Identification of Drugs Including a
Dopamine Receptor Antagonist that
Selectively Target Cancer Stem Cells
Eleftherios Sachlos,1Ruth M. Risuen ˜o,1Sarah Laronde,1Zoya Shapovalova,1Jong-Hee Lee,1Jennifer Russell,1
Monika Malig,1Jamie D. McNicol,1Aline Fiebig-Comyn,1Monica Graham,1Marilyne Levadoux-Martin,1Jung Bok Lee,1
Andrew O. Giacomelli,2John A. Hassell,2Daniela Fischer-Russell,1Michael R. Trus,3Ronan Foley,3Brian Leber,3
Anargyros Xenocostas,4Eric D. Brown,2Tony J. Collins,1and Mickie Bhatia1,2,*
1McMaster Stem Cell and Cancer Research Institute, Faculty of Health Sciences, McMaster University, 1280 Main Street West, MDCL 5029,
Hamilton, ON L8S4L8, Canada
2Department of Biochemistry and Biomedical Sciences, Faculty of Health Sciences, McMaster University, 1280 Main Street West, Hamilton,
ON L8S4L8, Canada
3Department of Pathology and Molecular Medicine, McMaster University, Hamilton, ON L8S4L8, Canada
Selective targeting of cancer stem cells (CSCs) offers
promise for a new generation of therapeutics.
However, assays for both human CSCs and normal
stem cells that are amenable to robust biological
screens are limited. Using a discovery platform that
reveals differences between neoplastic and normal
human pluripotent stem cells (hPSC), we identify
small molecules from libraries of known compounds
that induce differentiation to overcome neoplastic
self-renewal. Surprisingly, thioridazine, an antipsy-
chotic drug, selectively targets the neoplastic cells,
and impairs human somatic CSCs capable of in vivo
leukemic disease initiation while having no effect on
normal blood SCs. The drug antagonizes dopamine
receptors that are expressed on CSCs and on breast
cancer cells as well. These results suggest that
dopamine receptors may serve as a biomarker for
diverse malignancies, demonstrate the utility of
drugs, and provide support for the use of differentia-
tion as a therapeutic strategy.
Current chemotherapy treatments define rates of cancer patient
survival that have remained largely unchanged over the past
3 decades (Jemal et al., 2010). This implies that alternative
conceptual and practical approaches are required to treat
human cancer. Increasing evidence suggests that cancer/tumor
development is due to a rare population of cells, termed cancer
stem cells (CSCs) (Dick, 2009; Reya et al., 2001) that uniquely
initiates and sustains disease. In addition, experimental
evidence indicates that conventional chemotherapeutics are
actually ineffective against human CSCs (Guan et al., 2003).
This resistance to chemotherapeutics in the human is coupled
with indiscriminate cytotoxicity that often affects healthy stem
and progenitor cells, leading to dose restriction and necessi-
tating supportive treatment (Smith et al., 2006). Recent exam-
ples include selective induction of apoptosis (Atkinson et al.,
2011; Gupta et al., 2009) that remains to be tested in normal
SCs or in the human system, or driven by a specific oncogenic
factor. Accordingly, assays for human CSCs and normal SCs
that allow identification of agents that target CSCs alone are
now critical to provide truly selective anticancer drugs that could
be useful for preclinical testing.
Normal and neoplastic SCs are functionally defined by
a tightly controlled equilibrium between self-renewal versus
differentiation potential. In the case of CSCs, this equilibrium
shifts toward enhanced self-renewal and survival leading to
limited differentiation capacity allowing tumor growth. In
contrast to direct toxic effects that equally affect normal SCs,
an alternative approach, eradication of CSCs, may be achiev-
able by overcoming the differentiation block of CSCs. First
proposed in the 1970s (Sachs, 1978), this approach led to the
identification and clinical use of all-trans-retinoic acid (ATRA)
(Breitman et al., 1980) and arsenic trioxide (ATO) (Niu et al.,
1999) for patients with acute promyelocytic leukemias (APLs).
Subsequent studies have demonstrated that both of these
agents exert their effects by targeting the PML-RARa oncogenic
fusion protein that is specific to APL (Yoshida et al., 1996). In
contrast to the human, murine studies indicate degradation of
PML-RARa leads to APL remission, whereas ATRA’s differenti-
ation-inducing properties may be secondary (Nasr et al., 2008).
Despite these murine observations, combined ATRA and ATO
therapy allows for > 93% remission rates with a 5 year overall
patient survival rate approaching 100% (Wang and Chen,
2008), exemplifying how understanding differentiation induction
can be used in cancer.
1284 Cell 149, 1284–1297, June 8, 2012 ª2012 Elsevier Inc.
Cancer differentiation therapy has not been translated to the
treatment of other cancers, let alone other acute myeloid
leukemia (AML) subtypes (Estey et al., 1999). This may be in
part due to the absence of robust in vitro assays that can inter-
rogate human CSC differentiation. To this end, we have previ-
ously described a variant human pluripotent stem cell (hPSC)
line that faithfully reproduces, in vitro and in vivo, neoplastic
properties of somatic CSCs, including enhanced self-renewal
and survival and aberrant block in differentiation capacity
(Werbowetski-Ogilvie et al., 2009). Based on these shared prop-
erties with somatic CSCs, we examined whether neoplastic
hPSCs could serve as a surrogate for somatic CSCs that would
be amenable for high content screening in vitro.
Differentiation Response of Neoplastic hPSC Models
of Human Cancer
The inability of neoplastic hPSCs to terminally differentiate
while retaining self-renewing capacity represents key in vivo
features used to functionally define somatic CSC development
in human tumors (Werbowetski-Ogilvie et al., 2009). Distinct
from pluripotent teratomas generated from normal hPSCs, tera-
tomas generated from neoplastic hPSCs show features of more
primitive lineage-specific cells and retain a primitive Oct4+ pop-
ulation (Figures 1A and 1B) (Werbowetski-Ogilvie et al., 2009).
Furthermore, only neoplastic hPSC-derived teratomas possess
5K cells / well
stem cell &
stem cell &
10K cells / well
Oct4 / Hoechst
GFP / Hoechst
Neoplastic hPSC cell count
Normal hPSC cell count
Neoplastic hPSC GFP intenisty per cell
Normal hPSC Oct4 intensity per cell
(B) Neoplastic hPSCs refractory to
loss of pluripotency (LOP)
(C) LOP for normal
and neoplastic hPSCs
(A) No response
Endoderm MesodermEctodermEndodermMesoderm Ectoderm
Figure 1. Neoplastic hPSCs Recapitulate Cancer Development and Can Be Induced to Differentiate
(A) Histological assessment of normal H9-generated teratoma displaying derivatives of all three germ layers.
(B) Neoplastic v1H9-Oct4-GFP cells generated teratomas displaying populations of Oct4+ cells.
(C) Flow cytometry showing frequency of Oct4+ in normal H9 with or without BMP4 treatment.
(D) Flow cytometry of v1H9-Oct4-GFP cells showing changes in GFP and Oct4 levels with or without BMP4 treatment.
(E) Schematic representation of differential screening with H9 and v1H9-Oct4-GFP cells.
(F) Cell counts measured by high content analysis of H9 versus v1H9-Oct4-GFP challenged with 51 defined stem cell and anticancer activity compounds. Toxic
compounds are defined by a reduction in cell count bellow 2,500 cells per field of view for H9 and 900 cells per field of view for v1H9-Oct4-GFP. Each point n = 3;
mean ± SEM.
(G) The mean Oct4 intensity per cell of H9 and mean GFP intensity per cell of v1H9-Oct4-GFP was measured by pixel quantification and plotted. Note the three
distinct regions A, B and C outlined. Each point n = 3; mean± SEM.
(H) Summary of the number of responses seen with the 51 defined compounds.
See also Figure S1 and Tables S1 and S2.
Cell 149, 1284–1297, June 8, 2012 ª2012 Elsevier Inc. 1285
secondary teratoma generation capacity (Table S1), demon-
strating unique self-renewal capacity versus normal hPSCs
despite being subjected to the strong in vivo differentiation-
inducing environment. In vivo similarities of neoplastic hPSCs
to functional features of somatic CSCs (Reya et al., 2001),
coupled with the feasibility of culturing hPSCs, formed the basis
to examine whether neoplastic hPSCs and their normal hPSC
counterparts could serve as a surrogate-model system to
identify and validate selective anti-CSC agents based on differ-
Differentiation of human SCs occurs only by initial attenuation
of regulators that maintain a self-renewing state. In the case of
hPSCs, suppression of differentiation is governed by master
pluripotency transcription factors of which Octamer 4 (Oct4)
and SRY-box 2 (Sox2) (Boyer et al., 2005) are central. Loss of
Oct4 is one of the first identifiable steps of commitment from
pluripotent state toward differentiation (Nichols et al., 1998).
Bone Morphogenetic Protein 4 (BMP4) induces differentiation
of hPSCs (Chadwick et al., 2003; Desbordes et al., 2008; Xu
et al., 2002), resulting in immediate reduction of Oct4 in normal
(Figure 1C) and neoplastic (Figure 1D and Figure S1A available
online) hPSCs. Similarly, BMP4 treatment reduced Sox2 in
normal (Figure S1B) and neoplastic (Figures S1C and S1D)
hPSCs, confirming loss of the pluripotent state. BMP4-induced
loss of pluripotency concomitantly resulted in downregulation
of a network of genes associated with pluripotency, and upregu-
lation of > 20 genes associated with differentiation programming
(Figure S1E) in both normal (Figure S1F) and neoplastic (Fig-
ure S1G) hPSCs relative to treated controls (Figures S1H and
S1I). These data indicate that reduction of Oct4 and Sox2 in
viable cells provides a reliable indicator of loss of the self-renew-
ing pluripotent state accompanied by differentiation induction of
normal and neoplastic hPSCs.
To providea straightforward methodfor detecting lossof Oct4
or Sox2, we generated green fluorescent protein (GFP) reporter
lines by transduction of neoplastic hPSCs with the EOS-GFP
reporter (v1H9-Oct4-GFP and v1H9-Sox2-GFP, respectively)
(Hotta et al., 2009). GFP intensity correlated with Oct4 and
Sox2 expression in treatments that favored self-renewal stability
and conditions that induced differentiation with the addition of
BMP4 (Figure 1D and Figure S1D). This response was consis-
tently found by using an additional neoplastic hPSC line, v2H9
(Werbowetski-Ogilvie et al., 2009) transduced with the same
EOS-lentivirus GFP reporter (v2H9-Oct4-GFP; Figure S1J), as
well as a Sox2 reporter line (v1H9-Sox2-GFP; Figure S1D). The
uniform response to differentiation and maintenance of pluripo-
tency in all hPSC lines generated also reveals that viral integra-
tion or clonal selection by EOS reporter construct insertion is
irrelevant to responsiveness.
These results suggest that compounds that induce differenti-
ation can be identified based on the reduction of GFP intensity
in neoplastic hPSC reporter lines and could be exploited for
chemical screening. To that end, conditions for automated
high content microscopy (HCS) and fluorimetric-based high-
throughput plate reader screening (PRS) were used to detect
reductions in pluripotency marker expression of hPSCs. Micro-
scopic analysis of normal hPSCs showed that distinct Oct4+
cells are lost after BMP4 treatment (Figures S1K and S1L). Simi-
larly, the reduction in both GFP and Oct4 due to BMP4 treatment
of neoplastic Oct4-GFP hPSCs was quantified by HCS (Figures
S1M and S1N) and PRS (Figures S1O and S1P). BMP4-treated
(+BMP4) versus standard culture conditions (?BMP4) were
demarked ‘‘differentiated-state’’ and ‘‘pluripotent-state’’ thresh-
olds, respectively (Figures 1C, 1D,and 1K–1N), and used for
evaluation of chemical libraries containing compounds of
unknown function on human SCs. These data provide the
parameters for a screening platform employing both normal
and neoplastic hPSCs to identify anticancer agents that repre-
sent inducers of neoplastic cell differentiation.
Both HCS and PRS assays were optimized for 96-well plate
screening (Figures S1Q–S1V, see Extended Experimental
Procedures for optimization and definition of this hPSC culture
platform). This hPSC-screening platform was validated by using
an assembled library of 51 defined compounds with established
stem cell and anticancer activity (Figure 1E and Table S2).
The level of pluripotency marker expression and cell count
(defined by nuclei stained with Hoechst) for each compound
were recorded by using automated microscopy. Comparison
of normal versus neoplastic hPSC cell counts revealed
12 compounds cytotoxic to normal hPSCs of which 11 were
also cytotoxic to neoplastic hPSCs (Figure 1F). Such nonselec-
tive cytotoxic compounds that equally target normal SCs are
of limited value and were excluded from further evaluation.
Instead, preferred compounds were selected for the ability to
reduce Oct4 by selectively inducing differentiation of neoplastic
hPSCs, whereas having little effect on normal hPSCs. To identify
such an ideal candidate compound that fulfills the above crite-
rion, differentiation of both normal and neoplastic hPSCs in
response to compound treatment was assessed in parallel (Fig-
ure 1G). Using these 51 compounds, three distinct regions
of response could be resolved based on cellular response: (1)
no discernable reduction in pluripotency for differentiation
(38 compounds); (2) differentiation of normal hPSCs alone (nine
compounds); and (3) differentiation induction of both normal
and neoplastic hPSCs (four compounds) (Figures 1G and 1H).
Importantly, active compounds from regions B and C were not
detectable by using conventional cell lines utilized in drug
screening, e.g., HeLa cells (Figures S1W–S1Z), illustrating the
read-out in this hPSC-based platform.
Compounds that can induce differentiation of neoplastic
hPSCs are potential anticancer candidates. Region C maps
four such compounds; salinomycin, rapamycin, indolactam,
and PMA. These compounds also differentiate normal hPSCs
(Figure 1G) and are useful in verifying, defining, and validating
our screen. However, inducers of normal SC differentiation
were deemed undesirable as anti-CSC agents due to the likely
exhaustion of the normal SC population through overstimulation
ofdifferentiation. Instead,the idealcompoundshould onlydiffer-
entiate neoplastic hPSCs while not affecting normal hPSCs.
Identification of Compounds that Induce Differentiation
of Neoplastic hPSCs
Given the validation of our screening platform (Figure 1), we
extended our tests to chemical libraries composed of 590 well-
established annotated compounds from the NIH Clinical Collec-
tion and Canadian Compound Collection (Table S3). PRS and
1286 Cell 149, 1284–1297, June 8, 2012 ª2012 Elsevier Inc.
HCS platforms give equivalent measurements (Figures S1Y–
S1Z). PRS was selected for more rapidly screening compound
libraries (Figure 2A). Of the 590 compounds screened (at
10 mM based on previous studies), 11 compounds were identi-
fied to induce differentiation as indicated by a reduction in
both GFP percentage residual activity (%RA) and Hoechst
%RA(Figures 2Band2C).Four ofthesecompounds (indatraline,
thioridazine, azathioprine, and mefloquine) were identified as
candidate compounds based on clustering and levels of
Hoechst %RA in excess of 30% (Figure 2B). Secondary high
content analysis did not validate indatraline as a candidate and
was thus excluded, whereas high content and PRS analyses
dually confirmed thioridazine, azathioprine, and mefloquine as
candidate compounds (Figure 2D) and were thus selected for
further testing (Figures 2E–2G). When compared to control-
treated hPSCs (Figures 2E and 2F), each compound appeared
to induce distinct morphological changes in neoplastic hPSCs
(Figure 2E). Reduction in GFP intensity was confirmed by using
image analysis (Figure 2F) and further assessed over a wide
range of doses to calculate the half-maximal effective concen-
tration (EC50) for each compound (Figures 2G and Figure S2).
Only thioridazine and mefloquine were found to possess EC50
values lower than the 10 mM target threshold (Figure 2G) and
thus defined as candidates for further in-depth evaluation by
using neoplastic hPSCs and somatic CSCs from patients.
Thioridazine Selectively Induces Neoplastic hPSC
Differentiation and Reduces AML-Blasts without
Affecting Normal HSPCs
The responses to thioridazine and mefloquine were evaluated in
both normal (Figure 3A) and neoplastic hPSCs (Figure 3B) at
three concentrations by using quantitative flow cytometry to
detect the loss of Oct4 and reveal the degree of differentiation.
Salinomycin, a reported selective inhibitor of breast CSCs
(Gupta et al., 2009), was included for comparison. At 10 mM, all
compounds reduced the number of cells, but the levels of
Oct4inremaining normalhPSCs were notbelow levels observed
with BMP4 treatment (Figure 3A). This same response was repli-
cated in fibroblast-derived human iPSCs (Figure S3A), repre-
senting an additional normal hPSC line from a distinct (adult)
origin, indicating the effects are not specific to embryonic sour-
ces. When the same compounds were used to treat neoplastic
hPSCs, mefloquine and thioridazine treatments caused reduc-
tions in cell number and levels of Oct4. Only thioridazine was
overcome the neoplastic hPSC differentiation block. To identify
compounds that selectively differentiate neoplastic hPSCs
quantitatively, the ratio of normalized percentage of Oct4+ cells
between normal and neoplastic hPSCs in response to these
compounds was determined. For example, a ratio of 1 suggests
equivalent differentiation whereas a ratio > 1 defines relatively
more differentiation in neoplastic hPSCs versus normal hPSCs.
Only thioridazine, at both 1 mM and 10 mM, had a significant
impact on inducing differentiation of neoplastic hPSCs over
normal hPSCs (Figure 3C). Rapid accumulation of the cell stress
marker p53 (Figure 3D) and its transcriptional target p21 (Fig-
ure 3E) were used to further distinguish differentiation induction
from cellular toxicity. Treatment of neoplastic hPSCs with the
toxic chemotherapeutic agent etoposide resulted in high levels
of p53 and p21 after 24 hr. However, treatment with 10 mM thio-
ridazine or BMP4, unlike agents that induce toxicity alone,
resulted in no accumulation of p53 or p21, consistent with
induced differentiation rather than stress-response programs.
tiation genes quantified by TaqMan Low-Density Array-qPCR in
neoplastic hPSCs. An upregulation in 21 of 50 differentiation-
associated genes (Figure 3F) was observed in treated neoplastic
hPSCs consistent with the differentiation-inducing effects of
To examine the potential similarities in chemical response of
neoplastic hPSCs to somatic CSCs, we assessed normal and
neoplastic populations of the human hematopoietic system by
using powerful and well established in vitro and in vivo assays
(Bhatia et al., 1997; Bonnet and Dick, 1997). Lineage-depleted
umbilical cord blood (CB lin?) is highly enriched for hematopoi-
etic stem-progenitor cells (HSPCs) and is a reliable source of
normal somatic SCs capable of self-renewal and multi-lineage
differentiation to all blood lineages. AML is a hematological
neoplasia characterized by a block in mature myeloid differenti-
ation that is sustained by a self-renewing leukemic stem cell
(LSC) (Lapidot et al., 1994). As such, progenitor assays in meth-
ylcellulose were conducted with HSPCs and five AML patient
samples, each treated with thioridazine, mefloquine, or salino-
mycin in order to assess each compound’s impact on in vitro
clonogenic and multilineage
Representative cell pellets of the total colony-forming units
(CFUs) generated from HSPCs (Figure 3G) and AML (Figure 3H)
treated with each compound are shown. Thioridazine treatment
resulted in a reduction in AML proliferation/clonogenic capacity
while retaining HSPC multilineage differentiation (Figure S3C).
Changes in multilineage differentiation were quantified based
on the enumeration of CFUs generated following treatment of
HSPCs (Figure 3I) and AML patient (Figure 3J) samples with
thesecompounds.At both 1 mM and 10mM salinomycin reduced
AML-blast CFU potential (Figure 3J) but also reduced HSPC
CFU potential over all doses tested (Figure 3I), indicative of
nonspecific toxicity in the hematopoietic system. In contrast,
mefloquine and thioridazine reduced AML-blast CFU formation
(Figure 3J) while having little effect on HSPC CFU potential (Fig-
ure 3I) and multilineage composition (Figure S3D), indicating that
mefloquine and thioridazine do not alter normal hematopoiesis.
We calculated the ratio between total CFUs generated from
HSPCs versus AML-blasts to reveal the highest selectivity for
targeting AML (Figure 3K). A ratio of 1 suggests equivalent
normal to neoplastic progenitor potential whereas a ratio > 1
defines a compound that selectively reduces AML-blast CFU
potential. Salinomycin (1 mM), mefloquine (10 mM), and thiorida-
zine (10 mM) doses yielded the highest ratio values for each
compound (Figure 3K) and were thus selected for in vivo evalu-
ation. Thioridazine 10 mM, in particular, demonstrated the high-
est ratio of all compounds but most importantly was the only
compound to show significantly lower AML-blast CFU potential
relative to normal HSPC CFU potential (Figure 3K). To address
whether thioridazine’s specificity was due to induction of differ-
entiation, the frequency of CD11b, a marker of granulocytic
Cell 149, 1284–1297, June 8, 2012 ª2012 Elsevier Inc. 1287
Figure 2. Chemical Screening for Compounds that Differentiate Neoplastic hPSC Identifies Mefloquine and Thioridazine
(A) Schematic of screening strategy.
(B)XYscatterplotof percent residual activity(%RA) of GFPand Hoechst signals of the590compound screen.Regionoutlineddemonstrates lossof pluripotency
(LOP) as defined by reduced GFP and Hoechst. Each point n = 3, mean ± SD.
(C) Summary of responses seen with 590 compounds.
(D) Chemical structure of candidate compounds; thioridazine, azathioprine, and mefloquine.
(E) Representative GFP, Hoechst, and merged microscopic images of v1H9-Oct4-GFP cells treated with candidate compounds at 10 mM compared to controls
without (?BMP4) or with BMP4 (+BMP4) treatment.
(F) Histogram of GFP intensity of these images.
(G) Dose-response curves of v1H9-Oct4-GFP treated with candidate compounds and calculation of EC50. Each point n = 3; mean ± SEM.
See also Figure S2 and Table S3.
1288 Cell 149, 1284–1297, June 8, 2012 ª2012 Elsevier Inc.
maturation in patient AML cells, was assayed in response to
thioridazine treatment (Figure 3L). A marked increase in the
frequency of granulocytic AML-blast cells was observed with
treatment duration (Figure 3L), indicating that thioridazine
exhibits its specific targeting of AML cells through induction of
differentiation. This finding is analogous to differentiation-induc-
tion demonstrated in neoplastic hPSCs (Figures 3A–3F) and
confirms the robust readout of this screening platform toward
identifying agents able to differentiate neoplastic cells. This
result also suggests that thioridazine may represent the best
candidate for specific targeting of AML CSCs, which requires
testing using in vivo human-mouse xenograft assays.
Identification of Antileukemic Agents by Using
Neoplastic hPSC Screening
To reaffirm our screening approach and specificity to identify
thioridazine-like compounds, we expanded the chemical matter
used to screen neoplastic hPSC response to include 2,446
compounds (Figure 4A). Thioridazine, along with two other
phenothiazine compounds, fluphenazine and prochlorperazine,
were identified as hits among a list of 26 compounds identified
(Figures 4A–4C). Similarly, rapamycin, an mTOR-inhibitor, and
lestaurtinib, a tyrosine kinase inhibitor, were both independently
ties (Figures 4A–4C). Rapamycin’s antileukemic effects have
been studied in mouse models (Yilmaz et al., 2006) and demon-
strated in human AML patients (Re ´cher et al., 2005). Lestaurti-
nib’s antileukemic properties have been associated with differ-
entiation-induction (Zheng et al., 2002) and also tested in
human AML patients (Smith et al., 2004). Further assessment
of fluphenazine, prochlorperazine, rapamycin, and lestaurtinib
by using high content analysis revealed distinct morphological
changes in neoplastic hPSCs (Figure 4D) relative to control-
treated cells (Figure S4A). Reduction in GFP intensity was
confirmed by using image analysis (Figure 4E) and further
assessed over a wide range of doses to calculate the EC50for
each compound (Figure 4F). Of the three phenothiazines identi-
fied in the screens, thioridazine (confirmed by chemical analyses
Figures S4B–S4H) exhibited the lowest EC50 in neoplastic
phenothiazine of those tested for targeting of AML CSCs by
using human-mouse xenograft assays as in vivo readouts. The
identification of rapamycin and lestaurtinib, two established clin-
ically-relevant antileukemic agents, reinforces the detection
capacity of our screen by using the neoplastic hPSC assays
designed in our study.
Thioridazine Reduces LSC Function while Sparing
Normal Human Repopulating Cells
To delineate whether the inhibition of AML-blasts detected
in vitro was due to the compounds affecting the neoplastic
stem cell compartment, xenotransplantation studies (Dick,
(HSCs) were conducted (Figures 5A–5E). Treatment of HSPCs
with salinomycin (1 mM) significantly reduced hematopoietic
engraftment to almost nondetectable levels (Figure S5A)
revealing that this compound interferes with normal hematopoi-
esis from HSPCs and was thus excluded from further evaluation
as it is unlikely to provide the selective anti-CSC therapeutic
targeting desired. In contrast, mefloquine (10 mM) treatment
displayed a slight, yet insignificant, reduction in HSC capacity
relative to controls (Figure 5A). However, mefloquine proved
ineffective in reducing AML LSC capacity and was thus discon-
In contrast to both salinomycin and mefloquine, treatment of
HSPCs with thioridazine (10 mM) displayed the same level of
bone marrow (BM) engraftment (Figure 5A) and splenic engraft-
ment (Figure S5B) as control vehicle-treated cells. Multilineage
reconstitution capacity was identical between control- and thio-
(Figure 5B), erythroid (Figure S5D), and megakaryocytic devel-
opment (Figure S5D) completely unaffected. As measured by
secondary serial transplantation, thioridazine treatment did not
affect HSC self-renewal as compared to self-renewal in
control-treated samples (Figure S5F). Thioridazine treatment
was able to significantly reduce leukemic disease-initiating
AML LSCs (Figures 5C and5D and Figures S5C and S5E). Calcu-
lating the ratio of HSPC normal hematopoietic regeneration
(%hCD45+) to AML leukemogenesis (%CD33+hCD45+ blasts)
revealed that thioridazine significantly reduced LSC function
of thioridazine, no difference in the level of leukemic engraftment
of secondary transplant recipients was observed (Figure S5G).
This suggests that continued exposure to this drug is necessary
to inhibit leukemogenesis in secondary recipients. These data
demonstrate that thioridazine selectively targets somatic CSCs
while having no effect on normal SC properties in vivo.
Dopamine Receptor Expression Provides a Potential
Target of Human CSCs
Thioridazine is known to act through the dopamine receptors
(DR1-5) (Seeman and Lee, 1975). To assess whether the mech-
anism of thioridazine action to selectively interfere with human
CSCs versus normal SCs is via DR antagonism, we analyzed
DR cell-surface expression. To date, five DRs have been identi-
fied and divided into D1-family (D1 and D5) and D2-family (D2,
D3, and D4) receptors (Sibley and Monsma, 1992). Normal
hPSCs expressing the pluripotent marker SSEA3 were devoid
of DR expression (Figure 6A and Figures S6A and S6B). In
contrast, neoplastic hPSCs expressed all five DRs (Figure 6B).
The observed differential expression of DRs and the selective
inhibition of thioridazine for neoplastic hPSCs suggest that inhi-
bition of DR signaling may play a role in selective targeting of
human CSCs versus normal SCs.
To expand the potential role of DRs in CSCs based on the
functional role of thioridazine treatment, we examined whether
DR antagonism could account for the loss of LSC function
following thioridazine treatment. Expression of DR1-5 was
analyzed in HSPCs (Figure 6C) and human hematopoietic mono-
nuclear cells from normal CB (Figures S6C–S6F) and AML
patient samples (Figure 6D and Figure S6G). DRs were not
observed in the primitive HSCs or progenitor populations of
CB (identified as the CD34+38? or CD34+38+ fractions, respec-
tively [Bhatia et al., 1997]) (Figure 6C) indicating that HSCs and
progenitors do not express the targets for thioridazine. Similarly,
Cell 149, 1284–1297, June 8, 2012 ª2012 Elsevier Inc. 1289
Figure 3. The Effect of Salinomycin, Mefloquine, and Thioridazine on norMal and Neoplastic Populations
(A and B) Flow cytometry analysis of frequency of Oct4+ cells in (A) H9 and (B) v1H9-Oct4-GFP cells treated with salinomycin (SAL), mefloquine (MQ), and
thioridazine (THIO) at 10?7? 10?5M. Each bar represents n = 3; mean ± SD. Values are normalized to DMSO-treated control samples; (–) DMSO mean, (?) mean
minus one SD, (?) level of %Oct4+ in BMP4-treated samples.
1290 Cell 149, 1284–1297, June 8, 2012 ª2012 Elsevier Inc.
megakaryocytic (Figure S6C), and lymphoid cells (Figure S6D).
Only monocytes defined as CD14+ and approximately half the
population of granulocytes defined as CD15+ expressed DRs
(Figures S6E and S6F). All of the 13 AML patient samples
analyzed contained a population of DR+ blasts with varying
levels of all five receptors (Figure 6D) and were predominately
detected in CD34+/CD14+ cells (Figure S6G). However, unlike
normal HSCs, CD34+ cells do not correlate with LSC capacity
in human AML (Taussig et al., 2008). Recently, LSCs have
been identified in numerous subfractions devoid of CD34 or
CD38 (Eppert et al., 2011). Similar to malignant hematopoietic
tissue, somatic CSCs have been identified and validated in
human breast tumors and have a CD44+CD24?/lophenotype
(Al-Hajj et al., 2003). Using primary human breast tumors that
(PR?), and human epidermal receptor 2 (HER2?) that are asso-
ciated with the poorest prognostic outcomes (Dent et al., 2007)
we reveal DR colocalization on the CD44+CD24?/lobreast
CSCs (three patients) (Figures 6E and 6F and Figure S6H). This
finding is consistent with the low levels of DRs found in normal
mammary gland tissue, whereas benign breast tumors show
intermediate levels and breast cancers display high levels of
these receptors (Carlo et al., 1986). Our observations of differen-
tial DR expression between normal and neoplastic patient
samples strongly suggest human CSCs are heterogeneous
and drug targeting should be based on molecular pathways
instead of surrogate phenotypic markers.
We investigated whether the DR expression in AML-blasts
was correlative to incidence of LSCs in AML patients. AML
samples with a large fraction of DRD3+ blasts (Figure 6G) and
DRD5+ blasts (Figure 6H) contain LSCs as they are able to
initiate leukemia in xenotransplantation recipients, unlike AML
patient samples with significantly lower levels of DRs that do
not contain LSCs. Samples from AML patients containing
LSCs have been correlated to poor prognostic outcome while
non-LSC samples demonstrate a good prognosis (Eppert
et al., 2011). High levels of DR expression correlate with poor
prognosis, whereas low levels demonstrate good prognosis
(Figures 6G and 6H) suggesting that DR assessment has prog-
nostic biomarker applications and is less complicated than
complex molecular signatures or LSC readouts in mice for
each AML patient.
Thioridazine Inhibits Human AML and Augments Effects
To better understand the functional role of DRs in human AML,
two AML cell lines (AML-OCI2 and AML-OCI3) derived from
patients that express each DR1-5 (Figure 7A) at elevated levels
were utilized. Both AML lines were treated with thioridazine
and compared to other known DR antagonists, clozapine and
chlorpromazine (Seeman and Lee, 1975). All three reduced the
number of AML cells upon treatment (Figure 7B). To further eval-
uate the specificity of DR targeting on human AML cells, patient
AML samples were divided into DR+ and DR? subfractions by
using fluorescence-activated cell sorting before being treated
with a DMSO vehicle or thioridazine for 24 hr and then assayed
for blast-CFU content. A reduction in blast-CFU generation
was only observed in the DR+ subfraction treated with thiorida-
zine (Figure S7A), whereas no reduction was observed in
the DR? subfraction treated with thioridazine (Figure S7B).
Conversely, the addition of a DR D2-family agonist, 7OH-
DPAT, increased the number of AML cells (Figure 7C). DR D2-
family and D1-family exert opposing actions on intracellular
signaling, leading to differential biological effects (Self et al.,
1996). Treatment with a DR D1-family agonist, SKF38393, re-
sulted in a significant reduction in AML cell number, confirming
that D2-family signaling is necessary for AML cell survival (Fig-
ure 7D). These combined results suggest the mechanism of thio-
due to off-target effects, and identifies an alternative avenue of
CSC targeting via DR signaling. Although thioridazine is able to
bind to DR2 receptors at low nanomolar concentrations (See-
man and Ulpian, 1983), we show that at least 10 mM is required
to generate an anti-CSC effect. This dose is within the clini-
cally-tolerable range required to induce antipsychotic effects in
suggests that thioridazine may activate other undefined targets
that may work in synergy or independently of DR signaling for
overall anti-CSC impact.
Upon establishing thioridazine’s anti-LSC effect at clinically-
tolerable doses (Figure S7C), we investigated whether this
drug could be combined with conventional AML chemotherapy
using cytarabine (AraC). Although AraC is the standard chemo-
therapeutic used in both induction and consolidation therapy
of adult human AML, this treatment poses significant morbidity
(C) Ratio of normalized %Oct4+ cells in H9 per v1H9-Oct-GFP with same compound at the same concentration.
(D and E) Percent of neoplastic hPSC staining positive for (D) p53 and (E) p21 following 24 hr treatment with 10 mM etoposide, 10 mM thioridazine (THIO), BMP4
and DMSO-treated (CTRL) controls. Each bar represents n = 3; mean ± SD. Representative images of etoposide and thioridazine-treated cells included. Arrows
show p53+ and p21+ in etoposide-treated cells versus thioridazine-treated cells.
(F)Differentiation-associatedgeneswith>2-foldincrease followingthioridazine treatmentofneoplastic hPSC.Genesdividedintorespective lineages,endoderm
(ENDO), mesoderm (MESO), germ cell (GERM), neural (NEURO), and trophoblast (TROPH). Each bar represents the mean of two separate experiments.
(G–K)Hematopoietic multilineageand clonogenicpotential inresponse tocompoundtreatment detected by usingmethycelluloseassays.Representative colony
forming unit (CFU) pellets of (G) hematopoietic stem and progenitor cells (HSPC) versus (H) AML-blast CFUs pellets following compound treatment. (I and J)
Quantification of respective CFUs and blast-CFUs generated from (I) HSPC and (J) AML-blast cells following compound treatment. Values were normalized to
DMSO-treated control samples; (–) DMSO mean, (–) mean minus one SEM. Each HSPC bar represents n = 7 individual samples, mean ± SEM. Each AML bar
represents at least n = 5 individual patient samples, mean ± SEM. (K) Ratio of normalized HSPC CFUs per AML-blast CFUs with same compound at the same
(L)Frequency of normalized CD11b granulocytic cellsincultured patientAMLcells treatedwiththioridazine 10mM(THIO10mM)or DMSOvehicle(CTRL) forup to
96 hr. Each bar represents n = 3, mean ± SD (*) p < 0.05, (**) p < 0.01, (***) p < 0.001, (****) p < 0.0001. See also Figure S3.
Cell 149, 1284–1297, June 8, 2012 ª2012 Elsevier Inc. 1291
normal HSPC versus AML-blast detection, at concentrations
R 1 mM, AraC induced complete toxicity of AML CFU blasts;
however, it was equally sufficient at eliminating normal HSPCs
(Figure 7E). Using various doses we identified AraC’s effective
concentration (ECAraC), as defined by the concentration that
reduced AML-blast-CFU while retaining HSPC function, to be
at 100 nM (Figure 7E). However, the combination of thioridazine
at 10 mM with AraC reduced the effective concentration
(ECAraC+Thio) to 1 nM (Figure 7F) representing a 100-fold reduc-
Thioridazine-Like and Antileukemic Agents
(A) XY-scatter plot of GFP mean intensity and
cell counts of extended screen with 2,446
pluripotency (LOP) as defined by reduced mean
GFP intensity and cell count. Thioridazine’s data
point is outlined, along with other selected hits.
Each point mean of n = 3.
(B) Summary of toxic and LOP responses seen
with 2,446 compounds.
(C) Chemical structure of other phenothiazine
compounds; fluphenazine and prochlorperazine,
and antileukemic agents; rapamycin and les-
(D) Representative GFP, Hoechst and merged
microscopic images of v1H9-Oct4-GFP cells
treated with selected hit compounds at 10 mM.
(E) Histogram of GFP intensity of these images.
(F) Dose-response curves of v1H9-Oct4-GFP
treated with candidate compounds and calcula-
tion of EC50. Each point n = 3; mean ± SEM.
See also Figure S4.
4. Extended Screen Identifies
tion in the AraC dosage required. Alterna-
tively, the combination of thioridazine at
10 mM with AraC at 100 nM demonstrates
almost complete elimination of AML-
blast-CFUs while preserving HSPC func-
tion (Figure 7F), suggesting that these
remission and prevent relapse of AML in
patients. Collectively, these data show
the synergistic benefit of combining an
anti-LSC agent (thioridazine) with an anti-
proliferative agent (AraC) currently used
as a single first-line treatment for human
AML and to targeting CSCs in addition
to other cells in the leukemogenic hier-
archy. This combined effect with thiorida-
zine is likely to have significant benefit to
AML patients as it can reduce the severe
cytotoxic effects associated with high-
dose AraC therapy, as illustrated in
Unlike other assays that are limited to
using CSCs alone, murine systems,
cancer cell lines, or cancer initiation driven by overexpression
of a unique oncogene, our platform has the capacity to compare
each candidate compound with normal SC and CSC counter-
parts to capture compounds that selectively target CSCs in the
human. Based on our study, we propose the following model
to depict the role of differentiation-inducing drugs as anti-CSC
therapeutics (Figure 7G). Differentiation-inducers capable of
uniquely acting on CSCs may be capable of antagonizing
unusual pathways usurped for survival of CSCs and tumor
1292 Cell 149, 1284–1297, June 8, 2012 ª2012 Elsevier Inc.
generation that are not expressed by normal SCs in the same
tissue. By identifying and inhibiting these pathways, CSCs can
be made responsive to differentiation and produce differentiated
progeny that enter into normal cellular life cycles (Figure 7G). By
example, Thioridazine, an FDA-approved antipsychotic DR
antagonist of the phenothiazine group, represents one such
candidate compound and pathway that was revealed by using
our human CSC-screening platform.
The identification of thioridazine and its ability to selectively
approachareconsistentwith theideathatgeneexpression anal-
stem cell-like signature that correlates with poor clinical
outcome (Ben-Porath et al., 2008). Complementary to this
notion, neoplastic hPSCs may possess pathways shared with
all somatic CSCs and/or harbor unique lineage-specific CSC
pathways based on their pluripotent nature. Our experimental
findings support these concepts and provide experimental
support foraninvitro CSCdifferentiation model thatuseshPSCs
and is also amendable to automated drug-screening methodol-
ogies and mechanistic studies. For example, by representing
a selective inducer of CSC differentiation, we show that thiorid-
azine exerts its anti-CSC activity via antagonism of D2-family
DRs differentially expressed on neoplastic SCs. Equally impor-
tant, we show that thioridazine does not affect normal human
somatic stemcells—thus satisfying anunmetneed foridentifica-
tion of selective anti-CSC agents that can be used at escalating
doses in the clinic. DR signaling therefore represents a drugable
receptor pathway yet to be described in the context of human
CSCs and is, to the best of our knowledge, the only drugable
receptor pathway specific to CSCs in the human.
CB lin- w/ compound
or w/ DMSO
AML w/ compound
or w/ DMSO
% hCD45+ CD33+
AML CD33 hCD45
Normalized % of
human AML blasts in BM
(% hCD45+ CD33+)
CTRLTHIO 10 MMQ 10 M
p = 0.0127
Normalized % of
human blood cells
in BM (% hCD45)
CTRLTHIO 10 MMQ 10 M
Figure 5. Thioridazine’s Effect on HSC and LSC Engraftment
(A)Frequency ofhumanCD45+cellsinthebonemarrow followingHSPCtreatmentwiththioridazine 10mM(THIO10mM)ormefloquine10mM(MQ10mM).Values
normalized to DMSO-treated HSPC control (CTRL) samples. Total of two HSPC samples evaluated. Mean ± SEM.
(B) Representative flow cytometry plots of side scatter (SSC) versus myeloid (CD33) or lymphoid (CD19) markers within the hCD45+ population.
(C) Frequency of CD45+ CD33+ AML-blast cells in the bone marrow (BM) following treatment of AML with thioridazine 10 mM (THIO 10 mM) or mefloquine 10 mM
(MQ 10 mM). Values normalized to DMSO-treated AML control (CTRL) samples. Total of two AML patient samples evaluated.
(D) Representative flow plots of CD33 versus CD45 in DMSO-treated control (CTRL) populations versus thioridazine treated (THIO 10 mM).
(E) Ratio of normalized percent hCD45 HSPC engraftment per normalized percent CD45 CD33 AML-blast engraftment. (*) p < 0.05.
See also Figure S5.
Cell 149, 1284–1297, June 8, 2012 ª2012 Elsevier Inc. 1293
The anticancer effects of thioridazine and other phenothia-
zines have been reported in lymphoblastic leukemia cell lines
although they are less toxic to normal lymphocytes (Zhelev
et al., 2004). This selective response was attributed to inhibition
of mitochondrial DNA polymerase and decreased ATP produc-
tion. Although mitochondrial DNA polymerase was present
for lymphoblastic leukemic cell lines, there was no evidence
that mitochondrial DNA polymerase was inactive in normal
lymphocytes, casting doubt as to whether thioridazine’s effects
are due to this mechanism and limited to the lymphoid
system. Here, we suggest and demonstrate an alternative
mechanism by using primary human patient AML cells versus
normal HSPCs, which indicates that the selective expression
of DRs on neoplastic stem cells can be targeted by using DR
Several DR antagonists are already in clinical use. Interest-
ingly, schizophrenic patients receiving DR antagonist medica-
tion at doses deemed effective for schizophrenia were reported
to have a reduced incidence of rectum, colon, and prostate
cancer compared to the general population (Dalton et al.,
Figure 6. Dopamine Receptors Expressed on Neoplastic Stem Cells
(A and B) Flow cytometry of (A) normal H9 and (B) neoplastic v1H9-Oct4-GFP cells stained with SSEA3 and all five dopamine receptor (DR) subtypes.
DR expression in the SSEA3+ fraction is shown.
(C) Flow cytometry of lineage-depleted cord blood (HSPC) stained with CD34, CD38, and all five DR subtypes. DR expression is presented in the gated pop-
(D) Flow cytometry of 13 AML patient samples stained for all five DRs along with associated FAB classification.
(E) Colocalization of DRD5 in triple-negative (ER?, PR?, and HER2?) primary human breast tumor stained with CD44 and CD24.
(F) The frequency of triple-negative breast CSC (CD44+CD24?/lo) within the DRD3 and DRD5 population. Each bar is composed of three primary triple-negative
breast tumors; mean ± SEM.
(G and H) Frequency of AML-blast cells (CD33+CD45+) from patient samples which are also positive for (G) DRD3 and (H) DRD5. A total of eight AML patient
samples were assessed for leukemic-initiation potential in xenotransplantation recipients. Leukemic-initiating was defined as human engraftment > 0.1% of
17 mice. Total n = 8 AML samples; mean ± SEM.
See also Figure S6.
1294 Cell 149, 1284–1297, June 8, 2012 ª2012 Elsevier Inc.
2005); this suggests that development of some cancers may be
DR-dependent. These findings are further corroborated by the
lower cancer incidence rates observed in dopaminergic-
deficient Parkinson’s patients (Driver et al., 2007) as Parkinson’s
disease itself can be considered to be functionally akin to
disease-induced DR antagonism. It is conceivable that non-
CSC populations within the tumor may be resistant to anti-
CSC agents, and their elimination must be taken into consider-
ation when formulating drug treatments.
date whether DR upregulation is a more generalizable biomarker
of tissue-specific CSCs (outside of brain tumors), and what
survival benefit is derived from expressing these receptors. We
believe an understanding of the prominence of DR signaling in
human cancer and cancer progression would permit develop-
ment of targeted anti-CSC therapies.
Generation of Neoplastic hPSC EOS-GFP Lines
Neoplastic v1H9 or v2H9 hPSC cells (Werbowetski-Ogilvie et al., 2009) were
transduced with lentivirus bearing the EOS-C3+ or EOS-S4+ vectors provided
by Dr. James Ellis (Hotta et al., 2009). After lentiviral transduction cells were
selected by using Puromycin and subsequently sorted as single cells into
a 96-well plate based on GFP expression by using a FASCAria II (Becton
[7OH-DPAT] ( M)
DR D2-family agonist
Normalized # of cells in
0 0.11100 0.1110
Normalized # of cells
[SKF38393] ( M)
DR D1-family agonist
Normalized # of cells in
Single drug treatmentCombined drug treatment
self-renewaldifferentiation cell death
normal stem cell
Figure 7. Thioridazine Inhibits Dopamine Receptor Signaling in AML
(A) DR expression of AML-OCI2 and AML-OCI3 cell lines.
(B) Cell counts of AML-OCI2 and AML-OCI3 cells treated with three DR antagonist drugs. Values are normalized to DMSO-treated control samples. Each bar
represents n = 3; mean ± SD.
(CandD)Viablecellcounts(7AAD?,Hoechst+)ofsamecelllinestreated with(C)7OH-DPAT,aDRD2-familyagonist, or(D)SKF38393,aDRD1-familyagonist, in
serum-free conditions. Values are normalized to DMSO-treated control samples. Each bar represents n = 3; mean ± SD.
(E and F) Single versus combined drug treatment of AML and HSPC. (E) Single drug treatment of patient AML and HSPC with thioridazine (Thio 10 mM) or
cytarabine (AraC) followed by CFU generation and enumeration. (F) Combined thioridazine and AraC treatment of the same patient samples and CFU generation
and enumeration. The normalized ratio of HSPC:AML CFUs is calculated for each concentration and displayed above the appropriate bar pairs. HSPC bar
represents n = 4, two CB lin? samples; AML bars represent n = 4 AML patient samples; mean ± SEM.
(G) Proposedmodel of normal versus cancerstem cellbehavior toward anticancer differentiationtherapy. (*)p < 0.05, (**) p< 0.01, (***) p < 0.001,(****) p< 0.0001.
See also Figure S7.
Cell 149, 1284–1297, June 8, 2012 ª2012 Elsevier Inc. 1295
Dickinson). Colonies generated from single cell clones were used to establish
the v1H9-Oct4-GFP (EOS-C3+), v2H9-Oct4-GFP (EOS-C3+) and v1H9-Sox2-
GFP (EOS-S4+) lines.
The H9 hESC, v1H9, v1H9-Oct4-GFP, v2H9-Oct4-GFP, v1H9-Sox2-GFP, and
fibroblast-derived iPSCs were cultured as previously described (Chadwick
et al., 2003; Werbowetski-Ogilvie et al., 2009).
Screening with Neoplastic and Normal hPSCs
Chemical screens involved v1H9-Oct4-GFP cells seeded at 5,000 cells per
well in mouse embryonic fibroblast conditioned media (MEFCM) supple-
mented with 8 ng/ml bFGF. Twenty-four hours later the media was exchanged
for MEFCM with compounds at 10 mM and 0.1% DMSO, 0.1% DMSO
(?BMP4) or 100 ng/ml of BMP4 and 0.1% DMSO (+BMP4) for 48 hr then
exchanged with fresh media with compound for a further 24 hr (total
compound treatment time 72 hr) prior to being fixed and prepared for auto-
mated imaging and plate reader analysis. Confluent H9 and fibroblast-derived
iPSC were seeded at 10,000 cells per well in MEFCM supplemented with 8 ng/
ml bFGF. Twenty-four hours later the cells were treated with the same
compound formulation used for v1H9-Oct4-GFP. Fresh MEFCM supple-
mented with compounds was exchanged daily for 5 days. On day 5, hPSCs
were fixed and prepared for automated imaging and plate reader analysis.
Images were acquired at 103 0.45 N.A with an Arrayscan HCS VTI Reader
(Cellomics) by means of epifluorescence illumination and standard filter sets.
See Extended Experimental Procedures for further details.
Image analysis was performed by using custom scripts in Acapella software
(Perkin Elmer). Nuclear objects were segmented from the Hoechst signal.
For neoplastic cell lines, object intensity analysis was performed on GFP-posi-
tive cells only. For normal cell lines, the fraction of Alexa-Fluor-647-positive
cells was quantified. Images and well-level data were stored and analyzed in
a Columbus Database (Perkin Elmer) and further data analysis, compounds
registration and hit identification in ActivityBase (IDBS).
between groups were determined via unpaired two-way or one-way Students’
Supplemental Information includes Extended Experimental Procedures,
seven figures, and three tables and can be found with this article online at
This work was supported by grants from the Canadian Institute of Health
Research, Canadian Cancer Society Research Institute (CCSRI) to M.B, and
foremost by the Ontario Ministry of Economic Development and Innovation’s
Ontario Consortium for Regeneration inducing Therapeutics. M.B. is sup-
ported by the Canadian Chair Program and holds the Canada Research Chair
in human stem cell biology. R.M.R is supported by a CCS-TFF fellowship. The
McMaster University Animal Care Council approved all in vivo procedures and
protocols. We thank Ryan Mitchell, Dr. Tamra Werbowetski-Ogilvie, and
Dr. DongCheng Wu from the Bhatia lab for technical assistance and construc-
tive input; Dr. Mark D. Minden of the Department of Medical Oncology and
Hematology, Princess Margaret Hospital, Toronto, ON, Canada, for providing
AML cell lines (OCI-AML2, OCI-AML3); and members of the Labour and
Delivery clinic at the McMaster Children’s Hospital for supplying cord blood
samples.Wealso thank PeterCmorej and RobinHallet from Dr.John Hassell’s
lab for technical assistance in breast tumor preparation and guidance with
breast tumor biology, respectively; Dr. Eric Brown, Jan Blanchard, and staff
of the McMaster HTS laboratory and Centre for Microbial Chemical Biology
for supplying a portion of the Canadian Compound Collection library;
Dr. Rima Al-war from the Ontario Institute of Cancer Research for supplying
the OICR kinase library; Dr. Ahmed Aman from the Al-war lab for assistance
with HPLC analysis; Dr. Gerry Wright from the Institute for Infectious
Disease Research, McMaster University, and Dr. Geoff Tranmer and Inga
Kireeva from the Wright lab for their assistance with NMR and HRMS analysis.
E.S and R.M.R. designed and performed experiments, analyzed data, and
wrote the paper; S.L, Z.S, J.H.L, J.R, M.M, J.M designed and performed
experiments and analyzed data; A.F and M.G performed experiments;
M.L.M. sorted samples and assisted with flow cytometry analysis; M.T, R.F
and B.L supplied AML samples; A.O.G and J.A.H supplied breast tumor
samples. T.J.C and D.F.R designed experiments, analyzed data, and wrote
the paper. M.B. designed experiments, analyzed data, supervised the project
and wrote the paper.
Received: October 2, 2011
Revised: January 20, 2012
Accepted: March 29, 2012
Published online: May 24, 2012
Al-Hajj, M., Wicha, M.S., Benito-Hernandez, A., Morrison, S.J., and Clarke,
M.F. (2003). Prospective identification of tumorigenic breast cancer cells.
Proc. Natl. Acad. Sci. USA 100, 3983–3988.
Atkinson, J.M., Shelat, A.A., Carcaboso, A.M., Kranenburg, T.A., Arnold, L.A.,
Boulos, N., Wright, K., Johnson, R.A., Poppleton, H., Mohankumar, K.M., et al.
(2011). An integrated in vitro and in vivo high-throughput screen identifies
treatment leads for ependymoma. Cancer Cell 20, 384–399.
Ben-Porath, I., Thomson, M.W., Carey, V.J., Ge, R., Bell, G.W., Regev, A., and
Weinberg, R.A. (2008). An embryonic stem cell-like gene expression signature
in poorly differentiated aggressive human tumors. Nat. Genet. 40, 499–507.
Bhatia, M.,Wang, J.C.,Kapp, U.,Bonnet,D.,andDick,J.E.(1997).Purification
of primitive human hematopoietic cells capable of repopulating immune-
deficient mice. Proc. Natl. Acad. Sci. USA 94, 5320–5325.
Bonnet, D., and Dick, J.E. (1997). Human acute myeloid leukemia is organized
as a hierarchy that originates from a primitive hematopoietic cell. Nat. Med. 3,
Boyer, L.A., Lee, T.I., Cole, M.F., Johnstone, S.E., Levine, S.S., Zucker, J.P.,
Guenther, M.G., Kumar, R.M., Murray, H.L., Jenner, R.G., et al. (2005). Core
transcriptional regulatory circuitry in human embryonic stem cells. Cell 122,
Breitman, T.R., Selonick, S.E., and Collins, S.J. (1980). Induction of differenti-
ation of the human promyelocytic leukemia cell line (HL-60) by retinoic acid.
Proc. Natl. Acad. Sci. USA 77, 2936–2940.
Carlo, R.D., Muccioli, G., Bellussi, G., Portaleone, P., Ghi, P., Racca, S., and
Carlo, F.D. (1986). Steroid, Prolactin, and Dopamine Receptors in Normal
and Pathologic Breast Tissue. Ann. N Y Acad. Sci. 464, 559–562.
Chadwick, K., Wang, L., Li, L., Menendez, P., Murdoch, B., Rouleau, A., and
Bhatia, M. (2003). Cytokines and BMP-4 promote hematopoietic differentia-
tion of human embryonic stem cells. Blood 102, 906–915.
Dalton, S.O., Mellemkjaer, L., Thomassen, L., Mortensen, P.B., and Johansen,
C. (2005). Risk for cancer in a cohort of patients hospitalized for schizophrenia
in Denmark, 1969-1993. Schizophr. Res. 75, 315–324.
Dent, R., Trudeau, M., Pritchard, K.I., Hanna, W.M., Kahn, H.K., Sawka, C.A.,
Lickley, L.A., Rawlinson, E., Sun, P., and Narod, S.A. (2007). Triple-negative
breast cancer: clinical features and patterns of recurrence. Clin. Cancer Res.
Desbordes,S.C., Placantonakis, D.G., Ciro, A.,Socci,N.D.,Lee,G.,Djaballah,
H., and Studer, L. (2008). High-throughput screening assay for the identifica-
tionof compoundsregulatingself-renewaland differentiation inhumanembry-
onic stem cells. Cell Stem Cell 2, 602–612.
1296 Cell 149, 1284–1297, June 8, 2012 ª2012 Elsevier Inc.
Dick, J.E. (2009). Looking ahead in cancer stem cell research. Nat. Biotechnol.
Driver, J.A., Logroscino, G., Buring, J.E., Gaziano, J.M., and Kurth, T. (2007). A
prospective cohort study of cancer incidence following the diagnosis of
Parkinson’s disease. Cancer Epidemiol. Biomarkers Prev. 16, 1260–1265.
Eppert, K., Takenaka, K., Lechman, E.R., Waldron, L., Nilsson, B., van Galen,
P., Metzeler, K.H., Poeppl, A., Ling, V., Beyene, J., et al. (2011). Stem cell gene
expression programs influence clinical outcome inhumanleukemia.Nat. Med.
Estey, E., and Do ¨hner, H. (2006). Acute myeloid leukaemia. Lancet 368, 1894–
Estey, E.H., Thall, P.F., Pierce, S., Cortes, J., Beran, M., Kantarjian, H., Keat-
ing, M.J., Andreeff, M., and Freireich, E. (1999). Randomized phase II study of
fludarabine + cytosine arabinoside + idarubicin +/- all-trans retinoic acid +/-
granulocyte colony-stimulating factor in poor prognosis newly diagnosed
acute myeloid leukemia and myelodysplastic syndrome. Blood 93, 2478–
Guan, Y., Gerhard, B., and Hogge, D.E. (2003). Detection, isolation, and
stimulation of quiescent primitive leukemic progenitor cells from patients
with acute myeloid leukemia (AML). Blood 101, 3142–3149.
Gupta, P.B., Onder, T.T., Jiang, G., Tao, K., Kuperwasser, C., Weinberg, R.A.,
and Lander, E.S. (2009). Identification of selective inhibitors of cancer stem
cells by high-throughput screening. Cell 138, 645–659.
Hotta, A., Cheung, A.Y., Farra, N., Vijayaragavan, K., Se ´guin, C.A., Draper,
J.S., Pasceri, P., Maksakova, I.A., Mager, D.L., Rossant, J., et al. (2009).
Isolation of human iPS cells using EOS lentiviral vectors to select for pluripo-
tency. Nat. Methods 6, 370–376.
Jemal, A., Siegel, R., Xu, J., and Ward, E. (2010). Cancer statistics, 2010.
CA Cancer J. Clin. 60, 277–300.
Lapidot, T., Sirard, C., Vormoor, J., Murdoch, B., Hoang, T., Caceres-Cortes,
J., Minden, M., Paterson, B., Caligiuri, M.A., and Dick, J.E. (1994). A cell
initiating human acute myeloid leukaemia after transplantation into SCID
mice. Nature 367, 645–648.
Nasr, R., Guillemin, M.C., Ferhi, O., Soilihi, H., Peres, L., Berthier, C.,
Rousselot, P., Robledo-Sarmiento, M., Lallemand-Breitenbach, V., Gourmel,
B., et al. (2008). Eradication of acute promyelocytic leukemia-initiating cells
through PML-RARA degradation. Nat. Med. 14, 1333–1342.
Nichols, J., Zevnik, B., Anastassiadis, K., Niwa, H., Klewe-Nebenius, D.,
Chambers, I., Scho ¨ler, H., and Smith, A. (1998). Formation of pluripotent
stem cells in the mammalian embryo depends on the POU transcription factor
Oct4. Cell 95, 379–391.
Niu,C.,Yan,H., Yu,T.,Sun,H.P., Liu,J.X., Li,X.S.,Wu,W., Zhang,F.Q.,Chen,
Y., Zhou, L., et al. (1999). Studies on treatment of acute promyelocytic
leukemia with arsenic trioxide: remission induction, follow-up, and molecular
monitoring in 11 newly diagnosed and 47 relapsed acute promyelocytic
leukemia patients. Blood 94, 3315–3324.
Re ´cher, C., Beyne-Rauzy, O., Demur, C., Chicanne, G., Dos Santos, C., Mas,
kemic activity of rapamycin in acute myeloid leukemia. Blood 105, 2527–2534.
Reya, T., Morrison, S.J., Clarke, M.F., and Weissman, I.L. (2001). Stem cells,
cancer, and cancer stem cells. Nature 414, 105–111.
Sachs, L. (1978). Control of normal cell differentiation and the phenotypic
reversion of malignancy in myeloid leukaemia. Nature 274, 535–539.
Seeman, P., and Lee, T. (1975). Antipsychotic drugs: direct correlation
between clinical potency and presynaptic action on dopamine neurons.
Science 188, 1217–1219.
Seeman, P., and Ulpian, C. (1983). Neuroleptics have identical potencies in
human brain limbic and putamen regions. Eur. J. Pharmacol. 94, 145–148.
Self, D.W., Barnhart, W.J., Lehman, D.A., and Nestler, E.J. (1996). Opposite
modulation of cocaine-seeking behavior by D1- and D2-like dopamine
receptor agonists. Science 271, 1586–1589.
Sibley, D.R., and Monsma, F.J., Jr. (1992). Molecular biology of dopamine
receptors. Trends Pharmacol. Sci. 13, 61–69.
Smith, B.D., Levis, M., Beran, M., Giles, F., Kantarjian, H., Berg, K., Murphy,
K.M., Dauses, T., Allebach, J., and Small, D. (2004). Single-agent CEP-701,
a novel FLT3 inhibitor, shows biologic and clinical activity in patients with
relapsed or refractory acute myeloid leukemia. Blood 103, 3669–3676.
Smith, T.J., Khatcheressian, J., Lyman, G.H., Ozer, H., Armitage, J.O., Bal-
ducci, L., Bennett, C.L., Cantor, S.B., Crawford, J., Cross, S.J., et al. (2006).
2006 update of recommendations for the use of white blood cell growth
factors: an evidence-based clinical practice guideline. J. Clin. Oncol. 24,
Taussig, D.C., Miraki-Moud, F., Anjos-Afonso, F., Pearce, D.J., Allen, K.,
Ridler, C., Lillington, D., Oakervee, H., Cavenagh, J., Agrawal, S.G., et al.
(2008). Anti-CD38 antibody-mediated clearance of human repopulating cells
masks the heterogeneity of leukemia-initiating cells. Blood 112, 568–575.
Wang, Z.Y., and Chen, Z. (2008). Acute promyelocytic leukemia: from highly
fatal to highly curable. Blood 111, 2505–2515.
Werbowetski-Ogilvie, T.E., Bosse ´, M., Stewart, M., Schnerch, A., Ramos-
Mejia, V., Rouleau, A., Wynder, T., Smith, M.J., Dingwall, S., Carter, T., et al.
(2009). Characterization of human embryonic stem cells with features of
neoplastic progression. Nat. Biotechnol. 27, 91–97.
Thomson, J.A. (2002). BMP4 initiates human embryonic stem cell differentia-
tion to trophoblast. Nat. Biotechnol. 20, 1261–1264.
Yilmaz, O.H., Valdez, R., Theisen, B.K., Guo, W., Ferguson, D.O., Wu, H., and
Morrison, S.J. (2006). Pten dependence distinguishes haematopoietic stem
cells from leukaemia-initiating cells. Nature 441, 475–482.
Yoshida, H., Kitamura, K., Tanaka, K., Omura, S., Miyazaki, T., Hachiya, T.,
Ohno, R., and Naoe, T. (1996). Accelerated degradation of PML-retinoic acid
receptor alpha (PML-RARA) oncoprotein by all-trans-retinoic acid in acute
promyelocytic leukemia: possible role of the proteasome pathway. Cancer
Res. 56, 2945–2948.
Zhelev, Z., Ohba, H., Bakalova, R., Hadjimitova, V., Ishikawa, M., Shinohara,
Y., and Baba, Y. (2004). Phenothiazines suppress proliferation and induce
apoptosis in cultured leukemic cells without any influence on the viability of
normal lymphocytes. Phenothiazines and leukemia. Cancer Chemother.
Pharmacol. 53, 267–275.
Zheng, R., Friedman, A.D., and Small, D. (2002). Targeted inhibition of FLT3
overcomes the block to myeloid differentiation in 32Dcl3 cells caused by
expression of FLT3/ITD mutations. Blood 100, 4154–4161.
Cell 149, 1284–1297, June 8, 2012 ª2012 Elsevier Inc. 1297