Stem cells in breast tumours: Are they ready for the clinic?
Matthew P. Ablett, Jagdeep K. Singh, Robert B. Clarke
Breast Biology Group, School of Cancer and Enabling Sciences, University of Manchester, Wilmslow Road, Manchester M20 4BX, UK
Available online 26 April 2012
KEYWORDS
Breast cancer
Stem cells
Therapy resistance
Cancer stem cells
Biomarkers
Abstract The concept of stem-like cells in cancer has been gaining currency over the last
decade or so since evidence for stem cell activity in human leukaemia and solid tumours,
including breast cancer, was rst published. The evidence established that sub-populations
of cells identied by antibodies to cell surface markers behaved like developmental stem cells
in their capacity to re-grow the human tumour for several generations in experimental
immune-decient hosts. The experiments established that cells with tumourigenic capacity
expressed cancer stem cell (CSC) markers and that activity could also be measured by
self-renewal of tumour sphere colonies in culture. In breast and other cancers, there is good
evidence that CSCs are relatively resistant to radio- and chemotherapy indicating that novel
CSC-targeted therapies are needed. Several pathways are promising targets in breast CSCs.
There are several ways of combating CSC activity including inducing their apoptosis, inhibit-
ing stem cell self-renewal to either stop their division or to promote their differentiation, or
targeting the CSC niche that supports them. The rst challenge for developing novel CSC
therapies is to ascertain which of these CSC properties is being targeted. The second challenge
is to determine suitable CSC biomarkers to measure the efcacy of the novel CSC therapies.
We propose using biomarkers as a means to identify and assess CSC activity in clinical trials.
This is likely to be demanding but feasible in the near future. Thus, we asked if CSCs are ready
for the clinic, however, the emerging question becomes: is the clinic ready for cancer stem
cells?
2012 Elsevier Ltd. All rights reserved.
1. Introduction
Over the last 8 years, there has been increasing evi-
dence for the existence of a cellular hierarchy in many
tumour types. These data have emerged from the inves-
tigation of a developmental biological paradigm of
tumour behaviour, which proposes that tumour cells,
like normal tissue cells, are organised such that stem
cells present in low numbers produce numerous dieren-
tiated cells which make up the bulk of the cancer. Nor-
mal breast epithelial tissue consists mainly of either
luminal cells destined to produce milk or basal myoepi-
thelial cells that have a contractile function during lacta-
tion. In tumours, dierentiation is patently aberrant
compared to the normal tissue with very few human
tumours containing myoepithelial dierentiation and
0959-8049/$ - see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ejca.2012.03.019
Corresponding author: Tel.: +44 161 446 3210; fax: +44 161 446
3109.
E-mail address: robert.clarke@manchester.ac.uk (R.B. Clarke).
European Journal of Cancer (2012) 48, 21042116
Avai l abl e at www. s ci e nce di r ect . c om
j our nal homepage: www. ej conl i ne. com
only partial luminal dierentiation (no milk expressed).
On the other hand, an infrequent sub-population of
human breast tumour cells expressing surface markers
similar to normal breast epithelial stem cells can be iden-
tied and enriched by ow cytometry. Importantly, their
transplantation into immune-compromised mouse mod-
els has demonstrated that they are capable of growing
human breast tumours. These ndings have several
implications for the understanding of human breast can-
cer biology and a major impact on treatment paradigms
in both curative and palliative settings.
2. Evidence for the identication of stem-like cells in
human breast cancers
The cancer stem cell hypothesis posits that cancers
are maintained and re-populated by stem-like cells
within the tumour, termed cancer stem cells (CSCs). A
widely accepted denition of a CSC is . . .a cell within
a tumour that possesses the capacity to self-renew and
to cause the heterogeneous lineages of cancer cells that
comprise the tumour.
1
The rst strong in vivo evidence in support of the CSC
concept came from classical implantation studies in
human leukaemia by Bonnet and Dick.
2
They used uo-
rescence-activated cell sorting (FACS) to isolate a spe-
cic population of cells from acute myeloid leukaemia
(AML) patients which were able to initiate AML follow-
ing implantation into non-obese diabetic mice with
severe combined immunodeciency (NOD/SCID). The
leukaemia-initiating cells were dened by expression of
the cell surface antigen CD34 and displayed self-
renewal, dierentiative and proliferative capacities simi-
lar to normal haematopoietic stem cells.
The rst evidence for the existence of CSCs in solid
human tumours came from studies in breast cancer.
3
Al-Hajj and colleagues prospectively isolated a tumouri-
genic population of cells from primary human breast
cancers using FACS based on the phenotype ESA
+
/
CD44
+
/CD24
/low
/lineage
-
. When injected into the
mammary fat pads of NOD/SCID mice, as few as 200
cells with this phenotype consistently formed tumours
whereas 20,000 CD44
+
/CD24
+
cells failed to form
tumours and 10,000 unsorted cells only formed tumours
in 25% of cases. Enhanced tumour forming capacity of
ESA
+
/CD44
+
/CD24
/low
/lineage
-
cells has subse-
quently been reported by many other groups,
4,5
however
purication of a single breast CSC able to initiate a
tumour is lacking. Some tumour biologists argue that
the CSC hypothesis is too simplistic and propose a more
complex model of cancer development, termed clonal
evolution.
6
The clonal evolution model proposes that tumours
arise from an aberrant normal cell clone which prolifer-
ates uncontrollably due to accumulation of genetic
mutations. The progeny of this clone would accumulate
further mutations, hence the heterogeneous cellular nat-
ure of tumours.
7,8
Critics of the CSC hypothesis claim
that the current gold standard method for assessing
CSCs by injecting cells into immuno-compromised mice
may introduce a selection bias for human cells capable
of surviving and proliferating in the mouse microenvi-
ronment with foreign growth factors and cytokines.
9,10
Recently it has been proposed that a merging of the clo-
nal evolution model and the CSC hypothesis may
explain more fully how cancer is maintained and pro-
gresses. This merged model predicts that the frequency
of CSCs in each patient will vary dramatically and be
dependent on the type of breast cancer and the domi-
nant mutations, gene amplications and deletions. Fur-
thermore, dominant clonal CSCs could emerge during
tumour progression as resistant CSCs are preferentially
selected for by current therapy. It is therefore important
to develop reliable methods to identify and isolate breast
CSCs in the clinical setting.
3. How do we identify breast cancer stem cells and
measure their activity?
The current gold standard method of assessing
breast CSC activity is the ability of these cells to re-grow
tumours following injection into immuno-compromised
mice. A successful enrichment of a CSC population is
demonstrated by signicant improvement in tumourige-
nicity compared to unsorted cells of the original popula-
tion. One could argue that this approach assays for stem
cell activity or a state of stemness since a CSC entity
has not been successfully isolated at the single cell level.
Two methods currently used to enrich for breast CSC
activity include FACS using antibodies to specic cell
surface markers or intracellular enzymes such as alde-
hyde dehydrogenase (ALDH), using the ALDEFLUOR
assay, and the mammosphere assay. The mammosphere
assay is a non-adherent in vitro colony forming assay
originally developed by Dontu et al.
11
Under these
non-adherent culture conditions dierentiated cells
undergo cell death (anoikis) whereas cells with self-
renewal capacity are able to survive and proliferate
forming individual colonies termed mammospheres.
Both the mammosphere assay and tumour forming
capacity in vivo may measure stem cell activity; by de-
nition, they do not measure dormant stem cells. This is
an important distinction that researchers need to
acknowledge when studying CSCs, because lack of
detectable CSC activity does not necessarily equate to
eradication of the CSC population. This has important
clinical implications because dormant CSCs could
potentially become activated at a later date thereby lead-
ing to disease recurrence.
To date, there is no universal cell surface antigen, or
combination of antigens, for the purication of breast
CSCs by antibody techniques. This may be because of
M.P. Ablett et al. / European Journal of Cancer 48 (2012) 21042116 2105
the heterogeneous nature of breast cancers and the exis-
tence of dierent populations of breast CSCs within a
single tumour. The most common combination of cell
surface markers used to enrich for breast CSCs is
ESA
+
/CD44
+
/CD24
/low
. First established by Al-Hajj
et al.,
3
this combination of markers has been utilised
by many groups to enrich for breast CSCs from primary
breast carcinomas as well as breast cancer cell lines.
3,1216
Other markers used to identify breast CSCs (see Table 1)
demonstrate the high variability in enriching for breast
CSCs, even when the same markers are used. There is
also considerable variability between cell lines and
primary samples, which could be due to many factors,
including: the use of dierent uorescently-conjugated
antibodies, alternative gating strategies and selection
bias when establishing cell lines and primary samples
using standard adherent culture conditions.
17
The mammosphere assay has also been used to enrich
for breast CSCs. Ponti et al. (2005) showed that cells
within a mammosphere may be relatively undierenti-
ated and more stem-like since they did not express spe-
cic markers of breast lineages such as CK14 and
CK18 for myoepithelial cells or ESA for luminal cells.
12
Harrison et al. (2010) harvested anoikis-resistant cells
at early time points during the mammosphere assay
and demonstrated CSC enrichment in vitro, as quantied
by mammosphere formation and increased tumour-initi-
ating capacity in vivo.
18
Immunohistochemistry (IHC) using similar markers
to those used for FACS has also been used to identify
breast CSCs.
16
However, this approach is problematic
as it cannot establish the function of cells expressing
CSC markers. The advantage of FACS over IHC is that
sorted cells remain viable and can be used in functional
assays to demonstrate the presence of CSCs and to test
ecacy of targeted drugs. The disadvantage of func-
tional CSC assays is the large sample of fresh cells
required, whereas IHC requires only a small xed sam-
ple and can readily be carried out in the clinical setting.
These factors must be considered when attempting to
transfer CSC assays into the clinic. Although methods
of identifying breast CSCs are improving, measuring
the success of a CSC-directed treatment in patients is
challenging.
4. Are breast CSCs resistant to current anticancer
treatments?
Breast cancer management involves a combination of
surgery, chemotherapy and radiotherapy with or with-
out hormonal therapy to control loco-regional and met-
astatic disease. Although chemotherapy is able to reduce
tumour bulk and confers a survival benet, the majority
of patients with metastatic disease and a quarter of all
patients with early disease will eventually relapse despite
an initial response to treatment.
19
Similarly although
radiotherapy reduces local recurrence rate, a proportion
of patients will relapse. Local recurrence and distant
metastases, despite therapy, indicate that breast CSCs
are able to evade the eects of chemotherapy and radio-
therapy and thereby repopulate the tumour following
treatment. Laboratory and clinical evidence that sup-
ports this conclusion are described below and summa-
rised in Table 2.
4.1. Chemo-resistance
Using three breast cancer cell lines (SUM159, SUM1315
and MDA-MB-231), tumour initiating CD44
+
/CD24
/
ESA
+
cells were found to be enriched up to 30-fold follow-
ing a 6 day treatment in vitro with Paclitaxel (taxol) or 5-
urouracil compared with control.
5
Ginestier et al. (2010)
found that docetaxel treatment increased the proportion
of ALDEFLUOR
+
CSCs in xenograft mouse models of
human breast cancer compared to control mice.
20
Enrichment of breast CSCs has been further demon-
strated in numerous clinical studies investigating the
eects of chemotherapy. Using the mammosphere assay,
Yu et al. (2007) showed that mammosphere formation
of cells derived from 5 patients treated with neoadjuvant
Table 1
The frequency of breast cancer stem cells isolated from cell lines and primary breast tissue using either uorescence-activated cell sorting (FACS) or
the mammosphere assay. There is a large disparity in the percentage of breast cancer stem cell (CSCs) isolated when dierent cell surface markers
are used. To note, this variation is also observed when dierent laboratories use the same markers.
Isolation technique % Breast cancer stem cells No. of sorted cells required for growth in vivo Reference
Cell lines Primary
FACS
Aldeuor+ 0100% 8% 5 10
3
20,90
CD44
+
/CD24
cells by
3-fold and signicantly increased the number of cells
able to generate mammospheres 4-fold. By establishing
human breast cancer xenografts from patients before
and after treatment they showed that chemotherapy
doubled tumour forming capacity in SCID/Beige mice
suggesting an enrichment of CSCs. Using a gene expres-
sion signature common to both CD44
+
/CD24
/low
and
mammosphere forming cells, Creighton et al. (2009)
showed that these signatures were more pronounced in
tumour tissue from 12 patients following neoadjuvant
treatment with docetaxel.
23
Together these studies sug-
gest that breast CSCs are relatively chemo-resistant,
and such therapy can enrich for this cell population.
Potential chemo-resistance mechanisms in CSCs include
increased expression of antiapoptotic proteins, increased
drug eux transporters and increased eciency of DNA
repair.
24
Furthermore, it is possible that CSCs are less
sensitive to antimitotic agents due to a low rate of pro-
liferation. Chemotherapy may thus destroy non-CSCs in
a proportion of tumours, leaving the CSCs to poten-
tially re-seed disease.
4.2. Radio-resistance
Philips et al. (2006) demonstrated that mammo-
spheres derived from MCF-7 and MDA-MB-231 cell
lines were more radio-resistant than cells grown as
monolayer cultures.
13
More importantly they demon-
strated that fractionated doses of irradiation increased
the proportion of tumour-initiating CD44
+
/CD24
/low
cells in the non-adherent cell populations of MCF-7
monolayer cultures. The authors proposed that reduced
induction of reactive oxygen species (ROS), reduced
double strand DNA breaks and induction of the
Notch-1 pathway were responsible for the observed rel-
ative radio-resistance of breast CSCs and greater
tumour re-growth during radiotherapy treatment gaps.
Similarly Diehn et al. (2009) proposed that CSCs
Table 2
Summary of preclinical, animal model and clinical evidence of breast cancer stem cell (CSC) resistance to chemotherapy, radiotherapy and
endocrine therapy.
CSC isolation technique Source Type of treatment Enrichment or preferential
survival
Reference
Chemo-
resistance
ESA
+
CD44
+
/CD24
low
Cell lines 5-Flurouracil or Paclitaxel In vitro FACS 530-fold " 5
CD44
+
/CD24
-
Cell lines Doxorubicin-selected MCF7s In vitro FACS 30% " and "
tumours in vivo
103
CD44
+
/CD24
mammosphere
assay
Clinical
samples
Neoadjuvant 5-urouracil,
epirubicin & cyclophosphamide
In vivo FACS 9.5-fold "
In vivo MS " 0.5% to 5.9%
21
CD44
+
/CD24
mammosphere
assay
Clinical
samples
Neoadjuvant docetaxel or
doxorubicin & cyclophosphamide
In vivo FACS " 5% to 14%
In vitro MS " 5-fold
22
CD44
+
/CD24
/mammosphere
gene expression signature
Clinical
samples
Neoadjuvant docetaxel In vivo " gene signature
following chemotherapy
23
Radio-
resistance
Mammosphere assay Cell lines Radiation single & fractionated In vitro clonogenic assay 2-
fold " in survival
13
CD44
+
/CD24
/low
Cell lines Radiation fractionated In vitro FACS, up to 3-fold
"
104
Sca1+ BALB/c mice Mouse Radiation single dose In vivo FACS 3-fold " 26
CD24
+
Thy1
+
Lin
MMTV-
Wnt1
Mouse Radiation fractionated In vivo FACS, 2-fold " 25
Lin
CD29
+
CD24
+
P53-
nullmouse
Mouse Radiation single dose In vitro clonogenic assay up
to 10-fold "
105
Endocrine
resistance
CK5+ Cell lines Tamoxifen or Fulvestrant In vitro - up to 3.4-fold " in
CK5 protein expression
32
Clinical
samples
Neoadjuvant tamoxifen +/
exemestane
In vivo 2-fold " in CK5
expression
In vivo number of
CK5 + cells/eld " 2.6 to 30.4
CD44
+
/CD24
/Mammosphere
gene expression signature
Clinical
samples
Neoadjuvant letrozol In vivo " gene expression
signature
23
M.P. Ablett et al. / European Journal of Cancer 48 (2012) 21042116 2107
contain lower ROS due to increased expression of free
radical scavenging systems.
25
Depletion of ROS scav-
engers using pharmacological inhibitors decreased clo-
nogenicity and resulted in radiosensitisation.
25
Using the Hoechst dye eux method, Woodward
et al. (2007) found that the side population in hyperplas-
tic tissue from mouse mammary tumour virus (MMTV)
driven Wnt-1 transgenic mice was more than 2-fold
greater than matched controls and could also be selec-
tively enriched following exposure to radiation.
26
Together these studies provide evidence that breast
CSCs are endowed with mammary stem/progenitor cell
properties which confer protection from radiation-
induced cell death. This radiation resistant population
is thereby liable to re-seed tumours during and after
fractionated radiotherapy.
However a recent study suggests that the problem of
radio-resistant CSCs may be overcome by localised
hyperthermia therapy using nanoparticle technology.
27
Using genetic mouse model of breast cancer and human
breast cancer xenograft mouse models, Atkinson et al.
(2010) showed that radio-resistant CSCs could be
radio-sensitised by hyperthermia therapy resulting in a
greater reduction in tumour size and reduction of self-
renewal, measured by serial transplantation studies.
27
Although these ndings are promising, the clinical e-
cacy of combining radiotherapy with nanoparticle
hyperthermia therapy is yet to be determined.
4.3. Endocrine-resistance
Oestrogen deprivation is an important therapeutic
strategy in the management of breast cancers that
express the oestrogen receptor-a (ER-a). However,
despite an initial response to endocrine therapy, 25%
of patients with early breast cancer and all patients with
metastatic breast cancer eventually become resistant to
therapy.
19
In the normal mouse mammary gland the
stem/progenitor cell population is reported to be ER
negative (ER).
28
Using CD24
lo
and CD29
+
to sort
for stem cells Asselin-Labat et al. (2006) demonstrated
that less than 0.01% of these cells expressed ER com-
pared to 37% of luminal cells.
28
Sleeman et al. (2007)
subsequently showed that the greatest in vivo clonogenic
capacity lies within the ER/CD24
lo
cell population.
29
In contrast to the normal mouse and human breast
where there is a dissociation of ER expression and prolif-
eration, actively dividing ER + cells are frequently
observed in breast hyperplasia and breast cancer.
30
The
level of ER expression is predictive of response rate to
endocrine therapy. However, the high rate of relapse
and poor response in ER + metastatic breast cancer
patients suggests that endocrine therapy does not target
the cells driving tumourigenesis. For example, using the
ER+/PR + human breast cancer cell line T47D, Horwitz
et al. (2008) showed that within the tumour-initiating cell
fraction resides a rare population of ERPR
CK5 + cells.
31
In vitro, oestrogen combined with tamox-
ifen or fulvestrant led to a decrease in ER expression, but
an increase in CK5 expression compared with oestrogen
alone, suggesting that anti-oestrogen therapy enriches for
the ER-/CK5 + phenotype.
32
In a clinical study using
paired tissue samples, Kabos et al. (2008) reported that
neoadjuvant endocrine therapy resulted in a decrease in
ER gene expression and an increase in both CK5 gene
expression and the number of CK5 + cells in post-treat-
ment tumours compared to pre-treatment tumours.
32
Thus, an emerging mechanism of resistance to ER-
targeted therapy is the presence of an ER- CSC popula-
tion which can dierentiate to form treatment-sensitive
ER + cells, but which can escape endocrine treatment
and survive to subsequently repopulate the tumour.
Another proposed mechanism of endocrine resistance is
via enhanced growth factor tyrosine kinase signalling path-
ways. An inverse relationship between ER and epidermal
growth factor receptor (EGFR)/HER2 expression has been
demonstrated in cell line models of acquired endocrine
resistance and patients. Using an endocrine resistant
MCF-7 cell line McClelland et al. (2001) demonstrated that
as endocrine resistance develops, EGFR expression
increases in parallel with a decrease in ER expression.
33
Similarly, using tamoxifen-resistant MCF-7 cell lines,
Knowlden et al. (2003) demonstrated an increase in expres-
sion of HER2 compared to wild type MCF-7 cells.
34
In
both studies cell proliferation was reduced by targeting
the EGFR/HER2 signalling pathways. Clinical studies in
patients with recurrent breast cancer have demonstrated
that expression of EGFR/HER2 can be associated with
decreased sensitivity to endocrine therapy and a poorer
prognosis.
35
The acquisition of enhanced EGFR/HER2 signalling
and endocrine resistance could potentially arise from the
selectionof more stem-like cells.
36,37
Using the ALDEFLU-
OR assay, Korkaya et al. (2008) demonstrated that over-
expression of HER2 expanded the proportion of cells with
stem-like properties.
38
Furthermore, using HER2 over-
expressing breast cancer cell lines, Magnico et al. (2009)
showed that mammosphere-forming breast cancer cells
had higher expression of HER2 compared with cells cul-
tured under dierentiating conditions.
39
HER2 over-expression is associated with increased
expression of genes known to be important in regulating
stem cell function and is under the control of Notch
1.
38,39
Over-expression of HER2 in breast CSCs
increased their invasive capacity and tumourgenicity
when transplanted into NOD/SCID mice.
38
Moreover,
inhibition of HER2 signicantly decreased the propor-
tion of breast CSCs, decreased tumour forming capacity
and decreased invasion.
38
Together these studies indi-
cate that breast CSC activity is regulated by HER2
and that this population can be depleted by HER2 inhi-
bition. A pre-surgical window trial by Li et al. (2008)
2108 M.P. Ablett et al. / European Journal of Cancer 48 (2012) 21042116
demonstrated that treatment with lapatinib (a dual
EGFR/HER2 inhibitor) led to a decrease in the
CD44
+
/CD24
lo
CSC fraction and mammosphere form-
ing eciency of the residual tumour.
22
Targeting the
CSC population in HER2 overexpressing breast cancers
may partially explain the therapeutic success of anti-
HER2 therapy using trastuzumab (Herceptin). However
there is also evidence that HER2 inhibition may be ben-
ecial in HER2 negative cancers.
40
A synergistic eect of
chemotherapy and trastuzumab is supported by clinical
studies which have shown a signicant improvement in
3 year disease-free survival of HER2 positive patients
treated with trastuzumab combined with conventional
chemotherapy.
41,42
5. How do we target breast CSCs?
As described above, there is good evidence that most
standard therapies target the bulk of tumour cells reduc-
ing tumour size, but often do not eliminate CSCs which
eventually may lead to tumour recurrence. It is vital to
develop agents which target the CSCs specically. Stan-
dard therapies in combination with CSC-targeted thera-
pies may potentially provide a more eective treatment
strategy by de-bulking the tumour mass and preventing
recurrence. Combination therapy would also overcome
the possibility that tumour cells could de-dierentiate
to a more stem-like phenotype induced by extrinsic fac-
tors, as recently demonstrated in cell lines.
43,44
There are
at least three potential ways to target breast CSCs: (1)
inhibition of self-renewal signalling pathways thereby
inducing dierentiation or apoptosis, (2) targeting resis-
tance mechanisms or (3) targeting of the CSC niche
(Fig. 1bd).
Monoclonal antibodies raised against specic compo-
nents of signalling pathways or cell surface antigens
present on CSCs have been used to target these cells spe-
cically (Fig. 1b). For example, novel therapeutic anti-
bodies targeting the Notch signalling pathway and
stem cell surface antigens EpCAM and CD44 are cur-
rently in clinical trials (Table 3).
5.1. CSC signalling pathways
Targeting specic components of the Notch signal-
ling pathway has proven to be an eective treatment
against breast CSCs in pre-clinical studies both in vitro
and in vivo. The Notch pathway is implicated in mainte-
nance of both normal and malignant stem cell activ-
ity.
45,46
Aberrant activation of this pathway has been
reported in breast cancer, specically in the breast
CSC population.
36
Studies in pre-invasive ductal carci-
noma in situ and invasive breast cancer cell lines and
clinical samples have shown that inhibition of the Notch
signalling pathway using a c-secretase inhibitor, DAPT,
signicantly reduces CSC activity.
18,36
DAPT prevents
the cleavage and nuclear translocation of the intracellu-
lar domain of the Notch receptor, thus abrogating
Notch signalling. The c-secretase inhibitors MK-0752
(Merck) and RO4929097 (Roche) are currently in phase
I/II clinical trials in breast cancer
47
(Table 3).
Aberrant Hedgehog signalling has been demon-
strated in many dierent types of cancer and is another
potential CSC therapeutic target. Hedgehog signalling
regulates stem cell fate and proliferation during develop-
ment as well as in adult mammalian systems.
48
In breast
cell lines, Liu et al. (2006) showed that members of the
Hedgehog signalling pathway (PTCH1, Gli1 and Gli2)
were highly expressed in normal and malignant stem
cells.
49
Expression was down-regulated when the cells
were induced to dierentiate. A current clinical study
combines the Roche Notch inhibitor RO4929097, with
a Hedgehog inhibitor from Genetech, GDC-0049, based
on evidence that both of these pathways play an impor-
tant role in self-renewal and also interact with each
other.
47,50
Some unorthodox treatments for breast cancer have
shown the importance of other signalling pathways in
breast CSC activity. The dietary polyphenols, curcumin
and piperine have been shown to modulate self-renewal
of breast CSCs in vitro using the mammosphere assay.
51
Both of these compounds inhibited Wnt signalling,
which has been shown to be dysregulated in many can-
cers including breast cancer. Previous studies demon-
strated the importance of the NF-kappaB pathway in
leukaemia stem cell survival.
52,53
This work has been
extended to breast CSCs by Zhou et al. (2008), who
found that three dierent inhibitors of the NF-kappaB
pathway preferentially inhibited mammosphere forma-
tion and proliferation of the side population of MCF-
7 cells sorted using FACS.
37
Epidemiological studies indicate that diabetes is asso-
ciated with an increased risk of breast cancer.
54
Metfor-
min is commonly used to treat type 2 diabetes and has
been shown to reduce cancer risk.
55
In vivo, metformin
reduces tumour growth of triple-negative breast cancer
cell lines.
56
When combined with conventional therapy
such as doxorubicin or tratuzumab, metformin appears
to further suppress tumour growth by selectively target-
ing the breast CSCs.
5759
5.2. Dierentiation therapy
Another method to target CSCs is to induce them to
dierentiate, and thus lose their self-renewal potential.
To achieve this, the CSCs must exit their quiescent state
and become actively cycling in order to divide and dier-
entiate. Dierentiation therapy has been used to treat
some haematological cancers successfully. Rexinoids
and retinoids, strong inducers of dierentiation, are
the current standard treatment for some types of leukae-
mia and have been shown to promote dierentiation of
M.P. Ablett et al. / European Journal of Cancer 48 (2012) 21042116 2109
Traditional therapy Traditional therapy
(a)
Loss of
CSC re-grows
Relapse
tumour bulk
tumour
CSC therapies alongside traditional therapy p g py
Loss of CSC and
diff ti t d Tumour shrinkage with
(b)
differentiated
cancer cells
Tumour shrinkage with
traditional therapy
Cure
Targeting CSC resistance Targeting CSC resistance
(c)
Differentiation
Tumour shrinkage with
traditional therapy
Cure
of CSC
traditional therapy
Cure
Targeting CSC self-renewal
L f CSC d
(d)
L f CSC i h
Loss of CSC and
tumour shrinkage with
Cure
Loss of CSC niche
g
traditional therapy
Cure
Targeting CSC niche Cancer Stem Cell Differentiated Tumour Cell Endothelium Fibroblast Collagen Fibres g g g
Fig. 1. The possible eects of dierent cancer stem cell (CSC) therapies on tumour growth. Traditional therapy (a) targets the dierentiated cells
but not does aect all of the CSCs, leading to tumour relapse. Three strategies for CSC therapy are shown: (b) targeting CSC resistance, (c)
targeting CSC self-renewal pathways or (d) targeting components of the CSC niche. Each of these strategies would be predicted to reduce the
capacity for the CSCs to self-renew and repopulate the tumour. Use of these CSC therapies alongside traditional therapies should further reduce
breast cancer recurrence rates.
Table 3
Clinical progress of novel agents currently in development for the treatment of breast cancer. These agents target specic components of pathways
or surface integrins shown to be important in breast cancer progression. All agents listed and their clinical status refers to their use in trials
concerning breast cancer only.
Target Agent Company Clinical status Reference
Notch MK-0752 c-secretase inhibitor Merk Phase I/II 47
RO4929097 c-secretase inhibitor Roche Phase I 47
Dll4 specic Abs N/A Pre-clinical 106
Hedgehog Cyclopamine SMO antagonist N/A Pre-clinical 107
GDC-0449
a
Genetech Phase I 47
Bcl-2 G3139 antisense oligodeoxynucleotide Genasense Phase I/II 79
CD44 P245 CD44 specic Ab N/A Pre-clinical 81
EpCAM MT110 EpCAM specic Ab Micromet Phase I www.micromet.de
MT201 (Adecatumumab) Micromet Phase II www.micromet.de
EpCAM specic Ab
RAR & RXR Alitretinoin
b
pan-RAR, pan-RXR N/A Phase I 63
ATRA
b
N/A Phase I/II 61
PARP Olaparib (AZD-2281) PARP inhibitor Astra-Zenecaase Phase I 72
Veliparib (ABT-888) PARP inhibitor Abbott Phase II 73
Dll4 delta like 4; Ab antibody; SMO smoothened homologue; EpCAM Epithelial cell specic adhesion molecule; RAR retinoic acid
receptor; RXR rexinoid receptor; ATRA all-trans retinoic acid; PARP Poly (ADP-ribose) polymerase; N/A not applicable, as these agents
are still at the pre-clinical stage or commercially available.
a
Used in combination with Roche Notch c-secretase inhibitor RO4929097.
b
Used in combination with tamoxifen.
2110 M.P. Ablett et al. / European Journal of Cancer 48 (2012) 21042116
breast CSCs in vitro.
60
These agents have also shown
some promising results in clinical trials for the treatment
of breast cancer (Table 3).
6163
Breast CSCs display
increased ability to eux drugs due to high levels of
ATP-binding cassette (ABC) transporters and multidrug
resistance (MDR) proteins.
64
Side population studies
have shown that breast cancer cells capable of euxing
the uorescent dye Hoechst are enriched for CSCs.
65,66
Schinel et al. (1994) demonstrated that transgenic mice
lacking specic ABC transporters showed increased
drug sensitivity.
67
Thus, inhibition of drug transporters
may provide a novel therapeutic means of re-sensitising
CSCs. However, drug transporters such as BCRP and
ABCG2 (both ABC transporters) have been identied
on the blood brain barrier thereby conferring a risk of
adverse neurotoxic eects if administered with current
cancer therapies.
68,69
5.3. Targeting DNA repair mechanisms and apoptotic
resistance
The ability of a stem cell to maintain genomic and
chromosomal stability over time requires highly active
DNA repair mechanisms. We have listed many examples
of CSC resistance to both chemotherapy and radiother-
apy (Table 2). This intrinsic resistance can be attributed
to enhanced DNA repair mechanisms, upregulated cell
cycle control and increased free-radical scavengers.
70,71
Compounds aimed at disrupting repair mechanisms
may re-sensitise CSCs to chemotherapy or radiotherapy.
In breast cancer, clinical trials are currently underway
using inhibitors of an enzyme responsible for DNA
repair, Poly (adenosine disphosphate-ribose) polymerase
(PARP). Phase II trials have demonstrated signicant
improvement in response rate, tumour shrinkage and
overall survival when PARP inhibitors are used to treat
triple negative, metastatic breast cancer, either alone, or
in combination with conventional chemotherapy
72,73
(Table 3). However, the rst Phase III clinical trial to
combine the PARP1 inhibitor, Iniparib (BSI 201), with
chemotherapy showed no signicant improvement com-
pared to chemotherapy alone.
74
It has been suggested
that the poor ecacy of Iniparib in this trial was due to
inclusion of PARP-negative breast cancers in the cohort
and future trials should take PARP expression into
account.
75
As well as targeting DNA repair mechanisms, CSCs
could be sensitised by inhibiting pathways involved in
apoptotic resistance, such as NF-kappaB and Bcl2. Tar-
geting these anti-apoptotic proteins may improve treat-
ment of ER cancers.
76
Breast CSCs, identied by
CD44
+
expression, have higher expression of Bcl-2 than
dierentiated cells in human primary breast cancer sam-
ples.
77
The tolerability of chemotherapy in combination
with a Bcl-2 antisense oligodeoxynucleotide oblimersen
(Genasense, G3139) in neo-adjuvant systemic treatment
of primary breast cancer has been shown in phase I
clinical trials.
78,79
This gene-silencing strategy to target
Bcl-2 shows promise in the development of apoptosis-
modulating therapies.
5.4. CSC microenvironment
The signicance of the CSC microenvironment
should not be overlooked in the pursuit of CSC-targeted
therapies. However, there are few therapies currently in
development to specically target the CSC niche
(Fig. 1d). The notion of targeting components of the
CSC niche is not new. Miyake et al. (1990) demon-
strated that anti-CD44 inhibited the growth of murine
bone marrow progenitors in vitro by preventing interac-
tion with hyaluronic acid in the extra cellular matrix.
80
Since then, interest in CD44 as a CSC target in many
dierent types of cancers, including breast cancer, has
grown. Marangoni et al. (2009) examined the eects of
a monoclonal antibody against CD44 (P245) on human
breast cancer xenografts, with and without conventional
chemotherapy.
81
P245 dramatically inhibited the growth
of human breast cancer xenografts when given alone,
however the tumours re-grew when treatment was
stopped suggesting a cytostatic eect of anti-CD44 treat-
ment without complete eradication of all malignant
cells.
81
However when given in combination with adria-
mycin and cyclophosphamide tumour re-growth was
decreased demonstrating the therapeutic benet of tar-
geting cells which make up the bulk of the tumour in
combination with CSC-targeted treatment.
81
Integrins also mediate cell-extracellular matrix
interactions and are key cell surface proteins used for
enriching breast CSCs.
82
Elevated levels of the intracel-
lular signalling mediator, focal adhesion kinase (FAK),
have been linked with increased invasion.
83
Upregula-
tion of FAK in ductal carcinoma in situ (DCIS) suggests
that it is not limited to the invasive phenotype.
84
A
direct link between integrin signalling through FAK
and breast CSCs has been suggested by Luo et al.
(2009). They demonstrated that ablation of FAK
reduced the number of CSCs in primary tumours
obtained from FAK-knockout mice.
85
A link between
FAK signalling and breast CSC maintenance, tumour
progression and metastasis has been shown by several
other groups.
8688
Although there are several FAK
inhibitors in clinical trials, none of these are currently
in use to treat breast cancer.
89
These data suggest that targeting CD44 or integrin
related proteins associated with the CSC niche (e.g.
FAK) may provide an alternative strategy in breast can-
cer therapy. Many of the self-renewal pathways men-
tioned above play a role in the CSC niche. The Notch,
Hedgehog and Wnt signalling pathways all require
cell-to-cell contact for canonical signalling to occur.
Therefore these pathways should also be investigated
M.P. Ablett et al. / European Journal of Cancer 48 (2012) 21042116 2111
when considering novel therapies aimed at targeting
cells within the CSC niche.
In addition to intrinsic factors, extrinsic factors within
the tumour microenvironment can also regulate breast
CSC activity and thus may represent novel therapeutic
targets. Recent evidence suggests that cytokines such as
stromal derived factor-1, interleukin-6 (IL-6) and
interleukin-8 (IL-8) are important in regulating CSC
activity and may play a role in treatment resistance as
recently demonstrated by Max Wichas group in the case
of IL-6 and trastuzumab resistance.
9093
Studies using
human breast cancer xenograft mouse models have
shown that inhibition of IL-8 signalling specically
reduces CSC self-renewal, tumour growth and metasta-
ses.
94
In addition to autocrine production of cytokines
which promote CSC activity, other cells within the
tumour microenvironment, such as broblasts, macro-
phages and mesenchymal stem cells can also secrete these
cytokines which may further promote CSC activity via
paracrine signalling.
95
Clinical evaluation of cytokine
receptor inhibitors is needed to determine the ecacy
of targeting these signalling pathways.
6. How can we measure breast CSCs in clinical trials?
Breast cancer clinical trials can be divided into neoad-
juvant therapy trials, pre-surgical window trials and
adjuvant therapy trials. In order to assess breast CSCs
in clinical trials, we need to develop novel therapeutic
endpoints which reect their expression and/or activ-
ity/function (see Figs. 2 and 3). Fig. 2 highlights the
need for reassessment of endpoints used in clinical trials
when assessing the ecacy of novel CSC therapies.
These endpoints are currently based on tumour volume
which may not correlate with CSC frequency. Novel
clinical tests need to be developed in order to address
this issue and be able to measure CSC frequency as well
as tumour volume.
A clinical test must have a high sensitivity and spec-
icity, be acceptable to the patient, logistically feasible,
quick to perform and cost-eective. As discussed
above, the gold standard methods for assessing breast
CSC activity experimentally are in vivo tumour forma-
tion and serial transplantation assays. These methods
however are technically challenging, lengthy, expensive,
have ethical implications and would be impractical to
perform in a clinical trial setting.
Alternative in vitro methods such as colony-forming
assays and identication of cell surface markers such as
CD44
+
/CD24