Expression-based screening identifies the combination of histone deacetylase inhibitors and retinoids for neuroblastoma differentiation
The discovery of new small molecules and their testing in rational combination poses an ongoing problem for rare diseases, in particular, for pediatric cancers such as neuroblastoma. Despite maximal cytotoxic therapy with double autologous stem cell transplantation, outcome remains poor for children with high-stage disease. Because differentiation is aberrant in this malignancy, compounds that modulate transcription, such as histone deacetylase (HDAC) inhibitors, are of particular interest. However, as single agents, HDAC inhibitors have had limited efficacy. In the present study, we use an HDAC inhibitor as an enhancer to screen a small-molecule library for compounds inducing neuroblastoma maturation. To quantify differentiation, we use an enabling gene expression-based screening strategy. The top hit identified in the screen was all-trans-retinoic acid. Secondary assays confirmed greater neuroblastoma differentiation with the combination of an HDAC inhibitor and a retinoid versus either alone. Furthermore, effects of combination therapy were synergistic with respect to inhibition of cellular viability and induction of apoptosis. In a xenograft model of neuroblastoma, animals treated with combination therapy had the longest survival. This work suggests that testing of an HDAC inhibitor and retinoid in combination is warranted for children with neuroblastoma and demonstrates the success of a signature-based screening approach to prioritize compound combinations for testing in rare diseases.
Expression-based screening identifies the combination
of histone deacetylase inhibitors and retinoids for
Cynthia K. Hahn*, Kenneth N. Ross
, Ian M. Warrington*, Ralph Mazitschek
, Cindy M. Kanegai*, Renee D. Wright*,
Andrew L. Kung*, Todd R. Golub*
, and Kimberly Stegmaier*
*Department of Pediatric Oncology, Dana-Farber Cancer Institute and Children’s Hospital Boston, Harvard Medical School, Boston, MA 02115;
Institute of Harvard University and Massachusetts Institute of Technology, Cambridge, MA 02142;
Howard Hughes Medical Institute, Chevy Chase,
MD 20815; and
Department of Biological Chemistry and Molecular Pharmacology, Harvard Medical School, Boston, MA 02115
Edited by Edward M. Scolnick, The Broad Institute, Cambridge, MA, and approved April 17, 2008 (received for review November 1, 2007)
The discovery of new small molecules and their testing in rational
combination poses an ongoing problem for rare diseases, in par-
ticular, for pediatric cancers such as neuroblastoma. Despite max-
imal cytotoxic therapy with double autologous stem cell trans-
plantation, outcome remains poor for children with high-stage
disease. Because differentiation is aberrant in this malignancy,
compounds that modulate transcription, such as histone deacety-
lase (HDAC) inhibitors, are of particular interest. However, as single
agents, HDAC inhibitors have had limited efﬁcacy. In the present
study, we use an HDAC inhibitor as an enhancer to screen a
small-molecule library for compounds inducing neuroblastoma
maturation. To quantify differentiation, we use an enabling gene
expression-based screening strategy. The top hit identiﬁed in the
screen was all-trans-retinoic acid. Secondary assays conﬁrmed
greater neuroblastoma differentiation with the combination of an
HDAC inhibitor and a retinoid versus either alone. Furthermore,
effects of combination therapy were synergistic with respect to
inhibition of cellular viability and induction of apoptosis. In a
xenograft model of neuroblastoma, animals treated with combi-
nation therapy had the longest survival. This work suggests that
testing of an HDAC inhibitor and retinoid in combination is war-
ranted for children with neuroblastoma and demonstrates the
success of a signature-based screening approach to prioritize
compound combinations for testing in rare diseases.
chemical genomics 兩 small molecule screening
ose intensification has improved outcome for some malignan-
cies, such as neuroblastoma, the most common pediatric solid
tumor. Despite this advance, progression-free survival for children
with advanced disease is 45% at best, even with double autologous
stem cell transplantation (1). For those children whose disease is
cured, morbidity from intensive chemotherapy is significant. New
approaches to the treatment of this malignancy are needed.
One targeted class of compound of particular interest is the
histone deacetylase (HDAC) inhibitor. Histone acetylation is a
critical regulatory mechanism of gene expression. Aberrant
deacetylation of critical genes involved in differentiation, apoptosis,
or cell cycle arrest has been reported to contribute to the patho-
genesis of malignancy. Indeed, the development of neuroblastoma
is believed to be related, in part, to defects in neural crest cell
differentiation, and hence aberrant transcriptional regulation.
Chromatin-modifying compounds, such as HDAC inhibitors, are
therefore attractive agents for testing in neuroblastoma. Small
molecules targeting the HDACs have been shown to have prodif-
ferentiating and apoptotic effects in numerous cancer subtypes in
model systems, including neuroblastoma (2, 3). Recently, there has
been a marked increase in the number of HDAC inhibitors in
clinical development. However, as single agents, these molecules
have had limited activity, with the exception of the treatment of
cutaneous T cell lymphoma (CTCL). In 2006, the Food and Drug
Administration (FDA) approved the HDAC inhibitor suberoyla-
nilide hydroxamic acid (SAHA) for adults with CTLC (4). It is likely
that the broad efficacy of HDAC inhibitors will only be realized in
combination with other drugs.
Combination therapy has been a hallmark of successful cancer
treatment. With few exceptions, curative systemic cancer therapy
has required the use of multiple drugs. The vast majority of
combination strategies have derived from testing an agent with
components of existing standards of care in the clinic, as opposed
to rationally identifying compound combinations in the laboratory
based on mechanism of action or performance in a small-molecule
library screen. In a time where an extraordinary number of small
molecules are in development, the rational selection of compounds
for clinical testing becomes even more important. This is particu-
larly problematic in the pediatric malignancies where the diseases
are rare, precluding testing of all possible new agents, let alone all
possible combinations. The development of small-molecule library
screening approaches to facilitate the rational selection of com-
pound combinations is critical.
In the current study, rather than focusing on cytotoxicity, we
have focused on neuroblast differentiation. Prior clinical data
suggest that differentiation therapy may play a complementary
role in the treatment of neuroblastoma when used in combina-
tion with other therapies (5). However, high-throughput assays
to quantitatively measure neuroblastoma differentiation as a
phenotypic endpoint have been limited. We have developed a
potential solution to this challenge in which a multigene signa-
ture is used as a surrogate for highly complex cellular states: gene
expression-based high-throughput screening (GE-HTS) (6).
First, gene expression signatures are defined for each biological
state of interest by using genomewide expression profiling. Then,
a high-throughput, low-cost assay is performed to evaluate up to
100 marker genes (7).
This work develops a general approach to the identification of
combinations of small molecules for therapeutic consideration.
Specifically, we now apply GE-HTS to the identification of small
molecules that enhance the effects of HDAC inhibitors on neuro-
Author contributions: C.K.H., A.L.K., T.R.G., and K.S. designed research; C.K.H., I.M.W.,
C.M.K., R.D.W., and K.S. performed research; R.M. contributed new reagents/analytic tools;
C.K.H., K.N.R., I.M.W., C.M.K., A.L.K., and K.S. analyzed data; and K.S. wrote the paper.
The authors declare no conﬂict of interest.
This article is a PNAS Direct Submission.
Data deposition: The raw microarray data have been deposited at http://www.broad.mit.
To whom correspondence should be addressed. E-mail: Kimberly㛭stegmaier@dfci.
This article contains supporting information online at www.pnas.org/cgi/content/full/
© 2008 by The National Academy of Sciences of the USA
July 15, 2008
Gene Expression-Based Small-Molecule Library Screen Identifies the
Combination of an HDAC Inhibitor and Retinoid for Neuroblastoma
Differentiation. The goal in this work was to screen for compounds
that act in concert with HDAC inhibitors to induce neuroblastoma
differentiation. Although HDAC inhibitors are weak inducers of
differentiation as single agents, we hypothesized that together with
other compounds, we might identify synergistic combinations. First,
we needed to identify an appropriate HDAC inhibitor to use as the
enhancer. We selected valproic acid (VPA) because it was the only
HDAC inhibitor with FDA-approval status at the time that we
began the screen. VPA has a long history of clinical use in treating
patients with seizures and bipolar disease, even before its charac-
terization as an HDAC inhibitor. However, since the initiation of
this screen, SAHA and the depsipeptide FR901228 have completed
pediatric phase I testing, opening up possibilities for future clinical
trials. We chose to work with the neuroblastoma cell line BE (2)-C.
Presumably, with N-myc amplification, these cells most closely
resemble high-stage neuroblastoma where N-myc is frequently
amplified and a poor prognostic marker. To facilitate clinical
translation, we focused our screening efforts on a collection of small
molecules highly enriched for FDA-approved drugs. We screened
in triplicate the National Institute of Neurological Disorders and
Stroke (NINDS) small-molecule collection containing 1,040 com-
pounds, three-quarters of which are FDA-approved.
To perform an expression-based screen for neuroblastoma dif-
ferentiation, we needed to identify an expression signature of the
mature neuroblast. Although there is no perfect approach to
inducing neuroblastoma differentiation, we chose to use two com-
pounds, all-trans-retinoic acid (ATRA) and phorbol 12-myristate
13-acetate (PMA), and two neuroblastoma cell lines, BE (2)-C and
SHSY5Y, to ensure that we did not overfit the signature to any one
compound or cell line. By using Affymetrix U133A DNA microar-
rays, we profiled BE (2)-C and SHSY5Y differentiated for 5 days
versus a vehicle-treated control in triplicate for each condition.
Differentiation was confirmed by morphology and/or expression of
the mature neuronal marker neurofilament medium (NF-M) (data
not shown). After standard preprocessing of the data, we identified
genes meeting the following criteria: (i) at least a 2-fold variation
between vehicle versus differentiation agent in one of the cell lines
with P ⬍ 0.1 by t test and (ii) the appropriate direction of change
in the other cell line. Nine top genes meeting these criteria were
selected for the neuroblastoma differentiation signature and one
additional gene, IGFBP7, with decreased expression in BE (2)-C
treatment but poor overall expression in SHSY5Y was included
(Fig. 1A). Importantly, these 10 genes were chosen as a surrogate
for the end state of differentiation, not as bona fide targets of the
differentiation process. In addition, four genes with stable expres-
sion across the conditions were included in the signature as control
genes, glyceraldehyde-3-phosphate dehydrogenase (GAPD), tubulin,
gamma 1 (TUBG1), general transcription factor IIIA (GTF3A), and
heterogenous ribonuclear protein A/B (HNRPAB). These genes were
used to correct for well-to-well variability in the small-molecule
library screen [see supporting information (SI) Table S1 and Table
S2 for raw data file mapping and marker gene probe sequence].
We first tested VPA at the reported HDAC inhibitory concen-
tration of 1 mM in BE (2)-C cells and confirmed an increase in
acetylation of histone H3 and H4 in BE (2)-C cells. Next, we
confirmed that significant induction of differentiation, as measured
by our neuroblastoma 14-gene expression signature, is not induced
by VPA alone. BE (2)-C cells were treated with compounds at a
final approximate concentration of 20
M for 3 days. Signature
genes were then quantified with a low-cost, high-throughput assay
enabling the detection of up to 100 genes. As previously reported,
this assay used ligation-mediated amplification (LMA) to amplify
marker genes and then a barcode, fluorescent-bead-based system to
detect and quantify the amplicons (7). The assay is highly quanti-
tative; it closely recapitulates differential gene expression identified
by microarray (7), and the neuroblastoma differentiation signature
shows a tight dose–response for control compounds (Fig. S1).
Several scoring methods were used to identify hits: a summed-score
metric combining expression data for each of the marker genes
assuming each gene contributes equal weight, a weighted summed
score that weights each gene based on the signal-to-noise ratio
determined from the differentiated positive and undifferentiated
negative controls, a naı¨ve Bayes classifier, and a KNN classifier. We
looked at compounds scoring across multiple scoring metrics in
determining hits. Only four compounds scored as differentiated
(ATRA, flumequine, cytarabine, and penicillamine) with a prob-
ability ⬎0.9 based on the summed score or weighted summed score
(Fig. 1B and Table S3). When evaluated more closely, for two of
these compounds (flumequine and penicillamine), only one of the
three replicates was driving the classification as differentiated,
suggesting that these two compounds were false positives. Only one
compound, ATRA, scored in all four statistical assessments. We
tested each of these compounds with the original assay at doses up
M in dose–response studies. Only ATRA was confirmed to
induce the signature (Fig. S2). ATRA is a reported inducer of dif-
ferentiation in neuroblastoma, as well as other solid tumors and
hematological malignancies in vitro (8–11). As such, perhaps it was
not surprising that ATRA scored so highly in the screen. However,
as the top hit identified, it raised the important question of whether
Ve hicle AT RA
Ve hicle AT RA PMA
2 04 06 08 0 150
0 -5 0 0 0 0 1 0 5 1 0 0 2 0 5 2
Hit #1: ATRA
Hit #2: Cytarabine
Hit #3: Flumequine
Hit #4: Penicillamine
Weighted Summed Score
Fig. 1. GE-HTS identiﬁes the combination of VPA and ATRA for the induction of neuroblastoma differentiation. (A) Ten neuroblastoma differentiation marker
genes were chosen to distinguish undifferentiated neuroblastoma from differentiated cells. Blue represents poorly expressed genes and red depicts highly
expressed genes. (B) Distribution of the screening data is shown for the summed-score and weighted summed-score metrics. Undifferentiated BE (2)-C controls
(DMSO treated) included in the screen are shown in yellow and differentiated controls (5
M ATRA treated) in gray. The black dots represent the compounds
that did not score as hits with either of these two measurements. Hits scoring as differentiated based on the overall scores of all replicates are ATRA (red),
cytarabine (blue), ﬂumequine (green), and penicillamine (purple). Each unﬁltered replicate is depicted in the ﬁgure.
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0710413105 Hahn et al.
the HDAC inhibitor enhances the known prodifferentiation effects
The Combination of VPA and ATRA Enhances Neuroblastoma Differ-
entiation. To address whether ATRA potentiates the differentiat-
ing effects of VPA, we measured the original differentiation
signature with vehicle, single agent VPA or ATRA, or both
compounds in combination at 3 days in BE (2)-C cells. With
low-dose ATRA (LD-ATRA) (10 nM) and high-dose ATRA
M) there was a greater induction of the original
differentiation gene expression signature in combination with VPA
than with ATRA alone (Fig. 2A). Next, we determined the extent
of differentiation with more standard assays of neurite maturation.
ATRA in combination with VPA induced more dramatic morpho-
logical evidence of differentiation with neurite extension and
extensive branching than either drug alone (Fig. 2B). Furthermore,
both compounds when used in combination induced greater ex-
pression of neurofilament medium (NF-M), a protein expressed by
mature neuronal cells. VPA and HD-ATRA single-agent treatment
induced minimal induction of NF-M (Fig. 2C).
The Combination of VPA and ATRA Synergize to Inhibit Neuroblastoma
Cell Viability and Induce Apoptosis. With the induction of terminal
differentiation programs, many cells will ultimately undergo apo-
ptosis. In light of their cooperative induction of differentiation, we
hypothesized that the combination of an HDAC inhibitor and
retinoid would enhance inhibition of neuroblastoma cell viability
and the induction of apoptosis. To this end, we determined the
effects of ATRA and VPA on cellular viability, measured with an
ATP-based assay. In BE (2)-C and in another N-myc amplified cell
line, IMR-32, the combination treatment had synergistic effects on
viability as determined by isobologram (Fig. 3A). We next ad-
dressed whether this effect on cellular viability was, at least in part,
related to apoptosis. LD-ATRA and VPA had an additive effect on
apoptosis induction, whereas HD-ATRA and VPA had a syner-
gistic effect based on the excess over Bliss independence (Fig. 3B;
+ + +
Weighted Summed Score
Vehicle VPA 1 mM
ATRA 5 µM
VPA 1 mM,
ATRA 5 µM
Fig. 2. The combination of VPA and ATRA induces
increased differentiation. (A) BE (2)-C cells were
treated with combinations of 1 mM VPA, 10 nM LD-
ATRA, and 5
M HD-ATRA and the neuroblastoma
differentiation signature evaluated. A box-and-
whisker plot demonstrates the distribution of
weighted summed scores for each sample type where
the heavy lines inside the box show the median, the
boxes show the quartiles, and the whiskers show the
extremes of the observed distribution of scores. (B)
May Grunwald Giemsa staining of BE (2)-C cells treated
for 5 days reveals maximal differentiation with both
agents in combination. Images were acquired with an
Olympus CK40 microscope, 400⫻ magniﬁcation, and
Qcapture software. (C) Western immunoblot of BE
(2)-C cells treated with either vehicle, ATRA (10 nM or
M), VPA (1 mM), or both agents in combination for
5 days and analyzed with antibody to neuroﬁlament
medium (NF-M) and GAPD as a control.
Vehicle ATRA 10 nM
ATRA 5 µM
VPA 1 mM
ATRA 10 nM, VPA 1 mM
ATRA 5 µM, VPA 1 mM
13.6% 5.6% 5.8%
Fig. 3. VPA and ATRA show synergistic effects on cell viability and cell death. (A) The combined effect of VPA and ATRA on BE (2)-C (Left) and IMR-32 (Right)
cell viability at 5 days, as determined by ATP level, is shown by isobologram. Synergy appears as points below the line of additivity. (B) BE (2)-C cells were treated
with compounds for 3 days. Combination treatment induced increased annexin V positive cells consistent with apoptosis with an additive interaction at the
low-dose ATRA and synergistic interaction with high-dose ATRA as evaluated by excess over Bliss independence. (C) Western blot analysis of histone acetylation
at6hinBE(2)-C cells treated with VPA 1 mm (V), ATRA 10 nM (A), and both agents in combination. HeLa cell controls ⫾ butyrate are included.
Hahn et al. PNAS
July 15, 2008
defined in SI Materials and Methods) (12, 13). The combination of
HD-ATRA and VPA exceeded Bliss independence with an excess
of 20% induction of apoptosis. These enhanced effects were not
related to further increases in hyperacetylation with the addition of
retinoids. ATRA alone did not increase histone H3 or H4 acety-
lation and did not enhance the hyperacetylation induced by the
HDAC inhibitor (Fig. 3C).
Differentiation Precedes Apoptosis in HDAC Inhibitor and Retinoid
Combination Therapy. To address whether cells terminally differen-
tiate and then die, we performed parallel evaluation of apoptosis
and differentiation in a time course experiment. BE (2)-C cells were
treated with vehicle or with the combination of 1 mM VPA and 5
M ATRA in triplicate for 6, 24, and 72 h. At each time point, cells
were evaluated by flow cytometry for Annexin V/FITC and PI
staining patterns indicative of early and late apoptosis and for
induction of the differentiation signature. As early as 6 h, evidence
of differentiation was seen that increased over the 3 days (Fig. 4A).
However, apoptosis was not identified until 72 h (Fig. 4 B and C).
Although we cannot exclude the possibility that these two are
unrelated, the temporal nature of this finding suggests that differ-
entiation occurs and then death by apoptosis follows.
The Combination of Other HDAC Inhibitors and Retinoids Enhances
Both Differentiation and Cell Death.
VPA is not a potent HDAC
inhibitor, with doses of 0.5 to 1 mM necessary to achieve HDAC
inhibitory activity. At this high dose, it is possible that the effects on
differentiation, in combination with ATRA, are related to a
non-HDAC-related mechanism of activity. For example, the anti-
epileptic effects of VPA are thought to be related to increases in the
-aminobutyric acid (GABA), an inhibitory neurotrans-
mitter (14). To address this issue, we tested two more potent HDAC
inhibitors in combination with ATRA: LAQ824 and SAHA. Both
of these HDAC inhibitors had synergistic effects on ATRA-related
decrease in neuroblastoma cell viability (Fig. S3 A and B) and
synergistic effects on ATRA-induced apoptosis (Fig. S3C). Com-
bination treatment also enhanced maturation effects of low-dose
and high-dose ATRA in neuroblastoma, measured by the quanti-
tative gene expression signature used in the original screening assay
(Fig. S3 D and E). Similarly, 13-cis retinoic acid (13-cis-RA) had
synergistic effects with HDAC inhibitors on cellular viability and
enhanced differentiation effects (Fig. S4 A and B).
The Combination of SAHA and ATRA Has Enhanced Activity in a
Xenograft Model of Neuroblastoma.
Combination testing was next
extended to an in vivo model of neuroblastoma. Neuroblastoma
BE-2(C) xenografts were established s.c. in NCr nude mice. For the
in vivo work, we chose to use SAHA. Animals were treated with
either ATRA at 2.5 mg/kg i.p. (IP) daily, SAHA 25 mg/kg IP daily,
both drugs, or vehicle for up to 21 days. Tumor was measured with
calipers and volume calculated per standard as V ⫽ 0.5 ⫻ length ⫻
width ⫻ width. A statistical difference in survival was demonstrated
with SAHA versus vehicle but not with ATRA versus vehicle. The
combination of SAHA and ATRA versus any of the other three
arms (vehicle, ATRA alone, SAHA alone) showed a significant
difference in survival. The longest surviving animals were treated
with both drugs (Fig. 5 A and B). We next attempted to address
whether the combination treatment was inducing differentiation,
cell death, or both in vivo. BE (2)-C xenografts were established in
NCr nude mice until tumor volume reached 100 mm
, divided into
treatment cohorts, and received either vehicle, ATRA 2.5 mg/kg IP
daily, SAHA 25 mg/kg IP daily, or a combination of ATRA and
SAHA for 4 days. Tumors were then harvested, formalin fixed, and
stained with hematoxylin and eosin (two from each treatment arm).
Those treated with combination therapy demonstrated such
marked increased cell death by pathological evaluation that it was
difficult to determine the exact cause of death in vivo (Fig. 5C). With
that said, we attempted to measure gene expression changes in the
neuroblastoma maturation signature by using extant RNA. We
found the greatest induction of the differentiation signature in the
combination-treated tumors, suggesting that differentiation may in
part be responsible for the enhanced in vivo effects of combination
treatment on survival (Fig. 5D).
Numerous obstacles exist in the development of therapies for
pediatric cancer. The rarity of these diseases is a considerable
problem. Neuroblastoma is the most common extracranial pediatric
solid tumor, yet only 650 children and adolescents will be diagnosed
each year in the United States (15). This is in sharp contrast to the
most common adult solid tumor, lung cancer, with more than
200,000 new cases diagnosed annually (16). With so few people
affected by this disease, there is a reduced market incentive for
industry-based drug development. A second challenge, even with
drug in hand, is the development of clinical trials adequately
powered to address the question of efficacy. When only a limited
number of drugs can be tested, rational selection for testing
Veh 6h Veh 24h Veh 72hV+A 6h V+A 24h V+A 72h
Mean Weighted Summed Score
% Annexin V + / PI -
% Annexin V + / PI +
Veh 6h Veh 24h Veh 72h V+A 6h V+A 24h V+A 72h
Veh 24h Veh 72hV+A 6h V+A 24h V+A 72h
Fig. 4. Differentiation precedes apoptosis in combination-treated neuro-
blastoma cells. (A) BE (2)-C cells were treated with either vehicle (veh) or the
combination of 5
M ATRA (A) and 1 mM VPA (V) for 6, 24, or 72 h and the
effects on the 14-gene differentiation signature were evaluated.
signiﬁcance in a pairwise t test comparing vehicle with drug treatment at each
time point. Duplicate biological replicates and 16 technical replicates for each
time point and condition were evaluated. BE (2)-C cells were treated in
triplicate as above and the effects on early apoptosis (annexin V-FITC positive
and PI negative) (B) and late apoptosis (annexin V-FITC positive and PI positive)
(C) were evaluated.
, statistical signiﬁcance in a pairwise t test comparing
vehicle with drug treatment at each time point.
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0710413105 Hahn et al.
becomes critical. Testing combinations of compounds adds yet
another layer of complexity.
Despite both in vitro and clinical trial data suggesting that
differentiation is an alternative therapeutic approach for neuro-
blastoma, limited effort has been placed on the discovery of new
differentiation agents. With existing assays it has been difficult to
perform quantitative assessments of neuroblastoma differentiation.
In fact, we are not aware of any published neuroblastoma differ-
entiation screens, underscoring the limitations of traditional phe-
notypic and target-based screening. No single marker is yet suffi-
cient to identify neuroblast differentiation in a high-throughput
screen. For example, although NF-M is generally considered a
marker of neuroblastoma differentiation, in our microarrray pro-
filing, NF-M decreased in SHSY5Y neuroblastoma cells differen-
tiated with PMA rather than increased as would be predicted.
Furthermore, with limitations to resources, industry is unlikely to
focus on the development of new assays for rare diseases. GE-HTS
offered a potential solution to these challenges. Because it is a
generic approach, it can be applied to many specific questions
without the need to invent an assay each time a small-molecule
library screen is performed. Second, the use of a complex signature
provides specificity and sensitivity advantages over conventional
single-gene reporter assays (17).
The concentrations of drugs used in combination can have
significant effects on gene expression as shown in the work by Lamb
et al. (18) and Cheok et al. (19) potentially complicating the
incorporation of gene expression assays into small-molecule screen-
ing. However, the work by Lamb et al. also suggests that with a
complex signature, one can see gene expression connections among
compounds even without elaborate optimization of compound
concentration. This underscores the importance of a sensitive,
information-rich primary screening assay (i.e., multigene signature)
as opposed to a single gene signature because one cannot evaluate
all possible doses in a primary screen. Careful dose–response
studies will then be characterized in secondary assays after hits are
identified by the primary screen. Here, GE-HTS enabled the
quantitative measurement of a complex neuroblastoma differenti-
One of the more striking recent findings in clinical oncology has
been the narrow activity of HDAC inhibitors (CTCL) despite their
predicted broad effects in cancer cells. An ongoing challenge is to
identify combinations of compounds that will broaden the efficacy
of this drug class. Because neuroblastoma is associated with defects
in neural crest differentiation, aberrant transcription is likely to play
a causal role in the pathogenesis of the tumor, implicating this
malignancy as a rational testing ground for new HDAC inhibitor
combination therapy. Our screen identified the combination of an
HDAC inhibitor and a retinoid as prodifferentiating. Earlier studies
demonstrated the activity of HDAC inhibitors on neuroblastoma
growth in vitro and in vivo xenograft models (20–25) but fewer have
evaluated the effects on differentiation (26). More recently, a
limited number of reports have evaluated the combination of
HDAC inhibitors and retinoids on cell growth in neuroblastoma in
vitro (25, 27) and one reported efficacy in combination in a
xenograft model of neuroblastoma (28). Our work confirms that the
combination indeed has greater effects on cell viability, and ulti-
mately apoptosis, than does either compound alone. Moreover, the
presented data reveal an indication for this combination in pro-
moting neuroblastoma differentiation supported by quantitative
expression-based assays and traditional assays of neuroblastoma
maturation. Multiple HDAC inhibitors (the short-chain fatty acid
VPA and the hydroxamic acids SAHA and LAQ824) with either
ATRA or 13-cis-RA had differentiating effects when used in
combination, suggesting that activity is truly related to these
compound classes rather than to an off-target effect. Moreover, the
combination of SAHA and ATRA had enhanced activity in a
xenograft model of neuroblastoma with prolonged survival of
Based on current understanding of retinoic acid receptor regu-
lation, an interaction between HDAC inhibitors and retinoids
might be predicted. In the absence of ligand, retinoic acid receptors
recruit coregulator complexes with HDAC activity leading to the
Fig. 5. ATRA and SAHA have enhanced activity in an in vivo model of neuro-
blastoma. (A) BE (2)-C xenografts were established for 10 days in NCr nude mice.
Animals received vehicle, ATRA 2.5 mg/kg IP daily, SAHA 25 mg/kg IP daily, or a
combination of ATRA and SAHA for up to 21 days. Error bars show standard error
of the mean across six or seven replicates. The x axis represents the days since the
initiation of treatment. (B) Percentage of surviving animals is shown. The x axis
represents the days since the initiation of treatment. The two-tailed P values of
the survival curves were determined by logrank test for pairwise comparisons:
vehicle vs. ATRA ⫽ NS, vehicle vs. SAHA ⫽ 0.05, vehicle vs. combo ⫽ 0.003, ATRA
vs. SAHA ⫽ NS, ATRA vs. combination ⫽ 0.001 and SAHA vs. combination ⫽ 0.009.
(C) BE (2)-C xenografts in NCr nude mice were treated with either vehicle, ATRA
2.5 mg/kg IP daily, SAHA 25 mg/kg IP daily, or a combination of ATRA and SAHA
for 4 days. Tumors treated with the combination of drugs demonstrated marked
increased areas of cell death by hematoxylin and eosin staining. Images were
acquired with an Olympus BX41 microscope, 40⫻ magniﬁcation, and Qcapture
software. (D)AsinC, BE (2)-C xenografts were established and treated. Three to
ﬁve tumors from each class were harvested, and the neuroblastoma differenti-
ation signature measured for each sample with 16 technical replicates. All drug
treatments were statistically elevated compared with vehicle, and combination
treatment was statistically elevated compared to single agent treatment (P ⬍
0.001 by t test).
Hahn et al. PNAS
July 15, 2008
repression of gene transcription (29). Several preclinical studies
have demonstrated an enhanced effect of retinoids with HDAC
inhibitors in acute myeloid leukemia. In the acute leukemias,
several transcription factor rearrangements result in the abnormal
recruitment of HDACs, hypoacetylation, and the repression of
transcription of prodifferentiation genes (30, 31). HDAC inhibitors
relieve this repression and presumably facilitate the activity of the
retinoid (32). For example, in retinoic acid-resistant acute promy-
elocytic leukemia (APL) with the PLZF-RAR
the addition of an HDAC inhibitor has restored ATRA respon-
siveness in in vitro and in vivo models (33). Furthermore, patient
responsiveness to this combination has also been reported (34).
A national Children’s Oncology Group phase I study of SAHA
for children with recurrent or refractory solid tumors and leukemias
followed by a phase I study of SAHA in combination with 13-
cis-RA for patients with selected recurrent/refractory solid tumors
was recently completed. Our data suggest that a phase II trial testing
SAHA and retinoid for patients with high-risk neuroblastoma is
warranted. Moreover, this work further develops a generalizable
approach to small-molecule library screening, one that can bring
compound discovery to many orphan diseases, including the pedi-
Materials and Methods
Cell Culture. Neuroblastoma cell lines were purchased from the American Type
Culture Collection and grown in Dulbecco’s Modiﬁed Eagle Medium (DMEM)
(Cellgro) supplemented with 1% penicillin-streptomycin-glutamine (Cellgro) and
10% FCS (Sigma-Aldrich).
Marker Gene Selection. RNA was extracted with TRIzol per the manufacturer’s
protocol (Invitrogen) from BE (2)-C and SHSY5Y neuroblastoma cells treated in
triplicate with vehicle (ethanol) versus 5
M all-trans-retinoic acid (ATRA) for 5
days and in SHSY5Y treated with16 nM phorbol 12-myristate 13-acetate (PMA)
for 5 days. Gene expression was evaluated with Affymetrix U133A DNA microar-
rays (see SI Materials and Methods for full details). Raw microarray data are
available at http://www.broad.mit.edu/cancer/pub/Neuroblastoma㛭GE-HTS.
Small-Molecule Library Screen Methods. The GE-HTS assay was carried out as
described in ref. 7 and as detailed in the SI Materials and Methods. BE (2)-C
cells were plated in 384-well format with 2,000 cells per well and treated with
1 mM VPA. Compounds were added at a ﬁnal approximate concentration of
M in DMSO and incubated for 3 days. We screened in triplicate the NINDS
small-molecule collection (http://www.broad.harvard.edu/chembio/platform/
Viability Assay. Viability experiments were performed in 96-well format in
duplicate, and then the experiment repeated two to three times, by using the
Promega Cell-Titer Glo ATP-based assay per the manufacturer’s instructions.
Synergy was assessed by analyzing the IC
of one drug over a range of concen-
trations of the other drug and vice versa. The resulting concentration pairs were
visualized by isobologram (35).
Morphological Evaluation. BE (2)-C cells were evaluated by May–Grunwald
Giemsa staining with an Olympus CK40 microscope and Q-capture software.
Western Blot Analysis and Histone Extraction. All proteins were detected by
chemiluminescence and antibodies to NF-M (SC-20013, Santa-Cruz) GAPD (Ab
22556, Abcam), anti-acetyl-histone H3 (06-599, Upstate), and anti-acetyl-histone
H4 (06-866, Upstate). Control histones were untreated HeLa cell acid extract
(13-112, Upstate) and sodium butyrate-treated HeLa cell acid extract (13-113,
Upstate). See SI Materials and Methods for histone extraction and Western blot
Apoptosis Assay. Annexin V FITC/PI staining was performed with 500,000 cells by
using the Annexin V: FITC Apoptosis Detection Kit I (BD PharMingen). Cells were
analyzed by ﬂow cytometry with a FACScan ﬂow cytometer (Becton Dickinson)
and CELLQuest analytical software.
Chemicals. Chemicals were obtained from the following sources: VPA, ATRA, and
PMA (Sigma), LAQ-824 (Novartis), and SAHA (Broad Chemistry Program).
Xenograft Studies. Xenograft studies were performed with male NCr nude mice.
The four treatment groups were vehicle, ATRA alone (2.5 mg/kg), SAHA alone (25
mg/kg), or a combination of ATRA and SAHA, and animals were treated for either
21 or 4 days. Caliper measurements were used to calculate tumor volume by using
the formula: Volume ⫽ 0.5 ⫻ length ⫻ (width)
. All animal experiments were
performed after approval from the Institutional Animal Care and Use Committee.
See SI Materials and Methods for full details.
ACKNOWLEDGMENTS. We thank Paul Clemons, Nicola Tolliday, and Stuart
Schreiber for their guidance, David Peck for GE-HTS platform support,Zi Peng Fan
for technical support, and Jonathan Jesneck for computational discussions. This
work was funded by the Friends for Life Fellowship Program, a Children’s Oncol-
ogy Group Young Investigator Award, and the Claudia Adams Barr Program in
Cancer Research (K.S.).
1. George RE, et al. (2006) High-risk neuroblastoma treated with tandem autologous
peripheral-blood stem cell-supported transplantation: Long-term survival update.
J Clin Oncol 24:2891–2896.
2. Furchert SE (2007) Inhibitors of histone deacetylases as potential therapeutic tools for high-risk
embryonal tumors of the nervous system of childhood. Int J Cancer 120:1787–1794.
3. Bolden JE, Peart MJ, Johnstone RW (2006) Anticancer activities of histone deacetylase
inhibitors. Nature Rev 5:769 –784.
4. Duvic M, et al. (2007) Phase 2 trial of oral vorinostat (suberoylanilide hydroxamic acid,
SAHA) for refractory cutaneous T-cell lymphoma (CTCL). Blood 109:31–39.
5. Matthay KK, et al. (1999) Treatment of high-risk neuroblastoma with intensive che-
motherapy, radiotherapy, autologous bone marrow transplantation, and 13-cis-
retinoic acid. Children’s Cancer Group. N Engl J Med 341:1165–1173.
6. Stegmaier K, et al. (2004) Gene expression-based high-throughput screening(GE-HTS)
and application to leukemia differentiation. Nat Genet 36:257–263.
7. Peck D, et al. (2006) A method for high-throughput gene expression signature analysis.
Genome Biol 7:R61.
8. Pahlman S, Odelstad L, Larsson E, Grotte G, Nilsson K (1981) Neuron speciﬁc enolase: a marker
for differential diagnosis of neuroblastoma and Wilms’ tumor. Int J Cancer 28:583–589.
9. Reynolds CP, et al. (1991) Response of neuroblastoma to retinoic acid in vitro and in
vivo. Prog Clin Biol Res 366:203–211.
10. Sidell N (1982) Retinoic acid-induced growth inhibition and morphologic differentia-
tion of human neuroblastoma cells in vitro. J Natl Cancer Inst 68:589 –596.
11. Guzhova I, et al. (2001) Interferon-gamma cooperates with retinoic acid and phorbol
ester to induce differentiation and growth inhibition of human neuroblastoma cells.
Int J Cancer 94:97–108.
12. Bliss CI (1939) The toxicity of poisons applied jointly. Ann Appl Biol 26:585–615.
13. Keith CT, Borisy AA, Stockwell BR (2005) Multicomponent therapeutics for networked
systems. Nat Rev 4:71–78.
14. Johannessen CU (2000) Mechanisms of action of valproate: a commentatory. Neuro-
chem Int 37:103–110.
15. Goodman MT, Gurney JG, Smith MA, Olshan AF (1999) Cancer Incidence and Survival
among Children and Adolescents: United States SEER Program 1975-1995, eds Ries
LAG, et al. (National Cancer Institute, Bethesda, MD), pp 65–72.
16. Ries LAG, et al., eds (2006) SEER Cancer Statistics Review, 1975-2004 (National Cancer
Institute., Bethesda, MD).
17. Hieronymus H, et al. (2006) Gene expression signature-based chemical genomic pre-
diction identiﬁes a novel class of HSP90 pathway modulators. Cancer Cell 10:321–330.
18. Lamb J, et al. (2006) The Connectivity Map: Using gene-expression signatures to
connect small molecules, genes, and disease. Science 313:1929 –1935.
19. Cheok MH, et al. (2003) Treatment-speciﬁc changes in gene expression discriminate in
vivo drug response in human leukemia cells. Nat Genet 34:85–90.
20. Cinatl J, Jr, et al. (1997) Sodium valproate inhibits in vivo growth of human neuro-
blastoma cells. Anticancer Drugs 8:958 –963.
21. Jaboin J, et al. (2002) MS-27–275, an inhibitor of histone deacetylase, has marked in vitro
and in vivo antitumor activity against pediatric solid tumors. Cancer Res 62:6108 – 6115.
22. Glick RD, et al. (1999) Hybrid polar histone deacetylase inhibitor induces apoptosis and
CD95/CD95 ligand expression in human neuroblastoma. Cancer Res 59:4392– 4399.
23. de Ruijter AJ, et al. (2004) The novel histone deacetylase inhibitor BL1521 inhibits prolif-
eration and induces apoptosis in neuroblastoma cells. Biochem Pharmacol 68:1279 –1288.
24. Ouwehand K, de Ruijter AJ, van Bree C, Caron HN, van Kuilenburg AB (2005) Histone
deacetylase inhibitor BL1521 induces a G1-phase arrest in neuroblastoma cells through
altered expression of cell cycle proteins. FEBS Lett 579:1523–1528.
25. De los Santos M, Zambrano A, Aranda A (2007) Combined effects of retinoic acid and
histone deacetylase inhibitors on human neuroblastoma SH-SY5Y cells. Mol Cancer
26. Cinatl J, Jr, et al. (2002) Induction of differentiation and suppression of malignant
phenotype of human neuroblastoma BE(2)-C cells by valproic acid: enhancement by
combination with interferon-alpha. Int J Oncol 20:97–106.
27. Coffey DC, et al. (2000) Histone deacetylase inhibitors and retinoic acids inhibit growth
of human neuroblastoma in vitro. Med Pediat Oncol 35:577–581.
28. Coffey DC, et al. (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.
29. Nagy L, et al. (1997) Nuclear receptor repression mediated by a complex containing
SMRT, mSin3A, and histone deacetylase. Cell 89:373–380.
30. Lutterbach B, et al. (1998) ETO, a target of t(8;21) in acute leukemia, interacts with the
N-CoR and mSin3 corepressors. Mol Cell Biol 18:7176–7184.
31. Lin RJ, Nagy L, Inoue S, Shao W, Miller WH, Jr, Evans RM (1998) Role of the histone
deacetylase complex in acute promyelocytic leukaemia. Nature 391:811– 814.
32. Ferrara FF, et al. (2001) Histone deacetylase-targeted treatment restores retinoic acid
signaling and differentiation in acute myeloid leukemia. Cancer Res 61:2–7.
33. He LZ, et al. (2001) Histone deacetylase inhibitors induce remission in transgenic models of
therapy-resistant acute promyelocytic leukemia. J Clin Invest 108:1321–1330.
34. Warrell RP, Jr, He LZ, Richon V, Calleja E, Pandolﬁ PP (1998) Therapeutic targeting of
transcription in acute promyelocytic leukemiabyuse of an inhibitor of histone deacety-
lase. J Natl Cancer Inst 90:1621–1625.
35. Gessner PK (1995) Isobolographic analysis of interactions: an update on applications
and utility. Toxicology 105:161–179.
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0710413105 Hahn et al.