TGF-Β -FOXO signalling maintains leukaemia-initiating cells in chronic
Naka, Kazuhito; Hoshii, Takayuki; Muraguchi, Teruyuki; Tadokoro, Yuko;
Ooshio, Takako; Kondo, Yukio; Nakao, Shinji; Motoyama, Noboru; Hirao,
CitationNature, 463(7281): 676-680
Text version author
; TGF-b–FOXO signalling maintains leukaemia-
initiating cells in chronic myeloid leukaemia
Kazuhito Naka1*, Takayuki Hoshii1*, Teruyuki Muraguchi1, Yuko Tadokoro1, Takako Ooshio1,2, Yukio Kondo3,
Shinji Nakao3, Noboru Motoyama4& Atsushi Hirao1,2
Chronic myeloid leukaemia (CML) is caused by a defined genetic
kinase1. It is widely believed that BCR-ABL activates Akt signalling
that suppresses the forkhead O transcription factors (FOXO), sup-
portingtheproliferation or inhibiting theapoptosis of CMLcells2–4.
through for CML therapy, imatinib does not deplete the leukaemia-
initiating cells (LICs) that drive the recurrence of CML5–8. Here,
using a syngeneic transplantation system and a CML-like myelo-
proliferative disease mouse model, we show that Foxo3a has an
essential role in the maintenance of CML LICs. We find that cells
tion are enriched in the LIC population. Serial transplantation of
ability of LICs to cause disease is significantly decreased by Foxo3a
tion of TGF-b inhibition,Foxo3a deficiencyand imatinibtreatment
of human CML LICs with a TGF-b inhibitor impaired their colony-
TGF-b–FOXO pathway in the maintenance of LICs, and strengthen
our understanding of the mechanisms that specifically maintain
CML LICs in vivo.
Although tyrosine kinase inhibitor (TKI) therapy of CML patients
efficiently induces the death of leukaemia cells5–8, LICs in these
patients can survive this therapy. To understand the molecular
mechanisms maintaining CML LICs, we characterized LICs in vivo
using a mouse model for CML-like myeloproliferative disease
(MPD)9. Consistent with previous reports10–13, we found that a rare
c-Kit1Lineage2(Lin2)Sca-11(KLS1) population of CML cells (that
is, bearing markers of normal haematopoietic stem cells (HSCs))
induced efficient CML development in recipient mice (Supplemen-
tary Fig. 1). In contrast, neither c-Kit1Lin2Sca-12(KLS2) cells
(which correspond to normal progenitors), nor other CML cell
populations expressing differentiation markers, induced CML.
importantdownstreamtargets of PI3K-Akt signalling,are essentialfor
themaintenanceof self-renewal capacity innormalHSCs14–16. When a
growth factor binds to the appropriate receptor, Akt is activated and
phosphorylates Foxo proteins, resulting in their nuclear export and
subsequent degradation in the cytoplasm. In the absence of growth
factor stimulation, Foxo proteins are retained in an active state in the
ABL is thought to activate PI3K-Akt signalling that leads to nuclear
export of Foxo factors and suppression of their transcriptional
activity2–4. However, we found that, whereas most KLS2cells (non-
LICs) showed the expected cytoplasmic localization of Foxo3a, KLS1
cells (LICs) enriched cells with nuclear localization of Foxo3a (Fig. 1a
and Supplementary Fig. 2), as observed in normal HSCs15,17(Sup-
plementary Fig. 3). LICs with nuclear Foxo3a also exhibited decreased
levels of phosphorylated Akt (p-Akt) compared to most non-LICs,
ing Foxo3a-deficientCML model. Weisolatedimmature bone marrow
cells from Foxo3a1/1and Foxo3a2/2littermates15, infected these cells
syngeneic recipients (first bone marrow transplantation (BMT)). Both
recipient groups showed the same symptoms of CML-like MPD,
(Fig. 1b, c, left and Supplementary Figs 4 and 5). Thus, Foxo3a was
dispensable for the generation of CML-like disease. After a first BMT,
the absolute numbers of Foxo3a2/2LICs in recipient spleen and
bone marrow were significantly higher than the absolute numbers of
Foxo3a1/1LICs present in recipient organs (Fig. 1d, left and Sup-
Foxo3a1/1and Foxo3a2/2LICs was comparable (Fig. 1e, left).
When we transplanted Foxo3a1/1and Foxo3a2/2LICs from first
BMT mice into a new set of recipients (second BMT), there was no
difference in CML development (Fig. 1b, c, centre and Supplemen-
tary Figs 4 and 5). However, the in vitro colony-forming ability of
third BMT, relatively mild CML-like MPD developed in recipients
within one month. Foxo3a deficiency prevented the propagation of
Supplementary Figs 4 and 5). Furthermore, the absolute numbers of
Foxo3a2/2LICs in the organs of third BMT recipients were much
lower than in third BMT recipients that had received Foxo3a1/1
LICs (Fig. 1d, right and Supplementary Fig. 6). Thus, Foxo3a2/2
LICs may retain sufficient function to cause disease in second BMT
recipients but succumb to exhaustion in third BMT recipients.
Although CML-like MPD developed in mice that had received either
Foxo3a1/1or Foxo3a2/2LICs within 40days of a third BMT (Sup-
plementary Fig. 5a, b), recipients of Foxo3a1/1LICs also developed
acute lymphocytic leukaemia (ALL), as well as CML, 40days after the
observe development of ALL or CML in recipients of Foxo3a2/2LICs
45daysaftera thirdBMT(Supplementary Fig.5c), suggesting that the
Foxo3a2/2LICs lose their potential to generate malignancies. This
Nature nature08734.3d23/12/09 13:38:05
*These authors contributed equally to this work.
Evolution Science and Technology (CREST), Japan Science and Technology Agency (JST), Chiyoda-ku, Tokyo 102-0075, Japan.3Cellular Transplantation Biology, Division of Cancer
Medicine, Kanazawa University, Graduate School of Medical Science, Kanazawa, Ishikawa 920-8641, Japan.4Department of Geriatric Medicine, National Institute for Longevity
Sciences, National Center for Gerontology and Geriatrics, 36-3 Gengo, Morioka, Obu, Aichi 474-8522, Japan.
inability of Foxo3a2/2LICs to induce or sustain leukaemia resulted in
the reduced lethality of these animals (Fig. 1c, right). Thus, Foxo3a is
CML LICs impaired LIC function in vivo (Supplementary Fig. 7), the
defective maintenance of Foxo3a2/2LICs is probably due to a dimi-
nution in LIC self-renewal activity rather than a defect in the original
HSCs. Furthermore, the Foxo4 protein expression pattern in LICs is
very similar to that of Foxo3a (Supplementary Fig. 8), suggesting an
overlap in Foxo functions.
We next determined how Foxo3a deficiency impaired the main-
tenance of the CML-initiating potential of LICs. Loss of Foxo3a did
not affect LIC differentiation potential (Fig. 2a) or cell cycle status
(Fig. 2b), although LICs with nuclear localization of Foxo3a showed
lower expression of the Ki67 antigen (Supplementary Fig. 9).
apoptotic cells were significantly increased in histological sections of
bone marrow and spleen from Foxo3a2/2CML-affected mice
compared to controls (Fig.2c, d).To confirm that this apoptosis was
occurring in LICs, we purified KLS1cells and showed that the fre-
quency of Annexin-V1or TUNEL1cells among Foxo3a2/2KLS1
cells was higher than among Foxo3a1/1KLS1cells (Fig. 2e, f). Thus,
Foxo3a is required for LIC survival because it mediates suppression
An intriguing question in current CML research is how Akt is
inactivated (and thus Foxo is activated) mainly in LICs, despite
BCR-ABL expression in all CML cells. The fact that TGF-b regulates
Akt activation (and thus nuclear Foxo3a) in normal HSCs prompted
us to assess whether TGF-b signalling controls Foxo localization in
CML LICs19. We first examined the phosphorylation of Smad2/3
proteins, which are downstream effectors in the TGF-b signalling
pathway20. Phosphorylation of Smad2/3 in the nuclei of CML
KLS1cells was higher than that in KLS2cells (Fig. 3a and Sup-
plementary Fig. 10), indicating that TGF-b signalling was activated
in CML LICs. We then treated CML KLS1cells in vitro for 2h with
Foxo3a localization. TGF-b1 treatment increased Foxo3a nuclear
localization, whereas Ly364947 promoted nuclear export of Foxo3a
(Fig. 3b–d and Supplementary Fig. 11). Furthermore, KLS1cells
treated with TGF-b1 showed high Smad2/3 and low Akt phosphor-
ylation levels, whereas cells treated with Ly364947 exhibited the
opposite pattern (Fig. 3b, c and Supplementary Figs 11 and 12). The
activity of mTOR complex 1 (mTORC1), which was determined by
assessing the phosphorylation of S6 ribosomal protein22, correlated
positively with p-Akt levels in these KLS1cells (Fig. 3d and Sup-
plementary Fig. 13). We propose that TGF-b signalling controls Akt
activity in LICs, leading to retention of Foxo3a in the nucleus.
Naturenature08734.3d 23/12/09 13:38:05
020 103008 12 1410
P = 0.424
P = 0.005
P = 0.046
P = 0.007
Time after BMT (day)
Time after BMT (day)
P = 0.394
P = 0.017
No. of WBC in PB
(×104 per ml)
(n = 17)
(n = 24)
Foxo3a+/+ (n = 22)
Foxo3a–/– (n = 23)
(n = 15)
(n = 11)
P = 0.134
P = 0.770
P = 0.002
Survival rate (%)
Absolute no. of LICs
P = 0.022
n = 4
P = 0.396
n = 3
P = 0.007
n = 3
P = 0.125
n = 8
P = 0.011
n = 8
1st BMT2nd BMT3rd BMT
1st BMT2nd BMT3rd BMT
1st BMT2nd BMT1st BMT2nd BMT3rd BMT
No. of colonies
per 2,000 LICs
+/+–/– +/+ –/–+/+ –/–+/+–/–+/+–/–
Figure 1 | Maintenance of CML LICs depends on Foxo3a. a, Decreased Akt
phosphorylation and increased Foxo3a nuclear localization in CML LICs.
BCR-ABL-GFP1KLS2and KLS1cells were immunostained to detect p-Akt
and Foxo3a. Nuclei were visualized using 49,6-diamidino-2-phenylindole
(DAPI; blue). Scale bars, 10mm. b, Numbers of white blood cells (WBC) in
peripheral blood (PB) in Foxo3a1/1(open circles) and Foxo3a2/2(closed
squares) CML-affected mice at first to third BMT. Data shown are the mean
WBC number6s.d. (n54). c, d, Survival rate (c) and absolute number of
LICs (d) in CML-affected mice. Data are expressed as percentage survival
(c) and absolute number of LICs6s.d. in spleen (d). e, In vitro colony-
forming ability of LICs. Data shown are the mean colony number6s.d.
P = 0.193
Frequency of progenitors
in GFP+ cells (%)
Frequency of BrdU+ cells
in GFP+KLS+ cells (%)
Frequency of Annexin-V+ cells
in GFP+KLS+ cells (%)
P = 0.418
P = 0.163
P = 0.133
P = 0.011
Figure 2 | Essential role of Foxo3a in suppression of LIC apoptosis.
show the mean percentage6s.d. of CMP (common myeloid progenitors)-,
GMP (granulocyte-macrophage progenitors)-, and MEP (megakaryocyte-
erythroid progenitors)-like cells in BCR-ABL-GFP1bone marrow cells
(n53). b, Data are the mean frequency of BrdU1cells6s.d. (n53).
c–f, Increased apoptosis of Foxo3a2/2LICs. c, d, Histological sections of
bone marrow (c) and spleen (d) from Foxo3a1/1or Foxo3a2/2CML-
affected mice at second BMT were analysed by TUNEL staining (red) and
DAPI. Scale bars, 100mm. e, Mean percentage6s.d. of Annexin-V1cells
among GFP1KLS1cells (n53). f, LICs were subjected to TUNEL staining.
Scale bars, 50mm.
Notably, as TGF-b did not efficiently induce nuclear localization of
Foxo3a in non-LICs (KLS2) cells (Supplementary Figs 14–16), there
seem to be differences in how TGF-b-stimulated Foxo3a activation is
regulated in LICs and non-LICs.
Interestingly, TGF-b inhibitors (Ly364947, SD208) suppressed LIC
colony-forming ability after co-culture with OP-9 stromal cells,
which are used to maintain normal HSCs ex vivo and mimic their
in vivo environment23(Fig. 3e and Supplementary Fig. 17a).
with OP-9 cells reduced the colony-forming capacity of Foxo3a1/1
CML LICs, the addition of TGF-b inhibitor significantly enhanced
and Supplementary Figs 17b, c and 18a), consistent with a report in
which TGF-b inhibition enhanced imatinib-induced cell death24.
ability (Fig. 1e, left), significantly fewer colonies arose from
Foxo3a2/2LICs treated with imatinib compared to treated
Foxo3a1/1LICs (Fig. 3f). When imatinib was combined with
Foxo3a deficiency and Ly364947, there was no further inhibitory
effect compared to either Foxo3a deficiency or Ly364947 treatment
ability ofnormalKLS1cellsunder thesameexperimental conditions
(Supplementary Fig. 18b). These data indicate that Foxo3a is an
important downstream effector in the TGF-b signalling pathway
driving the survival of LICs exposed to TKI therapy, and that inhibi-
To test this hypothesis in vivo, we administered imatinib to second
BMT mice bearing Foxo31/1or Foxo3a2/2CML LICs. Without
imatinib, 80% of both groups died within 60days (Fig. 4a). Admini-
stration of imatinib alone delayed CML onset and reduced recipient
lethality, but 60% of the treated mice that had received Foxo3a1/1
deficiency significantly increased the survival of CML-affected mice.
imatinib-treated recipients, we found that Foxo3a deficiency had
reduced both the number and colony-forming ability of these LICs
(Supplementary Fig. 19). Thus, Foxo3a is essential for the ability of
CML LICs to survive imatinib therapy.
functions. Administration of Ly364947 alone led to increased p-Akt
levels and decreased nuclear Foxo3a in LICs (Fig. 4b and
Supplementary Fig. 20), demonstrating that TGF-b is a critical regu-
lator of Akt and Foxo in LICs in vivo. Unlike the suppressive effect
exerted by TGF-b in vitro, administration of Ly364947 alone did not
extend the survival of second BMT Foxo3a1/1mice with overt CML.
However, Ly364947 combined with imatinib significantly reduced
recipient lethality, decreased CML infiltration in lung (Fig. 4c and
Supplementary Fig. 21a), and decreased LIC frequency (Supplemen-
tary Fig. 21b, c). These in vivo data suggest that CML LICs are more
sensitive than normal HSCs/progenitors to TGF-b inhibitors.
inhibitors on mouse CML LIC function are also observed in human
CML LICs. We isolated CD341CD382Lin2cells as human CML
LICs12from bone marrow cells of human CML patients, and co-
cultured them on OP-9 cells with or without Ly364947. Ly364947
reduced the number of colonies formed by human CML LICs
(Fig. 4d) and enhanced the inhibitory effects of imatinib on human
CML LICs (Fig. 4e), suggesting that TGF-b–FOXO signalling also
governs the behaviour of human CML LICs.
Our study indicates a model in which Foxo3a has opposite effects
on the survival of LICs and non-LICs (Fig. 4f and Supplementary
Discussion 1). In non-LICs, BCR-ABL drives strong Akt activation
that forcefully represses Foxo3a functions. When TKIs block BCR-
ABL and reduce Akt activation, activation of Foxo3a leads to apop-
despite BCR-ABL expression in vivo, leading to nuclear localization
of Foxo3a. Although a previous study suggested that Foxo3a may
contribute to the acquisition of dormancy in leukaemia cells after
exposure to anti-leukaemic agents2, our data demonstrate that LICs
have properties that allow them to resist various stress in vivo in a
Foxo-dependent manner. Notably, we found that the role of Foxo3a
in a CML blast crisis model was different from that in this MPD
can mediate different regulatory mechanisms in different phases and
types of leukaemia.
It remains unclear precisely how Akt and Foxo signalling is con-
trolled in LICs in vivo. The presence of TGF-b1 in CML LICs
(Supplementary Fig. 24) and the in vitro effects of TGF-b inhibitor
on Foxo3a localization in LICs suggest that a cell-autonomous effect
of TGF-b may control Foxo3a function in vivo. However, when we
cultured LICs with TGF-b inhibitors in a stroma-free system, colony
the survival and/or proliferation of LICs depends not only on TGF-b
produced by LICs themselves but also on TGF-b in the surrounding
microenvironment. Alternatively, TGF-b signalling in niche in vivo
In any case, it appears that LICs are controlled by a complicated
network of TGF-b signalling.
Although our study suggests that inhibition of the TGF-b–FOXO
application of TGF-b or FOXO inhibitors to the treatment of haema-
ing how the LICs are maintained in vivo.
p-Smad2/3 DAPIFoxo3a Merge
No. of colonies per
1,000 LICs (×103)
No. of colonies per
P = 0.0012
P < 0.0001
P < 0.0001
P < 0.00001
Ly36494702 1020 (µM)
P = 0.111
P < 0.0001
Figure 3 | TGF-b–Foxo signalling is required for the colony-forming
KLS1and KLS2CML cells were immunostained to detect Foxo3a and
phosphorylated-Smad2/3 (p-Smad2/3). b–d, Control of Foxo3a localization
2h followed by immunostaining to detect Foxo3a and (b) p-Smad2/3,
TGF-b decreasesLICcolony-formingability. Representative data shownare
the mean LIC colony number6s.d. (n53). f, TGF-b inhibition or Foxo3a
deficiency enhances imatinib-mediated inhibition of LIC colony-forming
capacity. Representative data shown are the mean colony number6s.d.
For our mouse CML model, normal immature bone marrow cells (KLS1) from
C57BL/6 mice were infected with retrovirus carrying MSCV-BCR-ABL-ires-
irradiated (9.5Gy) C57BL/6 congenic mice along with 53105bone marrow
mononuclear cells from C57BL/6 mice. The development of CML-like MPD
was confirmed by morphological analysis and marker determinations. Toanalyse
was purified and transplanted into syngeneic recipients along with 53105bone
Foxo3a2/2and Foxo3a1/1littermates (C57BL/6; F4) were subjected to the above
protocol. For Akt activity, TGF-b signalling and Foxo3a localization, freshly iso-
nostained with anti-p-Akt (Cell Signaling), anti-phospho-S6 (Cell Signaling),
anti-phospho-Smad2/3 (Millipore), or anti-Foxo3a (Sigma). To examine
colony-forming ability in vitro, GFP1KLS1cells were cultured in semi-solid
methylcellulose medium (stroma-free)15, or were co-cultured for 5days on OP-9
stromal cells with Ly364947 (10mM). For experiments involving imatinib,
gift from Novartis), followed by washing and transfer to semi-solid medium.
CML patients (Cureline, Inc. and AllCells). These cells were co-cultured on OP-9
vivo BrdU incorporation (12h). Apoptosis was assayed by TUNEL (Roche) or
Annexin-V (Abcam) staining. Recipient mice bearing Foxo3a1/1or Foxo3a2/2
CML LICs received either imatinib by oral gavage twice a day for 80days at
tion for 80days at 10 mgkg21(of body weight) every 2days.
Note added in proof: During the reviewing process, a paper demonstrating that
human CML stem cells show predominant nuclear localization of FOXO
proteins was published25. The study supports our conclusion about the import-
ant roles of FOXO in CML LICs.
Full Methods and any associated references are available in the online version of
the paper at www.nature.com/nature.
Received 9 March; accepted 4 December 2009.
Published online XX 2010.
1.Ren, R. Mechanisms of BCR-ABL in the pathogenesis of chronic myelogenous
leukaemia. Nature Rev. Cancer 5, 172–183 (2005).
2.Komatsu, N. et al. A member of Forkhead transcription factor FKHRL1 is a
downstream effector of STI571-induced cell cycle arrest in BCR-ABL-expressing
cells. J. Biol. Chem. 278, 6411–6419 (2003).
Ghaffari, S., Jagani, Z., Kitidis, C., Lodish, H. F. & Khosravi-Far, R. Cytokines and
BCR-ABL mediate suppression of TRAIL-induced apoptosis through inhibition of
forkhead FOXO3a transcription factor. Proc. Natl Acad. Sci. USA 100, 6523–6528
Essafi, A. et al. Direct transcriptional regulation of Bim by FoxO3a mediates
STI571-induced apoptosis in Bcr-Abl-expressing cells. Oncogene 24, 2317–2329
Graham, S. M. et al. Primitive, quiescent, Philadelphia-positive stem cells from
Michor, F. et al. Dynamics of chronic myeloid leukaemia. Nature 435, 1267–1270
Roeder, I. et al. Dynamic modeling of imatinib-treated chronic myeloid leukemia:
functional insights and clinical implications. Nature Med. 12, 1181–1184 (2006).
Jørgensen, H. G., Allan, E. K., Jordanides, N. E., Mountford, J. C. & Holyoake, T. L.
Nilotinib exerts equipotent antiproliferative effects to imatinib and does not
induce apoptosis in CD341CML cells. Blood 109, 4016–4019 (2007).
Daley, G. Q., Van Etten, R. A. & Baltimore, D. Induction of chronic myelogenous
leukemia in mice by the P210bcr/ablgene of the Philadelphia chromosome. Science
247, 824–830 (1990).
10. Hu,Y.etal.Targeting multiplekinasepathwaysinleukemicprogenitorsandstem
cells is essential for improved treatment of Ph1leukemia in mice. Proc. Natl Acad.
Sci. USA 103, 16870–16875 (2006).
11.Neering, S. J. et al. Leukemia stem cells in a genetically defined murine model of
blast-crisis CML. Blood 110, 2578–2585 (2007).
12. Ito, K. et al. PML targeting eradicates quiescent leukaemia-initiating cells. Nature
453, 1072–1078 (2008).
13. Zhao, C. et al. Hedgehog signalling is essential for maintenance of cancer stem
cells in myeloid leukaemia. Nature 458, 776–779 (2009).
14. Tothova, Z. et al. FoxOs are critical mediators of hematopoietic stem cell
resistance to physiologic oxidative stress. Cell 128, 325–339 (2007).
cell pool. Cell Stem Cell 1, 101–112 (2007).
16. Yalcin, S. et al. Foxo3 is essential for the regulation of ataxia telangiectasia
J. Biol. Chem. 283, 25692–25705 (2008).
17. Yamazaki, S. et al. Cytokine signals modulated via lipid rafts mimic niche signals
and induce hibernation in hematopoietic stem cells. EMBO J. 25, 3515–3523
18. Pear,W.S.etal.Efficientandrapidinduction ofachronic myelogenousleukemia-
like myeloproliferative disease in mice receiving P210 bcr/abl-transduced bone
marrow. Blood 92, 3780–3792 (1998).
P = 0.018
control (n = 14)
control (n = 15)
control (n = 16)
imatinib (n = 16)
(n = 15)
P = 0.022
(n = 12)
(n = 19)
(n = 16)
Time after BMT (day)
90 30 600
Time after BMT (day)
Survival rate (%)
P = 0.005
P = 0.005
P < 0.0001
P = 0.003
cell cycle arrest
No. of colonies per
No. of colonies per
Ly364947 – +
Survival rate (%)
Figure 4 | Inhibition of TGF-b–Foxo3a signalling in combination with TKI
imatinib-treated CML mice in a second BMT. At 10days after BMT, mice
transplanted with Foxo3a1/1or Foxo3a2/2LICs (1.53104) received either
vehicle (control) or imatinib for 80days, and the percentage survival was
determined. b, Ly364947 administration to CML-affected mice (25mgkg21
10mm. c, Combined administration of Ly364947 and imatinib prolongs the
survival of CML-affected mice. At 10days after BMT, recipients received
dimethyl sulphoxide (DMSO) or Ly364947 plus vehicle (control) or
imatinib for 80days, and the percentage survival was determined.
d, e, Inhibition of TGF-b decreases the in vitro colony-forming ability of
human CML LICs. CD341CD382Lin2cells from bone marrow of CML
patients were co-culturedon OP-9 cells with either DMSO (2) or Ly364947
the mean colony number6s.d. (n53). f, Scheme outlining the proposed
distinct roles of Foxo3a in non-LICs versus CML LICs.
19. Yamazaki, S. et al. TGF-b as a candidate bone marrow niche signal to induce
hematopoietic stem cell hibernation. Blood 113, 1250–1256 (2009).
20. Massague ´, J. TGFb in Cancer. Cell 134, 215–230 (2008).
21. Sawyer, J. S. et al. Synthesis and activity of new aryl- and heteroaryl-substituted
pyrazole inhibitors of the transforming growth factor-b type I receptor kinase
domain. J. Med. Chem. 46, 3953–3956 (2003).
22. Bhaskar, P. T. & Hay, N. The two TORCs and Akt. Dev. Cell 12, 487–502 (2007).
23. Kodama, H., Nose, M., Niida, S. & Nishikawa, S. Involvement of the c-kit receptor
in the adhesion of hematopoietic stem cells to stromal cells. Exp. Hematol. 22,
24. Møller, G. M., Frost, V., Melo, J. V. & Chantry, A. Upregulation of the TGFb
signalling pathway by Bcr-Abl: implications for haemopoietic cell growth and
chronic myeloid leukaemia. FEBS Lett. 581, 1329–1334 (2007).
25. Pellicano, F. et al. FOXO transcription factor activity is partially retained in
quiescent CML stem cells and induced by tyrosine kinase inhibitors in CML
progenitor cells. Blood. doi:10.1182/blood-2009-06-226621 (1 December 2009).
Supplementary Information is linked to the online version of the paper at
Acknowledgements We thank H. Honda for BCR-ABL cDNA, C. A. Schmitt for
Plat-E retroviral packaging cells, T. Suda, N. Komatsu and K. Miyazono for
discussions, and M. Sakae and T. Hatakeyama for expert technical support. We
grant-in-aid for Scientific Research (C), and A.H. was supported by grants-in-aid
for Scientific Research (B) and Creative Scientific Research (17GS0419), from the
Ministry of Education, Culture, Sports, Science and Technology, Japan.
Author Contributions K.N. designed research, performed experiments, analysed
data, and co-wrote the paper. T.H., T.M., Y.T, T.O. and N.M. performed
experiments. Y.K. and S.N. provided technical support for the human cell
experiments. A.H. designed research, analysed data and co-wrote the paper.
Author Information Reprints and permissions information is available at
www.nature.com/reprints. The authors declare no competing financial interests.
Correspondence and requests for materials should be addressed to K.N.
(email@example.com) or A.H.
Nature nature08734.3d23/12/09 13:38:12
Mice. Foxo3a2/2and littermate Foxo3a1/1mice of the C57BL/6 (F4) genetic
background15were used in this study. C57BL/6 congenic mice were purchased
from Sankyo-Lab Service.Animal care in our laboratory was in accordance with
the guidelines for animal and recombinant DNA experiments of Kanazawa
University. Imatinib (Glivec; purchased from Novartis) was administered to
mice by oral gavage twice a day (200mgkg21of body weight per day in water).
Ly364947 (Merck Chemicals Ltd) was prepared as a 5mgml21stock solution in
DMSO and intraperitoneally administered in saline to mice for 80days at
10mgkg21(of body weight) every 2days, or for 7 days at 25mgkg21(body
weight) per day.
(RB6-8C5), Mac1 (M1/70), IL7Ra chain (B12-1), FccIII/II receptor (2.4G2) and
CD34 (RAM34, pacific blue-conjugated) (all from BD Biosciences). Anti-c-Kit
(ACK2) monoclonal antibody was from eBiosciences. A mixture of monoclonal
antibodies recognizing CD4, CD8, B220, TER119, Mac1 and Gr-1 was used to
Preparation of retrovirus. The cDNA encoding human BCR-ABL (gift from H.
Honda) was cloned into the EcoRI site of the MSCV or MSCV-ires-GFP vector.
Retroviral packaging cells (Plat-E) were transiently transfected with the MSCV-
plantation into mice as described later.
mate Foxo3a1/1mice were purified by flow cytometry and cultured in serum-
free S-Clone SF-03 medium (Sanko Junyaku) supplemented with 10ngml21
human TPO (thrombopoietin; PeproTech) plus 10ngml21mouse SCF (stem
cell factor; Wako Pure Chemical). To generate our CML-like MPD mouse
model, normal KLS1cells were infected with the above retrovirus carrying
MSCV-BCR-ABL-ires-GFP using CombiMag (OZ Bioscience). The transduced
KLS1cells were transplanted intravenously into lethally irradiated (9.5Gy)
C57BL/6 congenic mice along with 53105bone marrow mononuclear cells
from C57BL/6 mice.
For the prospective isolation of LICs, mononuclear cells were isolated from
bone marrow or spleen of CML-affected recipient mice at 12–14days post-
transplantation, and GFP1subpopulations (33104) were recovered by flow
as described earlier. For serial transplantations, GFP1KLS1cells (33104or
1.53104) were collected and pooled from five BMT mice and transplanted into
a second set of lethally irradiated congenic recipient mice along with 53105
normal bone marrow mononuclear cells from C57BL/6 mice. The absolute
number of CML LICs in the spleen of a recipient mouse was calculated as (total
3 frequency of GFP1KLS1cells (%) 3 1/100).
Mouse cell colony-forming assay. Mouse cells were cultured in semi-solid
medium (stroma-free) containing the cytokines SCF, IL-3, IL-6 and erythro-
air containing 5% CO2for 7days. For co-culture assays, cells (13103) were co-
cultured on OP-9 stromal cells23under hypoxic conditions (5% O2) for 5days.
After collecting and washing with PBS, colony formation was assessed in semi-
solid medium as described earlier. For inhibition of TGF-b signalling, cells were
20mM Ly364947 or SD208 (Sigma) for 5days, followed by washing and transfer
to semi-solid medium. For the combination treatment of TGF-b inhibitor plus
imatinib, cells were first treated with DMSO or a TGF-b inhibitor on OP-9
stromal cells. At 24h after starting culture, the cells were treated with further
DMSO or 5mM imatinib (STI571; provided by Novartis) and incubated for
another 4days (total 5days). Thereafter, the cells were washed with PBS and
transferred to semi-solid medium. The number of colonies was counted under
medium without cytokines for 30min on poly-L-lysine (Sigma)-coated glass
slides26at 37uC under hypoxic conditions (5% O2). To examine the effects of
TGF-b signalling, cells were treated with 5ngml21TGF-b1 (R&D Systems), 10
staining, the permeabilized cells were incubated with mouse or rabbit anti-
FKHRL1 (Foxo3a) (FR1 or F2178; Sigma), rabbit anti-Foxo4 (9472; Cell
Signaling), rabbit anti-phospho-Smad2/3 (Millipore), rabbit anti-phospho-
Akt (D9E; Cell Signaling), or rabbit anti-phospho-S6 ribosomal protein (2211;
incubating the cells with AlexaFluor 546- or AlexaFluor 647-conjugated goat
with the DNA marker DAPI (Sigma). Stained slides were mounted using
Fluoromount Plus (Diagnostic Biosystems), and fluorescent images were
acquired using a Fluoview 1000 laser confocal microscope (Olympus) and
Photoshop software (Adobe). Fluorescence intensities were quantified using
ImageJ software. To evaluate the subcellular localization of Foxo3a, approxi-
mately 100 cells per group were counted under the microscope.
Cell cycle analysis and differentiation potential. To determine the cell cycle
status of LICs in vivo, second BMT CML mice were administered BrdU intra-
peritoneally (100mgkg21of body weight in saline; Sigma) for 12h. Numbers of
BrdU1LICs were assessed by immunostaining with anti-BrdU antibody (3D4;
BD Biosciences) and flow cytometry. To examine the cell cycle status of LICs in
rabbit anti-FKHRL1 antibody(F2178;Sigma) todetect Foxo3a,and anti-mouse
Ki67 antibody (B56; BD Biosciences).
To evaluate differentiation potential, the frequencies of CMP-like
(CD341FccIII/II receptor2), GMP-like (CD341FccIII/II receptor1), and
MEP-like (CD342FccIII/II receptor2) cells among GFP1bone marrow cells
obtained from Foxo3a1/1and Foxo3a2/2CML mice at first BMT were analysed
by flow cytometry.
Detection of apoptosis. To determine apoptosis in vivo, freshly isolated spleens
and bones were immediately fixed with 4% paraformaldehyde, and tissue
sections were prepared and stained using the TUNEL method (Roche).
Isolated GFP1KLS1cells from mice that had been transplanted with
using the TUNEL method (Roche).
Inhibition of endogenous Foxo by dominant-negative Foxo. Retroviruses
carrying a dominant negative (dn)Foxo vector (MSCV-dnFoxo-ires-GFP27) or
a control GFP vector (MSCV-ires-GFP) were generated as described earlier. To
obtain dnFoxo CML LICs, normal KLS1cells (wild-type C57BL/6 CD45.1) were
infected with retrovirus carrying the BCR-ABL gene (without GFP). These
infected KLS1cells were transplanted into irradiated recipient mice (C57BL/6
(unfractionated) were transplantedintravenously into lethallyirradiated C57BL/
6 mice. At 14days after transplantation, absolute numbers of GFP1KLS1cells
were measured in recipient spleens. To examine LIC colony-forming ability in
vitro, these GFP1CML LICs were cultured in semi-solid medium (stroma-free),
or were co-cultured on OP-9 stromal cells as described earlier.
Quantitative real-time RT–PCR analysis. RNA samples were purified from
fractionated GFP1KLS1and GFP1KLS2cells (1.03105) using the RNeasy kit
(QIAGEN) and reverse-transcribed using the Advantage RT-for-PCR kit
following primers were used: Tgfb1, 59-TATGCTAAAGAGGTCACCCGCG-39
and 59-TGCTTCCCGAATGTCTGACG-39; Alk5 (also known as Tgfbr1), 59-GA
TCGCCCTTTCATTTCAGAGG-39 and 59-AAACCGACCTTTGCCAATGC-39;
Tgfbr2, 59-GAGAGCATGAAAGACAGTGTGC-39 and 59-CCAGCACTCGGTC
AAAGTCT-39; Actb, 59-AGGTCATCACTATTGGCAACGA-39 and 59-CACTT
CATGATGGAATTGAATGTAGTT-39. The following cycle parameters were
used: denaturation at 95uC for 10s, and annealing and elongation at 57uC for
Actb, and 60uC for Tgfb1, Tgfbr1 and Tgfbr2.
Analyses of primary human CML samples. Viable bone marrow mononuclear
cells from patients with chronic phase CML were purchased from Cureline, Inc.
(no. 16-122;) and AllCells LLC (no. 06-255 and 06-620). Cells were stained with
CD19 (SJ25C1), anti-CD20 (L27), anti-CD14 (MwP9), and anti-CD56
(NCAM16.2) antibodies (all from BD Biosciences). A mixture of monoclonal
antibodies recognizing CD3, CD16, CD19, CD20, CD14 and CD56 was used to
identify Lin2cells. CD341CD382Lin2cells were purified by cell sorting. To
examine the effects of treatment with Ly364947 alone or a combination of
Ly364947 plus imatinib, CD341CD382Lin2cells were cultured on OP-9
stromal cells as described earlier. After collecting and washing in PBS, the col-
ony-forming ability of LICs was evaluated by culture in semi-solid medium
containing SCF, GM-CSF, IL-3, IL-6, G-CSF and erythropoietin (Methocult
GF1H4435; Stem Cell Technologies).
Generation of CML blast crisis model. A mouse model of CML blast crisis was
with a retrovirus carrying MSCV-BCR-ABL-ires-GFP and a retrovirus carrying
Naturenature08734.3d 23/12/09 13:38:13
MSCV-NUP98/HOXA9. The transduced Lin2Sca-11(LS1) cells11were trans-
53105bone marrow mononuclear cells from C57BL/6 mice.
Statistical analyses. Statistical differences were determined using the unpaired
Student’s t-test for P-values, and the long-rank non-parametric test for survival
26. Ema, H. et al. Adult mouse hematopoietic stem cells: purification and single-cell
assays. Nature Protocols 1, 2979–2987 (2006).
27. Bouchard, C. et al. FoxO transcription factors suppress Myc-driven
lymphomagenesis via direct activation of Arf. Genes Dev. 21, 2775–2787 (2007).
28. Dash, A. B. et al. A murine model of CML blast crisis induced by cooperation
Nature nature08734.3d23/12/09 13:38:15
Title: TGF-b–FOXO signalling maintains leukaemia-
?initiating cells in chronic myeloid leukaemia
nature08734): Figs 1–4 (colour); Tables: None; Boxes: None. Please check all figures very carefully as they
have been re-labelled, re-sized and adjusted to Nature’s style.
2Nature Proofreader: Please update/confirm the tentative publication date
For Nature office use only:
Methods (if applicable)
AOP (if applicable)
Supp info (if applicable)
Author contribs (if applicable)
Naturenature08734.3d 23/12/09 13:38:15