Companion Diagnostic, Pharmacogenomic, and Cancer Biomarkers

PIM Kinases Are a Potential Prognostic Biomarkerand Therapeutic Target in NeuroblastomaDiede Brunen, Romy C. de Vries, Cor Lieftink, Roderick L. Beijersbergen, andRen�e Bernards

Abstract

The majority of high-risk neuroblastoma patients are refrac-tory to, or relapse on, current treatment regimens, resultingin 5-year survival rates of less than 50%. This emphasizes theurgent need to identify novel therapeutic targets. Here, wereport that high PIM kinase expression is correlated with pooroverall survival. Treatment of neuroblastoma cell lines withthe pan-PIM inhibitors AZD1208 or PIM-447 suppressed pro-liferation through inhibition of mTOR signaling. In a panel ofneuroblastoma cell lines, we observed a marked binary

response to PIM inhibition, suggesting that specific geneticlesions control responses to PIM inhibition. Using a genome-wide CRISPR-Cas9 genetic screen, we identified NF1 loss as themajor resistance mechanism to PIM kinase inhibitors. Treat-ment with AZD1208 impaired the growth of NF1 wild-typexenografts, while NF1 knockout cells were insensitive. Thus,our data indicate that PIM inhibition may be a novel targetedtherapy in NF1 wild-type neuroblastoma. Mol Cancer Ther; 17(4);849–57. �2018 AACR.

IntroductionNeuroblastoma is a pediatric malignancy originating from

precursor cells of the sympathetic nervous system (1). It accountsfor 8% to 10% of all childhood cancers and is the most commontype of cancer diagnosed in infants (2). Tumor behavior is highlyheterogeneous, varying from spontaneous regression to aggres-sivemetastatic disease. Patients with high-risk tumors have a poorprognosis due to refractory or relapsed disease, with 5-yearsurvival rates of less than 50% (3). Treatment options are mainlylimited to surgery, radiation, chemotherapy, and stem cell trans-plantation (3). Unfortunately, compared with more commonmalignancies, relatively few targeted agents are being evaluatedfor the treatment of neuroblastoma. The only FDA-approvedtargeted therapy is amAb (dinutuximab) targeting the cell surfaceglycolipid GD2, which is administered in combination withgranulocyte macrophage colony-stimulating factor (GM-CSF),IL2, and 13-cis retinoic acid (isotretinoin; ref. 4). It is thereforeessential to identify novel targeted therapies for the treatment ofneuroblastoma.

PIM serine/threonine kinases (PIM1, -2, and -3) are overex-pressed in many hematologic malignancies and several solidcancers (5). Because of a short half-life, the activity of theseconstitutively active kinases is mainly regulated at a transcrip-tional level by the JAK/STAT pathway (6–9). Many PIM kinasesubstrates are commonly regulated by AKT/mTOR signaling,including PRAS40 (10), TSC2 (11), 4EBP1 (12), the proapoptotic

protein BAD (13–15), and the cell-cycle–regulating proteinsp21CIP1 (16) and p27KIP1 (17). PIM kinases are highly homolo-gous and share a unique hinge region that results in a distinct ATP-binding site (9, 18), which has led to the development of highlyspecific pan-PIM kinase inhibitors, such as AZD1208 (19) andPIM-447 (20). Furthermore, Pim1-, Pim2-, and Pim3-deficientmice are viable and only display mild phenotypic changes, suchas reduced body size, which might indicate a favorable toxicityprofile for PIM kinase inhibitors in human patients (21).

Although PIM kinases are, thus far, mostly studied in hema-tologic malignancies, an increasing number of reports demon-strate a role for these kinases in solid cancers. For instance,PIM1 was recently found to be overexpressed in triple-negativebreast cancer, in which high mRNA expression correlated withpoor clinical outcome (22, 23). In addition, MYC-positivetriple-negative cell lines and patient-derived xenografts werefound to be sensitive to PIM inhibitors. Here, we explorePIM kinase inhibition as a potential strategy for the treatmentof neuroblastoma.

Materials and MethodsPatient cohort

Data from two independent neuroblastoma microarray data-sets were derived from the online tool R2 (http://r2.amc.nl). TheVersteeg dataset consisted of 88 samples and has previously beendescribed. The Kocak dataset consisted of 643 samples, of whichfor 476 samples, follow-up datawere available andwere thereforeused for the analysis. Patient characteristics are described inSupplementary Table S1. Patient studies were conducted in com-pliance with the Helsinki declaration.

Cell culture, transfection, and lentiviral transductionHEK293, HT230, SK-N-SH, NB90-4, NB90-6, NB90-9, NB69,

SK-N-AS, and SK-N-BE cell lines were cultured in DMEM. LAN5,CLB-MA, GI-ME-N, and KELLY cell lines were cultured inRPMI1640. IMR32 was cultured in Minimum Essential Medium.SH-SY5Y was cultured in DMEM/F12 medium. Medium was

Division of Molecular Carcinogenesis and Cancer Genomics Center Netherlands,The Netherlands Cancer Institute, Amsterdam, the Netherlands.

Note: Supplementary data for this article are available at Molecular CancerTherapeutics Online (http://mct.aacrjournals.org/).

Corresponding Author: Ren�e Bernards, The Netherlands Cancer Institute,Plesmanlaan 121, Amsterdam 1066 CX, the Netherlands. Phone: 312-0512-6973; Fax: 312-0512-1954; E-mail: r.bernards@nki.nl

doi: 10.1158/1535-7163.MCT-17-0868

�2018 American Association for Cancer Research.

MolecularCancerTherapeutics

www.aacrjournals.org 849

on February 20, 2020. © 2018 American Association for Cancer Research. mct.aacrjournals.org Downloaded from

Published OnlineFirst February 13, 2018; DOI: 10.1158/1535-7163.MCT-17-0868

supplemented with 8% FCS, 1% glutamine, and 1% penicillin/streptomycin, and cells were grown at 37�C and 5% CO2. Celllines were available within the Netherlands Cancer Institute(Amsterdam, the Netherlands), were STR profiled when referenceSTR datawere available for the particular cell line, and all cell lineswere routinely tested by PCR for mycoplasma contamination andwere found negative.

PEI transfection of HEK293 cells was used to produce lentiviralsupernatant. Target cells were seeded 1 day before infection.Lentivirus was added to the medium in the presence of 5 mg/mLpolybrene. Cells were selected for successful lentiviral integrationusing 2 mg/mL puromycin. For individual knockout of genes, thefollowing gRNAs were cloned into the lentiCRISPR version 2.1(LC2.1) vector by Gibson cloning:

gNT 50-ACGGAGGCTAAGCGTCGCAA -30

gNF1 #1 50-TCTTTAGTCGCATTTCTACC-30

gNF1 #2 50-ACACTGGAAAAATGTCTTGC-30

gTSC1 50-ATGAAAGAGTGCGTACACAC-30

gDEPDC5 #1 50-TGTTAATGTCGTAGACCCTA-30

gDEPDC5 #2 50- GGAACACTTGAGTCACAGTC-30

gPTEN 50-ACCGCCAAATTTAATTGCAG-30

gSTK11 50- GAAGGGGAGCTACGCCATCC-30

CRISPR-Cas9 resistance screenThe genome-wide GECKO version 2 library A, consisting of

65,383 gRNAs in lentiviral vectors, was used to infect KELLY cellswith 200-fold coverage. After 48 hours, viral supernatant wasreplaced by medium containing puromycin to select for infectedcells. Forty-eight hours after puromycin selection, cells wereharvested, a T0 sample was taken, and the rest of the cells werereseeded and cultured in the absence or presence of drug in twobiological replicates for 30 days, followed by genomic DNAisolation using the DNeasy Blood & Tissue Kit (Qiagen). gRNAsequences were amplified using Phusion High Fidelity DNApolymerase (Thermo Fisher Scientific) by performing a two-stepPCRamplification: (i) 98�C, 2minutes; (ii) 98�C, 30 seconds; (iii)60�C, 30 seconds; (4) 72�C, 1 minutes; (v) to step 2, 20 cycles(PCR1) or 15 cycles (PCR2); (vi) 72�C, 5 minutes; (vii) 4�C. Twomicroliters of the pooled PCR1 product was used for thesubsequent PCR2 reaction. The abundance of each gRNA inthe treated versus untreated pools was determined by Illuminanext-generation sequencing. Statistical analysis was performedusing DESeq version 1.8.3. We considered gRNAs with a log2fold change �5 and a Padj � 0.1 as hits (adjusted for multipletesting using the Benjamini–Hochberg procedure). Becauseseveral of the hits were gRNAs targeting negative regulators ofAKT/mTOR signaling, we used these gRNAs for subsequentvalidation experiments. The data of the CRISPR screen can befound in Supplementary Table S2.

Protein lysate preparation and Western blot analysisCells were lysed using RIPA buffer containing 150 mmol/L

NaCl, 50 mmol/L Tris pH 8.0, 1% NP-40, 0.5% sodium deox-ycholate, and 0.1% SDS supplemented with protease inhibitors(Complete, Roche) and phosphatase inhibitor cocktails II and III(Sigma). Samples were equalized by measuring relative proteinconcentrations using Pierce BCAProtein Assay Kit (Thermo FisherScientific). Bolt LDS Sample Buffer (Thermo Fisher Scientific) and10� Bolt Sample Reducing Agent (Thermo Fisher Scientific) wereadded, lysates were boiled for 10 minutes, and equal amounts of

sample were subjected to SDS gel electrophoresis, followed byWestern blotting. Primary antibodies against HSP90 (sc-7947),GAPDH (sc-20357), p-ERK1/2 (sc-7383), ERK1 (sc-93), ERK2(sc-154), STAT5 (sc-835), and MYC (sc-40) were purchased fromSanta Cruz Biotechnology. Antibody against NF1 (A300-140A)was purchased from Bethyl Laboratories. Antibodies againstclPARP (#5625), PIM1 (#3247), PIM2 (#4730), PIM3 (#4165),pSTAT3 (#9145), STAT3 (#4904), pSTAT5 (#4322), N-MYC(#9405), p-AKT (#4060), AKT (#2920), p-S6 (#2215), S6(#2217), p-4EBP1 (#9456), 4EBP1 (#9452), STK11 (#3050),PTEN (#9559), and TSC1 (#4906) were from Cell SignalingTechnology. Secondary antibodies were obtained from Bio-RadLaboratories. AZD1208 and MK2206 were purchased from Sell-eckchem. Cobimetinib was purchased from ApexBio. PIM-447was purchased from MedChem Express.

CellTiter-Blue viability assayRespectively, 625 (GI-ME-N), 800 (HT230), 2,000 (IMR32),

3,000 (NB69), 5,000 (NB90-6, SK-N-SH, SH-SY5Y, SK-N-BE, SK-N-AS, KELLY, NB90-4), 10,000 (NB90-9), and 15,000 (LAN5,CLB-MA) cells per well were seeded in a 96-well plate. After 24hours, drugs were added to the medium in 2-fold serial dilutionsusing a HP Direct Digital Dispenser. After 5 days of cultureCellTiter-Blue (Promega) was added. The conversion of resazurininto resorufin was measured by using an EnVision MultilabelReader. Treatment with 10 mmol/L phenyl arsenic oxide, resultingin complete cell death, was used as a baseline for viability.Nonlinear regression analysis in GraphPad Prism v6.0 was usedfor curve fitting.

RNAi and quantitative RT-PCRKELLY cells (10,000 cells/well) and SH-SY5Y (4,000 cells/well)

were seeded in triplicates in in a 96-well plate. After 24 hours, cellswere transfected with Dharmacon siGENOME SMARTPool siR-NAs against PIM1 (M-003923-00), PIM2 (M-005359-00), PIM3(M-032287-02), or nontargeting siRNA Pool #1 (D-001206-13)using DharmaFECT1 (#T-2001; KELLY), or Lipofectamine RNAi-Max (#13778075). Phenyl arsenic oxide (10 mmol/L), resulting incomplete cell death, was used as a baseline for viability. After 5days of culture, viability was measured using CellTiter-Blue (Pro-mega) as described above.

To determine the level of knockdown, 5 � 105 KELLY cells perwell and 1 � 106 SH-SY5Y cells per well were seeded in a 6-wellplate. After 24 hours, cells were transfected as indicated previous-ly. After 48 hours, RNAwas isolated using a ISOLATE II RNAMiniKit (Bioline, # BIO-52073). Subsequent cDNA synthesis wasperformed using Maxima Universal First Strand cDNA SynthesisKit (Thermo Fisher Scientific, # K1661).

The 7500 Fast Real-Time PCR System from Applied Biosystemswas used to measure mRNA levels. mRNA expression levels werenormalized to expression of RPL-13. The following primersequences were used in the SYBR Green Master Mix:

RPL-13_forward 50-GAGACAGTTCTGCTGAAGAACTGAA-30;RPL-13_reverse 50-TCCGGACGGGCATGAC-30;PIM1_forward 50-CGACATCAAGGACGAAAACATC-30;PIM1_reverse 50- ACTCTGGAGGGCTATACACTC-30;PIM2_forward 50-TCGAGGCCGAGTATCGACT-30;PIM2_reverse 50-ATTCCGGGGAATCACTTTG-30;PIM3_forward 50-CTGCTCAAGGACACGGTCTACAC-30;PIM3_reverse 50-CCCCACACACCATATCGTAGAGA-30;

Brunen et al.

Mol Cancer Ther; 17(4) April 2018 Molecular Cancer Therapeutics850

on February 20, 2020. © 2018 American Association for Cancer Research. mct.aacrjournals.org Downloaded from

Published OnlineFirst February 13, 2018; DOI: 10.1158/1535-7163.MCT-17-0868

Colony formation assay

Respectively, 20,000 (NB90-4, NB90-6, NB90-9, GI-ME-N,HT230), 40,000 (NB69, SK-N-BE, SK-N-AS, SK-N-SH, KELLY),and 100,000 (IMR32, CLB-MA, LAN5, SH-SY5Y) cells per wellwere seeded in 6-well plates and cultured both in the absence andpresence of drugs as indicated. When untreated cells reachedconfluency, cells were washed with PBS, fixed with 4% formal-dehyde, and stainedwith 0.1% crystal violet. For SH-SY5Y and SK-N-SH, nontargeting cells were stained at a later time point (whenuntreatedwell reached confluence) thanNF1KOcells, as the lattergrew significantly faster.

Mouse xenograftsEight-week-old male/female NMRI nude mice were housed

in a specific pathogen-free facility in individually ventilatedcages at the Animal Core Facilities of the Netherlands CancerInstitute. All mouse experiments were performed in compliancewith protocols approved by the local Animal Ethics Committee,which conform to institutional and national regulatory stan-dards on experimental animal usage. KELLY cells in PBS/Matri-gel mixture were subcutaneously implanted into the flanks ofmice (1 � 106 cells were used in the 90 mg/kg experiment and5 � 106 cells were used in the 30 mg/kg experiment). Whentumor size reached approximately 80 mm3 (90 mg/kg exper-iment) or approximately 200 mm3 (30 mg/kg experiment),mice were randomly assigned and treated once daily with 30 or90 mg/kg AZD1208 by oral gavage. Control group receivedvehicle (0.5% hydroxypropyl methylcellulose and 0.1% Tween80). All groups were composed of 8 mice, although not all miceformed tumors. In the 90 mg/kg AZD1208 experiment, thegroups consisted, therefore, of 7 (gNT vehicle), 6 (gNTAZD1208), 7 (gNF1 #1 vehicle), 7 (gNF1 #1 AZD1208), 7(gNF1 #2 vehicle), and 7 (gNF1 #2 AZD1208) mice, respec-tively. Grubbs' analysis was used to exclude outliers. Tumorvolume was measured three times per week with calipers andcalculated as tumor volume ¼ (length � width2) � 0.5.

Statistical analysisKaplan–Meier survival curves were compared using a log-

rank test in GraphPad Prism v6.0. HRs were calculated usingboth a univariate and multivariate Cox regression analysis inIBM SPSS Statistics version 21. Statistical analysis of PIMexpression in MYCN-amplified versus MYCN-nonamplifiedtumors was performed using two-tailed equal variance Studentt tests in GraphPad Prism v6.0. PIM expression levels of eachtumor stage were compared with stage IV tumors using a one-way ANOVA test with Dunnett correction for multiple compar-isons in GraphPad Prism v6.0. Differences in viability uponPIM knockdown were analyzed using a one-way ANOVA testwith Dunnett correction for multiple comparisons in GraphPadPrism v6.0. The mean � SEM is shown in figures. A P valueof less than 0.05 was considered statistically significant andis denoted by �. P � 0.01 is denoted by ��, P � 0.001 by ���, andP � 0.0001 by ����.

ResultsHigh PIM expression is associated with poor clinical outcome

To determine whether PIM expression is associated with sur-vival in neuroblastoma patients, we analyzed microarray datafrom a large cohort (n¼ 476) of neuroblastoma patients (24).We

classified patients using median PIM expression as a cut-off value(Fig. 1A–C). Interestingly, high expression of either PIM1, -2, or -3was significantly correlated with poor overall survival in thiscohort. Because PIM kinases are highly redundant, we hypothe-sized that patients with low expression of all PIM genes wouldhave a better overall survival than patients with high expression ofat least one of the PIMs. This classifier further improved theseparation between patients with good and poor overall survival(Fig. 1D). To determine whether high PIM expression is a prog-nostic biomarker, we calculated HRs using a Cox regressionanalysis (Table 1). Importantly, high expression of at least oneof the three PIM kinases versus low expression of all PIMs wassignificantly associatedwithpoor overall survival in amultivariatemodel and is therefore a potential prognostic biomarker.

We also determined the level ofPIM expression across all tumorstages (Fig. 1E–G). PIM1 and -2 expression was significantlyhigher in stage IV tumors compared with the other stages. StageIV disease indicates the occurrence of metastasis to distant lymphnodes and organs. Stage IVS represents "special neuroblastoma":These metastatic tumors often spontaneously regress and there-fore have favorable prognosis. We furthermore found statisticallysignificant differential expression of PIM2 and -3 betweenMYCN-amplified and nonamplified tumors (Fig. 1H). The same trendsbetween PIM expression and outcome were observed in a secondcohort of 88 patients (Supplementary Fig. S1; ref. 25), but not alldifferences reached statistical significance, most likely due to thesmaller cohort size.

PIMkinases are a potential therapeutic target in neuroblastomaTo investigate PIM kinases as a potential therapeutic target in

neuroblastoma, we first analyzed PIM protein levels in a panel of14 neuroblastoma cell lines (Fig. 2A). The majority of cell linesexpressed high levels of PIM1 and -3 and to a lesser extent PIM2.We observed no differential PIM expression between MYCN-amplified and nonamplified cell lines. Furthermore, several celllines displayed high phospho-STAT3 levels, one of the upstreamregulators of PIM.

To determine the dependency of neuroblastoma cells on PIMkinases, we performed siRNA-mediated depletion of PIM1, -2, or-3 in KELLY and SH-SY5Y cell lines (Fig. 2B and C; SupplementaryFig. S2A and S2B). Even though PIM kinases are described to beredundant, suppression of PIM1 was sufficient to reduce theviability of both cell lines. This is in line with findings in tri-ple-negative breast cancer, in which PIM1 knockdown was suffi-cient to reduce cell viability (22, 23).Wenext investigatedwhetherhigh PIM expression can be exploited therapeutically by treatingall 14 cell lines with the selective pan-PIM inhibitor AZD1208 inboth a short-term CellTiter-Blue cell viability assay (Supplemen-tary Fig. S2C and S2D) and a long-term colony formation assay(Fig. 2D). Interestingly, approximately one third of the cell linesresponded to treatment,whereas theother cell lineswere resistant.We did not observe a clear correlation between PIM protein levelsand PIM inhibitor sensitivity, possibly due to the relative smallnumber of cell lines (Fig. 2A andD).We validated these results byrepeating the colony formation assays with another pan-PIMinhibitor (PIM-447) in 3 AZD1208-sensitive and 3 AZD1208-resistant cell lines (Supplementary Fig. S2E and S2F).

To gain insights into the biochemical differences underlyingthese responses, we treated 3 sensitive and 3 resistant celllines with 2 mmol/L AZD1208 for 0, 4, 24, and 48 hours andanalyzed the activity of downstreamcomponents of PIM signaling

PIM Kinases: A Potential Therapeutic Target in Neuroblastoma

www.aacrjournals.org Mol Cancer Ther; 17(4) April 2018 851

on February 20, 2020. © 2018 American Association for Cancer Research. mct.aacrjournals.org Downloaded from

Published OnlineFirst February 13, 2018; DOI: 10.1158/1535-7163.MCT-17-0868

(Fig. 2E). Only sensitive cell lines demonstrated reduced phos-phorylation of p-70, S6, and 4EBP1 upon AZD1208 treatment,while MYC protein levels remained unaffected upon treatment.Again, we observed similar responses upon treatment with PIM-447 (Supplementary Fig. S2G).

CRISPR-Cas9 screen identifies genes whose loss confers PIMinhibitor resistance

The binary response of the neuroblastoma cell lines to the PIMinhibitors suggests the presence of an underlying mutation thatconfers resistance, which might be used as a biomarker forresponse. To identify genes whose loss renders cells insensitiveto AZD1208 treatment, we performed a loss-of-function genome-wide CRISPR-Cas9 genetic screen (Fig. 3A). We infected the PIMinhibitor–sensitive KELLY cell line with lentivirus containing theGECKO library A (containing 3 gRNAs for each gene), cultured

cells in the presence or absence of 2 mmol/L AZD1208 for 30 days,followed by genomic DNA isolation, PCR amplification of thegRNAs, and next-generation sequencing. To assess the efficacy ofour screen, we analyzed the abundance of gRNAs targeting non-essential and essential genes in the untreated cells versus timepoint zero (T0; Supplementary Fig. S3). Whereas the relativeabundance of gRNAs targeting nonessential genes remained sta-ble over time, gRNAs targeting essential geneswere depleted in theuntreated cells comparedwith T0, thus validating the screen. Next,we analyzed which gRNAs were enriched in the treated versusuntreated samples (Supplementary Table S2). Interestingly, with-in this list of 28 hits, we found a gRNA targeting the RAS-GAPNF1and gRNAs targeting negative regulators of the AKT/mTOR path-wayPTEN, TSC1,DEPDC5, and STK11 to be significantly enriched(Fig. 3B; Supplementary Table S2). We validated these hits bycreating polyclonal knockout cell lines, in whichwe subsequently

Table 1. Cox regression survival analysis of the Kocak cohort

Univariate analysis Multivariate analysisCovariate HR (95% CI) P HR (95% CI) P

MYCN gene amplification 9.23 (6.05–14.07) 5.4e�25 4.06 (2.48–6.65) 2.5e�8Stage III/IV 9.59 (5.11–18.02) 2.1e�12 3.40 (1.71–6.77) 0.001Age > 18 months 9.79 (5.77–16.61) 2.7e�17 3.63 (1.99–6.65) 2.8e�5High PIM1 expression 2.14 (1.38–3.31) 0.001 1.06 (0.65–1.72) 0.827High PIM2 expression 2.00 (1.29–3.08) 0.002 1.06 (0.65–1.73) 0.812High PIM3 expression 2.31 (1.49–3.59) 2.0e�4 0.89 (0.53–1.50) 0.666High PIM1, -2, or -3 expression 5.29 (2.15–13.03) 2.9e�4 3.05 (1.03–8.97) 0.043

Abbreviation: CI, confidence interval.

Figure 1.

PIM expression is a prognostic biomarker in neuroblastoma. A–D, High PIMmRNA expression correlates with poor overall survival in neuroblastoma. Kaplan–Meiersurvival curves of the Kocak cohort (476 patients) demonstrate overall patient survival, using median PIM expression as a cut-off value for high versus lowexpression. High PIM1/2/3 indicates patients with high expression of either PIM1, -2, or -3. P values were calculated using a log-rank test. E–G, PIM1 and -2 expressionis highest in stage IV neuroblastoma patients (Kocak cohort). P values were calculated using a one-way ANOVA and Dunnett test. P � 0.05 (�), P � 0.01(��), P � 0.001 (���), and P � 0.0001 (����). H, PIM expression in MYCN-amplified and nonamplified tumors (Kocak cohort). P values were calculated using anunpaired t test. P � 0.01 (��).

Brunen et al.

Mol Cancer Ther; 17(4) April 2018 Molecular Cancer Therapeutics852

on February 20, 2020. © 2018 American Association for Cancer Research. mct.aacrjournals.org Downloaded from

Published OnlineFirst February 13, 2018; DOI: 10.1158/1535-7163.MCT-17-0868

Figure 2.

PIM kinases are expressed in neuroblastoma and are a potential therapeutic target. A, Neuroblastoma cell lines express PIM kinases. Western blot analysis oflysates generated from 14 neuroblastoma cell lines. B andC,Knockdown of PIM reduces cell line viability. KELLY and SH-SY5Y cell lines were plated in 96-well plates,transfected with a scrambled siRNA (siCTRL) or siRNAs targeting PIM1, -2, or -3, cultured for 5 days, and viability was measured using CellTiter-Blue (n ¼ 3).D, Neuroblastoma cell lines demonstrate differential sensitivity to AZD1208 in a colony formation assay. Cells were plated in 6-well plates and treated withincreasing concentrations of AZD1208. Drug-containing medium was refreshed every 3 days. Cells were fixed, stained, and scanned when untreated wellsreached confluency (n ¼ 3). E, AZD1208-resistant cell lines demonstrate sustained mTOR signaling upon treatment. Cells were treated with 2 mmol/LAZD1208 for 0, 24, and 48 hours, followed by Western blot analysis (n ¼ 3).

PIM Kinases: A Potential Therapeutic Target in Neuroblastoma

www.aacrjournals.org Mol Cancer Ther; 17(4) April 2018 853

on February 20, 2020. © 2018 American Association for Cancer Research. mct.aacrjournals.org Downloaded from

Published OnlineFirst February 13, 2018; DOI: 10.1158/1535-7163.MCT-17-0868

determined the response to AZD1208 using colony formationassays (Fig. 3C).Whereas KELLY cells infectedwith a nontargetinggRNA remained sensitive to AZD1208, knockout of either of thehit genes conferred resistance to PIM inhibitor treatment. Knock-out was assessed by Western blot analysis if respective antibodieswere available (Fig. 3D). Because all hit genes are known toregulate mTOR signaling, we analyzed the activity of S6 and4EBP1 upon AZD1208 treatment in the context of the particularknockouts (Fig. 3D). Similar to our observations in neuroblas-toma cell lines that are intrinsically resistant to PIM inhibitortreatment (Fig. 2D), knockout of the genes identified in the

genetic screen resulted in persistent mTOR signaling upon PIMinhibition.

NF1 protein expression is a potential predictive biomarker forPIM inhibitor treatment

NF1 protein is estimated to be lost in approximately 1 of 3neuroblastoma tumors (26), whereas mutations in PTEN,TSC1, DEPDC5, and STK11 rarely occur or have not beenreported in neuroblastoma. Therefore, we analyzed the cellline panel for NF1 protein levels to investigate whether NF1loss could be a potential biomarker to predict PIM inhibitor

Figure 3.

A CRISPR-Cas9 genetic screen identifies mediators of PIM inhibitor resistance. A, Schematic outline of the resistance screen. KELLY cells were infected withthe genome-wide GECKO library A and cultured in the presence or absence of 2 mmol/L AZD1208. After 30 days, genomic DNAwas isolated, gRNA sequences werePCR amplified, and the abundance of each gRNA in the treated versus untreated condition was determined by next-generation sequencing. B, The screen identifiesloss of NF1 and genes regulating AKT/mTOR signaling as potential resistance mechanisms. The x-axis depicts the average number of sequencing reads in theuntreated versus treated samples. The y-axis depicts the log2 fold change in abundance of gRNAs in the treated versus untreated population. A log2 foldchange of 5 was used as cutoff to select hits. C, CRISPR-Cas9–mediated knockout of NF1, PTEN, TSC1, STK11, or DEPDC5 induces resistance to AZD1208. KELLY cellswere plated in 6-well plates and treated with AZD1208. Drug-containing medium was refreshed every 3 days. Cells were fixed, stained, and scanned whenuntreated wells reached confluency (n ¼ 3). D, CRISPR-Cas9–mediated knockout of NF1, PTEN, TSC1, STK11, or DEPDC5 results in sustained mTOR signaling uponPIM inhibition. KELLY cells were treated with 2 mmol/L AZD1208 for 48 hours, followed by Western blot analysis (n ¼ 3). E, NF1 loss correlates with AZD1208resistance in a panel of neuroblastoma cell lines. Western blot analysis of lysates generated from 14 neuroblastoma cell lines.

Brunen et al.

Mol Cancer Ther; 17(4) April 2018 Molecular Cancer Therapeutics854

on February 20, 2020. © 2018 American Association for Cancer Research. mct.aacrjournals.org Downloaded from

Published OnlineFirst February 13, 2018; DOI: 10.1158/1535-7163.MCT-17-0868

resistance (Fig. 3E). Five of 14 cell lines demonstrated completeloss of NF1 and were resistant to AZD1208. Furthermore, theresistant cell line SK-N-AS demonstrated partial loss of NF1,which has previously been reported and, in addition, carries anNRASQ61 mutation (26). The cell lines IMR32, NB90-6, and

CLB-MA were resistant to AZD1208 despite being NF1 wildtype, which suggests that other genes can also modulate theresponse to PIM inhibitors. Consistent with a causal role forNF1 in PIM inhibitor resistance, all AZD1208-sensitive celllines in the panel expressed NF1 protein.

Figure 4.

NF1 loss induces PIM inhibitor resistance. A–D, NF1 loss results in resistance to AZD1208 and PIM-447 in a colony formation assay. KELLY and SH-SY5Y cellswere plated in 6-well plates and treated with increasing concentrations of AZD1208 or PIM-447. Drug-containing medium was refreshed every 3 days. Cells werefixed, stained, and scanned when untreated wells of each particular cell line reached confluency (n ¼ 3). E, NF1 knockout results in enhanced MAPK and AKTsignaling and sustained mTOR signaling upon PIM inhibition. KELLY cell lines were treated with 2 mmol/L AZD1208 for 0, 24, and 48 hours, followed by Westernblot analysis (n¼ 3). F, PIM inhibition suppresses tumor growth of NF1 wild-type KELLY cells in a xenograft model. KELLY cells (1� 106 cells) were subcutaneouslyimplanted in NMRI mice. Once tumors were established (80 mm3), animals were treated with vehicle or AZD1208 (90 mg/kg).

PIM Kinases: A Potential Therapeutic Target in Neuroblastoma

www.aacrjournals.org Mol Cancer Ther; 17(4) April 2018 855

on February 20, 2020. © 2018 American Association for Cancer Research. mct.aacrjournals.org Downloaded from

Published OnlineFirst February 13, 2018; DOI: 10.1158/1535-7163.MCT-17-0868

To further assess the effect of NF1 loss on AZD1208 sensitivity,we created NF1 knockout cells using two different gRNAs inKELLY, SH-SY5Y, and SK-N-SH cell lines. These NF1 knockoutcells were subsequently treated with AZD1208 or PIM-447 in acolony formation assay (Fig. 4A–D; Supplementary Fig. S4A andS4B). In concordance with the initial results, NF1 loss inducedresistance to both AZD1208 and PIM-447. To gain insights intothe biochemical response underlying the resistance, we treatedcontrol and NF1 knockout cell lines with AZD1208 for 0, 4, 24,and 48 hours (Fig. 4E; Supplementary Fig. S4C and S4D). In linewith NF1 being a negative regulator of RAS, we observed strongphosphorylation of downstream ERK and AKT in NF1 knockoutcells. Whereas PIM inhibition suppressed phosphorylation of S6and 4EBP1 in cell lines infected with a nontargeting gRNA, theiractivity was maintained upon NF1 loss.

Because NF1 knockout cells display increased activation ofboth MAPK and AKT, we wondered whether the addition of aMEK or AKT inhibitor could alleviate resistance to PIM inhibitors.To assess the potential of this combination strategy, we cotreatedcontrol and NF1 knockout KELLY, SH-SY5Y, and SK-N-SH cellswith AZD1208, the MEK inhibitor cobimetinib, or the AKTinhibitor MK2206 (Supplementary Fig. S5A–S5C; ref. 27). Inter-estingly, MEK inhibitor treatment sensitized NF1 knockout celllines to AZD1208 in KELLY and SH-SY5Y cells. SK-N-SH cellsdisplayed increased sensitivity to cobimetinib monotreatmentuponNF1 loss and, therefore, demonstrated no additional benefitof PIM inhibitor treatment. Although AKT inhibition partiallyalleviated resistance inKELLY cells, it did not restore PIM inhibitorsensitivity in SH-SY5Y or SK-N-SH cell lines, suggesting thatMAPK activation is the main driver of resistance upon NF1 loss.We subsequently investigated whether dual PIM/MEK inhibitortreatment could also prevent resistance to PIM inhibition in the"naturally" NF1-deficient cell lines from our cell line panel(Supplementary Fig. S5D–S5F). However, in line with a previousreport (28), theseNF1-deficient cells were highly sensitive toMEKinhibitor monotreatment, due to a dependency on MAPK signal-ing, which was not enhanced by AZD1208.

To validate the efficacy of PIM kinase inhibition in neuroblas-toma in vivo, we subcutaneously injected control andNF1 knock-out KELLY cells into nude mice and treated mice once daily withvehicle, 30 mg/kg AZD1208, or 90 mg/kg AZD1208 (Fig. 4F;Supplementary Fig. S6A–S6D). High-dose AZD1208 did notaffect mouse body weight (Supplementary Fig. S6E). WhereasNF1 knockout tumors were found to be insensitive to AZD1208,the survival of mice harboring NF1 wild-type tumors was signif-icantly prolonged upon PIM inhibitor treatment.

DiscussionEven though neuroblastoma is the most common malignancy

in infants, the number of treatment options is limited comparedwith more frequently occurring types of cancer in adults. Becauseof the low absolute number of patients, there is a lack of financialincentive for pharmaceutical companies to develop compoundsthat are solely used in neuroblastoma. To overcome this problem,novel therapeutic targets in neuroblastoma should ideally betargetable by drugs that are currently in use or development formore frequent adult malignancies.

There are several recurrent mutations in neuroblastoma thatmight represent vulnerabilities that can be targeted pharmaco-logically. For instance, ALK mutations occur in 8% to 12% of all

neuroblastoma (29, 30). This observation led to a phase I clinicaltrial assessing the efficacy of the ALK inhibitor crizotinib inneuroblastoma patients. However, of the 11 patients with ALK-mutant tumors, only 1 patient demonstrated a complete responseand 2 patients displayed stable disease upon crizotinib treatment(31). Another frequently occurring aberration is NF1 deficiency.We and others have previously reported the presence of NF1mutations and the loss of NF1 protein in neuroblastoma (26, 28).This is potentially clinically actionable due to the resulting acti-vation of the RAS–MAPK pathway. Indeed, in childhood neuro-fibromatosis type I–related plexiform neurofibromas, the MEKinhibitor selumetinib proved to be very effective, with partialresponses in 17 of 24 patients (32). Consistent with this, NF1-deficient neuroblastoma cells are more sensitive to MEK inhibi-tion than their wild-type counterparts, providing a rationale toevaluate the clinical efficacy of MEK inhibitors in NF1-deficientneuroblastoma (Eleveld and colleagues, 2015; Woodfield andcolleagues, 2016).

However, ALK and NF1 mutations are present in only a smallfraction of neuroblastomas. In the current study, we find that PIMkinase inhibitors may elicit clinical benefit in NF1 wild-typetumors. This is of particular relevance, given that our data alsoindicate that high expression of PIM kinases is associated withpoor clinical outcome. Furthermore, high PIM expressionremained significantly associated with poor clinical outcomewhen tumors with low NF1 expression were excluded (Supple-mentary Fig. S6F). Thus, there is a significant unmet need to findbetter treatments for such patients.

The long-term benefit of targeted therapies is generally ham-pered by the development of resistance. In line with thisobservation, AZD1208 initially suppressed growth of KELLYcells, but did not prevent long-term tumor outgrowth in vivo(Fig. 4F). Additional research is necessary to identify rationallydesigned drug combinations with PIM inhibitors to prevent theoutgrowth of resistant neuroblastoma cells. Current treatmentregimens for (especially high risk) neuroblastoma mostlyinclude a combination of chemotherapeutic drugs. Interesting-ly, it has previously been demonstrated that PIM kinase inhi-bitors can enhance the efficacy of chemotherapeutic agents intriple-negative breast cancer xenografts (23). Similar observa-tions were made in cell lines derived from different tumortissues, in which PIM1 expression antagonized chemotherapy-induced apoptosis (33). It is therefore interesting to speculatethat the addition of a PIM inhibitor might improve the efficacyof current treatment regimens.

In conclusion, we have identified PIM kinases as a noveltherapeutic target in neuroblastoma. Furthermore, NF1 loss canbe used as a biomarker to identify patients that do not benefitfrom PIM inhibitor treatment.

Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.

Authors' ContributionsConception and design: D. Brunen, R. BernardsDevelopment of methodology: D. Brunen, R.L. BeijersbergenAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): D. BrunenAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis):D.Brunen,R.C. deVries,C. Lieftink,R.L. Beijersbergen,R. BernardsWriting, review, and/or revision of the manuscript: D. Brunen, R. Bernards

Brunen et al.

Mol Cancer Ther; 17(4) April 2018 Molecular Cancer Therapeutics856

on February 20, 2020. © 2018 American Association for Cancer Research. mct.aacrjournals.org Downloaded from

Published OnlineFirst February 13, 2018; DOI: 10.1158/1535-7163.MCT-17-0868

Administrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): R.C. de VriesStudy supervision: R. Bernards

AcknowledgmentsThis studywas supported by grants from theDutchCancer Society, theCenter

for Cancer Genomics, and Stand Up to Cancer (SU2C-AACR-DT0912). Wethank all members of the Bernards Laboratory for helpful support and discus-sions. We thank the intervention unit of The Netherlands Cancer Institute for

assistance with the xenograft experiments. We thank Dr. R. Versteeg for makingthe R2 tool publicly available.

The costs of publication of this articlewere defrayed inpart by the payment ofpage charges. This article must therefore be hereby marked advertisement inaccordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Received September 6, 2017; revisedNovember 1, 2017; accepted January 10,2018; published first February 13, 2018.

References1. Marshall GM, Carter DR, Cheung BB, Liu T, Mateos MK, Meyerowitz JG,

et al. The prenatal origins of cancer. Nat Rev Cancer 2014;14:277–89.2. Cheung NK, Dyer MA. Neuroblastoma: developmental biology, cancer

genomics and immunotherapy. Nat Rev Cancer 2013;13:397–411.3. Maris JM. Recent advances in neuroblastoma. N Engl J Med 2010;362:

2202–11.4. Yu AL, Gilman AL, Ozkaynak MF, London WB, Kreissman SG, Chen HX,

et al. Anti-GD2 antibody with GM-CSF, interleukin-2, and isotretinoin forneuroblastoma. N Engl J Med 2010;363:1324–34.

5. Nawijn MC, Alendar A, Berns A. For better or for worse: the role of Pimoncogenes in tumorigenesis. Nat Rev Cancer 2011;11:23–34.

6. Miura O, Miura Y, Nakamura N, Quelle FW, Witthuhn BA, Ihle JN, et al.Induction of tyrosine phosphorylation of Vav and expression of Pim-1correlates with Jak2-mediated growth signaling from the erythropoietinreceptor. Blood 1994;84:4135–41.

7. Stout BA, BatesME, Liu LY, FarringtonNN, Bertics PJ. IL-5 and granulocyte-macrophage colony-stimulating factor activate STAT3 and STAT5 andpromote Pim-1 and cyclin D3 protein expression in human eosinophils.J Immunol 2004;173:6409–17.

8. Matikainen S, Sareneva T, Ronni T, Lehtonen A, Koskinen PJ, Julkunen I.Interferon-alpha activates multiple STAT proteins and upregulates prolif-eration-associated IL-2Ralpha, c-myc, and pim-1 genes in human T cells.Blood 1999;93:1980–91.

9. Qian KC, Wang L, Hickey ER, Studts J, Barringer K, Peng C, et al. Structuralbasis of constitutive activity and a unique nucleotide binding mode ofhuman Pim-1 kinase. J Biol Chem 2005;280:6130–7.

10. Zhang F, Beharry ZM,Harris TE, LillyMB, Smith CD,Mahajan S, et al. PIM1protein kinase regulates PRAS40 phosphorylation and mTOR activity inFDCP1 cells. Cancer Biol Ther 2009;8:846–53.

11. Lu J, Zavorotinskaya T, Dai Y, Niu XH, Castillo J, Sim J, et al. Pim2 isrequired for maintaining multiple myeloma cell growth through modu-lating TSC2 phosphorylation. Blood 2013;122:1610–20.

12. Tamburini J, GreenAS, Bardet V,ChapuisN, Park S,Willems L, et al. Proteinsynthesis is resistant to rapamycin and constitutes a promising therapeutictarget in acute myeloid leukemia. Blood 2009;114:1618–27.

13. Yan B, ZemskovaM,Holder S, Chin V, Kraft A, Koskinen PJ, et al. The PIM-2kinase phosphorylates BAD on serine 112 and reverses BAD-induced celldeath. J Biol Chem 2003;278:45358–67.

14. Aho TL, Sandholm J, Peltola KJ,MankonenHP, LillyM, Koskinen PJ. Pim-1kinase promotes inactivation of the pro-apoptotic Bad protein by phos-phorylating it on the Ser112 gatekeeper site. FEBS Lett 2004;571:43–9.

15. Macdonald A, Campbell DG, Toth R, McLauchlan H, Hastie CJ, Arthur JS.Pim kinases phosphorylate multiple sites on Bad and promote 14-3-3binding and dissociation from Bcl-XL. BMC Cell Biol 2006;7:1.

16. Zhang Y, Wang Z, Magnuson NS. Pim-1 kinase-dependent phosphoryla-tion of p21Cip1/WAF1 regulates its stability and cellular localization inH1299 cells. Mol Cancer Res 2007;5:909–22.

17. Morishita D, Katayama R, Sekimizu K, Tsuruo T, Fujita N. Pim kinasespromote cell cycle progression by phosphorylating and down-regulatingp27Kip1 at the transcriptional and posttranscriptional levels. Cancer Res2008;68:5076–85.

18. Bullock AN, Debreczeni J, Amos AL, Knapp S, Turk BE. Structure andsubstrate specificity of the Pim-1 kinase. J Biol Chem 2005;280:41675–82.

19. Keeton EK, McEachern K, Dillman KS, Palakurthi S, Cao Y, Grondine MR,et al. AZD1208, a potent and selective pan-Pim kinase inhibitor, demon-strates efficacy in preclinical models of acute myeloid leukemia. Blood2014;123:905–13.

20. Burger MT, Nishiguchi G, Han W, Lan J, Simmons R, Atallah G, et al.Identification of N-(4-((1R,3S,5S)-3-Amino-5-methylcyclohexyl)pyridin-3-yl)-6-(2,6-difluorophenyl)- 5-fluoropicolinamide (PIM447), a potentand selective proviral insertion site of moloney murine leukemia (PIM)1, 2, and3kinase inhibitor in clinical trials for hematologicalmalignancies.J Med Chem 2015;58:8373–86.

21. Mikkers H, Nawijn M, Allen J, Brouwers C, Verhoeven E, Jonkers J, et al.Mice deficient for all PIM kinases display reduced body size andimpaired responses to hematopoietic growth factors. Mol Cell Biol2004;24:6104–15.

22. Horiuchi D, Camarda R, Zhou AY, Yau C, Momcilovic O, Balakrishnan S,et al. PIM1 kinase inhibition as a targeted therapy against triple-negativebreast tumors with elevated MYC expression. Nat Med 2016;22:1321–9.

23. Braso-Maristany F, Filosto S, Catchpole S, Marlow R, Quist J, Francesch-Domenech E, et al. PIM1 kinase regulates cell death, tumor growth andchemotherapy response in triple-negative breast cancer. Nat Med 2016;22:1303–13.

24. Kocak H, Ackermann S, Hero B, Kahlert Y, Oberthuer A, Juraeva D, et al.Hox-C9 activates the intrinsic pathway of apoptosis and is associated withspontaneous regression in neuroblastoma. Cell Death Dis 2013;4:e586.

25. Molenaar JJ, Koster J, Zwijnenburg DA, van Sluis P, Valentijn LJ, van derPloeg I, et al. Sequencing of neuroblastoma identifies chromothripsis anddefects in neuritogenesis genes. Nature 2012;483:589–93.

26. Holzel M, Huang S, Koster J, Ora I, Lakeman A, Caron H, et al. NF1 is atumor suppressor in neuroblastoma that determines retinoic acid responseand disease outcome. Cell 2010;142:218–29.

27. Hirai H, Sootome H, Nakatsuru Y, Miyama K, Taguchi S, Tsujioka K, et al.MK-2206, an allosteric Akt inhibitor, enhances antitumor efficacy bystandard chemotherapeutic agents or molecular targeted drugs in vitroand in vivo. Mol Cancer Ther 2010;9:1956–67.

28. Eleveld TF, Oldridge DA, Bernard V, Koster J, Daage LC, Diskin SJ, et al.Relapsed neuroblastomas show frequent RAS-MAPK pathway mutations.Nat Genet 2015;47:864–71.

29. Mosse YP, Laudenslager M, Longo L, Cole KA, Wood A, Attiyeh EF, et al.Identification of ALK as a major familial neuroblastoma predispositiongene. Nature 2008;455:930–5.

30. Bresler SC, Weiser DA, Huwe PJ, Park JH, Krytska K, Ryles H, et al. ALKmutations confer differential oncogenic activation and sensitivity to ALKinhibition therapy in neuroblastoma. Cancer Cell 2014;26:682–94.

31. Mosse YP, LimMS, Voss SD,Wilner K, Ruffner K, Laliberte J, et al. Safety andactivity of crizotinib for paediatric patients with refractory solid tumours oranaplastic large-cell lymphoma: a Children's Oncology Group phase 1consortium study. Lancet Oncol 2013;14:472–80.

32. Dombi E, Baldwin A, Marcus LJ, Fisher MJ, Weiss B, Kim A, et al. Activity ofselumetinib in neurofibromatosis type 1-related plexiformneurofibromas.N Engl J Med 2016;375:2550–60.

33. Chen J, Kobayashi M, Darmanin S, Qiao Y, Gully C, Zhao R, et al. Pim-1plays a pivotal role in hypoxia-induced chemoresistance. Oncogene2009;28:2581–92.

www.aacrjournals.org Mol Cancer Ther; 17(4) April 2018 857

PIM Kinases: A Potential Therapeutic Target in Neuroblastoma

on February 20, 2020. © 2018 American Association for Cancer Research. mct.aacrjournals.org Downloaded from

Published OnlineFirst February 13, 2018; DOI: 10.1158/1535-7163.MCT-17-0868

2018;17:849-857. Published OnlineFirst February 13, 2018.Mol Cancer Ther   Diede Brunen, Romy C. de Vries, Cor Lieftink, et al.   Target in NeuroblastomaPIM Kinases Are a Potential Prognostic Biomarker and Therapeutic

  Updated version

  10.1158/1535-7163.MCT-17-0868doi:

Access the most recent version of this article at:

  Material

Supplementary

  http://mct.aacrjournals.org/content/suppl/2018/02/13/1535-7163.MCT-17-0868.DC1

Access the most recent supplemental material at:

   

   

  Cited articles

  http://mct.aacrjournals.org/content/17/4/849.full#ref-list-1

This article cites 33 articles, 13 of which you can access for free at:

  Citing articles

  http://mct.aacrjournals.org/content/17/4/849.full#related-urls

This article has been cited by 1 HighWire-hosted articles. Access the articles at:

   

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

  Subscriptions

Reprints and

  .pubs@aacr.org

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

  Permissions

  Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://mct.aacrjournals.org/content/17/4/849To request permission to re-use all or part of this article, use this link

on February 20, 2020. © 2018 American Association for Cancer Research. mct.aacrjournals.org Downloaded from

Published OnlineFirst February 13, 2018; DOI: 10.1158/1535-7163.MCT-17-0868