Ovarian cancer side population defines cells with stem-like characteristics and Mullerian Inhibiting Substance responsiveness
The recent identification of “side population” (SP) cells in a number of unrelated human cancers and their normal tissue sources has renewed interest in the hypothesis that cancers may arise from somatic stem/progenitor cells. The high incidence of recurrence attributable to multidrug resistance and the multiple histologic phenotypes indicative of multipotency suggests a stem cell-like etiology of ovarian cancer. Here we identify and characterize SP cells from two distinct genetically engineered mouse ovarian cancer cell lines. Differential efflux of the DNA-binding dye Hoechst 33342 from these cell lines defined a human breast cancer-resistance protein 1-expressing, verapamil-sensitive SP of candidate cancer stem cells. In vivo, mouse SP cells formed measurable tumors sooner than non-SP (NSP) cells when equal numbers were injected into the dorsal fat pad of nude mice. The presence of Mullerian Inhibiting Substance (MIS) signaling pathway transduction molecules in both SP and NSP mouse cells led us to investigate the efficacy of MIS against these populations in comparison with traditional chemotherapies. MIS inhibited the proliferation of both SP and NSP cells, whereas the lipophilic chemotherapeutic agent doxorubicin more significantly inhibited the NSP cells. Finally, we identified breast cancer-resistance protein 1-expressing verapamil-sensitive SPs in three of four human ovarian cancer cell lines and four of six patient primary ascites cells. In the future, individualized therapy must incorporate analysis of the stem cell-like subpopulation of ovarian cancer cells when designing therapeutic strategies for ovarian cancer patients. • cancer stem cells • breast cancer-resistance protein 1
Ovarian cancer side population defines cells with
stem cell-like characteristics and Mullerian Inhibiting
Paul P. Szotek*, Rafael Pieretti-Vanmarcke*, Peter T. Masiakos*, Daniela M. Dinulescu
, Denise Connolly
, David Dombkowski
, Frederic Preffer
, David T. MacLaughlin*, and Patricia K. Donahoe*
*Pediatric Surgical Research Laboratories, Department of Surgery, and
Flow Cytometry Laboratory, Department of Pathology and Center for Regenerative
Medicine, Massachusetts General Hospital, Harvard Medical School, 185 Cambridge Street, Boston, MA 02114;
Fox Chase Cancer Center, 7701 Burholme
Avenue, Philadelphia, PA 19111;
Department of Pathology, Eugene Braunwald Research Center, Brigham and Women’s Hospital, Harvard Medical School,
221 Longwood Avenue, Room 401a, Boston, MA 02115; and
Department of Medicine, Division of Hematology兾Oncology, Massachusetts General Hospital,
Harvard Medical School, 70 Blossom Street, Boston, MA 02114
Contributed by Patricia K. Donahoe, May 4, 2006
The recent identiﬁcation of ‘‘side population’’ (SP) cells in a number
of unrelated human cancers and their normal tissue sources has
renewed interest in the hypothesis that cancers may arise from
somatic stem兾progenitor cells. The high incidence of recurrence
attributable to multidrug resistance and the multiple histologic
phenotypes indicative of multipotency suggests a stem cell-like
etiology of ovarian cancer. Here we identify and characterize SP
cells from two distinct genetically engineered mouse ovarian
cancer cell lines. Differential efﬂux of the DNA-binding dye
Hoechst 33342 from these cell lines deﬁned a human breast
cancer-resistance protein 1-expressing, verapamil-sensitive SP of
candidate cancer stem cells. In vivo, mouse SP cells formed mea-
surable tumors sooner than non-SP (NSP) cells when equal num-
bers were injected into the dorsal fat pad of nude mice. The
presence of Mullerian Inhibiting Substance (MIS) signaling path-
way transduction molecules in both SP and NSP mouse cells led us
to investigate the efﬁcacy of MIS against these populations in
comparison with traditional chemotherapies. MIS inhibited the
proliferation of both SP and NSP cells, whereas the lipophilic
chemotherapeutic agent doxorubicin more signiﬁcantly inhibited
the NSP cells. Finally, we identiﬁed breast cancer-resistance protein
1-expressing verapamil-sensitive SPs in three of four human ovar-
ian cancer cell lines and four of six patient primary ascites cells. In
the future, individualized therapy must incorporate analysis of the
stem cell-like subpopulation of ovarian cancer cells when designing
therapeutic strategies for ovarian cancer patients.
cancer stem cells 兩 breast cancer-resistance protein 1
ecently, two human primary cancers, leukemia and breast, and
several human cancer cell lines, such as central nervous system,
gastrointestinal tumors, and retinoblastoma, were shown to possess
‘‘side population’’ (SP) cells that have been described as cancer stem
cells (1–5). Cancer stem cells, like somatic stem cells, are thought
to be capable of unlimited self-renewal and proliferation. Multipo-
tent cancer stem cells may explain the histologic heterogeneity
often found in tumors (6–9). In addition, cancer progression and
metastasis may involve tumor stem cell escape from innate somatic
niche regulators. Quiescent somatic stem cells residing in specific
tissue niches until activation by injury or other stimuli have been
described in skin and hair follicles, mammary glands, intestines, and
other organs (10). The evolving evidence that somatic stem cells
contribute to normal tissue repair and regeneration suggests the
potential for multipotent somatic stem cells in the ovary responsible
for regulated surface epithelial repair after ovulatory rupture and
possibly the generation of oocyte nurse cells for folliculogenesis
(11). Ovarian somatic stem cells would be expected to divide
asymmetrically, yielding both a daughter cell that proceeds to
terminal differentiation for epithelial repair and an undifferenti-
ated self-copy. Repeated asymmetric self-renewal sets the stage for
somatic stem cells or their immediate progenitors to accrue mu-
tations over time, which might ultimately lead to their transforma-
tion into cancer stem cells and malignant progression.
Epithelial ovarian cancer, thought to emanate from the surface
epithelium of the ovary (12, 13), consists of various histologic
subtypes of Mullerian origin (serous, mucinous, and endometrioid),
affects ⬎22,000 women in North America per year, and accounts
for ⬎16,000 deaths per year with a projected 5 year mortality rate
exceeding 70% (14). Aggressive surgical cytoreduction followed by
chemotherapy results in complete clinical response in 50–80% of
patients with stage III and IV disease. However, the majority of
patients will relapse and become drug-resistant (14–16). Various
types of membrane-spanning ATP-binding cassette transporters,
such as the multidrug-resistant gene 1 and breast cancer-resistance
protein 1 (BCRP1), contribute to the drug resistance of many
cancers, including ovarian cancer, by pumping lipophilic drugs out
of the cell (17). Within bone marrow, researchers have defined a
subset of verapamil-sensitive BCRP1-expressing cells with the
ability to efflux the lipophilic dye Hoechst 33342. This subset has
been described as the SP (18). The functional and phenotypic
characteristics of ovarian cancer predict a stem cell etiology, and the
availability of tumor cells in ascites permits their study by flow
Here we show that distinct histologic types of genetically engi-
neered mouse ovarian cancer cells (MOVCAR 7 and 4306) have a
proportionately large SP, making them a model to study ovarian
cancer stem cell biology. A similar, albeit very small, SP was also
identified in human ovarian cancer cell lines (IGROV-1, SK-OV3,
and OVCAR-3) and in patient primary ascites cells. We used the
MOVCAR 7 cell line to demonstrate that SP cells can reconstitute
colonies in vitro, form tumors earlier than NSP cells in vivo, and
remain re sponsive to Mullerian Inhibiting Substance (MIS). Our
findings suggest that the SP phenotype may be a marker for ovarian
cancer stem cells and that one of the advantages of MIS may be its
ability to inhibit proliferation of both stem and nonstem cancer cells
as compared with the lipophilic chemotherapeutic doxorubicin,
which more effectively inhibited the NSP. These findings, if cor-
roborated in further studies of human specimens, may provide an
explanation for the ability of transporter substrates such as anthra-
cyclines to cytoreduce but essentially never cure recurrent ovarian
cancer. More importantly, identification of the ovarian cancer stem
Conﬂict of interest statement: No conﬂicts declared.
Abbreviations: BCRP1, human breast cancer-resistance protein 1; Bcrp1, mouse breast
cancer-resistance protein 1; MIS, Mullerian Inhibiting Substance; MISRII, MIS type II recep-
tor; MTT; methylthiazoletetrazolium; NSP, non-SP; SP, side population.
To whom correspondence should be addressed. Email: donahoe.patricia@mgh.
© 2006 by The National Academy of Sciences of the USA
July 25, 2006
no. 30 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0603672103
cell would provide a critical step in advancing the development of
novel therapeutic strategies in the management of this disease.
Identification of SPs in Mouse Ovarian Cancer Cell Lines. To deter-
mine whether mouse ovarian cancer cell lines contain candidate
cancer stem cells, Hoechst 33342 was used to sort for the SP
phenotype. The serous adenocarcinoma-recapitulating MOVCAR
7 and 8 cell lines were developed by using the MIS type II receptor
(MISRII) promoter to drive the SV40 T antigen (19). The endo-
metrioid carcinoma-recapitulating 4306 cell line was developed
from conditional LSL-K-ras
mice in which the
ovarian surface epithelium was infected with adenovirus expressing
Cre recombinase (20). Flow cytometry demonstrated a very high
percentage of Hoechst
SP cells in the MOVCAR 7 and 4306 cell
lines (Fig. 1 A and B), whereas SP was not detected in MOVCAR
8. Verapamil, a BCRP1 inhibitor (18), effectively eliminated the SP
in both MOVCAR 7 and 4306 cells (Fig. 1 C and D). The average
first sort percentage of SP cells was 6.28% (n ⫽ 6) for MOVCAR
7 and 1.83% (n ⫽ 4) for 4306 cells, which is elevated relative to the
SP found in other somatic and malignant sources (3, 18, 21).
Colocalization of Hoechst
and Bcrp1 immunoreactive MOV-
CAR 7 and 4306 cells confirmed the pre sence of SP cells (Fig. 1
E–J). Bcrp1 mRNA was detected by qualitative RT-PCR in SP cells
(data not shown). Thus, MOVCAR 7 and 4306 cells possess SPs
with Hoechst efflux characteristics reminiscent of those defined in
hematopoietic stem cells.
Growth of SP and NSP Cells. Growth characteristics of the SP
and NSP cells were consistent with previous findings for cancer
stem cells (21). MOVCAR 7 and 4306 cells were sorted by flow
cytometry and equal numbers of SP and NSP cells cultured. SP cells
from both cell lines formed characteristic compact circular colonies
with a cobblestone appearance and survived numerous passages
(Fig. 2 A and C; n ⫽ 9). NSP cells from both cell lines were sparse
and failed to proliferate beyond 1–2 weeks (Fig. 2 B and D; n ⫽ 9).
These differences were not a consequence of prolonged Hoechst
retention in the NSP cells because propidium iodide was used to
gate out all nonviable cells. Serial sorting and reanalysis (total
passages ⫽ 3) of SP cells demonstrated enrichment of the SP and
the presence of NSP cells (Fig. 2 E–J), suggesting asymmetric
division occurred during culture (Fig. 2 F, G, I, and J). Thus, SP cells
are able to self-renew, be enriched, and produce NSP cells when
recovered and serially sorted in culture.
SP Cells Are in G
Cell Cycle Arrest and Resistant to Doxorubicin
. By definition, SP cells should express high levels of BCRP1
and thus be able to efflux the lipophilic dye Hoechst 33342 and
some lipophilic anticancer drugs, including those used in the
treatment of ovarian cancer (18, 22). The lipophilic anticancer drug
doxorubicin is a substrate of the BCRP1 transporter, whereas the
Fig. 1. Identiﬁcation of SP cells in established mouse ovarian cancer cell lines.
MOVCAR 7 and 4306 cell lines were labeled with Hoechst 33342 dye and
analyzed by ﬂow cytometry before (A and B) and after (C and D) treatment
with verapamil. MOVCAR 7 and 4306 cells were examined for colocalization of
Brcp1 immunoreactivity and Hoechst dye uptake. Hoechst
cells (E and F;
dashed circle and arrows) show Bcrp1 immunoreactivity (G and H; arrows).
cells colocalize with Bcrp1-positive cells (I and J; arrows).
Fig. 2. Growth characteristics of mouse SP cells. MOVCAR 7 and 4306 SP (A and C) and NSP (B and D) cells recovered in culture and photographed with an inverted
⫻10 phase-contrast microscope. SP cells from both cell lines form tight colonies after 4 days in culture, whereas NSP cells are scattered and do not proliferate.
MOVCAR 7- and 4306-sorted SP cells (E and H) were cultured for 7–10 days, resorted by ﬂow cytometry (F and I), recovered for an additional 7–10 days, and then
reanalyzed by ﬂow cytometry (G and J). Each successive sort demonstrated the enrichment of SP cells and the presence of NSP cells.
Szotek et al. PNAS
July 25, 2006
lipophilic microtubule inhibitor paclitaxel is not (23). To investigate
the functional significance of the Bcrp1 transporter found in
MOVCAR 7 SP cells, we tested their response to doxorubicin and
paclitaxel, as compared with that observed in the NSP, by meth-
ylthiazoletetrazolium (MTT) proliferation assays (Fig. 3).
MOVCAR 7 SP cells were only inhibited by 30% after treatment
with doxorubicin, whereas the NSP cells showed 81% inhibition
(Fig. 3A). In contrast to doxorubicin, MOVCAR 7 SP and NSP cells
were almost equally inhibited by the Bcrp1-resistant paclitaxel (Fig.
3B;SP⫽ 85% inhibition; NSP ⫽ 83% inhibition).
Quiescence is one of the defin ing characteristics of somatic
stem cells (24). Cell cycle analysis of three sorted populations,
NSP, and Hoechst
NSP (Fig. 3C),
revealed that the Hoechst
NSP cells had a
higher percentage of cells in S phase (Fig. 3 D and E), compared
SP (Fig. 3F). In contrast, Hoechst
demonstrate a predominance of cells in the G
phase (Fig. 3F)
compared with the Hoechst
NSP cells (Fig. 3D).
Growth Characteristics of MOVCAR 7 SP and NSP Cells. To
assess in vivo tumorgenicity of MOVCAR 7 SPs and NSPs, viable
propidium iodide-negative SP and NSP cells were sorted and
injected into the dorsal fat pad of nude mice (Fig. 4A and Table 1,
which is published as supporting information on the PNAS web
site). Tumors appeared in three of three animals at 10 weeks after
injection of 5.0 ⫻ 10
SP cells, whereas animals injected with an
equal number of NSP cells had no detectable tumors (zero of three)
at that time (Fig. 4B). Tumors appeared in two of three of the NSP
animals only after 14 weeks. Tumors appeared at 7 weeks in animals
injected with 7.5 ⫻ 10
SP cells, whereas NSP-injected animals had
no detectable tumors (zero of three) at that time and only appeared
after 10 weeks in two of three animals (Table 1). To investigate
whether the appearance of tumors in the NSP could possibly be
explained by incomplete sorting, we reanalyzed the sorted popu-
lations by using identical gating and found 82.6% SP cell purity (Fig.
4C; NSP contamination ⫽ 2.63% or ⬇13,150 SP cells in a total of
5 ⫻ 10
cells per animal; 19,750 SP cells in a total of 7.5 ⫻ 10
per animal) and 92.3% NSP cell purity (Fig. 4D; SP contamina-
tion ⫽ 1.72% or ⬇8,600 SP cells in 5 ⫻ 10
cells per animal; 12,900
SP cells in 7.5 ⫻ 10
cells per animal). In addition, the NSP tumors
dissected from animals at euthanization showed verapamil-
sensitive SP cells (Fig. 4 E and F), suggesting that SP cells have the
potential to initiate earlier tumor growth at lower numbers. In a
parallel experiment, preinjection analysis demonstrated an SP
fraction equal to 0.21% (⬇12,600 SP cells per animal) of the 6 ⫻
unsorted cells per animal injected into 50 nude mice (data not
shown). The average time to appearance of these tumors was ⬇9
weeks; in close agreement with our 7.5 ⫻ 10
NSPs (⬇12,900 SP
cells) injected animals and corroborating our speculation that a very
small population of SP cells has the potential to initiate tumor
growth in vivo.
MOVCAR 7 and 4306 SP Cells Respond to MIS
. MIS has been
shown to inhibit MOVCAR 7 both in vitro and in vivo (25). Thus,
we investigated whether MIS inhibits MOVCAR 7 and 4306 SP
and兾or NSP cells in vitro. We first confirmed that SP and NSP cells
possess an intact MIS signal transduction pathway, previously
shown to be required for MIS responsiveness in the embryonic
urogenital ridge (26). By using anti-MISRII antibody we observed
that MOVCAR 7 and 4306 cells express the MISRII receptor by
epifluorescent and confocal microscopy (Fig. 5 B and C; 4306 cells
Fig. 3. SP cells demonstrate decreased inhibition by doxorubicin and G
cycle arrest. MOVCAR 7 cells were sorted for MTT growth-inhibition analysis
against doxorubicin and paclitaxel. SP cells showed 30% inhibition (A)by
, P ⬍ 4.2 ⫻ 10
) and 85% inhibition (B) by paclitaxel (
, P ⬍
6.7 ⫻ 10
) compared with vehicle-treated controls. NSP cells were inhibited
by doxorubicin and paclitaxel by 81% and 88% versus vehicle-treated controls
, P ⬍ 3.2 ⫻ 10
, P ⬍ 5.1 ⫻ 10
(B)]. (A) NSP cells were signiﬁcantly
more inhibited by doxorubicin than by SP cells (81% versus 30% growth
, P ⬍ 1.6 ⫻ 10
). Cell cycle analysis of three populations was
performed as shown in C. Hoechst
NSP and Hoechst
cells (D and E)
demonstrate a predominance of S phase, 69.3% (average ⫽ 45.3%) and 68.9%
(average ⫽ 51.5%), respectively, and decreased G
-arrested cells, 23% (aver-
age ⫽ 53%) and 15.9% (average ⫽ 39%), compared with Hoechst
[P ⬍ 0.0407 (F)]. Hoechst
SPs demonstrate a predominance of G
cells, 63% (average ⫽ 65.8%), and decreased S phase replicating cells, 33.4%
(30.57%). All experiments were performed in triplicate.
Fig. 4. In vivo growth characteristics of MOVCAR 7 SP and NSP cells. MOVCAR
7 cells were sorted for SP and NSP (A), and nude mice were injected with equal
numbers of SP and NSP cells [group I, 5 ⫻ 10
cells per animal (B); group II, 7.5 ⫻
cells per animal (data not shown)]. Measurable tumors were detected in
SP-injected group I animals at 10 weeks (three of three) (B) and SP-injected
group II animals at 7 weeks (Table 1) after implantation, whereas group I (zero
of three) (B) and group II NSP-injected animals did not demonstrate tumors at
the ﬁrst appearance of SP tumors. The appearance of NSP tumors was delayed
in group I and II NSP tumors until 14 and 11 weeks after injection. Sorting
purity analysis (C and D) showed an ⬇93% purity in both SP and NSP sorts,
identifying contamination by NSP sorts with SP cells. NSP tumors harvested
after euthanization revealed the presence of a verapamil-sensitive SP of
similar percentage to that initially injected due to incomplete sorting (E and F).
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0603672103 Szotek et al.
not shown). We then confirmed the presence of MISRII, MISRI
(Alk 2 and 3), and Smad 1兾5兾8 mRNA by RT-PCR in sorted SP
and NSP cells (SP in Fig. 5D; NSP and 4306 cells are the same but
not shown), suggesting these cells would likely respond to MIS.
MOVCAR 7 and 4306 SP and NSP cells were sorted, incubated
for 24 h, and treated with 10
g兾ml MIS for MTT proliferation
assays. MOVCAR 7 SP and NSP cells responded to MIS after
initial sorting of the neat population. MOVCAR 7 SP cells were
inhibited by 86%, whereas NSP cells were inhibited by 93%
compared with vehicle controls (Fig. 5 E). In contrast, only 4306 SP
cells showed a significant inhibition of 37% by MIS (Fig. 7, which
is published as supporting information on the PNAS web site)
However, because NSP cells could not reliably be maintained in
culture for serial sorting, we evaluated the ability of MIS to inhibit
the SP alone after enrichment in both cell lines. MOVCAR 7 serial
sorting followed by MTT showed 93% inhibition after sort 2 and
94% inhibition after sort 3 (Fig. 5 F and G). Serial sorting of 4306
cells followed by MTT showed 60% inhibition after sort 2 (Fig. 7),
and no inhibition after sort 3 was observed (17% inhibition; P ⫽
0.054). Thus, MIS inhibits MOVCAR 7 and 4306 SP cells in vitro.
Human Ovarian Cancer Cell Lines and Primary Patient Ascites Cells
To determine the prevalence of SP cells in human ovarian
cancer, we evaluated the cell lines OVCAR 3, OVCAR 8, SK-
OV-3, and IGROV-1, as well as ascites from six ovarian cancer
patients (see cell line and patient demographics in Table 2, which
is published as supporting information on the PNAS web site).
Patient ascites cells were obtained directly from the operating
theatre and analyzed within 96 h. We detected verapamil-sensitive
SP cells in IGROV-1 (Fig. 6A), OVCAR 3 (data not shown), and
SK-OV-3 (21), but not in OVCAR 8 (Fig. 6B). Viable human
ascites cells, selected as CD45⫺兾CD31⫺, were found to exhibit
verapamil-sensitive SP cells in four of six patients (Fig. 6 C and D).
Thus, an appreciable number of human ovarian cancer cell lines and
primary ovarian cancer ascites cells possess SP cells.
Mouse and Human Ovarian Cancer Cell Surface Phenotype. To inves-
tigate whether ovarian cancer cells express somatic and cancer stem
cell surface markers, as well as to identify differential expression
between SP and NSP cells, we analyzed mouse and human ovarian
cancer cells by flow cytometry. All mouse and human SP cells were
gated as negative for CD45 (common leukocyte antigen) and CD31
(platelet endothelial cell adhesion molecule 1兾endothelial cells).
Compared with NSP cells, the MOVCAR 7 SP cells were enriched
in number of cells and intensity of expression of c-kit兾CD117 (stem
cell factor receptor), whereas 4306 and human SP and NSP cells did
not express c-kit. MOVCAR 7 SP and NSP cells strongly express the
tumor metastasis marker CD 44 (hyaluronic acid receptor),
Fig. 5. MOVCAR 7 SP cells respond to MIS in
vitro. MOVCAR 7 cells express the MISRII by
epiﬂuorescent and confocal microscopy (A–
C). RT-PCR evaluation of SP cells demon-
strated the presence of an intact MIS signaling
pathway (D, left to right: SMAD 1, SMAD 5,
SMAD 8, MIS type I receptors Alk 2 and Alk 3,
and the MISRII). The proliferation of MOVCAR
7 SP and NSP cells were analyzed after the ﬁrst
sort by MTT assay (E) and demonstrate inhibi-
tion of both SP (86%) and NSP (93%) cells by
MIS versus vehicle. SP cells were serially sorted
two more times, demonstrating that enriched
SP cells remain responsive to MIS (F, 93% in-
hibition; G, 94% inhibition). All experiments
were performed in triplicate.
Fig. 6. Human ovarian cancer cell lines and primary ascites
from patients have SPs and express the BCRP1 transporter.
Human ovarian cancer cell line IGROV-1 had a verapamil-
sensitive SP (A and E), whereas OVCAR-8 did not (B and F).
Human serous adenocarcinoma ascites patients 215 and 216
have small verapamil-sensitive SPs (C, D, G, and H). Immuno-
ﬂuorescent analysis of IGROV-1 and the ascites cells of patients
215 and 216 demonstrate colocalization of BCRP1 with
cells (I, K, and L; white arrows), whereas OVCAR-8
did not express Hoechst
or BCRP1-positive cells (J).
Szotek et al. PNAS
July 25, 2006
whereas 4306 cells and most human ovarian cancer cells do not.
MOVCAR 7 and 4306 SP and NSP cells did not express CD24,
CD34, CD105, CD133, or Sca-1 (Table 3, which is published as
supporting information on the PNAS web site). Human cell lines
and ascites cells showed variable expression of the epithelial cell
marker epithelial-specific antigen兾Ep-CAM (epithelial specific an-
tigen) and CD24 (Table 2 and Fig. 8, which is published as
supporting information on the PNAS web site). These markers,
aside from c-kit in MOVCAR 7, did not add to the consistent SP
phenotype and Bcrp1 immunostaining we have observed in iden-
tifying putative ovarian cancer stem cells in both mouse and human.
The hypothesis that rare ‘‘embryonic rests’’ are responsible for
malignancy was suggested ⬎100 years ago (24), but recent advances
in somatic stem cell identification has rejuvenated the investigation
of this premise (4, 27–29). The unique asymmetric self-renewal
capacities of somatic stem cells make it plausible and probable that
mutations in these cells are perpetuated and over time lead to
malignancy. Like somatic stem cells, cancer stem cells have the
properties of self-renewal, heterologous descendent cells, slow
cell-cycle times, and, unlike somatic stem cells, enriched tumor
formation (8, 24). Here we demonstrate these properties within a
subpopulation of mouse ovarian cancer cells that were isolated by
SP sorting. MOVCAR7 and 4306 SP cells are able to self-renew and
produce heterologous descendent NSP cells in culture, MOVCAR
7 SP cells are predominantly G
cell cycle arrested, and the in vivo
time to appearance of tumors in animals injected with equal
numbers of MOVCAR 7 cells may be shorter in those receiving SP
cells. We speculate that the number of SP cells required to initiate
tumor formation in vivo is likely quite low, as evidenced by the
appearance of tumors in NSP-injected animals at the same time as
animals injected with unsorted cells possessing approximately the
same number of SP cells. Thus, these isolated mouse SP cells posses
the properties ascribed to cancer stem cells and provide a model for
comparison with human ovarian cancer.
Ovarian cancer patients in itially respond well to surgical
c ytoreduction and chemotherapy. Chemotherapy alone can yield
several logs of tumor cytoreduction but seldom a cure. The
majorit y of patients who respond to primary chemotherapy
ultimately develop recurrent, usually dr ug-resist ant, disease that
is conceivably due to the ability of ovarian cancer stem cells to
escape these drugs. BCRP1, other wise known as the ABCG2
transporter, confers the ability to not only define a stem cell-like
Hoechst 33342-excluding SP but, perhaps more import antly, the
dr ug resistance-associated efflux of many lipophilic chemother-
apeutic agents such as mitoxantrone, daunorubicin, doxor ubicin,
indolcarbazole, and others (22). Here we demonstrate that
candidate mouse ovarian cancer stem cells, defined as Hoechst-
ef fluxing, verapamil-sensitive, and BCRP1
SP cells, are more
resistant to doxorubicin, c onfirming these stem cell-like charac-
teristics as a potential mechanism for drug resistance. In addi-
tion, we identified a similar subpopulation of cells in both human
ovarian cancer cell lines and primary human ascites cells that
c ould be defined as Hoechst-ef fluxing, verapamil-sensitive,
SP cells. We propose that these ‘‘markers’’ might be
used to detect and isolate patient primary ovarian cancer stem
cells for further characterization.
We cannot with certainty assert that SP cells are cancer stem
cells; however, a subpopulation of cells found in the mouse SPs
demonstrate some of the properties of cancer stem cells. Regardless
of whether SP cancer cells are truly cancer stem cells or early
progenitor cells, expression of the drug-resistance transporter
BCRP1 or other multidrug-resistance proteins (30–33) may have a
profound impact on selection of individual treatment strategies,
clinical outcome, and the design or selection of the next generation
of chemotherapeutic agents. For example, the fact that MIS could
inhibit human anchorage-independent Mullerian tumors in soft
agarose (34) indicated that MIS might act on cancer stem cell-like
populations and led us to investigate its efficacy against these
candidate cancer stem cells. The evidence that MIS inhibits
MOVCAR 7 SP cells in vitro suggests that MIS has the potential to
function as an effective adjuvant to current ovarian cancer chemo-
therapeutic regimens because of its ability to attack this elusive
subpopulation of cancer cells. However, MIS inhibits MOVCAR 8
and OVCAR 8 (25, 35), indicating that response to MIS is not
dependent on the presence of an SP. Currently, the evaluation of
a wide range of chemotherapeutic and molecular-targeted agents
are tested in nonselected in vitro culture systems or animal xeno-
grafts with efficacy being scored on cell death of what is likely the
dominant, drug-sensitive, and perhaps biologically irrelevant NSP
cells. A clinical testing model of SP cells or even purer subsets within
the SP fraction are predicted to yield a more reliable insight into the
development of effective therapeutic agents. Further work is
needed and underway to more clearly define primary human
ovarian cancer stem cells and their response to MIS in vivo.
Flow Cytometry. Flow cytometry was performed in the Department
of Pathology and Center for Regenerative Medicine Flow Cytom-
etry Laboratory according to their published protocols (36). Mouse
and human ovarian cancer SP sorting and immunophenotyping
were performed as described in Supporting Methods, which is
published as supporting information on the PNAS web site. When
testing SPs for multidrug resistance-like BCRP1 sensitivity, vera-
g兾ml; Sigma) was also added.
For cell cycle analysis, MOVCAR 7 cells were harvested, sorted
, and Hoechst
SP cells, and
fixed with 70% ethanol for 24 h. Cells were washed in PBS, stained
g兾ml propidium iodide and 1 mg兾ml RNase (Type IIA;
Sigma), and collected on a Life Sciences Research flow cytometer
CELLQUEST PRO software (BD Biosciences, Frank-
lin Lakes, NJ).
Cell Lines and Culture. Mouse ovarian cancer cell lines, MOVCAR
7 and 8, were developed by D.C. by using the MISRII promoter
to drive the SV40 T antigen (19). The OVCAR 3 and OVCAR
8 human ovarian cancer cell lines were developed by Thomas
Hamilton (Fox Chase Cancer Center) (37). The 4306 cell line
was developed by D.M.D. from conditional LSL-K-ras
mice after infection of ovarian surface epithelium
with adenovirus expressing Cre recombinase. These mice de-
veloped invasive endometrioid ovarian cancers 7 weeks after
infection, and the 4306 cell line was est ablished from ascites cells
(20). IGROV-1 and SK-OV-3 were obtained from American
Type Culture Collection (ATCC). Cell lines were maintained in
4% female FBS (MIS-free) and DMEM with added
1% pen icillin兾streptomycin, and 1% insulin-transferrin-
selen ium (ITS; GIBCO) at 37°C, 5% CO
, in T175 flasks within
a humidified chamber. All cells rec overed f rom sorting were
grown in the same media.
Human Primary Ascites Cell Isolation. Primary ascites cells were
analyzed from five stage III ovarian cancer patients and one
(AC-01) patient with recurrent ascites, who ranged in age from 54
to 71 years (mean, 62.2 years). The study was approved by the
Human Studies Committee of Massachusetts General Hospital
(Protocol No. FWA0003136), and consent was obtained from each
patient on the Gynecology Oncology Service at the time of
outpatient paracentesis or before surgery. Ascites harvested at
laparotomy or ultrasound-guided paracentesis were placed on ice,
centrifuged to isolate the cellular component, and resuspended in
media. Erythrocytes were lysed, and cells were cultured in RPMI
with 10% female FBS, 1% penicillin兾streptomycin, and 1% fun-
gizome. Cells were analyzed by flow cytometry within 96 h for the
presence of an SP and surface markers.
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0603672103 Szotek et al.
Immunostaining of Cultured Cells. Anit-MISRII rabbit polyclonal
antibodies (153p兾MISRII) were developed for Western blot
analysis in the Pediatric Surgical Research Laboratories (35).
Immunofluorescence was performed on MOVCAR 7 and 4306
cells by using 153p as described (25). Images were obtained by
using either epifluorescent (Nikon Eclipse E400 microscope,
SPOT camera, and
SPOT ADVANCE software) or confocal mi-
crosc opy (Leica TCS NT confocal microscope,
C ONFOCAL soft-
ware Version 2.5 Build 1227, and krypton 568-nm laser; Leica,
For BCRP1 immunostaining, cells were double-labeled in sus-
pension with Hoechst 33342 and BCRP1 antibody as described in
ref. 5 and as detailed in Supporting Methods.
Reverse Transcriptase PCR. Total RNA from MOVCAR 7 and 4306
SP and NSP cells was extracted by using the Qiagen (Valencia, CA)
RNeasy Mini Kit (catalog no. 74104) according to the manufac-
turer’s instructions, and 0.5
g of RNA was reverse transcribed into
cDNA by using Superscript II reverse transcriptase as directed by
the manufacturer (Invitrogen). All RT-PCRs were run for 30 cycles
with an annealing temperature of 57°C and separated on 2%
agarose gels. Mouse PCR primers are as in Table 4, which is
published as supporting information on the PNAS web site.
Growth Inhibition by MIS
. MTT assay was used to assess
proliferation inhibition. MOVCAR 7 and 4306 cells were har-
vested, sorted for SPs and NSPs, and plated in the inner wells of
96-well plates at 1,000 cells per well in 200
l of medium per well.
Twenty-four hours after plating, each set of 10 wells of SP or NSP
cells was treated with 10
g兾ml recombinant human MIS (25), 4 nM
paclitaxel (6 mg兾ml; NovaPlus, Irving, TX), a 4-nM doxorubicin
hydrochloride injection (2 mg兾ml; NovaPlus), or media alone. At
day 5 or 7 of incubation, cell viability was quantified by measuring
mitochondrial activity (38) on an ELISA plate reader at an absor-
bance of 550 nm to generate an OD proportional to the relative
abundance of live cells in a given well.
Growth of MOVCAR 7 SP Cells
. MOVCAR 7 SP and NSP cells
were sorted and injected into T and B cell-deficient 6-week-old
female Swiss nude mice in equal numbers (first experiment, 5.0 ⫻
; second experiment, 7.5 ⫻ 10
) into the dorsal fat pad between
the scapulae. Mice were housed in the Edwin L. Steele Laboratory
for Tumor Biology under American Association for Laboratory
Animal Science guidelines with the approval of the MGH Animal
Care and Use Committee (protocol no. 2005N000384).
Purification of Recombinant Human MIS. The human MIS gene was
transfected into CHO cells, amplified, purified, and maintained in
a dedicated facility in the Pediatric Surgical Research Laboratories
for use in this study as described in ref. 39. MIS levels were
measured by using human MIS-specific ELISA (40). MIS was
purified by a combination of lectin affinity chromatography and
FPLC anion-exchange chromatography (39). The MIS purified by
this method causes regression in the organ culture bioassay for MIS
Statistical Analysis. In MTT assays, MIS-, doxor ubicin-, and
paclit axel-treated and untreated samples were analyzed by using
the univariant two-tailed Student t test, with P ⱕ 0.05 c onferring
st atistical significance. All experiments were performed in
We thank Dr. Rakesh K. Jain for per mitting us to perform the nude mice
experiments in the Edwin L. Steele L aboratory for Tumor Biology; Dr.
Thomas Flotte and Ms. Margaret E. Sher wood for guidance in perfor m-
ing confocal microscopy; Dr. Tyler Jacks for access to the 4306 cell line
developed by D.M.D.; and Drs. Michael Seiden, Jose Teixeira, and David
Scadden for many suggestions and an internal review of the manuscript.
Human tissues were provided through the Ovarian Cancer Tumor Bank
at the Massachusetts General Hospit al (MGH), supported by the
Dana–Farber Harvard Cancer Center Ovarian SPORE (1P50CA105009)
and grants from the OCEAN Foundation at MGH and the Harvard Stem
Cell Institute. P.P.S. is supported by Ruth L. Kirchstein National
Research Service Award T32 Training Grant in Cancer Biology
5T32CA071345-10. D.M.D. is supported by the Burroughs Wellcome
Fund Career Award. D.C. is supported by National Institutes of Health
(NIH) Grants P50 CA083638 and U01 CA084242. P.K.D. and D.T.M.
are supported by NIH Grants CA17393 and HD32112. D.T.M. is also
supported by the Edmund C. Lynch, Jr. Cancer Fund through Dr. Jeff rey
Gelfand. This work was also supported by contributions to the Pediatric
Surgical Research Laboratories from the McBride Family and the W.
Gerald Austen Funds.
1. Al-Hajj, M., Wicha, M. S., Benito-Hernandez, A., Morrison, S. J. & Clarke, M. F.
(2003) P roc. Natl. Acad. Sci. USA 100, 3983–3988.
2. Haraguchi, N., Utsunomiya, T., Inoue, H., Tanak a, F., Mimori, K., Barnard, G. F. &
Mori, M. (2006) Stem Cells (Dayton, Ohio) 24, 506–513.
3. Kondo, T., Setoguchi, T. & Taga, T. (2004) Proc. Natl. Acad. Sci. USA 101, 781–786.
4. Wulf, G. G., Wang, R. Y., Kuehnle, I., Weidner, D., Marini, F., Brenner, M. K.,
Andreeff, M. & Goodell, M. A. (2001) Blood 98, 1166–1173.
5. Seigel, G. M., Campbell, L. M., Narayan, M. & Gonzalez-Fernandez, F. (2005) Mol.
Vis. 11, 729–737.
6. Poste, G. & Greig, R. (1982) Invasion Metastasis 2, 137–176.
7. Poste, G., Tzeng, J., Doll, J., Greig, R., Rieman, D. & Zeidman, I. (1982) Proc. Natl.
Acad. Sci. USA 79, 6574–6578.
8. Reya, T., Morrison, S. J., Clarke, M. F. & Weissman, I. L. (2001) Nature 414, 105–111.
9. Tu, S. M., Lin, S. H. & Logothetis, C. J. (2002) Lancet Oncol. 3, 508–513.
10. Li, L. & Xie, T. (2005) Annu. Rev. Cell Dev. Biol. 21, 605–631.
11. Bukovsky, A., Caudle, M. R., Svetlikova, M. & Upadhyaya, N. B. (2004) Reprod. Biol.
Endocrinol. 2, 20.
12. Scully, R. E. (1970) Hum. Pathol. 1, 73–98.
13. Young, R. H. (2005) Mod. Pathol. 18 2, S3–S17.
14. Jemal, A., Murray, T., Ward, E., Samuels, A., Tiwari, R. C., Ghafoor, A., Feuer, E. J.
& Thun, M. J. (2005) CA Cancer J. Clin. 55, 10–30.
15. Cannistra, S. A. (1993) N. Engl. J. Med. 329, 1550–1559.
16. Parker, S. L., Tong, T., Bolden, S. & Wingo, P. A. (1997) CA Cancer J. Clin. 47, 5–27.
17. Gottesman, M. M., Fojo, T. & Bates, S. E. (2002) Nat. Rev. Cancer 2, 48–58.
18. Goodell, M. A., Brose, K., Paradis, G., Conner, A. S. & Mulligan, R. C. (1996) J. Exp.
Med. 183, 1797–1806.
19. Connolly, D. C., Bao, R., Nik itin, A. Y., Stephens, K. C., Poole, T. W., Hua, X.,
Harris, S. S., Vanderhyden, B. C. & Hamilton, T. C. (2003) Cancer Res. 63, 1389–1397.
20. Dinulescu, D. M., Ince, T. A., Quade, B. J., Shafer, S. A., Crowley, D. & Jacks, T.
(2005) Nat . Med. 11, 63–70.
21. Hirschmann-Jax, C., Foster, A. E., Wulf, G. G., Nuchtern, J. G., Jax, T. W., Gobel,
U., Goodell, M. A. & Brenner, M. K. (2004) Proc. Natl . Acad. Sci. USA 101,
22. Doyle, L. A. & Ross, D. D. (2003) Oncogene 22, 7340–7358.
23. Gottesman, M. M. & Past an, I. (1993) Annu. Rev. Biochem. 62, 385–427.
24. Sell, S. (2004) Crit . Rev. Oncol . Hematol. 51, 1–28.
25. Pieretti-Vanmarcke, R., Donahoe, P. K., Szotek, P., Manganaro, T., Lorenzen, M. K.,
Lorenzen, J., Connolly, D. C., Halpern, E. F. & MacLaughlin, D. T. (2006) Clin.
Cancer Res. 12, 1593–1598.
26. Zhan, Y., Fujino, A., MacLaughlin, D., Manganaro, T., Szotek, P., Arango, N.,
Teixeira, J. & Donahoe, P. (2006) Development (Cambridge, U.K.) 133, 2359–2369.
27. Behbod, F. & Rosen, J. M. (2005) Carcinogenesis 26, 703–711.
28. Gale, R. P. & Butturini, A. (1992) Leukemia 6, Suppl. 1, 80–85.
29. Makino, S. (1956) Ann. N. Y. Acad. Sci. 63, 818–830.
30. Duan, Z., Brak ora, K. A. & Seiden, M. V. (2004) Mol. Cancer Ther. 3, 833–838.
31. Duan, Z., Lamendola, D. E., Yusuf, R. Z., Penson, R. T., Preffer, F. I. & Seiden,
M. V. (2002) Anticancer Res. 22, 1933–1941.
32. Penson, R. T., Oliva, E., Skates, S. J., Glyptis, T., Fuller, A. F., Jr., Goodman, A. &
Seiden, M. V. (2004) Gynecol. Oncol. 93, 98–106.
33. Seiden, M. V., Swenerton, K. D., Matulonis, U., Campos, S., Rose, P., Batist, G., Ette,
E., Garg, V., Fuller, A., Harding, M. W. & Charpentier, D. (2002) Gynecol. Oncol.
34. Fuller, A. F., Jr., Krane, I. M., Budzik, G. P. & Donahoe, P. K. (1985) Gynecol. Oncol.
35. Ha, T. U., Segev, D. L., Barbie, D., Masiakos, P. T., Tran, T. T., Dombkowsk i, D.,
Glander, M., Clarke, T. R., Lorenzo, H. K., Donahoe, P. K. & Maheswaran, S. (2000)
J. Biol . Chem. 275, 37101–37109.
36. Preffer, F. I., Dombkowsk i, D., Sykes, M., Scadden, D. & Yang, Y. G. (2002) Stem
Cells (Dayton, Ohio) 20, 417– 427.
37. Schilder, R. J., Hall, L., Monks, A., Handel, L. M., Fornace, A. J., Jr., Ozols, R. F.,
Fojo, A. T. & Hamilton, T. C. (1990) Int. J. Cancer 45, 416–422.
38. Denizot, F. & L ang, R. (1986) J. Immunol. Methods 89, 271–277.
39. Lorenzo, H. K., Teixeira, J., Pahlavan, N., L aurich, V. M., Donahoe, P. K. &
MacLaughlin, D. T. (2002) J. Chromatogr. B Biomed. Sci. Appl. 766, 89–98.
40. Hudson, P. L., Dougas, I., Donahoe, P. K., Cate, R. L., Epstein, J., Pepinsky, R. B.
& MacL aughlin, D. T. (1990) J. Clin. Endocrinol. Metab. 70, 16–22.
41. Donahoe, P. K., Ito, Y., Marfatia, S. & Hendren, W. H. (1976) Biol. Reprod. 15,
42. Donahoe, P. K., Ito, Y. & Hendren, W. H. (1977) J. Sur g. Res. 23, 141–148.
Szotek et al. PNAS
July 25, 2006