HDAC2 Inhibitors Induce Neuroblastoma Differentiation

Chemistry & Biology

Article Selective HDAC1/HDAC2 Inhibitors Induce Neuroblastoma Differentiation Stacey M. Frumm,1,3 Zi Peng Fan,1,3 Kenneth N. Ross,4 Jeremy R. Duvall,4 Supriya Gupta,4 Lynn VerPlank,4 Byung-Chul Suh,4 Edward Holson,4 Florence F. Wagner,4 William B. Smith,2 Ronald M. Paranal,2 Christopher F. Bassil,1,3 Jun Qi,2 Giovanni Roti,1,3 Andrew L. Kung,5 James E. Bradner,2,6 Nicola Tolliday,4 and Kimberly Stegmaier1,3,4,* 1Department

of Pediatric Oncology of Medical Oncology Dana-Farber Cancer Institute, Boston, MA 02215, USA 3Division of Hematology/Oncology, Boston Children’s Hospital, Boston, MA 02215, USA 4Broad Institute of Massachusetts Institute of Technology and Harvard University, Cambridge, MA 02142, USA 5Department of Pediatrics, Columbia University Medical Center, New York, NY 10032, USA 6Department of Medicine, Harvard Medical School, Boston, MA 02115, USA *Correspondence: [email protected] http://dx.doi.org/10.1016/j.chembiol.2013.03.020 2Department

SUMMARY

While cytotoxic chemotherapy remains the hallmark of cancer treatment, intensive regimens fall short in many malignancies, including high-risk neuroblastoma. One alternative strategy is to therapeutically promote tumor differentiation. We created a gene expression signature to measure neuroblast maturation, adapted it to a high-throughput platform, and screened a diversity oriented synthesisgenerated small-molecule library for differentiation inducers. We identified BRD8430, containing a nine-membered lactam, an ortho-amino anilide functionality, and three chiral centers, as a selective class I histone deacetylase (HDAC) inhibitor (HDAC1 > 2 > 3). Further investigation demonstrated that selective HDAC1/HDAC2 inhibition using compounds or RNA interference induced differentiation and decreased viability in neuroblastoma cell lines. Combined treatment with 13-cis retinoic acid augmented these effects and enhanced activation of retinoic acid signaling. Therefore, by applying a chemical genomic screening approach, we identified selective HDAC1/HDAC2 inhibition as a strategy to induce neuroblastoma differentiation. INTRODUCTION To date, cytotoxic agents have dominated the arsenal of drugs used to treat patients with cancer. Great progress has been made in treating patients with these compounds, but progress has slowed and alternate approaches will be needed to continue to advance patient care. Many cancer types have defects in both proliferation and differentiation, with the former being the target of current chemotherapies. Markedly less effort has gone into identifying compounds that target the differentiation defect, although some prodifferentiating agents have already proven

efficacious in the clinic. For example, all-trans retinoic acid (ATRA) differentiation therapy has revolutionized the care of patients with acute promyelocytic leukemia (APL) (Abdel-Wahab and Levine, 2010; Ade`s et al., 2010). We thus embarked on the task of identifying additional differentiation therapies. Historically, much of phenotype-based screening has focused on the endpoint of cell death. Screening for induction of differentiation is a markedly more challenging task because of the complexity of the target phenotype. In most cases, a single marker gene cannot be used as a read-out for differentiation. Current high-content imaging techniques have enabled screening for morphological changes. Here, however, we opted to use gene expression-based high-throughput screening (GE-HTS), a method that uses gene expression signatures as proxies for biologic state switches. The gene expression signatures can be detected in a high-throughput screening platform that involves ligation-mediated amplification of the genes of interest and a fluorescent bead-based detection (Peck et al., 2006; Stegmaier et al., 2004). At the same time, small-molecule libraries and screening capabilities in academic centers have continued to evolve over the past decade. In the past, standard combinatorial libraries were largely populated by planar, achiral compounds, possibly due to the ease with which these compounds could be made. However, recent evidence has shown that complexity (as measured by sp3 content) and the inclusion of chiral centers are important factors in the transition from discovery through drug development (Lovering et al., 2009). Indeed, many compounds known to disrupt key protein-protein interactions are structurally complex natural products (Koehn and Carter, 2005). Diversity-oriented synthesis (DOS) is a strategy that yields collections of small molecules with structural complexity and diversity mimicking that of natural products (Schreiber, 2000). We thus opted to screen a DOS library of small molecules (Marcaurelle et al., 2010). Furthermore, the DOS library selected for screening was biased for chromatin modification by the incorporation of moieties that bind zinc. Broad transcriptional changes regulate differentiation, and epigenetic alterations have been implicated in the differentiation block observed in cancer cells (Helman et al., 2012; Lotem and Sachs, 2006;

Chemistry & Biology 20, 713–725, May 23, 2013 ª2013 Elsevier Ltd All rights reserved 713

Chemistry & Biology HDAC1/2 Inhibitors Differentiate Neuroblastoma

Ramı´rez and Hagman, 2009; Scaffidi and Misteli, 2010), suggesting that chromatin-modifying small molecules might effectively induce cellular differentiation. For our study, we chose to look at neuroblastoma, a disease where differentiation therapy has been successful but not yet fully explored. Neuroblastoma is the most common extracranial pediatric solid tumor (Modak and Cheung, 2010). Although cure rates for patients with low-risk disease are greater than 90%, the prognosis for patients with high-risk neuroblastoma remains dismal, with cure rates as low as 35% despite the incorporation of aggressive chemotherapy, surgery, radiation, transplant, and biological therapy (Modak and Cheung, 2010). The differentiating agent 13-cis retinoic acid (cisRA) is presently used to treat minimal residual disease in the high-risk patient group after autologous stem cell transplantation (Matthay et al., 1999). However, the full therapeutic benefit of prodifferentiating agents has not been thoroughly explored. Here, we report the development of a robust gene expression signature for neuroblastoma differentiation. We screened a DOS library for induction of the neuroblastoma differentiation signature. We identified a prodifferentiating compound, BRD8430, and subsequently characterized it as a selective inhibitor of class I histone deacetylases (HDACs; HDAC1 > 2 > 3). Further investigation honed in on selective HDAC1/HDAC2 inhibition as important in inducing neuroblastoma differentiation and cell death. RESULTS A Gene Expression Signature for Neuroblastoma Differentiation Identifying small-molecule inducers of neuroblastoma differentiation via high-throughput screening is a challenging task because of the complexity of the target phenotype. Therefore, we developed a gene expression signature representing differentiation and adapted it for GE-HTS, a previously described method that uses ligation-mediated amplification and fluorescent bead-based detection to measure gene expression levels (Peck et al., 2006; Stegmaier et al., 2004). Two neuroblastoma cell lines, BE(2)-C, which harbors MYCN amplification, and SHSY5Y, MYCN nonamplified, were treated with the prodifferentiating agent(s) cisRA and/or phorbol 12-myristate 13-acetate (PMA) or the appropriate vehicle control. Whole-genome expression analysis was then performed using Affymetrix microarrays. A 59-gene signature for neuroblastoma differentiation was derived; it included 40 upregulated genes in the chemically differentiated cells relative to the undifferentiated neuroblastoma cells, 11 downregulated genes in the differentiated cells, and 8 reference genes with stable expression across the two biological states (see Supplemental Experimental Procedures available online for a full description of the signature creation and Table S1 for a list of the signature genes and probes). Two known differentiation agents, ATRA and cisRA, were confirmed to induce the differentiation signature in a dose-dependent manner in BE(2)-C cells after 2 days of treatment (Figure 1A). The combined expression of the signature genes can be represented by a single value (i.e., the weighted summed score), which will heretofore be referred to as the ‘‘differentiation score.’’ Absolute scores are not directly comparable across experiments, but compound perfor-

mance can be evaluated within an experiment relative to positive and negative controls. The differentiation signature was further validated by treating the SH-SY5Y cell line with both cisRA and PMA and by treating three additional MYCN amplified neuroblastoma cell lines not used in the development of the GE-HTS signature (Kelly, LAN-1, and NGP) with cisRA (Figures S1A– S1D). Finally, we confirmed in BE(2)-C cells that at doses which induced the differentiation signature, differentiation was detected by another experimental approach: immunofluorescent labeling for the differentiation marker NF-M (neurofilament medium; Figure S1E) in extended neurite projections. Screening of a DOS Library Identifies an Inducer of Neuroblastoma Differentiation The compound library used to screen for the induction of the differentiation signature was created through DOS (Marcaurelle et al., 2010) and biased for chromatin modification via incorporation of a zinc-chelating ortho-amino anilide group. BE(2)-C cells were treated in duplicate with 10 mM of 1,916 members of the DOS library. DMSO treatment was used as the negative control, and 1 mM of cisRA as the positive control. After incubating for 2 days, the GE-HTS assay was performed. The ability of each compound to induce differentiation was evaluated by five complementary scoring methods (Figure S1F). Their performance in two of the five methods is shown in Figure 1B. The 32 topscoring compounds were selected to be rescreened across a range of concentrations. To confirm the activity observed in the primary screen, eight concentrations of each compound were evaluated in duplicate in BE(2)-C cells after 2 days of incubation. Again, all 32 compounds induced the differentiation score at 10 mM but differed in their overall concentration-response profiles, with the best-performing compound, BRD8430, significantly inducing the signature at all eight concentrations tested and the worst-performing compound, BRD3259, significantly inducing the signature with five of the eight concentrations (Figures 1C and 1D). To incorporate the performance across the full concentration range, a curve was fitted to the differentiation score across the eight concentrations and the area under the curve (AUC) was calculated (Figures S1G and S1H). BRD8430 had the highest AUC value of the 32 compounds evaluated. Key Stereochemical and Structural Features of BRD8430 Elucidated through Analog Testing One advantage of working with a small molecule derived from a DOS pathway is the ease in accessing and evaluating stereochemical and structural variants. Therefore, we were able to investigate features of BRD8430 that were important for its prodifferentiating activity by evaluating stereochemical and structural analogs using the GE-HTS assay. In addition to the ortho-amino anilide functionality, BRD8430 contains a nine-membered lactam, a para-ether dimethylaniline, and three stereocenters, two within the macrocycle and one outside of the ring (Figure 2A). BRD8430 and its stereoisomers were synthesized as outlined in the Supplemental Experimental Procedures and as illustrated by the scheme in Figure S2A. The full complement of eight stereoisomers was tested across a range of compound concentrations. The differentiation score across the concentration range was fitted with a curve and the

714 Chemistry & Biology 20, 713–725, May 23, 2013 ª2013 Elsevier Ltd All rights reserved

Chemistry & Biology HDAC1/2 Inhibitors Differentiate Neuroblastoma

nM cisRA

B

20

DMSO

15

1 μM cisRA Compounds (not retested)

10

“Hits” (compounds retested)

5 0 -5 -10 -15 -20 -25 -30

-25

-20

-15

-10

-5

0

5

Summed score

D

* *

60 50

50 40 30 20

*

10

* *

* *

**

* *

**

**

**

0

40

**

30 20 10

*

0

*

* *

**

-10

-10 -20

* *

-20 0.16 0.31 0.63 1.25 2.5

5

μM BRD8430

10

20

cisRA

60

DMSO

C

Differentiation score

ABCB1 ADD3 ARHGEF3 CDKN1A CRABP2 CRH CTSB CYP26A1 DDAH2 DDEF2 DOCK4 DPYSL3 DUSP6 EGR1 FLJ10357 FLJ12584 FLNB HEBP HOXD4 IER3 ISGF3G JARID2 KCTD13 MAFF MMP111 OIP106 PELO PLAT PLEKHA6 PLK2 PRKCH RAB20 RBP1 RGS16 ST6GAL1 SYNJ2 SYT11 TGM2 TPST1 VGF ASCL1 KCTD12 IGFBP5 CHGA SLC29A1 PPIF GATA3 Summed Score

Weighted summed score

0

cisRA

0 1000

DMSO

nM ATRA 1000

Differentiation score

A

0.16 0.31 0.63 1.25 2.5

5

10

20

μM BRD3259

Figure 1. A DOS Library Was Screened for the Induction of a 59 Gene Neuroblastoma Differentiation Signature (A) BE(2)-C cells were treated with either ATRA or cisRA across a range of doses. After 48 hr, the cells were lysed and the expression levels of the 59 signature genes (see Table S1 for gene list) were measured using GE-HTS. The heat map depicts the relative gene expression with blue denoting low expression and red denoting high expression (fold change range: 3 to 3). Four poorly performing genes were removed from the analysis and the eight reference genes are not shown. (B) In BE(2)-C cells, 1,916 DOS compounds were screened in duplicate. DMSO served as the negative control and 1 mM cisRA as the positive control. The summed score and weighted summed score for each compound or control are plotted. Compounds in red were retested in the primary assay across a range of compound concentrations. (C and D) Dose-responsive induction of the differentiation score (weighted summed score) for the strongest and weakest performing compounds in the confirmatory screen (i.e., BRD8430 and BRD3259, respectively). Error bars represent the mean ± SD for two replicates per compound condition, 143 1 mM cisRA replicates, and 336 DMSO replicates. *p < 0.01, **p < 0.001 calculated using a one-way ANOVA, with Bonferroni correction, comparing each condition to DMSO. See also Figure S1.

AUC was calculated and used as a metric of the overall compound performance. BRD4586, which differs from BRD8430 only at the exo-cyclic stereocenter, retained the full differentiation-inducing activity (Figures 2B–2D; Table S2). The six other stereoisomers displayed markedly reduced activity (Figure 2B; Table S2). We also assessed the differentiation-inducing activity of two structural analogs of BRD8430: BRD6819, which contains an eight-membered lactam; and BRD6332, which also contains an eight-membered lactam but with an ortho-ether dimethylaniline. BRD6819 induced the differentiation signature in a manner similar to BRD8430, while BRD6332 exhibited greatly dampened differentiation activity (Figures 2E and 2F). The three most active compounds (i.e., BRD8430, BRD4586, and BRD6819) were confirmed to induce the differentiation signature in a second cell line, SH-SY5Y (Figures S2B–S2D). BRD8430 Induces Differentiation and Decreases Viability across Neuroblastoma Cell Lines Upon the identification of BRD8430 as a top-scoring compound, it was repurified to R95% purity (by LC/MS) and retested in four neuroblastoma cell lines: Kelly, LAN-1, BE(2)-C, and SH-SY5Y. BRD8430 was confirmed to induce the differentiation signature in a concentration-dependent manner in all four cell lines at

both the 2 day and 5 day time points (Figures 3A and S3). Furthermore, BRD8430 induced morphologic changes, including neurite outgrowth, consistent with differentiation (Figure 3B). BRD8430 also reduced the viability of these four lines after 5 days of treatment, an expected finding when neuroblastoma cells undergo differentiation (Figure 3C; Table S3). BRD8430 Is a Selective HDAC Inhibitor BRD8430 includes an ortho-amino anilide metal chelation moiety making it a putative inhibitor of HDACs, which require zinc atoms for their enzymatic activity. Presumably, the orthoamino anilide serves as the zinc-chelating moiety and the nine-membered central ring extends to the solvent-exposed surface-binding region of the HDAC enzyme (Bressi et al., 2010). Indeed, in BE(2)-C cells, BRD8430 increased histone H3 acetylation in a concentration-dependent manner after 6 hours of treatment (Figure 4A). To determine if this HDAC inhibitory activity is required for the induction of neuroblastoma differentiation, an analog lacking the chelation functionality, BRD8703, was synthesized and tested in BE(2)-C cells. The analog did not induce the differentiation signature at concentrations of up to 20 mM (Figure 4B). Three analogs in which the chelation moiety was substituted also failed to induce the signature (Figures S4A–S4C).

Chemistry & Biology 20, 713–725, May 23, 2013 ª2013 Elsevier Ltd All rights reserved 715

Chemistry & Biology HDAC1/2 Inhibitors Differentiate Neuroblastoma

A

B

Figure 2. SSARs for BRD8430 Analogs

Stereochemistry (C1,C2C3)

O N H N

Me Me 2

O 3 Me

H 2N

R,SS

S,RR

R,RR

S,SS

R,SR

S,RS

R,RS

S,SR

AUC Key

N 1

N Me

O

-11.0

OH

Me

-23.3 -30.0

(C1,C2C3) = (R,SR)

D

BRD8430 (9-para; R,SR) N H Me Me

O

30 Me

10

** * ** *

N

N Me

O

Me

-10

50

H 2N

N

OH

* * *

-30

Me Me

O Me

10

N

N Me

O

Me

-10

OH

* * *

H 2N

** * * ** **

-30

**

N H

50

H2 N

20

5

10

2.5

1.25

0.63

0.31

0.16

0.08

cisRA

70

Me Me

O

μM BRD4586

BRD6332 (8-ortho; R,SR)

O N

DMSO

20

5

2.5

1.25

0.63

0.16

0.31

cisRA

DMSO

10

F

BRD6819 (8-para; R,SR)

Differentiation score

N H N

30

μM BRD8430

70 50

O

**

-50

-50

E

Differentiation score

50

**

BRD4586 (9-para; S,SR) 70

O

0.08

Differentiation score

70

Differentiation score

C

(A) The structure of BRD8430 with its three stereocenters denoted 1, 2, and 3. This compound has a nine-membered lactam, para-ether dimethylaniline, and R,SR stereochemistry (9-para; R,SR). (B) BE(2)-C cells were treated in duplicate with eight concentrations of BRD8430 and its seven stereoisomers for 2 days. Cells were then screened for induction of the differentiation signature. The heatmap represents AUC values calculated for each compound, with red denoting high AUC, yellow denoting intermediate AUC, and green denoting low AUC. (C and D) The signature induction for stereoisomers BRD8430 (9-para; R,SR) and BRD4586 (9-para; S,SR). (E) Differentiation score for BRD6819, a structural analog of BRD8430 with an eight-membered lactam (8-para; R,SR). (F) Differentiation score for BRD6332, a structural analog of BRD8430 with an eight-membered lactam and ortho-ether dimethylaniline (8-ortho; R,SR). In (C–F), error bars represent the mean ± SD of two replicates for each compound dose, 261 replicates for DMSO, and 34 replicates for cisRA at 1 mM. *p < 0.01, **p < 0.001 calculated using a one-way ANOVA, with Bonferroni correction, comparing each condition to DMSO. See also Figure S2 and Table S2.

O

**

Me

N

Me

N H

N

Me Me

O

H2 N

models (Coffey et al., 2000, 2001; Hahn et al., 2008; Mu¨hlethaler-Mottet et al., 10 OH Me ** OH 2008; Panicker et al., 2010). The majority ** -10 -10 * of these studies have focused on the -30 -30 so-called HDAC pan-inhibitors, which are small molecules that nonselectively bind -50 -50 to most of the 11 zinc-dependent HDACs. μM BRD6819 μM BRD6332 Given the intriguing HDAC-selectivity profile observed with BRD8430, we evalThere are 11 zinc-dependent HDACs. To determine which are uated the ability of a diverse collection of HDAC inhibitors inhibited by BRD8430, the compound was evaluated in a with different selectivity profiles to decrease viability and biochemical HDAC activity assay (Bradner et al., 2010b). induce differentiation in neuroblastoma cell lines. This collecBRD8430 was determined to inhibit HDAC1 with a half-maximal tion included pan-HDAC inhibitors (having activity against inhibitory concentration (IC50) of 0.069 mM (95% confidence in- more than HDACs 1, 2, and 3): ITF-2357 (Leoni et al., 2005), terval [CI]: 0.062–0.077), HDAC2 with an IC50 value of 0.56 mM vorinostat (Kelly and Marks, 2005), apicidin (Darkin-Rattray (95% CI: 0.47–0.66), and HDAC 3 with an IC50 value of 1.3 mM et al., 1996), and BRD7914 (Jones et al., 2006); selective (95% CI: 1.1–1.6). Across the same concentration range, we HDAC1/2/3 inhibitors: MS-275 (Suzuki et al., 1999) and were not able to determine IC50 values for inhibition of the other CI-994 (el-Beltagi et al., 1993); selective HDAC1/2 inhibitors: HDACs (Figures 4C and S4D). Ortho-amino anilides are known BRD5298 (Wilson et al., 2011), compound 60 (Methot et al., to be selective binders for HDACs 1, 2, and 3 (Moradei et al., 2008; Moradei et al., 2007), and BRD5100 (Wilson et al., 2007), but BRD8430 further demonstrated a narrow margin of 2011); a selective HDAC6 inhibitor: BRD8148 (Wilson selectivity for HDACs 1 and 2 over HDAC3, and for HDAC1 in et al., 2011); a selective HDAC8 inhibitor: PCI-34051 (Balasuparticular. This observation prompted us to further investigate bramanian et al., 2008); and a selective HDAC3 inhibitor: inhibthe effects of selective HDAC inhibitors on neuroblastoma cell itor 106 (Chou et al., 2008). lines. Two neuroblastoma cell lines, BE(2)-C and Kelly, were treated with each inhibitor at four concentrations, and cell Selective Inhibition of HDAC1 and HDAC2 Decreases viability was measured at days 2 and 5 (Figures 4D and S4E). Viability and Induces Differentiation of Neuroblastoma The selective HDAC6 and HDAC8 inhibitors did not signiCell Lines ficantly decrease viability, and the selective HDAC3 inhibitor For more than a decade, HDAC inhibitors have been known only decreased viability at the highest concentration (10 mM) to decrease viability and induce differentiation in neuroblastoma at both time points. All three selective HDAC1/2 inhibitors 30

N

20

10

5

2.5

1.25

0.63

0.31

0.08

0.16

cisRA

O

DMSO

20

10

5

2.5

Me

1.25

0.08

cisRA

DMSO

10

* ** * **

N

O

0.63

N Me

0.31

Me

0.16

30

716 Chemistry & Biology 20, 713–725, May 23, 2013 ª2013 Elsevier Ltd All rights reserved

Chemistry & Biology HDAC1/2 Inhibitors Differentiate Neuroblastoma

Kelly 160 ** 140 120

**

100

**

LAN-1 60 **

**

**

40

**

* *

0.16

0.31

30

60

20

40

0.31

0.63

1.25

BE(2)-C 250

140

**

120

200 150

0.63

1.25

**

**

0.63

1.25

SH-SY5Y **

100

* *

*

DMSO

DMSO

0 0.16

cisRA

10

20 0

80

* *

*

Figure 3. BRD8430 Induces Differentiation and Decreases Viability in Four Neuroblastoma Cell Lines (A) Concentration-dependent induction of the differentiation score after a 5-day incubation with BRD8430 in four cell lines. Error bars represent the mean ± SD of four replicates. *p < 0.01, **p < 0.001 calculated using a one-way ANOVA, with Bonferroni correction, comparing each condition to DMSO. See Figure S3 for day 2 time point. (B) BE(2)-C cells were treated with vehicle (DMSO) or BRD8430 at the indicated concentrations and labeled for NF-M (red) and DAPI (blue). (C) Cells were treated with BRD8430 across a 9-point concentration range for 5 days and then viability was measured using an ATP-based assay. The values relative to the DMSO-treated controls are displayed with error bars representing the mean ± SEM of four replicates. See Table S3 for IC50 values.

60

100

40 50

DMSO

cisRA

0 0.16

0.31

0.63

1.25

B

DMSO

20 0

cisRA

Differentiation score

** **

50

80

cisRA

Differentiation score

A

0.16

0.31

BRD8430

Vehicle

0.01 μM

C

1.4

0.1 μM

1 μM

5 μM

BE(2)-C LAN-1 SH-SY5Y

1.2

Kelly

Relative viability

1.0

0.8

0.6

0.4

0.2

0

0

0.02

0.04

0.08

0.16

0.31

0.63

1.25

2.50

μM BRD8430

had little effect on viability after 2 days of treatment but greatly decreased viability after 5 days of treatment. In contrast, the effects of the pan-HDAC inhibitors and the selective HDAC1/2/3 inhibitors were comparable at all time points analyzed. The ability of several of the compounds to induce the differentiation signature in BE(2)-C cells after 2 days of treatment was also assessed. Neither the selective HDAC6 inhibitor nor

the selective HDAC8 inhibitor significantly induced the signature at any of the concentrations tested, and the selective HDAC3 inhibitor only significantly induced the signature at the highest concentration (10 mM) (Figure 4E). Vorinostat, a pan-HDAC inhibitor, compound 60, a selective HDAC1/2 inhibitor, and MS-275 and CI-994, both selective HDAC1/2/3 inhibitors, all significantly induced differentiation at multiple concentrations (Figure 4F). Compound 60 is a commonly used tool compound that selectively inhibits both HDAC1 and HDAC2. We confirmed the selective activity profile of this compound (Figure 5A) and demonstrated that, similar to BRD8430, compound 60 induced the differentiation signature across multiple neuroblastoma cell lines at multiple time points (Figures 5B and S5). It also decreased the viability of these cell lines after 5 days of treatment (Figure 5C; Table S4).

Simultaneous Genetic Knockdown of HDAC1 and HDAC2 Inhibits 5.00 Neuroblastoma Growth and Induces Differentiation To confirm the sensitivity of neuroblastoma cell lines to HDAC1 and HDAC2 inhibition and to determine if the inhibition of HDAC1 or HDAC2 is preferentially responsible for the observed differentiation phenotype, we used small interfering RNA (siRNA). Two cell lines, Kelly and BE(2)-C, were transfected with a nontargeting siRNA pool, an siRNA pool targeting HDAC1, an siRNA pool targeting HDAC2, or the HDAC1 and HDAC2 targeting pools combined (Figure 6A). The differentiation signature was only induced upon

Chemistry & Biology 20, 713–725, May 23, 2013 ª2013 Elsevier Ltd All rights reserved 717

Chemistry & Biology HDAC1/2 Inhibitors Differentiate Neuroblastoma

O

B

C

N H

600

Vinculin

N

N Me

300

O

Me

OH

*

*

200 100

cisRA

0

D

Me Me

O Me

400

DMSO

AcH3

Differentiation score

2

1

μM BRD8430 0.5

DMSO

N

500

125

Percent enzyme activity

A

HDAC1 HDAC2 HDAC3

100

75

50

25

0 0.08 0.16 0.31 0.63 1.25 2.5

5

10

20

-10

-8

μM BRD8703

-6

-4

Log[Conc BRD8430 (M)]

1.4

pan-HDAC inhibitors: ITF-2357 (1-9) vorinostat (1, 2, 3, 6, 8) apicidin (1, 2, 3, 8) BRD7914 (1, 2, 3, 6)

Day 2

Relative viability

1.2 1 0.8 0.6

HDAC1/2/3-selective:

0.4

MS-275 (1, 2, 3) CI-994 (1, 2, 3)

0.2 0

HDAC1/2-selective: BRD5298 (1, 2) compound 60 (1, 2) BRD5100 (1, 2)

1.4

Day 5

Relative viability

1.2 1 0.8

HDAC6-selective: BRD8148 (6)

0.6

HDAC8-selective: PCI-34051 (8)

0.4

HDAC3-selective:

0.2

inhibitor 106 (3)

0 0

0.01

0.10

1.00

10.00

0

0.01

0.10

1.00

10.00

μM HDAC inhibitor

100

0

DMSO

50

200

150

0.10

1.00

10.00

μM HDAC inhibitor

* *

100

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

** * * **

** ** **

50

0 0.01

vorinostat (1, 2, 3, 6, 8) MS-275 (1, 2, 3) CI-994 (1, 2, 3) compound 60 (1, 2)

** *

DMSO

** *

150

250

cisRA

200

F

inhibitor 106 (3) BRD8148 (6) PCI-34051 (8)

** *

cisRA

Differentiation score

250

Differentiation score

E

0.01

0.10

1.00

10.00

μM HDAC inhibitor

Figure 4. BRD8430 Is a Selective HDAC Inhibitor (A) Immunoblot for acetylated histone H3 (AcH3) in nuclear extracts from BE(2)-C cells treated with BRD8430 or DMSO for 6 hours. (B) The differentiation score in BE(2)-C cells treated for 2 days with BRD8703, a BRD8430 analog lacking the metal chelation functionality. Error bars represent the mean ± SD of six replicates. See Figures S4A–S4C for additional compounds with substitutions for the chelation functionality. (C) Effects of BRD8430 on the enzymatic activity of HDACs 1, 2, and 3. See Figure S4D for effects on the activity of HDACs 4-9. Error bars represent the mean ± SD of three replicates. (D) Viability relative to DMSO-treated controls for BE(2)-C cells treated with one of the 12 HDAC inhibitors for 2 (top) or 5 (bottom) days. Error bars represent the mean ± SD of four replicates. The HDACs selectively inhibited by each compound are indicated in parentheses beside each compound name in the key. A panHDAC inhibitor is defined as an inhibitor of HDACs 1, 2, 3, and other HDACs. See Figure S4E for effects of the HDAC inhibitor panel on the viability of Kelly cells. (E and F) The differentiation score of BE(2)-C cells treated for 2 days with the indicated compounds. Error bars represent the mean ± SD of four replicates for each HDAC inhibitor dose, 72 DMSO replicates, and 24 replicates of 1 mM cisRA. (B, E, and F) *p < 0.05, **p < 0.01, ***p < 0.001 calculated using a one-way ANOVA, with Bonferroni correction, comparing each condition to DMSO.

concurrent knockdown of both HDAC1 and HDAC2, with the majority of the marker genes’ expression levels changing in the expected direction in both cell lines (Figures

6B and 6C). Simultaneous knockdown of HDAC1 and HDAC2 also significantly decreased cell growth in both lines (Figure 6D).

718 Chemistry & Biology 20, 713–725, May 23, 2013 ª2013 Elsevier Ltd All rights reserved

Chemistry & Biology HDAC1/2 Inhibitors Differentiate Neuroblastoma

Kelly

1.4 BE(2)-C LAN-1 SH-SY5Y Kelly

Relative viability

1.2 1.0 0.8 0.6 0.4

0.63

1.25

μM compound 60

0.02 0.04 0.08 0.16 0.31 0.63 1.25 2.50 5.00

DMSO

0.31

*

0.16

0.31

0.63

1.25

μM compound 60

SH-SY5Y 140

*

120

*

*

*

100

*

80

*

*

*

0.31

0.63

60

100

40

50

20 0

0

0 0.16

150

0

*

10

200

0.2

*

20

BE(2)-C 250

Differentiation score

C

* *

30

20

Log[Conc BRD8430 (M)]

50 40

40

-4

60

0 0.04

0.08

0.16

0.31

μM compound 60

DMSO

-6

*

60

0 -8

*

80

0 -10

*

100

*

cisRA

25

120

DMSO

50

140

DMSO

75

LAN-1 70

*

cisRA

160

cisRA

100

B Differentiation score

Percent enzyme activity

HDAC1 HDAC2 HDAC3

cisRA

125

A

0.08

0.16

μM compound 60

μM compound 60

Figure 5. Compound 60, a Selective HDAC1/2 Inhibitor, Induces Differentiation and Decreases Viability in Four Neuroblastoma Cell Lines (A) Effects of compound 60 on the enzymatic activity of HDACs 1, 2, and 3. Error bars represent the mean ± SD of three replicates. (B) Concentration-dependent induction of the differentiation score after a 5-day incubation with compound 60 in four cell lines. Error bars represent the mean ± SD of four replicates. *p < 0.001 calculated using a one-way ANOVA, with Bonferroni correction, comparing each condition to DMSO. See Figure S5 for day 2 time point. (C) Cells were treated with compound 60 across a range of concentrations for 5 days and then cell viability was measured using an ATP-based assay. The values relative to the DMSO-treated controls are displayed with error bars representing the mean ± SEM of four replicates. See Table S4 for IC50 values.

Combined Treatment with Selective HDAC Inhibitors and cisRA Enhances Differentiation, Reduces Viability, and Activates Retinoic Acid Signaling Combining the pan-HDAC inhibitor vorinostat with cisRA was previously shown to strongly induce differentiation and synergistically affect viability (Hahn et al., 2008). To determine if the combination of more selective HDAC inhibitors and cisRA is likewise effective, Kelly and BE(2)-C cells were treated with compound 60 plus cisRA and BRD8430 plus cisRA. These combinations caused significantly greater induction of the differentiation signature than any of the single agents (Figures 7A and S6A; Table S5), and it had a synergistic effect on viability (Figures 7B and S6B). The retinoic acid (RA) pathway is a reported target of HDAC inhibition, and the combination of a pan-HDAC inhibitor plus ATRA synergistically activates RA signaling (Epping et al., 2007). To determine if this pathway is targeted in neuroblastoma cells, the activation of a RA receptor-dependent reporter, RARE3-tkluc, was assessed. The combinations of compound 60 plus cisRA and BRD8430 plus cisRA displayed greater activity on RARE3-tk-luc than the effects of any agent alone (Figures 7C and S6C). DISCUSSION Neuroblastoma is the most common extracranial solid tumor of childhood. It is a highly heterogeneous disease; some cases resolve without intervention, while others remain refractory

despite aggressive treatment with chemotherapy, surgery, radiation, autologous stem cell transplantation, and consolidation with cisRA and immunotherapy (Modak and Cheung, 2010). Clinically, we have reached toxicity limits in using standard cytotoxic chemotherapy, thus alternative modalities for treating patients with high-stage neuroblastoma are greatly needed. Differentiation therapy is a strategy that has shown promise but has not been fully explored in neuroblastoma. It has been demonstrated that cisRA, a prodifferentiating agent, is beneficial when administered to patients with minimal residual disease following consolidation chemotherapy and/or autologous stem cell transplantation (Matthay et al., 1999). However, it has not proven beneficial in patients with measurable disease (Kohler et al., 2000). In another disease, APL, ATRA differentiation therapy has been a highly effective front-line treatment (Abdel-Wahab and Levine, 2010; Ade`s et al., 2010). Consequently, the identification of additional differentiating agents to be used in treating patients with neuroblastoma, either alone or in combination with cisRA, is warranted. However, efforts to identify these differentiation agents have been hampered due to minimal industry interest in leading such efforts, lack of high-throughput phenotypic assays to measure neuroblast maturation, and limited chemical diversity in current combinatorial libraries. We sought to overcome these challenges by integrating the screening of a structurally diverse, small-molecule library with a gene expression-based differentiation assay. We identified a 59-gene signature for the

Chemistry & Biology 20, 713–725, May 23, 2013 ª2013 Elsevier Ltd All rights reserved 719

Chemistry & Biology HDAC1/2 Inhibitors Differentiate Neuroblastoma

Kelly

A siHDAC1: siHDAC2:

-

+ -

+

+ +

siHDAC1: siHDAC2:

+

+ +

HDAC2

HDAC2

Vinculin (1)

Vinculin (1)

Vinculin (2)

Vinculin (2)

BE(2)-C

160

***

140

140

*** **

120

120

100

100

80

80 60

60

40

40

20

20

0

NT siHDAC1 siHDAC2 siHDAC1 + siHDAC2

siHDAC1

NT siHDAC1 siHDAC2 siHDAC1 + siHDAC2

NT

ABCB1 ADD3 ARHGEF3 CRABP2 CYP26A1 DDEF2 DOCK4 DUSP6 EGR1 FLNB HOXD4 IER3 ISGF3G JARID2 KCTD13 MAFF MMP111 OIP106 PLAT PLEKHA6 PLK2 PPKCH RAB20 RBP1 TGM2 TPST1 VGF FLJ12584 ASCL1 CHGA DLK1 SLC29A1 PPIF GATA3 Diff. score

siHDAC2

siHDAC1 + siHDAC2

NT

siHDAC1

siHDAC2

siHDAC1 + siHDAC2

D 6

Viability relative to day 0

Differentiation score

+ -

HDAC1

0

C

-

HDAC1

Kelly

B

Figure 6. Genetic Knockdown of HDAC1 and/ or HDAC2 in Neuroblastoma Cell Lines

BE(2)-C

5

Kelly NT siHDAC1 siHDAC2 siHDAC1 + siHDAC2

(A) Immunoblots illustrating knockdown of HDAC1 and HDAC2 in Kelly and BE(2)-C cells 2 days after transfection with siRNAs targeting these genes. Control cells were transfected with pooled nontargeting siRNAs. (B) Differentiation score at day 4 (BE(2)-C) or day 6 (Kelly) after transfection with nontargeting siRNAs (NT), siRNAs targeting HDAC1, siRNAs targeting HDAC2, or the combined HDAC1- and HDAC2targeting siRNA pools. Error bars represent the mean ± SEM of six replicates. ***p < 0.001 calculated using a one-way ANOVA, with Bonferroni correction, comparing each siRNA condition to NT. (C) Heatmaps displaying the relative expression of the 34 differentiation signature genes changing in the anticipated direction in both Kelly cells at day 6 posttransfection and BE(2)-C cells at day 4 posttransfection. Blue indicates low expression and red indicates high expression (fold change range: 3 to 3). (D) Viability of Kelly and BE(2)-C cells relative to the viability at the time of seeding (day 2 posttransfection). Error bars represent the mean ± SD of six replicates. *p < 0.05, **p < 0.01, ***p < 0.001 calculated using a two-way repeated-measures ANOVA, with Bonferroni correction, comparing each condition to NT at each time point.

4

at the two endo-cyclic stereocenters is critical for BRD8430’s activity; however, 2 the stereochemistry at the exo-cyclic stereocenter is not. Furthermore, the size 1 of the central ring can be reduced by one 0 atom without affecting the activity, but 0 2 4 changing the substitution pattern of Day the dimethylaniline markedly decreased BE(2)-C activity. Gene expression data were thus 9 successfully used to detail how structural 8 changes to a compound influence its ability * 7 to induce a complex phenotype. 6 *** These studies revealed that the HDAC 5 inhibitory activity of BRD8430 is essential 4 to induce the differentiation phenotype. *** 3 The acetylation status of amino acids in 2 histone tails dictates how tightly DNA is ** *** 1 wound and thus how accessible the DNA 0 is to transcription factors and other gene 0 2 4 Kelly BE(2)-C regulatory elements (Gregory et al., 2001). Day High histone acetylation is associated with active transcription, while low histone differentiation of neuroblastoma cells, adapted it to the GE-HTS acetylation is associated with gene silencing. Acetyl groups are platform, and screened a DOS library of small molecules. added by histone acetyl transferases (HATs) and removed by Through our screening efforts, we identified a prodifferentiating HDACs, which includes 18 enzymes divided into four classes (I–IV). It is hypothesized that defects in neuroblastoma differcompound, BRD8430. The structure of BRD8430 includes a nine-membered lactam, entiation are related to aberrant transcriptional regulation, para-ether dimethylaniline, defined stereochemistry at three and HDAC inhibitors have previously been shown to induce positions, and an ortho-amino aniline functional group that differentiation in neuroblastoma, presumably by enhancing imparts HDAC inhibitory activity. The stereo/structure activity the active transcription of genes critical to the differentiation relationship (SSAR) analysis revealed that the stereochemistry program (Hahn et al., 2008). HDAC inhibitors also hold promise ***

Viability relative to day 0

3

720 Chemistry & Biology 20, 713–725, May 23, 2013 ª2013 Elsevier Ltd All rights reserved

Chemistry & Biology HDAC1/2 Inhibitors Differentiate Neuroblastoma

HDAC inhibitor: 0.3 μM compound 60 400

300

300

200

200

100

100

0

0

0.6

+ -

+

+ +

-

+ -

+ -

+ + -

+

+ +

Dose additivity line cisRA BRD8430

3.0

0.2

0.3

0.4

μM BRD8430

0.5

Dose additivity line cisRA compound 60

0

0

0.5

0.1

μM compound 60

+ + -

2.5

0.5 μM cisRA:

+ -

2.0

0.01 μM cisRA:

-

1.5

HDAC inhibitor:

B

Figure 7. Combined Treatment of Selective HDAC1/2 Inhibitors and cisRA in BE(2)-C Cells

HDAC inhibitor: 0.6 μM BRD8430

400

1.0

Differentiation score

A

0

10

20

30

40

0

50

μM cisRA

10

20

30

40

50

60

70

μM cisRA

Relative luciferase:renilla ratio

C 80

*

250

70 60 50

150

40 30

**

200

(A) BE(2)-C cells were treated with an HDAC inhibitor (compound 60 or BRD8430) and cisRA alone and in combination for 2 days and then the differentiation score was measured. The whiskers indicate the maximum and minimum values of four replicates per condition. Each condition was significantly different from DMSO and each combination was significantly different from the corresponding single-agent treatments; the relevant comparisons are included in Table S5. (B) Isobolograms of the effect of the combined treatment of compound 60 and cisRA or BRD8430 and cisRA on the viability of BE(2)-C cells at 4 or 5 days, respectively. Each circle on the plot represents one of the compounds at a fixed dose and the concentration of the second compound in combination to produce 50% viability compared to the vehicle control. Synergy appears as points below the line of additivity. (C) Activation of a retinoic acid signaling reporter, RARE3-tk-luc, with compound 60 ± cisRA or BRD8430 ± cisRA treatment in BE(2)-C cells. Displayed are the normalized values, luciferase:renilla, relative to the ratio in the DMSO-treated control cells. Error bars represent the mean ± SD of two replicates for the compound 60 experiment, and the mean ± SD of three replicates for the BRD8430 experiment. *p < 0.05, **p < 0.001 calculated using a one-way ANOVA, with Bonferroni correction, comparing each condition to DMSO. See Figure S6 for results of these experiments in a second cell line, Kelly.

**

100

increasing isoenzyme selectivity of compounds. Indeed, there is a general move0 ment toward more selective, targeted 0 0.5 μM 0.6 μM 0.5 μM cisRA 0.5 μM 0.5 μM 0.5 μM cisRA DMSO DMSO BRD8430 + compound 60 + therapies for patients with cancer, such cisRA cisRA 0.6 μM 0.5 μM BRD8430 compound 60 as PI3K isoform-specific inhibitors (Kong and Yamori, 2009) and the inhibition of BRAF V600E by PLX4720 (Tsai et al., as effective anticancer agents more broadly. Indeed, the 2008). Moreover, there has been a call for HDAC1/2 selective HDAC inhibitors vorinostat and romidepsin are approved molecules for the treatment of patients with sickle cell disease therapies for cutaneous T-cell lymphoma (Mann et al., 2007; (Bradner et al., 2010a). Combination therapy has generally been the mode for curative Piekarz et al., 2009; Whittaker et al., 2010), and several other HDAC inhibitors are currently in clinical trials (Rasheed et al., treatment of patients with cancer. CisRA is a logical candidate to nominate for combination testing with selective HDAC inhibition 2008). However, the present inhibitors used clinically are largely for patients with neuroblastoma. CisRA differentiation therapy nonspecific, having activity against multiple HDACs across one has already been incorporated into the standard of care for or more classes. Vorinostat, which has been previously reported patients with high-risk neuroblastoma, and intriguingly, in a to decrease viability and induce differentiation in neuroblastoma, phase I trial evaluating vorinostat ± cisRA to treat pediatric inhibits HDACs 1, 2, 3, and 6 at subnanomolar concentrations patients with solid tumors, the one complete response was and HDAC8 at a low micromolar concentration. BRD8430, observed in a patient with neuroblastoma who received the comalthough a less potent inhibitor, exhibits greater selectivity for bination therapy (Fouladi et al., 2010). Retinoids and HDAC inhibitors have nonindependent mechaclass I HDACs (HDAC1 > 2 > 3) over all other HDACs, and it prompted further investigation through which we uncovered nisms of action. In the absence of ligand, retinoid receptors bind HDAC1 and HDAC2 as the critical targets for the induction corepressors (NCoR and SMRT), which are found within a comof neuroblastoma differentiation and attenuation of cellular plex containing HDAC activity, and mediate transcriptional viability. Because the use of relatively nonselective HDAC inhib- repression. When ligand bound, retinoid receptors recruit itors causes patient toxicity, morbidity may be mitigated by coactivators with histone acetylase activity (HATs) leading to 20

50

10

Chemistry & Biology 20, 713–725, May 23, 2013 ª2013 Elsevier Ltd All rights reserved 721

Chemistry & Biology HDAC1/2 Inhibitors Differentiate Neuroblastoma

activated transcription (Minucci and Pelicci, 1999). Thus, both retinoic acid receptors (RARs) and HDACs are involved in the regulation of RA target genes. A relevant clinical example of RA receptors and HDACs acting in concert is the defining lesion of APL (i.e., PML-RARa fusion) that blocks cellular differentiation by recruiting the NCoR/HDAC complex and represses transcription (Minucci and Pelicci, 1999). More recently, Epping et al. (2007) identified RA signaling as a target of HDAC inhibitors using a functional genetic screen. They reported that HDAC inhibition activates RA signaling, which is repressed by the expression of either RARa or preferentially expressed antigen of melanoma (PRAME), a known repressor of RA signaling (Epping et al., 2007). Moreover, they observed a synergistic induction of RAR transactivation with the combination of ATRA and an HDAC inhibitor in a heterologous reporter assay. Similarly, we observed that in neuroblastoma cells, the combination of selective HDAC1/2 and HDAC1/2/3 inhibitors plus cisRA displayed greater activity on an RA receptor-dependent reporter than the effects of either agent alone. These studies reveal the power of integrating expressionbased screening with DOS chemistry to provide new biological insights. They illuminate the role of HDAC1/2 selective inhibition in activating RAR signaling to induce neuroblastoma differentiation. Our findings, in conjunction with other studies, support the need to develop HDAC1/2 inhibitors for clinical application. SIGNIFICANCE The intersection of genomics with chemical screening has provided new opportunities in the search for smallmolecule perturbagens of complex biologic state switches. We leveraged a gene expression-based screening approach to identify compounds that induce differentiation in a highly malignant pediatric solid tumor, neuroblastoma. The integration of cisRA differentiation therapy into consolidation treatment for these pediatric patients has shown promise, but the application of differentiation therapy has not been thoroughly explored. Focusing on a collection of DOS molecules, we demonstrate the success of this chemical genomics approach in both prioritizing lead compounds for inducing neuroblastoma differentiation and as a read-out in SSAR evaluation. This screen identified BRD8430 as a prodifferentiating agent. BRD8430 was characterized as a class I HDAC inhibitor (HDAC1 > 2 > 3). Through investigation of more selective compounds, HDACs 1 and 2 were demonstrated to be the key targets for induction of differentiation in neuroblastoma. Furthermore, the combination of a selective HDAC1/2 inhibitor and cisRA had an augmented effect on differentiation and a synergistic effect on viability, with enhanced activation of retinoic acid signaling as a contributing mechanism to the observed synergy. The HDAC inhibitors currently used clinically are largely pan-inhibitors, having activity against many of the 11 zinc-dependent HDACs. This work supports the need to develop selective HDAC1/2 inhibitors for clinical application and, more broadly, illuminates the power of integrating expression-based screening with DOS chemistry to provide new biological insights.

EXPERIMENTAL PROCEDURES Cell Culture and Viability Assays Neuroblastoma cell lines were maintained in DMEM (Cellgro, Manassas, VA, USA) supplemented with 10% fetal bovine serum (Sigma-Aldrich, St. Louis, MO, USA) and 1% penicillin-streptomycin with glutamine (Cellgro). BE(2)-C and SH-SY5Y were obtained from the American Type Culture Collection, and Kelly and LAN-1 were kindly provided by Dr. Rani George. The identity of all lines was verified by small tandem repeat profiling performed by the Shannon McCormack Advanced Molecular Diagnostics Laboratory at the Dana-Farber Cancer Institute (Boston, MA, USA). Cellular viability was assessed using CellTiter Glo (Promega, Madison, WI, USA) per the manufacturer’s instructions. Compounds ATRA, cisRA, PMA, and VPA were purchased from Sigma-Aldrich. Compound 60 was synthesized by Dr. Jun Qi. DOS compounds were synthesized by the Broad Institute Chemical Biology Platform; synthetic protocols are included in Supplemental Experimental Procedures. Compounds included in the panel of nonselective and selective HDAC inhibitors (ITF-2357, vorinostat, apicidin, BRD7914, MS-275, CI-994, BRD5298, compound 60, BRD5100, BRD8148, PCI-34051, and inhibitor 106) were provided by the Broad Institute Chemical Biology Platform and Stanley Center for Psychiatric Research. Gene Expression-Based High-Throughput Screening Gene expression-based high-throughput screening (GE-HTS) studies were performed as previously described (Peck et al., 2006; Stegmaier et al., 2004), including mRNA capture, reverse transcription, ligation-mediated amplification of the signature genes, and fluorescent bead-based detection (FlexMap Technology, Luminex, Austin, TX, USA). Probe sequences for the 59 neuroblastoma signature genes are included in Table S1. Signature performance was primarily evaluated using a weighted summed score (differentiation score) metric, calculated by dividing the expression levels of the signature genes by the mean expression of the control genes and then summing the ratios with weights and signs determined by the signal-to-noise ratio of cisRA-treated positive controls and DMSO-treated negative controls. As with other expression profiling platforms, scores can differ across experiments secondary to multiple features, including instrument variability, the dynamic nature of transcriptional alterations, and the bidirectional nature of the changes measured. Furthermore, genes that perform poorly in control conditions are omitted from the analysis on an experiment-by-experiment basis, leading to variation in the signature assessed. Finally, data from the low-throughput GE-HTS experiments did not undergo the same normalization as data from the larger screens, leading to variability in the absolute differentiation scores between experiments. DOS Library Screen BE(2)-C cells were plated at 2,000 cells per well in 384-well plates with 2 nM cisRA added as a differentiation enhancer; 1,916 compounds from the DOS library were added in duplicate by pin transfer to a final concentration of 10 mM. Each plate included 32 wells treated with 1 mM cisRA as the positive control and 32 wells treated with DMSO as the negative control. In addition, a control plate with a cisRA dose response was included in each run. After 2 days of incubation, the GE-HTS assay was performed. The screen data were analyzed as follows. The control gene N4BP1 was used to eliminate poorly performing wells from the analysis. For the plateby-plate analysis, we removed wells with an N4BP1 level greater than three median absolute deviations (MADs) from the median N4BP1 value of the DMSO-treated wells in that plate. For median-scaling analysis, we removed wells with an N4BP1 level greater than two MADs from the median N4BP1 value of the DMSO-treated wells across all the plates from each run. We then made a ratio of each marker gene to the mean of the eight reference genes within that well. For median-scaling analysis, plate-to-plate variability was corrected by median scaling. We normalized all of the plates by scaling each up- and downregulated gene independently based on its median expression in the cisRA-treated and DMSO-treated wells, respectively, on each plate. The summed scores (sum of the expression ratios with the sign determined by the expected direction of regulation) were individually transformed into robust

722 Chemistry & Biology 20, 713–725, May 23, 2013 ª2013 Elsevier Ltd All rights reserved

Chemistry & Biology HDAC1/2 Inhibitors Differentiate Neuroblastoma

Z scores by using the median and MAD of the chemical- and DMSO-treated wells together, including either all of the wells in a plate (plate-by-plate analysis) or all of the wells in a run (median-scaling analysis). To identify initial hits, we selected compounds for which both replicates had a robust Z score higher than three. Five additional metrics were used to identify prodifferentiating compounds: summed score, weighted summed score, K-nearest neighbor, naive Bayes, and support vector machine (Hahn et al., 2009). Thirty-two DOS compounds were selected to be retested at eight concentrations in duplicate in BE(2)-C cells with 2 nM cisRA as a differentiation enhancer, 1 mM cisRA as the positive control (32 wells per plate) and DMSO as the negative control (36 wells per plate). A control plate with a cisRA dose response was included in each run. The GE-HTS assay was performed after 2 days of incubation. The data were analyzed as described previously. Area under the Curve Calculation To quantize the relative efficacy of the compounds, total activity was summarized using the AUC, defined as the area between the response curve and the zero score level (defined by the average differentiation score in DMSO-treated negative control wells). AUC was calculated using the R system for statistical computing (R Development Core Team, 2006; http:// www.R-project.org) by taking the cubic natural spline for 50 points spread evenly across the log-dose curve, summing up the areas of the 49 trapezoids formed between the 50 spline points and the zero response line, and normalizing the value by dividing by the range of log concentrations. Because the AUC is a measurement summarizing the differentiation score across a concentration range, absolute AUC values suffer the same challenges as the differentiation score and cannot be compared directly between experiments, but can be used within an experiment to assess the relative activity of compounds. Protein Extraction and Immunoblotting For whole cell protein extraction, cells were lysed with Cell Signaling Lysis Buffer (Cell Signaling Technology, Danvers, MA, USA) containing complete, EDTA-free protease inhibitor cocktail (Roche Diagnostics, Mannheim, Germany) and PhosSTOP phosphatase inhibitor (Roche Diagnostics). Nuclear protein extractions were performed using the cytoplasmic and nuclear extraction kit from Thermo Scientific (Rockford, IL, USA). Immunoblots were run as previously described (Banerji et al., 2012). Primary antibodies included anti-HDAC1 (Abcam ab7028, Cambridge, MA, USA), anti-HDAC2 (Abcam ab7029), anti-Vinculin (Abcam ab18058), and anti-acetylated histone H3 (Millipore 06-599, Billerica, MA, USA). HDAC Activity Assay HDAC1-9 activity was determined in vitro using an optimized homogenous assay performed in a 384-well plate (Bradner et al., 2010b). Compound was incubated with recombinant, full-length HDAC protein (BPS Biosciences, San Diego, CA, USA) for 3 hr prior to enzymatic reactions. The reactions were carried out with fluorophore conjugated substrate, MAZ1600 and MAZ1675, at Km = [S] (MAZ1600: 21 mM for HDAC1, 22 mM for HDAC2, 9 mM for HDAC3, 9 mM for HDAC6; MAZ1675: 10 mM for HDAC4, 40 mM for HDAC5, 22 mM for HDAC7, 282 mM for HDAC8, and 26 mM for HDAC9). Reactions were performed in assay buffer (50 mM HEPES, 100 mM KCl, 0.001% Tween-20, 0.05% BSA, 200 mM TCEP, pH 7.4) and followed for fluorogenic release of 7-amino-4-methylcoumarin from substrate upon deacetylase and trypsin enzymatic activity. Fluorescence measurements were obtained every 5 minutes using a multilabel plate reader and plate stacker (Envision; PerkinElmer, Waltham, MA, USA). Each plate was analyzed by plate repeat, and the first derivative within the linear range was imported into analytical software (Spotfire DecisionSite and GraphPad Prism). Replicate experimental data from incubations with inhibitor were normalized to DMSO controls ([DMSO] < 0.5%). IC50 is determined by logistic regression with unconstrained maximum and minimum values. Immunofluorescence BE(2)-C cells were treated in duplicate with either vehicle (DMSO) or the indicated compound concentration. After 72 hr, cell were washed, fixed (4% formaldehyde), permeabilized (0.2% Triton X-100), and stained with a monoclonal antibody to NF-M (Santa Cruz Biotechnology sc20013) and

species-specific secondary antibody linked to Alexa Fluor 568 (Invitrogen, Carlsbad, CA, USA). Slides were mounted with Prolong Gold antifade reagents and counterstained with DAPI (Invitrogen). siRNA Studies BE(2)-C and Kelly cells seeded in six-well plates were transfected with siGENOME SMARTpool siRNAs from Dharmacon targeting HDAC1 (Thermo Scientific M-003493-02-0005), HDAC2 (Thermo Scientific M-003495-020005), or a negative control, nontargeting siRNA pool (Thermo Scientific D-001206-13-05) following the manufacturer’s protocol. Briefly, we used 4 ml of Dharmafect Transfection Reagent I (Thermo Scientific T-2001) per well and a final concentration of 45 nM siRNA. After 48 hr, cells were harvested, counted, and replated in 384-well plates for viability and GE-HTS assays. Protein was collected to evaluate knockdown. Dual Reporter Assay BE(2)-C or Kelly cells were cotransfected with RARE3-tk-luc (de The´ et al., 1990) (kindly provided by Dr. Daniel G. Tenen) and pUbC-RL plasmids using Fugene 6 transfection reagent per the manufacturer’s instructions (Roche). In the compound 60 experiments, the cells were harvested, counted, and replated in a 96-well plate 24 hr posttransfection, and then compounds were added. In the BRD8430 experiments, the cells were transfected in 96-well plates and thus were not replated before compound addition. Twenty-four hours later, levels of firefly luciferase and renilla luciferase were detected using the Dual Luciferase Reporter Assay System (Promega). Statistical Analyses Statistical significance was determined by one-way ANOVA with Bonferroni correction when comparing more than two unmatched groups, and by twoway repeated-measures ANOVA with Bonferroni correction when comparing more than two unmatched groups over a time course. Analyses were done in GraphPad Prism 5. ACCESSION NUMBERS The GEO accession number for the genome-wide expression profiling data reported in this paper is GSE45587. SUPPLEMENTAL INFORMATION Supplemental Information includes six figures, five tables, and Supplemental Experimental Procedures and can be found with this article online at http:// dx.doi.org/10.1016/j.chembiol.2013.03.020. ACKNOWLEDGMENTS We thank Rani George for providing neuroblastoma cell lines, Daniel G. Tenen for providing the reporter plasmid, and Brian Crompton and Aaron Thorner for their technical assistance. This work was funded by the Friends for Life Fellowship (to K.S.), the Claudia Adams Barr Program in Cancer Research (to K.S.), and a Howard Hughes Medical Institute Physician-Scientist Early Career Award (to K.S.). The project has been funded in part with federal funds from the National Cancer Institute’s Initiative for Chemical Genetics, National Institutes of Health (NIH), under contract no. N01-CO-12400. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Service, nor does mention of trade names, commercial products, or organizations imply endorsement by the United States government. This work was funded in part by the NIGMS-sponsored Center of Excellence in Chemical Methodology and Library Development (Broad Institute CMLD; P50 GM069721), as well as the NIH Genomics Based Drug Discovery U54 grants Discovery Pipeline RL1CA133834 (administratively linked to NIH Grants RL1HG004671, RL1GM084437, and UL1DE019585). Received: September 11, 2012 Revised: March 7, 2013 Accepted: March 22, 2013 Published: May 23, 2013

Chemistry & Biology 20, 713–725, May 23, 2013 ª2013 Elsevier Ltd All rights reserved 723

Chemistry & Biology HDAC1/2 Inhibitors Differentiate Neuroblastoma

REFERENCES Abdel-Wahab, O., and Levine, R.L. (2010). Recent advances in the treatment of acute myeloid leukemia. F1000 Med. Rep. 2, 55. Ade`s, L., Guerci, A., Raffoux, E., Sanz, M., Chevallier, P., Lapusan, S., Recher, C., Thomas, X., Rayon, C., Castaigne, S., et al.; European APL Group. (2010). Very long-term outcome of acute promyelocytic leukemia after treatment with all-trans retinoic acid and chemotherapy: the European APL Group experience. Blood 115, 1690–1696. Balasubramanian, S., Ramos, J., Luo, W., Sirisawad, M., Verner, E., and Buggy, J.J. (2008). A novel histone deacetylase 8 (HDAC8)-specific inhibitor PCI-34051 induces apoptosis in T-cell lymphomas. Leukemia 22, 1026–1034. Banerji, V., Frumm, S.M., Ross, K.N., Li, L.S., Schinzel, A.C., Hahn, C.K., Kakoza, R.M., Chow, K.T., Ross, L., Alexe, G., et al. (2012). The intersection of genetic and chemical genomic screens identifies GSK-3a as a target in human acute myeloid leukemia. J. Clin. Invest. 122, 935–947. Bradner, J.E., Mak, R., Tanguturi, S.K., Mazitschek, R., Haggarty, S.J., Ross, K., Chang, C.Y., Bosco, J., West, N., Morse, E., et al. (2010a). Chemical genetic strategy identifies histone deacetylase 1 (HDAC1) and HDAC2 as therapeutic targets in sickle cell disease. Proc. Natl. Acad. Sci. USA 107, 12617–12622. Bradner, J.E., West, N., Grachan, M.L., Greenberg, E.F., Haggarty, S.J., Warnow, T., and Mazitschek, R. (2010b). Chemical phylogenetics of histone deacetylases. Nat. Chem. Biol. 6, 238–243. Bressi, J.C., Jennings, A.J., Skene, R., Wu, Y., Melkus, R., De Jong, R., O’Connell, S., Grimshaw, C.E., Navre, M., and Gangloff, A.R. (2010). Exploration of the HDAC2 foot pocket: Synthesis and SAR of substituted N(2-aminophenyl)benzamides. Bioorg. Med. Chem. Lett. 20, 3142–3145. Chou, C.J., Herman, D., and Gottesfeld, J.M. (2008). Pimelic diphenylamide 106 is a slow, tight-binding inhibitor of class I histone deacetylases. J. Biol. Chem. 283, 35402–35409. Coffey, D.C., Kutko, M.C., Glick, R.D., Swendeman, S.L., Butler, L., Rifkind, R., Marks, P.A., Richon, V.M., and LaQuaglia, M.P. (2000). Histone deacetylase inhibitors and retinoic acids inhibit growth of human neuroblastoma in vitro. Med. Pediatr. Oncol. 35, 577–581. Coffey, D.C., Kutko, M.C., Glick, R.D., Butler, L.M., Heller, G., Rifkind, R.A., Marks, P.A., Richon, V.M., and La Quaglia, M.P. (2001). The histone deacetylase inhibitor, CBHA, inhibits growth of human neuroblastoma xenografts in vivo, alone and synergistically with all-trans retinoic acid. Cancer Res. 61, 3591–3594. Darkin-Rattray, S.J., Gurnett, A.M., Myers, R.W., Dulski, P.M., Crumley, T.M., Allocco, J.J., Cannova, C., Meinke, P.T., Colletti, S.L., Bednarek, M.A., et al. (1996). Apicidin: a novel antiprotozoal agent that inhibits parasite histone deacetylase. Proc. Natl. Acad. Sci. USA 93, 13143–13147. de The´, H., Vivanco-Ruiz, M.M., Tiollais, P., Stunnenberg, H., and Dejean, A. (1990). Identification of a retinoic acid responsive element in the retinoic acid receptor beta gene. Nature 343, 177–180. el-Beltagi, H.M., Martens, A.C., Lelieveld, P., Haroun, E.A., and Hagenbeek, A. (1993). Acetyldinaline: a new oral cytostatic drug with impressive differential activity against leukemic cells and normal stem cells—preclinical studies in a relevant rat model for human acute myelocytic leukemia. Cancer Res. 53, 3008–3014. Epping, M.T., Wang, L., Plumb, J.A., Lieb, M., Gronemeyer, H., Brown, R., and Bernards, R. (2007). A functional genetic screen identifies retinoic acid signaling as a target of histone deacetylase inhibitors. Proc. Natl. Acad. Sci. USA 104, 17777–17782.

based screening identifies the combination of histone deacetylase inhibitors and retinoids for neuroblastoma differentiation. Proc. Natl. Acad. Sci. USA 105, 9751–9756. Hahn, C.K., Berchuck, J.E., Ross, K.N., Kakoza, R.M., Clauser, K., Schinzel, A.C., Ross, L., Galinsky, I., Davis, T.N., Silver, S.J., et al. (2009). Proteomic and genetic approaches identify Syk as an AML target. Cancer Cell 16, 281–294. Helman, E., Naxerova, K., and Kohane, I.S. (2012). DNA hypermethylation in lung cancer is targeted at differentiation-associated genes. Oncogene 31, 1181–1188. Jones, P., Altamura, S., Chakravarty, P.K., Cecchetti, O., De Francesco, R., Gallinari, P., Ingenito, R., Meinke, P.T., Petrocchi, A., Rowley, M., et al. (2006). A series of novel, potent, and selective histone deacetylase inhibitors. Bioorg. Med. Chem. Lett. 16, 5948–5952. Kelly, W.K., and Marks, P.A. (2005). Drug insight: Histone deacetylase inhibitors—development of the new targeted anticancer agent suberoylanilide hydroxamic acid. Nat. Clin. Pract. Oncol. 2, 150–157. Koehn, F.E., and Carter, G.T. (2005). The evolving role of natural products in drug discovery. Nat. Rev. Drug Discov. 4, 206–220. Kohler, J.A., Imeson, J., Ellershaw, C., and Lie, S.O. (2000). A randomized trial of 13-Cis retinoic acid in children with advanced neuroblastoma after high-dose therapy. Br. J. Cancer 83, 1124–1127. Kong, D., and Yamori, T. (2009). Advances in development of phosphatidylinositol 3-kinase inhibitors. Curr. Med. Chem. 16, 2839–2854. Leoni, F., Fossati, G., Lewis, E.C., Lee, J.K., Porro, G., Pagani, P., Modena, D., Moras, M.L., Pozzi, P., Reznikov, L.L., et al. (2005). The histone deacetylase inhibitor ITF2357 reduces production of pro-inflammatory cytokines in vitro and systemic inflammation in vivo. Mol. Med. 11, 1–15. Lotem, J., and Sachs, L. (2006). Epigenetics and the plasticity of differentiation in normal and cancer stem cells. Oncogene 25, 7663–7672. Lovering, F., Bikker, J., and Humblet, C. (2009). Escape from flatland: increasing saturation as an approach to improving clinical success. J. Med. Chem. 52, 6752–6756. Mann, B.S., Johnson, J.R., Cohen, M.H., Justice, R., and Pazdur, R. (2007). FDA approval summary: vorinostat for treatment of advanced primary cutaneous T-cell lymphoma. Oncologist 12, 1247–1252. Marcaurelle, L.A., Comer, E., Dandapani, S., Duvall, J.R., Gerard, B., Kesavan, S., Lee, M.D., 4th, Liu, H., Lowe, J.T., Marie, J.C., et al. (2010). An aldol-based build/couple/pair strategy for the synthesis of medium- and large-sized rings: discovery of macrocyclic histone deacetylase inhibitors. J. Am. Chem. Soc. 132, 16962–16976. Matthay, K.K., Villablanca, J.G., Seeger, R.C., Stram, D.O., Harris, R.E., Ramsay, N.K., Swift, P., Shimada, H., Black, C.T., Brodeur, G.M., et al.; Children’s Cancer Group. (1999). Treatment of high-risk neuroblastoma with intensive chemotherapy, radiotherapy, autologous bone marrow transplantation, and 13-cis-retinoic acid. N. Engl. J. Med. 341, 1165–1173. Methot, J.L., Chakravarty, P.K., Chenard, M., Close, J., Cruz, J.C., Dahlberg, W.K., Fleming, J., Hamblett, C.L., Hamill, J.E., Harrington, P., et al. (2008). Exploration of the internal cavity of histone deacetylase (HDAC) with selective HDAC1/HDAC2 inhibitors (SHI-1:2). Bioorg. Med. Chem. Lett. 18, 973–978. Minucci, S., and Pelicci, P.G. (1999). Retinoid receptors in health and disease: co-regulators and the chromatin connection. Semin. Cell Dev. Biol. 10, 215–225. Modak, S., and Cheung, N.K. (2010). Neuroblastoma: Therapeutic strategies for a clinical enigma. Cancer Treat. Rev. 36, 307–317.

Fouladi, M., Park, J.R., Stewart, C.F., Gilbertson, R.J., Schaiquevich, P., Sun, J., Reid, J.M., Ames, M.M., Speights, R., Ingle, A.M., et al. (2010). Pediatric phase I trial and pharmacokinetic study of vorinostat: a Children’s Oncology Group phase I consortium report. J. Clin. Oncol. 28, 3623–3629.

Moradei, O.M., Mallais, T.C., Frechette, S., Paquin, I., Tessier, P.E., Leit, S.M., Fournel, M., Bonfils, C., Trachy-Bourget, M.C., Liu, J., et al. (2007). Novel aminophenyl benzamide-type histone deacetylase inhibitors with enhanced potency and selectivity. J. Med. Chem. 50, 5543–5546.

Gregory, P.D., Wagner, K., and Ho¨rz, W. (2001). Histone acetylation and chromatin remodeling. Exp. Cell Res. 265, 195–202.

Mu¨hlethaler-Mottet, A., Meier, R., Flahaut, M., Bourloud, K.B., Nardou, K., Joseph, J.M., and Gross, N. (2008). Complex molecular mechanisms cooperate to mediate histone deacetylase inhibitors anti-tumour activity in neuroblastoma cells. Mol. Cancer 7, 55.

Hahn, C.K., Ross, K.N., Warrington, I.M., Mazitschek, R., Kanegai, C.M., Wright, R.D., Kung, A.L., Golub, T.R., and Stegmaier, K. (2008). Expression-

724 Chemistry & Biology 20, 713–725, May 23, 2013 ª2013 Elsevier Ltd All rights reserved

Chemistry & Biology HDAC1/2 Inhibitors Differentiate Neuroblastoma

Panicker, J., Li, Z., McMahon, C., Sizer, C., Steadman, K., Piekarz, R., Bates, S.E., and Thiele, C.J. (2010). Romidepsin (FK228/depsipeptide) controls growth and induces apoptosis in neuroblastoma tumor cells. Cell Cycle 9, 1830–1838. Peck, D., Crawford, E.D., Ross, K.N., Stegmaier, K., Golub, T.R., and Lamb, J. (2006). A method for high-throughput gene expression signature analysis. Genome Biol. 7, R61. Piekarz, R.L., Frye, R., Turner, M., Wright, J.J., Allen, S.L., Kirschbaum, M.H., Zain, J., Prince, H.M., Leonard, J.P., Geskin, L.J., et al. (2009). Phase II multi-institutional trial of the histone deacetylase inhibitor romidepsin as monotherapy for patients with cutaneous T-cell lymphoma. J. Clin. Oncol. 27, 5410–5417. Ramı´rez, J., and Hagman, J. (2009). The Mi-2/NuRD complex: a critical epigenetic regulator of hematopoietic development, differentiation and cancer. Epigenetics 4, 532–536. Rasheed, W., Bishton, M., Johnstone, R.W., and Prince, H.M. (2008). Histone deacetylase inhibitors in lymphoma and solid malignancies. Expert Rev. Anticancer Ther. 8, 413–432. Scaffidi, P., and Misteli, T. (2010). Cancer epigenetics: from disruption of differentiation programs to the emergence of cancer stem cells. Cold Spring Harb. Symp. Quant. Biol. 75, 251–258.

Schreiber, S.L. (2000). Target-oriented and diversity-oriented organic synthesis in drug discovery. Science 287, 1964–1969. Stegmaier, K., Ross, K.N., Colavito, S.A., O’Malley, S., Stockwell, B.R., and Golub, T.R. (2004). Gene expression-based high-throughput screening (GEHTS) and application to leukemia differentiation. Nat. Genet. 36, 257–263. Suzuki, T., Ando, T., Tsuchiya, K., Fukazawa, N., Saito, A., Mariko, Y., Yamashita, T., and Nakanishi, O. (1999). Synthesis and histone deacetylase inhibitory activity of new benzamide derivatives. J. Med. Chem. 42, 3001– 3003. Tsai, J., Lee, J.T., Wang, W., Zhang, J., Cho, H., Mamo, S., Bremer, R., Gillette, S., Kong, J., Haass, N.K., et al. (2008). Discovery of a selective inhibitor of oncogenic B-Raf kinase with potent antimelanoma activity. Proc. Natl. Acad. Sci. USA 105, 3041–3046. Whittaker, S.J., Demierre, M.F., Kim, E.J., Rook, A.H., Lerner, A., Duvic, M., Scarisbrick, J., Reddy, S., Robak, T., Becker, J.C., et al. (2010). Final results from a multicenter, international, pivotal study of romidepsin in refractory cutaneous T-cell lymphoma. J. Clin. Oncol. 28, 4485–4491. Wilson, A.J., Holson, E., Wagner, F., Zhang, Y.L., Fass, D.M., Haggarty, S.J., Bhaskara, S., Hiebert, S.W., Schreiber, S.L., and Khabele, D. (2011). The DNA damage mark pH2AX differentiates the cytotoxic effects of small molecule HDAC inhibitors in ovarian cancer cells. Cancer Biol. Ther. 12, 484–493.

Chemistry & Biology 20, 713–725, May 23, 2013 ª2013 Elsevier Ltd All rights reserved 725