pri- SMO-dependent full activation of HH signaling and simul- mary cilia results in a constitutive low-level output of HH taneously eliminates GLI2-R formation, resulting in low but signaling in both tumor and developmental contexts. persistent GLI2 activity and HH pathway transcriptional output. This trade-off enables cilia mutant cells to escape Sufu Heterozygosity Synergizes with Cilia Loss to drug inhibition and maintain a “persister” state by surviv- Enhance HH Signaling Output ing on this persistent low output of HH signaling. In stark Although primary cilia are absent in both R25 and R14 due contrast to cilia mutants, Sufu-mutant cells have drastically to mutations in Ofd1, these clones exhibit different levels of low levels of both full-length and repressor forms of GLI2 HH signaling output and growth rates. R25 cells have lower (Fig. 4C), indicating that a key role of SUFU is to sequester levels of GLI1 expression and grow much more slowly than and stabilize GLI protein (23–25). These data highlight a R14 cells (Fig. 5A and B). Parental cells depleted of OFD1 mechanistic difference between Sufu mutation– and Ofd1 or KIF3A exhibited a phenotype similar to R25, with low mutation–mediated activation of HH signaling. activation of HH signaling and slow growth (Supplementary To examine applicability of these effects to normal devel- Fig. S5A and S5B), indicating that loss of primary cilia alone opment, we performed CRISPR/Cas9-mediated knockout of is sufficient to confer resistance by enabling tumor cells to Ofd1 in cerebellar granule cell precursors (GCP) from wild- attain the “persister” state. As persister cells have the poten- type and Ptch+/− mice, and measured HH signaling activity tial for further genetic alteration and tumor evolution, we based on Gli1 expression levels. In the absence of primary cilia postulated that R14 may harbor a relevant alteration in addi- (sgOfd1 conditions), Gli1 levels were significantly reduced in tion to cilia loss. Because we observed frequent genomic loss both wild-type and Ptch+/− cells compared with SAG-stimu- of the second Sufu allele in Sufu-mutant clones, we examined lated activation (Supplementary Fig. S5C). However, these the genomic copy number of Sufu in Ofd1-mutant clones. levels of Gli1 were significantly higher than those in unstimu- Quantitative PCR revealed only one copy of the Sufu gene in lated control sgGFP cells, and there were no significant differ- R14, whereas two copies remained in R25 (Fig. 5C). Moreov er, ences between wild-type and Ptch+/− GCPs, indicating that this heterozygosity for Sufu in R14 is associated with reduced pro- low but persistent activation of HH signaling is independent tein levels of SUFU (Fig. 5D). Based on these observations,

December 2017 CANCER DISCOVERY | 1441

15-CD-17-0281_p1436-1449.indd 1441 11/17/17 2:34 PM Published OnlineFirst September 18, 2017; DOI: 10.1158/2159-8290.CD-17-0281

RESEARCH ARTICLE Zhao et al.

A BC D Sufu Ofd1 0.5 CRISPR Kif3a Parental R13 R14 R16 sgEGFP sg SonidegibParental Sufu Ofd1 0.4 Sonidegib −+−+−+− + Sonidegib −+ −+ r r r r r

Pa Pa Pa Pa Pa R13 R1 R14 R14 R16 R25 0.3 GLI2-F GLI2-F Gli1 Gli2 Boc 0.2 Mycn Sfrp1 0.1 Ptch1 Enrichment score GLI2-R Ptch2 GLI2-R Ccnd1 0.0 Bcl2 KIF3A Hhip Actin −3 +3 Sufu Ofd1 Parental Actin

EFGH Parental Ofd1 (R14) shLucshGli2-2 shGli2-3 DMSO Sonidegib DMSO Sonidegib 120 1.2 100 1.0 WC Cyto. Nuc. Chr. WC Cyto. Nuc. Chr. WC Cyto. Nuc. Chr.CWC Cyto. Nuc. hr. 80 0.8

GLI2-F ival *** *** *** *** Vinculin 60 0.6 *** *** SFPQ 40 % Su rv 0.4 Gli1 mRNA H2B 20 0.2 00 Parental Sufu Ofd1 Parental Sufu Ofd1 (R13) (R14) (R13) (R14)

Figure 4. Loss of primary cilia enables tumor cells to evade drug inhibition and maintain a “persister” state. A, Expression of HH pathway signature genes in parental, Sufu-mutant, and Ofd1-mutant cells by RNA-seq analysis. B, Enrichment scores of the HH pathway by GSEA. C and D, Formation of GLI2 repressor is abolished in cilia mutant cells regardless of sonidegib treatment. Immunoblot to assess full-length GLI2 (GLI2-F) processing into repressor (GLI2-R) with DMSO or 1 μmol/L sonidegib in parental, Sufu-mutant, and Ofd1-mutant cells (C), or in cells with CRISPR/Cas9-mediated Kif3a depletion (D). E and F, Subcellular fractionation of GLI2 in parental or Ofd1-mutant cells. Full-length GLI2 is present and maintained in the cytoplasmic, nuclear, and chromatin fractions of Ofd1-mutant cells treated with DMSO or 1 μmol/L sonidegib for 24 hours. WC, whole cell; Cyto., cytoplasm; Nuc., nuclear; Chr., chromatin. G and H, Mutant cells depend on HH signaling for survival. G, Survival analysis of parental, Sufu-mutant, and Ofd1-mutant cells after shRNA-mediated depletion of Gli2. H, qRT-PCR analysis of Gli1 mRNA levels in parental, Sufu-mutant, and Ofd1-mutant cells after Gli2 knockdown (mean ± SEM, n = 3 independent experiments; ***, P < 0.0005, two-way ANOVA with Dunnett correction).

we hypothesize that in the absence of primary cilia, greater GLI1 expression in Sufu+/− than in Sufu+/+ cells (Sup- sequestration of GLI2 by SUFU becomes a bottleneck that plementary Fig. S6C). Taken together, these data indicate limits GLI2 activation and nuclear translocation. Therefore, that Sufu heterozygosity, which is not sufficient to trigger losing one copy of the Sufu gene in this sensitized condition activation of HH signaling on its own, can synergize with drastically enhances GLI2 activation and causes a “persister” loss of primary cilia to augment pathway activation (Fig. state of slow-growing resistant cells to evolve into rapidly 5I). This surprising synergy provides a new route for robust growing resistant tumors. To test this hypothesis, we system- activation of downstream HH signaling in the absence of atically adjusted SUFU levels in Ofd1 mutant cells. Notably, primary cilia and may play a role in a range of developmental a 4-fold increase in SUFU suppressed cell proliferation rate and disease contexts. and dampened HH signaling in Ofd1 (R14) cells, whereas proliferation rate and HH pathway activity of parental cells Cilia Loss and Ciliary Mutations in Preclinical were not perturbed by a similar increase in SUFU (Fig. 5E and and Clinical Resistant Samples F). Furthermore, redistribution of GLI2-F from the nucleus into To determine if cilia mutant cells generate resistant tumors the cytoplasm was observed upon increase of SUFU expres- in vivo, Ofd1-mutant cells were orthotopically transplanted sion only in Ofd1 (R14) mutant cells (Fig. 5G and H), result- into mice. Both parental and Ofd1-mutant cells initiate ing in reduction of GLI2-F in the nucleus and decreased GLI1 tumors in the brain. Notably, although sonidegib abrogates expression. Together, these data indicate that in the absence growth of parental cells, Ofd1-mutant cells exhibit com- of cilia, SUFU acts in a quantitative, dose-dependent man- plete resistance to sonidegib treatment (Fig. 6A). We next ner in limiting GLI2-mediated HH signaling output in examined whether loss of primary cilia occurs in preclinical tumor cells. To examine whether this synergistic interaction in vivo models in the context of acquired resistance. SMB21 between loss of cilia and Sufu heterozygosity occurs in a broad parental cells were transplanted into nude mice and treated range of contexts, we turned to a genetically defined system, with sonidegib. After the initial robust response to son- Sufu+/+ and Sufu+/− mouse embryonic fibroblast (MEF) cells. idegib treatment, resistant tumors developed spontaneously As observed for SMB cells, heterozygosity of Sufu significantly in all animals (Fig. 6B). Immunostaining with cilia mark- reduced SUFU protein and mRNA levels (Supplementary ers revealed that resistant tumors had a greater percentage Fig. S6A and S6B). Importantly, depletion of OFD1 induces of unciliated cells than tumors that were not treated with

1442 | CANCER DISCOVERY December 2017 www.aacrjournals.org

15-CD-17-0281_p1436-1449.indd 1442 11/17/17 2:34 PM Published OnlineFirst September 18, 2017; DOI: 10.1158/2159-8290.CD-17-0281

Cilia Loss as New Paradigm of Resistance to Smo Inhibitors RESEARCH ARTICLE

A BCD Parental Parental_sonidegib Parental R16 DMSO R14 R14_sonidegib R14 R25 Sonidegib R25 R25_sonidegib 1.5 150 2.5 ) 4 *** 2.0 Parental Ofd1 (R14) Ofd1 (R25) 1.0 100 Sonidegib −+ −+−+ * 1.5 SUFU 1.0 0.5 50 ** Actin 0.5 Cell number (10 0.0 0 0.0 gene copy number Sufu gene copy Relative Gli1 mRNA levels Relative Parental R14 R25 02 5 Day

E FG H +GFP Parental + GFP R14 + GFP +GFP +Sufu Cytoplasm +Sufu Parental + SUFU R14 + SUFU WCNWCN Nucleus

GLI2 R14 Parental 5 150 n.s ** *** n.s. ) 1.0 1.0 4 LMNB1 4 n.s 0.8 *** 100 GAPDH 3 *** 0.6 WCNWCN 2 0.5 0.4 50 *** GLI2 action of GLI2 0.2 1 Fr Cell number (10 LMNB1 0.0 0.0 0 0 G SGS Relative Gli1 mRNA levels Relative Relative Sufu mRNA levels Relative Parental R14 02 5 GAPDH Parental R14 Parental R14 Day

Sonidegib Ptch−/− Ptch−/− Ptch−/− Ptch−/− Cilia−/− Cilia−/− SMO Sufu+/−

GLI2 GLI2 SUFU SUFU SMO GLI2 GLI2 SUFU SUFU

R GLI2A GLI2A

OFF HH signaling output Maximum ON

Figure 5. Sufu heterozygosity synergizes with cilia loss to enhance HH signaling output. A and B, R14 cells have higher pathway activity and prolifer- ate faster than R25 cells. A, qRT-PCR analysis of Gli1 mRNA levels in parental and Ofd1-mutant (R14, R25) cells treated with DMSO or 1 μmol/L sonidegib for 24 hours (mean ± SD, n = 3 independent experiments; *, P < 0.05; ***, P < 0.005, two-way ANOVA with Sidak correction). B, Growth curve of cells treated with DMSO or 1 μmol/L sonidegib (mean ± SD, n = 3 technical replicates, representative of three independent experiments; **, P < 0.01, one- way ANOVA with Tukey correction). C, Sufu genomic copy number of parental and Ofd1-mutant cells determined by qPCR (mean ± SD, n = 3 experiments). D, Immunoblot comparing levels of SUFU in parental and Ofd1 mutants treated with DMSO or 1 μmol/L sonidegib for 24 hours. E, qRT-PCR analysis of Sufu and Gli1 mRNA levels in parental- and Ofd1-mutant cells with or without overexpression of Sufu (mean ± SD, n = 3 technical replicates, representative of three independent experiments; ***, P < 0.001, one-way ANOVA with Tukey correction). F, Growth curve of parental and Ofd1 (R14) cells with or without overexpression of SUFU (mean ± SD, n = 3 technical replicates, representative of three independent experiments; ***, P < 0.001, one-way ANOVA with Tukey correction). G, Subcellular fractionation of full-length GLI2 in parental or Ofd1-mutant cells with or without overexpression of Sufu. W, whole cell; C, cytoplasm; N, nuclear. H, Quantification of GLI2 subcellular localization inG (mean ± SD, n = 3 experiments; **, P < 0.01, one-way ANOVA with Tukey correction). I, Schematic overview of HH signaling in Ptch−/− tumors with or without primary cilia in the presence or absence of primary cilia. The primary cilium is required for Smo-dependent conversion of GLI2 into GLI2-A, achieving maximal activation of HH signaling (right-most plot). The cilium is also required for conversion of GLI2 into GLI2-R and complete shutdown of HH signaling in the presence of Smo inhibitors (left-most plot). In the absence of primary cilia, Smo is inactive and Smo inhibitors are ineffective, and so both GLI2-A and GLI2-R conversions are compromised. Cilia mutant cells maintain a persistent low signaling output by relying on low levels of GLI2 activation, which is constrained by SUFU sequestration. In Sufu heterozygotes, reduced SUFU levels synergize with cilia loss to markedly enhance the pathway output.

December 2017 CANCER DISCOVERY | 1443

15-CD-17-0281_p1436-1449.indd 1443 11/17/17 2:34 PM Published OnlineFirst September 18, 2017; DOI: 10.1158/2159-8290.CD-17-0281

RESEARCH ARTICLE Zhao et al.

A BCParental vehicle Parental Smo (D477G) D

Vehicle Sonidegib Parental sonidegib Acetylated tubulin γ -tubulin DAPI Vehicle-treated Smo (D477G) vehicle Sonidegib-resistant Parental Ofd1 (R14) Smo (D477G) sonidegib 10,000 2,500 80 100 *** *** 8,000 2,000 *** 60 6,000 1,500 40 4,000 1,000 50 % Cilia

2,000 500 % Survival 20 % Bioluminescence 0 0 0 0 00714 21 714 21 02040 60 80 Parental Smo Days after treatment Days after treatment Days after treatment (D477G) Sonidegib-resistant Vehicle-treated EF GHIJ Parental Mutated cilia genes Parental 5 100 2,000 2,000 125 Sufu (R1) 125 Ofd1 (R14) Sufu (R13) Sufu (R13) 4 80 USH2A 1,500 100 Ofd1 (R14) 100 SMO DNAH10 PTCH1 1,000 1,000 Ofd1 (R16) ival ival 75 3 60 500 75 Ofd1 (R25) MYO15A 2 40 CEP290 RP1 0 0 50 50 % Su rv PCDH15 −500 % Su rv 1 20 CDH23 Resistant 25 25 OFD1 −1,000 Sensitive −1,000 % Resistant samples % Cilia gene mutations 0 0 −1,500 Normal skin 0 0 20 40 60 80 Ciliomor signature scores Ciliomor signature scores 012 34 012 34 % Untreated samples ATO [log (nmol/L)] JQ1 [log (nmol/L)] Sensitive UntreatedResistant Resistant

Figure 6. Ciliary mutations in preclinical and clinical resistant samples. A, Luciferase-expressing parental and Ofd1 (R14)-mutant cells were orthotopically transplanted to mice in the cerebellum. Mice were then treated with sonidegib or vehicle. Tumor growth was assessed by measuring bioluminescence (mean ± SEM, n = 8). B, Kaplan–Meier survival analysis of nude mice that received subcutaneous transplants of parental cells or parental cells expressing Smo (D477G). All sonidegib-treated mice engrafted with parental cells gradually developed resistance. Two mice were euthanized due to tumor burden, and 3 mice with tumors less than 2 cm were collected at the end of the study; n = 5 parental vehicle; n = 5 parental sonidegib; n = 4 Smo (D477G) vehicle; n = 5 Smo (D477G) sonidegib. C, Immunostaining for primary cilia in sonidegib-resistant versus vehicle-treated subcutaneous tumors (acetylated tubulin in green, γ-tubulin in red, DAPI in blue). Scale bar, 20 μm. Scale bar in the magnified region, 5μ m. Arrows indicate magnified regions. D, Quantification of cilia immunostaining shown in C, based on percentage of DAPI-positive cells with cilia (mean ± SD, n = 4–5 tumors; ***, P < 0.001, one- way ANOVA with Tukey correction). E, Analysis of cilia gene mutations in vismodegib-resistant clinical BCC samples compared with untreated samples. Percentages are quantified based on total number of cilia gene mutations normalized to total number of mutations in each sample. Mann–Whitney test, P < 0.05 (n = 48 untreated and 11 resistant samples). F, Recurrent mutations in cilia genes quantified based on total number of samples with a given mutated cilia gene normalized to the total number of samples in the population (48 untreated and 11 resistant). G, ssGSEA was applied to generate an enrichment score of Ciliomor signature in each BCC patient sample. H, Box plot analysis was used to visualize expression of Ciliomor signature in resistant clinical BCC samples (n = 9) compared with a group of normal skin (blue dots, n = 8) and sensitive samples (purple dots, n = 4). Unpaired t test, P = 0.0034. I and J, Survival analysis for parental, Sufu (R1, R13)-mutant, and Ofd1 (R14, R16, R25)-mutant cells treated with ATO or JQ1 (mean ± SEM, n = 3–5 independent experiments).

sonidegib (Fig. 6C and D). As a control, we examined the set of patients may lead to rapidly growing resistant tumors. prevalence of cilia in SMB cells expressing SmoD477G, a SMO Together, these results provide clinical evidence that loss of mutant resistant to sonidegib. Xenografts of these resistant primary cilia, alone or in combination with additional altera- cells have the same level of cilia formation before and after tions, constitutes a route for resistance to SMO inhibition. treatment with sonidegib (Fig. 6C and D). Together, these data indicate that loss of primary cilia confers resistance to Gene Signatures Associated with Cilia Loss in SMO inhibitors in vivo. Clinical Resistance Samples To determine whether loss of primary cilia occurs in cancer To identify global changes in cell state that are associated patients treated with SMO inhibitors, we analyzed datasets with loss of cilia and might be indicative of resistance to of BCC, a common cancer that is the FDA-approved indica- SMO inhibitors, we compared expression profiles of cilia tion for SMO inhibitors (10). Analysis of sequencing data mutant cells with parental SMB cells in our model system. A from 11 resistant and 48 untreated patients (“resistant” and total of 236 gene signatures were evaluated by GSEA, includ- “untreated”) revealed many mutations in ciliary genes (Sup- ing 186 gene sets derived from the Kyoto Encyclopedia of plementary Table S2). Importantly, the incidence of mutations Genes and Genomes (KEGG) pathway database and 50 hall- in ciliary genes is significantly greater in resistant samples mark gene sets from the Molecular Signatures database (26). compared with untreated tumors (Fig. 6E and F). Further- Genes involved in myogenesis and cytokine–cytokine recep- more, in 1 patient, we identified both aSUFU mutation and tor interaction are among the most upregulated signatures, an OFD1 mutation in the same resistant sample. In addition, whereas genes involved in oxidative phosphorylation and resistant samples from 2 other patients showed a loss of one ribosome function are among those most downregulated copy of chromosome 10q including the SUFU locus, as well as (Supplementary Fig. S7A). These data suggest that metabolic mutations in ciliary genes. Notably, loss of one copy of chro- state and protein translation are altered by cilia loss, and mosome 10q is one of the most common copy-number aberra- these changes can potentially be explored as an indicator for tions in the SHH subgroup of medulloblastoma (37/133, 28%; clinical resistance. To evaluate the clinical relevance of this ref. 17). Mutations in any one of the ciliary genes in this large expression profile, we constructed a gene signature based

1444 | CANCER DISCOVERY December 2017 www.aacrjournals.org

15-CD-17-0281_p1436-1449.indd 1444 11/17/17 2:34 PM Published OnlineFirst September 18, 2017; DOI: 10.1158/2159-8290.CD-17-0281

Cilia Loss as New Paradigm of Resistance to Smo Inhibitors RESEARCH ARTICLE

on the top 50 differentially upregulated genes and the top son insertion events that mutate both alleles, or else depend on 50 downregulated genes, defined as the “Ciliomor” (Cili- a collaborating genomic event to inactivate the second allele. opathy tumor) signature (Supplementary Fig. S7B). We first The exception is for genes on the X chromosome. Ofd1 is on validated this signature in medulloblastoma cells indepen- the X chromosome, and parental SMB21 cells are derived from dently created by CRISPR/Cas9-mediated mutations in Ofd1 a male mouse. Thus, one transposon insertion is sufficient to and Kif3a. Using single-sample GSEA (ssGSEA), a Ciliomor cause complete loss of Ofd1 function. signature score was calculated for each sample. Both Kif3a Recurrent mutations in Sufu (13) and Ofd1 (3) suggest and Ofd1 knockout cells are indeed associated with higher that our screen has achieved a good degree of coverage of the scores of Ciliomor signature compared with control cells haploid genome. Even though our screen did not identify any (Supplementary Fig. S7C). Using this signature, we applied resistant clone in which both alleles are mutated by transpo- the same ssGSEA analysis to a cohort of patients with BCC sons, the screen was able to detect the transposon insertions treated with SMO inhibitors to generate an enrichment that cooperate with other genetic events. In the majority of 13 score in each patient sample (9). This analysis showed resist- Sufu-mutant clones, the transposon-mutated allele in combi- ant samples preferentially associated with higher scores of nation with genomic loss of the second Sufu allele resulted in Ciliomor signature (Fig. 6G and H), supporting the notion complete loss of SUFU. We speculate that the frequent loss of that cilia loss might be associated with clinical resistance the second Sufu allele is partially due to an unstable genome to SMO inhibitors. Due to lack of resistant medulloblas- in parental SMB21 cells and is a favorable event selected by toma samples, we have also applied ssGSEA analysis in a drug inhibition. Notably, loss of Sufu genomic region is one cohort of treatment-naïve medulloblastoma samples (17). of the most frequent copy-number variation events in human We found that 3 of 73 cases exhibited high Ciliomor sig- medulloblastoma, hence constituting a physiologically rel- nature scores (Supplementary Fig. S7D). This preliminary evant event. Intriguingly, we found more Sufu mutants (13) observation raises a possibility of preexisting resistance to than Ofd1 mutants (3). This ratio suggests that the Sufu locus SMO inhibitors among these samples. Examination of pre/ may be a hotspot for more frequent transposon landing, or posttreatment patient cohorts will be necessary to confirm Sufu mutations may enable more robust resistance. the implication in medulloblastoma. A surprising finding of our study is identification of loss To identify therapeutic strategies that overcome this new of primary cilia as a new mechanism of resistance to SMO type of resistance, we explored epigenetic and transcrip- inhibitors. The primary cilium is notable for supporting tional targeting strategies. Parental, Sufu-mutant, and Ofd1- maximal activation of HH signaling and is essential for SMO mutant cells all responded to arsenic trioxide (ATO; Fig. activation and conversion of GLI2 into GLI2-A. Paradoxically, 6I), a reported Gli inhibitor (27), in accordance with GLI2 we find that loss of primary cilia causes resistance to SMO knockdown results shown above. Preclinical studies indicated inhibitors. We show that cells lacking primary cilia express that a BET bromodomain inhibitor, JQ1, which affects the HH pathway signature genes and maintain a persistent low- epigenetic landscape (28), is effective in treating the SHH level activation of HH signaling. Our results further indicate subgroup of medulloblastoma (29). We found that treatment that primary cilia are required for conversion of GLI2 into with JQ1 significantly reduced cell proliferation and survival GLI2-R and for the complete cessation of HH signaling in in parental, Sufu-mutant, and Ofd1-mutant cells (Fig. 6J). the presence of SMO inhibitors, suggesting that cells with- Together, these data suggest that intervention with epige- out cilia evade the effect of SMO inhibitors by eliminating netic and transcriptional inhibitors has the potential to over- GLI2-R. Consequently, cilia mutant cells grow slowly, but can come diverse mechanisms of resistance to SMO inhibitors. survive as persister cells during drug treatment (Fig. 7). Inter- estingly, Ift80 and Rpgrip1l were additional ciliogenesis genes with a single hit in our screen (Supplementary Table S1). Fur- DISCUSSION ther experimental verification would be needed to determine Here, we developed a piggyBac transposon system as a new whether these mutations are driver or passenger events. screening platform for identification of drug resistance mecha- We also uncovered a new evolutionary path wherein cilia nisms. This approach provides a fast, efficient, and systematic mutant tumors in the persister state can progress into a fast- approach parallel and complementary to library-based func- growing state. Loss of one copy of Sufu synergizes with the tional screens such as CRISPR, RNAi, and ORFeome expres- loss of primary cilia to accelerate resistant tumor regrowth. sion library. Compared with library-based approaches, this Mechanistically, in the absence of cilia-dependent GLI2 pro- approach has several unique advantages. First, interrogation of cessing, cytoplasmic sequestration of GLI2 by SUFU becomes the genome is not limited by the design and content of librar- a bottle-neck step that restrains GLI2 activation. Therefore, ies, and thus has the potential to uncover functionally relevant any reduction of SUFU protein in this sensitized condi- regions previously overlooked. Second, DNA transposons can be tion synergizes with loss of primary cilia to enhance GLI2 easily engineered to incorporate new features such as selection activation and proliferation, resulting in the evolution of markers or regulatory elements to facilitate both gain-of-func- more aggressive tumor growth. It is conceivable that other tion and loss-of-function screens. Third, transposon-mediated alterations such as activation or amplification ofGli2 may be mutations may easily be reversed to validate causal relationship, an alternative way to synergize with loss of primary cilia to which is key for successful functional screens. However, one promote resistance to SMO inhibitors. major limitation of insertional mutagenesis methods is their In clinical sequencing data from BCC tumors, we identi- low probability of mutating both alleles of a gene in the same fied many mutations in ciliary genes (Supplementary Table cell. Homozygous mutations require two independent transpo- S2) and demonstrated that these mutations are enriched in

December 2017 CANCER DISCOVERY | 1445

15-CD-17-0281_p1436-1449.indd 1445 11/17/17 2:34 PM Published OnlineFirst September 18, 2017; DOI: 10.1158/2159-8290.CD-17-0281

RESEARCH ARTICLE Zhao et al.

SMO inhibitor Treatment Tumor size Tumor Smo (mut) Time

SMO inhibitor Tumor size Tumor Sufu−/− Time

SMO inhibitor

Persister state Tumor size Tumor Cilia−/− Time

SMO inhibitor

Aggressive state Tumor size Tumor Persister state Cilia−/− Cilia−/− Cilia−/−; Sufu+/− Time

Figure 7. Models for acquired resistance due to loss of primary cilia. Point mutations in Smo directly revert drug action and restore tumor growth. Homozygous loss of Sufu can similarly reverse drug inhibition by directly activating downstream HH signaling. In contrast, loss of primary cilia costs cells full activation of HH signaling, but allows cells to survive and grow slowly as “persister” cells under drug selection. Importantly, cilia mutant cells are sus- ceptible for oncogenic transformation by additional mutations, such as Sufu heterozygosity, which normally would not be sufficient to cause oncogenic transformation. Sufu heterozygosity, therefore, acts as a second hit to elevate HH signaling and transform resistant cells into aggressively growing recurrent tumors.

drug-resistant samples. Moreover, we identified co-occur- as Sufu reduction, may be a potent oncogenic driver. Indeed, rence of SUFU and OFD1 mutations in one resistant sample. removal of primary cilia has been reported to cooperate with Together, these results provide clinical evidence that loss of constitutive activation of GLI2 to promote tumor initiation primary cilia, alone or in combination with additional muta- in mouse models (30, 31). Because GLI2 amplification and tions, constitutes a route for bypassing SMO inhibition. loss of SUFU genomic region are common events in human Given that hundreds of genes are critical for primary cilia medulloblastoma and BCC, sequencing analyses of sufficient formation (20), mutations in ciliary genes are likely to rep- number of clinical tumor samples should be able to deter- resent a broad new class of therapeutic resistance that shares mine whether loss of primary cilia synergizes with other key pathologic characteristics. alterations and thereby contributes to oncogenesis. Do mutations in ciliogenesis genes cause cancer in general? Finally, this study reveals two different strategies to over- HH pathway–dependent tumors are predominantly initiated come diverse mechanisms of resistance to SMO inhibitors by loss of Ptch or activating mutations in Smo that cause that could inform future therapies. Consistent with evi- robust activation of HH signaling. Our study demonstrated dence that GLI2 is necessary for the growth of resistant the significance of cilia loss in the context of drug resistance. tumors that lack primary cilia, ATO, which is reported to Given that this is a general mechanism for low-level activa- abrogate GLI2 function, can be used to treat these resist- tion of HH signaling in both tumor resistance and develop- ant tumors. Therapeutic intervention of epigenetic states by mental settings, it is conceivable that this mechanism could bromodomain inhibitors such as JQ1 provides an alternative also play a role in tumorigenesis. Our findings predict that approach to counter resistance that will enable enhanced cilia loss in combination with additional alterations, such treatment of patients with HH-dependent tumors.

1446 | CANCER DISCOVERY December 2017 www.aacrjournals.org

15-CD-17-0281_p1436-1449.indd 1446 11/17/17 2:34 PM Published OnlineFirst September 18, 2017; DOI: 10.1158/2159-8290.CD-17-0281

Cilia Loss as New Paradigm of Resistance to Smo Inhibitors RESEARCH ARTICLE

METHODS (Invitrogen; 1:250). Secondary antibodies were conjugated to Alexa Fluor 488 or 568 (Invitrogen; 1:200) or Cy5 (Jackson ImmunoRe- Additional details are provided in the Supplementary Data. search Laboratories; 1:200). For immunoblots, cell lysates were made in RIPA buffer Animals (50 mmol/L Tris-HCl, pH 7.4, 150 mmol/L NaCl, 0.25% deoxycholic

All experimental procedures were done in accordance with the acid, 1% NP-40, 1 mmol/L DTT, 10 mmol/L NaF, 1 mmol/L NaVO3, and National Institutes of Health guidelines and were approved by the 1 mmol/L PMSF) with protease inhibitor cocktail (Roche). Lysates were Dana-Farber Cancer Institutional Animal Care and Use Committee. quantified (Bradford assay), normalized, reduced, denatured (95°C), The nu/nu mice were obtained from Charles River Laboratories. and resolved by SDS gel electrophoresis on 4% to 12% Bis-Tris Glycine gels (Invitrogen). Proteins were transferred to PVDF membranes (Bio- SMB Cell Culture Rad) and probed with primary antibodies recognizing GLI1 (Cell Sign- SMB cell lines (SMB21, SMB55, and SMB56) were derived from aling Technology; 1:1,000), SUFU (Cell Signaling Technology; 1:1,000), spontaneous medulloblastoma tumors in Ptch+/− mice. They were Actin (Sigma; 1:10,000), GLI2 (R&D; 1:3,000), KIF3A (Sigma; 1:1,000), established in our lab in 2010 and were described and character- IFT88 (Sigma; 1:1,000), OFD1 (a gift from Jeremy Reiter; 1:1,000; ref. ized in ref. 16. SMB cells exhibited LOH for Ptch and mutations 32), Vinculin (Sigma-Aldrich; 1:10,000), SFPQ (Abcam; 1:1,000), and in Trp53 as described. The propagated cells were authenticated by Histone H2B (Cell Signaling Technology; 1:1,000). Secondary anti- genotyping Ptch locus and testing sensitivity to sonidegib. SMB cells bodies are either horseradish peroxidase–linked (Bio-Rad) or infrared were cultured as neurospheres at 37°C in a humidified incubator dye–labeled (LI-COR). Proteins were visualized using LI-COR Odyssey Imager, film, or ImageQuant LAS 4000 imager. with 5% CO2 in DMEM/F12 media (2% B27, 1% Pen/Strep). To pas- sage cultures, cells were dissociated with accutase and plated 1:3 in fresh media. Subcutaneous and Orthotopic Transplantation and In Vivo Treatment Transposon Mutagenesis Screen Cells (5 × 106) in 100 μL were injected subcutaneously into the Transfection was achieved using 5 μg of mPBase plasmid, 5 μg of right flank ofnu/nu mice (6–8 weeks old). Tumor volumes were meas- 2 PB transposon plasmid, and 30 μL of Fugene6 (Roche) per million ured twice a week and calculated using the formula V = 0.5 × a × b , SMB21 cells. After 1 week of selection with puromycin (1 μg/mL), where a and b are the shortest and longest perpendicular tumor 3 30 million cells were plated in soft agar. After 4 to 8 weeks of 1 μmol/L diameters, respectively. When tumors reached 150 mm , animals sonidegib selection, 29 individual resistant clones were isolated. were randomly separated into treatment groups (5 mice per group). Twenty-seven clone cell lines were successfully established. Transpo- Sonidegib was administered at 80 mpk by oral gavage once daily. son insertion identification was described in Supplementary Methods. Sonidegib was formulated as diphosphate salt in 0.5% methylcel- lulose and 0.5% Tween 80 (Fisher). Mice were euthanized when 3 Rescue by Transposon Cargo Removal with tumors exceeded 2,000 mm . Tumor progression data were initially Lentiviral-Delivered Flpo reported (16). The Kaplan–Meier survival analysis and cilia analysis are reported in this study. Transposon cargo was flanked by two FRT sites. Its removal Cells (5 × 105) in 2 μL were injected orthotopically into the cerebel- was achieved using lentiviral-delivered Flpo. Removal of transposon lum of nu/nu mice (6–8 weeks old). Injection coordinates are (relative cargo rendered cells sensitive to puromycin. For transposon cargo to bregma): anterior, 6.5 mm; lateral, 1.0 mm; depth, 2 mm. After 6 removal, 0.5 × 10 SMB cells were infected with 1 mL of viral super- engraftments were confirmed, mice were randomly separated into natant of LV-tdTomato or LV-Flp. Forty-eight hours after infection, treatment or control groups (8 mice per group). Animal were treated blasticidin (2 μg/mL) was added, and cells were selected for 3 weeks. with vehicle control, sonidegib (80 mpk daily p.o.), or JQ1 (50 mpk daily i.p.). Tumor growth and response to drug treatments were mon- Cell Survival Assays with Pharmacologic Inhibition itored by measurement of whole-body bioluminescence once a week. SMB cells were seeded in 96-well plates (3 × 104 cells per well). Animals were euthanized once they entered moribund status (end Serial dilutions of the relevant compound in DMSO were used, point for survival analyses). Tumor growth and metastatic spread yielding final drug concentrations ranging from 10 μmol/L to were quantified by measuring bioluminescence imaging intensity 0.01 nmol/L. In all cases, the final volume of DMSO did not exceed over the head or over the spinal cord, respectively. 1%. Cells were incubated for 72 hours following addition of sonidegib, vismodegib, or cyclopamine, or for 96 hours following the BCC Human Data Analysis and the SYSCILIA addition of ATO or JQ1. Cell viability was measured using CellTiter Gold-Standard List of Ciliary Genes 96 Aqueous One Solution (Promega) and calculated as a percentage Publicly available whole-exome sequencing of resistant and of control (DMSO-treated cells). A minimum of three replicates were untreated BCC samples was analyzed based on reported single-nucle- performed for each cell line and drug combination. Survival curves otide variants and insertions/deletions (original data doi: 10.1016/j. were modeled using a nonlinear regression curve fit with a sigmoid ccell.2015.02.001; ref. 10). dose–response, and displayed using GraphPad Prism 5. Note that Mutations in ciliary genes were determined using the SYSCILIA 10 mmol/L stocks of compounds were made in DMSO and stored gold-standard list of 303 genes for known ciliary components (20). at −20°C. Sonidegib (LDE225), vismodegib (GDC-0449), and cyclopamine were purchased from Selleck Chemicals. ATO was purchased Ciliomor Signature and ssGSEA of Patients with BCC from Sigma Aldrich. JQ1 was purchased from Cayman Chemicals. The Ciliomor signature was defined by the top 50 differentially upregulated genes together with the top 50 differentially down- Immunocytochemistry and Immunoblotting regulated genes in cilia mutant cells compared to parental sensitive For immunostaining, cells were fixed in 4% paraformaldehyde, cells. Expression heat map of Ciliomor signature genes is shown in permeablized with PBS containing 0.2% Triton X-100, blocked with Supplementary Fig. S7. PBS containing 0.1% Triton X-100, 5% NGS, and then incubated with ssGSEA, an extension of GSEA, calculates an enrichment score primary and secondary antibody solutions and DAPI. Primary anti- for each sample against one gene set (33). The ssGSEA enrichment bodies used: Gamma-Tubulin (Sigma; 1:250) and Acetylated Tubulin score represents the degree to which the genes in this gene set

December 2017 CANCER DISCOVERY | 1447

15-CD-17-0281_p1436-1449.indd 1447 11/17/17 2:34 PM Published OnlineFirst September 18, 2017; DOI: 10.1158/2159-8290.CD-17-0281

RESEARCH ARTICLE Zhao et al.

are coordinately upregulated or downregulated within a sample. ssGSEA was applied to generate an enrichment score of Ciliomor References signature in each patient sample. Gene expression data of nor- 1. Rubin LL, de Sauvage FJ. Targeting the Hedgehog pathway in cancer. mal skin (n = 8), sensitive (n = 4), and resistant BCC (n = 9) sam- Nat Rev Drug Discov 2006;5:1026–33. ples were obtained from GEO (GSE58375; ref. 9). Gene expression 2. Oro AE, Higgins KM, Hu Z, Bonifas JM, Epstein EH Jr, Scott MP. data of treatment-naïve medulloblastoma samples were from GEO Basal cell carcinomas in mice overexpressing sonic hedgehog. Science (GSE49243; ref. 17). 1997;276:817–21. 3. Xie J, Murone M, Luoh SM, Ryan A, Gu Q, Zhang C, et al. Activating Statistical Analysis Smoothened mutations in sporadic basal-cell carcinoma. Nature 1998;391:90–2. Statistical analyses were performed with the Student t test, one- 4. Taylor MD, Northcott PA, Korshunov A, Remke M, Cho YJ, Clifford way ANOVA with Dunnett or Bonferroni post hoc test as indicated. SC, et al. Molecular subgroups of medulloblastoma: the current con- A P value of 0.05 or less was considered statistically significant. All sensus. Acta Neuropathol 2012;123:465–72. data analyses were performed using Microsoft Excel or GraphPad 5. Goodrich LV, Milenkovic L, Higgins KM, Scott MP. Altered neural Prism 5. cell fates and medulloblastoma in mouse patched mutants. Science 1997;277:1109–13. Accession Numbers 6. Hahn H, Wicking C, Zaphiropoulous PG, Gailani MR, Shanley S, Chidambaram A, et al. Mutations of the human homolog of Dros- RNA-seq data are accessible through GEO series accession number ophila patched in the nevoid basal cell carcinoma syndrome. Cell GSE103538. 1996;85:841–51. 7. Johnson RL, Rothman AL, Xie J, Goodrich LV, Bare JW, Bonifas JM, Disclosure of Potential Conflicts of Interest et al. Human homolog of patched, a candidate gene for the basal cell J.F. Kelleher is Head of Discovery Biology at Kymera Therapeutics nevus syndrome. Science 1996;272:1668–71. and has ownership interest (including patents) in the same. No 8. Peukert S, Miller-Moslin K. Small-molecule inhibitors of the hedge- potential conflicts of interest were disclosed by the other authors. hog signaling pathway as cancer therapeutics. ChemMedChem 2010;5:500–12. Authors’ Contributions 9. Atwood SX, Sarin KY, Whitson RJ, Li JR, Kim G, Rezaee M, et al. Smoothened variants explain the majority of drug resistance in basal Conception and design: X. Zhao, E. Pak, J.F. Kelleher, R.A. Segal cell carcinoma. Cancer Cell 2015;27:342–53. Development of methodology: X. Zhao, E. Pak, K.J. Ornell, 10. Sharpe HJ, Pau G, Dijkgraaf GJ, Basset-Seguin N, Modrusan Z, Janu- T. Ponomaryov ario T, et al. Genomic analysis of smoothened inhibitor resistance in Acquisition of data (provided animals, acquired and managed basal cell carcinoma. Cancer Cell 2015;27:327–41. patients, provided facilities, etc.): X. Zhao, E. Pak, K.J. Ornell, 11. Robinson GW, Orr BA, Wu G, Gururangan S, Lin T, Qaddoumi I, M.F. Pazyra-Murphy, E.L. MacKenzie, E.J. Chadwick et al. Vismodegib exerts targeted efficacy against recurrent sonic Analysis and interpretation of data (e.g., statistical analysis, hedgehog-subgroup medulloblastoma: results from phase II pediat- biostatistics, computational analysis): X. Zhao, E. Pak, K.J. Ornell, ric brain tumor consortium studies PBTC-025B and PBTC-032. J Clin M.F. Pazyra-Murphy, E.L. MacKenzie, R.A. Segal Oncol 2015;33:2646–54. Writing, review, and/or revision of the manuscript: X. Zhao, 12. Yauch RL, Dijkgraaf GJ, Alicke B, Januario T, Ahn CP, Holcomb T, E. Pak, K.J. Ornell, T. Ponomaryov, J.F. Kelleher, R.A. Segal et al. Smoothened mutation confers resistance to a Hedgehog path- Administrative, technical, or material support (i.e., reporting way inhibitor in medulloblastoma. Science 2009;326:572–4. or organizing data, constructing databases): X. Zhao, E. Pak, 13. Rudin CM, Hann CL, Laterra J, Yauch RL, Callahan CA, Fu L, et al. K.J. Ornell, M.F. Pazyra-Murphy Treatment of medulloblastoma with hedgehog pathway inhibitor Study supervision: X. Zhao GDC-0449. N Engl J Med 2009;361:1173–8. 14. Atwood SX, Li M, Lee A, Tang JY, Oro AE. GLI activation by atypical Acknowledgments protein kinase C iota/lambda regulates the growth of basal cell carcinomas. Nature 2013;494:484–8. We thank Stephan Teglund for Sufu+/− and Sufu+/+ MEF cells, 15. Buonamici S, Williams J, Morrissey M, Wang A, Guo R, Vattay Jeremy F. Reiter for OFD1 antibody, Anthony E. Oro for sharing A, et al. Interfering with resistance to smoothened antagonists by unpublished information, Harvard Medical School EM facility inhibition of the PI3K pathway in medulloblastoma. Sci Transl Med for help with TEM images, Lurie Family Imaging Center at Dana- 2010;2:51ra70. Farber Cancer Institute for preclinical studies, Molecular Biology 16. Zhao X, Ponomaryov T, Ornell KJ, Zhou P, Dabral SK, Pak E, et al. Core Facility at Dana-Farber Cancer Institute for NGS service, RAS/MAPK activation drives resistance to Smo inhibition, metasta- Rodent Histopathology Core at Dana-Farber/Harvard Cancer sis, and tumor evolution in Shh pathway-dependent tumors. Cancer Center (supported by NIH 5 P30 CA06516), and members of the Res 2015;75:3623–35. Segal laboratory for scientific discussion and critical reading of 17. Kool M, Jones DT, Jager N, Northcott PA, Pugh TJ, Hovestadt V, et al. the manuscript. Genome sequencing of SHH medulloblastoma predicts genotype-related response to smoothened inhibition. Cancer Cell 2014;25:393–405. Grant Support 18. Ferrante MI, Zullo A, Barra A, Bimonte S, Messaddeq N, Studer M, The work was funded by grants from the NIH: P01 CA142536, et al. Oral-facial-digital type I protein is required for primary cilia Alex’s Lemonade Stand Foundation, and Emerald Foundation (to formation and left-right axis specification. Nat Genet 2006;38:112–7. 19. Ferrante MI, Giorgio G, Feather SA, Bulfone A, Wright V, Ghiani M, R.A. Segal), and F31CA183145 (to E. Pak). et al. Identification of the gene for oral-facial-digital type I syndrome. The costs of publication of this article were defrayed in part by Am J Hum Genet 2001;68:569–76. the payment of page charges. This article must therefore be hereby 20. van Dam TJ, Wheway G, Slaats GG, Group SS, Huynen MA, Giles marked advertisement in accordance with 18 U.S.C. Section 1734 RH. The SYSCILIA gold standard (SCGSv1) of known ciliary compo- solely to indicate this fact. nents and its applications within a systems biology consortium. Cilia 2013;2:7. Received March 16, 2017; revised July 14, 2017; accepted Septem- 21. Eggenschwiler JT, Anderson KV. Cilia and developmental signaling. ber 11, 2017; published OnlineFirst September 18, 2017. Annu Rev Cell Dev Biol 2007;23:345–73.

1448 | CANCER DISCOVERY December 2017 www.aacrjournals.org

15-CD-17-0281_p1436-1449.indd 1448 11/17/17 2:34 PM Published OnlineFirst September 18, 2017; DOI: 10.1158/2159-8290.CD-17-0281

Cilia Loss as New Paradigm of Resistance to Smo Inhibitors RESEARCH ARTICLE

22. Huangfu D, Anderson KV. Signaling from Smo to Ci/Gli: conser- stability of the Gli2 transcriptional effector. Proc Natl Acad Sci U S A vation and divergence of Hedgehog pathways from Drosophila to 2010;107:13432–7. vertebrates. Development 2006;133:3–14. 28. Filippakopoulos P, Qi J, Picaud S, Shen Y, Smith WB, Fedorov O, et al. 23. Chen MH, Wilson CW, Li YJ, Law KK, Lu CS, Gacayan R, et al. Selective inhibition of BET bromodomains. Nature 2010;468:1067–73. Cilium-independent regulation of Gli protein function by Sufu in 29. Tang Y, Gholamin S, Schubert S, Willardson MI, Lee A, Hedgehog signaling is evolutionarily conserved. Genes Dev 2009; Bandopadhayay P, et al. Epigenetic targeting of Hedgehog pathway 23:1910–28. transcriptional output through BET bromodomain inhibition. Nat 24. Humke EW, Dorn KV, Milenkovic L, Scott MP, Rohatgi R. The Med 2014;20:732–40. output of Hedgehog signaling is controlled by the dynamic associa- 30. Han YG, Kim HJ, Dlugosz AA, Ellison DW, Gilbertson RJ, tion between Suppressor of Fused and the Gli proteins. Genes Dev Alvarez-Buylla A. Dual and opposing roles of primary cilia in medul- 2010;24:670–82. loblastoma development. Nat Med 2009;15:1062–5. 25. Tukachinsky H, Lopez LV, Salic A. A mechanism for vertebrate 31. Wong SY, Seol AD, So PL, Ermilov AN, Bichakjian CK, Epstein EH Hedgehog signaling: recruitment to cilia and dissociation of SuFu- Jr, et al. Primary cilia can both mediate and suppress Hedgehog Gli protein complexes. J Cell Biol 2010;191:415–28. pathway-dependent tumorigenesis. Nat Med 2009;15:1055–61. 26. Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, 32. Singla V, Romaguera-Ros M, Garcia-Verdugo JM, Reiter JF. Ofd1, Gillette MA, et al. Gene set enrichment analysis: a knowledge-based a human disease gene, regulates the length and distal structure of approach for interpreting genome-wide expression profiles. Proc Natl centrioles. Dev Cell 2010;18:410–24. Acad Sci U S A 2005;102:15545–50. 33. Barbie DA, Tamayo P, Boehm JS, Kim SY, Moody SE, Dunn IF, et al. 27. Kim J, Lee JJ, Kim J, Gardner D, Beachy PA. Arsenic antagonizes the Systematic RNA interference reveals that oncogenic KRAS-driven Hedgehog pathway by preventing ciliary accumulation and reducing cancers require TBK1. Nature 2009;462:108–12.

December 2017 CANCER DISCOVERY | 1449

15-CD-17-0281_p1436-1449.indd 1449 11/17/17 2:34 PM Published OnlineFirst September 18, 2017; DOI: 10.1158/2159-8290.CD-17-0281

A Transposon Screen Identifies Loss of Primary Cilia as a Mechanism of Resistance to SMO Inhibitors

Xuesong Zhao, Ekaterina Pak, Kimberly J. Ornell, et al.

Cancer Discov 2017;7:1436-1449. Published OnlineFirst September 18, 2017.

Updated version Access the most recent version of this article at: doi:10.1158/2159-8290.CD-17-0281

Supplementary Access the most recent supplemental material at: Material http://cancerdiscovery.aacrjournals.org/content/suppl/2017/09/16/2159-8290.CD-17-0281.DC1

Cited articles This article cites 33 articles, 13 of which you can access for free at: http://cancerdiscovery.aacrjournals.org/content/7/12/1436.full#ref-list-1

Citing articles This article has been cited by 12 HighWire-hosted articles. Access the articles at: http://cancerdiscovery.aacrjournals.org/content/7/12/1436.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerdiscovery.aacrjournals.org/content/7/12/1436. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.